Textbook of

Clinical
Epidemiology

For Medical Students

Chongjian Wang
Fen Liu
Editors

123
Textbook of Clinical Epidemiology
Chongjian Wang • Fen Liu
Editors

Textbook of Clinical
Epidemiology
For Medical Students
Editors
Chongjian Wang Fen Liu
Department of Epidemiology and Department of Epidemiology and Health
Biostatistics, College of Public Health Statistics, School of Public Health
Zhengzhou University Capital Medical University
Zhengzhou, Henan, China Beijing, China

ISBN 978-981-99-3621-2 ISBN 978-981-99-3622-9 (eBook)

https://doi.org/10.1007/978-981-99-3622-9

Jointly published with Zhengzhou University Press

The print edition is not for sale in China (Mainland). Customers from China (Mainland) please order the
print book from: Zhengzhou University Press.

© Zhengzhou University Press 2023

This work is subject to copyright. All rights are solely and exclusively licensed by the Publisher, whether
the whole or part of the material is concerned, specifically the rights of reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
or information storage and retrieval, electronic adaptation, computer software, or by similar or
dissimilar methodology now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publishers, the authors, and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publishers nor the
authors or the editors give a warranty, expressed or implied, with respect to the material contained
herein or for any errors or omissions that may have been made. The publishers remain neutral with
regard to jurisdictional claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Singapore Pte Ltd.
The registered company address is: 152 Beach Road, #21-01/04 Gateway East, Singapore 189721,
Singapore
Foreword

As a newly developed basic discipline of modern medicine, clinical epidemiology

attempts to apply epidemiological and statistical theories and methods to resolve the
various problems in medical practice and guides clinical practice. With the rapid
development of social mobility, computer availability, readily accessible powerful
software, and the development of statistical methods, clinical epidemiology has
increasingly become an important approach to clinical medicine and practice. This
is because of its ingenious application of theories and methods of epidemiology and
health statistics to clinical research to continuously enrich and optimize methodol-
ogy for clinical research. Clinical epidemiology also provides tools used to obtain
observable evidence from clinical trials and contributes to enhancing the develop-
ment of clinical diagnosis and treatment. Clinical epidemiology is a useful tool for
clinical practitioners undertaking clinical practice and scientific research, adequately
learning and applying its principles will help clinicians enhance their knowledge and
increase their efficiency through acquiring reliable information that is needed for
decision-making.
The most important feature of the knowledge economy era is the continuous
acquisition and updating of information. This book was therefore compiled in order
to reflect the need to nurture students in this new era, meet the demand for talented
individuals in contemporary society, and promote the exchange of science and
technology between China and the West through “the Belt and Road Initiative.”
Several topics pertaining to more than one aspect of science are discussed in various
sections of this text. The first ten chapters of this book concentrate on the basic
principles, concepts, and methodology used in clinical epidemiology while the
remaining chapters are composed of its practical applications. Each chapter starts
with a few key points and ends with short questions.
This book aims to provide an overview of the principles of clinical epidemiology,
which is not only as a reference but also as a tool for several daily tasks. When
writing a paper or reviewing articles and reports, it can aid in checking the appro-
priateness and the implications of concepts and words; when teaching and lecturing,
it may assist in preparing notes and visual aids. Individual chapters may provide a

v
vi Foreword

fresh perspective on familiar topics. Furthermore, this book can be used as a

textbook for graduate and undergraduate students in medical schools, as well as a
reference book for medical teachers and practitioners. Although we have attempted
to be as accurate as possible, we acknowledge that any work of this scope could
contain mistakes and omissions, thus, any suggestions on the improvement of this
book would be appreciated.

Department of Epidemiology and Guangcai Duan

Biostatistics, College of Public Health,
Zhengzhou University, Zhengzhou,
Henan, People’s Republic of China
Contents

1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
Chongjian Wang
1.1 Brief History of Clinical Epidemiology . . . . . . . . . . . . . . . . . . . 1
1.1.1 Early Clinical Epidemiology . . . . . . . . . . . . . . . . . . . . 2
1.1.2 Clinical Epidemiology . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.3 Modern Clinical Epidemiology . . . . . . . . . . . . . . . . . . 3
1.2 Definition of Clinical Epidemiology . . . . . . . . . . . . . . . . . . . . . 4
1.3 Roles of Clinical Epidemiology . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.1 To Provide Scientific Ideas and Methods for Clinical
Medical Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.2 To Provide Scientific Methods and Means for the
Evaluation of Clinical Diagnosis and Treatment . . . . . . 6
1.3.3 To Provide a Scientific Methodology and Evidence for
Clinical Decision-Making and Practice of Evidence-
Based Medicine . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.3.4 It is Possible to Train Clinicians and Medical
Scientists with Excellent Knowledge, Skills, and
Quality Under the Modern Medical Model . . . . . . . . . . 7
1.4 Methodology of Clinical Epidemiologic Study . . . . . . . . . . . . . 7
1.4.1 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4.1.1 Clarification of Study Aim . . . . . . . . . . . . . 8
1.4.1.2 Determination of Study Methods . . . . . . . . . 8
1.4.1.3 Identification of the Study Subjects . . . . . . . 9
1.4.1.4 Determination of the Groups . . . . . . . . . . . . 10
1.4.1.5 Determination of Study Indicators . . . . . . . . 10
1.4.1.6 Determination Methods for Data Collection
and Analysis . . . . . . . . . . . . . . . . . . . . . . . 11
1.4.1.7 Determination of Study Quality Control
Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

vii
viii Contents

1.4.2 Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.4.3 Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.4.3.1 Evaluate the Validity and Reliability of the
Study Results . . . . . . . . . . . . . . . . . . . . . . . 12
1.4.3.2 Evaluate the Importance of Study Results . . . 12
1.5 Characteristics of Clinical Epidemiology . . . . . . . . . . . . . . . . . 13
1.5.1 Group . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5.2 Comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
1.5.3 Probability Theory and Mathematical Statistics . . . . . . 14
1.5.4 Social Psychology . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.5.5 Integrating Medicine . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.5.6 Development . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2 Distribution of Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Shan Zheng
2.1 Measures of Disease Frequency . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.1 Frequency Measures . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.1.1 Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.1.2 Proportion . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.1.3 Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.2 Morbidity Frequency Measures . . . . . . . . . . . . . . . . . . 18
2.1.2.1 Incidence Rate . . . . . . . . . . . . . . . . . . . . . . 18
2.1.2.2 Attack Rate . . . . . . . . . . . . . . . . . . . . . . . . 19
2.1.2.3 Secondary Attack Rate . . . . . . . . . . . . . . . . 20
2.1.2.4 Prevalence . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.1.3 Mortality Frequency Measures . . . . . . . . . . . . . . . . . . 21
2.1.3.1 Mortality Rate . . . . . . . . . . . . . . . . . . . . . . 21
2.1.3.2 Case Fatality Rate . . . . . . . . . . . . . . . . . . . . 22
2.1.3.3 Survival Rate . . . . . . . . . . . . . . . . . . . . . . . 22
2.2 Epidemic Disease Occurrence . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.2.1 Sporadic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.2.2 Epidemic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.2.3 Outbreak . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.3 Distribution of Disease by Time, Place, and Person . . . . . . . . . . 23
2.3.1 Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3.1.1 Rapid Fluctuation . . . . . . . . . . . . . . . . . . . . 24
2.3.1.2 Seasonality . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3.1.3 Periodicity . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.3.1.4 Secular Trend . . . . . . . . . . . . . . . . . . . . . . . 25
2.3.2 Place . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.3.2.1 Comparisons Among and Within
Countries . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.3.2.2 Urban-Rural Comparisons . . . . . . . . . . . . . . 29
2.3.2.3 Endemic Clustering . . . . . . . . . . . . . . . . . . . 30
2.3.2.4 Endemic Disease . . . . . . . . . . . . . . . . . . . . 30
Contents ix

2.3.3 Person . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.3.3.1 Age . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.3.3.2 Gender . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.3.3.3 Ethnic and Racial Group . . . . . . . . . . . . . . . 33
2.3.3.4 Occupation . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.3.3.5 Marital Status . . . . . . . . . . . . . . . . . . . . . . . 34
2.3.3.6 Behavior and Lifestyles . . . . . . . . . . . . . . . . 34
2.3.4 Combinations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3 Descriptive Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
Zhenxing Mao and Wenqian Huo
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.1.1 Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.1.2 Characteristics of Descriptive Studies . . . . . . . . . . . . . 38
3.1.3 Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.2 Case and Case Series Report . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.1 Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.2 Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.2.2.1 Identifying New Diseases . . . . . . . . . . . . . . 39
3.2.2.2 Establishing the Diagnosis . . . . . . . . . . . . . . 39
3.2.2.3 Forming an Etiological Hypothesis . . . . . . . 40
3.2.2.4 Identifying Early Disease Outbreaks and
Epidemics . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2.3 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2.3.1 Estrogen Chemical Bisphenol a and Breast
Cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2.3.2 Occupational Exposure to Vinyl Chloride
and Hepatic Hemangioma . . . . . . . . . . . . . . 41
3.2.3.3 AIDS Discovery Case . . . . . . . . . . . . . . . . . 41
3.2.4 Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.2.5 Limitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3 Cross-Sectional Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3.1 Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.3.2 Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.3.3 Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.3.3.1 Census . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.3.3.2 Sampling Survey . . . . . . . . . . . . . . . . . . . . 45
3.3.4 Design and Implementation . . . . . . . . . . . . . . . . . . . . . 46
3.3.4.1 Clarifying the Purpose and Type of
Investigation . . . . . . . . . . . . . . . . . . . . . . . . 46
3.3.4.2 Identifying Subjects . . . . . . . . . . . . . . . . . . 46
3.3.4.3 Determining Sample Size . . . . . . . . . . . . . . 46
3.3.4.4 Determining the Sampling Method . . . . . . . . 49
3.3.4.5 Data Collection, Collation, and Analysis . . . 51
x Contents

3.3.5 Bias and Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

3.3.5.1 Selection Bias . . . . . . . . . . . . . . . . . . . . . . . 54
3.3.5.2 Information Bias . . . . . . . . . . . . . . . . . . . . . 54
3.3.5.3 Confounding Bias . . . . . . . . . . . . . . . . . . . . 54
3.3.6 Strengths and Limitations . . . . . . . . . . . . . . . . . . . . . . 55
3.3.6.1 Strengths . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3.6.2 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3.7 Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.3.7.1 Purpose and Type of Study . . . . . . . . . . . . . 56
3.3.7.2 Subjects and Sample Size . . . . . . . . . . . . . . 56
3.3.7.3 Research Content and Data Collection,
Collation, and Analysis . . . . . . . . . . . . . . . . 56
3.3.7.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.4 Ecological Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.4.1 Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.4.2 Type of Study Design . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.4.2.1 Ecological Comparison Study . . . . . . . . . . . 57
3.4.2.2 Ecological Trend Study . . . . . . . . . . . . . . . . 57
3.4.3 The Main Application . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.4.4 Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.4.5 Strengths and Limitations . . . . . . . . . . . . . . . . . . . . . . 59
3.4.5.1 Strengths . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.4.5.2 Limitations . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.4.6 Case . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4 Cohort Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Li Liu
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.1.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.1.1.1 Observational Study . . . . . . . . . . . . . . . . . . 63
4.1.1.2 Setting up a Comparison Group . . . . . . . . . . 63
4.1.1.3 From “Cause” to “Outcome” . . . . . . . . . . . . 63
4.1.2 Types of Cohort Study . . . . . . . . . . . . . . . . . . . . . . . . 63
4.1.2.1 Prospective Cohort Study . . . . . . . . . . . . . . 63
4.1.2.2 Retrospective Cohort Study (Historical
Cohort Study) . . . . . . . . . . . . . . . . . . . . . . . 63
4.1.2.3 Ambispective Cohort Study . . . . . . . . . . . . . 64
4.2 Design of a Cohort Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2.1 Selection of the Cohort . . . . . . . . . . . . . . . . . . . . . . . . 65
4.2.1.1 Choice of the Exposure Population . . . . . . . 65
4.2.1.2 Choice of Control Population . . . . . . . . . . . 66
4.2.2 Determine the Sample Size . . . . . . . . . . . . . . . . . . . . . 68
4.2.2.1 Matters to Be Considered when Calculating
Sample Size . . . . . . . . . . . . . . . . . . . . . . . . 68
4.2.2.2 Four Factors Affecting Sample Size . . . . . . . 68
4.2.2.3 Calculation of Sample Size . . . . . . . . . . . . . 68
Contents xi

4.2.3 Follow-Up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
4.2.3.1 Purpose of Follow-Up . . . . . . . . . . . . . . . . . 69
4.2.3.2 Follow-up Methods . . . . . . . . . . . . . . . . . . . 69
4.2.3.3 Follow-up Contents . . . . . . . . . . . . . . . . . . 69
4.2.3.4 Endpoint of Observation . . . . . . . . . . . . . . . 70
4.2.3.5 Follow-Up Interval . . . . . . . . . . . . . . . . . . . 70
4.2.3.6 The Termination Time of Observation . . . . . 70
4.2.4 Quality Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.2.4.1 Selection and Training of the Investigators . . 70
4.2.4.2 Preparation of an Investigator’s Handbook . . 71
4.2.4.3 Supervision during the Follow-Up . . . . . . . . 71
4.3 Data Collection and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.3.1 Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.3.2 Measures of Outcome Frequency . . . . . . . . . . . . . . . . . 72
4.3.2.1 The Basic 2 × 2 Tables Summarizing the
Results of a Cohort Study . . . . . . . . . . . . . . 72
4.3.2.2 Person-Time . . . . . . . . . . . . . . . . . . . . . . . . 73
4.3.2.3 Standardized Mortality Ratio (SMR) . . . . . . 75
4.3.2.4 Statistical Tests . . . . . . . . . . . . . . . . . . . . . . 75
4.3.3 Measures of Association . . . . . . . . . . . . . . . . . . . . . . . 76
4.3.3.1 Relative Risk (RR) . . . . . . . . . . . . . . . . . . . 76
4.3.3.2 Attributable Risk (AR) and Attributable
Fraction (AF) . . . . . . . . . . . . . . . . . . . . . . . 77
4.3.3.3 Population Attributable Risk (PAR) and
Population Attributable Fraction (PAF) . . . . 78
4.3.3.4 Dose-Effect Relationship . . . . . . . . . . . . . . . 79
4.4 Common Bias and Controlling . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.4.1 Selection Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
4.4.2 Information Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.4.3 Confounding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.5 Advantages and Disadvantages of Cohort Studies . . . . . . . . . . . 81
4.5.1 Advantages of Cohort Studies . . . . . . . . . . . . . . . . . . . 81
4.5.2 Disadvantages of Cohort Studies . . . . . . . . . . . . . . . . . 81
4.6 Example of a Cohort Study . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5 Case-Control Studies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
Qian Wu
5.1 Overview of Case-Control Studies . . . . . . . . . . . . . . . . . . . . . . 83
5.1.1 History . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.1.2 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
5.1.3 Type of Design Case-Control Studies . . . . . . . . . . . . . 85
5.1.4 Characteristics of Case-Control Study . . . . . . . . . . . . . 86
5.1.5 Application . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.1.5.1 Example of a Case-Control Study . . . . . . . . 86
xii Contents

5.2 Design of Case-Control Studies . . . . . . . . . . . . . . . . . . . . . . . . 87

5.2.1 Basic Principles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
5.2.2 Selection of Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.2.3 Selection of Controls . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.2.3.1 Population Controls . . . . . . . . . . . . . . . . . . 89
5.2.3.2 Hospital or Disease Registry Controls . . . . . 89
5.2.4 Matching . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
5.2.4.1 Matching Type . . . . . . . . . . . . . . . . . . . . . . 90
5.2.4.2 Overmatching . . . . . . . . . . . . . . . . . . . . . . . 91
5.2.5 Exposure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
5.2.6 Sample Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
5.3 Data Collection and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.3.1 Main Analysis Objectives . . . . . . . . . . . . . . . . . . . . . . 94
5.3.2 Descriptive Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.3.3 Statistical Inference . . . . . . . . . . . . . . . . . . . . . . . . . . 94
5.3.3.1 Unmatched (Frequency Matching) Design . . 95
5.3.3.2 Matched Design . . . . . . . . . . . . . . . . . . . . . 98
5.4 Common Bias and Controlling . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.4.1 Selection Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
5.4.1.1 Prevalence-Incidence Bias . . . . . . . . . . . . . . 101
5.4.1.2 Unmasking Bias . . . . . . . . . . . . . . . . . . . . . 101
5.4.1.3 Subject Refuses Participation . . . . . . . . . . . . 101
5.4.2 Information Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.4.3 Confounding Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
5.5 Strengths and Weaknesses of Case-Control Studies . . . . . . . . . . 103
5.5.1 Advantage of the Case-Control Study . . . . . . . . . . . . . 103
5.5.2 Disadvantage of the Case-Control Study . . . . . . . . . . . 104
6 Experimental Epidemiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
Xing Liu
6.1 Basic Ideas of Experimental Study . . . . . . . . . . . . . . . . . . . . . . 106
6.1.1 Study Question . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.1.2 Choice of Intervention . . . . . . . . . . . . . . . . . . . . . . . . 107
6.1.3 Choice of Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.1.3.1 Standard Control . . . . . . . . . . . . . . . . . . . . . 107
6.1.3.2 Placebo Control . . . . . . . . . . . . . . . . . . . . . 108
6.1.3.3 Self-Control . . . . . . . . . . . . . . . . . . . . . . . . 108
6.1.3.4 Cross-Over Control . . . . . . . . . . . . . . . . . . . 108
6.1.4 Randomization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
6.1.4.1 The Randomization Process . . . . . . . . . . . . . 110
6.1.4.2 Simple Randomization . . . . . . . . . . . . . . . . 110
6.1.4.3 Blocked Randomization . . . . . . . . . . . . . . . 110
6.1.4.4 Stratified Randomization . . . . . . . . . . . . . . . 111
Contents xiii

6.1.5 Blinding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

6.1.5.1 Single-Blind . . . . . . . . . . . . . . . . . . . . . . . . 112
6.1.5.2 Double-Blind . . . . . . . . . . . . . . . . . . . . . . . 112
6.1.5.3 Triple-Blind . . . . . . . . . . . . . . . . . . . . . . . . 112
6.1.6 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112
6.1.7 Sample Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
6.1.7.1 Sample Size Calculation for Dichotomous
Response Variables . . . . . . . . . . . . . . . . . . . 114
6.1.7.2 Sample Size Calculation for Continuous
Response Variables . . . . . . . . . . . . . . . . . . . 114
6.1.7.3 Sample Size Calculation for
“Time to Failure” . . . . . . . . . . . . . . . . . . . . 114
6.2 Clinical Trial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.2.1 Basic Ideas of Clinical Trial . . . . . . . . . . . . . . . . . . . . 114
6.2.2 Phases of Clinical Trial . . . . . . . . . . . . . . . . . . . . . . . . 115
6.2.2.1 Phase I Studies . . . . . . . . . . . . . . . . . . . . . . 115
6.2.2.2 Phase II Studies . . . . . . . . . . . . . . . . . . . . . 115
6.2.2.3 Phase III Studies . . . . . . . . . . . . . . . . . . . . . 116
6.2.2.4 Phase IV Studies . . . . . . . . . . . . . . . . . . . . . 116
6.2.3 Case Study of Clinical Trial . . . . . . . . . . . . . . . . . . . . 116
6.3 Field Trial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
6.3.1 Basic Ideas of Field Trial . . . . . . . . . . . . . . . . . . . . . . 117
6.3.2 Design and Implementation . . . . . . . . . . . . . . . . . . . . . 117
6.3.2.1 A specified Question . . . . . . . . . . . . . . . . . . 117
6.3.2.2 Inclusion and Exclusion Criteria . . . . . . . . . 118
6.3.2.3 Choice of Intervention . . . . . . . . . . . . . . . . . 118
6.3.2.4 Time and Interval of Follow-up . . . . . . . . . . 118
6.3.3 Case Study of Field Trial . . . . . . . . . . . . . . . . . . . . . . 118
6.4 Community Trial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
6.4.1 Basic Ideas of Community Trial . . . . . . . . . . . . . . . . . 119
6.4.2 Case Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
7 Screening and Diagnostic Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
Fen Liu
7.1 Design a Screening or Diagnostic Test . . . . . . . . . . . . . . . . . . . 124
7.1.1 Gold Standard (Reference Standard) . . . . . . . . . . . . . . 124
7.1.2 Study Subjects . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
7.1.3 Sample Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125
7.2 Evaluation of a Screening Test . . . . . . . . . . . . . . . . . . . . . . . . . 125
7.2.1 Validity of a Screening Test . . . . . . . . . . . . . . . . . . . . 126
7.2.1.1 Sensitivity . . . . . . . . . . . . . . . . . . . . . . . . . 126
7.2.1.2 Specificity . . . . . . . . . . . . . . . . . . . . . . . . . 126
7.2.1.3 Youden’s Index . . . . . . . . . . . . . . . . . . . . . 127
7.2.1.4 Likelihood Ratio . . . . . . . . . . . . . . . . . . . . . 127
xiv Contents

7.2.2 Evaluation of the Reliability of a Test . . . . . . . . . . . . . 128

7.2.2.1 Coefficient of Variation . . . . . . . . . . . . . . . . 128
7.2.2.2 Agreement Rate and Kappa Statistic . . . . . . . 128
7.2.3 Predictive Value . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
7.2.4 Determination of Cutoff Point for a Screening Test . . . . 132
7.2.4.1 ROC Curve . . . . . . . . . . . . . . . . . . . . . . . . 133
7.2.4.2 The Area under ROC Curve . . . . . . . . . . . . 134
7.3 Improving the Efficiency of Screening and Diagnostic Tests . . . 134
7.3.1 Selecting Population with a High Prevalence . . . . . . . . 134
7.3.2 Use of Multiple Tests . . . . . . . . . . . . . . . . . . . . . . . . . 134
7.3.2.1 Simultaneous Testing . . . . . . . . . . . . . . . . . 134
7.3.2.2 Sequential Testing . . . . . . . . . . . . . . . . . . . 135
7.4 Potential Bias in Screening Tests . . . . . . . . . . . . . . . . . . . . . . . 136
7.4.1 Volunteer Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.4.2 Lead-Time Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136
7.4.3 Length-Time Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
8 Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
Lu Long
8.1 Introduction of Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139
8.2 Selection Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
8.2.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
8.2.2 Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
8.2.2.1 Self-Selection Bias . . . . . . . . . . . . . . . . . . . 140
8.2.2.2 Berksonian Bias . . . . . . . . . . . . . . . . . . . . . 141
8.2.2.3 Detection Signal Bias . . . . . . . . . . . . . . . . . 142
8.2.2.4 Neyman Bias . . . . . . . . . . . . . . . . . . . . . . . 143
8.2.2.5 Loss of Follow-Up . . . . . . . . . . . . . . . . . . . 143
8.2.3 Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143
8.2.3.1 Scientific Research Design . . . . . . . . . . . . . 143
8.2.3.2 Develop Strict Inclusion and Exclusion
Standards . . . . . . . . . . . . . . . . . . . . . . . . . . 144
8.2.3.3 Maximize Response Rates . . . . . . . . . . . . . . 144
8.2.3.4 Randomization Principle . . . . . . . . . . . . . . . 144
8.3 Information Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
8.3.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
8.3.2 Classification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
8.3.2.1 Differential Misclassification . . . . . . . . . . . . 145
8.3.2.2 Nondifferential Misclassification . . . . . . . . . 146
8.3.3 Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147
8.3.3.1 Material Collection . . . . . . . . . . . . . . . . . . . 147
8.3.3.2 Objective Research Indicators . . . . . . . . . . . 148
8.3.3.3 Investigation Skills . . . . . . . . . . . . . . . . . . . 148
Contents xv

8.4 Confounding Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

8.4.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148
8.4.2 Confounding . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
8.4.3 Control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
8.4.3.1 Random Allocation . . . . . . . . . . . . . . . . . . . 150
8.4.3.2 Restrict . . . . . . . . . . . . . . . . . . . . . . . . . . . 150
8.4.3.3 Matching . . . . . . . . . . . . . . . . . . . . . . . . . . 151
8.4.3.4 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . 151
9 Cause of Disease and Causal Inference . . . . . . . . . . . . . . . . . . . . . . 153
Li Ye
9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153
9.2 Cause of Disease in Epidemiology . . . . . . . . . . . . . . . . . . . . . . 154
9.2.1 The Concept of Cause in Epidemiology and its
Development History . . . . . . . . . . . . . . . . . . . . . . . . . 154
9.2.2 Classification of Cause . . . . . . . . . . . . . . . . . . . . . . . . 155
9.2.3 Causation Models . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156
9.2.3.1 Triangle Model . . . . . . . . . . . . . . . . . . . . . . 156
9.2.3.2 Wheel Model . . . . . . . . . . . . . . . . . . . . . . . 157
9.2.3.3 Chain of Causation Model . . . . . . . . . . . . . . 157
9.2.3.4 Web of Causation Model . . . . . . . . . . . . . . . 158
9.2.4 Sufficient Cause and Necessary Cause . . . . . . . . . . . . . 158
9.3 Epidemiologic Methods of Causation . . . . . . . . . . . . . . . . . . . . 160
9.3.1 Epidemiologic Study Designs for Causation . . . . . . . . . 160
9.3.1.1 Descriptive Studies . . . . . . . . . . . . . . . . . . . 160
9.3.1.2 Analytical Studies (Case-Control Studies,
Cohort Studies) . . . . . . . . . . . . . . . . . . . . . . 161
9.3.1.3 Experimental Studies . . . . . . . . . . . . . . . . . 161
9.3.2 Mill’s Canons-the Logical Basis of Causation . . . . . . . 162
9.3.2.1 Method of Agreement . . . . . . . . . . . . . . . . . 162
9.3.2.2 Method of Difference . . . . . . . . . . . . . . . . . 162
9.3.2.3 Joint Methods of Agreement and
Difference . . . . . . . . . . . . . . . . . . . . . . . . . 163
9.3.2.4 Method of Concomitant Variations . . . . . . . 163
9.3.2.5 Method of Residue . . . . . . . . . . . . . . . . . . . 164
9.4 Causal Inference . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
9.4.1 Association Vs. Causation . . . . . . . . . . . . . . . . . . . . . 164
9.4.1.1 Chance Association . . . . . . . . . . . . . . . . . . 165
9.4.1.2 Spurious Association . . . . . . . . . . . . . . . . . 165
9.4.1.3 Noncausal Association . . . . . . . . . . . . . . . . 166
9.4.1.4 Causal Association . . . . . . . . . . . . . . . . . . . 166
9.4.2 Evaluating Causal Association—Hill’s Criteria . . . . . . . 167
9.4.2.1 Temporal Relationship . . . . . . . . . . . . . . . . 167
9.4.2.2 Strength of Association . . . . . . . . . . . . . . . . 168
xvi Contents

9.4.2.3 Dose-Response Relationship . . . . . . . . . . . . 168

9.4.2.4 Consistency . . . . . . . . . . . . . . . . . . . . . . . . 169
9.4.2.5 Biologic Plausibility . . . . . . . . . . . . . . . . . . 169
9.4.2.6 Reversibility . . . . . . . . . . . . . . . . . . . . . . . . 169
9.4.2.7 Specificity . . . . . . . . . . . . . . . . . . . . . . . . . 169
9.4.2.8 Analogy . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
9.4.2.9 Experimental Evidence . . . . . . . . . . . . . . . . 170
9.4.3 An Example of Causal Inference Using Hill’s
Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
9.4.3.1 Temporality of Association . . . . . . . . . . . . . 171
9.4.3.2 Strength of Association . . . . . . . . . . . . . . . . 171
9.4.3.3 Consistency . . . . . . . . . . . . . . . . . . . . . . . . 171
9.4.3.4 Dose-Response Relationship . . . . . . . . . . . . 171
9.4.3.5 Biologic Plausibility . . . . . . . . . . . . . . . . . . 171
9.4.3.6 Experimental Evidence . . . . . . . . . . . . . . . . 171
10 Disease Prevention and Surveillance . . . . . . . . . . . . . . . . . . . . . . . . 173
Chunhua Song
10.1 Prevention Strategies and Measures . . . . . . . . . . . . . . . . . . . . . 173
10.1.1 Strategy and Implementation for Prevention . . . . . . . . . 173
10.1.2 Disease Prevention . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
10.1.2.1 The Definition of Disease Prevention . . . . . . 174
10.1.2.2 The Development of Disease . . . . . . . . . . . . 174
10.1.2.3 The Three Levels of Prevention of Disease . . 175
10.1.3 Health Protection and Promotion . . . . . . . . . . . . . . . . . 177
10.1.3.1 Health Protection . . . . . . . . . . . . . . . . . . . . 177
10.1.3.2 Health Education . . . . . . . . . . . . . . . . . . . . 178
10.1.3.3 Health Management . . . . . . . . . . . . . . . . . . 178
10.1.3.4 Health Promotion . . . . . . . . . . . . . . . . . . . . 179
10.1.3.5 Global Health Strategies and Practice . . . . . . 179
10.2 Public Health Monitoring . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
10.2.1 Introduction of Public Health Surveillance . . . . . . . . . . 181
10.2.1.1 The Basic Concept of Public Health
Surveillance . . . . . . . . . . . . . . . . . . . . . . . . 181
10.2.1.2 The Purpose and Application of Public
Health Surveillance . . . . . . . . . . . . . . . . . . . 182
10.2.2 Categories of Public Health Surveillance . . . . . . . . . . . 184
10.2.2.1 Surveillance of Disease . . . . . . . . . . . . . . . . 184
10.2.2.2 Symptom Surveillance . . . . . . . . . . . . . . . . 186
10.2.3 Methods of Public Health Surveillance . . . . . . . . . . . . 187
10.2.3.1 Surveillance Methods . . . . . . . . . . . . . . . . . 187
10.2.3.2 Surveillance Methods and Techniques . . . . . 189
10.2.3.3 Attention in Public Health Surveillance . . . . 191
Contents xvii

10.2.4 Procedures and Assessment of Public Health

Surveillance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192
10.2.4.1 Basic Procedures of Public Health
Surveillance . . . . . . . . . . . . . . . . . . . . . . . . 192
10.2.4.2 Evaluation of Public Health Surveillance
System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
11 Communicable Diseases Epidemiology . . . . . . . . . . . . . . . . . . . . . . 197
Rongguang Zhang
11.1 Infection Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
11.1.1 Infection Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . 198
11.1.2 Spectrum of Infection . . . . . . . . . . . . . . . . . . . . . . . . . 198
11.2 Epidemic Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
11.2.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199
11.2.2 Three Links in the Epidemic Process . . . . . . . . . . . . . . 199
11.2.2.1 Sources of Infection . . . . . . . . . . . . . . . . . . 199
11.2.2.2 Routes of Transmission . . . . . . . . . . . . . . . . 200
11.2.2.3 Herd Susceptibility . . . . . . . . . . . . . . . . . . . 201
11.2.3 Two Factors Affecting the Epidemic Process . . . . . . . . 201
11.2.3.1 Natural Factors . . . . . . . . . . . . . . . . . . . . . . 202
11.2.3.2 Social Factors . . . . . . . . . . . . . . . . . . . . . . . 202
11.2.4 Epidemic Focus and Epidemic Process . . . . . . . . . . . . 202
11.2.4.1 Epidemic Focus . . . . . . . . . . . . . . . . . . . . . 202
11.2.4.2 Epidemic Process . . . . . . . . . . . . . . . . . . . . 203
11.3 Strategy and Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . 203
11.3.1 Strategies for Control of Communicable Diseases . . . . . 203
11.3.1.1 Population Strategy . . . . . . . . . . . . . . . . . . . 204
11.3.1.2 High-Risk Strategy . . . . . . . . . . . . . . . . . . . 204
11.3.2 Measures for Control of Communicable Diseases . . . . . 205
11.3.2.1 Surveillance of Communicable Diseases . . . . 205
11.3.2.2 Measures on Sources of Infection . . . . . . . . 205
11.3.2.3 Measures on Routes of Infection . . . . . . . . . 206
11.3.2.4 Measures on Susceptible Populations . . . . . . 206
11.4 Immunization Program and Effectiveness . . . . . . . . . . . . . . . . . 207
11.4.1 Immunization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
11.4.2 Immunization Program . . . . . . . . . . . . . . . . . . . . . . . . 208
11.4.3 Evaluation of Immune Effectiveness . . . . . . . . . . . . . . 208
11.5 Emerging Communicable Diseases . . . . . . . . . . . . . . . . . . . . . . 209
11.5.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
11.5.2 Main Emerging Communicable Diseases . . . . . . . . . . . 209
11.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
12 Epidemiology of Noncommunicable Diseases . . . . . . . . . . . . . . . . . . 213
Jie Yang and Man Li
12.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
12.1.1 Definition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213
12.1.2 The Influence of NCDs on Health and Society . . . . . . . 214
xviii Contents

12.2 Epidemiological Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 214

12.2.1 Overall Global NCDs Outlook . . . . . . . . . . . . . . . . . . 214
12.2.2 Epidemiological Features of the Risk Factors of
NCDs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 216
12.2.2.1 Tobacco Use . . . . . . . . . . . . . . . . . . . . . . . 216
12.2.2.2 Alcohol Use . . . . . . . . . . . . . . . . . . . . . . . . 217
12.2.2.3 Unhealthy Diet . . . . . . . . . . . . . . . . . . . . . . 217
12.2.2.4 Physical Inactivity . . . . . . . . . . . . . . . . . . . 217
12.2.2.5 Raised Blood Pressure . . . . . . . . . . . . . . . . 218
12.2.2.6 Overweight/Obesity . . . . . . . . . . . . . . . . . . 218
12.3 Risk Factors of Several Common NCDs . . . . . . . . . . . . . . . . . . 218
12.3.1 Cardiovascular and Cerebrovascular Disease . . . . . . . . 218
12.3.1.1 Stroke . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
12.3.1.2 Coronary Heart Disease . . . . . . . . . . . . . . . . 219
12.3.2 T2DM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220
12.3.2.1 Genetic Factor . . . . . . . . . . . . . . . . . . . . . . 220
12.3.2.2 Overweight/Obesity . . . . . . . . . . . . . . . . . . 221
12.3.2.3 Physical Inactivity . . . . . . . . . . . . . . . . . . . 221
12.3.2.4 Unhealthy Diet . . . . . . . . . . . . . . . . . . . . . . 221
12.3.2.5 Malnutrition . . . . . . . . . . . . . . . . . . . . . . . . 222
12.3.2.6 Impaired Glucose Tolerance (IGT) . . . . . . . . 222
12.3.2.7 Insulin Resistance . . . . . . . . . . . . . . . . . . . . 222
12.3.2.8 Maternal Diabetes . . . . . . . . . . . . . . . . . . . . 222
12.3.3 Cancer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
12.3.3.1 Physical Factors . . . . . . . . . . . . . . . . . . . . . 223
12.3.3.2 Tobacco Use . . . . . . . . . . . . . . . . . . . . . . . 223
12.3.3.3 Alcohol Use . . . . . . . . . . . . . . . . . . . . . . . . 223
12.3.3.4 Dietary Factors . . . . . . . . . . . . . . . . . . . . . . 223
12.3.3.5 Occupational Exposures . . . . . . . . . . . . . . . 224
12.3.3.6 Biological Factors . . . . . . . . . . . . . . . . . . . . 224
12.3.3.7 Genetic Factors . . . . . . . . . . . . . . . . . . . . . . 224
12.3.3.8 Other Factors . . . . . . . . . . . . . . . . . . . . . . . 224
12.4 Prevention and Control of NCDs . . . . . . . . . . . . . . . . . . . . . . . 225
12.4.1 Prevention Strategy . . . . . . . . . . . . . . . . . . . . . . . . . . 225
12.4.2 Prevention Measures . . . . . . . . . . . . . . . . . . . . . . . . . 226
13 Epidemiology of Public Health Emergencies . . . . . . . . . . . . . . . . . . 227
Hong Zhu
13.1 Basic Conception of Public Health Emergencies . . . . . . . . . . . . 228
13.1.1 Definition of Public Health Emergencies . . . . . . . . . . . 228
13.1.2 Characteristics of Public Health Emergencies . . . . . . . . 228
13.1.3 Classification of Public Health Emergencies . . . . . . . . . 229
13.1.4 Phases of Public Health Emergencies . . . . . . . . . . . . . . 230
13.1.5 Harm Caused by Public Health Emergencies . . . . . . . . 231
Contents xix

13.2 Basic Principles and Application of Epidemiology in Public

Health Emergencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232
13.2.1 Role of Epidemiology in Public Health Emergencies . . 232
13.2.2 Key Epidemiological Indicators . . . . . . . . . . . . . . . . . . 232
13.2.3 Outbreak Investigation . . . . . . . . . . . . . . . . . . . . . . . . 233
13.2.3.1 Purpose of Outbreak Investigation . . . . . . . . 233
13.2.3.2 Three Elemental Epidemiological Designs
in an Outbreak Investigation . . . . . . . . . . . . 234
13.2.3.3 Key Steps in Carrying Out Outbreak
Investigation . . . . . . . . . . . . . . . . . . . . . . . . 234
13.2.4 Disaster Investigation . . . . . . . . . . . . . . . . . . . . . . . . . 236
13.2.4.1 Purpose of Disaster Investigation . . . . . . . . . 236
13.2.4.2 Key Steps in Carrying Out Disaster
Investigation . . . . . . . . . . . . . . . . . . . . . . . . 237
13.3 Public Health Emergency Preparedness . . . . . . . . . . . . . . . . . . 238
13.3.1 Definition of Public Health Emergency Preparedness . . 238
13.3.2 Significance of Public Health Emergency
Preparedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
13.3.3 The Main Activities of Public Health Emergency
Preparedness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 239
13.4 Public Health Emergency Response . . . . . . . . . . . . . . . . . . . . . 242
13.4.1 Definition of Public Health Emergency Response . . . . . 242
13.4.2 Significance of Public Health Emergency Response . . . 242
13.4.3 The Main Activities of Public Health Emergency
Response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242
13.4.3.1 Ensuring Availability of Preventive and
Emergency Medical Treatment . . . . . . . . . . 242
13.4.3.2 Preventing Secondary Public Health
Emergencies After Disaster . . . . . . . . . . . . . 243
13.4.3.3 Interrupting the Route of Transmission . . . . . 243
13.4.3.4 Remediating of Environmental Health
Conditions . . . . . . . . . . . . . . . . . . . . . . . . . 243
13.4.3.5 Performing Laboratory Analyses to Support
Epidemiology and Surveillance . . . . . . . . . . 244
13.4.3.6 Communicating with Media and Delivering
Message to the Public . . . . . . . . . . . . . . . . . 244
13.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244
14 Molecular Epidemiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
Hui Wang
14.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
14.1.1 Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 247
14.1.2 Characteristic . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
14.2 Classes of Biomarkers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
14.2.1 Biomarkers of Exposure . . . . . . . . . . . . . . . . . . . . . . . 248
14.2.2 Biomarkers of Effects . . . . . . . . . . . . . . . . . . . . . . . . . 249
xx Contents

14.2.3 Biomarkers of Susceptibility . . . . . . . . . . . . . . . . . . . . 250

14.2.4 Biomarker Selection . . . . . . . . . . . . . . . . . . . . . . . . . . 250
14.3 Main Research Methods Used in Molecular Epidemiology . . . . . 252
14.3.1 Study Design in Molecular Epidemiology . . . . . . . . . . 252
14.3.1.1 Cross-Sectional Studies . . . . . . . . . . . . . . . . 252
14.3.1.2 Case-Control Studies . . . . . . . . . . . . . . . . . . 253
14.3.1.3 Cohort Studies . . . . . . . . . . . . . . . . . . . . . . 253
14.3.2 Main Molecular Methods Used in Molecular
Epidemiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
14.3.2.1 Electrophoretic Mobility Shift Assay
(EMSA) . . . . . . . . . . . . . . . . . . . . . . . . . . . 254
14.3.2.2 Dual Luciferase Reporter Assay . . . . . . . . . . 255
14.3.2.3 The Comet Assay . . . . . . . . . . . . . . . . . . . . 256
14.3.2.4 Micronucleus (MN) Assay . . . . . . . . . . . . . 256
14.3.3 Genome-Wide Association Studies (GWAS) . . . . . . . . 257
14.3.4 Mendelian Randomization (MR) . . . . . . . . . . . . . . . . . 258
14.4 Application and Prospection . . . . . . . . . . . . . . . . . . . . . . . . . . 259
14.4.1 Control and Prevention of Infectious Diseases . . . . . . . 259
14.4.1.1 Outbreak Investigation . . . . . . . . . . . . . . . . 259
14.4.1.2 Trace Dissemination of a Specific Subtype
of Pathogen Across Time and Space . . . . . . 260
14.4.1.3 Determine the Origin of an Epidemic . . . . . . 260
14.4.1.4 Follow the Emergence and Spread of New
Infections . . . . . . . . . . . . . . . . . . . . . . . . . . 261
14.4.1.5 Identify Previously Unknown or
Uncultivable Infectious Microbes . . . . . . . . . 261
14.4.2 Control and Prevention of Chronic Diseases . . . . . . . . . 262
14.4.2.1 Improving the Understanding of Mechanism
of Pathogenesis . . . . . . . . . . . . . . . . . . . . . . 262
14.4.2.2 Evaluating the Susceptibility of Individual
and Defining the Risk Population . . . . . . . . . 263
14.4.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263
15 Pharmacoepidemiology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 265
Xiaotian Liu and Jian Hou
15.1 A Brief History and Definition . . . . . . . . . . . . . . . . . . . . . . . . . 265
15.1.1 A Brief History of Pharmacoepidemiology . . . . . . . . . . 265
15.1.2 Definition of Pharmacoepidemiology . . . . . . . . . . . . . . 267
15.1.3 Drug-Related Concepts . . . . . . . . . . . . . . . . . . . . . . . . 267
15.2 Main Research Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 268
15.2.1 Drug Safety Evaluation . . . . . . . . . . . . . . . . . . . . . . . . 268
15.2.2 Drug Effectiveness Evaluation . . . . . . . . . . . . . . . . . . 269
15.2.3 Drug Utilization Study . . . . . . . . . . . . . . . . . . . . . . . . 269
15.2.4 Pharmacoeconomic Study . . . . . . . . . . . . . . . . . . . . . . 269
Contents xxi

15.3 Aims and Significances . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 270

15.3.1 Aims of Pharmacoepidemiology . . . . . . . . . . . . . . . . . 270
15.3.2 Significances of pharmacoepidemiology . . . . . . . . . . . . 271
15.3.2.1 Improve the Quality of Premarketing
Clinical Trials . . . . . . . . . . . . . . . . . . . . . . . 271
15.3.2.2 Post-Marketing Study of Drug . . . . . . . . . . . 271
15.4 Methods of Pharmacoepidemiology . . . . . . . . . . . . . . . . . . . . . 272
15.4.1 Case Report and Case Series Study . . . . . . . . . . . . . . . 272
15.4.2 Ecological Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 272
15.4.3 Cross-Sectional Study . . . . . . . . . . . . . . . . . . . . . . . . . 274
15.4.4 Case-Control Study . . . . . . . . . . . . . . . . . . . . . . . . . . 274
15.4.5 Cohort Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274
15.4.6 Experimental Study . . . . . . . . . . . . . . . . . . . . . . . . . . 275
15.4.7 Systematic Review and Meta-Analysis . . . . . . . . . . . . . 276
15.4.8 Real-world Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . 276
15.4.9 Newly Derived Study Design . . . . . . . . . . . . . . . . . . . 276
15.5 Data Collection and Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . 277
15.5.1 Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 277
15.5.1.1 Routine Data . . . . . . . . . . . . . . . . . . . . . . . 277
15.5.1.2 The Literature . . . . . . . . . . . . . . . . . . . . . . . 278
15.5.1.3 ADR Monitoring and Reporting System . . . . 278
15.5.2 Data Processing and Analysis . . . . . . . . . . . . . . . . . . . 278
15.5.2.1 Mining and Analysis of ADR Monitoring
Database . . . . . . . . . . . . . . . . . . . . . . . . . . 279
15.5.2.2 Mining and Analysis of Prescription
Database . . . . . . . . . . . . . . . . . . . . . . . . . . 279
16 Evidence-Based Medicine and Systematic Review . . . . . . . . . . . . . . 281
Qi Gao and Huiping Zhu
16.1 Evidence-Based Medicine . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
16.1.1 Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281
16.1.2 Development of EBM . . . . . . . . . . . . . . . . . . . . . . . . . 282
16.1.3 Categories of EBM Practice . . . . . . . . . . . . . . . . . . . . 284
16.1.4 Procedures of Practicing EBM . . . . . . . . . . . . . . . . . . 285
16.2 Systematic Review and Meta-Analysis . . . . . . . . . . . . . . . . . . . 287
16.2.1 Systematic Review . . . . . . . . . . . . . . . . . . . . . . . . . . . 287
16.2.1.1 Cochrane Systematic Review . . . . . . . . . . . . 288
16.2.1.2 Importance of Systematic Review . . . . . . . . 288
16.2.1.3 The Difference Between Systematic Review
and Traditional Review . . . . . . . . . . . . . . . . 289
16.2.1.4 How to Do a Systematic Review? . . . . . . . . 289
16.2.1.5 Evaluation and Application of Systematic
Review . . . . . . . . . . . . . . . . . . . . . . . . . . . 294
16.2.1.6 Methods of Evaluating System Review . . . . 298
16.2.1.7 Application of Systematic Review . . . . . . . . 299
xxii Contents

16.3 Meta-Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301

16.3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 301
16.3.1.1 Basic Concepts . . . . . . . . . . . . . . . . . . . . . . 301
16.3.2 Steps to Perform a Meta-Analysis . . . . . . . . . . . . . . . . 301
16.3.2.1 Data Extraction . . . . . . . . . . . . . . . . . . . . . . 301
16.3.2.2 Data Types and Effect Size . . . . . . . . . . . . . 302
16.3.2.3 Heterogeneity Test . . . . . . . . . . . . . . . . . . . 302
16.3.2.4 Combining Effect Size Estimates and
Hypothesis Testing . . . . . . . . . . . . . . . . . . . 303
16.3.3 Fixed Effect Model and Random Effect Model . . . . . . . 303
16.3.3.1 Fixed Effect Model . . . . . . . . . . . . . . . . . . . 303
16.3.3.2 Random Effect Model . . . . . . . . . . . . . . . . . 304
16.3.4 Evaluating the Result of a Meta-Analysis . . . . . . . . . . . 304
16.3.4.1 Heterogeneity Test . . . . . . . . . . . . . . . . . . . 304
16.3.4.2 Robustness of Meta-Analysis Results . . . . . . 304
16.3.4.3 Applicability of the Meta-Analysis Results . . 305
17 Disease Prognosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 307
Fang Wang
17.1 Basic Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 308
17.1.1 Concept of Prognosis . . . . . . . . . . . . . . . . . . . . . . . . . 308
17.1.2 Natural History and Clinical Course of Disease . . . . . . 308
17.1.3 Prognostic Factors . . . . . . . . . . . . . . . . . . . . . . . . . . . 309
17.2 Design of a Prognosis Study . . . . . . . . . . . . . . . . . . . . . . . . . . 310
17.2.1 Research Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
17.2.2 The Patient Sample . . . . . . . . . . . . . . . . . . . . . . . . . . 310
17.2.3 Determine the Starting Time Point . . . . . . . . . . . . . . . . 311
17.2.4 Determine the Outcomes . . . . . . . . . . . . . . . . . . . . . . . 311
17.2.5 Sample Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
17.2.6 Follow-Up . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
17.3 Describing Prognosis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
17.3.1 Case Fatality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
17.3.2 Remission Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312
17.3.3 Recurrence Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
17.3.4 Disability Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
17.3.5 Quality of Life . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313
17.4 Analysis for Prognosis Study Data . . . . . . . . . . . . . . . . . . . . . . 314
17.4.1 Calculating Survival Rate . . . . . . . . . . . . . . . . . . . . . . 314
17.4.1.1 Five-Year Survival . . . . . . . . . . . . . . . . . . . 314
17.4.1.2 Life Tables . . . . . . . . . . . . . . . . . . . . . . . . . 314
17.4.1.3 Kaplan–Meier Analysis . . . . . . . . . . . . . . . . 315
17.4.2 Several Points About Interpreting Survival Curves . . . . 316
17.4.3 Comparison of the Survival Curves . . . . . . . . . . . . . . . 316
17.4.4 Dealing with Multiple Prognostic Factors . . . . . . . . . . . 317
Contents xxiii

17.5 Common Bias and Controlling . . . . . . . . . . . . . . . . . . . . . . . . . 317

17.5.1 Bias in Prognosis Study . . . . . . . . . . . . . . . . . . . . . . . 317
17.5.1.1 Assembly Bias . . . . . . . . . . . . . . . . . . . . . . 317
17.5.1.2 Migration . . . . . . . . . . . . . . . . . . . . . . . . . . 317
17.5.1.3 Loss to Follow-Up . . . . . . . . . . . . . . . . . . . 318
17.5.1.4 Survival Cohort Bias . . . . . . . . . . . . . . . . . . 318
17.5.1.5 Zero Bias . . . . . . . . . . . . . . . . . . . . . . . . . . 318
17.5.1.6 Measurement Bias . . . . . . . . . . . . . . . . . . . 318
17.5.2 Control of Bias . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319
17.5.2.1 Randomization . . . . . . . . . . . . . . . . . . . . . . 319
17.5.2.2 Matching . . . . . . . . . . . . . . . . . . . . . . . . . . 319
17.5.2.3 Restriction . . . . . . . . . . . . . . . . . . . . . . . . . 319
17.5.2.4 Stratification . . . . . . . . . . . . . . . . . . . . . . . . 320
17.5.2.5 Standardization . . . . . . . . . . . . . . . . . . . . . . 320
17.5.2.6 Multivariable Analysis . . . . . . . . . . . . . . . . 320
17.5.2.7 Other Methods to Control Bias in
Prognosis . . . . . . . . . . . . . . . . . . . . . . . . . . 321
18 Nosocomial Infections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
Zhijiang Zhang
18.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323
18.2 Definition and Diagnostic Standards . . . . . . . . . . . . . . . . . . . . . 324
18.3 Nosocomial Infection Sites . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
18.3.1 Surgical Sites . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
18.3.2 Respiratory System . . . . . . . . . . . . . . . . . . . . . . . . . . 325
18.3.3 Bacteremia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325
18.3.4 Urinary Tract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
18.4 Microorganisms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
18.4.1 Normal Microorganisms in Nosocomial Infections . . . . 326
18.4.1.1 Bacteria . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
18.4.1.2 Viruses . . . . . . . . . . . . . . . . . . . . . . . . . . . 326
18.4.1.3 Parasites and Fungi . . . . . . . . . . . . . . . . . . . 326
18.4.2 Antimicrobial Resistance and Nosocomial Infections . . 327
18.5 Categories of Nosocomial Infections . . . . . . . . . . . . . . . . . . . . 327
18.5.1 Endogenous Infections . . . . . . . . . . . . . . . . . . . . . . . . 327
18.5.2 Exogenous Infections . . . . . . . . . . . . . . . . . . . . . . . . . 327
18.6 Epidemic Process of Nosocomial Infection . . . . . . . . . . . . . . . . 328
18.6.1 Source of Infection . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
18.6.2 Route of Transmission . . . . . . . . . . . . . . . . . . . . . . . . 328
18.6.3 Susceptible Population . . . . . . . . . . . . . . . . . . . . . . . . 328
18.7 Prevention of Nosocomial Infections . . . . . . . . . . . . . . . . . . . . 329
18.7.1 Preventing Human-to-Human Transmission . . . . . . . . . 329
18.7.1.1 Hand Decontamination . . . . . . . . . . . . . . . . 329
18.7.1.2 Clothing . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
18.7.1.3 Masks . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
xxiv Contents

18.7.1.4Gloves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
18.7.1.5Safe Injection and Other Skin-Piercing
Practice . . . . . . . . . . . . . . . . . . . . . . . . . . . 330
18.7.2 Preventing Transmission from Environment . . . . . . . . . 330
18.7.2.1 Routine Cleaning . . . . . . . . . . . . . . . . . . . . 330
18.7.2.2 Disinfection of Equipment . . . . . . . . . . . . . . 330
18.7.2.3 Sterilization . . . . . . . . . . . . . . . . . . . . . . . . 331
18.8 Surveillance of Nosocomial Infections . . . . . . . . . . . . . . . . . . . 331
18.8.1 Objectives of Surveillance Programs . . . . . . . . . . . . . . 331
18.8.2 Implementation of Surveillance Programs . . . . . . . . . . 331
18.8.3 Evaluation of Surveillance Program . . . . . . . . . . . . . . . 332
18.8.3.1 Strategy Evaluation . . . . . . . . . . . . . . . . . . . 332
18.8.3.2 Feedback Evaluation . . . . . . . . . . . . . . . . . . 332
18.8.3.3 Evaluation of Data Quality . . . . . . . . . . . . . 333
19 Epidemiology Design in Clinical Research . . . . . . . . . . . . . . . . . . . . 335
Yi Wang
19.1 Design and Implementation of Clinical Research . . . . . . . . . . . . 336
19.1.1 Forming Research Questions . . . . . . . . . . . . . . . . . . . . 336
19.1.2 Commonly Used Epidemiological Design in Clinical
Research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
19.1.3 Collection and Analysis of Clinical Research Data . . . . 338
19.1.3.1 Data Collection . . . . . . . . . . . . . . . . . . . . . . 339
19.1.3.2 Data Analysis . . . . . . . . . . . . . . . . . . . . . . . 342
19.1.4 Preparing Papers for Publication . . . . . . . . . . . . . . . . . 343
19.1.4.1 Choose Target Journal(s) . . . . . . . . . . . . . . . 343
19.1.4.2 Choose a Clear Message . . . . . . . . . . . . . . . 344
19.1.4.3 Achieve High Quality in Writing . . . . . . . . . 345
19.1.5 Common Problems in Clinical Research Design . . . . . . 345
19.2 Reporting Guidelines for Clinical Research Reports . . . . . . . . . 346
19.2.1 Observational Studies Reporting Guidelines . . . . . . . . . 346
19.2.2 Diagnostic/Prognostic Studies Reporting Guidelines . . . 347
19.2.3 Clinical Trials Reporting Guidelines . . . . . . . . . . . . . . 347
19.2.4 Systematic Reviews Reporting Guidelines . . . . . . . . . . 348
19.3 Real-World Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
19.3.1 Definition of Real-World Study . . . . . . . . . . . . . . . . . . 349
19.3.2 The Difference Between RWS and RCT . . . . . . . . . . . 349

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
Chapter 1
Introduction

Chongjian Wang

Key Points
• Clinical epidemiology concerns with the application of epidemiological princi-
ples and methods in specified populations to observe, analyze, and explain issues
in clinical medicine such as diagnosis, screening, treatment, prognosis as well as
the cause of disease with the purpose of providing scientific evidence for clinical
decision.
• The methods of clinical epidemiological studies include descriptive studies,
analytical studies, experimental studies, and mathematical statistics
• The core contents of clinical epidemiology are design, measurement, and evalu-
ation in clinical research.

1.1 Brief History of Clinical Epidemiology

Clinical epidemiology is a newly developed basic science that integrates clinical

medicine with epidemiology by concentrating on the scientific research of clinical
medicine. The development of clinical epidemiology occurred over a long period of
time, with numerous doctors and persons contributing to its progress. When disease
occurred in the population, the people always attempted to identify the cause of the
disease and tried to control the spread of the disease. These efforts contributed to the
development of clinical epidemiology.

C. Wang (✉)
College of Public Health, Zhengzhou University, Zhengzhou, China

© Zhengzhou University Press 2023 1

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_1
2 C. Wang

1.1.1 Early Clinical Epidemiology

Many observations were made in early epidemiology. Today, these observations

may seem simple or not rigorous enough in design, but they provided very useful
information toward describing diseases and their control in those days.
Hippocrates (circa 400 B.C.), the pioneer of epidemiology, attempted to explain
disease occurrence from a rational rather than a supernatural viewpoint. In his essay
entitled “On Airs, Waters, and Places,” Hippocrates described the epidemics of
disease and suggested that environmental and host factors such as behaviors might
influence the development of disease.
In the mid-fifteenth century, ships from outside Europe were asked to stay in an
isolated area for 40 days before arrival at the port to protect the community against
the Black Death (plague) in Europe, which was the origin of “quarantine.”
John Graunt, a London haberdasher, published his landmark analysis of mortality
data in 1662. He was the first to quantify patterns of birth, death, and disease
occurrence, noting men-women disparities, high infant mortality, urban-rural differ-
ences, and seasonal variations. He put forward the need to establish a comparative
group when studying the rule of death and the quality of death data. His contribution
was to introduce statistics into epidemiology.

1.1.2 Clinical Epidemiology

Later, in 1747, the British surgeon James Lind found that vitamin C deficiency was
the cause of scurvy by dividing 12 sailors with scurvy into six groups for a
comparative treatment trial, which was the first epidemiological experiment and
marked the beginning of clinical epidemiology in human history.
In 1796, Edward Jenner, a British doctor, carried out a vaccination in order to
prevent smallpox, which effectively controlled the spread of smallpox and pioneered
active immunization for the control of infectious diseases. In the 18th century, the
French Revolution had a profound impact on the development of epidemiology.
Pierre Charles Alexandre Louis, one of the pioneers of modern epidemiology,
explored the curative effect of bloodletting therapy on inflammatory diseases
through comparative observation and studied the genetic effect on tuberculosis
using life tables. In 1838, Pierre Charles Alexandre Louis and his student, William
Farr, considered the father of modern vital statistics and surveillance, began to
systematically collect and analyze Britain’s mortality data. They developed many
of the basic practices used today in vital statistics and disease classification, extended
the epidemiologic analysis of morbidity and mortality data, and looked at the effects
of marital status, occupation, and attitude. They also developed many epidemiologic
concepts and techniques, such as life tables and the standardization of rates that are
still in use today.
1 Introduction 3

During the mid-1880s, “the father of field epidemiology” named John Snow
conducted a series of investigations using a dot distribution map method of case
distribution to explore the prevalence of cholera in the Broad Street of London. He
also analyzed the mortality rates of cholera in different water supply areas. He was
the first to put forward the famous scientific statement that "cholera is transmitted
through water," and successfully controlled further spread of the epidemic by
stopping the water supply from the suspected pump. Twenty years before the
development of the microscope, Snow’s studies of cholera outbreaks led to the
discovery of contaminated drinking water as the cause of the disease and brought
about effective measures to prevent its recurrence. This became a classic example of
epidemiological field investigation, analysis, and control.

1.1.3 Modern Clinical Epidemiology

Later in the 1800s, many researchers in Europe and the United States began to apply
epidemiological methods to investigate disease occurrence. At that time, most
investigators focused on acute infectious diseases. Around World War II and later,
epidemiologists extended their methods to chronic noncommunicable diseases such
as cancer and cardiovascular diseases. A series of studies, including case-control and
prospective cohort studies by Richard Doll and Austin Bradford Hill, suggested a
very significant association between cigarette smoking and lung cancer. The Fra-
mingham Heart Study is another classic example, which was started in 1948 when
heart disease had become the leading cause of death in the United States. Finally,
during the 1960s and early 1970s, health workers applied epidemiological methods
to eradicate smallpox worldwide. This was an unprecedented achievement in applied
epidemiology.
In 1951, Jerome Cornfield put forward the relative risk, odds ratio, and other
measurement indicators. In 1959, Nathan Mantel and William Haenszel proposed
stratified analysis, which was one of the most cited epidemiological study methods.
In the discipline of infectious diseases, field trials of the polio vaccine, organized by
Jonas Edward Salk in 1954, involving more than 1.5 million children in grades 1 to
3 in the United States, Canada, and Finland, not only confirmed the protective effect
of the vaccine but also laid the foundation for the eventual eradication of polio. In
1979, Sackett summarized 35 possible biases that might occur in analytical studies.
In 1985, Miettinen proposed a bias classification that included comparison, selec-
tion, and information bias.
However, the study of etiology does not solve all the problems of disease
prevention and treatment. For example, epidemiologic studies neglected many
problems in clinical medicine, such as research on medical and health needs,
evaluation of clinical treatment effects, screening and early diagnosis of diseases,
prediction of natural disease history and prognosis, and so on. In this context, many
clinicians began to focus on rigorous design, measurement, and evaluation (DME) in
clinical medical research, and epidemiologists collaborated with clinicians,
4 C. Wang

meanwhile, randomized control trials (RCTs) were proposed in a clinical study.

Selection bias and confounding bias were eliminated using the randomization of
group. The blinding principles of intervention drugs or preventive measures can
eliminate the information bias in the trial process, and then ensure the authenticity of
research results, thus becoming the signature method of clinical epidemiology
research. As the most reliable way to assess causality in a population, RCTs became
the gold standard for evaluating the effectiveness of medical interventions. Since
then, as an independent discipline, clinical epidemiology began to step into modern
medicine. Several representative clinical epidemiology textbooks were also
published, such as Clinical Trials: A Practical Approach (Stuart J Pocock, 1983),
Clinical Epidemiology: The Architecture of Clinical Research (Alvan R Feinstein,
1985), Clinical Epidemiology (David L Sackett, 1985) and Clinical Trials: Design,
Conduct, and Analysis (Curtis L Meinen, 1986).
Today, epidemiological methods of investigation have become tools for answer-
ing questions in medicine and public health regarding their biological and social
facets. Analyses of large databases and complicated calculations have become
feasible due to the collaborations and integration with other disciplines, especially
the use of computers. Clinical epidemiology has contributed to the understanding of
diseases in the population, the study of etiology, and controls of some health
problems, including prevention or treatment of important diseases such as cardio-
vascular diseases, especially ischemic heart disease, asthma, and some cancers.
Regarding the identification of the possible causal risk factors for some emerging
infectious diseases such as acquired immunodeficiency syndrome (AIDS), severe
acute respiratory syndrome (SARS), and coronavirus disease 2019 (COVID-19),
epidemiology plays a significant role.
In summary, how to systematically summarize the evidence and make clinical
and prevention decisions based on the current best research results under the
circumstances of limited resources is imminent, which also creates opportunities
for the development of clinical epidemiology.

1.2 Definition of Clinical Epidemiology

The term “clinical epidemiology” is derived from the combination of clinical

medicine and epidemiology. It is “clinical” because it seeks to answer clinical
questions and guide clinical decision-making with the best available evidence. It is
“epidemiology” because many epidemiological methods are used to answer these
questions and the care of individual patients is seen in the context of the larger
population of which the patient is a member. Many definitions have been proposed,
but the following definition from Siyan Zhan captures the underlying principles and
the public health aspect of clinical epidemiology: “clinical epidemiology is the
science of the application of epidemiological principles and methods in specified
populations to observe, analyze, and explain issues in clinical medicine such as
1 Introduction 5

diagnosis, screening, treatment, prognosis, as well as the cause of disease with the
purpose of providing scientific evidence for clinical decision-making.”

1.3 Roles of Clinical Epidemiology

Clinical epidemiology is based on clinical practice and aims to resolve clinical

questions. Following the gradual development of scientific technology, especially
the explosive development of biology and information science, new diagnostic
techniques and therapeutic strategies are constantly emerging. Clinical epidemiology
creatively applies the theories and methods of epidemiology and health statistics to
clinical research and thus continuously enriches and optimizes the methodology for
clinical research. Clinical epidemiology also provides tools for obtaining optimal
evidence from clinical trials and contributes to enhancing the level of clinical
diagnosis and treatment. Clinical epidemiology is a useful tool for clinical practi-
tioners undertaking clinical practice and scientific research; correctly learning and
applying its principles will help clinicians improve their academic level and increase
their efficiency in acquiring reliable information for decision-making.

1.3.1 To Provide Scientific Ideas and Methods for Clinical

Medical Research

Humans are the research object of medicine, which has dual attributes of nature and
society. For a difference in genetic traits, growth, and living environments, the
clinical manifestations of one disease can vary greatly, and the drug of choice may
either be effective or ineffective in a particular patient groups, which leads to endless
difficulties in diagnosis and treatment. The following problems need to be solved
through clinical medical research: how to improve the level of diagnosis and clinical
differential diagnosis; how to strengthen the evaluation of drug safety and effective-
ness, and improve the level of clinical treatment. Clinical epidemiology is to provide
clinical workers with scientific research methods from three aspects of design,
measurement, and evaluation. DME is the core content of clinical epidemiology,
summarized by clinical epidemiologists at MacMaster University in Canada, and has
been recognized by peers around the world.
6 C. Wang

1.3.2 To Provide Scientific Methods and Means

for the Evaluation of Clinical Diagnosis and Treatment

Problems in clinical practice include four aspects: etiology, diagnosis, treatment, and
outcome. While traditional epidemiology studies focus on etiology, RCTs have
solved the important problem of evaluation of clinical diagnosis and treatment,
thus driving the overall development of clinical research methodology and promot-
ing the emergence of clinical epidemiology. RCTs serve as the gold standard for
exploring causal relationship and evaluating clinical outcomes in populations in
epidemiology studies due to their effective control of selection, information, and
confounding biases.

1.3.3 To Provide a Scientific Methodology and Evidence

for Clinical Decision-Making and Practice
of Evidence-Based Medicine

The core explanation of evidence-based practice and clinical decision-making

according to David L Sackett, a Canadian academic is "Evidence-based medicine
is the conscious, unambiguous, and deliberate utilization of the best available
evidence to make decisions about the care of individual patients. Practicing
evidence-based medicine means doctors need to take into account the best available
research evidence, clinical experience, and patient opinion.” Clinical epidemiology,
with RCTs as the basic method, provides a scientific methodology for solving
various clinical problems. By the 1970s, a number of RCTs had been completed,
and new studies were still being published. However, how to systematically sum-
marize and disseminate the results of these RCTs, use the evidence to guide medical
practice, and improve the quality and efficiency of medical and health services
became a great challenge for medical workers at that time. Therefore, clinical
epidemiologists propose that clinicians should continuously obtain evidence from
published clinical research papers or generate evidence through their own research to
support clinical decision-making and improve the ability of literature retrieval,
analysis, evaluation, and correct use of the latest research results. It is also further
proposed that: how to propose problems that need to be solved clinically; how to
retrieve and collect the best scientific evidence; how to evaluate the quality of this
evidence; whether the effect is good or bad, and extrapolation of the results; how to
formulate a reasonable patient diagnosis and treatment plan combining existing
evidence and reference for other related factors, and according to the effect of the
practice continuously improve the diagnosis and treatment plan, which is a complete
evidence-based decision-making based on scientific thought. The development of
clinical epidemiology has not only catalyzed and followed the development of
evidence-based medicine theory and practice but has also triggered a medical
practice revolution.
1 Introduction 7

1.3.4 It is Possible to Train Clinicians and Medical Scientists

with Excellent Knowledge, Skills, and Quality Under
the Modern Medical Model

The modern medical model has changed from the traditional socio-psychological
and biomedical model to the environmental ecological public health model. The core
of this model requires modern doctors to have a comprehensive decision-making
ability. In order to improve the scientific nature of clinical decision-making, it is
necessary to take various clinical probabilities as the basis, guided by the theory of
probability and applied strategy theory, through certain analysis and calculation, to
quantify complex clinical problems, and then choose a reasonable diagnosis and
treatment plan. At the same time, complex factors such as bioethics, health econom-
ics, and social value orientation should be considered to make safe, effective, and
affordable clinical diagnosis and treatment decisions. Clinical epidemiology is based
on clinical medicine and epidemiology, which is characterized by: under the envi-
ronmental ecological public health model, it permeates and integrates with epide-
miology, biostatistics, health economics, and social medicine based on clinical
practice. The research object is expanded from focusing on individual cases to the
corresponding whole disease population. The study site is extended from individual
patients in the hospital to the comprehensive prevention and treatment of diseases in
the community. The research content is transformed from the research and discus-
sion of early detection, diagnosis, and treatment of diseases to the law of disease
occurrence, development, and outcome to formulate complete clinical research ideas
and improve the clinical diagnosis and treatment level. Scientific method and
thinking of clinical epidemiology not only improve the ability of clinical doctors
in medical research but also make them master clinical decision-making thoughts
and methods, which is beneficial to clinical medicine development, improving the
diagnosis and treatment level, and training a group of high-quality clinical doctors.

1.4 Methodology of Clinical Epidemiologic Study

Clinical epidemiology investigates disease in the population by observing and

inquiring, describing the frequency and distribution of disease, developing hypoth-
eses through induction, synthesis, and analysis, and then testing the hypothesis
through analytical study, and finally verifying the hypothesis through experimental
study. After understanding the occurrence of the disease, a mathematical model is
used to predict the incidence or prevalence of the disease. Thus, clinical epidemiol-
ogy may be classified into three general categories: observational study, experimen-
tal study, and mathematical statistics.
The core of clinical epidemiology is to design, measure, and evaluate (DME).
Any scientific study requires rigorous design, accurate measurement, and reasonable
8 C. Wang

evaluation; this is the essence of clinical scientific study, and it applies to any
subject. The basic aspects of DME are represented below.

1.4.1 Design

Clinical research should have clear study aims. Additionally, research hypotheses
should be put forward according to study aims, and appropriate research objects and
methods should be determined to verify or test the hypotheses. This process is the
clinical study design. It generally includes the following.

1.4.1.1 Clarification of Study Aim

The study’s aim is the core basis of the design. It may be the problems encountered
in clinical work, the unsolved problems of previous work, the scientific enlighten-
ment, and problems concluded from a literature review, and some clinical problems
that need to be solved by superiors. A clear and specific aim is the foundation of all
designs.

1.4.1.2 Determination of Study Methods

In clinical research, an appropriate design method is extremely important according

to different study aims and the nature of the clinical research topic. As all research
methods have both advantages and disadvantages, appropriate and practical methods
should be selected according to different clinical subject areas and goals.
① Observational Study
In an observational study, the study factors of the groups are predetermined
variables, not controlled or influenced by the epidemiologist. Past exposure to
risk factors, lifestyle, personal behaviors, environmental factors, immunization
status, and genetics all affect the status of health and susceptibility to disease in
the groups, and these factors are not influenced by the epidemiologist. That is to
say, the investigator in an observational study measures the factor or exposure but
does not intervene. Observational studies include cross-sectional study, surveil-
lance study, ecological study, case-control study (retrospective study), and cohort
study (prospective or follow-up study).
Observational investigations may be divided into descriptive studies or ana-
lytical studies. A descriptive study (cross-sectional study, surveillance study, and
ecological study) is the first step in an epidemiological investigation and is used to
validate the measurement of the health conditions and health-related characteris-
tics of populations, typically in terms of person, place, and time. This information
provides essential contextual information with which to develop hypotheses,
1 Introduction 9

study design, and interpret results and serves as the foundation for studying
populations. As a particular type of descriptive study, surveillance can monitor
the changes in the occurrence of disease over time. Descriptive approaches are
also useful in clinical epidemiology, which includes assessing the performance of
diagnostic and screening approaches and clinical decision-making. Analytical
epidemiology (case-control study and cohort study) is often used to systemati-
cally evaluate the suspected relationships between an exposure and a health
outcome, and provide stronger evidence about particular relationships between
exposure variables and health status in the general population or in a specific
population.
② Experimental Study
In an experimental study, the subjects are randomly assigned to either an
experimental group or a control group. In the experimental group, an active
attempt is adopted to change a disease, condition, or death determinant, such as
an exposure or behavior, or the progress of a disease through treatment, and then
assess the cause-and-effect relationships statistically. The experimental study
design includes randomized controlled trials using patients as subjects (clinical
trial), field trials, and community trials in which the participants are general
people.
An experimental study is useful in establishing a sound cause-and-effect
relationship between a factor, intervention, or agent for therapy or prevention of
a disease, condition, or death. However, the implementation of experimental
studies often involves practical and ethical issues. Analytical epidemiologic
studies can offer a realistic approach to testing hypotheses of exposure-disease
relationships, and provide more accurate information and useful insights into the
effects of diseases or conditions.
③ Mathematical Statistics
After understanding the occurrence rule of the disease, we can use a mathe-
matical model (theoretical epidemiology) to predict the incidence or prevalence
of the disease. Figure 1.1 summarizes the hierarchy of clinical epidemiological
study design.
The appropriate design applies not only to attaining study goals but also allows
for making effective use of human effort, materials, and time. It reflects the
scientific accuracy of the observed results and therefore should also aim to be
cost-effective and scientifically rigorous, accurate, and reliable.

1.4.1.3 Identification of the Study Subjects

The study subjects include population and sample. The research population is the
whole group of subjects determined according to the study aim. The sample is a
representative part selected from the whole population, which is often used in
practical work. This requires sampling randomization, a sufficient number of sam-
ples, and clear diagnostic criteria for samples (cases). Three criteria can guarantee
the reliability of the research.
10 C. Wang

Cross-sectional study

Descriptive
Surveillance
epidemiology Develop hypotheses
Ecological study
Observational
study
Case-control study
Analytical
Test hypotheses
epidemiology
Cohort study

Methods of
Clinical trial
clinical
epidemiologic Experimental Experimental Field trial Verify hypotheses
study epidemiology
studies
Community trial

Mathematical Theoretical
statistics epidemiology

Fig. 1.1 Hierarchy of clinical epidemiological studies design

1.4.1.4 Determination of the Groups

In clinical scientific research, the subjects are divided into experimental and control
groups for comparison. Experimental groups can be new diagnostic methods, new
drugs, or preventive measures. Comparison is one of the characteristics of clinical
epidemiology. The method of grouping can be random or non-random, such as
grouping by different times, places, or by certain features. However, randomization
allows for an equal chance of selection among all qualified subjects, and guarantees
that samples are representative and free of selection bias.

1.4.1.5 Determination of Study Indicators

Study indicators are determined according to study aims. In order to evaluate clinical
diagnostic trials, it is necessary to select a recognized index of clinical diagnostic
method or equipment as the gold standard. To evaluate the safety and effectiveness
of clinical new drugs, clinical indicators, such as effective rate, improvement rate,
fatality rate, and incidence of adverse reactions, are used to verify the effect. With the
aim of evaluating the effect of vaccine prevention, the protection rate and the rate of
change of serum antibody level should be the evaluation criteria. The complication
rate, disability rate, and recovery rate should be considered as the indicators of
disease prognosis. When measuring these indicators, the measurement methods used
and the authenticity and reliability of indicators need to be carefully considered and
clearly defined in the research and design stage.
1 Introduction 11

1.4.1.6 Determination Methods for Data Collection and Analysis

In clinical scientific research, the results usually involve the confounding influence
of known and unknown factors that are not the direct object of study. To attain
scientifically rigorous results and conclusions, the study design should comply with
some basic principles, e.g., control, randomization, and blinding principles.
Control principle: The setting of control is a core idea of epidemiology. Proper
identification can be derived only through comparison.
Randomization principle: Randomization is an important method and a basic prin-
ciple of clinical studies. Randomization allows for an equal chance of selection
among all qualified subjects and guarantees that samples are representative and
free of selection bias.
Blinded principle: This principle avoids the influence of subjectivity on the part of
the investigator and prevents the investigator from knowing the details of group-
ing and intervention. This practice safeguards the accuracy of the clinical condi-
tions and observed results. Based on the level of understanding regarding the
grouping details and relationships amongst the subject, investigator, and data
analyzer, this can be divided into single-blinded (only the subject does not know
what the intervention is), double-blinded (neither the subject nor the investigator
performing the study knows how the intervention is grouped), and triple-blinded
trials (neither the subject nor the investigator nor the data analyzer knows how the
intervention is grouped).

1.4.1.7 Determination of Study Quality Control Methods

The common biases in clinical epidemiology include selection bias due to inconsis-
tent diagnostic criteria when patients are enrolled, information bias due to collection
of clinical information, and confounding bias due to non-strict randomization of
grouping. In addition to the above bias, the inconsistent equipment, different batches
of diagnostic reagents, inconsistent information acquisition time (such as blood
pressure measurement), inconsistent recognition of diagnostic standards for different
doctors, and different adherence of the patient to the doctor's orders, will bring some
bias for the results of the study. Therefore, it is very important to adopt quality
control methods against the above possible bias at the study design stage.

1.4.2 Measurement

In clinical studies, the effect of drugs needs to be evaluated through the measurement
of certain indices, which can be quantitatively and qualitatively measured. However,
high sensitivity and specificity are required for any method. Some can be objectively
expressed in terms of specific units or values and are called hard indices or objective
12 C. Wang

indices, which include factors such as frequency indices (incidence, prevalence,

mortality, fatality rate, etc.), effect indices (absolute risk, relative risk, and attribut-
able risk, dose-response relationship, etc.), and objective indices (obtained from
objective methods or instruments, such as heart rate, blood pressure, height, weight,
morbidity, and death, etc.). Others are difficult to express in terms of specific
measurement units and are termed soft indices or subjective indices. These include
factors such as pain, quality of life, nausea, dizziness, fatigue, and some psycholog-
ical determinations. Subjective indices are obtained from the individual impressions
of the subject while objective indices are obtained from clinical observation or
measurement. Objective indices are not influenced by subjective factors and are
therefore more accurate and reliable.
Usually, measurements in clinical research are conducted by different medical
staff, so there are many possible influencing factors resulting in bias. For example,
some bias due to the inconsistent equipment, different batches of diagnostic
reagents, inconsistent recognition of diagnostic standards for different doctors, and
variation adherence of the patient to the doctor's orders will bring uncertainty to the
results of the study.

1.4.3 Evaluation

Evaluation is the application of basic theory and methods of epidemiology to assess

the accuracy and scientific basis of clinical data and results through statistical
analysis. It involves the final selection of valid results and the discarding of false
results. The content of the evaluation is mainly reflected in the following aspects.

1.4.3.1 Evaluate the Validity and Reliability of the Study Results

Clinical epidemiological methods are used to evaluate the design, the accuracy of
diagnostic methods, the short-term and long-term efficacy of treatments, the preven-
tion and control measures of bias, the source, representativeness, and compliance of
study subjects, etc., to test the validity and reliability.

1.4.3.2 Evaluate the Importance of Study Results

① Evaluate the Clinical Significance of the Results

According to the strict evaluation standards of etiology, diagnosis, prevention,
treatment, the prognosis of clinical epidemiology and evidence-based medicine,
and related indicators of clinical significance, combined with professional and
clinical practice, the clinical value of the results is evaluated for improving
clinical medical level.
② Evaluate the Statistical Significance of the Results
1 Introduction 13

If the results of the study have clinical significance, the correct statistical
methods must be used to test the significance of the results to evaluate the true
degree of clinical differences, i.e., the probability of true positive and true
negative results as well as the level and confidence interval (CI) range of test
effectiveness, so as to obtain the evaluation of the true degree of clinical study
results.
③ Evaluate the Health Economic Implications of the Results
The results of clinical medical research should evaluate the social and eco-
nomic benefits (including the cost-effectiveness, cost-benefit, and cost-utility) by
applying the principles and methods of health economics, compare, and evaluate
to affirm those research results with low cost and good effect so that they can be
popularized and applied.
In summary, the main research content and methods of clinical epidemiology are
design, measurement, and evaluation. Effective control of various biases should be
adopted to ensure the validity and reliability of clinical research results.

1.5 Characteristics of Clinical Epidemiology

As a field of inquiry, clinical epidemiology is a logical discipline that proceeds by

way of a sequence of reasoning from empirical data. Clinical epidemiology involves
a systematic set of methods and procedures for developing knowledge about health-
related events and the relationships between them. It effectively deals with chance,
bias, and error. Clinical epidemiology has the following characteristics.

1.5.1 Group

The research object of clinical medicine is the individual diagnosis, treatment, and
outcome, while the object of clinical epidemiology is a group of populations with a
particular clinical disease according to the purpose of the study.

1.5.2 Comparison

Epidemiology is a comparative discipline. It asks questions such as how much

disease is in different populations, places, and times and why the disease is distrib-
uted this way. It likewise compares the frequency of possible risk factors between
groups that have a particular disease (cases) and those without the disease (controls).
Key comparative measurements in epidemiology are prevalence, incidence, and risk
ratio. Clinical epidemiology also requires the comparison of diagnostic methods or
14 C. Wang

drugs used in the two groups of cases so as to judge the effectiveness of the
diagnostic methods or drugs.

1.5.3 Probability Theory and Mathematical Statistics

Epidemiology uses frequency, rather than absolute numbers, to describe the distri-
bution of disease because absolute numbers do not show the intensity of disease or
the risk of death in the population. Epidemiology, particularly clinical epidemiology,
emphasizes that probability or frequency is actually a probability needing the right
denominator. In addition, epidemiological work requires a reasonable sample size,
and the final sample depends on statistical principles and varies from case to case.

1.5.4 Social Psychology

Health is closely related to environmental factors. The occurrence of disease is not

only related to the health status of the human body but it is also affected by the
natural environment and social environment. The biological, psychological, and
social conditions of people should be taken into consideration when studying the
etiology and risk factors of disease.

1.5.5 Integrating Medicine

At present, chronic non-communicable diseases (NCDs), such as cardiovascular and

cerebrovascular diseases, tumors, diabetes mellitus, and respiratory diseases, are the
main health problems affecting most clinical patients. Chronic NCDs not only rank
first among the causes of death in the Chinese population but also endanger the
health of the labor force and cause the rapid rise of medical expenses. Therefore,
controlling the rapid rise of NCDs cannot be achieved by only adhering to the
approach of clinical treatment. This requires the integration of prevention and
treatment: the integration of residents' health and disease management in a commu-
nity hospital, with disease management in superior general and specialized hospitals,
and the establishment of benign referral; the integration of etiology prevention, early
diagnosis and early treatment with active treatment, rehabilitation, reduction or delay
of complications, and prevention of disability; the integration of control of behav-
ioral risk factors with clinical drug therapy; the integration of physiological, patho-
logical and psychological therapy; the integration of traditional Chinese medicine
with modern Western medicine treats the patient as an organism (to treat both
symptoms and root causes) and applies high and new technology to improve the
1 Introduction 15

quality of life. All of these will be reflected in the study of clinical epidemiology and
will play an increasingly important role.

1.5.6 Development

The definition and tasks of epidemiology are developed according to the major
health problems during different periods. In recent years, the epidemiological
methods have also improved with the development of other disciplines. Traditional
epidemiology focuses on the study of three links and two factors of infectious
diseases, modern epidemiology focuses on the study of social, psychological, and
environmental factors of disease, while clinical epidemiology concentrates on
design, measurement, and evaluation. From ecological research of macro
epidemiology to molecular biology of micro epidemiology, from RCTs in clinical
epidemiology to production, evaluation, and use of medical scientific evidence in
evidence-based medical research. All these indicate that development is one of the
characteristics of clinical epidemiology.
Chapter 2
Distribution of Disease

Shan Zheng

Key Points
• Frequency measurement is an effective method to quantitatively study the char-
acteristics of disease distribution. Common measures of disease frequency
include morbidity, prevalence, and mortality.
• The level of disease is usually expressed by sporadic, outbreak, epidemic, and
pandemic, which refers to the change of incidence rate and the association
between cases in a certain population during a certain period of time.
• The distribution of disease is used for describing the status of disease in specified
populations, areas and time, which will reveal the epidemic characteristics of the
disease and its potential risk factors.
By describing the incidence, prevalence, and death, the epidemiological character-
istics of diseases may be presented based on the combination of the distribution by
people, time, and place (commonly known as the three-dimensional distribution in
epidemiology). The study of the distribution of disease is the foundation of finding
possible risk factors or understanding etiology, which may have an important
contribution in determining the core problems in public health and high-risk groups.
Of course, the application of this knowledge would provide scientific evidence for
the planning and evaluation of healthcare systems. Briefly, the study of the distri-
bution of disease is not only the starting point and basis of epidemiological research
but also an indispensable part of studying epidemic patterns and etiology of diseases.
In this chapter, some basic elements, concepts, and tools of epidemiology are
discussed: the basics of measurement and comparison, the level of disease, and
distribution of disease.

S. Zheng (✉)
School of Public Health, Lanzhou University, Lanzhou, China
e-mail: [email protected]

© Zhengzhou University Press 2023 17

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_2
18 S. Zheng

2.1 Measures of Disease Frequency

2.1.1 Frequency Measures

2.1.1.1 Ratio

The value obtained by dividing one quantity by another. In a ratio, the values of
x and y may be completely independent, or x may be included in y.

x
Ratio = ðx is completely independent of y or x is part of yÞ ð2:1Þ
y

2.1.1.2 Proportion

A type of ratio in which the numerator is included in the denominator.

x
Proportion = ðx is part of yÞ ð2:2Þ
y

2.1.1.3 Rate

A measure of the frequency of occurrence of a phenomenon. In epidemiology, a rate

is an expression of the frequency with which an event occurs in a defined population,
usually in a specified period.

Number of cases or events occurring during a given time period

Rate = × K ð2:3Þ
Population at risk during the same time period

2.1.2 Morbidity Frequency Measures

2.1.2.1 Incidence Rate

The incidence rate is the number of new cases per population at risk in a given time
period. The numerator is the number of new events in a defined period or other
physical spans. The denominator is the population at risk of experiencing the event
during this period. The calculating formula for incidence rate follows:
2 Distribution of Disease 19

Number of new cases in specified period

Incidence rate = ×K
Average number of person exposed to risk during this period
K = 100%, 1000‰, 10, 000 per 10, 000 . . . . . .
ð2:4Þ

To calculate incidence rate, a year is usually chosen as a study period. You can
also define the study period by the characteristics of diseases or events.
The numerator of an incidence rate should be the new cases of disease which
occurred or were first diagnosed during the observation period. Those cases which
occurred or were diagnosed earlier should not be included in the numerator. If a
person has multiple episodes of illness during the observation period, they should be
all counted as new cases, such as influenza and diarrhea, which can occur more than
once in a year.
Notice that the denominator is the population at risk, which means that the
denominator is the number of people exposed and at risk for the observed disease
in the population of an area during the observation period. Persons who are already
ill and are not likely to become new cases during the observation period should not
be included in the exposed population. For example, in calculating the incidence of
measles, people who already have measles cannot be included in the denominator. In
addition, theoretically, people who have received the measles vaccine and gained
immunity should not be included in the denominator, but it is not easy to divide in
practice. The denominator is usually the average population of the area during the
observation period.
Incidence rates are useful in the study of disease etiology because they are
informative about the risk of a disease process in different population groups. By
comparing the difference in incidence rates, some possible or potential causation
could be found and proposed. It is usually used to measure the risk of acute disease
or conditions but is also used for chronic diseases.

2.1.2.2 Attack Rate

An attack rate is a variant of an incidence rate applied to an outbreak of disease

among a narrowly defined population during a short period of time. It is calculated
by the same formula as incidence rate, but it is observed over a shorter period of
time. The attack rate is often used in outbreaks and epidemics of food poisoning,
occupational poisoning, or infectious diseases.

Attack rate
Number of new cases among the population during the period
= × 100% ð2:5Þ
Population at risk at the beginning of the period
20 S. Zheng

2.1.2.3 Secondary Attack Rate

A secondary attack rate is a measure of the frequency of new cases of a disease

among the contacts of known cases. The formula is as follows:

Secondary attack rate

Number of cases among contacts of primary cases during the period
= × 100%
Total number of contacts
ð2:6Þ

This index is always used to measure the contagiosity of infectious diseases and
to evaluate the effectiveness of prevention measures. It is worth noting that the cases
occurring without the incubation period following exposure to a primary case are
generally not included in the numerator and denominator.

2.1.2.4 Prevalence

Prevalence is a proportion of persons in a population who have a particular disease

(including new cases and pre-existing cases) at a specified point in time (point preva-
lence) or over a specified period (period prevalence). The formula for prevalence is:

New cases and pre‐existing cases during a given time period

Prevalence = ×K
Population during the same time period
K = 100%, 1000‰, 10, 000 per 10, 000 . . . . . .
ð2:7Þ

The prevalence of a disease in the population is influenced by many factors; the

common factors are shown in Table 2.1. Among these factors, incidence and
duration of the disease play the most significant effect on prevalence. When the
incidence and duration of a disease in a certain place are stable for a long time, the
relationship between prevalence, incidence, and duration of the disease is shown as:
Prevalence = incidence × duration of disease.

Table 2.1 Factors affecting prevalence

Increased by Decreased by
Longer duration of the disease Shorter duration of the disease
Prolongation of life of patients without a cure Increased case-fatality rate
Increase in new cases (increase in incidence) Decrease in new cases (decrease in incidence)
In-migration of cases In-migration of healthy people
Out-migration of healthy people Out-migration of cases
In-migration of susceptible people Improved cure rate of cases
Improved diagnostic level –
Better reporting rate
2 Distribution of Disease 21

The prevalence is usually used to describe the epidemiological characteristics of

diseases, especially for the chronic diseases with a longer duration. It can be used to
estimate the impact of the diseases to human health and to evaluate medical and
health work and health resource allocation conditions, such as hospital beds turnover
rate, sanitation, and the demand and supply of human resources, etc.

2.1.3 Mortality Frequency Measures

2.1.3.1 Mortality Rate

A mortality rate represents the proportion of total deaths in a population in a given

period of time. The numerator is the number of persons dying during the period; the
denominator is the average population during the same period. The annual death rate
or mortality rate from all causes in a population is generally calculated by the
following formula:

Annual mortality rate for all causes

Number of deaths from all causes in 1 year
= ×K ð2:8Þ
Average population in the same period
K = 100%, 1000‰, 10, 000 per 10, 000 . . . . . .

Generally, a mortality rate for all causes is also called the crude death rate. When
comparing mortality rates in different regions, mortality rates need to be standard-
ized. Similarly, comparisons of prevalence or incidence rates across regions need to
be standardized.
In addition, the crude death rate is an unadjusted mortality rate which shows all
causes of death for a population. When we calculate the mortality rate by a specific
disease, age, gender, race, and so on, this mortality rate will be named as cause-
specific mortality rate. For example, an age-specific mortality rate is defined as:

Total number of deaths occurring in a specific age group

of the population in a defined area during a specified period
×K ð2:9Þ
Estimated total population of the same age group
of the population in the same area during the same period

Mortality rate can be an indicator reflecting the total death level of a population,
which is usually used to measure the death risk of a population in a certain period and
a certain area. The specific mortality rate can provide information on the variation of
the mortality rate of a disease in population, time, and region and can be used to
explore the etiology and evaluate the prevention and treatment measures.
22 S. Zheng

2.1.3.2 Case Fatality Rate

Case fatality rate is used to measure the severity of disease and is defined as the
proportion of cases with a specified disease or condition who die within a specified
time.

Number of deaths from diagnosed cases

in a specified time
Case fatality rate = × 100% ð2:10Þ
Number of diagnosed cases of the disease
in the same period

2.1.3.3 Survival Rate

The proportion of patients who received treatment or those with a certain disease
who were still alive after following up for several years.

Number of cases who were still alive

after following up N years ð2:11Þ
Survival rate = × 100%
Number of all cases after following up N years
Where N is always equal to 1, 3, 5, and 10 years:

Survival rates are used to evaluate the severity and long-term outcomes of chronic
diseases such as tumors, cardiovascular diseases, and chronic infectious diseases
(like tuberculosis and AIDS).

2.2 Epidemic Disease Occurrence

The level of disease is defined as the variations in the trend of incidence of a

particular disease and the association between the cases in a given population during
a certain period. Terms of the level of disease include sporadic, epidemic, and
outbreak.

2.2.1 Sporadic

Sporadic means the incidence of a particular disease is in the range of the average
incidence of the disease over the past 3 years in that population in that area, which
refers to a disease that occurs infrequently and irregularly. For example, plague
occurs by chance in a grazing area.
2 Distribution of Disease 23

The cause of sporadic occurrence includes four aspects.

① The level of immunity in a population has increased as a large portion of the
population in the area has been vaccinated, or people have been infected with the
disease before.
② The disease is an inapparent infectious disease, such as poliomyelitis and
Japanese encephalitis.
③ The transmission mechanism of the disease is not easy to achieve.
④ Diseases with a long incubation period, such as leprosy.

2.2.2 Epidemic

Epidemic means the incidence of a disease is above the average incidence of the
disease over the past several years in that population in that area. When a disease
shows epidemic, it means there is an obvious association between all cases in time
and space. For example, in 2009, the epidemic of influenza A (H1N1) showed an
obvious characteristic, including person-to-person transmission and dissemination in
different areas. In addition, when an epidemic occurs on a scale which crosses
international boundaries, it is called a pandemic. For example, the 1918–1919
influenza pandemic.

2.2.3 Outbreak

As a group of people are all exposed to an infectious agent or a toxin from the same
source, an increase in the number of cases of a disease appear suddenly. The
epidemic is always limited to a localized area, e.g., in a village, town, or closed
institution.

2.3 Distribution of Disease by Time, Place, and Person

The distribution of disease is used for describing the status of disease by time, place,
and person, which will reveal the epidemic characteristics of the disease and
potential risk factors and even propose the hypothesis for the pathogenesis of
some disease process. Distribution of disease is the core of descriptive study and
the basis of analytical study.
24 S. Zheng

2.3.1 Time

Whether it is an infectious disease or a non-communicable disease, their epidemic

characteristics will change by time lapse. Analyzing the time variation in disease risk
is used to provide the basis for studying the cause of disease and drawing up
prevention and control measures. There are four forms of time variation used to
describe the time distribution of disease, including rapid fluctuation, seasonality,
periodicity, and secular trend.

2.3.1.1 Rapid Fluctuation

This refers to a phenomenon where the number of cases of a disease increases

suddenly in a certain institution or a fixed population. It is similar to an outbreak,
but an outbreak is suitable for a small population, while rapid fluctuations apply to
large population.
The causes of rapid fluctuation are generally clear, and the main cause is that a
number of people are exposed to the same pathogenic factors at the same time or
continually contacted, such as food poisoning in collective canteens or the outbreak
or epidemic of dysentery and measles. Besides, rapid fluctuation of a disease might
result from natural disasters, environmental pollution, or social politics.
Figure 2.1 shows the time distribution of viral meningitis outbreak in a school as
the pollution of drinking water. The first case occurred on June 26, and the number of
cases increased quickly on June 30 and decreased on July 2. The epidemic stopped
on July 18 and lasted 23 days.

18
16
14
Number of cases

12
10
8
6
4
2
0

Date of onset

Fig. 2.1 Time distribution of viral meningitis outbreak in a school

2 Distribution of Disease 25

2.3.1.2 Seasonality

Seasonality refers to a phenomenon that the incidence or mortality of a disease

increases in a certain season, which includes two different characteristics.

Strict Seasonality

The new cases of some infectious diseases only occur in some specific months of the
year; however, in other months, no cases are occurring. The epidemic of malaria and
encephalitis B is usually seen in summer and autumn, which is related to the life
cycle of the vector that causes disease transmission.

Seasonal Rise

Some diseases occur all year round, but the incidence increases only in certain month
(s). For example, enteric infectious diseases and respiratory infectious diseases occur
all year, but the incidence of enteric infectious diseases is most common during the
summer and autumn months, while the incidence of respiratory infectious diseases
always rises in spring and winter. Moreover, some non-communicable diseases also
show a characteristic seasonal rise, such as pollinosis occurs at the end of spring and
the beginning of summer; the incidence of stroke always increases in winter.

2.3.1.3 Periodicity

Periodicity refers to a phenomenon where the frequency of a disease increases in a

certain interval. Some respiratory infectious diseases often have regular periodic
epidemics. As shown in Fig. 2.2, the weekly measles cases in New York City
showed a clear periodicity from 1891 to 1963. There were two peaks during this
period, on the weekend of 31 January 1920 (4671 cases) and 22 March 1941 (5016
cases). But with the invention of the first measles vaccine in 1963, measles vaccines
have been used widely to immunize susceptible population against measles. So the
number of measles cases in this region declined rapidly.
The reasons for periodic epidemics of disease are associated with immunization
level of the population, duration of immunization, accumulation of susceptible
population, variation of pathogens, etc.

2.3.1.4 Secular Trend

Secular trend (Synonyms: temporal trend or long-term trend) denotes that the annual
cases or rate of a disease over a long period, generally years or decades, shows
26 S. Zheng

Fig. 2.2 Trends in measles cases in New York City from 1891 to 1984. Reprinted with permission
from Hempel, Karsten; Earn, David J D. A century of transitions in New York City’s measles
dynamics. J R Soc Interface. 2015 May 6;12(106):20150024. doi: 10.1098/rsif.2015.0024

long-term or secular trends in the occurrence of the disease. The probable causes of
secular trends include changes in etiology or pathogenic factors, changes in patho-
gens, improvements in medical treatment and prevention, and changes in social
policies. Therefore, these trends are commonly used to suggest or predict the future
incidence of a disease, to evaluate programs or policy decisions, and to reveal some
potential causes in the occurrence of a disease.
Figure 2.3 shows mortality rates for six diseases in men from 2000 to 2016.
According to the WHO, trachea, bronchus, and lung cancer showed a significant
decline during the 16-year period, the death rate (Age-standardized) was 33.41 per
100,000 population in 2016 with a - 38.4% of change relative to 2000. The reason
for this downward trend may be related to the decline in smoking rates among men,
and the implementation of smoking-related policies such as banning smoking in
public areas and raising tobacco taxes may also influence the change in the trend. In
addition, we also see slight increases in diabetes mellitus and pancreatic cancer,
which may also be related to changes in the lifestyle of the population.

2.3.2 Place

The occurrence of a disease is associated with natural and social environment, so

analyzing data by place can provide an important clue for revealing the cause of a
disease and also provide a basis for proposing prevention measures. The variation of
pathogenic factors in different places can result in the variation of the disease
distribution. Some natural environmental factors, e.g., special geographical location,
topography and geomorphology, meteorological conditions, and some social
 2
Distribution of Disease

Fig. 2.3 Trends in age-adjusted death rates from six diseases for males, United States, 2000–2016
27
28 S. Zheng

environmental factors, e.g., living habits and sociocultural background, can play a
significant role on disease distribution by place.
To study the distribution by place, there are two methods used to analyze the data
which are political boundaries and natural boundaries. The method of political
boundaries is used to divide places by continent, state, or region at the global level
and by province, city, or urban and rural in a country. This method is easy and
convenient, but it generally does not match the distribution of natural environmental
factors, which may cover up the ecological relationship between natural environ-
mental factors and the distribution of a disease. Another method used to analyze data
is natural boundaries, which is used to divide places by mountains, plains, lakes,
rivers, or grasslands, and so on. This method is better at revealing the correlation
between natural environmental factors and disease, but it is difficult to collect the
data and carry it out.

2.3.2.1 Comparisons Among and Within Countries

The Distribution of a Disease in Different Countries

Some diseases are spread all over the world, but the distribution is not uniform, and
the morbidity or mortality rate of these diseases may vary greatly. Both infectious
diseases and non-communicable diseases present various distributions between
countries. For example, yellow fever is only an epidemic in South America and
Africa. Cholera is common in India. The mortality rate of stomach cancer is higher in
countries such as Japan and Chile, while it is lower in Australia and the United
States. Liver cancer is common in Asia and Africa, but breast cancer and colon
cancer are common in Europe and North America.
According to the Global cancer statistics (2012) [1], prostate cancer is the second
most frequently diagnosed cancer in men worldwide. It is clear that the most
frequently diagnosed cancer among men was in more developed countries. Incidence
rates vary by more than 25-fold worldwide and are higher in Australia/New Zealand,
Northern America, Northern, and Western Europe, and some Caribbean nations, and
lower in Asia.

The Distribution of a Disease in Different Areas within a Country

The incidence or mortality rates of diseases also vary in different areas within a
country. For example, schistosomiasis only occurs in some provinces, south of the
Yangtze River in China, which is associated with the distribution of oncomelania in
the surrounding environment. Nasopharyngeal cancer is most common in Guang-
dong Province, China.
Based on the third national retrospective sampling survey of death causes in
China from 2004 to 2005, the highest breast cancer mortality rates were found in
Shanghai (5.21 per 100,000) and Heilongjiang Province (5.69 per 100,000), while
2 Distribution of Disease 29

the mortality rates in Liaoning, Jilin, Shandong, Guangxi, and Hunan (ranging from
4.53 to 4.84 per 100,000) were higher than most provinces.

2.3.2.2 Urban-Rural Comparisons

The variation of population density, natural environment, health level, living condi-
tions, economic level, and population mobility shows an obvious difference in kinds
of the disease, cause of death, incidence, and fatality of the disease between urban
and rural areas.
In urban areas, respiratory infections, such as chicken pox, mumps, and influenza,
are mostly common due to higher population density, small living space, traffic
congestion, and large mobility. In addition, more serious environmental pollution in
urban areas can increase the incidence and mortality rates of some diseases related to
air pollution.
In rural areas, intestinal infections and vector-borne infectious diseases are mostly
common due to poor sanitary conditions and local natural environment. Respiratory
infectious diseases are difficult to spread because of lower population density, but it
can also cause an outbreak once the infection sources enter.
According to the China Health Statistical Yearbook (2003, 2021) [2, 3], cardio-
vascular disease mortality rates among urban and rural residents in China increased
from 1990 to 2020 (Fig. 2.4). While the mortality rate increased rapidly in rural
areas, it ultimately surpassed the rate in urban areas by 2009. In recent years,
differences in the distribution of disease between urban and rural areas are narrowing
and getting bigger after 2013 due to economic development and urbanization in
China.

350
330 Urban
310
Rural
Mortality Rate per 100,000

290
270
250
230
210
190
170
150
1990
1995
2000
2002
2003
2004
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020

year

Fig. 2.4 Trends in cardiovascular diseases mortality among urban and rural residents in China,
1990–2020
30 S. Zheng

2.3.2.3 Endemic Clustering

Endemic clustering refers to the phenomenon of the morbidity or mortality of disease

in a locality being significantly higher than in the surrounding areas. A disease with
the characteristic of endemic clustering suggests that there are some specific path-
ogenic factors which have an adverse effect on human health in a certain place. The
term is normally used to discuss the cause of disease and evaluate the intervention
effects of some prevention measures. Ebola hemorrhagic fever has been found in the
Ebola River of southern Sudan and the Congo since 1976. At present, the disease
presents endemic clustering, which is confined to the central African rainforest and
the tropical savannah of southeast Africa. In February 2014, an Ebola hemorrhagic
fever broke out in West Africa and gradually spread to areas such as Sierra Leone
and Liberia. The process presents obvious regional aggregation.

2.3.2.4 Endemic Disease

Endemic disease presents that some diseases are confined to a certain area due to the
influence of natural environmental and social factors.
The standards for determining whether a disease is endemic include several
aspects.
① The incidence of the disease is high among the residents of the area.
② The incidence of the disease among similar populations in other regions is low
and even nonpathogenic.
③ After immigrating to the area, the incidence of the disease in immigrants is
consistent with that in the local population.
④ After the population has moved out of the area, the incidence is reduced or the
symptoms of the disease are reduced or self-healing.
⑤ Besides people, local susceptible animals have the same disease.

2.3.3 Person

The differences in morbidity and mortality are more influenced by personal charac-
teristics (age, sex, ethnic and racial group, occupation, behavior, etc.), and these
characteristic variables or their variation can be considered as the foremost risk
factors for many diseases. Therefore, the study of these demographic characteristics
is of great significance to explore the risk factors or epidemic characteristics of
disease or health status, which can provide clues to confirm high-risk population and
many hypotheses, importantly, provide evidence for policy-making.
2 Distribution of Disease 31

2.3.3.1 Age

Age is one of the most important demographic variables. The occurrence and
development of almost all diseases are associated with age. The age-specific patterns
also differ between and within diseases. Generally, the incidence of chronic diseases
increases with age, while the incidence of acute infectious diseases decreases.
However, some chronic diseases present a peak at younger ages as the change of
pathogenic factors, such as malignant tumor, diabetes, high blood pressure, etc.
Sometimes, the incidence and prevalence of diseases are not the same in the different
age groups. The incidence of acute lymphocytic leukemia is higher in children, and
the Hodgkin’s disease has two peaks at youth and old age. The incidence of gastric
cancer shows a rising trend with age. The effects of age are most commonly ascribed
to an individual’s cumulative exposure to environmental factors over a lifetime or to
the decline in immunological defenses.
There are two ways to analyze the association between diseases and age. One is
cross-sectional analysis, and another is birth cohort analysis.
1. Cross-sectional analysis is used to examine the age-specific disease rates of a
population at a particular period, which would include individuals born at
different time periods. For example, the age-specific incidence rates of some
acute infectious diseases are presented in a cross-sectional age curve at a partic-
ular period for different generations. The potential problem associated with a
cross-sectional analysis is that individuals in different age groups at a particular
calendar time were born in different years. Therefore, these individuals belong to
different generations or birth cohorts. Different generations always have different
disease risks due to different types or levels of exposure to disease risk factors.
Thus, the cross-sectional analysis, especially for chronic disease, may or may not
validly explain the association between disease and age.
2. Birth cohort analysis is based on age-specific incidence rates for a particular birth
cohort observed at successive points in calendar time as the cohort grows older,
which includes all persons born within a specified period. People from a gener-
ation or a birth cohort would carry, even throughout their lives, a relatively higher
or lower risk for certain diseases compared to other birth cohorts. If disease risk
changes with successive birth cohorts due to the increase or decrease in exposure
to risk factors, the age-birth cohort curves would be progressively higher or lower
than the previous birth cohorts. Thus, birth cohort analysis can reflect a real
change in disease rate with age through examining the disease rates from people
born in the same or different time periods as age increases, if disease risk indeed
varied by birth cohorts.
Figure 2.5 shows the relationship between age and lung cancer occurrence in a
city for 5-year periods (1995–1999, 2000–2004, 2005–2009, 2010–2014, and
2015–2019) in the cross-sectional age curves, a classic example of examining age
effect on disease. According to these curves, the incidences of lung cancer at first
increase with age, then the incidences are at peak in the 65–69 age groups and
32 S. Zheng

Fig. 2.5 Lung cancer incidence rates in a city based on cross-sectional age curves

subsequently decline as age advances. Meanwhile, the incidences of lung cancer

increased significantly during the 5-year survey periods; the interpretation of this
situation may be that the risk for lung cancer continues to increase. The question to
be considered is why the lung cancer incidence rates in the city show a decline after
ages 65–69 in the cross-sectional age curves. Actually, there seems to be no
biological basis for assuming that age-related changes result in a decrease in lung
cancer risk in the city based on the present etiology of lung cancer.
However, in Fig. 2.6, when the same lung cancer age-specific incidence rates are
presented in six generations (G1920, G1925, G1930, G1935, G1940, and G1945),
the rates generally showed a continuous increase with increasing age for each of the
six birth cohorts for the given age ranges. The incidence rates among people born in
latter generations at the same ages are higher than that in those people born in
previous generations and tend to be younger.

2.3.3.2 Gender

There are, for certain diseases, substantial differences in the rates between men and
women (sex ratio). For example, incidence and mortality rates from most malignant
neoplasms are substantially higher in men than in women.
The reason for the substantial differentials in gender may represent the difference
in endogenous factors, such as the sex hormones, physiological and anatomical
characteristics. The disparity may act as a promoter of disease pathogenesis or as a
protective factor. Thus, cervical cancer occurs in women, and prostate cancer occurs
in men. Besides endogenous factors, differences in exposure to environmental risk
factors may contribute to differences in risk between male and female. Hence, the
2 Distribution of Disease 33

Fig. 2.6 Lung cancer incidence rates in a city based on age-birth cohort curves

high male-female incidence disparity for leptospirosis and schistosomiasis is largely

a consequence of exposure to contaminated water in men through certain occupa-
tion, while much of the occupational poisoning and accidental fall in men relative to
women can be explained by contacting more exposure in some hazardous pro-
fessions. Additionally, differences in social, cultural, and behavioral habits between
men and women are also possible causes of the differences in disease distribution.
Lung cancer is largely associated with tobacco smoking in men relative to women.

2.3.3.3 Ethnic and Racial Group

Disparities in some diseases among racial and ethnic minorities largely reflect that
race and/or ethnicity is another important risk factor affecting the distribution of
disease in the population. Interpretation of ethnic differences in disease risk may
consider these factors, such as genetic background, geographical environment,
religious beliefs, customs, way of life, etc.
According to the data from American Cancer Society [4], Non-Hispanic blacks
have the highest incidence and mortality rates of cancer, while Asian and Pacific
Islanders have the lowest rates in both genders compared to other ethnic groups
(Table 2.2).

2.3.3.4 Occupation

There is a close relationship between occupation and diseases. The distribution and
severity of diseases are different in many occupation categories due to the different
34 S. Zheng

Table 2.2 Incidence and mortality rates for cancers by race and ethnicity, US, 2009–2013 (per
100,000)
Male Female
Race and ethnicity Incidence Mortality Incidence Mortality
Non-Hispanic White 519.3 204.0 436.0 145.5
Non-Hispanic Black 577.3 253.4 408.5 165.9
Asian and Pacific Islander 310.2 122.7 287.1 88.8
American Indian and Alaska Native 426.7 183.6 387.3 129.1
Hispanic/Latino 398.1 142.5 329.6 97.7
Data from the American Cancer Society. Cancer Facts & Figures 2017 [4]

exposure profiles. For example, benzene exposure in the shoemaking workers

increases susceptibility to leukemia; coal miners are predisposed to pneumoconiosis,
workers with fur processing and animal husbandry are susceptible to anthrax. The
cause of these differences in the distribution of disease is the differences in the
chance of exposure and/or pathogenic factors. Occupational factors, such as labor
conditions, physical labor intensity, mental stress, and social and economic status,
also contribute to differences in risk between occupations and diseases.

2.3.3.5 Marital Status

A number of epidemiological studies have analyzed the relationship between marital

status and disease risk. It is apparent that differences in marital status have an impact
on the risk of disease. Research confirms that married people tend to have lower
mortality rates than single people, and divorced people have the highest mortality
rates. Meanwhile, the incidences of suicide and violence tend to be higher among
divorced persons, which indicates that divorce events may cause a lot of adverse
effects on mental and psychological aspects, resulting in diseases. Moreover, cousin
marriage may contribute to the increased incidence of congenital malformation and
hereditary diseases, seriously influencing population quality. The main difference in
determining whether persons in marriage get health advantages or if there are certain
characteristics of good health or long life might be in favor of an individual’s
predisposition to marriage.

2.3.3.6 Behavior and Lifestyles

Many diseases are associated with bad behavior and lifestyles. Studies have shown
that about 60–70% of all chronic diseases, such as malignant tumor, cardiovascular
disease, and diabetes, were caused by unhealthy behavior and lifestyles. Common
adverse behaviors include smoking, alcoholism, drug addiction, lack of physical
activity, unsafe sex, and mental stimulation. Previous studies have confirmed that
smoking can cause lung cancer, oral cancer, pharyngeal cancer, laryngeal cancer,
2 Distribution of Disease 35

esophageal cancer, and bladder cancer, a significant dose-response relationship

between cancer risk and smoking was also identified. Intravenous drug use, homo-
sexual behavior, and unsafe sexual behavior may increase the spread of sexually
transmitted diseases. Thus, researches based on the distribution of behavior and
lifestyles are helpful to explore the etiology and formulate the strategies for preven-
tion and control.

2.3.4 Combinations

In epidemiologic studies, there is a need to comprehensively study the disease

distribution of time, place, and person to know the distribution characteristics and
potential risk factors and make relative control measures.
Migrant epidemiology provides useful insights into the relative importance of
environmental exposures and genetic factors in the etiology of disease. Migrant
refers to a group of people who have moved from their former residence to another
place of residence and have permanently changed their place of residence. Migrant
epidemiology is a method of etiological research that analyzes the causes of disease
by comparing the morbidity of migrants with the population of their country of
origin or migrants with the population of their new host country where they have
settled. Studies of disease risk in migrants may provide useful information on
whether the disease is mainly caused by genetic or environmental factors. The
principles of migrant studies are shown in the following two aspects.
① If the disease is mainly caused by environmental factors, the incidence or
mortality rate of the disease in migrants is different from that of the population
from which they originated but is close to those of the new host country in which
they have settled.
② If the disease is mainly caused by genetic factors, the incidence or mortality rate
of the disease in migrants is similar to that of the population from which they
originated but is different from those of the new host country in which they have
settled.
Generally, the risk of disease among migrants may be influenced by the changing
of the living environment and conditions. Meanwhile, the related factors, such as
medical and health care, the population characteristics including age, gender, edu-
cational level, socio-economic status, religious belief, the length of time for migrants
in the new country of residence, and the number of generations, should be consid-
ered in migrant epidemiology.
Chapter 3
Descriptive Study

Zhenxing Mao and Wenqian Huo

Key Points
• Descriptive studies are mainly used by observing, collecting, and analyzing
relevant data to describe the distribution of disease, health status, and exposure
and generate hypotheses for further investigations.
• Descriptive studies mainly include case and case series report, cross-sectional
studies, and ecological studies.
• The ability of descriptive studies to prove whether it is a causal association or
coincidental phenomenon between exposure and outcome is limited.
• Selection bias, information bias, confounding bias are three major sources of bias
in cross-sectional study. Ecological fallacy and confounding factors are the main
limitations in ecological study.
Descriptive study, also known as descriptive epidemiology, is the most basic type of
epidemiological research method. Descriptive studies are mainly used for describing
the distribution of disease, health status, and exposure and generating hypotheses for
further investigations but cannot tell causal relations between disease and exposure.
Descriptive studies are also mainly used for ascertaining high-risk individuals and
evaluating the effects of public health measures, etc. Descriptive studies mainly
include case and case series reports, cross-sectional studies, and ecological studies.

Z. Mao (✉) · W. Huo (✉)

College of Public Health, Zhengzhou University, Zhengzhou, China
e-mail: [email protected]

© Zhengzhou University Press 2023 37

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_3
38 Z. Mao and W. Huo

3.1 Introduction

3.1.1 Concept

Descriptive study is a research method that describes the distribution of diseases or

health status and their influencing factors at different times, regions, and populations
without changing the current disease status and exposure characteristics of the
subjects.

3.1.2 Characteristics of Descriptive Studies

1. Descriptive studies take observation as the main research method and do not
impose any intervention measures on research subjects. Only by observing,
collecting, and analyzing relevant data do descriptive studies analyze and sum-
marize the distribution of diseases, health conditions, relevant characteristics, and
exposure factors.
2. Descriptive studies generally do not set up a control group. And the ability to
prove whether it is a causal association or coincidental phenomenon between
exposure and outcome is limited. However, it could provide a preliminary
contribution to subsequent studies.
3. Descriptive studies have a shorter duration. The distribution of disease and health
status in a population is typically analyzed for transient or temporal characteris-
tics. However, it is easy to implement. The distribution of disease and risk factor
distribution can be obtained in a relatively short time.

3.1.3 Application

1. To describe the prevalence of disease in different regions and different population

characteristics. Continuous descriptive studies at different intervals can also
provide time trend data of disease.
2. To describe the regional, population, and temporal distribution of risk factors.
3. To provide etiological clues and form a preliminary etiological hypothesis.
4. Through descriptive research, patients at early or different stages can be found
and accept early treatment. At the same time, patients with different disease stages
and different infection patterns in the population can also be found. So it can be
used to study the natural history of diseases.
5. To provide baseline data as the basis for the longitudinal study.
6. Descriptive studies can evaluate the effectiveness of preventive and control
measures in the same population before and after the implementation of
interventions.
3 Descriptive Study 39

3.2 Case and Case Series Report

3.2.1 Concept

Case reports usually study a newly discovered or specific disease and its character-
istics. A complete case report includes the patient’s epidemiological data, such as
pre-onset lifestyle characteristics and history of exposure to suspected risk factors. In
the case of infectious diseases, attention should also be paid to investigating and
reporting the possible exposure to patients, animals, and the environment before and
after the onset of illness.
Case series reports are conducted on the basis of case reports used for describing a
series of clinical features or cases with similar diagnoses. The content of the case
series report is exactly the same as that of the case report mentioned above. However,
it should be emphasized that the case series reports should pay more attention to the
demographic characteristics of each case, especially the similarity of risk factor
exposures and clinical characteristics. Focusing on the chronological sequence of
cases and their interconnections is more conducive to forming etiological hypothe-
ses. Generally, case series reports often provide evidence better than case reports.
Hospitals are important places to detect the potential new and special cases; case
reports and case series reports are usually carried out by clinicians. Only those with
systematic epidemiological training and keen insight can catch abnormal cases in
daily diagnosis and treatment activities and report to the local CDC in time.
Measures are also taken to prevent further spread of the disease. Both approaches
are applicable to infectious and chronic non-communicable diseases.

3.2.2 Application

3.2.2.1 Identifying New Diseases

When a new disease occurs, it is necessary to describe the clinical, demographic, and
lifestyle characteristics of the patients, behavioral risk factors, the characteristics of
working and living environment in detail. Then, we explore the possible causes for
diagnosis and make prevention. On the basis of the above research contents,
clinicians can also evaluate treatment measures and effects and expand the research
contents.

3.2.2.2 Establishing the Diagnosis

Based on the clinical symptoms, signs, and laboratory examination results of the
patients provided by the case reports and case series reports, by combining with the
patient’s demographic characteristics and epidemiological data (lifestyle
40 Z. Mao and W. Huo

characteristics, targeted risk factor exposure history, time and place of onset, etc.),
summarizing the common clinical and epidemiological characteristics of patients,
diagnostic criteria can be established for the identification and diagnosis of subse-
quent similar diseases.

3.2.2.3 Forming an Etiological Hypothesis

From the characteristics of individual cases, it can be preliminarily speculated that

some characteristics may be associated with the onset of disease. The characteristics
of multiple patients can be obtained from the case series reports. Analyzing the
characteristics of these patients can provide more information about the relationship
between exposure and disease. On these bases, it can form a preliminary hypothesis
that a certain characteristic may be the cause of the disease. However, the power to
provide evidence is very weak because of the limitations of this approach. It is a very
preliminary etiology suggestion and further research using other epidemiological
methods and causal demonstration is needed to validate this etiological hypothesis.

3.2.2.4 Identifying Early Disease Outbreaks and Epidemics

The early manifestations of disease outbreaks and epidemics usually occur in one
case and then in several cases, followed by more cases of the same characteristics in
susceptible contacts. If the outbreak is not identified and controlled early, the disease
can continue spreading through a population, leading to outbreaks and epidemics.
Therefore, clinicians should have a keen epidemiological thought and vision. When
encountering unusual disease or disease symptom and sign, they should be very
vigilant. If this may be a sign of an early outbreak and epidemic of a certain
infectious disease, clinicians should report to the local CDC timely and take
corresponding preventive and control measures.

3.2.3 Case

3.2.3.1 Estrogen Chemical Bisphenol a and Breast Cancer

A case series report described 15 cases of breast cancer in young women. Nine of the
women reported consuming food packaged with estrogenic chemical bisphenol A
(BPA) at least once a week, and urine samples of nine patients demonstrated the
presence of BPA.
3 Descriptive Study 41

3.2.3.2 Occupational Exposure to Vinyl Chloride and Hepatic

Hemangioma

In 1974, Creech and Johnson reported that three of the workers in the vinyl chloride
plant were found to have hepatic hemangioma. Three of these patients are clearly
unusual in such a small population, and it is easy to form the cause hypothesis that
“the occupational exposure to vinyl chloride caused the occurrence of hepatic
hemangioma.” In the following year, this hypothesis was confirmed by data from
two analytical studies. If there is only one patient, it is not enough to form the cause
hypothesis.

3.2.3.3 AIDS Discovery Case

From October 1980 to May 1981, a report of pneumocystis pneumonia was found
among young, healthy gay men and women in Los Angeles, United States. This
series of reports was unusual because pneumocystis pneumonia previously only
occurred in elderly cancer patients with inhibition of the immune system due to
chemotherapy. At the beginning of 1981, many cases of Kaposi’s sarcoma were
found in young gay men, which is also a noteworthy new discovery. Because this
malignant tumor always occurs in the elderly, and the chances of men and women
are equal. As a result of these extraordinary discoveries, the US Centers for Disease
Control and Prevention immediately implemented monitoring to determine the
severity of the problem and developed diagnostic criteria for this new disease. It is
quickly noted by monitoring that homosexuals have a high risk of developing the
disease. Subsequent case reports and serial case reports indicate that AIDS can also
occur through blood transmission in intravenous drug users, blood transfusion
patients receiving blood transfusions, and hemophilia patients with blood products.
This descriptive data provided clues for the design and implementation of analytical
studies and subsequently identified a range of specific risk factors for AIDS. Serum
obtained from these cases and comparable controls helped identify the pathogen of
AIDS, the human immunodeficiency virus (HIV).

3.2.4 Bias

1. The results are of high promiscuity. The patient is in a natural clinical environ-
ment, and the doctor may not be able to control the patient’s ability to seek and
receive other treatment or control the patient’s diet and daily life, which may
affect the clinical outcome of the disease.
2. The absence of a control group precluded causal inference.
3. The results are less generalizable. Because cases and case series reports are
individual. Strictly, it is almost impossible to find other cases of the same
42 Z. Mao and W. Huo

condition in reality. Usually, based on their own knowledge and experience,

doctors would choose the case reports which have the most consistent key
characteristics for reference.
4. There is a serious publication bias.

3.2.5 Limitation

Although case reports and case series reports are useful in forming etiological
hypotheses, their limitations may overrule causal inference.
1. The incidence of disease cannot be obtained from case reports and case series
reports. The case report/case series report lacks the population of patients with a
disease that is necessary to calculate the disease rate. For example, when calcu-
lating the proportion or incidence of breast cancer in women exposed to BPA, the
total number of people exposed to BPA or the total number of years must be clear.
2. Case reports and case series reports lack a control group. In the above example,
60% (9/15) of the 15 breast cancer cases were exposed to BPA. The exposure rate
appears to be high, but what is the exposure rate in women who do not have breast
cancer? This comparison is key to the hypothesis that BPA may be the cause of
breast cancer, but it is absent in case reports and case series reports.
3. The cases described in case reports and case series reports are often highly
selective subjects, which could not represent the general population well. For
example, 15 cases of breast cancer may be from a community hospital with the
same severe air pollution or other potential carcinogen concentrations. In this
case, a reasonable estimate of the incidence of breast cancer in women in the same
community that is not exposed to BPA is needed to infer the relationship between
BPA and breast cancer so as to avoid overestimating the link between the two. At
the same time, these highly selected cases are highly likely to be reported early,
and more cases need to be accumulated, including atypical cases of clinical stages
(especially in the middle and late stages), to see the complete history of the
disease.
4. There is sampling variability in case reports and case series reports because there
might be large natural variations as the disease progresses. The number of cases
needs to be increased to estimate the incidence of disease accurately and eliminate
the effects caused by chance or sample variation.

3.3 Cross-Sectional Study

3.3.1 Concept

Cross-sectional study is an epidemiological study that describes the distribution of

disease or health status among a specific group of population at a specific time and
3 Descriptive Study 43

explores the relationship between variables and disease or health status. Cross-
sectional study can get the prevalence of diseases, so it is known as prevalence
study. Through cross-sectional study, the occurrence of certain diseases, abnormal-
ities, and vital events in the population can be learned about.

3.3.2 Application

1. To describe the distribution of diseases or health status and provide clues for
disease etiology study.
2. Identifying high-risk groups is the first step in early detection, diagnosis, and
treatment for chronic diseases.
3. Repeated cross-sectional surveys at different stable stages can not only obtain
baseline data of other types of epidemiological studies but also can evaluate the
effectiveness of disease monitoring, vaccination, and other prevention and con-
trol measures by comparing the prevalence differences at different stages.
Cross-sectional study is the basis and starting point of epidemiological research as
well as one of the foothold of public health decision-making. It is a prominent
position in epidemiology. Cross-sectional study could not only accurately describe
the distribution of disease or health status in a population but also explore the
relationship between multiple exposure and disease. But the statistical correlations
between disease and exposure revealed by cross-sectional study, which only pro-
vides clues to establish causal associations, are derived from analytical studies and
cannot be used to make causal inferences.

3.3.3 Classification

Cross-sectional study can be divided into census and sampling survey according to
the scope of research objects involved. In the actual work, the use of census or
sampling survey mainly depends on the purpose of the research, the characteristics
of the research topic, funds, manpower, material resources, and implementation
difficulty.

3.3.3.1 Census

Concept

Census refers to the survey of all the people in a specific time or period and within a
specific range as research objects. A specific time or period means a short time. It can
be a certain time or a few days. For too long, disease or health conditions in the
44 Z. Mao and W. Huo

population could change, which may affect census results. A specific range refers to
a particular area or population.

Purpose

① Early detection, diagnosis, and treatment can be achieved through census, such
as cervical cancer screening in women.
② The prevalence of chronic diseases and the distribution of acute infectious
diseases, such as the prevalence of hypertension among the elderly and the
distribution of measles in children, can be obtained through the census.
③ Through a census, the health status of local residents can be obtained, such as
residents’ diet and nutrition status survey.
④ The distribution of disease and its risk factors can be comprehensively under-
stood through census, and the relationship between risk factors and disease can be
preliminarily analyzed to provide clues for etiological research.
⑤ In a census, all subjects are investigated through a questionnaire or physical
examination. In this process, health education could be conducted to popularize
medical knowledge.
⑥ The normal range of index of all sorts of physiology and biochemistry of the
human body can be obtained, just like the measurements of teenage height and
weight.

Conditions for Carrying out the Census

① Sufficient manpower, material resources, and equipment are available for case
detection and treatment.
② The prevalence of diseases should be higher so that more patients can be found
and the benefits of census can be improved.
③ The disease detection method should be simple, easy to operate, and easy to
implement in the field. The experiment should have high sensitivity and
specificity.

Strengths and Limitations

Strengths
Census surveys all members of a defined population, and there is therefore no
sampling error in the census, and it is relatively simple to determine the respondents.
Census can provide a comprehensive understanding of the health status and the
distribution of diseases or risk factors in a population to establish physiological
reference values.
3 Descriptive Study 45

All cases in the population can be found through the census, which provides clues
for etiological analysis and research to help with prevention.
Through the census, a comprehensive health education and health promotion
activities can be carried out to publicize and popularize the medical knowledge.

Limitations
① It is not suitable for the investigation of disease with low prevalence and
complex diagnosis methods.
② Due to the heavy workload and short survey period, it is difficult to carry out an
in-depth and detailed investigation, and there may be missed diagnosis and
misdiagnosis. The proportion of no response may be high, affecting the repre-
sentativeness of the research results.
③ Due to the large number of staff participating in the census, the variety of their
proficiency in techniques and methods would increase the difficulties to control
the quality of the survey.
④ Only prevalence or positive rate can be obtained, but not incidence.
⑤ Due to the relatively large scope of population involved in census, more
manpower, material resources, and time are consumed in research.

3.3.3.2 Sampling Survey

Concept

Through random sampling, a representative sample of the population at a specific

time point and within a specific range is investigated, and the range of parameters is
estimated by the sample statistics, i.e., the overall situation of the population is
inferred through the investigation of the research subjects in the sample. In the actual
investigation work, there is no need to carry out the census if it is not for the purpose
of early detection and early treatment of patients but only to describe the distribution
of disease.
The basic requirement of the sampling survey is that the results obtained from the
sample can be extrapolated to the entire population. For this reason, the sampling
must be randomized, and the sample size must be sufficient (representative).

Strengths and Limits

Compared with the census, the sampling survey has the advantages of saving time,
manpower, and material resources. At the same time, due to the small scope of the
investigation, it is easy to do it in detail. However, the design, organization, and
implementation of the sampling survey and data analysis are complex. Duplications
and omissions are not easy to find, so they are not suitable for populations with large
variations. Sampling is also not appropriate for diseases with low prevalence because
46 Z. Mao and W. Huo

small samples could not provide the information required. In addition, if the sample
size is large enough up to 75% of the population, a census may be a better choice.

3.3.4 Design and Implementation

An excellent design scheme is the premise of successful implementation in a

research project. It is necessary to pay special attention to the representativeness of
the selected subjects in the sampling survey, which is the prerequisite for the overall
inference of the research results. Random sampling and avoiding selection bias are
important conditions to ensure the representativeness of the research objects.

3.3.4.1 Clarifying the Purpose and Type of Investigation

According to the research problems expected to be solved, the purpose of the survey
should be confirmed, such as to know the distribution characteristics of diseases and
the exposure of risk factors or to carry out group health examination. Then, census or
sampling survey can be determined to choose based on the specific research purpose
according to the specific research purpose.

3.3.4.2 Identifying Subjects

According to the research objective and the distribution characteristics, geographical

scope and time point of investigation, and the feasibility of carrying out the survey in
the target population, the research object was determined. In the design, the research
object can be defined as all or part of the residents in a certain area. It can also be
composed of the floating population at a certain point. It can also be selected as the
research object for some special groups. For example, the professional workers
exposed to a certain chemical substance can be collected to study skin cancer.

3.3.4.3 Determining Sample Size

The sample size is the minimum number of observations required to ensure

the reliability of the research results. The factors determining the sample size of
the present study come from many aspects, but the main influencing factors include
the following aspects:
1. Expected incidence (P) or standard deviation (S). The largest sample size is
required when the current incidence rate is 50%.
2. The accuracy of the results of the survey requirements, i.e., the greater the
allowable error (d), the smaller the sample size required.
3 Descriptive Study 47

3. Significance level (α) or the probability of the type I error. The smaller the test
level, the more samples are required. For the same test level, the sample content
required by the bilateral test is larger than that required by the unilateral test,
which is usually taken as 0.05 or 0.01.
Statistical variables are generally divided into two categories: numerical variables
and categorical variables. Therefore, the formula used to estimate the corresponding
sample size is also different.

Estimating Sample Size for Numerical Variables

The following formula is used to estimate the sample size for random sampling:

2
ZαS
n= ð3:1Þ
d

In the formula, n is the sample size, d is the allowable error, i.e., the difference
between the sample mean and the overall mean, which is determined by the survey
designer according to the actual situation. S is the standard deviation, Zα is the
normal critical value at the test level and α is usually taken as 0.05 and Zα = 1.96.

Estimating of Sample Size for Categorical Variables

The following formula is used to estimate the sample

t 2 PQ
n= ð3:2Þ
d2

In the formula, n is the sample size, P is the estimated overall prevalence,

Q = 1 - P, d is the allowable error, t is the statistic of hypothesis testing.
Assuming that d is a fraction of P, when the permissible error d = 0.1P, α = 0.05,
t = 1.96 ≈ 2, then formula (3.2) can be written as:

n = 400 × Q=P ð3:3Þ

The above sample size estimation formula only applies to the data of Binomial
distribution, i.e., np > 5, n(1 - p) > 5. Otherwise, it is advisable to estimate the
sample size by Poisson distribution. The expected value of Poisson distribution and
the confidence interval table can be used to estimate the sample size.
The sample size calculation method introduced above is only applicable to simple
random sampling. However, in field epidemiologic investigation, cluster sampling is
more commonly adopted because it is easy to organize and implement. The sampling
error of cluster sampling is large. If the sample size of simple random sampling is
48 Z. Mao and W. Huo

Table 3.1 The confidence 0.95 0.90

interval of expected value for
1 - 2α Lower Upper Lower Upper
the Poisson distribution
0 0.00 3.69 0.00 3.00
1 0.03 5.57 0.05 4.74
2 0.24 7.22 0.36 6.30
3 0.62 8.77 0.82 7.75
4 1.09 10.24 1.37 9.15
5 1.62 11.67 1.97 10.51
6 2.20 13.06 2.61 11.84
7 2.81 14.42 3.29 13.15
8 3.45 15.76 3.93 14.43
9 4.12 17.08 4.70 15.71
10 4.30 18.29 5.43 16.96
11 5.49 19.68 6.17 18.21
12 6.20 20.96 6.92 19.44
13 6.92 22.23 7.69 20.67
14 7.65 23.49 8.46 21.89
15 8.40 24.74 9.25 23.10
20 12.22 30.89 13.25 29.06
25 16.18 36.90 17.38 34.92
30 20.24 42.83 21.59 40.69
35 24.38 48.68 25.87 46.40
40 28.58 54.47 30.20 54.07
45 32.82 60.21 34.56 57.69
50 37.11 65.92 38.96 63.29

calculated to estimate the sample size of cluster sampling, the sample size will be
smaller. Therefore, it is advocated to multiply the sample size for simple random
sampling by 1.5 as the sample size for cluster sampling.
Example The estimated incidence of colorectal cancer in a city is 30/100,000. How
many people should be sampled?
If you take a random sample of 10,000 people, according to the incidence,
30/100,000, the expected number of survey cases is 3. As Table 3.1 shows, when
the expected number of cases (1 - 2α) is 2, the 95% confidence interval is
0.24–7.22, which means there may be no case. If there was at least 1 case with
colorectal cancer, the lower limit of 95% confidence interval is 1.09, and the
expected number of cases will be 4. In order to reach at least 4 cases of colorectal
cancer patients in the survey results, 4/X = 30/100,000, so X = 13,334. Finally,
13,334 people should be investigated at least.
In actual work, the sample size can be appropriately increased to avoid errors
between the estimated and actual incidence rate.
3 Descriptive Study 49

3.3.4.4 Determining the Sampling Method

There is non-random sampling and random sampling. The former includes a typical
investigation. A random sampling must follow the randomization principle, which
means ensuring that everyone in the population has a known, non-zero probability of
being selected as the research object to ensure representativeness. If the sample size
is large enough, the data are reliable, and the analysis is correct, the results can then
be extrapolated to the population.
The commonly used random sampling methods are simple random sampling,
systematic sampling, stratified sampling, cluster sampling, and multi-stage
sampling.

Simple Random Sampling

Simple random sampling is the simplest and most basic sampling method. The
important principle is that each subject is selected with the equal probability (n/N ).
The specific method is as follows: first, all observation objects are numbered to form
a sampling frame. Then, some observation objects are randomly selected from the
sampling frame by drawing lots or using random number table to form samples.
Simple random sampling is the most basic sampling method and the basis of other
sampling methods. However, when the overall number of the survey is large, it is
difficult to number each individual in the population. Moreover, the sample is
scattered, which is not easy to organize and implement. Therefore, it is rarely used
in epidemiological studies.

Systematic Sampling

Systematic sampling, also known as mechanical sampling, is to number individuals

in a population in order and then randomly select a number as the first survey
individual, while the others are selected according to some rules. The most com-
monly used systematic sampling is isometric sampling, in which all units within the
population are sorted and numbered. According to the sample size, the
corresponding individual samples are mechanically selected in specific sampling
space. The sample numbers taken are:

i, i þ k, i þ 2k, i þ 3k, . . . , i þ ðn - 1Þk ð3:4Þ

k is sampling space; n is sample size; i is the randomly selected starting number.

Example If there are 250,000 observation objects in a population, 1000 objects are
to be selected for investigation. Sampling can be carried out through systematic
sampling. Firstly, the sampling interval is k = 250,000/1000 = 250, and then one
number was randomly selected from the first unit by the simple random sampling
50 Z. Mao and W. Huo

method as the starting point. If i is 25, the individual numbers were selected
successively: 25, 275, 525, 775, 1025, etc.
Compared with simple random sampling, systematic sampling saves time, and
the sample distribution is more uniform and representative. However, the disadvan-
tage of systematic sampling is that when the observed individuals in the population
have a periodic increasing or decreasing trend, it would produce bias, and the
representativeness of the obtained samples will be declined, e.g., the seasonality of
diseases, periodic changes of investigation factors.

Stratified Sampling

Stratified sampling refers to dividing the population into several subpopulations

according to certain characteristics, and then conducting simple random sampling
from each subpopulation to form a sample. The smaller the intra-group variation and
the greater the inter-group variation, the better. Stratified sampling is more accurate
than simple random sampling. Moreover, it is more convenient for organization and
management.
Stratified sampling is divided into two categories: one is called proportional
allocation stratified random sampling, i.e., the sampling proportion within each
subpopulation is equal. The other is called optimum allocation stratified random
sampling, i.e., the sampling proportion within each subpopulation is unequal. The
sampling proportion with small inter-group variation is small, and with large inter-
group variation is large.

Cluster Sampling

Cluster sampling refers to dividing the population into several groups and selecting
some groups as observation samples. If all the selected groups are all the respon-
dents, it will be a simple cluster sampling. If some individuals are investigated after
sampling again, it is called two-stage sampling. The characteristics of cluster
sampling are as follows:
① It is easy to organize, convenient to try, and implement;
② The smaller the difference between groups, the more groups are extracted, and
the higher the accuracy will be;
③ The sampling error is large, so it is usually increased by 1/2 on the basis of the
simple random sample size estimation.
The above-mentioned four basic sampling methods are introduced. When the
sampling method is fixed, the order of sampling error is from large to small: cluster
sampling, simple random sampling, systematic sampling, and stratified sampling.
3 Descriptive Study 51

Multi-Stage Sampling

The sampling process is carried out in multi-stages, with each stage using a different
sampling method. Combined with the above sampling methods, multi-stage sam-
pling is commonly used in large epidemiological studies. For example, the
InterASIA Study (International Collaborative Study of Cardiovascular Disease in
Asia) has adopted the following multi-stage sampling method:
The first stage: the sampling unit is the province and city. Four economically and
geographically representative cities were drawn from the south and north, respec-
tively. Beijing and Shanghai were included in the northern and southern samples,
respectively, and a total of 10 provinces and municipalities were drawn. It should
be noted that in order to fully consider the geographical and economic level of
representation, random sampling is not used at this stage, but the random sam-
pling method is used in the next three stages.
The second stage: the sampling units are counties and urban areas. A county and an
urban area were randomly selected from the provinces and cities in the first stage,
and ten counties and ten urban districts were drawn.
The third stage: the sampling unit is a street, town, or township. Street or town
(or township) is randomly selected from each urban area and county, and a total of
ten streets and ten towns (or townships) are drawn.
The fourth stage: the sampling unit is an individual. The list of residents of all streets
or towns will be used as a sample source (limited to 35–74 years old), and each
site will have 400 male and female residents.
The above sampling methods used in four stages are called multi-stage sampling.
Multi-stage sampling can make full use of the advantages of various sampling
methods and overcome their shortcomings. The disadvantage of multi-stage sam-
pling is that the demographic data and characteristics of each sub-group should be
collected before sampling. Also, the statistical analysis of the data is complicated,
such as the sampling weight of the complex sampling design when calculating the
sampling error.

3.3.4.5 Data Collection, Collation, and Analysis

In a cross-sectional study, the method of data collection cannot be changed once it is

determined. Consistency must be maintained throughout the study to avoid hetero-
geneity of data. The data collection process should pay attention to the unification
definition of exposure and the criteria of disease. All personnel involved in the
inspection or testing must be trained to avoid measurement bias with unified
investigation and testing standards.
52 Z. Mao and W. Huo

Determining the Data to be Collected

The most basic principle of the cross-sectional study is whether the subject has a
certain disease or characteristics, and the investigator uses grading or quantitative
methods as much as possible. In addition, other information, such as social and
environmental factors, need to be collected to illustrate the distribution and related
factors. The relevant information collected generally includes the following:
① Basic information about the individual, including age, gender, ethnicity, edu-
cation level, marital status, per capita monthly income, etc.
② Occupational and exposure status, including nature, type, position, and working
years.
③ Lifestyle and health conditions, including diet, smoking history, drinking
history, physical exercise, depression, anxiety, medical history, disease
history, etc.
④ Women’s reproductive status, including menstrual and obstetrical histories, use
of contraceptives, and hormones.
⑤ Environmental information, expressed in objective and quantitative indicators.
⑥ Prevalence, infection rate, etc.

Investigator Training

Before the investigation, the investigators should be trained uniformly following a

standard protocol. The consistency of the methods and standards for collecting data
can be guaranteed. Investigator training is an important part to ensure the accuracy
of data.

Data Collection Methods

In a cross-sectional study, there are three methods for collecting data. The first one is
by laboratory measurement or examination, such as blood glucose detection, blood
lipid detection, etc. The second way is to investigate the subject through the use of a
questionnaire to obtain information on exposure or disease. The third way is to use
routine data. For example, get data from the Center for disease control (CDC) and
electronic disease records.

Data Collation and Analysis Methods

Data collation refers to checking the integrity and accuracy of the original data
carefully, filling in the missing items, deleting the duplicates, and correcting the
errors. Disease or a state of health is verified and classified according to clearly
3 Descriptive Study 53

defined criteria. Then it can be described according to different spaces, time, and the
distribution in the crowd.
In data analysis, the population can be further divided into exposed and
non-exposed groups or different levels of exposure population. The differences in
disease rate or health status between the groups can be compared and analyzed. The
subjects can also be divided into disease and non-disease groups to evaluate the
relationship between factors (exposure) and disease.
① Description of the demographic characteristics. A detailed description of the
demographic characteristics (e.g., gender, age, education level, occupation, mar-
ital status, and socioeconomic status) can help to easily understand the basic
characteristics of the research object and can be used to compare with other
studies.
② Analysis of the distribution characteristics of the disease: According to the
characteristics of the different subjects (gender, age, education level, occupation,
marital status, socioeconomic status, etc.), regional characteristics (urban, urban,
north, south, mountain, plain, or administrative division, etc.) and time charac-
teristics (season, month, year, etc.) are grouped, the prevalence of a disease or the
mean and sampling error of a certain variable are calculated and compared and the
correct statistical method is used to test the differences between the different
groups.
③ Analysis of the relationship between exposure factors and disease: Compare the
prevalence of a disease or the mean value of a numerical variable according to the
presence or absence of exposure factors or the level of exposure. It is also possible
to calculate an odds ratio (OR) to estimate the association and association strength
in an epidemiological method (such as a case-control study). Not only univariate
analysis but also multivariable adjustments to calculate the ORs are required.
What needs to be emphasized here is that cross-sectional study can only provide
preliminary clues to the cause.

3.3.5 Bias and Control

Bias is the systematic errors generated in the process from design, implementation,
to analysis, as well as the one-sidedness in the interpretation or inference of the
results, which leads to the tendency of a difference between the research results and
the true value, thus mischaracterizing the relationship between exposure and disease.
The common bias in cross-sectional studies includes selection bias, information bias,
and confounding bias.
54 Z. Mao and W. Huo

3.3.5.1 Selection Bias

Selection bias is the systematic error caused by the differences of the characteristics
between the included subjects and those who were not included in the study. It
mainly includes the following aspects:
1. Selective bias: In object selection process, due to not strictly sampling, the objects
are selected subjectively, which results in the deviation of the research samples
from the population. For example, when you want to know the prevalence of
hepatitis B in one city last year, if the sample were only information collected
from the hepatitis specialized hospital, the prevalence must be higher than the
actual rate in the general population.
2. Non-response bias: During the investigation, the subjects did not cooperate or
were unable or unwilling to participate for various reasons, resulting in a missed
investigation. If the response rate is too low (less than 80% or even 85%), it could
produce a non-response bias, and it is more difficult to apply the results to
estimate the source population.
3. Survivor bias: In cross-sectional study, survivors of disease are often selected as
subjects. Current cases and deaths may have different characteristics, which could
not summarize the overall situation. Therefore, the results have some limitations
and one-sidedness.

3.3.5.2 Information Bias

Information bias is a systematic error that occurs when information is obtained from
a research subject during the investigation process. Information bias can come from
subjects, investigators, measuring instruments, equipment, and methods.
1. Respondent bias includes recall bias and reporting bias: The subjects were biased
by unclear or completely forgotten disease history, drug application history, and
risk exposure history.
2. Investigator bias: The bias occurs in the process of collecting, recording, and
explaining information from respondents. One reason is, different investigators
have different results for the same subject, the other is the same investigator has
different results for several surveys of the same subject.
3. Measurement bias is a systematic error caused by inaccurate instrument and
incorrect operation procedure. For example, if the sphygmomanometer is not
calibrated, all measurement results are higher or lower than true value. The
methods of investigation used are not uniform, and bias may occur.

3.3.5.3 Confounding Bias

Confounding bias is caused by confounding factors. If the association between

exposure and disease is analyzed, then there will be confounding bias.
3 Descriptive Study 55

Bias can be avoided or reduced, so it is necessary to carry out quality control in

the research to minimize the occurrence of bias.
1. In the sampling process, keep the randomization principle strictly to ensure the
representativeness of the sample.
2. To improve the compliance and test rate of the respondents, each subject should
be investigated.
3. To correct the measuring instruments, equipment, and testing methods, including
the preparation of questionnaires.
4. To train the investigators, unify survey standards and conduct mutual supervision
and spot checks.
5. After the investigation, reviewing and checking the information is needed.
6. In the process of data collation, the correct statistical analysis method should be
selected, pay attention and identify confounding and influencing factors.

3.3.6 Strengths and Limitations

3.3.6.1 Strengths

1. The implementation time is short, and the results can be obtained quickly. Thus,
the research task can be completed in a short time.
2. Compared with other research types, a cross-sectional study is a relatively
inexpensive method.
3. It could investigate the association between disease and factors and establish a
preliminary etiological hypothesis.
4. A cross-sectional study can provide a basis for making disease prevention and
control plans.

3.3.6.2 Limitations

1. Prevalence, instead of incidence, can only be obtained from a cross-sectional

study.
2. Low-prevalence disease and its influencing factors are not suitable for cross-
sectional study.
3. The time sequence between exposure and disease cannot be determined, so there
is no causal association, and only preliminary etiological clues can be provided.

3.3.7 Cases

Due to the rapid development of the Chinese economy, the diet and lifestyle have
changed greatly. Diabetes, hypertension, hyperlipidemia, and many other diseases
56 Z. Mao and W. Huo

related to diet and lifestyle increased significantly. Li Liming et al. conducted a

survey on the situation among Chinese people in 2002.

3.3.7.1 Purpose and Type of Study

The purpose of this study is to investigate the nutrition and health status of Chinese
residents and to analyze the main factors affecting the nutrition and health status.
Therefore, the cross-sectional study was adopted. Compared with the census, the
sampling survey saves more time and cost. Therefore, multi-stage stratified cluster
sampling was selected.

3.3.7.2 Subjects and Sample Size

The target population was the national resident population. With 95% accuracy and
90% precision, the minimum sample size was estimated at 225,000. In addition,
assuming no response rate of 10%, the final sample size was 250,000.

3.3.7.3 Research Content and Data Collection, Collation, and Analysis

Data collection included questionnaires, medical examinations, laboratory tests, and

dietary surveys. Firstly, demographic characteristics, socioeconomic status, disease
history, smoking, drinking, and physical activity of the individuals were collected
through questionnaires. Secondly, the height, weight, waist circumference, and
blood pressure of the individuals were tested. Thirdly, laboratory tests were
performed on the serum indexes, including hemoglobin, TC, TG, HDL-C, and
LDL-C. Fourthly, the 24-hour retrospective method, food frequency method, and
weighing method were used to carry out the dietary survey.
Finally, 243,206 people were enrolled in this study. After adjusting for age and
region, the national adjustment rate was calculated by direct standardization method.
The results showed that the consumption of cereals was the maximum, and the
dietary structure showed a significant regional difference. The consumption of meat,
fruit, and vegetable oil in urban areas is higher than that in rural areas. While the
consumption of cereals, tubers, and vegetables in rural areas was higher than that in
urban areas. The overweight rate in China is 17.6%, and the obesity rate is 5.6%.
Both overweight and obesity rates increase with age. Obesity rates are higher in
cities than in rural areas among people over the age of 7. The prevalence of anemia is
15.2%, which is significantly higher in young and middle-aged women than in men.
The prevalence of diabetes among Chinese adults is 2.6%, which increases with age.
The prevalence in cities rises faster than in rural population.
3 Descriptive Study 57

3.3.7.4 Conclusion

The nutrition and health status of Chinese population are gradually changing. The
prevalence of anemia reflects the lack of trace elements like iron in Chinese
population. The prevalence of chronic diseases such as overweight, obesity, and
diabetes is increasing rapidly, which has been a threat affecting the health of the
Chinese people. In addition, disease prevention should be targeted because of the
differences in nutrition and health levels between urban and rural populations.

3.4 Ecological Study

3.4.1 Concept

Ecological study is also called correlational study. It is used to analyze the relation-
ship between exposures and diseases by describing the exposure and the frequency
of diseases in different populations.

3.4.2 Type of Study Design

3.4.2.1 Ecological Comparison Study

Ecological comparison study is often used to compare the relationship between the
exposure and the disease frequency in different population groups to provide clues
for the disease cause. For example, the National Cancer Research Center of the
United States has drawn a statistically significant regional difference in the
age-adjusted mortality map of oral cancer between 1950 and 1969. The highest
morality is in the urban areas which are dominated by men in the northeast and
women in the southeast. Smoking may be a risk factor for oral cancer from this
distribution, as smoking in the South is common. Later case-control studies also
supported this cause hypothesis. Immigration epidemiological research method can
also be applied in ecological studies. It is usually used to analyze the relationship
between genetic factors or environmental factors and diseases by comparing the
incidence of immigrants and their children with the incidence of residents of the
original place of residence and residents of the settlement in different areas.

3.4.2.2 Ecological Trend Study

The ecological trend study is to continuously observe the changes in exposure levels
and diseases in the population and describe their trends. The relationship between
58 Z. Mao and W. Huo

factors and disease is judged by comparing changes in disease before and after
exposure changes.
In the implementation of ecological studies, the above two types are often
combined. The suspicious etiology of the disease is explored by studying the
frequency of occurrence of a disease in multiple regions (multiple groups of com-
parisons) and at different times (time trends). For example, some researchers ana-
lyzed the relationship between water hardness and cardiovascular mortality for
gender and age in 63 towns in the UK from 1948 to 1964. It was found that
cardiovascular mortality and water hardness were negatively correlated in all gen-
ders and ages, especially in men. In urban areas with high water hardness, the
increase in cardiovascular mortality was less than in towns with low water hardness.

3.4.3 The Main Application

1. Etiological assumptions related to the disease distribution can be found through

ecological studies. Ecological studies have found that colorectal cancer was more
common in developed countries than in developing countries, considering that
dietary habits or environmental pollution might be related to the incidence of
colorectal cancer.
2. To provide positive or negative evidence for some existing disease causal
hypotheses.
3. It can be used to evaluate the effects of intervention experiments or field exper-
iments. For example, promote low sodium intake in the population and then
compare the changes in the average sodium intake level before and after the
promotion of low sodium salt and the trend of the average blood pressure of the
population to evaluate the effect of low sodium salt intervention.
4. To estimate trends in disease changes. Applying ecological trend studies in
disease surveillance to estimate trends in a disease can help prevent and control
disease. Between 1959 and 1966, the number of deaths from asthma in England
and Wales was associated with a simultaneous increase in sales of bronchodila-
tors. After the cessation of bronchodilators in the pharmacy in 1968, the mortality
rate of asthma decreased significantly. Therefore, the development of a ban on
bronchodilators without prescription was the result of ecological research.

3.4.4 Bias

Disease information in ecological studies is often derived from historical records

(cancer registers, medical records, etc.), while exposure information is often derived
from government agency data (tobacco and alcohol sales, etc.). The accuracy of
these data directly affects the reliability of research results.
3 Descriptive Study 59

3.4.5 Strengths and Limitations

3.4.5.1 Strengths

1. Ecological study can be carried out using existing routine data to save time and
money and then quickly yield results. It takes a long time to measure the
relationship between a biological indicator and a disease through a prospective
study. The preliminary study using an ecological method can narrow the
research risk.
2. For unknown disease etiology, etiological clues for further research can be
provided by an ecological study. This is the most striking feature of ecological
study.
3. In the field of environment or other research, an ecological study is the only
alternative research method when the cumulative exposure of an individual is not
easy to measure. For example, in the study of the relationship between urban air
pollution and lung cancer, it is difficult to estimate the amount of polluted air
inhaled by individuals accurately. At this time, multiple methods of ecological
comparison can be applied for research.
4. When the range of individual exposure in a population is not large enough, it is
difficult to estimate the relationship between exposure and disease. In this case,
ecological comparison studies with multiple populations are more appropriate.
For example, not only are high-fat diet habits similar, but also the intake is
generally high in Western countries. If the relationship between individual fat
intake and coronary heart disease were studied only in Western countries, it
would be difficult to find a relationship. If a comparative study of the low-fat
diet of the Eastern countries was chosen, meaningful results might be found.
5. Ecological studies are appropriate for evaluating population intervention mea-
sures. For example, folic acid deficiency in humans can lead to fetal neural tube
defect, which was first hypothesized through ecological studies. The addition of
folic acid in the pregnant population led to a significant decrease in the incidence
of fetal neural tube defects.

3.4.5.2 Limitations

1. Ecological fallacy: Etiological clues suggested by ecological studies may be

either a true or a false association between disease and exposure. The ecological
fallacy is a misinterpretation of the association between exposure and outcomes
due to an inaccurate assessment.
2. Confounding factors are often difficult to control, especially socio-demographic
and environmental variables. Multicollinearity may affect the correct analysis of
the relationship between disease and exposure.
3. Because the timing sequence between exposure and disease is not easy to
determine, it is difficult to determine the causal relationship between the two
variables.
60 Z. Mao and W. Huo

When conducting an ecological study, do not set too many research questions in a
study. The differences between population groups should be minimized. The inter-
pretation of the results should be compared with other non-ecological results as far as
possible and combined with professional knowledge for comprehensive analysis and
judgment.

3.4.6 Case

In order to analyze and evaluate the relationship between life expectancy and fine
particulate pollutants in the air, a study from the United States compiled the life
expectancy, socioeconomic status, and society of 211 counties in 51 urban areas in
1980 and 2000. Demographic characteristics and concentration of airborne fine
particle contaminants were analyzed using regression models to analyze the rela-
tionship between airborne fine particle concentration and life expectancy, after
adjusting socioeconomic status and demographic variables, as well as the prevalence
of smoking. The results of the study showed that a 10 ug/m3 reduction in the
concentration of fine particulate contaminants was associated with an increase in
life expectancy of 0.61 ± 0.20 years (P = 0.004). And after adjusting the multivar-
iate in the model, the results still remained significant. The results suggested that
reduced fine particulate contaminants in the air can increase life expectancy by 15%.
Chapter 4
Cohort Study

Li Liu

Key Points
• A cohort study is an observational study which begins with a group of people who
are free of an outcome of interest and classified into subgroups according to the
exposure to a potential cause of the outcome. Variables of interest are specified
and measured, and the whole cohort is followed up in order to see how the
subsequent development of new cases of the disease (or other outcomes) differs
between the exposed and unexposed groups.
• There are three types of cohort studies according to the time when information on
exposures and outcomes is collected, namely prospective cohort study, retrospec-
tive cohort study and ambispective cohort study.
• The measures of associations in cohort studies include relative risk, attributable
risk or attributable fraction, population attributable risk or population attributable
fraction, and dose-effect relationship.

4.1 Introduction

4.1.1 Definition

A cohort study is an observational study which begins with a group of people who
are free of an outcome of interest and classified into subgroups according to the
exposure to a potential cause of the outcome. Variables of interest are specified and
measured, and the whole cohort is followed up in order to see how the subsequent
development of new cases of the disease (or other outcomes) differs between the
exposed and unexposed groups.

L. Liu (✉)
School of Public Health, Tongji Medical College, Huazhong University of Science and
Technology, Wuhan, China

© Zhengzhou University Press 2023 61

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_4
62 L. Liu

Fig. 4.1 The design of a cohort study

Exposure means that the subject has been exposed to some substances (e.g.,
heavy metals) or has some characteristics (e.g., being a carrier of a particular
genotype) or behaviors (e.g., alcohol drinking).
The term “cohort” is derived from the Roman army, where it referred to a group
of about 480 soldiers, or one-tenth of a legion. Soldiers remained in the same cohort
throughout their whole military life, similar to members of epidemiologic cohorts.
According to the time of participants entering the study, the cohorts can be
classified into two types: the fixed cohort and dynamic cohort. Fixed cohort means
that all participants are enrolled in the cohort at a fixed time or in a short period of
time, and followed up until the end of the observation period. The participants have
not exited due to other reasons than the outcome, and no new members have joined
it. During the whole period, the cohort remains relatively stable. Dynamic cohort,
also known as an open cohort, refers to a cohort in which the original members
continue to withdraw, and/or new members can join in during the follow-up.
The simplest situation of a cohort study is to recruit one group of population with
a specific exposure and one group without that exposure and then follow up for a
period of time to see if the participants develop the outcome of interest (Fig. 4.1).
The participants must be free of the outcome of interest at the start of the follow-up,
which makes it easier to be sure that the exposure precedes the outcome. After a
period of time, the investigator compares the incidence rates of the outcome between
the exposed and unexposed group. The unexposed group serves as the reference
group, providing an estimate of the baseline amount of the outcome occurrence. If
the incidence rates are substantively different between the exposed and unexposed
groups, the exposure is said to be associated with the outcome. According to the
basic principles of cohort studies, there are some basic characteristics:
4 Cohort Study 63

4.1.1.1 Observational Study

The exposures in cohort studies are not given artificially, but objectively before the
study, which is an important aspect of the difference between cohort studies and
experimental studies.

4.1.1.2 Setting up a Comparison Group

Cohort studies usually set up an unexposed group for comparison during the
research design phase. The control group may come from the same population as
the exposed group or from different populations.

4.1.1.3 From “Cause” to “Outcome”

In the course of the cohort study, we usually know the “cause” (exposure factors)
first, and then look into the “outcome” (disease or death) through longitudinal
observation, which is consistent with experimental research.

4.1.2 Types of Cohort Study

There are three types of cohort studies according to the time when information on
exposures and outcomes is collected (Fig. 4.2).

4.1.2.1 Prospective Cohort Study

In prospective studies, data on exposures are collected at baseline and updated

during the follow-up. The outcomes are not available at the beginning of the cohort
and should be collected during the follow-up. The investigators could use the most
up-to-date measurements to address exposures of interest with minimized bias.
However, the investigators need to wait for a relatively long time until a sufficiently
large number of events occur. For rare outcomes, the follow-up period may span one,
or even several decades.

4.1.2.2 Retrospective Cohort Study (Historical Cohort Study)

Data on the exposures and outcomes are collected from existing records and can
immediately be analyzed. It relies on exposure measurements made before the study
set up, which may be available from demographic, employment, medical, or other
64 L. Liu

Fig. 4.2 Design of the three types of cohort studies

records. Compared with a prospective cohort study design, it is more useful for rare
diseases with a long natural history. By using existing data, the wait time for the
exposure to have any impacts on the risk of outcome could be largely reduced. A
retrospective cohort study is particularly useful in occupational and environmental
epidemiology because if there is a concern that a certain exposure may be a risk
factor, it is not reasonable to wait for a long time to confirm in a prospective cohort
study. The main disadvantage of a retrospective cohort study is that the exposure
data available in records are usually less detailed and accurate than if they were
collected prospectively. A retrospective cohort study can be particularly successful
when biological specimens were collected in the past so that up-to-date laboratory
techniques can be used to detect past exposures. This method could minimize
inaccurate exposure measurements in the past, but the stability of the biomarkers
during long periods of storage is largely unknown.

4.1.2.3 Ambispective Cohort Study

An ambispective cohort study is a design that combines prospective cohort study

with retrospective cohort study. In an ambispective study, a large proportion of
participants are still at risk of the outcome when the retrospective cohort are
4 Cohort Study 65

identified, and the follow-up period can be extended into the future to obtain the
maximum amount of information from the cohort. So an ambispective cohort study
combines the advantages of both retrospective and prospective cohort studies and to
some extent, makes up for their respective deficiencies.

4.2 Design of a Cohort Study

4.2.1 Selection of the Cohort

The study population includes both exposure and control (unexposed) population.
Depending on the purpose and the conditions of the study, the choice of study
population varies.

4.2.1.1 Choice of the Exposure Population

The choice of the exposure population depends on practical considerations and the
study hypotheses. There are usually four sources:

General Population

It refers to a well-defined region of the entire population or its samples. It is

composed of individuals with different exposure factors and is suitable for simulta-
neous observation of multiple exposures and their relationships with multiple dis-
eases. When the exposure group is chosen from the general population, there are two
points to be considered: ① Do not intend to pay attention to the incidence of special
population, but focus on the general population, so that the research results have
universal significance. ② The exposure factors and outcomes of interest are very
common in the general population, so there is no need to choose special populations
or no special population to choose from. The Framingham Study is a well-known
example of a general population cohort.

Occupational Population

If you want to study a suspicious occupational exposure factor and an outcome, you
should select the relevant occupational population as the exposure group. In addi-
tion, records on occupational exposures and diseases are often more comprehensive,
true, and reliable, so it is a very good source for retrospective cohort study.
66 L. Liu

Special Exposed Population

It refers to people with special exposure experiences, which is the only way to study
some rare special exposures, such as selecting people who have undergone radio-
therapy to study the relationship between radiation and leukemia. If an exposure
factor has pathogenic effects, the incidence or mortality of certain disease in special
exposed population should be higher than that that in the general population, which
could facilitate the identification of the association between the exposure and the
disease.

Organized Population

The organized population can be considered as a special form of the general

population, such as school and army members. The selection of such a population
mainly relies on the relevant organizational system to facilitate the efficient collec-
tion of follow-up information. Given the similar occupational experiences, the
occupational people are more comparable.

4.2.1.2 Choice of Control Population

The unexposed group should be comparable to the exposed group in the distribution
of factors that may be related to the outcome of interest except for the exposure. That
is, in case of no association between the exposure and outcome, the outcome would
have the same incidence in the exposed group and the unexposed group. The control
groups mainly include:

An Internal Comparison Group

An internal comparison group includes unexposed members from the same cohort.
Both exposed and unexpected groups are within the selected population. This is
usually the best comparison group since the subjects are similar in a lot of aspects.
For example, if we want to assess the association between yogurt consumption and
the risk of conventional and serrated precursors of colorectal cancer, subjects from
the cohort are categorized into groups according to the amount of yogurt consump-
tion, and the group with the lowest intake is used as an internal comparison group, to
which the other groups are compared [5]. Cohort studies should try to choose an
internal control because it is comparable with the exposed population, easy to
conduct, and able to understand the overall incidence of the study.
4 Cohort Study 67

An External Comparison Group

When it is impossible to take a well-defined cohort and divide it into the exposed and
unexposed groups, the comparison group should be selected outside the cohort,
which refers to “external comparison group.” A potential external comparison group
is another cohort with similar characteristics but without the exposure of interest.
The advantage of this approach is that the follow-up observation can be protected
from the exposure group. The disadvantage is the effort to organize another
population.

General Population as a Comparison Group

It can be considered as a kind of external control. The whole population of the

geographic area where the exposed cohort is located might be selected as a compar-
ison group (unexposed). Since it is highly impractical to follow up the entire
population of a geographic region, incidence rates in the general population are
typically derived from routine statistics, which represents an efficient approach
compared to studying an additional, unexposed cohort. It takes advantage of existing
incidence or mortality statistics across the region, as the morbidity or mortality of the
general population is relatively stable and readily available and can save significant
time and money, but the disadvantage is that information is often not very accurate.
The quality of the information can hardly be checked because it is not collected
directly by investigators. Information on potential confounders (other than age, sex,
and other basic demographic characteristics) is typically not available in the general
population, and confounding by factors such as smoking, cannot be controlled. It
should be noted that the control group may contain some exposed subjects, so the
total control population applies to a small proportion of the total exposure popula-
tion. In practice, instead of using a direct comparison of the incidence of the exposed
group and the general population, a standardized ratio is used. For example, the
standardized mortality rate (SMR) is the ratio of the number of expected morbidity
or mortality figures calculated from the incidence or death of exposed groups to the
total population.

Multiple Comparison Groups

Multiple comparison groups refer to that more than one group of people listed above
should be selected as control. It can reduce the bias caused by using only one kind of
control and enhance the reliability of the results. However, multiple comparison
groups undoubtedly increase the workload of the research.
68 L. Liu

4.2.2 Determine the Sample Size

4.2.2.1 Matters to Be Considered when Calculating Sample Size

1. In general, the sample size of the unexposed group should not be less than that of
the exposed group, usually the same amount. Small sample size may cause
increased standard deviation and unstable results.
2. Due to long-term follow-up of cohort studies, the loss of follow-up is inevitable.
An estimated rate of loss to follow-up in advance helps to prevent the analysis
from being affected by insufficient sample size in the later stage of the study.

4.2.2.2 Four Factors Affecting Sample Size

1. The incidence of the outcome of interest in the unexposed population p0.

2. The incidence of the outcome of interest in the exposed population p1. The greater
the difference between the two incidences of the exposed and unexposed
populations, the smaller the sample size requires. If the incidence of the exposed
group is not easy to obtain, one can try to get the estimate of the relative risk (RR)
and calculate p1 by the formula p1 =RR×p0.
3. Significance level α: That is the probability of the type I error when making a
hypothesis test. The smaller the probability of false positives, the greater the
sample size required. α is usually taken as 0.05 or 0.01.
4. Power (1 - β): β is the probability of the type II error in the hypothesis test.
Power of test refers to the ability to avoid false negatives when testing. The
smaller the β, the greater the sample size required. Typically, β is 0.10 and
sometimes 0.20.

4.2.2.3 Calculation of Sample Size

If the sample size for the exposed and unexposed groups is the same, the sample size
required for each group can be calculated using the following formula:
p p 2
Z α 2pq þ Z β p0 q0 þ p1 q1
n= ð4:1Þ
ð p1 - p0 Þ 2

p0 and p1in the formula represent the incidence of the unexposed and exposed
groups, respectively;p is the average of the two incidences; q = 1 - p; Zα and Zβ are
standard normal distribution limits, which can be found from the Standard Normal
Distribution Table.
4 Cohort Study 69

4.2.3 Follow-Up

The follow-up of participants is a very arduous and important work in a cohort study.
It should be planned and strictly implemented.

4.2.3.1 Purpose of Follow-Up

The purpose of follow-up includes three points: identifying whether a subject is still
under observation; identifying various outcomes (e.g., disease incidence) in the
study population; further collecting data on exposures and confounding factors.
For a variety of reasons, some participants are out of observation during follow-
up, a phenomenon known as loss to follow-up, which would have an impact on the
findings. When the loss rate is greater than 10%, measures should be taken to further
estimate its possible impact. If the loss rate is very high, the authenticity of the study
will be seriously questioned. Ensuring follow-up success is therefore one of the keys
to successful cohort studies.

4.2.3.2 Follow-up Methods

Follow-up methods include direct face-to-face interviews, telephone interviews,

self-administered questionnaires, periodic physical examinations, environmental
and disease monitoring, etc. The follow-up methods should be based on follow-up
contents, follow-up objects, and manpower, material, and financial resources. It
should be emphasized that the same follow-up method should be used for the
exposed and comparison groups, and the follow-up method should remain
unchanged throughout the follow-up.

4.2.3.3 Follow-up Contents

The contents of follow-up are generally consistent with the baseline data, but the
focus of follow-up is the outcome of interest. The specific items may be different
depending on the purpose and design of the study. In general, one should mainly
collect the following information: ① Study outcomes: whether the study population
has some kinds of research outcomes. Suspected patients found for the initial
examination should be further confirmed. ② Exposure data: what is the exposure
of the study subjects? Is there any change? For example, if the study aims to detect
the relationship between smoking and lung cancer, one should ask about the amount
of cigarette smoking at baseline and during the follow-up. ③ Other relevant
information of the study population: the same as the baseline items. ④ Changes in
population information: information on lost or retired population, or new arrivals
(dynamic cohorts).
70 L. Liu

4.2.3.4 Endpoint of Observation

The endpoint of observation means that the subjects develop the desired outcome.
For example, when the etiological factor of the disease is studied, often the outcome
is the occurrence of the disease or the death caused. When the study subjects develop
the outcome of interest, they are no longer observed. In general, the endpoint of the
observation is the disease or death, but may also be changes of certain indicators,
such as the emergence of serum antibodies and elevated blood lipids, according to
the study purpose.

4.2.3.5 Follow-Up Interval

In theory, follow-up should be carried out after the shortest induction period or
incubation period of the disease. The follow-up interval depends on the intensity of
exposure and the length of the incubation period of the disease. The weaker the
exposure or the longer the incubation period is, the longer the follow-up interval
needs. The induction or incubation period of chronic disease is not very clear. In
general, the follow-up interval of chronic diseases can be set for several.

4.2.3.6 The Termination Time of Observation

The termination time of observation refers to the deadline of entire research work,
and the expected time to get the result of interest. The termination time of observa-
tion is determined according to the length of the observation period, which depends
on the incubation period of the disease. In addition, one should also take into account
the amount of person-year. One should try to shorten the observation period on the
basis of these principles so as to save manpower and material resources and reduce
the number of loss to follow-up.

4.2.4 Quality Control

Cohort studies are by nature time-consuming and expensive. Therefore, the strict
implementation process, especially the quality control during data collection, is of
particular importance. Generally, the following quality control measures are taken:

4.2.4.1 Selection and Training of the Investigators

Investigators should maintain strict work ethic and scientific attitude. Honesty and
reliability are the basic qualities that investigators should possess. Generally,
4 Cohort Study 71

investigators should possess the expertise and knowledge required for the investi-
gation. The work ethic, scientific attitude, survey techniques of investigators, and the
experience of clinical doctors and laboratory technicians will affect the reliability
and authenticity of the survey. Therefore, before data collection, investigators should
be trained for better performance during the investigation.

4.2.4.2 Preparation of an Investigator’s Handbook

Due to the large number of investigators involved and the long duration of follow-up
in cohort studies, an Investigator’s Handbook, including operating procedures, pre-
cautions, and a complete description of the questionnaire is necessary.

4.2.4.3 Supervision during the Follow-Up

Common supervision measures include: repeating the survey among some partici-
pants by another investigator, checking numerical or logical errors, comparing the
distribution of variables collected by different investigators, analyzing temporal
trends of variables, and recording the interviews by using tape recorders, etc.

4.3 Data Collection and Analysis

4.3.1 Data Collection

The investigators should first collect the baseline information of every participant,
mainly including information on exposure status (e.g., the type, duration, and dose of
the exposure), personal characteristics (e.g., health status, age, gender, occupation,
educational level, marital status), and other circumstances (e.g., home environment,
lifestyle and family history of disease). Participants are followed over time, and
baseline information is compared with later follow-ups. It also works as a basis to
characterize baseline exposures (e.g., classify individuals into exposed or unexposed
group, ascertain degrees of exposure and potential confounders), and to obtain
tracking materials for follow-up and key information for inclusion or exclusion.
The major methods to collect baseline information include data records (e.g.,
employment, medical examinations, insurance), questionnaires or interviews, phys-
ical examinations and tests of biological samples, as well as environmental mea-
surements. Besides baseline information, data collection throughout the process of
follow-up is also important (e.g., changes of exposures and measurements of out-
comes over time). For more detailed information, please see the second section on
follow-up of this chapter.
72 L. Liu

4.3.2 Measures of Outcome Frequency

4.3.2.1 The Basic 2 × 2 Tables Summarizing the Results of a

Cohort Study

Disease incidence could be described by the cumulative incidence or incidence

density. Cumulative incidence is generated by dividing the number of incident
cases by the number of persons at risk in the cohort, as shown in Table 4.1:
The cumulative incidence in the exposed group = d1/n1.
The cumulative incidence in the unexposed group = d0/n0.
The cumulative incidence represents the individual risk of developing the disease
of interest with no unit. It is a proportion, not accounting for possible different
periods of follow-up time, thus mainly used in fixed cohorts. When studying acute
outcomes within a short period of follow-up, such as outbreaks, cumulative inci-
dence could be used to estimate the risk of the disease, given a fixed period of
follow-up. However, in most circumstances, such as chronic disease research, the
periods of follow-up are relatively long; thus, the cumulative incidence is no longer
appropriate since the follow-up time usually differ across cohort members. In this
situation, the outcome of interest is preferably described by rate, which is incidence
density, the other index to reflect disease incidence, and it is widely utilized in
dynamic cohorts. Incidence density is calculated by dividing the number of outcome
events by the person-time at risk, as shown in Table 4.2:
Incidence density in the exposed group = d1/T1.
Incidence density in the unexposed group = d0/T0.
One should note that a person “at risk” refers to the fact that the outcome of
interest can occur within the given time frame. Thus if subjects are immune, they are
no longer at risk of getting this disease. If on the other hand, the event of interest is
uterine cancer, a hysterectomized woman would not be “at risk.” Measurements of
risk and incidence of disease could provide valuable information related to the public
health burden of the outcome of interest, which is important for disease prevention
and public health management.

Table 4.1 Measures of cumulative incidence

Exposure status Cases Non-cases Total Cumulative incidence
Exposed d1 n1 - d1 n1 d1/n1
Unexposed d0 n0 - d0 n0 d0/n0

Table 4.2 Measures of incidence density

Exposure status Cases Person-time at risk Incidence density
Exposed d1 T1 d1/T1
Unexposed d0 T0 d0/T0
4 Cohort Study 73

4.3.2.2 Person-Time

In a dynamic cohort, study subjects have unequal periods of time from entry into the
cohort to disease occurrence or end of follow-up, and this must be taken into
account. Person-time is introduced to reflect the exposure experience of a subject
in this circumstance. Total person-time is the summation of the time at risk of
individual cohort members to develop the disease, which is often the denominator
of the incidence density. The common unit of person-time is person-year. As shown
in Table 4.3, people entered the cohort at different ages and experienced separate
lengths of time. Before the end of the follow-up, four subjects were diagnosed with
disease of interest. The person-years of each person are presented in the last column,
and the total person-time in this example is 91 person-years.
This exact computation method is based on the duration of participation of each
individual; however, for large cohorts, one may not obtain detailed information for
each participant, then approximation method is an alternative though with less
precision. The approximate person-years are considered as the average number of
the population multiplied by the number of years of observation. The average
number of the population refers to the average number of the population at the
beginning of two contiguous years or the number of the population in the middle of a
specific year. In a hypothetical cohort study which started on September 1, 2014, and
finished on September 1, 2017, the numbers of subjects were 15,262 in the begin-
ning, and 15,276 at the end, and more details are shown in Table 4.4. The average
population in the 20–29 age group are 26,203 persons: (8724 + 8736) /

Table 4.3 Data from a fictitious cohort

Person Age at Years of Age at end of Age at Person-years at
ID entry follow-up follow-up diagnosis risk
1 34 14 48 14
2 37 20 57 52 15
3 30 12 42 12
4 33 17 50 41 8
5 37 9 46 9
6 38 16 54 49 11
7 43 11 54 11
8 32 20 52 43 11
Total 120 91

Table 4.4 Numbers of subjects in a hypothetical cohort study at different times stratified by age
groups
Age groups 2014-09-01 2015-09-01 2016-09-01 2017-09-01
20–29 8724 8736 8740 8730
30–40 6538 6570 6554 6546
Total 15,262 15,306 15,294 15,276
74 L. Liu

Table 4.5 Data from a fictitious cohort study to calculate person-years with simple life table
No. of objects
At the Occurring Lost to
Observing beginning Entering the outcome events follow-up Pearson-
time (x) (Nx) cohort (Ex) (Dx) (Lx) years (Tx)
2011 1898 76 4 22 1923
2012 1948 70 6 18 1971
2013 1994 52 7 15 2009
2014 2024 30 5 19 2027
Total 7930

2 + (8736 + 8740)/2 + (8740 + 8730)/2 = 26,203. The average population is then

multiplied by the number of follow-up years to get the person-time.
Another method to calculate the person-time is to utilize simple life table. The
basic equations are as follows:

1
T x = Nx þ ðE x - Dx - Lx Þ ð4:2Þ
2

N xþ1 = N x þ E x - Dx - Lx ð4:3Þ

x refers to a certain period of time, usually representing 1 year; Tx is the person-

time during x time; Nx is the number of population at the beginning of x time; Ex is
the number of subjects entering the cohort during x time; Dx is the number of
occurring outcome events during x time; and Lx is the number of subjects who are
lost to follow-up.
According to the equations above, one can get a simple life table, and the total
person-years are the sum of every Tx.
For example, according to Table 4.5, the person-years in 2011 are

1
T 2011 = N 2011 þ ðE 2011 - D2011 - L2011 Þ
2
= 1898 þ ð76–4 –22Þ =2 = 1923

The number of population at the beginning of 2012 is

N 2012 = N 2011 þ E2011 - D2011 - L2011

= 1898 þ 76–4–22 = 1948

So the person-years in 2012 are

4 Cohort Study 75

T 2012 = 1948 þ ð70–6 – 18Þ =2 = 1971

By that analogy, person-years in 2011, 2012, 2013, and 2014 are 1923, 1971,
2009, and 2027, respectively, and the total person-years are 7930.

4.3.2.3 Standardized Mortality Ratio (SMR)

For cohorts with a general population comparison group, one usually estimates the
association between an exposure and an outcome by calculating standardized mor-
tality (or incidence) ratios (SMRs). The SMR is the ratio of the observed number of
deaths in the cohort and the expected number of deaths in the cohort, given the
age-specific mortality rates of a reference population and the age structure of the
cohort.

I
ni
i=1
SMR = I
ð4:4Þ
t i × ai
i=1

Where I stands for the age group, ni denotes the number of observed deaths of the
age group, ti denotes the number of person-years in the age group, and ai represents
the age-specific mortality rate of the age group from the reference population.
The SMR is commonly adjusted for age, calendar period, and other characteris-
tics like race. Example: There were 1000 workers aged between 40 and 50 in a
factory, and four of them died of lung cancer in 2000. Assuming that the mortality of
lung cancer among the total population aged between 40 and 50 is 2‰ in 2000, then
the expected number of death is 2, and we have known that the practical number of
deaths is 4; thus the SMR is 2 (4/2 = 2).

4.3.2.4 Statistical Tests

To test the statistical difference of incidence rate between the exposed and
unexposed groups, U test is commonly used in practice. However, there are some
noteworthy conditions to abide by relatively large sample size, not too small
p (incidence rate) and 1 - p (e.g., n × p and n × (1 - p) are both over five), and
approximately normal distribution of incidence rates.

p 1 - p0
u= ð4:5Þ
pc ð1 - pc Þ 1=n1 þ 1=n0
76 L. Liu

p1 and p0 are incidence rates in the exposed group and the unexposed group,
respectively; n1 and n0 are numbers of subjects in the exposed and unexposed
groups, respectively; and pc is incorporative sampling rate ( pc = Xn11 þX þn0 , X1 and X0
0

are the numbers of outcome events in the exposed and unexposed groups, respec-
tively). One should subsequently compare the U value with the standard U table,
then seek out the corresponding P value and make inference based on the significant
level.
Other statistical tests include probabilistic methods based on binomial or Poisson
distribution, Chi-Square test, or score test. Similarly, it is notable that each test has its
conditions.

4.3.3 Measures of Association

4.3.3.1 Relative Risk (RR)

RR refers to the ratio of the probabilities of an outcome in the exposed and

unexposed groups. Its value is a positive real number with a range from 0 to +1,
and could take the following form:

I1
RR = ð4:6Þ
I0

I1 and I0 refer to risk or rate of outcome in the exposed and unexposed groups,
respectively.
There are two alternative and equivalent expressions: the risk ratio and the rate
ratio.
Risk ratio is based on the cumulative incidence, with not accounting for person-
time. In Table 4.1, risk ratio could be expressed as:

d1=n
RR = d0=n
1
ð4:7Þ
0

Rate ratio is the most natural way to express relative risk. It uses incidence
density, which takes person-time into account. In Table 4.2, the rate ratio would
then be:

d 1=T
RR = d 0=T
1
ð4:8Þ
0

One can also estimate the 95% confidence interval (CI) of the RR using the Woolf
method based on the variance of RR. According to Table 4.1, the variance of ln RR
is computed as follows:
4 Cohort Study 77

Table 4.6 General criteria to Relative risk Strength of association

estimate the strength of asso-
1.0–1.1 0.9–1.0 None
ciation of relative risk
1.2–1.4 0.7–0.8 Weak
1.5–2.9 0.4–0.6 Moderate
3.0–9.9 0.1–0.3 Strong
10- <0.1 Infinite
Monson [6]

1 1 1 1
Var ðln RRÞ = þ þ þ ð4:9Þ
d0 d1 n0 - d0 n1 - d1

and

95%CI of ln RR = ln RR ± 1:96 Var ðln RRÞ ð4:10Þ

One could obtain the 95% CI of RR by taking the antinatural logarithm of 95% CI
of lnRR.
Risk ratio and rate ratio have the same epidemiological implication, but their
values are usually different in the same study. The interpretation of the relative risk is
as follows:
If RR > 1, the risk of disease for the exposure is increased compared with the
unexposed group;
If RR < 1, the risk of disease for the exposure is decreased compared with the
unexposed group;
If RR = 1, there is no association.
The risk in the reference group multiplied by the corresponding RR approximates
the risk in the exposed group. The value of RR reflects the level of association. Here
are the general criteria to estimate the correlation intensity (Table 4.6):

4.3.3.2 Attributable Risk (AR) and Attributable Fraction (AF)

The RR mainly measures the level of risk associated with the exposure to a risk
factor. It cannot reflect the impact of the factor in a population. To address this issue,
AR and AF are introduced. RR mainly provides clues for etiology, while AR and AF
are important for disease prevention and public health. AR, also known as the risk
difference or excess risk, is the measure of the rate of disease related to the exposure
to a risk factor. Attributable risk is applied to quantify risk in the exposed group
which could be attributable to the exposure.
78 L. Liu

d1 d
AR = I 1 - I 0 = - 0 ð4:11Þ
n1 n0

or

AR = I 1 - I 0 = RR × I 0 - I 0 = I 0 ðRR - 1Þ ð4:12Þ

AF is the proportion of the total number of cases related to the exposure to a risk
factor. It allows to calculate the proportion of disease attributable to the exposure in
the exposed group. This can also be viewed as the proportion of disease in the
exposed group that can be avoided through the elimination of the risk factor. It is
calculated by dividing the risk difference by the incidence of disease in the exposed
group and then multiplying it by 100 to convert it into a percentage

I1 - I0
AF = × 100% ð4:13Þ
I1

or

RR - 1
AF = × 100% ð4:14Þ
RR

AR and AF are both calculated from incidence rates. One should note that they
only make sense for a causal association of a risk factor with an outcome occurrence.
The underlying assumption is that no other potential confounders are involved in the
occurrence of the outcome.

4.3.3.3 Population Attributable Risk (PAR) and Population

Attributable Fraction (PAF)

PAR estimates the proportion of disease attributed to the exposure in the study
population. PAR can be looked at as the proportion of a disease that could be
prevented by eliminating a causal risk factor from the population. PAR tends to be
a function of time because both the prevalence of a risk factor and its effect on the
exposed population may change over time, as may the underlying risk of disease.
Definitions for PAR and PAF are given by

PAR = I t - I 0 ð4:15Þ

It - I0
PAF = × 100% ð4:16Þ
It
4 Cohort Study 79

Where It represents the incidence of disease in the total population, and I0

indicates the incidence of disease in the absence of exposure.
PAF is also given as:

Pe ðRR - 1Þ
PAF = × 100% ð4:17Þ
Pe ðRR - 1Þ þ 1

Where the prevalence of exposure “Pe” is the proportion of individuals exposed

to the risk factor.

4.3.3.4 Dose-Effect Relationship

In some circumstances, there may exist a dose-effect relationship between the

exposure and the outcome. To address this, one could stratify the exposure into
several levels, with defining the lowest level as a reference, and then calculate RRs of
other groups compared to the referent group. Taking Table 4.7 as an example, along
with the increase of serum cholesterol level, the relative risk of developing coronary
heart disease also increases, which indicates that there may exist a dose-effect
relationship between serum cholesterol levels and incidence of coronary heart
disease. If necessary, one can further make a trend test.

4.4 Common Bias and Controlling

4.4.1 Selection Bias

Selection bias occurs when the selection of the exposed and unexposed individuals is
related to the occurrence of the outcomes of interest. This is a major potential
problem in retrospective cohort studies, since knowledge about the exposure and
outcome is likely to differentially influence participants. However, it is generally not
a problem in prospective cohort studies, since the outcome of interest has not
occurred. A serious potential concern is loss to follow-up in prospective cohort
studies [7], which arises when study subjects refuse to participate in or cannot be

Table 4.7 The occurrence of coronary heart disease stratified by serum cholesterol levels in a
fictitious cohort study
Cholesterol level No. of participants No. of cases Risk Relative risk
Very low 200 2 0.01 1(reference)
Low 300 15 0.05 5
Intermediate 400 40 0.1 10
High 300 60 0.2 20
Very high 100 30 0.3 30
80 L. Liu

found for the data collection during follow-up. Retention of subjects might be
differentially related to both exposure and outcome, and this brings a similar effect
that can prejudice the results, causing either an underestimate or an overestimate of
an association. For example, if an exposed individual will develop the outcome in
the future, but she/he is more likely to be lost to follow-up, then the exposed
incidence will be underestimated, along with the RR tending towards the null.
Loss to follow-up can result in bias and reduce the statistical power. The primary
way to reduce this bias is to improve compliance and response rate of participants.

4.4.2 Information Bias

Similar to selection bias, information bias occurs in different ways under different
study designs. Reporting bias is one of the potential information biases in cohort
studies since the exposure status may influence the reporting of the outcome. For
example, in an investigation about occupational hazard, workers are more likely to
report having experienced various harmful exposures when this refers to labor
guarantee or benefits; thus, some associations may be overestimated. If possible, it
would be better to utilize some objective methods and sources of data, such as
medical records and laboratory tests, to ascertain the exposure and outcome status.
Another important form of information bias is detection bias. Detection bias occurs
when knowledge of exposure status differentially increases the likelihood of
detecting the outcome of interest among the exposed in cohort studies. A typical
example is that a medically relevant exposure could bring about more medical visits
and an increased possibility of a diagnostic evaluation, which increases the proba-
bility of detecting the outcome in the exposed group. An effective way to address
this issue is to apply blinding method to collect information.
Besides, other factors may also contribute to information bias. For example, in the
collection of laboratory data, the quality of instruments and reagents, selected
measurement standard, measuring conditions and technical competence of the oper-
ator are all potential factors influencing the results. Additionally, scientific question-
naires and complete records are also imperative.

4.4.3 Confounding

Except for selection bias and information bias, confounding is also an important
factor that can cause systematic bias in epidemiology, thus the investigators must
consider it from study design to data analysis. Confounding distorts the underlying
correlation of the exposure with the outcome of interest. The factors causing
confounding are called confounders. The criteria for a factor to become a confounder
are as follows: the factor must be related with both the exposure and the disease of
interest, and at the same time it must not be an intermediate variable in the causal
4 Cohort Study 81

chain between the exposure and the disease of interest. Directed Acyclic Graph
(DAG) is an effective method to distinguish a confounder and a collider. In the
following example:

BMI ė Diabetes ė Mortality

̨Ė̩
Smoking

Smoking is a confounder when exploring the association between BMI and the
prevalence of diabetes, or the association between the prevalence of diabetes and
mortality. However, when exploring the association between smoking and BMI,
diabetes acts as a collider (a variable directly affected by two or more other variables
with arrows pointing to itself in the DAG, but not the other way around).
In cohort studies, confounding occurs when risk factors are unevenly distributed
between the exposed group and the unexposed group. The major methods to control
confounding are restriction on inclusion criteria, randomization, and matching.
Besides, statistical procedures such as standardization, stratification analysis, and
multivariate analysis are also available.

4.5 Advantages and Disadvantages of Cohort Studies

4.5.1 Advantages of Cohort Studies

1. Strong ability to identify cause-effect association because of the temporal rela-

tionship between the exposure and the outcome, reliable data personally observed
by researchers and computable indicators reflecting relevance intensity such as
RR, AR, etc.
2. Helpful in understanding the natural history of disease in the population.
3. Unexpected outcome data are obtained to analyze the relationship between
multiple outcomes and a cause.
4. Able to study the effects of rare exposures.
5. Avoiding recall bias at enrollment.

4.5.2 Disadvantages of Cohort Studies

1. It is not suitable for disease with low morbidity because large sample size is
needed.
2. In a long follow-up period, lost to follow-up of subjects would cause bias.
3. A large amount of manpower, material resources, and financial resources are
required.
82 L. Liu

4. During the follow-up, the entry of unknown variables and the changes of known
variables could influence the outcome, making the analysis complicated.

4.6 Example of a Cohort Study

To facilitate the understanding of cohort studies, the design, implementation and

main results of a cohort study “Fresh Fruit Consumption and Major Cardiovas-
cular Disease in China [8]” is cited. This study is from The China Kadoorie
Biobank Study a nationwide, prospective cohort study involving 10 diverse locali-
ties (regions) in China. For more details, please see Du H, Li L, Bennett D, Guo Y,
et al. Fresh Fruit Consumption and Major Cardiovascular Disease in China [J]. N
Engl J Med. 2016;374(14):1332-1343.
Chapter 5
Case-Control Studies

Qian Wu

Key Points
• The case-control study population consisted of a case group selected from those
with the disease of interest and a control group selected from those who did not
have the disease.
• Case-control studies belong to observational studies. It set up a control group.
• In case-control studies, Odds Ratio was used to estimate the strength of the
association between disease and exposure factors.
• Selection bias, information bias, and confounding bias are major sources of bias
in case-control studies.

5.1 Overview of Case-Control Studies

The purpose of the case-control study is to evaluate the relationship between the
disease and the exposure factors suspected of causing the disease. Both cohort and
case-control studies are analytical studies, their main difference lies in the selection
of the study population. In a cohort study, the subjects do not have the disease when
entering the study and are classified according to their exposure to putative risk
factors, in contrast, subjects in case-control studies are grouped according to the
presence or absence of the disease of interest. Case-control studies are relatively easy
to conduct and are increasingly being applied to explore the causes of disease,
especially rare diseases. Case-control studies are used to estimate the relative risk
of disease caused by a specific factor. When the disease is rare, case control study
may be the only research method.

Q. Wu (✉)
School of Public Health, Xi’an Jiaotong University, Xi’an, China
e-mail: [email protected]

© Zhengzhou University Press 2023 83

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_5
84 Q. Wu

5.1.1 History

Case control study has a long history. In 1843, Guy compared male occupations with
lung diseases with those with other diseases. But it was not until 1926 that Janet
Lane Claypon first proposed a case-control study in a breast cancer research. Richard
Doll’s research on smoking and lung cancer in the 1950s gave a great impetus to the
applications of case-control study. Since then, case-control studies have become
more prominent in biomedical literature, and their design, implementation, and
analysis have become more standardized in methodology.

5.1.2 Definition

A case-control study involves cases from those individuals with disease of interest
and controls from those who are without the disease. Previous exposure histories of
case and control subjects were examined to evaluate the relationship between
exposure and diseases. If the exposure history of the case group and the control
group is different, it is possible to infer that the exposure may be related to the
disease. The difference in exposure between the case and control group helps to
identify potential risk factors. The purpose is to explore whether there are factors
related to the disease. The basic principle of a case-control study is shown in Fig. 5.1.
A case-control study is called a retrospective study because researchers need to
investigate the exposure factors of the subjects before the occurrence of the disease.
Sometimes retrospective studies are used to represent case-control studies. It may be
confusing because the terms retrospective and perspicacity are also used to describe
the time of data collection related to the current date. In this sense, case-control

Time

Exposed
Cases
(People with disease)

Unexposed Source
population

Exposed
Controls
(People without disease)
Unexposed

Direction of inquiry

Fig. 5.1 Design of case-control study

5 Case-Control Studies 85

studies can be retrospective, when all data are related to the past; it can also be
forward-looking, in which data collection continues over time. Therefore, retrospec-
tive study is not the essential characteristics of case-control study. The essence of a
case-control study is to divide the subjects into case and control groups according to
the presence or absence of the disease of interest.
Example Some researchers surveyed the relationship between plasma metal con-
centration and the incidence rate of coronary heart disease (CHD) [Yu Yuan, Yang
Xiao, Wei Feng, et al. Plasma Metal Concentrations and Incident Coronary Heart
Disease in Chinese Adults: The Dongfeng-Tongji Cohort. Environ Health Perspect.
2017,125(10): 107007.]. The researchers compared 1621 CHD cases with 1621
controls free of CHD in Shiyan City, Hubei Province, China, in 2013. All of the
participants were retired. Concentrations of aluminum, arsenic and barium, were
significantly higher in cases (57.41, 2.32, 40.53 μg/L) than controls (48.95, 1.96,
35.47 μg/L). The study presented the concentrations of aluminum; arsenic and
barium were higher in the cases than in the controls, indicating that circulating
metals were associated with an increased incidence of CHD.
For example, information of participants’ disease and their plasma metal was
extracted from previous studies. In 2013, investigators according to the interest
disease divided retirement employees into two groups. The case group is retirement
employees with CHD, while control group is free of major cardiovascular disease.
The researchers explored metal concentrations in plasma of participants from 2008
to 2013.
Firstly, the case-control study recruited patients according to their current disease
status. Exposure history was inquired for in each case and control. Data were mostly
collected after disease occurred, thus case-control study was considered retrospec-
tive, which was a limitation. Compared with cohort design, case-control study
design has weak support for causal hypothesis. However, it provides more powerful
evidence than cross-sectional studies in analyzing and interpreting the results. Case-
control study is one of the commonly used research designs. The reason is that the
implementation of case-control study is relatively simple and convenient compared
with other study designs.

5.1.3 Type of Design Case-Control Studies

There are three kinds of case-control studies. First is the traditional case-control
design. In this type, cases and controls are recruited from population. The case group
is assumed to include all cases that occurred in that hypothetical cohort up to the time
when the study is conducted. Control group is selected from those without the
disease of interest throughout the study period. There are three subgroups of
traditional case-control, which are unmatched, frequency matching, and individual
matching case-control studies. Next is the nest case-control design which is
conducted in a cohort population. At the beginning of nest case-control study (t0),
86 Q. Wu

members of the cohort are collected exposure factors. Cases and controls are
identified subsequently at time t1. The control group is selected from the cohort
members who do not meet the case definition at t1. Third is a case-cohort design. in
the first step, a population was identified as the cohort for the study, and a sample
within that cohort was selected as the control group using a randomized method. In
the whole cohort, all cases of the disease to be studied were collected as a case group.
Finally, the two groups were compared and analyzed to explore the factors affecting
disease onset, disease survival time, and prognosis.

5.1.4 Characteristics of Case-Control Study

Case-control studies belong to observational study. Case-control study draws infer-

ences from a sample to a population where the independent variable is not under the
intervention of the investigator because of ethical concerns or logistical constraints.
Case-control study set up a control group. The differences of exposure were
compared between case and control group.
Case-control studies are a special type of retrospective study. Investigators look
back in time and access prior exposure status between two groups.
The relative risk (RR) cannot be calculated directly in case-control studies, and
the Odds Ratio (OR) can be used to estimate the RR.

5.1.5 Application

Case-control studies are suitable for investigating rare diseases or diseases with a
long latency period, as subjects are selected from the outset based on their outcome
status. Therefore, compared to cohort studies, case-control studies are faster and
relatively less expensive to implement, require relatively fewer subjects, and allow
for multiple exposures or risk factors to be assessed for a single outcome.

5.1.5.1 Example of a Case-Control Study

In October 1989, physicians in the United States reported three patients with a newly
recognized disease characterized by marked peripheral eosinophilia with features of
scleroderma. After reporting this obvious association, more cases were found in the
United States and Europe. To illustrate a possible link between EMS and the
tryptophan manufacturing process, they conducted case-control studies to assess
potential risk factors, including the use of tryptophan from different manufacturers.
In early November 1989, they carried out a case-control study that demonstrated an
epidemiologic association between the consumption of tryptophan products and the
eosinophilia-myalgia syndrome (EMS). The case-control studies were used to
5 Case-Control Studies 87

evaluate potential risk factors, including the use of tryptophan from different man-
ufacturers. The investigators analyzed the tryptophan samples using high-
performance liquid chromatography to determine the other chemical component.
The results found that 29 of 30 case patients (97%) and 21 of 35 controls (60%) of
the subjects using tryptophan had consumed tryptophan produced by one company.
The EMS outbreak in 1989 was due to the ingestion of a chemical ingredient that
was associated with a specific tryptophan manufacturing condition in one company.
This study suggests several important characteristics of case-control studies.
Firstly, the design provides a suitable research method for studying this rare disease
of EMS. Case-control study is applicable to the etiology of rare diseases. Secondly,
case-control studies allow researchers to investigate several risk factors at the same
time. In this research, researchers explored the effect of tryptophan and other factors
on EMS. Finally, a case-control study usually does not “prove” causality, but it can
suggest a hypothesis. The researchers believe that more research is necessary to
identify the composition of the chemicals that trigger EMS and to clarify the
pathogenesis of the syndromes. Follow-up revealed that the removal of
tryptophan-containing products from the market resulted in the near elimination of
reported cases of EMS.

5.2 Design of Case-Control Studies

Case-control study is the most commonly used method of analytical epidemiology.

In its implementation, the selection of research objects is crucial. Especially the
selection of control group is difficult to master. It is usually required that the control
should represent the source population that generated the case.
The case-control study determines whether the subjects are case group or control
group according to the status at the beginning of the investigation. This status is
considered as the outcome variable of the study. The outcome may be whether the
subject has been diagnosed with a certain disease or has experienced a complication.
Once outcome status is identified and subjects are categorized as cases or controls.
Then, information on exposure to one or several risk factors is then collected
retrospectively, usually through interviews or surveys.

5.2.1 Basic Principles

There are three principles of case-control study design. First, it is the study popula-
tion, also called a source population. The source population may produce the cases
and controls. The selection of the control group should not be influenced by
exposure factors. Overall, the key issue is for the control group to be representative
of the population that generated the cases. The second is de-confounding principle.
De-confounding address issues that arise when the exposure of concern is associated
88 Q. Wu

with other possible risk factors. Confounding factors can be eliminated by getting rid
of the variability of that factor. For example, if gender is a possible confounding
factor, selecting only males would eliminate gender variability altogether. Finally,
the principle of comparability was introduced in the two investigation processes. The
precision of the exposure measurements was consistent between the control group
and the case group. For example, in studies on the effects of smoking on lung cancer,
researchers have used nicotine levels in urine to measure smoking in the case group,
while questionnaires to measure the controls group, which is inappropriate. Bias due
to different measurement methods between cases and controls should be eliminated.
The selection of controls and cases was determined based on the presence or
absence of interested disease and could not be influenced by exposure status. Cases
and controls do not have to be representative of everyone; in fact, they can be
restricted to any specific subgroup, such as elderly, male, or female.

5.2.2 Selection of Cases

Case groups for case-control studies should be representative of all cases in a

population. Case selection is based on interested disease and does not have to
consider exposure. Cases were available at the beginning of the study. Cases may
include new cases, existing cases, and deaths.
New cases are preferred when selecting cases to avoid the influence of survival
factors related to the etiology of the disease. Cases found in one clinic or treated by a
physician are alternative cases for case-control studies. The source population of
cases treated at a clinic is all those who may be seen at that clinic. Reviewing
previous studies, many case-control studies were conducted using one or a small
group of hospitals or clinics. This will help to obtain cases in a timely manner and
increase the possibility of cooperation, thus limiting selection bias. At the same time,
however, there may be problems in the definition of the population from who the
case originated.
Community-based population disease registries, particularly for cancer and birth
defects, are generally considered to be the best source of cases. This is because the
population at risk may be clearly defined by geographic or administrative
boundaries.

5.2.3 Selection of Controls

The most difficult task in case-control studies is the selection of the control group.
The control group should be selected from the population that generated the cases
with interested disease. Controls are personals without the disease. A key and
difficult aspect of population-based case-control studies is to identify a control
group in a more efficient way. Otherwise, it would be necessary to demonstrate
5 Case-Control Studies 89

that the population providing the control group had the same exposure distribution as
the population that was the source of the cases, a very stringent requirement that can
rarely be demonstrated. The control group should be selected independent of their
exposure status. There are four types of controls in case-control studies.

5.2.3.1 Population Controls

The best control group ensure that controls are random sample of all noncases in the
same population that produced the cases. Another way to ensure that cases and
controls are comparable is to draw from the same cohort which is called a nested
case-control study. The approach, relative to simply analyzing the data as a cohort
study, is that analyses are more efficient.
A control group is selected from the same institution or community. Neighbors or
friends were controls, and if these individuals showed results of interest, they would
be classified as cases. Selecting a control from a neighbor or friend of the case is also
a more feasible method. All households in the area surrounding the case were
censored and approached in random order until a suitable control was found. It is
important to note that the control was present while the case was being diagnosed.
The same difficulty is faced with the use of friend control, i.e., random selection from
the census of friends provided in each case. The main advantage of friend control is
the low level of non-response.

5.2.3.2 Hospital or Disease Registry Controls

The method of selecting controls from hospitals or clinics is more feasible, but it is
hardly representative of the source population. For example, a case-control study
investigates the relationship between depression and social and economic factors. A
particular clinic may be known to have the best depression specialists in a particular
area. If both cases and controls are selected from that clinic, then the depression
cases may represent the entire region, while the controls represent only the local
neighborhood. Cases and controls may then have different social and economic
characteristics. Therefore, cases and controls should be selected from multiple
diagnosis and treatment institutions to improve their representativeness.
Controls from a medical practice may be more appropriate than controls from
hospitals in an urban health center study. The control may have the same high
response level as the case. In the medical practice, they may be interviewed in the
hospital, which is an advantage from the perspective of the principle of comparable
accuracy. The likelihood of patients going to different hospitals varies. If a patient
has the disease being studied, the likelihood of going to a specific hospital will be
different from the likelihood of going to that hospital for patients with other diseases.
In addition, the exposure may be related to the diseases of some controls. Hospital-
based case-control studies generally believe that the disease of the control has not
associated with exposure. It is hoped that controls for these diseases will effectively
90 Q. Wu

form the basis of the study in a randomized sample. Because there is little certainty
about the independence of exposure and disease diagnosis, the standard recommen-
dation is to select controls with multiple diagnoses to ensure that failure of any of
them to meet the criteria will not affect the study. If a diagnosis is found to be related
to exposure, these controls can be excluded.

5.2.4 Matching

In case-control studies, matching is a common method to control confounding

factors. Matching means that the control group is similar to the case group in
some characteristics (such as age and sex).
The goal of matching is to control confounders and increase the efficiency of
study. If the factors used for matching are related to exposure, the matched control
sample usually has a more case-like exposure distribution than the unmatched
control sample. Matching eliminates differences in the distribution of certain
confounding factors between cases and controls, thus improving the efficiency of
the study. In this way, studies can achieve a strong statistical power with a smaller
sample size.
Matching begins with the identification of the case group. The investigator then
selects a control group from the source population. Matching is divided into two
types, depending on whether it is performed at the individual or group level.

5.2.4.1 Matching Type

Matching can be performed on a group of subjects, which is called group matching,

or on a subject-by-subject basis, which is called individual matching.

Group Matching

Group matching means that the matching factors are in the same proportion in the
case and control groups and is also referred to as frequency matching. For example,
the percentage of women in the case group was 45%, so we chose the control group
with 45% women as well. Keeping the control group and case group have the same
characteristics (e.g., proportion of male participants). Such that, a group of controls
is matched to a group of cases on a particular characteristic (e.g., gender).

Individual Matching

Investigators select a specific control for each case by matching variables. For
instance, if the first case enrolled in a study is a 40-year-old black woman, we will
5 Case-Control Studies 91

seek a 40-year-old back female control. Each case can be matched with more than
one control group. However, the ratio of controls to cases rarely exceeds 4:1, as the
higher the ratio the increasing difficulty of implementation.

5.2.4.2 Overmatching

If more variables are matched, it may be difficult to find appropriate controls. And
we were unable to explore possible associations of the disease with any of the
variables already matched in the cases and controls. In this way, overmatching
may happen.
An overmatch is a match that causes a loss of information in the study. There are
two types of overmatching. The first type is a match that impairs statistical effi-
ciency, such as a variable related to exposure but not to disease being matched. The
second type is a match that impairs validity, such as an intermediate variable
between exposure and disease being matched. If the investigator happens to match
on a factor that is itself related to the exposure, overmatching will appear. For
example, in a particular study of NSAIDs and renal failure, if arthritis symptoms
were matched in cases and controls, and arthritis symptoms were usually treated with
NSAIDs. Matching for arthritis may then affect NSAIDs. This overmatching can
decrease the association between exposure and disease.

5.2.5 Exposure

An important element of case-control studies is to determine the difference in past

exposure to a factor between cases and controls. The validity of case-control studies
also depends on measuring exposure. In the case-control design, the exposure status
of the case is usually investigated after the occurrence of the disease, usually by
asking the patient or relatives or friends. The purpose of measuring exposure is to
assess the extent of the subject’s exposure over a period of time prior to the onset of
the disease. The method of collecting exposure data should be the same for cases and
controls.
Most case-control studies use questionnaires or interviews to determine the
exposure of subjects. The validity of this information will depend in part on the
attitude of the subject. People are able to remember well some constant information,
such as where they lived in the past and what they did for a living. However, the
long-term memory of subjects for specific dietary information may be less reliable.
Exposure is sometimes measured by biochemical tests (e.g., calcium in the blood)
and may not accurately reflect relevant past exposures if not designed in advance.
For instance, lead in the blood of children at age 6 years is not a good indicator of
exposure at age 1–2 years. This problem can be avoided if exposure is estimated
from established record systems (routine blood tests or stored results from
92 Q. Wu

employment records) or if information is collected prospectively for case-control

studies so that exposure data can be collected before disease occurs.
Exposure information can sometimes be determined from historical records. For
example, a case-control study on the relationship between sinusitis and multiple
sclerosis determined their contact history by searching the general practitioner
records of patients and control groups. As long as the records are reasonable and
complete, this method is usually more accurate than the method relying on memory.

5.2.6 Sample Size

The sample size was calculated to ensure confidence in the findings and conclusions
of the study. Every researcher wants to complete a meaningful scientific study. The
estimation of the sample size is a necessary consideration in the study design. Should
an applicant receive funding from a funding agency if a sufficient number of subjects
are not enrolled in the study, resulting in no chance of finding a statistically
significant difference? Most funding agencies are concerned about sample size and
power in the studies they support and do not fund studies that would waste limited
resources.
There is also a problem with too large sample size. If the number of samples
recruited exceeds the required amount, the duration of the study will be extended.
Excessive sample size will also affect the quality of the investigation work and
increase the burden and cost of research.
Recognize that sample size is essential to ensure scientifically meaningful results
and proper management of financial, organizational, material, and human resources.
Let’s review how to determine statistical capacity and sampling size. Statistical
power is calculated with regard to a particular set of hypotheses.
Statistical power is calculated based on a set of assumptions. Epidemiological
hypothesis usually compares the observed proportion or ratio with the assumed
value. Statistical power refers to the probability that the null hypothesis will be
rejected if the specific alternative hypothesis is true. ß denotes the Type II error, i.e.,
the probability of not rejecting the null hypothesis when the alternative is true. A
study should be at least 80% power, and typically studies are designed to have
90–95% power to detect an outcome. What factors affect the power of a study? There
are α, β, effect size, variability, and n.
α is the probability of type I error, also known as the significance level of the test
hypothesis. This is often determined to be 5% or 1%, implying that the researcher is
willing to accept the risk of making a mistake in the alternative hypothesis.
Statistical power is related to effect size, sample size, and significance level. All
other factors being equal, an increase in effect size, sample size, or significance level
will yield more statistical power.
The sample size of case-control study is calculated according to Formula 5.1.
5 Case-Control Studies 93

2
Zα ð1 þ mÞp0 ð1 - p0 Þ þ Zβ p1 ð1 - p1 Þ þ mp0 ð1 - p0 Þ
n= ð5:1Þ
ð p1 - p0 Þ 2

p1 þ p0 =m p0 OR
p0 = p1 =
1 þ 1=m 1 þ p0 ðOR - 1Þ

Here n is that needed individuals in each group, α = alpha, β = 1 - power. OR is

the odds ratio which is the ratio of the exposure ratio between cases and controls.
“m” is ratio of the sample size of the control group to the sample size of the case
group. “p1”—probability of exposure in case, p0 can be estimated as prevalence of
exposure in the control group.
The formula gives the minimum number of cases needed to detect true odds ratio
or case exposure with power and bilateral type I error probability α.
Calculation of sample size for individual matched case-control studies.
The estimated case sample size for paired matched case-control studies was
calculated according to Eq. 5.2, and the control sample size was r × n.

2
n=½Z 1-α=2 ð1 þ 1=rÞ
pð1-
pÞ þ Z β p1 ð1-p1 Þ=r þ p0 ð1-p0 Þ =ðp1 -p0 Þ2 ð5:2Þ
p1 = ðOR × P0 Þ=ð1 - P0 þ OR × P0 Þ
P = ðP1 þ rP0 Þ=ð1 þ r Þ

Where α = alpha, β = 1 – power, P1, P0 denote the estimated exposure rates of

the case and control groups in the target population, respectively.

5.3 Data Collection and Analysis

When researchers have determined the outcomes (disease or health status) of interest
in the case-control study and the factors to be studied, they can develop methods for
collecting information. The data should include information about research out-
comes and factors. Data analysis involves two parts job. First is descriptive data.
Next is statistical inference and measure of association. The odds ratio represents an
indicator of the association between the disease and each factor of interest.
Researchers often consider data analysis to be the most enjoyable part of epide-
miological research. Because after all the hard work and waiting, they have a chance
to gain answers. The basic method of analysis in case-control studies is to compare
the proportion of exposure in the case and control groups and to calculate the OR.
94 Q. Wu

5.3.1 Main Analysis Objectives

Assess and refine data quality. Describe the study population and its relationship to
the target population. Assess potential bias. Estimate the frequency of exposure.
Estimate the strength of the association between exposure factors and disease.
A quality data analysis consists of three phases. In the first stage, the analyst
should review the recorded data for accuracy and completeness. Next, the analyst
should summarize the data in a concise form and perform descriptive analyses, such
as classifying observations according to key factors, using a contingency table.
Finally, the summarized data are used to estimate epidemiologic measures of
interest, usually expressed in terms of strength of association with appropriate
confidence intervals.

5.3.2 Descriptive Analysis

The number of study subjects and the composition of the various characteristics are
described. The exploration of the data reports the frequencies. These measures will
provide the basis for important subgroups. Standardization or other adjustment
procedures may be required to account for differences in age and other risk factor
distributions, duration of follow-up, etc. Compare whether certain basic character-
istics are similar between case and control groups.

5.3.3 Statistical Inference

The indicator that indicates the strength of the association between disease and
exposure in case-control studies is the odds ratio (OR). Data analysis included
calculating odds ratios as a measure of the association between the disease and the
interested factors. When analyzing data on the relationship between exposure and
disease variables, we usually have to make statistical inferences about relationship.
Several means were employed to avoid random errors, such as p-value and confi-
dence interval (CI) tests. But we should understand that the role of statistically
significant is limited. Statistical significance is usually based on the P-value:
depending on whether the P-value is less than or greater than the critical value,
usually 0.05. The critical value is then referred to as the alpha level of the test, and
the result is considered “significant” or “insignificant.”
The type of analysis used in case-control studies depends on whether controls are
sampled in an unmatched or matched manner. Different analysis methods are used
for different matching methods.
5 Case-Control Studies 95

5.3.3.1 Unmatched (Frequency Matching) Design

In case-control studies, researchers attempt to assess the strength of the association

between disease and study factors. The investigators analyzed the proportion of
exposure in the case and control groups. Data from unmatching or frequency
matching case-control studies are summarized in Table 5.1. For better understand-
ing, only two levels of exposure are discussed here. Each object can be divided into
four basic cells, which are defined by disease and prior exposure status.
A simple unmatched case-control study, such as that in Table 5.1, can be analyzed
by using OR (odds ratio) for association. In case-control studies, groupings are made
according to the presence or absence of disease. Therefore, we can’t measure health
outcomes or disease incidence rate. The proportion of persons in the study who have
the disease is no longer determined by risk of developing the disease, but rather by
the choice of investigator. So, investigators could not calculate RR (relative risk).
Investigators can obtain valid estimates of risk ratios by using OR. When the disease
interested is a rare disease, the odds ratio approximates the risk ratio or
RR. However, this is not always the case, researcher should be careful taken to
interpret the odds ratio appropriately.
χ 2 test and statistical inference (formula 5.3)

2
jad - bcj - N2 N
χ2 = ð5:3Þ
n 1 n0 m 1 m 0

Odds Ratio

The odds ratio (OR) is an index of the association between exposure and disease or
outcome. The odds ratio is the ratio of exposure in the case group divided by the ratio
of exposure in the control group. With the notation in Table 5.1, the odds of exposure
for case represent the probability that a case was exposed divided by the probability
that a case was not exposed. The odds are estimated by the following formula.

Exposed cases Unexposed cases a b a

Odds of case exposure = = = = =
All cases All cases aþb aþb b

Similarly, the odds of exposure among controls are estimated by the following
formula:

Table 5.1 The result of case- Case Control Total

control study
Exposed a b a + b (m1)
Unexposed c d c + d (m0)
Total a + c (n1) b + d (n0) a + b + c + d (n)
96 Q. Wu

c
Odds of control exposure =
d

The odds of exposure for cases divided by the odds of exposure for the controls
are expressed as the OR. Substituting from the preceding equations, the OR is
estimated by formula 5.4

odds of case exposure a c a×d

OR = = = = ð5:4Þ
odds of control exposure b d c×b

OR indicated “How many times more exposed are cases than no-case exposed?”
Since OR have a different scale of measurement than RR, the answer to this question
can sometimes differ from the answer to the corresponding question about
RR. However, case-control studies are concerned with rare diseases, for which RR
and OR are very similar.

Interpreting the Odds Ratio

A case-control study comparing the smoking habits of 58 lung cancer cases with
93 controls showed the following results (Table 5.2).

a × d 22 × 86
OR = = = 7:5
b×c 7 × 36

The proportion of lung cancer cases exposed to smoking was 7.5 times greater
than the proportion of controls who smoked. It is suggested that there is a strong
association between lung cancer and smoking. Smoking could thus be a factor that
increases the probability of having lung cancer.
As can be seen, we can determine the risk factors by calculating the OR. It is
important to recognize that case-control studies are comparing the odds of exposure
[(a/c)/(b/d)] between cases and controls. Conceptually, this is very different from
comparing the odds of illness [(a/b)/(c/d)] between exposed and unexposed indi-
viduals, which is the result we are really interested in.
Fortunately, in rare disease studies, the ratio [(a/c)/(b/d )] of the ratio of cases and
controls with exposure is equal to ad/bc. It can also be seen that the odds ratio [(a/b)/
(c/d)] in favor of disease in exposed and unexposed populations is also equal to ad/
bc.

Table 5.2 Results of a case-control study of lung cancer and smoking

Individuals with lung cancer (cases) Individuals without lung cancer (controls)
Smokers 22 (a) 7 (b)
Nonsmokers 36 (c) 86 (d )
Total 58 93
5 Case-Control Studies 97

Table 5.3 Study on the association between obesity and eating vegetables
Obese individuals (Cases) Non-obese individuals (controls)
Eat vegetables 121 171
Do not eat vegetables 129 79
Total 250 250

Table 5.4 Results of a study on depression and eating vegetables

Individuals with depression Individuals without depression
(cases) (controls)
Eat vegetables 80 80
Do not eat 120 120
vegetables
Total 200 200

Sometimes, the factors studied would reduce the probability of developing the
disease. Such factors are known as protective factors of the disease. For instance,
250 obese individuals (cases) in a case-control study were compared to
250 non-obese individuals (controls) in terms of vegetable consumption in their
diet. The results are shown below (Table 5.3).

a×d 121 × 79
OR = = = 0:43
b × c 129 × 171

The proportion of cases eating vegetables was 0.43 times greater than the
proportion exposed in the control group. Therefore, the proportion of eating vege-
tables in the case group was 48% lower than the exposure proportion in the control
group was 68%. The results of the case-control study showed that compared with the
control group, the case group were less likely to eat vegetables. Eating vegetables
may be a protective factor in reducing obesity.
Sometimes case-control studies did not find an association between study factors
and outcomes. In this case, the OR for the strength of the association between factors
and disease in the case-control study was 1.0. For example, in a case-control study,
200 people with depression were compared with 200 people without depression
regarding their vegetable consumption (Table 5.4).

a × d 80 × 120
OR = = = 1:00
b × c 80 × 120

The odds of eating vegetables among depressed patients were the same as the
odds in the control group. An OR of 1.00 was calculated, indicating a lack of
association between depression and eating vegetables. The results of the study did
not show an association between eating vegetables and suffering from depression.
98 Q. Wu

In summary, OR > 1 indicates that the factor may increase the risk of disease,
OR < 1 indicates that the factor may attenuate the risk of disease, and OR = 1
indicates no association.

Confidence Interval Estimation of Odds Ratio

An OR is a point value estimate, which may have a random error. The OR

confidence interval gives the range of estimates of the OR. The range of estimates
is calculated based on a given set of sample data. The OR confidence interval
reduces the random error generated by a single study. Ninety-five percent confidence
interval (CI) means a 95% probability which the interval includes the true OR. If
95% CI range includes “1,” it is not statistically significant since it could be either a
risk factor (OR≧1) or a protective factor (OR ≦ 1). If 95% CI range is greater than
1, the exposure is a significant risk factor (OR ≧ 1) with a probability of higher
than 95%.
An approximate 95% CI around the point estimate of OR for an unmatched case-
control study can be calculated using the formula (5.5).

1 1 1 1
OR95%CI = ðORÞexp ± 1:96 þ þ þ ð5:5Þ
a b c d

Where exp. is the natural logarithm, and a, b, c, and d represent the numerical
entries into the summary format in Table 5.1.

1 1 1 1
95%CI = ð7:5Þ exp ± 1:96 þ þ þ = ð7:5Þ expð ± 1:96 × 0:477Þ
22 36 7 86

Lower bound = ð7:5Þ expð- 1:96 × 0:477Þ = ð7:5Þ expð- 0:94Þ = 2:9
Up bound = ð7:5Þ expðþ1:96 × 0:477Þ = ð7:5Þ expðþ0:94Þ = 19:1

The CI provides two values, low (L ) and high (U ), with a specific confidence
level between these two values for the population parameter. A 95% confidence
interval means that if we conduct a study, there is a 95% probability that the results
will fall within the confidence interval. The above example illustrates that the
interval between 2.9 and 19.1 includes a probability of 0.95 for the true OR value.

5.3.3.2 Matched Design

In individually matched case-control studies, the analysis must take into account the
matched sampling scheme. When a control is matched to one case, summary data in
the format shown in Table 5.5 can appear. This table is different from the one that we
5 Case-Control Studies 99

Table 5.5 A 1:1 matched case-control study

Control exposed Control unexposed Total
Case exposed a c a+c
Case unexposed b d b+d
Total a+b c+d a+b+c+d

introduced in our previous group matching analysis. Each cell in Table 5.5 repre-
sents not one subject but a pair (one case and one control). Each case-control pair can
be classified as one of the exposure states. Just as Table 5.5, “a” means numbers of
pairs that both case and control exposed while “c” means numbers of pairs that case
exposed but control unexposed. “b” means numbers of pairs that case unexposed but
control exposed. “d” means number of pairs that both case and control unexposed. In
the analysis of individual matching studies, only pairs with inconsistent exposure
were used. Inconsistent pairs of exposures occur when the exposure status of the case
differs from that of the control group.

2 × 2 Table

χ2 Test and Statistical Inference

ðjb - cj - 0:5Þ2
χ2 =
bþc

OR and 95%CI

The OR of individual matched case-control study is calculated by simple ratio.

c
OR = ð5:6Þ
b
1 1
OR95%CI = ðORÞexp ± 1:96 þ ð5:7Þ
b c

The significance of individual matching OR is the same as that of group matching

case-control study. Endometrial cancer and estrogen are used as examples to illus-
trate the procedure for calculating OR in individual matched case-control studies.
The 390 pairs consisted of 390 patients with endometrial cancer and 390 controls,
100 Q. Wu

Table 5.6 Hypothetical matched case-control study

Control exposed Control unexposed Total
Case exposed 84 96 180
Case unexposed 48 162 210
Total 132 258 390

Table 5.6. Exposure is defined as women who have ever taken any estrogen. The OR
form the study is as below.

c 96
OR = = = 2:00 OR95%CI = ð1:40 to 2:89Þ
b 48

The calculation method of OR 95% confidence interval (CI) of individual

matched case-control study is the same as that of group matched case-control
study. Formula 5.7 gives the formula for calculating the OR 95% confidence interval
of individual matched case-control study. The approximate 95% CI for the OR is
1.40 to 2.89. This individually matched case-control study showed a moderate
association between endometrial cancer and estrogen use.

5.4 Common Bias and Controlling

A case-control study is an observational study in which subjects are enrolled based

on the presence or absence of the disease of interest. The exposure history of both
groups is then evaluated to determine the strength of the association between disease
and exposure. Case-control studies are susceptible to observational epidemiological
study bias. These biases include selection, information, or confounding biases.

5.4.1 Selection Bias

Selection bias is the most common bias in case-control studies. Selection bias may
exist if the control group is not from the source population that generated the cases.
For example, to study asthma, cases of asthma are drawn from high school students,
while people without asthma are drawn from the elderly population to form a control
group. The fact that the control and case groups are not a source population has the
potential to introduce serious bias. The factors that cause asthma may be different in
younger and older people. Thus, based on studies of such mismatched cases and
controls, many of the factors that may be found to be associated with asthma may
simply be due to the different ages of the two populations.
Sampling of controls and cases can sometimes be stratified, e.g., by sex and age
group. In addition to this, there should be randomization in subgroups of subjects
5 Case-Control Studies 101

with and without disease. However, researchers are often not randomly sampled, and
selection bias arises. This bias poses a significant impact on the validity of case-
control studies.
Bias does occur when the sampling fractions depend jointly on exposure and
disease, usually because exposed controls are more or less likely to be sampled than
non-exposed controls. When hospital patients are utilized as cases and controls, the
control is not a random sample of the target population because the control is a
subset of hospital patients. Cases in case group are only part patients in the hospital.
Patients and hospitals are mutually selective. The systematic differences in some
characteristics between the case group and the control group are unavoidable,
resulting in an admission rate bias. This is also known as Berkson bias.
The following factors contribute to selection bias.

5.4.1.1 Prevalence-Incidence Bias

More information might have been obtained if the survey respondents had chosen
existing cases, but much of this information was only relevant to survival and may
have overestimated the etiologic role of certain exposure factors. In addition,
survivors of a disease change their habits so as to reduce the level of a risk factor
or distort their pre-morbid habits when they are investigated, resulting in the
association of a factor with the disease being incorrectly estimated. This type of
bias is usually referred to as prevalence-incidence bias. Therefore, new cases should
be included in the investigation as much as possible to avoid the effect of prevalence-
incidence bias.

5.4.1.2 Unmasking Bias

Patients often seek medical attention for certain symptoms unrelated to the causative
agent, thereby increasing the detection rate of early cases and leading to an
overestimation of exposure. This systematic error is then referred to as
unmasking bias.

5.4.1.3 Subject Refuses Participation

In case-control studies, the most common reason is that subjects refuse to participate,
either by actively refusing to sign a consent form or by passively not returning
questionnaires or failing to attend laboratory tests at the specified time. Cases tended
to be highly motivated to participate, while controls selected from the population
were not willing to participate. Participation rates in the control group tended to
depend on a number of factors related. For example, rejection rates for telephone
surveys are higher for people who are older, less socially connected, less educated,
and have lower incomes.
102 Q. Wu

5.4.2 Information Bias

Information bias is a systematic bias in the process of collecting and organizing

information due to flaws in the methods used to measure exposure and outcome.
Even if the classification of subjects’ exposure and outcome is completely accurate,
bias may be introduced due to different choices in case-control studies. More
commonly, subjects are incorrectly classified in terms of exposure status or outcome,
and estimates of association can be biased. These errors are often referred to as
misclassification. Misclassification can be classified as differential misclassification
or non-differential misclassification.
Differential misclassification is also referred to as “recall bias.” Recall bias may
arise when cases remember past exposures more completely than controls. This often
happens because cases tend to try to find out the cause of their disease. As a result,
when they are interviewed, they tend to report more information about the past.
Control do not deliberately report information about past exposures.
The second type of information bias is non-differentiated misclassification.
Non-differential error classification means that the frequency of errors is similar in
the case and control groups. Misclassification of exposure status is more serious than
misclassification of outcome. However, both misclassifications can bias a study. For
example, a case-control study was conducted to explore the relationship between a
high-fat diet and coronary artery disease. Subjects with heart disease and controls
without heart disease were recruited and asked to fill out a questionnaire about their
dietary habits. Then they were determined whether to consume a high fat diet. It is
difficult to accurately assess the amount of fat in the diet from questionnaires.
Therefore, it would not be surprising if there were errors in the classification of
exposure. In such cases, misclassification may occur regardless of the final disease
status. When exposed is qualitative variables, non-differential misclassification
always favors the null. Or, if there is an association, whether positive or negative,
it tends to minimize it. For example, the OR between a high-fat diet and coronary
heart disease is 5.0, but a biased estimate might give an OR is 2.4 if about 20% of
exposed subjects are misclassified as “non-exposed” in both disease and control.
This implies that the bias tends towards the null.
If there are multiple exposure levels, non-differential misclassification may bias
the estimate toward or away from the null, which rely on the category to which the
subject was misclassified.

5.4.3 Confounding Bias

Confounding is that the relationship between exposure factors and outcomes

is distorted by external variables. The systematic error generated by this distortion
is the confounding bias. Confounding factors usually have three characteristics. One
is a variable associated with the exposure and independent of that exposure, and the
5 Case-Control Studies 103

third is a risk factor for the disease. The distortion introduced by confounding factors
can be significant, and it can even change the direction of the effect. However,
confounding bias can be adjusted for in the analysis, which is different from
selection and information bias. For example, the crude death rate in city A may be
higher than the crude death rate in city B, but after adjusting for age, there is no
difference in the adjusted death rate between cities A and B. The age-induced
deviation in crude death rates in two cities is known as the confounding bias.
There are two strategies for controlling confounding. Prevent confounding bias
from occurring in the first place, which can be done by limiting or matching during
the study design phase. Next is to deal with it when it occurs by using analytic
techniques such as stratification and statistical model. The effectiveness of all of
these strategies except randomization depends on the ability to identify and measure
any confounders accurately.

5.5 Strengths and Weaknesses of Case-Control Studies

5.5.1 Advantage of the Case-Control Study

Case-control studies save time, cost less, and are the most effective design. Case-
control studies are the preferred choice for rare disease research. This is because in a
cohort design, studies of rare diseases must follow many people to identify those
with outcomes. Case-control studies, on the other hand, do not have to worry about
no outcomes occurring. Case-control studies are also advantageous in studying
diseases with longer latency periods.
In addition, case-control studies have several other advantages. First, occurrence
of exposure in subjects retrospectively investigated in case-control studies. Investi-
gators do not have to follow study subjects over time as in cohort studies. Investi-
gators do not have to follow study subjects over time to collect exposure and disease
information as they do in cohort studies. Finally, the sample size of the case-control
study was small. Compared to cohort studies and experimental studies, case-control
studies are easier to implement. (Table 5.7).

Table 5.7 Advantages and disadvantage of case-control studies

Advantages Disadvantages
Suitable for research on rare diseases The relative risk of disease cannot be directly
estimated
Suitable for long latency chronic dis- Not suitable for studying rare exposures
ease studies
Smaller sample size required compared More susceptible to selection bias than alternative
to other types of studies designs
Less expensive than alternative designs Information on exposure may be less accurate than that
Save time over other types of study available in alternative designs.
designs
104 Q. Wu

5.5.2 Disadvantage of the Case-Control Study

Case-control studies are divided into case and control groups according to the
presence or absence of the disease of interest. Therefore, incidence rates could not
be calculated for either group. Without knowing the incidence, it is not possible to
calculate the relative risk in case-control studies. One can calculate the OR in a case-
control study, which is a measure of association that approximates relative risk under
certain condition.
The temporal sequence of exposure and disease may be difficult to determine in a
case-control study, so it may not be possible to know whether the exposure occurred
before the disease. For example, A case-control study of asthma in high school
students suggests an association between asthma and cat ownership. However, it
may be difficult to know whether high school students had cats first or whether they
had asthma attacks first. People usually choose newly diagnosed cases to overcome
this drawback.
Although case-control studies have advantages in studying rare diseases, they are
not suitable for studying rare exposures (Table 5.7). For example, we would like to
study the risk of asthma associated with working in a nuclear submarine shipyard
and would probably not prefer a case-control study because only a small percentage
of people with asthma would be exposed to this environmental factor.
Case-control studies are grouped by study disease, so they can only be used to
study one disease. However, it is possible to study the association between a disease
and multiple factors. If want to study more than one disease, you can consider a
cohort study design.
In conclusion, case-control studies are a more efficient research method, but the
results are susceptible to the influence of known and unknown confounding vari-
ables. Case-control studies are suitable for investigating the association between
diseases and factors, and the etiology of diseases. When there is limited evidence on
a topic, there are cost-effective ways to raise and investigate hypotheses before
conducting larger and more expensive studies. Sometime, they are often the only
choice of research method, especially when cohort studies or randomized controlled
trials are impractical. Case-control studies investigated information about each sub-
ject’s exposure up to a certain time period. Case-control studies require first defining
the case, then identifying the source population that generated the case, and finally
identifying the case group and control group. The studies have some strong charac-
teristics such as being cheap, efficient.
Chapter 6
Experimental Epidemiology

Xing Liu

Key Points
• Follow the ethical principles! Read the Declaration of Helsinki. Always remem-
ber: “The health of my patient will be my first consideration.” and “A physician
shall act in the patient’s best interest when providing medical care.”
• Experimental study serves as the “gold standard” in medical studies for causal
inference.
• Different from observational studies, researchers determine the status of exposure
of the participants in experimental studies.
• A successful randomization with a large sample size is powerful in eliminating
confounding due to known and unknown confounders at baseline. However, if
adherence to treatment is poor, or loss to follow-up is serious, new confounding
will arise. Loss to follow-up may also introduce selection bias.
An experimental study is the most powerful design in examining causal relation-
ships. The three major types of experimental study in humans include clinical trial,
field trial, and community trial, which differ by objectives, principles,
implementations, and target populations. Clinical trial aims to evaluate the treatment
effects of new drugs or therapies among patients to improve the prognosis. Field trial
aims to examine the potential preventive effect of the intervention in reducing
morbidity or mortality among healthy individuals. Community trial implements
the intervention among healthy people at the group level instead of at the individual
level. By performing experimental studies, researchers make causal inferences,
confirm the risk factors and protective factors for diseases, and evaluate the effects
of interventions in disease prevention and control.

X. Liu (✉)
School of Public Health, Fudan University, Shanghai, China
e-mail: [email protected]

© Zhengzhou University Press 2023 105

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_6
106 X. Liu

6.1 Basic Ideas of Experimental Study

An experimental study is a prospective study comparing the effect of an intervention

against a control in humans. In experimental studies, participants are assigned to
groups with different treatments or agents and followed up for a period of time to see
if the outcomes vary across groups.
An experimental study is generally expensive and ethically restricted, usually
focusing on a narrow question in a highly selected population with a well-defined
protocol. Therefore, the experimental studies are reserved for relatively mature
research questions suggested by previous observational studies appealing for further
confirmative evidence. The key difference between an experimental study and an
observational study is that the status of exposure is decided by researchers in an
experiment. Not all research question is amenable to the experimental design. It is
not ethical to expose people to harmful substances, and it is not always feasible to
study the long-term effect of an intervention.
In a classic two-group experiment, one group receives the treatment of interest,
and the other group does not. The experimental groups are expected to be identical
with respect to extraneous factors affecting the outcome. Thus, if the treatment of
interest does not have any effect on the outcome, an identical outcome frequency
with random variation would be observed between the two groups. In other words, if
the frequency of the outcome varies across the groups, the difference is attributable
to the treatment effect plus random variation. This objective can be achieved if all
extraneous factors and conditions which may have an effect on the outcome have
been controlled. However, in human studies, it is not possible to create completely
identical groups with respect to all extraneous factors. Instead, researchers expect the
groups to be comparable and exchangeable, with the net effect of extraneous factors
to be minimalized and much less than the effect of the treatment. In a classic
experimental study, a control group is always needed to provide a basis for compar-
ison, randomization is often employed to minimize the influence of confounding,
and blinding is used when possible, to eliminate the biases that arise from knowing
the treatment assignment.

6.1.1 Study Question

Each experimental study should have a specific question stated clearly and in
advance. This encourages proper design and enhances the credibility of the findings.
The primary question should be the one the researchers are most interested in and the
one that could be adequately answered. Generally, the primary question is based on
comparing outcomes across treatment groups. The outcome could be a beneficial
event including improved prognosis, prolonged survival, increased rate of cure,
released symptoms, reduced complications, or improved quality of life. There also
may be a series of secondary questions in an experimental study, which can be
6 Experimental Epidemiology 107

elucidated by the data collected. The secondary questions may comprise different
response variables and subgroup hypotheses. Both primary and secondary questions
should be relevant scientific questions, with important implications in medicine or
public health. Adverse events or side effects should also be collected through the
experimental study. Unlike the primary and secondary questions, adverse events and
side effects may not always be specified in advance. Investigators usually monitor a
variety of clinical and laboratory measurements and record the reports from partic-
ipants. The safety and well-being of participants are the most crucial concerns in
performing an experiment. Investigators should always monitor the balance of
benefits and risks, and be guided by the independent ethical review committees.

6.1.2 Choice of Intervention

The intervention techniques employed by an experimental study may be single or a

combination of diagnostic, preventive, or therapeutic biologics, drugs, regimens,
devices, or procedures. In an experimental study, the intervention a participant
receives is assigned by the investigator for the purpose of a study instead of the
subject’s need. Ethical constraints severely limit the types of interventions and
circumstances for an experiment to be performed on human subjects. Adherence
to the scientific protocol should not conflict with the subject’s best interests. Any
exposure given to participants should cause no known harm and should be limited to
potential prevention or cure of disease.

6.1.3 Choice of Control

The choice of the control group is an important design issue in experimental studies,
for it provides the basis to make a valid comparison. The methodological principle of
choice of control is that the distribution of extraneous factors is the same between the
intervention group and the control group to make the two groups comparable. The
ethical principle of choice of control is that if there is an optimal, known best therapy
or standard, usual care, the new intervention should be compared against it, or added
to it. The commonly used types of control groups include standard control, placebo
control, self-control, and cross-over control.

6.1.3.1 Standard Control

Standard control is the most commonly used control in clinical trials. The optimal or
standard treatment is assigned to the control group or to both groups, while a new
treatment or new therapy is assigned or added to the intervention group. The effect of
the new treatment should be compared against the standard care when the latter is
108 X. Liu

available, instead of against a placebo. Although comparing a new treatment to a

placebo or blank control might provide a larger effect size, the goal of the study is to
determine whether the new treatment is better than the one currently used, but not if
the new treatment has any effect.

6.1.3.2 Placebo Control

When a new intervention is added to standard care or usual care, it is compared

against that care plus a placebo. Or when there is no standard care available, the new
intervention is compared against a placebo. Placebo control is also commonly used
in field trials including the trial of vaccines. Using a placebo control has two major
benefits: helps in keeping the blinding, and helps in controlling the “placebo effect.”
For keeping the blinding successful, the formulation, size, and appearance of the
placebo should be identical to the new drug. Thus, the participants would not know
which study group they are in, improving their adherence to taking the treatment,
and preventing them from dropping out once they know they are not in the
intervention group. The researchers would also have no idea about which groups
of participants are taking the intervention therapy, preventing them from making
differential observations and data collection.
The placebo itself has no treatment or preventive effect at all. However, using any
form of the drug may induce certain “effect” in both the intervention group and the
control group. This kind of psychological benefit is called the placebo effect, even if
it occurs among participants in the intervention group. By using a placebo in the trial,
the psychological effects in both groups cancel out, and the real effect of the
intervention can be observed. However, if the drug or the treatment of intervention
has a certain side effect, the subjects might gradually realize the assigned group, and
blinding might be broken, thus the compliance and the control of the placebo effect
might be weakened.

6.1.3.3 Self-Control

The subjects themselves may serve as the control group before the intervention is
given. Or the contralateral body or organ may serve as the control when intervention
is assigned to one side. But researchers still have to pay attention if the extraneous
factors change before and after the intervention is given. If so, the estimate of the
effect might still be confounded.

6.1.3.4 Cross-Over Control

The cross-over design also allows the subject to serve as his or her own control,
while in this case, the study has more than one period. In the first period, each subject
receives either intervention or control treatment, and in the second period the
6 Experimental Epidemiology 109

alternative. The order in which intervention or control is given to the subject is

usually randomized. Depending on the characteristics of the intervention, a wash-out
period is required between the two periods. The use of the cross-over design is thus
limited to those interventions that the effects during the first period can be washed
out. A cross-over control may have more than two periods and may have more than
two arms.

6.1.4 Randomization

Observational studies are often used to compare the effects of different treatments
given to patients in clinical settings. However, when some of the manifestations
affect both the outcome and the treatment allocation, the effect estimates can be
biased. The patients in different treatment groups differ in many ways, and the
groups might be incomparable. For example, the general condition of a patient has
a definite impact on the disease progression and prognosis, and the general condition
also determines whether the doctor would choose surgery or a more conservative
treatment. In this kind of situation, the differences observed in the outcome between
groups may contribute to not only the potential treatment effect but also the
confounding brought by the severity of the disease. And this type of confounding
is called “confounding by indication.” Thus, the effect estimates gained from
observational studies are faced with uncontrolled confounding when the different
treatment groups are not comparable, since not all confounders can be realized,
identified, measured, and controlled.
The observed association in an observational study comprises the treatment
effect, systematic bias, and random error. An experimental study aims to eliminate
the part of systematic bias. First of all, it is crucial to reduce to the best extent of
incomparability in different treatment groups by balancing the extraneous factors
affecting the outcome. Ronald Aylmer Fisher and others developed the practice of
randomization to account accurately for extraneous variability in experimental
studies. A random assignment mechanism is used to assign treatments to subjects,
and the mechanism is unrelated to those extraneous factors that affect the outcome.
Thus, the difference in the outcomes across groups that is not attributable to
treatment effects could be attributed to chance. A study with random assignment
of exposure allows computing the probability of the observed association under the
hypothesis and making a statistical inference based on the compatibility between the
observation and the hypothesis. Randomization guarantees that statistical tests have
valid rates of false positive error.
Successful randomization with a sufficient sample size generates comparable
groups at baseline. Not only known confounding factors but also those unknown
confounders are balanced across groups. However, compliance with the follow-up
and adherence to treatment is critical during the study period to make the effect
estimate valid. If adherence to treatment is influenced by extraneous factors affecting
the outcome, confounding will arise and affect the effect estimate between exposure
110 X. Liu

received and the outcome. If the loss to follow-up is severe, it would not only affect
the study efficiency but also introduce selection bias and confounding if the loss to
follow-up is differential with regard to exposure, outcome, and extraneous factors.
Therefore, it is important to maintain a low rate of loss to follow-up and high
adherence to assigned treatments during the study period.

6.1.4.1 The Randomization Process

Randomization is a mechanism or process by which each subject has the same

chance of being assigned to either the intervention group or the control group.
Several methods of randomly assigning subjects are introduced here. The commonly
used methods for randomization include simple randomization, blocked randomiza-
tion, and stratified randomization.

6.1.4.2 Simple Randomization

Simple randomization or complete randomization is the most elementary form of

randomization. To toss an unbiased coin when a participant is eligible for random-
ization is an example. One might also use a random digit table on which the equally
likely digits from zero to nine are arranged by rows and columns. For larger studies,
one may use a random number-producing algorithm provided by most statistical
software to generate random numbers in the interval from 0.0 to 1.0 for each subject.
The procedure might assign subjects to group A with probability p and subjects to B
with probability 1-p. If the random number is between 0 and p, the subject would be
assigned to group A, otherwise to group B. And this procedure can be adapted to
more than two groups. The advantage of simple randomization is that it is easy to
implement. Simple randomization generates an anticipated proportion for the num-
ber of subjects in each arm in the long run with a large sample size. However, at any
point in the process of randomization, or when the sample size is small, there could
be a substantial imbalance. Although this kind of imbalance would not cause the
statistical tests to be invalid, it harms the statistical efficiency.

6.1.4.3 Blocked Randomization

Blocked randomization is used to avoid serious imbalance in the number of subjects

assigned to each group when the sample size is small. It also helps to have balanced
numbers at any point in the randomization procedure during enrollment. If partici-
pants are randomly assigned with equal probability to groups A or B, then for blocks
with even size, one-half of the participants would be assigned to each group. The
order in which the treatments are assigned in each block is randomized. For example,
if a block of size 4 is used, there are six possible combinations of treatment
assignments: AABB, ABAB, ABBA, BAAB, BABA, and BBAA. Select one
6 Experimental Epidemiology 111

from these arrangements randomly and apply it accordingly to the four participants
entering the study. Repeat for every consecutive group of four participants until all
are randomized. Advantage of the block randomization is that the number of
participants in each group is always balanced during the process of randomization,
at any time point, and with any sample size. The disadvantage is that strictly
speaking, data analysis is more complicated for blocked randomization than for
simple randomization. And the use of blocked randomization should be taken into
consideration during data analysis.

6.1.4.4 Stratified Randomization

Randomization balances extraneous factors that affect the outcome in studies with
large sample sizes and for small studies on average. However, for one single study
especially with a small sample size, it is possible that not all baseline characteristics
distribute evenly across groups. When there is the concern of imbalance for major
prognostic factors, one might employ stratified randomization within the strata of
those factors considered. If several factors are considered, the number of strata is the
product of the number of subgroups for each factor. Within each stratum, the
randomization process could be a simple randomization or a blocked randomization.

6.1.5 Blinding

Blinding is one of the solutions to reduce systematic biases in experimental studies.

Not knowing which group the participant is in, the adherence to the exposure, the
compliance to the follow-up, and the measurement of outcomes can be improved.
Thus, blinding is often employed when it is possible. Some kinds of trials can only
be conducted without blinding, including those that have surgical procedures,
changes in lifestyle, or behavioral interventions. The main disadvantage of an
unblinded experiment is that participants may report symptoms and side effects
differentially between intervention and control groups. Also, researchers may mea-
sure and collect these data differentially when knowing which group the subjects are
from. Moreover, the participants in the control group may have a higher possibility
of leaving the study, when knowing that there would not be any extra benefits.
There are at least four levels of blinding in a trial. First, the participant does not
know which treatment group he or she is in. The adherence and compliance to the
study would not be influenced, and the accuracy of the report of symptoms would
not be affected. Second, the staff assigning participants to different treatment groups
do not know which group the participants would be assigned to. This avoids
participant assignment based on the staff’s willingness. Third, the physicians taking
care of participants during the whole process do not know which treatment group the
participants are in. This ensures the health care provided and symptoms and clinical
manifestations recorded would not be affected. Fourth, the researchers including
112 X. Liu

statisticians do not know which treatment group the participants are in. This makes
sure that the measurements of treatment effects, the record of side effects, and data
analyses would not be affected.
There are three common methods for performing blinding in practice: single-
blind, double-blind, and triple-blind.

6.1.5.1 Single-Blind

In a single-blind study, the participants do not know which treatment group they are
in. Thus, the biased report of symptoms and side effects by subjects can be reduced.
However, the researchers can still influence the administration of treatment, data
collection, and analysis in a single-blind study.

6.1.5.2 Double-Blind

In a double-blind study, neither the participants nor the researchers know the
treatment assignment. The risk of bias is greatly reduced in a double-blind study.
The actions of investigators would occur equally to participants from both groups.
Double-blind studies are usually more complex to carry out than a single-blind or
unblinded study. An effective data monitoring protocol should be set up. And the
emergency unblinding procedures must be established.

6.1.5.3 Triple-Blind

A triple-blind study is an extension of a double-blind design. And it may have

different definitions under different circumstances. In some designs, it is the com-
mittee monitoring response variables that is not aware of the treatment assignment.
While in some designs, it is the group performing statistical analyses that has no idea
of the treatment assignment. Thus, the bias introduced during statistical analyses can
be avoided.

6.1.6 Data Analysis

Data analysis in experimental studies has special strategies. Noncompliance with the
assigned treatment results in a discrepancy between the treatment assigned and the
treatment actually received. The standard practice of data analysis in the experimen-
tal study is making comparisons based on the treatment assigned instead of received.
Such a practice is called the intent-to-treat analysis (ITT). Comparisons based on the
treatment received are called according-to-protocol (ATP) or per-protocol analysis
(PP). If the compliance to treatment is poor, or there is a considerable loss to
6 Experimental Epidemiology 113

Uncontrolled confounding

Randomization Treatment Outcome

Fig. 6.1 A causal diagram with valid instrument randomization, for the treatment—outcome effect.
If 1. Randomization affects outcome; 2. Randomization affects outcome only through treatment;
3. Randomization and outcome share no common causes; and then randomization can be taken as a
valid instrumental variable in examining the association between treatment and the outcome. The
association between treatment and outcome might have been affected by uncontrolled confounding,
however, the association between randomization and outcome has not been confounded

follow-up during the study, the association between the exposure received and the
outcome might be biased. ITT analysis preserves the validity of the test for the null
hypothesis of treatment effects.
In an intent-to-treat analysis, no matter how the compliance to the assigned
treatment is, the analysis takes the assigned treatment as the exposure to test the
null hypothesis between the exposure and the outcome. As mentioned earlier,
successful randomization is not affected by the extraneous factors affecting the
outcome. And randomization has an effect on the outcome through and only through
the actual treatment received. Although the association between the treatment
actually received and the outcome can be confounded, the association between the
treatment assigned (randomization) and the outcome will not be confounded. This
makes randomization a valid instrumental variable in examining the association
between the treatment and the outcome. The use of the instrumental variable protects
the validity of the test of the null hypothesis between treatment and outcome,
although the effect estimate might have been biased (Fig. 6.1).

6.1.7 Sample Size

An experimental study should have sufficient statistical power to detect the differ-
ences across treatment groups. The sample size of a study is decided by the
following aspects:
(1) The significance level, denoted as α. It is the probability of a false positive
finding, or Type I error.
(2) The probability of a false negative result, or Type II error, denoted as β. 1 - β is
the statistical power of the test.
(3) The difference between the measurements of the outcome across the groups.
114 X. Liu

6.1.7.1 Sample Size Calculation for Dichotomous Response Variables

2
Ζα 2Pð1 - PÞ þ Ζ β Pc ð1 - Pc Þ þ Pe ð1 - Pe Þ
N= ð6:1Þ
ðPc - Pe Þ2

Where N = the sample size for each group, Pc is the event rate for the control
group, Pe is the event rate for the treatment group, P = (Pc + Pe)/2, Ζα is the critical
value which corresponds to the significance level α, and Ζβ corresponds to the power
1 – β.

6.1.7.2 Sample Size Calculation for Continuous Response Variables

2
2 Ζ α þ Ζ β σ2
N= ð6:2Þ
d2

Where N = the sample size for each group, σ is the estimated standard deviation,
d is the estimated difference of the means, Ζα is the critical value which corresponds
to the significance level α, and Ζβ corresponds to the power 1-β.

6.1.7.3 Sample Size Calculation for “Time to Failure”

2
2 Ζα þ Ζβ
N= 2
ð6:3Þ
λc
ln λe

Where N = the sample size for each group, λ is called the hazard rate or force of
mortality, Ζα is the critical value which corresponds to the significance level α, and
Ζβ corresponds to the power 1 – β.

6.2 Clinical Trial

6.2.1 Basic Ideas of Clinical Trial

Clinical trial is an experimental study with patients as subjects. The goal of a clinical
trial is to evaluate a new drug or therapy for a disease to improve prognosis, reduce
mortality or improve the quality of life among patients. It also collects information
6 Experimental Epidemiology 115

on the adverse effects of a new treatment and provides evidence on the effectiveness
and safety of the treatment to enter clinical use.
Subjects in a clinical trial are patients with the disease in question. Participants in
a clinical trial should meet the criteria of eligibility well-defined in advance. Patients
who do not meet those criteria should not be enrolled. And subjects whose illness is
too severe or too mild usually will not be considered eligible since they are less likely
to permit the form of treatment or to complete the follow-up. Patients with compli-
cated conditions are usually excluded especially at earlier stages of the trial because
of the need to minimalize differences in the extraneous factors affecting the outcome
between treatment groups. Therefore, at earlier stages of the trial, the participants are
usually a highly selected population with restricted criteria for inclusion, affecting
the generalization of the conclusion.

6.2.2 Phases of Clinical Trial

When comparing the effectiveness of a new drug, several phases of clinical research
must be performed. Classically the trials of pharmaceutical agents involve phases I
to IV.

6.2.2.1 Phase I Studies

Phase I studies collect early data in humans after preclinical information is obtained
from in vitro or animal studies. Participants in phase I studies are generally healthy
volunteers with sample sizes ranging from 20 to 100. Phase I studies characterize
pharmacokinetics and pharmacodynamics and estimate the tolerability in humans.
The questions including bioavailability, body compartment distribution, and drug
activity are answered by phase I studies. The maximally tolerated dose, the safety
range of the dose, and the recommended dose is explored at this stage. Phase I also
collects data on side effects.

6.2.2.2 Phase II Studies

Phase II studies evaluate whether the drug has any biological activity or effect once
the dose or range of dose is determined with sample sizes ranging from 100 to 300. A
phase II study usually employs a randomized control design, compares the effect of
the new drug against the standard drug or a placebo, and evaluates the effectiveness
and safety of the new treatment. Phase II studies continue to collect side effects data,
evaluate the safety, and recommend the dose for clinical use.
116 X. Liu

6.2.2.3 Phase III Studies

Phase III studies are generally multi-center randomized controlled trials conducted in
different countries with sample sizes ranging from 300 to 3000 or more. Phase III
studies further evaluate the effectiveness and safety of the new drug or therapy
against the standard care, confirming the value in clinical use. Phase III collects data
on the adverse effects and the interaction of the drug with other drugs. The treatment
approved after phase III can be used in clinical settings.

6.2.2.4 Phase IV Studies

Phase IV studies are conducted after the new treatment is approved and used
clinically. All patients who received the new drug can be considered participants.
The participants enrolled before phase IV are generally highly selected with
restricted criteria for eligibility, which limits the generalizability of the study con-
clusions. Phase IV studies observe the drug efficacy in the real world, and those
patients with complex conditions may also be enrolled. Thus, the limitations of
earlier studies can be improved. Phase IV studies are generally open cohort studies,
monitoring drug efficacy, side effects, and interaction with other drugs at a large
scale in the long run. Phase IV studies can collect data on side effects especially the
ones that occur rarely or late.

6.2.3 Case Study of Clinical Trial

A randomized, phase II study examined the efficacy of carboplatin and pemetrexed

with or without pembrolizumab for advanced, non-squamous non-small-cell lung
cancer (NSCLC). The investigators enrolled 123 non-squamous NSCLC stage IIIB
or IV patients without former chemotherapy and targetable EGFR or ALK genetic
aberrations from 26 medical centers in the USA and Taiwan, China. A 1:1 ratio in
blocks of four randomization assigned 60 to the group of pembrolizumab plus
chemotherapy, and 63 to the group of chemotherapy alone. The primary endpoint
was the proportion of patients who had radiologically confirmed complete or partial
responses. Fifty-five percent (95% CI 42–68%) of patients in the pembrolizumab
plus chemotherapy group achieved this objective response compared with 29%
(18–41%) of patients in the chemotherapy alone group (treatment difference 26%
[95% CI 9–42%]; P = 0.0016). The incidence of grade 3 or worse treatment-related
adverse events was similar between groups (39% in the pembrolizumab plus che-
motherapy group and 26% in the chemotherapy alone group). The most common
grade 3 or worse adverse events in the pembrolizumab plus chemotherapy group
were anemia (12%) and decreased neutrophil count (5%). In the chemotherapy-alone
group, the most common were anemia (15%) and decreased neutrophil count,
6 Experimental Epidemiology 117

pancytopenia, and thrombocytopenia (3% each). 2% of patients in the

pembrolizumab plus chemotherapy group experienced treatment-related death com-
pared with 3% in the chemotherapy group. These results suggested that the combi-
nation of pembrolizumab plus chemotherapy may be an effective treatment for
patients with early, advanced non-squamous NSCLC.

6.3 Field Trial

6.3.1 Basic Ideas of Field Trial

Field trial differs from a clinical trial in the subjects. The participants in a clinical trial
are those patients diagnosed with the disease of interest in clinical settings; while the
participants in a field trial are those healthy people from the community. Field trial
often requires a larger sample size and recruit participants who are not under clinical
management. Therefore, they are often more expensive and difficult to conduct. A
field trial is limited to studying the prevention of common or severe diseases. The
interventions for field trials include health supplements, vaccines, and changes in
lifestyle. The principles of study design, control selection, randomization, and
blinding for experimental studies apply to field trials. Field trials are used to confirm
the causal relationship, risk factors, and preventive factors for diseases and to reduce
morbidity.

6.3.2 Design and Implementation

Participants in the field trial are free-living healthy people recruited from the
community. The management and conduct of a field trial would be more difficult
than a clinical trial. A well-designed feasible protocol on a solid scientific question is
crucial for a successful field trial.

6.3.2.1 A Specified Question

A clear scientific question should be stated in advance including the specific

intervention and anticipated outcome. The objective of the study should be based
on a clear research hypothesis. The intervention should be derived from a risk factor
for disease with relatively sufficient evidence from observational studies. And the
conclusions gained from the study should have benefits for individuals or public
health.
118 X. Liu

6.3.2.2 Inclusion and Exclusion Criteria

The participants for the field trial are healthy people from the community and are at
risk for disease of interest. The inclusion and exclusion criteria should be defined in
advance based on the study objective and should be implemented strictly. Partici-
pants can be enrolled from those communities with low mobility to avoid a substan-
tial loss to follow-up, otherwise, selection bias may arise and harm the validity of the
conclusion. Also, if the disease of interest is of low incidence rate in the population,
it is suggested to conduct a field trial in the population at higher risk for the disease to
save resources for long-term follow-up. Restricted inclusion criteria and highly
selected participants may have an influence on the generalizability of the research
conclusion.

6.3.2.3 Choice of Intervention

A clear definition and description of the intervention are necessary. The dose,
contents, method, frequency of application, etc. of the intervention should be
introduced clearly. Adherence to intervention is critical for field trial participants.

6.3.2.4 Time and Interval of Follow-up

The time of each visit and interval during follow-up are decided by the effect of an
intervention. Investigators balance the need for collecting necessary data,
maintaining participants in the follow-up, and the cost. During the study, it is
important to improve the compliance and adherence of the participants to avoid a
loss of follow-up and selection bias.

6.3.3 Case Study of Field Trial

Efficacy of a bivalent L1 virus-like particle vaccine in prevention of infection with

human papillomavirus types 16 and 18 in young women: a randomized
controlled trial
Genital human papillomavirus (HPV) infection leads to cervical cancer, a major
cause of cancer deaths in women worldwide. 230,000 die and 470,000 are diagnosed
due to cervical cancer annually. The most prevalent oncogenic HPV strains, HPV-16
and HPV-18, can be vaccinated to prevent up to 70% of cervical cancers from
developing. A bivalent HPV-16/18 L1 virus-like particle vaccine was tested for
efficacy, safety, and immunogenicity in a randomized, double-blind, controlled trial.
Between July and December 2000, 1,113 North American and Brazilian women
aged 15–25 were enrolled with an average age of 20. HPV infection was tested using
6 Experimental Epidemiology 119

self-obtained cervicovaginal samples and cervical cytology. After randomization,

560 of 1,113 women received the vaccine and 553 received the placebo. 958 women
completed the first phase through month 18, with similar rates of vaccination and
placebo dropouts. According-to-protocol HPV-16/18 vaccine effectiveness against
the incident and persistent infection was 91.6% and 100%, respectively. Intention-
to-treat analysis showed 95.1% efficacy against persistent infection. Neither the
vaccine nor the placebo groups experienced any vaccine-related adverse effects. In
this trial, the bivalent HPV vaccine proved efficacious, safe, well-tolerated, and
highly immunogenic.

6.4 Community Trial

6.4.1 Basic Ideas of Community Trial

Community trial conducts intervention among healthy people, and the interventions
are given at the population level instead of at the individual level. Community trial is
used to evaluate the effect of interventions that are not suitable to be given at the
individual level. For example, some interventions on dietary factors are easier to be
performed at the family level; changing the source of drinking water from the river to
tap water is easier to be conducted at the community level. These kinds of interven-
tions are not given individually.
Community trial often uses cluster randomization. The success of cluster ran-
domization depends on the relative sample size within each group compared to the
total sample size. If the number of clusters is large, randomization has a higher
possibility to be successful. If there are only two communities randomized, the
meaning of randomization is limited and the comparability of baseline characteristics
of the two communities has a great impact on the results. During the study,
investigators need to pay attention to the changes in extraneous factors including
mobility, economic changes, medical care conditions, and implementation of other
programs in the community.

6.4.2 Case Study

Research on prevention and control strategies of liver cancer in Qidong and the
effect of the community trial
Liver cancer is one of the most common malignant tumors in China, which has a
serious impact on people's health. According to a survey from 1990 to 1992, the
standardized mortality rate of liver cancer in China was 17.83/100,000 person-years,
accounting for about 18.8% of cancer deaths. Nationwide, the mortality rate of liver
cancer in the 90s was higher than in the 70s. The incidence of liver cancer increased
120 X. Liu

after the age of 40 and increases with age, and the age of onset was earlier in high-
incidence areas. The male-to-female sex ratio was close to 3:1.
The increase in the incidence of liver cancer may come from the improvement of
liver cancer diagnosis, the increase in the proportion of middle-aged and elderly
people, and the increase in the incidence of liver cancer caused by the increase of
environmental carcinogens. In the early 1970s of the twentieth century, the risk
factors for liver cancer were not yet clear, and health workers carried out a large
number of investigations and studies in Qidong. The earliest case-control study
carried out in 1973 included 100 cases of primary liver cancer, 100 cases of other
malignant tumors, and 100 cases of healthy people, and explored the association
between liver history, tumor history, pesticide exposure and poisoning history,
drinking water source and water quality, tobacco, alcohol and eating habits, family
history, and other factors and liver cancer. Patients with hepatitis, liver cirrhosis, and
respiratory diseases in Qidong People's Hospital since 1964 were followed up to
confirm that patients with liver disease had a high risk of liver cancer. Since 1976, a
prospective cohort study has been carried out in Qidong, and long-term follow-up of
nearly 15,000 people has been carried out, and the incidence of liver cancer among
hepatitis B surface antigen carriers was 361.55/100,000, the incidence rate of
non-carriers was 30.90/100,000, and the relative risk was 11.70, confirming the
association between hepatitis B virus and liver cancer. The evidence accumulated by
years of long-term research suggested and basically clarified that hepatitis B virus,
aflatoxin, drinking water pollution, and gene susceptibility were risk factors for liver
cancer in this population.
Therefore, the prevention and control strategy of liver cancer in Qidong area was
as follows: to carry out intervention research on the suspected causes of liver cancer.
By observing changes in the incidence and mortality of liver cancer, the effect of the
intervention was evaluated and the etiology was further verified. A range of inter-
vention strategies and specific interventions were identified and implemented. In the
early 1970s of the twentieth century, measures of "prevention and control of
hepatitis, improvement of drinking water, and prevention of mildew in food" were
proposed. Put forward the requirement of "hydration of drinking water wells" to
reduce residents' drinking of ditches and river water, and later formed a "deep well
tap water supply network" to improve the quality of drinking water; Corn harvesting
adopted "fast harvest and quick drying into the warehouse to remove mold", and
then changed the staple food to rice, changing the eating habits of residents and
reducing the intake of aflatoxin from corn. Various measures have been taken to cut
off the transmission of hepatitis B virus and protect susceptible people, and since
1983, large-scale neonatal hepatitis B vaccination has been carried out in Qidong to
reduce the epidemic of hepatitis B virus. The academic views and research decisions
based on etiology research have been responded to and supported by the govern-
ment, forming a comprehensive prevention and control strategy on the spot.
6 Experimental Epidemiology 121

Interventions and on-site implementation of major risk factors for liver cancer
include:
1. Anti-mildew and Detoxification
As the main chemo-preventive measure, it was important to reduce the intake
of food contaminated with aflatoxin by the population. A number of case-control
studies and food testing have found a significant association between mildew in
corn and the occurrence of liver cancer. Prevention interventions were
implemented at two levels: changing the structure of staple foods at the commu-
nity level to promote the use of rice, with 96.4% of the population switching to
rice by 1986; At the individual level, it was promoted to prevent the intake of
mildew corn, and preventive measures are taken in the "harvest, storage, and
eating" process. This greatly reduced the aflatoxin exposure of Qidong residents.
2. Improve Drinking Water
Based on Qidong's research, Professor Su Delong proposed that the high
incidence of liver cancer was related to drinking water pollution. The incidence
and mortality of liver cancer among residents with different types of drinking
water differed significantly: the incidence of liver cancer in drinking ditch water
could be as high as 141.40/100,000, and the incidence of drinking deep well
water was 0.23/100,000. Algal toxins, microcystins, and other substances in ditch
water are cancer-promoting factors of liver cancer and may interact with afla-
toxin. Although there was no direct evidence of carcinogenesis, the drinking
water improvement project has solved the problem of drinking water pollution for
Qidong residents, and by 2010, 99% of residents were drinking pipe water.
3. Prevention and Treatment of Hepatitis B
HBsAg was screened in blood donors in Qidong, and positive people were not
allowed to donate blood, cutting off the transmission route of the virus and
reducing the epidemic. A randomized controlled intervention trial of hepatitis B
vaccine immunization for the prevention of liver cancer in nearly 80,000 infants
between 1984 and 1990 reported a decrease in HBsAg positivity, reporting a
75.9% immune protection rate and a decrease in HBV carrier rate among vacci-
nated people. After more than 20 years of follow-up in the second phase,
vaccination was found to have sustained immunity against chronic HBV
infection.
4. Carry Out Research on Early Diagnosis and Early Treatment
For the secondary prevention of liver cancer prevention and treatment, strat-
egies and research on early diagnosis and treatment were carried out in the area. In
the first stage, a large-scale screening of alpha-fetoprotein—a biomarker of liver
cancer was carried out in 1.8 million people in the 1970s, and a large number of
early cases were detected and treated; The second stage was in the 1980s: the
high-risk group of liver cancer in Qidong was defined as HBsAg-positive men
aged 30–59 years; In the third stage, in the 1990s, periodic screening of high-risk
groups was carried out and the screening effect was evaluated. The screening
results showed that the early case detection rate of the screening team was high,
and the survival rate was higher than that of the control group; In the fourth stage,
122 X. Liu

Qidong was established as a national sample for early diagnosis and treatment of
liver cancer in 2006. Most of the long-term survivors of liver cancer in Qidong
were beneficiaries who were found through screening and resected surgically,
which shows that screening can detect early cases, and after receiving appropriate
treatment, survival can be extended or even cured.
The decrease in morbidity and mortality is the goal of tumor prevention
and treatment and an important indicator to test the effect of intervention strategies
and measures. After more than 40 years of efforts, the age-standardized incidence
and mortality of liver cancer in Qidong have decreased. Although the crude inci-
dence and mortality rate of liver cancer in Qidong have increased in the past
40 years, after controlling for the factors of population growth and the increase in
the proportion of the elderly population, the incidence of age-standardized liver
cancer decreased from 49.95/100,000 in 1972 to 38.22/100,000 in 1990 and 25.75/
100,000 in 2011.
The decline in the incidence and mortality of liver cancer in Qidong was
accompanied by significant evidence of changes in risk factors for liver cancer.
From 1989 to 2012, the level of aflatoxin adducts representing aflatoxin exposure
decreased significantly, from 19.2 pg/mg in 1989 to 2.3 pg/mg in 1999 and
undetectable in 2009. The drinking water of residents was changed from the easily
polluted house ditch water and the water from the Yangtze River to tap water from
deep wells and the Yangtze River, and the quality of drinking water was significantly
improved. After 2002, the vaccination rate of hepatitis B among newborns reached
100%, the short-term and medium-term efficacy of the vaccine was confirmed, and
the long-term effect and association with the decline in the incidence of liver cancer
have yet to be confirmed by long-term follow-up. The above facts and evidence from
changes in biomarkers, ecological changes, and changes in population
immunoprevention show that even if the mechanism of action of some risk factors
for liver cancer has yet to be elucidated, after controlling these risk factors, the
incidence and mortality of liver cancer in the population have indeed decreased
significantly, which is enough to prove that these preventive measures are effective.
Chapter 7
Screening and Diagnostic Tests

Fen Liu

Key Points
• Screening is the process of using quick and simple tests to identify and separate
persons who have an illness from apparently healthy people.
• The validity of a screening test is defined by its ability to correctly categorize
subjects who do or do not have a disease into corresponding groups. The
components of validity include sensitivity, specificity, Youden’s index, and
likelihood ratio. Reliability is an index that reflects the stability of the testing
results. That includes agreement rate and Kappa statistic. The PPV is defined as
the probability of the persons having the disease when the test is positive. The
NPV is the percentage of the persons not having the disease when the test is
negative. The position of the cutoff point for a screening test will determine the
number of true positives, false positives, false negatives, and true negatives. For
continuous measurement data of a screening test, the cutoff point is determined
mostly by the ROC curve.
• Screening the high-risk population or performing multiple tests increased the
validity of a screening test.
• Volunteer bias, lead-time bias, and length-time bias are three major sources of
bias in screening test.
Screening is an effective strategy for early detection of diseases and is considered a
secondary prevention program in public health; diagnostic tests are helpful in
confirming diagnoses of diseases and can help the doctors determine the therapeutic
plans for patients. Along with the progress in science and technology, novel

F. Liu (✉)
School of Public Health, Capital Medical University, Beijing, China
e-mail: [email protected]

© Zhengzhou University Press 2023 123

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_7
124 F. Liu

screening and diagnostic tests are continuously put forward. Thus, the quality of
screening and diagnostic tests is a critical issue. In this chapter, we will address the
questions on how to assess the quality of various screening and diagnostic methods,
in particular, the newly available ones, and how to make reasonable decisions on
their application.

7.1 Design a Screening or Diagnostic Test

Although the purpose, observational subjects, and requirement of screening and

diagnostic tests are different, the principle for the evaluation of these two types of
tests is similar. Therefore, we take a screening test assessment as an example to
discuss.
Screening is the process of using quick and simple tests to identify and separate
persons who have an illness from apparently healthy people. To evaluate a new
screening test, we need to compare the results of the test to that of a standard test,
which is called the “gold standard” via using the blinding method.

7.1.1 Gold Standard (Reference Standard)

A “gold standard” method refers to the most reliable method to diagnose a disease,
which is also referred to as standard diagnosis. Application of gold standard can
distinguish whether the disease is truly present or not. The gold standard can be
biopsy followed by pathological examination, surgical discovery, bacteria cultiva-
tion, autopsy, special examination, and imaging diagnosis; it also can be an inte-
grated combination of several diagnostic criteria (such as Jones diagnosis standard,
etc.). The outcomes of long-term clinical follow-up obtained by applying the
affirming diagnostic methods were also used for the gold standard.

7.1.2 Study Subjects

The subjects of a screening test include the case group who has a specific disease and
controls who do not have the disease. They should be representative of the target
population. Therefore, the case group should include various types of the studied
disease: mild, moderate, or severe; early, middle, or late stage; typical or atypical;
with or without complication; treated or untreated, in order to make the result of the
study more representative and applicable to the general population. In contrast, the
control group should include individuals without the studied disease, but with other
illnesses, particularly those that are not easily distinguishable from the studied
disease. The testing of study subjects should be kept within the same research period
7 Screening and Diagnostic Tests 125

through either continuous sampling or proportional sampling, rather than by the

researchers’ choice. Otherwise, a selection bias may be present, which influences the
validity and reproducibility of the test.

7.1.3 Sample Size

The sample size is determined based on the following factors: sensitivity, specificity,
permissible error, and alpha level. The formula for sample size calculation is as
follows:

Z α 2 p ð 1 - pÞ
n= ð7:1Þ
δ2

n is the sample size of abnormal or normal subjects in the study.

Zα is the Z value for normal distribution of cumulative probability, which is equal
to α/2.
δ is admissible error, usually, it is set at a 0.05 ~ 0.10 level.
p is the estimation of sensitivity or specificity of the test. Sensitivity is used to
calculate the sample size of the case group, while specificity is used for the control
group. This formula requires the sensitivity or specificity approaching 50%. When
the sensitivity or specificity ≤20% or ≥80%, the corrected formula is needed:

2
57:3Z α
n= ð7:2Þ
sin - 1 δ= pð1 - pÞ

7.2 Evaluation of a Screening Test

When evaluating a new screening test for a disease, the gold standard for the disease
should be used simultaneously. The subjects will be divided into two groups based
on the test results: case group and control group (non-disease group). The results of
the gold standard and the screening test are then compared. The first step of this
comparison is to generate a two-by-two table and calculate several indexes.
As shown in Table 7.1, in cell a, the disease of interest is present, and the
screening test result is positive, a true-positive result. In cell d, the disease is absent,
and the screening test result is negative, a true-negative result. In both a and d cells,
the screening test result agrees with the actual status of the disease. Cell b represents
individuals without the disease who have a positive screening test result. Since these
test results incorrectly suggest that the disease is present, they are considered to be
false positives. Subjects in cell c have the disease but have negative screening test
126 F. Liu

Table 7.1 Comparison of the results of a screening test with the gold standard
Gold standard
Screening test Patients Controls Total
Positive True positive (a) False positive (b) a+b
Negative False negative (c) True negative (d ) c+d
Total a+c b+d a+b+c+d

results. These results are designated false negatives because they incorrectly suggest
that the disease is absent.

7.2.1 Validity of a Screening Test

The validity of a screening test is defined by its ability to correctly categorize

subjects who do or do not have a disease into corresponding groups. The compo-
nents of validity include sensitivity, specificity, Youden’s index, and likelihood
ratio.

7.2.1.1 Sensitivity

The sensitivity of a screening test is defined as the proportion of persons with the
disease in the screened population who are identified as ill by the test. Sensitivity is
calculated as follows:

a
SensitivityðSenÞ = × 100% ð7:3Þ
aþc

If someone with the disease is incorrectly called “negative,” it is a false-negative

result. The false-negative rate is complementary to sensitivity.

7.2.1.2 Specificity

Specificity of a test is defined as the proportion of disease-free people who are so

identified by the screening test. Specificity is calculated as follows:

d
SpecificityðSpeÞ = × 100% ð7:4Þ
dþb

If some people without a disease are incorrectly called “positive,” it is a false-

positive result. The rate of false-positive is complementary to the specificity.
7 Screening and Diagnostic Tests 127

7.2.1.3 Youden’s Index

Youden’s index (YI) is also called the accuracy index, which is frequently used to
evaluate the overall performance of a test. The formula of the Youden’s index is:

YI = Sen þ Spe - 1 ð7:5Þ

It ranges from 0 to 1. The greater the index is, the better the validity.

7.2.1.4 Likelihood Ratio

The likelihood ratio (LR) reflects the validity of screening test; it is an integrative
index that can reflect the sensitivity and specificity altogether, i.e., the ratio of true-
positive or false-negative rates in disease group to the false-positive or true-negative
rates in the group without the disease. Using the results of the screening tests, we can
calculate all the LR of the tests, which thus reflect the overall validity of a
screening test.
The positive likelihood ratio of a screening test is the ratio of true-positive rate to
false-positive rate, and negative likelihood ratio is a ratio of false-negative rate to
true-negative rate. The computation formulas for positive likelihood and negative
likelihood ratios are as follows:

a=ða þ cÞ Sen
LRþ = = ð7:6Þ
b=ðb þ dÞ 1 - Spe
c=ða þ cÞ 1 - Sen
LR - = = ð7:7Þ
d=ðb þ dÞ Spe

The likelihood ratio is more stable than sensitivity and specificity, and it is less
influenced by prevalence.
There is an example that would be helpful in understanding the calculation of
these indices.
Example Suppose, we perform a diabetes screening test in a cohort of 1000 people,
of whom 20 are diabetic patients and 980 are not. A test is available that can yield
either positive or negative results. We want to use this test to distinguish subjects
who have diabetes from those who do not. The results are shown in Table 7.2. How
do we evaluate the validity of the screening test?
These results showed that of the study population, 90% were positive in the
screening test, but the remaining 10% were not diagnosed. Among the individuals
without diabetes, 95% tested negative with the screening, and 5% were
misdiagnosed in the screening.
128 F. Liu

Table 7.2 The results of a screening test and the gold standard test for diabetes
Gold standard
Results of screening Have the disease Don’t have the disease Total
Positive 18 49 67
Negative 2 931 933
Total 20 980 1000
Sensitivity = (18/20) × 100% = 90%
Specificity = (931/980) × 100% = 95%
False-negative rate = (2/20) × 100% = 10%, or 1 – 90% = 10%
False-positive rate = (49/980) × 100% = 5%, or 1 – 95% = 5%
Youden’s index = 0.90 + 0.95 – 1 = 0.85
LR+ = 0.90/0.05 = 18.00
LR- = 0.10/0.95 = 0.11

7.2.2 Evaluation of the Reliability of a Test

Reliability or repeatability is an index that reflects the stability of the testing results,
i.e., if the results are replicable when the test is repeated. In a study, almost all
variations of measured data stem from the observer’s variation (intra-observer and
inter-observer variation), measuring instruments, reagents variation, and research
object’s biological variation (intra-subject variations), etc.

7.2.2.1 Coefficient of Variation

For a continuous variable, the variations of data are commonly measured with
standard deviation (SD) and coefficient of variation. The coefficient of variation
(CV) is obtained by dividing the SD by mean (percentage).

SD
CV = × 100% ð7:8Þ
X

7.2.2.2 Agreement Rate and Kappa Statistic

Agreement (consistency) rate is also called accuracy rate, which is defined as the
proportion of the combined true positive and true negative number of the total
population evaluated by a screening test, i.e., the percentage of the results of a
screening test that is in accordance with those of the gold standard method. Below is
the formula for calculating accuracy rate:

Agreement rate = ½ða þ dÞ=ða þ b þ c þ dÞ × 100% ð7:9Þ

7 Screening and Diagnostic Tests 129

Table 7.3 Kappa value judg- Kappa value Consistency strength

ment standard
<0 Poor
0 ~ 0.2 Weak
0.21 ~ 0.40 Light
0.41–0.60 Moderate
0.61 ~ 0.80 High
0.81 ~ 1.00 Strong

For counted variable, the observation coincidence rate or kappa statistic is used to
determine data reliability (repeatability or precision).
This is the calculation of kappa:

Percent
agreement
Percent
expected
agreement -
by
observed
chance
alone
Kappa = ð7:10Þ
Percent
agreement
expected
100% -
by
chance
alone

Kappa is an index that judges consistency in levels between different observers.

Landis and Koch suggested that kappa greater than 0.75 represents an excellent
agreement beyond chance, while a kappa less than 0.40 shows poor agreement, and a
kappa of 0.40 to 0.75 represents intermediate to good agreement (Table 7.3). Testing
for the statistical significance of kappa, please refer to the relevant book.

7.2.3 Predictive Value

Sensitivity and specificity are indicators of the accuracy of a test, which can be
considered the characteristics of a screening or diagnostic test itself. However, the
predictive value is affected by both the sensitivity and specificity of the test and the
prevalence of the disease in the population to be tested. There are positive predictive
value (PPV or PV+) and negative predictive value (NPV or PV–).
The PPV is defined as the probability of the persons having the disease when the
test is positive. The PPV is calculated as follows:
130 F. Liu

a
PPV = × 100% ð7:11Þ
aþb

The NPV is the percentage of the persons not having the disease when the test is
negative.

d
NPV = × 100% ð7:12Þ
cþd

Take the data in Table 7.2 as an example again for the calculation of predictive
values:

18
PPV = × 100% = 26:87%
18 þ 49
931
NPV = × 100% = 99:79%
2 þ 931

The PPV of 26.87% means that 67 individuals are positive in screening, but
among them, the number of real patients is 18, accounting for 26.87% of the total
positive results. The NPV of 99.79% indicates that 933 persons have negative test
results, and among them, the number of individuals “not having the disease” is
931, accounting for 99.79% of the total negative results.
Predictive value is affected by the prevalence of a disease in a specific population,
or by the pretest probability of the presence of a disease in an individual. We can use
the formula derived from Bayesian theorem of conditional probability to show the
relationships of predictive value, sensitivity, specificity, and prevalence.

Sensitivity × Prevalence
PPV = ð7:13Þ
Sensitivity × Prevalence þ ð1 - SpecificityÞ × ð1 - PrevalenceÞ
Specificity × ð1 - PrevalenceÞ
NPV = ð7:14Þ
ð1 - SensitivityÞ × Prevalence þ Specificity × ð1 - PrevalenceÞ

The more sensitive a test is, the higher will be its negative predictive value (the
more confident clinicians can be that a negative test result rules out the disease being
sought). Conversely, the more specific the test is, the better will be its positive
predictive value (the more confident clinicians can be that a positive test confirms or
rules in the diagnosis being sought). Because predictive value is also influenced by
prevalence, it is not independent of the setting in which the test is used.
As the numbers in Table 7.4 show, positive results even for a very specific test,
when applied to patients with a low likelihood of having the disease, will be largely
false positives. Similarly, negative results, even for a very sensitive test, when
applied to patients with a high chance of having the disease, are likely to be false
negatives. In summary, the interpretation of a positive or negative result of a
 7

Table 7.4 The screening results of diabetes in populations with different values of sensitivity, specificity, and prevalence
Gold standard
Prevalence (%) Sensitivity (%) Specificity (%) Screening results Patients Non-patients Total PPV (%) NPV (%)
Screening and Diagnostic Tests

50 50 50 + 250 250 500 50 50

- 250 250 500
Total 500 500 1000
20 50 50 + 100 400 500 20 80
- 100 400 500
Total 200 800 1000
20 90 50 + 180 400 580 31 95
- 20 400 420
Total 200 800 1000
20 50 90 + 100 80 180 56 88
- 100 720 820
Total 200 800 1000
131
132 F. Liu

screening or diagnostic test is dependent on the setting in which the test is carried
out, in particular, the estimated prevalence of the disease in the target population.

7.2.4 Determination of Cutoff Point for a Screening Test

Ideally, the sensitivity and specificity of a screening test both should be 100%. In
practice, when we plot the value of a screening test for a disease group and
non-disease group on the same graph, the distribution often overlaps, the test does
not separate normal from diseased with 100% accuracy. Figure 7.1 is the schematic
graph showing the distributions of test results for patients with and without the
disease. The area of overlap indicates where the test cannot distinguish normal and
abnormal. We need to determine a balance by an arbitrary cutoff point (indicated by
A and B) between normal and disease. The position of the cutoff point will
determine the number of true positives, false positives, false negatives, and true
negatives. If we want to increase sensitivity and include all true positives, we can use
A as a cutoff point, but by doing this, we increase the number of false positives,
which means decreased specificity. Likewise, if we want to increase specificity by
using B as a cutoff point, it will lead to decreased sensitivity.
We can also use the blood sugar data in Table 7.5 as an example to illustrate how
changes in the cutoff point will affect the sensitivity and specificity of a screening
test.
To make decisions on the appropriate cutoff point for a screening test, the
following principles need to be taken into consideration. For a proven serious
disease that can be cured if diagnosed early, a high sensitivity may be suggested.
If a false-positive result would detrimentally affect a patient both mentally and
physically, such as cancers, which may put a patient at risk of surgery and chemo-
therapy, a test with high specificity would be required. If both the sensitivity and
specificity are important, the junction point of curves might be used as the cutoff
point.

Fig. 7.1 Blood sugar level Number

distribution in normal
people and diabetes patients Overlap

Normal diabetes patients

A B Blood sugar
7 Screening and Diagnostic Tests 133

Table 7.5 The effects of cut- Blood sugar (mg dL-1) Sensitivity (%) Specificity (%)
off points of 2 h after-meal
80 100.0 1.2
blood sugar on sensitivity and
specificity of the 90 98.6 7.3
screening test 100 97.1 25.3
110 92.9 48.4
120 88.6 68.2
130 81.4 82.4
140 74.3 91.2
150 64.3 96.1
160 55.7 98.6
170 52.9 99.6
180 50.0 99.8
190 44.3 99.8

Fig. 7.2 The ROC curve of 100

blood sugar in the diabetes
diagnosis 80
Sensitivity(%)

60

40

20

0
0 20 40 60 80 100
1-Specificity(%)

7.2.4.1 ROC Curve

For continuous measurement data of a screening test, the cutoff point is determined
mostly by the receiver operator characteristic (ROC) curve. ROC curve is a graphical
plot of true positive rate (sensitivity, Y-axis) against the false negative rate (1 –
specificity, X-axis) for different cutoff point. A ROC curve could reflect the rela-
tionship between the sensitivity and the specificity of a test (Fig. 7.2). By conven-
tion, the point nearest to the top-left corner of the ROC curve is set for optimal cutoff
point.
As shown in Fig. 7.2 and Table 7.5, when sensitivity is 88% and specificity is
68%, the sum of the false positive and false negative rates is the minimum.
Accordingly, the blood sugar level of 120 mg dL-1 can be set as the optimal cutoff
point for diabetes screening in this population.
134 F. Liu

7.2.4.2 The Area under ROC Curve

ROC curves can also be used to compare clinical values of two or more screening
tests, thus helping clinicians choose the best screening test. The area under the ROC
curve is a measure of the test’s accuracy. The larger the area under the ROC curve,
the better the diagnostic test. The maximum value for the area under the ROC curve
is 1, which indicates a perfect test; an area of 0.5, on the other hand, represents a
worthless test.
We can use statistical software, such as MedCalc, SPSS, and SAS, to compute the
area under the ROC curve and compare the areas under ROC curve between two or
more screening tests (for details, please refer to related statistics books).

7.3 Improving the Efficiency of Screening

and Diagnostic Tests

In order to increase the sensitivity and specificity of a screening test, several methods
can be used, such as screening high-risk population or performing multiple tests.

7.3.1 Selecting Population with a High Prevalence

The predictive value of a test is influenced by the sensitivity, specificity, and

prevalence of a disease. When sensitivity and specificity are constant, it is influenced
mainly by the prevalence rate. Since morbidity has larger influence on the positive
predictive value, the latter would have very low value if a screening test is carried out
in a population with a low prevalence rate of the disease to be tested. However, if a
high-risk population is screened, the positive predictive value can be significantly
increased.

7.3.2 Use of Multiple Tests

A method combining two or more tests is called multiple tests. In general, multiple
tests can be carried out in two ways, simultaneous testing and sequential testing.

7.3.2.1 Simultaneous Testing

In simultaneous testing (parallel tests), the sample is evaluated with more than one
screening test simultaneously; a positive result of any test is considered evidence for
7 Screening and Diagnostic Tests 135

Table 7.6 The results of Test results

simultaneous and sequential
Multiple tests Test A Test B Diagnosis
testing
Simultaneous testing + - +
- + +
+ + +
- - -
Sequential testing + + +
+ - -
- + -
- - -

Table 7.7 An example of screening results using multiple tests (%)

Screening methods Sensitivity Specificity PPV NPV
Test A 80 60 33 92
Test B 90 90 69 97
Simultaneous testing (A and B) 98 54 35 99
Sequential testing (A and B) 72 96 82 93

the target disease. Simultaneous testing can improve sensitivity and negative pre-
dictive value, but lower the specificity and positive predictive value (Table 7.6).

7.3.2.2 Sequential Testing

Sequential testing (serial testing) means multiple screening tests are used in series,
the individual is considered to be positive if all the test results are positive but is
stopped when the previous test result is negative. Sequential testing increases
specificity and positive predictive value but decreases sensitivity and negative
predictive value.
Take the hypothetical example in Table 7.7 as an example, in which a population
is screened for hepatocellular carcinoma using ultrasonography and serum alpha-
fetoprotein (AFP) level. If two tests with 80% and 90% sensitivity, respectively,
were used simultaneously, the sensitivity of the simultaneous testing will be
increased up to 98%. However, there is a loss of specificity (decreased to 70%)
compared to each test alone. In sequential testing, there is a gain in specificity
(increased up to 96%), but a loss in sensitivity (down to 72%).
From the results above, we can summarize the regular pattern of sensitivity and
specificity in different multiple tests. How to make the decision to choose either
simultaneous or sequential testing is based on the actual situation.
136 F. Liu

7.4 Potential Bias in Screening Tests

There are three major sources of bias, which are specified to each screening test.

7.4.1 Volunteer Bias

The characteristics may be different between people who attend a screening and
those who do not, especially when those factors are directly related to the survival of
patients. Individuals with a higher risk of a disease are more likely to voluntarily join
a screening program, as they might be more health-conscious and with higher
compliance tend to have a better prognosis. For example, women with a significant
family history of breast cancer are more likely to join a mammography program than
those without it. This tendency is reflected by a higher rate of diagnosis in a series of
screening tests than what is truly reflective of the population. Likewise, the screened
people tend to have a larger percentage of adverse clinical outcomes than it would be
in the general population.
The most effective way to avoid volunteer bias is to recruit a pool of volunteers
and then assign them randomly to receive screening or not to receive it.

7.4.2 Lead-Time Bias

Lead time refers to the duration from early detection of disease (usually by screen-
ing) to the presentation of clinical symptoms and thus being diagnosed in the
standard way. Especially for chronic diseases, the cases of which progress slowly,
therefore patients with those diseases are more likely to be detected by screening and
likely to have increased survival time than unscreened cases. In fact, the screening
has no effect on the outcome of the disease; it only resulted in an earlier diagnosis of
the disease when compared to traditional diagnostic methods. To illustrate the lead-
time bias, we take a cancer screening test (shown below) as an example. As shown in
the illustration, the tumor is detected at different ages with or without the screening
test, but the patients die at the same age (Fig. 7.3), indicating that the overall survival
of patients is not altered by the screening test.
So, unless we have some idea of the actual lead-time, perhaps from previous
studies, we should not use survival time after diagnosis to evaluate a screening
program. Instead, we should consider the effects on longer-term age-specific mor-
bidity or mortality rates of the disease. The survival rates are therefore less likely to
reflect the true benefits of early treatment better.
7 Screening and Diagnostic Tests 137

Fig. 7.3 An illustration of lead-time associated with screening and cancer development process

7.4.3 Length-Time Bias

Many screening programs are implemented to detect cancers. Doctors and

researchers hypothesize that tumors with low growth rates have better outcomes
than more aggressive types. However, it is found that screening is more likely to
detect slower-growing, less deadly tumors due to their longer preclinical stages. In
other words, patients with concealed, less fatal cancers may not know the fact before
their death from other diseases, if without screening. This example results in the
“length-time” bias associated with screening tests, which gives the appearance that
screening can benefit patients and prolong their life span, when, in fact, the test
selectively detects those diseases that progress slowly, thus allowing the patient to
live longer.
Chapter 8
Bias

Lu Long

Key Points
• Bias refers to various influencing factors in epidemiological research, including
design, implementation, analysis, and inference, also known as systematic error.
Three major threats to validity are selection bias, information bias, and
confounding bias.
• Selection bias occurs when the characteristics of the subjects are different from
the source population, which leads to the deviation of the research results from the
real situation.
• Information bias, known as observational bias, refers to the inaccuracy or incom-
pleteness of the exposure or outcome information obtained during the implemen-
tation of the research, which results in the misclassification of the exposure or
disease of the research subjects and affects the validity of the results.
• Confounding bias is due to the existence of one or more external factors that mask
or exaggerate the link between research factors and diseases, thus partially or
wholly distorting the actual association.

8.1 Introduction of Bias

Bias, also known as systematic error, refers to various influencing factors in epide-
miological research, including design, conduct, analysis, and inference.
The existence of these influencing factors, including design errors, data acquisi-
tion distortion, incorrect analysis, or not logical inference, leading to the association
between exposure and outcome is misestimated, and this actual relationship is
systematically distorted, which leads to the wrong conclusion. Bias is an important
issue that affects the authenticity of the results. Thus, we must fully understand the

L. Long (✉)
West China School of Public Health, Sichuan University, Chengdu, China

© Zhengzhou University Press 2023 139

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_8
140 L. Long

source of bias and its causes and minimize the occurrence of bias in our studies to
ensure the authenticity of the study. There are two directions of bias, i.e., positive
bias and negative bias. Positive bias means that the measured value of the study
overestimates the true value, on the contrary, it is negative bias.
We generally classify bias as selection bias, information bias, and
confounding bias.

8.2 Selection Bias

8.2.1 Definition

Selection bias means that the characteristics of the selected subjects are different
from those of the unselected, which leads to a deviation of the research results from
the real situation.

8.2.2 Classification

8.2.2.1 Self-Selection Bias

Self-selection bias, or volunteer bias, is one source of selection bias. Self-selection

bias is one type of bias that results from individuals disproportionately selecting
themselves to join a group. For example, researchers selected soldiers from the
Smoky Atomic Test in Nevada to investigate the leukemia incidence. In this
study, 76% of the soldiers are members of the cohort with known outcomes, and
the remaining 24% were identified as the cohort without known outcomes. Those
who knew the outcomes, 82% were traced by the investigators, while others reached
out to surveyors. We ordinarily consider self-reported subjects as a threat to validity
because self-reporting may be related to the study results.
In the Smoky Atomic Test study, among the 62% (2% × 76%) of cohort
members, investigators traced four target cases and the 14% (18% × 76%) of cohort
members also traced four target cases who reported themselves. We assume the
leukemia incidence without known outcomes (24%) is similar to that of the subjects
traced by the investigators. In that case, we should expect that only (24%/
62%) × 4 = 1.5, meaning that about one or two cases occurred among this 24% of
the members without known outcomes, only a total of nine or ten cases in the entire
cohort. If instead, we suppose that the 24% without known outcomes had the same
incidence of leukemia as subjects with known outcomes. We would calculate that
8(24%/76%) = 2.5, meaning that about two or three cases occurred among this 24%,
in the entire cohort we will observe 10 or 11 cases. However, among the 24% + 14%
of the cohort, all cases were untraced among the self-reported, leaving no case
among those without known outcome. T. The total number of cases will be only
8 Bias 141

8 in the entire cohort. This example indicates that self-selection bias is a small but a
real problem in research.

8.2.2.2 Berksonian Bias

Berkson’s bias or Berksonian bias is also known as admission rate bias. It usually
occurs in a hospital-based case-control study because the selected case or controls
represent only a subset of patients with a disease rather than an unbiased sample of
the corresponding target population. Affected by medical conditions, residence,
socio-economy, education, and other factors, patients have specific selectivity to
hospitals, and hospitals also have a specific selectivity to patients, which results in
problems in sample representativeness and bias in a hospital-based case-control
study.
For example, hospital-based case-control study was used to explore the relation-
ship between birth control pills and thrombophlebitis. Cases were recruited from
people with thrombophlebitis in a hospital. And randomly selected as patients
without thrombophlebitis in a certain ward of the same hospital as a control group.
Suppose there are 5000 patients with thrombophlebitis and 5000 patients without
thrombophlebitis. Oral contraceptive accounts for 15% in each of them. It is
assumed that admission rates for these three conditions are relatively independent
(Table 8.1).
It can be calculated from Table 8.1 that the correlation of thrombophlebitis and
oral contraceptive, OR = (750 × 4250)/(4250 × 750) = 1.0, which indicates that
there is no correlation among oral contraceptive and thrombophlebitis.
Now assume the admission rate of case group was 25% while control group was
60%, and the admission rate of oral contraceptive was 40%. The composition of the
comparative study samples is shown in Table 8.2.
The admission rate of the 750 patients with thrombophlebitis and exposure to
contraceptive was 25%, So the number of thrombophlebitis hospitalizations were
750 × 25% = 187.5 ≈ 188; and 40% of the remaining were hospitalized due to
exposure to contraceptive, and the number of hospitalized patients was
(750–750 × 25%) × 40% = 225, and the total hospitalizations was 413.

Table 8.1 Exposure and dis- Oral contraceptive

ease distribution in the total
Group Yes No Total
population
Case 750 4250 5000
Control 750 4250 5000

Table 8.2 Distribution of Oral contraceptive

exposure and disease in the
Group Yes No Total
hospital
Case 413 1063 1476
Control 570 2550 3120
142 L. Long

The admission rate of the 4250 patients with thrombophlebitis rather than expo-
sure to contraceptive was 25%, So the number of cases group was
4250 × 25% = 1062.5 ≈ 1063.
The admission rate of the 750 patients without thrombophlebitis who were
exposed to contraceptives was 60%, so the hospitalizations were
750 × 60% = 450, and 40% of the remaining patients were hospitalized because
of exposure to contraceptives, the hospitalizations was
(750–750 × 60%) × 40% = 120, with a total hospitalization of 570.
The admission rate of the 4250 patients without thrombophlebitis and the oral
contraceptives was 60%, So the number of total hospitalizations was
4250 × 60% = 2550.
According to the above data, OR = (2250 × 413)/(570 × 1063) = 1.53, oral
contraceptive was positively correlated with thrombophlebitis.
There was no association between oral contraceptives and thrombophlebitis in the
total population, but a case-control study using hospital samples found a positive
correlation. The degree of the association was influenced by the admission rate,
which deviated from the true association in the population. This is Berksonian Bias.

8.2.2.3 Detection Signal Bias

Detection signal bias, known as unmasking bias, is also a common selection bias. If
the exposure factor to be studied has no turel causal relationship to the disease,
however, its presence may cause the subject to develop symptoms or sighs related to
the disease to be studied, leading to earlier or more frequent visits to the doctor,
which increases the detection rate of the disease and makes it more likely to be
included as a case in the study. Suppose these patients are taken as case groups in
case-control studies. In those cases, there will be systematic differences in certain
characteristics (such as exposure factors) between admitted patients and
non-admitted patients, leading to misestimating the true associations between expo-
sure factors and outcomes. For example, several studies found that oral estrogen was
associated with endometrial cancer and believed that oral estrogen was a risk factor
for endometrial cancer. However, many scholars later proposed that estrogens do not
cause cancer to occur, but only allow cancer to be diagnosed. Because estrogen can
stimulate the growth of the endometrium, making the uterus prone to bleeding. The
women who take estrogen are more likely to seek medical attention, this made early-
stage endometrial cancer patients easier to be identified. In contrast, while case-
control studies with such patients as case group led to an increased proportion of oral
androgens in endometrial cancer patients, thereby overestimating the association
between estrogen and endometrial cancer.
8 Bias 143

8.2.2.4 Neyman Bias

Neyman bias, called prevalence-incidence bias, was first described by Neyman in

1955 and occurred in the case-control study design. When carrying out case-control
studies, we can select three cases: cases-incident cases, prevalent cases, and death
cases. If all the admitted cases are survived cases, especially the cases with a long
disease course, may be related to the survival, but not to the onset of the disease. It
may, thus misestimate the etiological effect of these factors. On the other hand,
survivors of disease may change some of their existing exposure. When they are
investigated, they may mistake these changed exposure characteristics as their
disease conditions, resulting in errors in the correlation between these factors and
the disease.

8.2.2.5 Loss of Follow-Up

Cohort studies, clinical trials, and clinical prognosis studies generally require follow-
up of subjects. For the long observation period, the follow-up process cannot avoid
the absence of outcome events due to relocation of subjects, death due to other
reasons (competitive risk), or withdrawal from the study due to poor treatment
effects, adverse reactions, and other reasons. Loss of follow-up will affect the
representativeness of the research objects, thus affecting the authenticity of the
results. Therefore, this bias is called loss of follow-up bias.

8.2.3 Control

It is difficult to eliminate or correct its effects on the results once select bias occurs.
Therefore, scientific research design should be performed to reduce and avoid
such bias.

8.2.3.1 Scientific Research Design

In the research process, we(researchers) should clear the global and the sample
population and predict the various bias that may be generated in the sample selection
process based on the nature of the study. In the case-control study, we should avoid
selecting cases in a single hospital, and we can set up community control and
hospital control at the same time. Even if the cases can only be selected from the
hospital, they should also be randomly sampled in the different areas and different
levels of hospitals. In the cohort study, we can establish various controls, including
comparing incidence in exposed populations and all populations or compare inci-
dence in exposed populations and other unexposed populations, to reduce the effects
of selective bias.
144 L. Long

8.2.3.2 Develop Strict Inclusion and Exclusion Standards

In both observational research and experimental research, we must have developed a

strict, clear unified standard about inclusion and exclusion, including disease diag-
nostic criteria and exposure criteria, enabling the selected research object to better
represent the overall. After the exclusion standard determines the selection, it strictly
complies with the study’s implementation phase and cannot be changed casually.

8.2.3.3 Maximize Response Rates

Various measures should be taken to obtain the cooperation of the subjects as far as
possible, improve the response rate, reduce or prevent the occurrence of loss of
follow-up, and control the selection bias. During the study, we should increase the
subjects’ understanding of the significance of the study through various ways. When
the non-response rate or loss of follow-up rate is more than 10%, we should be
cautious in analyzing the research results. A random sampling survey should be
conducted on the non-responders or lost respondents if possible, and the results of
the sampling survey should be compared with those responders. If there is no
significant difference, it shows that the non-response or loss of follow-up has little
effect on the results; oppositely, we should explain appropriately. Strategies to
reduce loss to follow-up include: screening of willingness prior to registration,
detailed collection of participants’ contact information, using effective incentives,
and maintaining regular contact with participants. In addition, the sample size can be
appropriately increased to reduce the impact of the loss of follow-up or non-response
on the results after the corresponding sample size is calculated in the design stage.

8.2.3.4 Randomization Principle

Randomization can be divided into two different forms of random sampling and
random allocation. Random sampling means the opportunity of each target object
extracted into the study queue is equal, making the research sample representative,
avoiding bias due to the subjective, arbitrary choice of research objects; random
allocation is the equivalent opportunity for participants to be assigned to the exper-
imental group or control group without the effect of researchers and participants’
subjective wishes or unconscious objective reasons. The purpose of random distri-
bution is to make the non-research factors evenly distributed in each group and to
increase the transferability among groups.
8 Bias 145

8.3 Information Bias

8.3.1 Definition

Information bias, known as observational bias, refers to the inaccuracy or incom-

pleteness of the exposure or disease information obtained during the implementation
of the research, which results in the wrong classification of the exposure or outcome
of the research subjects and affects the authenticity of the results. Information bias
generally occurs when there are errors in the measurement, which is also known as
classification error or misclassification for discrete variables. Misclassification can
divide into differential misclassification and nondifferential misclassification. Com-
pared with nondifferential misclassification, differential misclassification has a
greater impact on study results. Due to the directions of differences in the
misclassification among groups, the effect value may be overestimated or
underestimated.

8.3.2 Classification

8.3.2.1 Differential Misclassification

Differential misclassification refers to classification errors that rely on the actual

values of other variables. The most common differential misclassification is recall
bias. Suppose an interview of congenital malformations in a case-control study, we
generally obtain the etiological information from the mother. We selected mothers
who have recently given birth to a deformed baby as a case, whereas mothers who
had recently given birth to an apparently healthy baby as a control. The mothers of
deformed infants are better able to recall exposures than mothers of healthy infants,
leading to a kind of differential misclassification, referred to as recall bias. Because
the birth of a deformed infant can stimulate the mother to recall all events that may
have played some role in the unfortunate outcome. The difference produced by this
recall bias is an apparent effect unrelated to any biological effect. Recall bias is likely
to arise in any case-control study that requires recall of past experiences. Klemetti
and Saxen [9] considered time as a critical indicator of recall accuracy.
When establishing or verifying a research hypothesis, if personal biased views are
reflected in the process of data collection, it will lead to interviewer bias. The
resulting inducement bias is also classified as interviewer bias if the researcher
intentionally induces the subject to provide the required information. In cohort
studies or experimental epidemiological studies, more detailed examination of
exposure or intervention group may be performed if the investigator has previously
assumed that the exposure or intervention is associated with the occurrence of
outcome. It leads to a misjudgment of the study results.
146 L. Long

Not all misclassification will exaggerate the association under study, but exam-
ples of the opposite can also be found. When investigating sensitive issues with the
subjects, they will deliberately minimize the information. For example, patients with
sexually transmitted diseases such as syphilis and gonorrhea may be reluctant to let
investigators know about their history of exposure to unprotected sex because of
stigma, and the resulting bias may underestimate the association between unpro-
tected sex and sexually transmitted diseases.

8.3.2.2 Nondifferential Misclassification

Classification error that is independent of other variables is called nondifferential

misclassification.
Presumably, bias due to independent nondifferential misclassification of expo-
sure or disease is predictable in the direction, i.e., toward the null. Some researchers
have used complex procedures to demonstrate that misclassification is
nondifferential, Unfortunately, decomposing continuous or categorical data into
fewer categories can transform non-differential errors into differential misclassifica-
tions even under blinding is accomplished or in cohort studies where disease out-
comes have not yet emerged. Non-differentially alone does not guarantee a bias
toward the null. Even if nondifferential misclassification is implemented, it may
come at the cost of increasing the total bias.
Both disease and exposure can occur nondifference misclassification. When the
proportion of subjects misclassified by disease does not depend on the subject’s
status with respect to other variables in the analysis, including exposure, it will occur
non-differential disease misclassification. Similarly, when the proportion of subjects
misclassified by exposure does not depend on subject status to other variables in the
analysis, including disease, it will occur nondifferential exposure misclassification.
We will give an example to illustrate how an independent nondifferential disease
misclassification with full specificity does not bias the risk ratio estimate but rather
biases the absolute magnitude of the risk difference estimate downward by a factor,
equal to the probability of false negatives. Suppose there is a cohort study in which
30 cases occur in 300 unexposed subjects and 60 cases occur among 200 exposed
subjects. The actual risk ratio is 3, and the actual risk difference is 0.20. Assumes no
false positives for disease detection, sensitivity is only 70% for both exposure
groups. The expected numbers of exposure cases detected will be 0.70 × 60 and
unexposed cases detected will be 0.70 × 30, which means that the expected risk ratio
is estimated to be ((0.70 × 60)/200)/((0.70 × 30)/300) = 3 and the expected risk
difference is estimated to be (0.70 × 60)/200 – (0.70 × 30)/300 = 0.14. Thus,
although disease misclassification did not bias the risk ratio but the expected risk
difference estimate was 0.14/0.20 of the actual risk difference.
The effects of nondifferential misclassification of exposure are similar to the
effect of nondifferential misclassification of disease. We hypothesized a cohort study
comparing the incidence of liver cancer in smokers with the incidence among
nonsmokers to explore nondifferential exposure misclassification. The incidence
8 Bias 147

rate was assumed to be 0.01% per year for nonsmokers, and 0.05% per year for
smokers. We surppose2/3 of the study population are smokers, but only 50% admit
this. This would then result in only 1/3 of subjects being identified as smokers with a
disease incidence of 0.05% per year. And the remaining 2/3 of the population is
made up of equal numbers of smokers and nonsmokers. Among those classified as
nonsmokers, their average incidence would be 0.03% per year rather than 0.01% per
year. The rate difference has been reduced by misclassification from 0.04% to
0. 02%, while the rate ratio has been reduced from 5 to 1.7.
These examples present how a nondifferential misclassification of a dichotomous
exposure will produce a bias toward the null value (no relationship) if the
misclassification is unrelated to other errors. The association will be completely
obliterated and the direction of association will be reversed by bias, if the
misclassification is severe enough (although the reversal will only occur if the
classification method is worse than randomly classifying people as “exposed” or
“unexposed”).
We cannot dismiss a study simply because of the presence of substantial
non-differential misclassification of exposure, it is incorrect. This is because the
implications may be greater if there is no misclassification, which provides a
probability of misclassification that applies uniformly to all subjects. Thus, the
impact of nondifferential misclassification depends heavily on whether the study is
considered positive or negative. Emphasizing measurement rather than qualitative
descriptions of study results can reduce the likelihood of misinterpretation, but even
so, it is important to keep in mind the direction and possible magnitude of bias.

8.3.3 Control

Whether differential misclassification or nondifferential misclassification is mainly

due to problems in measurement or data collection methods, resulting in errors in
acquired data. Therefore, we mainly adopt the following methods to control
information bias.

8.3.3.1 Material Collection

The main purpose of the survey design is to standardize the tables in the study, which
is crucial for internal validity, so that valid, reliable, and complete data could be
collected efficiently. In addition, pretesting survey instrument in populations similar
to the study population can identify flaws in the survey design and instruments
before full data collection begins. We’d better use the blinding method to collect data
to avoid the influence of subjective psychology of research objects and investigators
on the survey results.
148 L. Long

8.3.3.2 Objective Research Indicators

Try to use objective indicators or quantitative indicators to avoid information bias,

such as applying laboratory examination results and consulting the medical records
or health examination records of the subjects as the source of investigation infor-
mation. Suppose it is necessary to collect data by means of inquiry. In that case, we
should adopt closed questions and answers as far as possible to prevent the occur-
rence of report bias and measurer bias. For questionnaires concerning lifestyle and
privacy, the respondents should be informed in advance that all responses are
confidential and will be properly kept appropriately.

8.3.3.3 Investigation Skills

The investigative skills of investigators are particularly important when obtaining

information, especially the research that requires the participation of investigators.
We can improve their investigation level by training investigators and formulating
investigators’ manuals to reduce information bias.

8.4 Confounding Bias

8.4.1 Definition

Confounding bias is due to one or more external factors that mask or exaggerate the
link between research factors and diseases, thus partially or entirely distorting the
actual relationship between them. Confounding is produced by confounders (expo-
sures, interventions, treatments, etc.).
Taking Stark and Mantel’s study on neonatal Down’s syndrome as an example.
Population monitoring data indicated that Down’s syndrome was associated with
birth order. Assume the incidence of Down’s syndrome in the first-born child was
0.06% while in the fifth-born child was 0.17%. The risk of Down’s syndrome
increased with the increase of birth sequence, which seemed birth order to be a
risk factor for Down’s syndrome. However, we should consider maternal age at
delivery as a confounder, closely related to birth order and Down’s syndrome risk.
Further study found that the incidence of Down’s syndrome in children delivered by
pregnant women younger than 20 years old was 0.02%, and gradually increased with
the age of delivery, and the incidence of Down’s syndrome in children delivered by
pregnant women over 40 years old was as high as 0.85%. The study indicates that the
maternal age at childbirth is related to the occurrence of the disease. Therefore, it is
suggested that the association of birth sequence with Down’s syndrome risk may be
influenced by the confounding factor of maternal age at birth.
8 Bias 149

In this part, we briefly refer to confounding bias, but we will discuss confounding
and how to control it in the next part.

8.4.2 Confounding

When estimating the effect of an exposure on exposed individuals, Confounding can

occur when the exposed and nonexposed subgroups of the population have different
background disease risks. These subgroups can have different disease risks even if
they are not exposed to any of the effects in both subpopulations. More generally,
confounding occurs when the exposed and unexposed groups are not fully compa-
rable or “exchangeable” in terms of exposure response, i.e., the exposed and
unexposed groups may exhibit different risks even if both experience the same
level of exposure. In general, a factor associated with both the exposure and the
outcome could be a confounder. The following are three necessary but not sufficient
conditions to be a confounder of the effect of the exposer.
First, confounder must be predictors of the disease without the exposure under
study. Confounders are not necessarily the genuine cause of the disease under study.
However, they are only “predictive” within the level of exposure apart from casual
relations. For example, race, age, gender, etc., may be considered as potential
confounders. Thus, one almost always sees adjustments made for age and sex.
Second, the confounder must be related to the study exposure. For example,
confounder should be related to exposures in the control group in case-control study.
If the factor is not associated with exposure in the control group, an association
between cases may still occur because both the study factor and the potential
confounder are risk factors for disease, but this is a consequence of those effects
and therefore does not cause confounding.
Third, confounder cannot be intermediate variables between exposure and out-
come. In other words, confounders cannot be intermediates in the causal pathway
between exposure and disease, or a condition caused by the outcome. To do
otherwise would introduce a serious bias. Hypothetically, in a study of overweight
and the risk of cardiovascular disease, it would be inappropriate to control for
diabetes as confounder if diabetes was a consequence of being overweight and is
also a part of the causal chain leading to overweight and cardiovascular disease. On
the other hand, assuming diabetes is studied directly as a primary interest, over-
weight would be regarded as a potential confounder if it also involved exposure to
other risk factors for cardiovascular disease.
We discussed the misclassification of disease and exposure in information bias.
Here we need to refer to the misclassification of confounders. The ability to control
confounding in the analysis will be hindered If a confounder is misclassified.
Although independent nondifferential misclassification of exposure or disease usu-
ally causes the study results to be biased in the direction of the null hypothesis,
independent nondifferential misclassification of a confounder usually reduces the
degree of control for confounding, which may lead to bias in either direction. For this
150 L. Long

reason, misclassification of confounder can be a serious concern. If the confounding

is robust and the exposure–disease relationship is weak or zero, misclassification of
the confounder can yield highly misleading results, even if such misclassification is
independent and non-differential.

8.4.3 Control

In the study design and analysis, confounding bias can be controlled by adjusting for
all confounders or a sufficient subset of them at the same time. There are usually
three methods to control for bias during the design stage.

8.4.3.1 Random Allocation

The first method is randomization, where participants are randomly assigned to

exposure categories (applicable to experiments only). Ideally, we can create study
cohorts with the equal incidence rate and eliminate the potential for confounding.
But it must be practically and ethically feasible to assign exposure subjects. If just a
few factors determine incidence, and the investigation personnel are aware of these
factors, the ideal plan might call for exposure assignment that would result in the
identical, balanced distributions of these disease causes in each group. Nonetheless,
in studies of human disease, there are always immeasurable causes of disease that
cannot be forced to be balanced amongst treatment groups. Randomization is one
approach that permits one to probabilistically limit the confounding of unmeasured
factors and to quantitatively account for the potential residual confounding arising
from these unmeasured factors. However, this is usually only one alternative that
may be beneficial for potential exposures. For instance, it is impractical and
unethical to conduct randomized trials of the health effects of smoking, and therefore
randomized trials may fail to prevent all confounding.

8.4.3.2 Restrict

The second control method is restriction, i.e., limiting the conditions of the study
subject to a narrow range of values of the potential confounders. If a variable is
prohibited from changing, it will not generate confounding if it is prohibited from
varying. The restriction is a promising way to prevent or at least reduce confounding
by known factors, it is both extremely effective and inexpensive. However, the
advantages of restricting the study must be balanced against the disadvantages of
reducing the study population when potential subjects are less plentiful. This
approach has several conceptual and computational advantages, but may severely
reduce the number of study subjects available and ultimately limit the extrapolation
of results.
8 Bias 151

8.4.3.3 Matching

The third control method is matching, where study subjects are matched on the basis
of potential confounders. Matching may be done by subject to subject, called
individual matching, or for groups to groups, called frequency matching. Individual
matching refers to the selection of one or more reference subjects with equal
matching factor values to those of the index subject, whereas frequency matching
refers to the selection of a whole stratum of reference subjects with similar matching-
factor values to that of a stratum of index subjects. Individual matching would
prevent age-gender-race confounding in cohort studies but is seldom done because
it is very labor-intensive. In addition, matching does not completely eliminate
confounding but does facilitate its control in case-control studies because matching
for strong confounder will usually improve the precision of effect estimates. We
have to discuss the concept of overmatching, which is often occurred in matched
studies. In case-control studies, matching may be less accurate if the match factor
related to exposure is only a weak risk factor for the disease of interest. When the
number of matching factors exceeds 3, finding a suitable control becomes increas-
ingly difficult.

8.4.3.4 Data Analysis

The above three control methods are usually implemented during the design phase.
The analysis phase can also employ a number of methods to control for confounding
bias. In the most straightforward situation, controlling for confounding in the
analysis includes stratifying the data according to the level of confounders and
calculating an effect estimate that summarizes the association between the strata of
confounding factors. In a stratified analysis, it is usually not possible to control for
more than two or three confounders at the same time, because finer stratification
often results in many strata that contain no exposed or non-exposed individuals.
Such strata are noninformative; therefore, a stratification that is too fine is a waste of
information. In addition, we can use multi-factor analysis and standardized analysis
to control confounding bias.
Chapter 9
Cause of Disease and Causal Inference

Li Ye

Key Points
• In epidemiology, cause and causal inference are used to explore the etiology of
risk factors for diseases at a population level.
• A causal model is a concise and conceptual graphic that describes the relationship
between cause and disease.
• Most epidemiologic study designs can be used for evaluating causation. The
strength of these designs to evaluate causation varies.
• Mill’s canons represent logical strategies for inferring a causal relationship.
• Hill’s criteria are a list of guidelines to distinguish causal and noncausal associ-
ations; these criteria have been widely used and are the best known criteria for
assessing causal inference.

9.1 Introduction

One of the major focuses of epidemiology is to find the causes of diseases or events.
Understanding the causes of diseases is important not only for correct diagnoses and
treatments but also for effective prevention and control strategies. Therefore, cause
of disease and causal inference—the process by which we identify the cause of
disease are essential in both clinical and preventive medicine. In epidemiology,
cause and causal inference are used to explore the etiology of or risk factors for
diseases as well as their impact on the development of disease at the population level,
which can provide unique insights into the etiology of the disease and lead to a
population-level understanding of the disease. This chapter describes the epidemi-
ologic concept of cause and the approaches to causal inference.

L. Ye (✉)
School of Public Health, Guangxi Medical University, Nanning, China

© Zhengzhou University Press 2023 153

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_9
154 L. Ye

9.2 Cause of Disease in Epidemiology

9.2.1 The Concept of Cause in Epidemiology and its

Development History

There are many definitions of cause in epidemiology. The following widely accepted
definition is from Abraham Lilienfeld: a causal relationship would be recognized to
exist whenever evidence indicates that the factors from part of the complex of
circumstances that increase the probability of the occurrence of disease and that a
diminution of one or more of these factors decrease the frequency of that disease.
Another definition from Kenneth Rothman and Greenland [10] is also widely
accepted due to its simplicity and clarity: an event, condition or characteristic, or a
combination of these factors that play an essential role in producing an occurrence of
the disease. Cause is an important concept in epidemiology. There are many other
synonyms to describe cause, including causal agency, determinant, risk factor,
exposure, etiological factor, etiological agent, etc. In epidemiology, cause is often
referred to as a risk factor, which means the factor that increases the risk of disease.
The cause of disease has long been explored. The most ancient idealism attributed
the occurrence of diseases to the god or devil. In the fourth century BC, Hippocrates,
the father of medicine, considered that diseases occurred because of the imbalance of
“four body humors.” In the fifth century, Chinese ancestors founded a materialistic
view of the cause, and they proposed diseases were from the imbalance of “Yin-
Yang” or “Five elements (wood, earth, water, metal, fire).”
In the later nineteenth century, at the height of the era of germ theory, Robert
Koch, the founder of modern bacteriology, proposed Koch’s postulates, which
include four generalized principles for determining whether a specific microorgan-
ism causes a specific disease. Koch’s postulates contributed greatly to the formation
of the concept of cause in epidemiology because identifying the microorganism was
equivalent to identifying the cause of the disease. In fact, the discipline of epidemi-
ology as well as the concept of the cause of disease originated from etiology and
epidemic studies on communicable diseases, among which the germ theory and
Koch’s postulates represent landmark achievements.
However, Koch’s postulates cannot explain the causes for most diseases, espe-
cially noncommunicable diseases, which have replaced communicable diseases as
the main threat to human health since the middle of the twentieth century. More
recently, the epidemiologic studies have focused more on the probability and
multicausality of the occurrence of diseases, which finally led to the formation of a
modern concept of cause, as described at the beginning of this subsection.
9 Cause of Disease and Causal Inference 155

9.2.2 Classification of Cause

In modern epidemiology, the concept of cause actually means “multicausality”.

Most diseases, whether communicable or noncommunicable, have more than one
cause. Since the definition of cause, either by Lilienfeld or by Rothman, means that
any “factor” or any “event, condition or characteristic” plays a role in affecting the
occurrence of the disease, the cause in epidemiology covers a wide range of factors,
including individual genetics, physiological influences, environmental influences,
social structure, etc. According to the source of these factors, we can divide the
causes into two general categories: host factors and environmental factors
(Table 9.1). Host factors refer to various characteristics that are related to people
or an individual, such as genetics, immune status, age, sex, race, and behavior.
(Table 9.1). Environmental factors mainly include biological, physical, chemical,
and social factors (Table 9.1).

Table 9.1 Classification of the cause of disease

Factor (cause) Description
Host factors
1. Genetics Chromosomal disorder, single gene disorder, polygenetic disorders, etc.
2. Immune status It involves in the occurrence of most diseases, both communicable and
noncommunicable
3. Age and sex People of different age or sex may be susceptible to different diseases
4. Race Occurrence of disease has difference in race
5. Personality Temperament, psychological status, psychiatric status, etc. may have effects
on the occurrence or progression of diseases
6. Behavior Bad behaviors or habits such as smoking, drinking, poor diet, lack of
exercise, unsafety sexual behaviors, drug abuse, noncompliance with traffic
laws, etc.
Environmental
factors
1. Biological Pathogenic microorganisms (bacteria, viruses, rickettsiae, mycoplasmas,
chlamydiae, spirochetes, actinomyces, etc.); parasites (protozoa, worms,
insects, etc.); venomous animals and poisonous plants (snakes, ergot,
mushrooms, etc.)
2. Physical Temperature, humidity, altitude, noise, light, vibration, radiation, dust, fire,
etc.
3. Chemical Pollution, agricultural chemicals, food additives, microelement, heavy
metal, etc.
4. Social Social system, socioeconomic level, war, disaster, education, religion, living
condition, lifestyle, occupation, family relationship, etc.
156 L. Ye

9.2.3 Causation Models

A causal model is a concise and conceptual graphics that describes the relationship
between cause and disease. During the development of etiology, different causal
models were proposed based on contemporary understanding of the diseases in
different historical periods, which made a great contribution to the formation of
the modern concept of cause. Casual models can be used to illuminate the associ-
ation between cause and disease as well as the relationship between multiple causes
of a disease and to provide direction or clues to find a new cause. Essentially, the aim
of causal models is to find causes and elucidate the dominant cause and ultimately
determine the best prevention or intervention strategy. The most representative
causal models are the triangle model, the wheel model, the chain of causation
model, and the web of causation model.

9.2.3.1 Triangle Model

In 1954, John Gordon summarized the knowledge about the epidemiologic etiology
of diseases at that time and put forward an epidemiologic triangle model (epidemi-
ologic triad) to describe the relationships between multifactorial causes and a
disease, especially communicable disease. The model considers that host factors
(age, sex, race, genetic profile, immune status, etc.), agents (biologic pathogens,
chemical, physical, nutritional agents, etc.), and environmental factors (temperature,
humidity, crowding, housing, water, food, radiation, pollution, noise, etc.) are the
troika of a disease. These three aspects are indispensable for the occurrence of a
disease and have an equal role in the occurrence of disease. Hence, the relationships
can be described as an equilateral triangle (Fig. 9.1). The three kinds of factors
interact and restrict each other, and thus, a dynamic balance exists that makes the
occurrence of disease in a stable state. Once the balance is disturbed, the occurrence
of disease increases or decreases. The triangle model is helpful even today for
finding the cause of communicable diseases and controlling the epidemic. However,
it is basically unsuitable for the description of noncommunicable diseases.

Fig. 9.1 Epidemiologic

triangle model
9 Cause of Disease and Causal Inference 157

9.2.3.2 Wheel Model

In the middle of the twentieth century, noncommunicable diseases became the main
threat to humankind. However, there is no obvious or absolute agent for most
noncommunicable diseases, as a pathogen agent for communicable diseases. It is
very difficult to describe the relationship between cause and noncommunicable
disease using a triangle model. In 1985, Mansner and Kramer proposed the wheel
model based on the triangle model. In this model, host factors play the core role in
the occurrence of disease and are located in the center of the wheel, with genetic
factors as the core of the center (Fig. 9.2). The host center is surrounded by three
kinds of environmental factors, including biological, social, and physical environ-
mental factors. The main difference between the wheel model and the triangle model
is that the wheel model considers that different factors have different importance for
the occurrence of disease. Therefore, the area sizes of the center (host factors) and
surrounding parts in the wheel (biological, social, and physical factors, respectively)
can be adjusted to reflect the importance of different factors. The wheel model
emphasizes the core role of host factors as well as the influencing effects of
environmental factors. It is considered to be better than the triangle model and
suitable for both communicable and noncommunicable diseases. However, the
wheel model came from etiology knowledge in the 1980s and could not truly reflect
the complex interactions between various factors. It is still limited for many
noncommunicable diseases, especially chronic diseases.

9.2.3.3 Chain of Causation Model

In multicausality theory, there are multiple causes of communicable and

noncommunicable diseases. The multiple causes or risk factors can always be
displayed in the form of a chain. Some factors are direct or proximal or most
immediate causes, and others are indirect or distal causes. Some factors are

Fig. 9.2 Causation wheel

model
158 L. Ye

Fig. 9.3 Chain of causation (diabetes)

independent causes; however, others are dependent causes that interact with other
causes. To interpret the association of different causes and final disease as well as the
complex relationship among multiple causes, a model of “chain of causation” was
proposed to describe the causes in the form of a chain. For example, an accelerated
life tempo can lead to an unhealthy diet or less exercise and then obesity, followed by
insulin resistance, which often results from obesity and further results in elevated
blood glucose. Finally, diabetes occurs when blood levels of glucose become
chronically elevated (Fig. 9.3). It is worth mentioning that removing any factor in
the chain can block the whole chain and thus prevent the occurrence of disease
(Fig. 9.3).

9.2.3.4 Web of Causation Model

In some cases, a disease has several and interrelated chains of causation. The
causation chains of the disease link and interplay with each other, and thus constitute
a complex network. MacMahon proposed the web of causation model to describe the
complex relationships between causes and disease as well as the interlacing chains of
causation.
For example, the causation of liver cancer can be described as a network or a web.
The four chains of causes of liver cancer consist of biological factors, physical and
chemical factors, behavioral factors, and genetic factors (hereditary susceptibility).
Multiple factors of the four chains also interact with each other and form a network,
thus ultimately leading to the occurrence of liver cancer (Fig. 9.4).

9.2.4 Sufficient Cause and Necessary Cause

Modern epidemiology considers that the relationships between cause and effect are
multiple and complex. A given disease can be caused by many factors; however, a
single factor is not enough to cause the disease, as joint action from other causes is
necessary. Obviously, the role and importance of different factors are different in the
occurrence of a disease. From the logical view of cause and effect, all effects have
sufficient and necessary conditions; thus, cause in epidemiology can also be divided
into sufficient and necessary cause.
In 1976, Kenneth Rothman used a sufficient-component causal model (Fig. 9.5)
to explain the complex relationship between cause and effect. Rothman proposed
that a sufficient cause is a factor or a combination of several factors that will
9 Cause of Disease and Causal Inference 159

Fig. 9.4 Web of causation (liver cancer)

Fig. 9.5 Sufficient cause

and necessary cause

Sufficient cause I Sufficient cause II Sufficient cause III

inevitably cause disease. A component cause is a factor that contributes to the

occurrence of disease but is not sufficient to cause disease on its own. A necessary
cause is any agent that is required for the occurrence of disease (for example, cholera
bacillus for cholera occurring); without necessary cause, the disease will not occur.
For instance, the three sufficient causes (I, II, III) shown in Fig. 9.5, comprise
4 component causes. In this figure, there are three sufficient causes (I, II, III),
and A, B, C, D, E, F, G, H, and I are component factors. Because A is present in
all three sufficient causes, it is a necessary cause.
The sufficient-component causal model interprets two paradoxes of the causation
theory in epidemiology. First, why does a given disease occur without a specific
cause? For example, alcohol abuse is the cause of cirrhosis; however, individuals
who never drink may also develop cirrhosis. The possible reason is that cirrhosis
develops through other sufficient causes, such as hepatitis B virus infection
(Fig. 9.4). Second, why does a disease fail to occur in the presence of a specific
cause? For example, smoking causes lung cancer; however, many smokers never
develop lung cancer in their lifetime. This phenomenon can be explained by the fact
that smoking is not a sufficient cause of lung cancer. In reality, most identified causes
for noncommunicable chronic diseases are neither necessary nor sufficient. For
example, hypertension is neither a necessary cause nor a sufficient cause for cardio-
vascular disease. Nevertheless, every component cause is necessary for sufficient
cause that contains it. Removal of any component cause is equal to the removal of
160 L. Ye

the sufficient cause that contains this component cause, which is an important
strategy for the prevention of a disease.

9.3 Epidemiologic Methods of Causation

9.3.1 Epidemiologic Study Designs for Causation

Etiological studies in epidemiology usually contain several common steps. The first
step is to find the influencing factors that are associated with a disease or to develop a
cause-and-effect hypothesis, usually by descriptive or analytical epidemiologic
studies; the second step is to test the hypothesis often using analytical studies; and
the last step is to verify the hypothesis, usually by experimental epidemiologic
studies.
Most epidemiologic study designs can be used for evaluating or establishing
causation. However, the strength of these designs to evaluate causation is different.
Table 9.2 outlines the relative strength of the different study designs in establishing
causation. These study designs have been introduced in prior chapters, and their use
in providing evidence for causation will be described as follows.

9.3.1.1 Descriptive Studies

The start of causation is to develop a cause-and-effect hypothesis. Descriptive

studies are always used to generate hypotheses. Descriptive studies mainly include
case reports, case series, cross-sectional studies, and ecological studies. Case reports
and case series are useful for developing a hypothesis based on analysis of the
characteristics of patients or case groups. Cross-sectional studies are always used to
describe the distribution of disease in different populations, and the pattern or trend
of disease occurrence over time or by geographic area, which can provide the clues
regarding influencing factors. Ecologic studies explore the association of influencing
factors and disease at the population or region level, which can also provide clues for
influencing factors that cannot be measured at the individual level, for example, air

Table 9.2 The strength of evidence for causation by different epidemiologic study designs
Type of study design Strength of evidence in causation
Randomized controlled trials Strong
Nonrandomized controlled studies Moderate
Cohort studies Moderate
Case-control studies Moderate
Cross-sectional studies Weak
Ecologic studies Weak
Case reports Weak
9 Cause of Disease and Causal Inference 161

pollution. Generally, the strength of descriptive studies to evaluate causation is weak

compared with analytical studies or experimental studies due to a lack of evidence on
the time sequence of events. Of all descriptive studies, the weakest for causation is
case reports because they have neither defined populations nor comparison groups.
Nevertheless, when causal relationships have already been established, well-
designed descriptive studies, especially cross-sectional studies with multiple time
points or time series studies, can be very useful to quantify the effects of cause.

9.3.1.2 Analytical Studies (Case-Control Studies, Cohort Studies)

Analytical studies, mainly including case-control studies and cohort studies, are
more reliable methods to form a hypothesis than descriptive studies. Analytical
studies can also be used for hypothesis testing. Case-control studies, which are
mainly used to confirm the association between factors and disease, compare the
exposure levels between the case group and the control group. Because the research
direction of case-control studies is from effects (diseases) to causes (factors), this
study design is vulnerable to various biases. Cohort studies are either prospective or
retrospective, and they can test hypotheses more effectively by comparing the
incidence rates of exposure groups with control groups, and directly calculating
the relative risk (RR) of factors in the temporal order of cause and effect. Well-
conducted cohort studies are a better design for causation than case-control studies
because the former can minimize various biases, including selection, information,
and confounding biases.

9.3.1.3 Experimental Studies

Experimental studies include clinical trials, field trials, and community trials. Clin-
ical trials are most frequently conducted among patients, with the aim of evaluating
the efficacy of a new treatment or medicine. Therefore, clinical trials are known as a
robust and reliable method to test or verify hypotheses, especially clinical random-
ized controlled trials (RCTs), which are considered the gold standard to evaluate a
new treatment or medicine and the most rigorous method for hypothesis testing.
Nevertheless, RCTs are subjected to many constraints, such as ethical issues, strict
inclusion criteria, parallel control, and strict randomization, which greatly limit the
feasibility of RCTs in causation studies. Quasi-experiments that lack parallel control
or randomized assignment are also used in epidemiologic etiology, with less strength
to evaluate causation. Other experimental studies, including field and community
trials, are seldom used to study causation. Field trials mainly involve people who are
disease-free, with the aim of preventing the occurrence of diseases. Community trials
are conducted at the level of the community instead of the individual level. There-
fore, although experimental studies have strong strength to test and verify hypoth-
eses, most of the causative evidence so far has not come from this study design but
comes from observational studies such as descriptive and analytical studies. For
162 L. Ye

example, most of evidence about the effects of smoking on health comes from case-
control and cohort studies.

9.3.2 Mill’s Canons-the Logical Basis of Causation

The causa models mentioned in Sect. 9.2.3 of this chapter are mainly used to
describe the relationships between the factors and diseases or between different
risk factors. They cannot be used as a method to find causes or test causation.
Mill’s canons were proposed by philosopher John Stuart Mill in 1843, intended to
illuminate a causal relationship between a circumstance and a phenomenon, which
provides the logical basis of causation studies. In epidemiology, Mill’s canons
provide certain guidance for causation, especially the development of hypotheses.
The canons with minor adjustments constitute five methods for the induction of
hypotheses.

9.3.2.1 Method of Agreement

This method means that factors in common among different instances of a disease
are perhaps the cause or a necessary part of the cause of the given disease. In other
words, if two or more instances of a disease under investigation have only one factor
in common, which is likely to be the cause of the given disease. For example, one
school had an outbreak of diarrhea. It was found that all the students with diarrhea
had consumed soy milk in the same canteen in the morning, so soy milk may be the
cause of diarrhea.
However, in actual conditions, it is difficult to obtain only “one common factor”.
There may be a few other factors shared by patients with the same disease, but most
of the factors are not the cause of the disease. In the example mentioned above, most
students with diarrhea may have also eaten another food in common in the same
cafeteria that morning. Therefore, the method of agreement in this example is
actually not sure that soy milk may be the cause of outbreaks of diarrhea. Generally,
a hypothesis cannot be formed by one method.

9.3.2.2 Method of Difference

If some instances in which the disease occurs, and other instances in which the
disease does not occur, they have all other factors in common except one existing
only in the former. That one may be the cause or a necessary part of the cause of the
disease. The method of agreement concerns whether patients share certain common
factors. The method of difference compares the differences in certain characteristics
between patients and nonpatients. For example, if one student had diarrhea while the
9 Cause of Disease and Causal Inference 163

other did not, the only different food that two students consumed in the canteen was
soy milk, and the soy milk may be the cause of diarrhea.
Similar to the situation of the method of agreement, in the method of difference,
the assumption that “all other factors are the same” between patients and nonpatients
is hard to make in practice. In the above example, the two students may be different
from each other in many other aspects. The method of difference cannot exclude
other factors and hypothesize that only soy milk is the cause of diarrhea.

9.3.2.3 Joint Methods of Agreement and Difference

This method is actually a combination of the method of agreement and the method of
difference, however, it is not a simple combination but alternately contains
multiround use of two methods. Briefly, if two or more instances in which the
disease occurs have only one factor in common, while two or more instances in
which the disease does not occur have nothing in common except the absence of the
factor commonly existing in the former instances, then, the factor may be the cause,
or a necessary part of the cause, of the disease. Generally, the joint methods of
agreement and difference are much more likely to find a risk factor than the method
of agreement or difference alone. The main reason is that joint methods of agreement
and difference essentially introduce contrast in the investigation, which greatly
increases the logicality.
We return to the example of soy milk and diarrhea. If all of the students with
diarrhea had consumed soy milk, the students without diarrhea must have not
consumed soy milk in the same canteen. This is the joint method to indicate that
soy milk was likely to be the cause of diarrhea.

9.3.2.4 Method of Concomitant Variations

According to the method of concomitant variations, whatever one event varies in any
manner whenever another event varies in some particular manner. The former event
is either a cause or an effect of the latter; in other words, these two events are
connected through cause-and-effect association. In essence, method of concomitant
variations emphasizes dose-dependent relationships. When there is a dose-
dependent relationship between two events, cause-and-effect associations are more
likely exist.
In the above examples, if students who consumed more soy milk had more severe
diarrhea, there was a dose-dependent relationship, namely, concomitant variations,
so the probability of the soy milk as the cause of diarrhea was higher.
164 L. Ye

9.3.2.5 Method of Residue

Suppose a disease is caused by many factors; when you remove the previously
known factors as well as the instances of disease caused by those factors, the residue
of factors may be the cause for the remaining instances of disease. For example, in
1972, a number of dermatitis cases occurred in Shanghai, China. Possible factors,
including industrial waste gas, plant pollen, blood-sucking arthropods, and poison-
ous moths, as well as dermatitis cases caused by these factors were excluded. The
residual factor, Euproctis similis, emerged. Therefore, researchers suspected that this
outbreak of dermatitis was caused by Euproctis similis, and finally confirmed this
hypothesis.
Although Mill’s canons are considered as a logical basis for causation studies and
still have certain guidance significance for current epidemiologic etiologies, they
have great limitations in actual practice in causation studies. Generally, Mill’s
canons are suitable to find both sufficient and necessary causes, such as acute
infectious agents and to judge strong causal associations. However, for most dis-
eases, especially noncommunicable chronic diseases, the risk factors are almost all
nonsufficient and nonnecessary causes, and one disease always has multiple suffi-
cient causes. In these cases, the Mill’s canons are not suitable for effectively
assessing the causal association. At most, the canons just play a role in the formation
of a hypothesis.

9.4 Causal Inference

Causal inference is the term used for the process by which we identify the cause of
disease. In other words, it is used to determine whether the observed association is
causal. In essence, the relationship between cause and disease is a kind of cause-and-
effect association in philosophy. Various risks or exposure factors are the cause, and
diseases are the effect. The term association is another important concept in
epidemiology.

9.4.1 Association Vs. Causation

Two events, the suspected cause and the effect, obviously must be associated if they
are to be determined as causally related. However, not all associations are causal,
namely, cause-and-effect associations. Various other associations, including chance
association, spurious association, and noncausal association, which are caused by
various reasons such as random error, bias, or confounding, should be excluded
before a causal association is assessed. Figure 9.6 outlines various associations
caused by different reasons.
9 Cause of Disease and Causal Inference 165

Fig. 9.6 Association and causation

9.4.1.1 Chance Association

First, we need to judge whether an association between two events, for example, an
exposure and an outcome, is a statistically significant association rather than a
chance association due to random error, e.g., sampling error and random measure-
ment error (Fig. 9.6). In epidemiology, the strength of association can be expressed
as rate ratios, odds ratios, or attributable risks, which are introduced in prior chapters.
The exclusion of chance association is based on statistical comparison of these
indicators between the exposed group (or case group) and the control group.
When the P value was less than 0.05, we considered that there was a statistical
significant difference, namely, a statistically significant association. The A value less
than 0.05 is defined as a low-probability event, which means that the result is reliable
in at over 95% probability, and there is less than a 5% chance that the result is caused
by random error. Because random error cannot be avoided and exists in all study
designs as well as each step of a study, well-designed and well-conducted studies are
essentially important to effectively reduce chance association caused by random
error, despite almost no study being perfect in either design or conduct practice.

9.4.1.2 Spurious Association

If the association is statistically significant and the probability of random error is

very limited, then we could evaluate whether the association is spurious, which
generally comes from nonrandom systematic errors known as selection or informa-
tion bias (Fig. 9.6). Spurious association means that the association truly exists, but
is not true due to selection or information bias. Thus, it is also called as false
association. Selection bias is generally caused by the difference in exposure- or
outcome-related characteristics between subjects selected for study and those not
166 L. Ye

selected or between the exposed (or case) group and the control group. For example,
in a case-control study on birth defects, the case group of newborns with birth
defects, while the control group contained consists of those without birth defects.
The collection of exposure information was primarily based on the memory of
mothers of the newborns. In information collection, the mothers of newborns with
birth defects, who were stimulated by adverse pregnancy were able to recall various
exposures during pregnancy in detail, such as taking over-the-counter drugs, fever,
or cold. However, mothers in the control group were less likely to make an effort to
recall the details and did not respond carefully to the relevant exposure events,
because no adverse pregnancy occurred. Therefore, the results obtained may be
influenced by recall bias. The association between potential exposure factors and
neonatal birth defects may be overestimated, and it may be a false association.

9.4.1.3 Noncausal Association

Even though a true association exists, it is still necessary to know whether the
association occurs indirectly by another extraneous factor (called a confounding
factor), which always leads to a noncausal association. If confounding was not
found, a noncausal association could be excluded and a causal association may
exist (Fig. 9.6). Confounding bias is caused by an extraneous factor (confounding
factor) that is closely related to both exposure and outcome but not an intermediate
link in the causal chain of the exposure and the outcome. Confounding bias can lead
to underestimation or overestimation of the association between the exposure and the
outcome. For example, investigation may find an association between smoking and
alcoholic liver, obviously, which is not reasonable in biological plausibility. The link
between smoking and alcoholic liver is a noncausal association because alcohol
drinking consumption is a confounding factor, which that is often associated with
smoking and directly related to alcoholic liver. Confounding occurs commonly in
epidemiologic studies. However, it can be well controlled through careful designs
such as matching, restriction, and randomization or through analyses such as stan-
dardization, stratification, and multivariate analysis.

9.4.1.4 Causal Association

After excluding chance association, spurious association, and noncausal association

caused by random error, selection/information bias, and confounding bias, the
association between the exposure and the outcome is likely to be causal, and still
needs to be further assessed by various judgments. We call this process causal
inference (Fig. 9.6). Among various judgments, the best known and widely used is
Hill’s criteria. The details of Hill’s criteria are introduced in the next subsection.
9 Cause of Disease and Causal Inference 167

9.4.2 Evaluating Causal Association—Hill’s Criteria

The existing association between two events or two variables after exclusion of
chance, spurious, and noncausal associations is just probably to be a cause-and-
effect association, which is required for further judgment based on totality of
evidence. Judgment of causal association is neither simple nor straightforward,
and various sets of guidelines have been proposed for the judgment. In 1965, the
British statistician Sir Austin Bradford Hill proposed a list of nine guidelines to
evaluate causal association, which has been widely used and is certainly the best
known set of criteria for the considerations of causation, sometimes with modifica-
tions (Table 9.3).

9.4.2.1 Temporal Relationship

In causal inference, temporal relationship is essential: the cause must precede the
effect; or, in other words, an exposure to cause a disease must precede the develop-
ment of the disease. Of all Hill’s guidelines, this is an absolute requirement.
Different study designs have different strengths to provide evidence of temporal
relationships. Cohort and experimental studies have obvious temporal relationships
because they are performed prospectively. However, in cross-sectional studies,
difficulty may arise in judging temporal relationships because the proposed cause
and effect are measured at the same time point. In case-control studies, sometimes it
is assumed that one event precedes another without actually establishing the order,
and in other cases, it may be difficult to determine which the first is. Nevertheless,
there are some strategies to find evidence supporting temporal relationships. For
example, when the cause is an exposure that can be divided into different levels, it is
essential that a sufficiently high level should be reached before the disease occurs.
Repeated measurement of exposure at multiple time points or in different locations

Table 9.3 Hill’s criteria for causation

Criteria Comments
Temporality The cause precedes the effect (essential criterion)
Strength The strength of association between the cause and effect (odds ratio, relative
risk)
Dose-response Increased exposure is associated with increased effect
Consistency Similar results are shown in other studies
Biologic The association is consistent with biologic mechanism
Plausibility
Reversibility Removal or reduction of exposure is followed by decreased effect
Specificity One cause leads to one effect, and vice versa
Analogy Exposure and effect are similar to those in a well-established causal
association
Experiment Evidence from animal, intervention or mechanism studies
168 L. Ye

may also strengthen the evidence of temporal relationships. Although temporal

relationships are necessary for causal inference, an existing temporal sequence
alone is weak evidence for causation. Many things occur before an event: however,
they have no relationship with the event. For example, someone may sneeze in
Beijing city, and 30 minutes later, there’s a heavy rain fall in Nanjing city. Obvi-
ously, there is no causal link between these two events.

9.4.2.2 Strength of Association

The stronger an association is, which is usually expressed by the relative risks (odds
ratio, OR; relative risk, RR), the more likely a causal association is. A strong
association less likely comes from either bias or confounding. Thus, the 13.70-fold
higher risk of lung cancer among male smokers compared with nonsmokers is much
stronger evidence than the finding that smoking is related to coronary heart disease,
for which RR is only 2.00. What is a “weak’ or “strong” association? There is no
universal standard, but epidemiologists generally consider a relative risk (OR, RR)
greater than 2.0 (or less than 0.5) to be moderately strong and a risk greater than 5.0
(or less than 0.2) to be strong. Nevertheless, a weak association does not mean that it
can be overlooked for causal inference. Sometimes the strength of an association
may depend on the prevalence of other possible causes. For example, the relation-
ship between diet and coronary heart disease is a cause-and-effect association;
however, the diets in populations are rather homogeneous, although greater variation
may be observed among different individuals or in different stages of one person. In
addition, a weak association, when combined with other guidelines, for example,
consistently observed in different designs or in different settings, may also provide
stronger evidence than a strong association that is only found in one or two studies.

9.4.2.3 Dose-Response Relationship

When changes in the level of possible cause are associated with corresponding
changes in the incidence or prevalence of the disease, a dose-response relationship
exists. Generally, the presence of a dose-response relationship in unbiased studies is
considered strong evidence for causation. However, the absence of a dose-response
relationship does not mean that the association is noncausal, because not all causal
associations exhibit a dose-response relationship. For instance, there may be a
“threshold” effect in which any exposure above a certain level will lead to disease.
In addition, although a dose-response relationship is a strong evidence for causation,
it cannot exclude confounding factors.
9 Cause of Disease and Causal Inference 169

9.4.2.4 Consistency

Consistency means that, when several studies are conducted at different times in
different settings or among different patient groups, the same or similar results are
derived. Consistency is a kind of evidence strengthened for causal inference because
the possibility that all different studies make the same “mistake” is minimized.
However, a lack of consistency does not exclude a causal association. Different
results may come from variation in study design or quality or different exposure
levels and other conditions that may affect the impact of a causal factor on the effect.

9.4.2.5 Biologic Plausibility

A causal association generally should have biological rationality. The association

between cause and effect is consistent with the current biologic knowledge and often
enhances convincing causal inference. However, the lack of biological plausibility
does not deny a causal association, which may simply reflect a lack of scientific
knowledge or evidence. Increasing knowledge of biological mechanisms may sup-
port this association in the future. In other words, biologic plausibility, when present,
enhances evidence for causation; when absent, other evidence for causation should
be sought.

9.4.2.6 Reversibility

When the removal of a factor that is likely to be a cause of disease results in a

decreased risk of disease, there is a greater possibility that the association is causal.
An example is that people giving up smoking decreases their risk of lung cancer
compared with people who continue to smoke. Reversible associations are strong but
not infallible evidence for causation because they cannot exclude confounding
factors, which can also conceivably account for a reversible association.

9.4.2.7 Specificity

Specificity refers to the strict corresponding relationship between a cause and a

disease: that is, a certain factor can only cause one certain disease, and vice versa, the
disease is just caused by a certain factor. This guideline is currently only applicable
for some acute communicable or genetic diseases because for most diseases, either
communicable or noncommunicable, there are many risk factors for the same effect
or many diseases come from one factor. Therefore, specificity is considered the
weakest evidence of all the guidelines for causation.
170 L. Ye

9.4.2.8 Analogy

Analogy is sometimes used in causal inference. Suppose there is a well-established

cause-and-effect association: for instance, factor A leads to effect B. If a similar
association is observed between factor C and effect D, which are also similar to
factor A and effect B, respectively, we can consider that factor C is likely to be the
cause of effect D. In general, analogy is weak evidence for causation.

9.4.2.9 Experimental Evidence

Experimental data, from studies in animals or other organisms, from intervention

studies in humans, or from mechanistic studies, may also provide evidence
supporting causal associations. In medicine, evidence from a well-conducted exper-
imental clinical trial is always considered the strongest evidence for causation.
However, in epidemiology, the results from a single or few experiments are gener-
ally not considered to be convincing or strong evidence for causation.
Among the nine guidelines described above, temporal relationships, and strength
of association are necessary conditions to judge causality. That means that if there is
a causal association, temporal relationships and statistically significant associations
must exist; otherwise, causal associations can be denied. The other seven guidelines
belong to unnecessary conditions, which are just general criteria for causal inference.
A lack of any one or even all seven guidelines does not preclude causal association.
Moreover, it is worth mentioning that all nine guidelines are not sufficient conditions
for causation. Thus, even though a relationship between two events satisfies all nine
guidelines, we cannot absolutely draw a conclusion that the relationship is causal.
Causal inference is always tentative, and judgment must be made on the basis of the
available evidence. Although there are no completely reliable criteria for determin-
ing whether an association is causal or not, Hill’s criteria are widely accepted and
have been applied in practice. Nevertheless, Hill’s criteria are essentially guidelines
for causal inference but not a “gold standard” to judge a causal association.

9.4.3 An Example of Causal Inference Using Hill’s Criteria

Primary hepatocellular carcinoma (HCC) is one of the most common malignancies

worldwide. Many researchers have investigated the causes of HCC and indicated
that alcohol abuse (habitual heavy drinking) is likely to be a risk factor for HCC. The
Causal inference was performed and summarized as follows.
9 Cause of Disease and Causal Inference 171

9.4.3.1 Temporality of Association

Cohort studies have indicated that habitual heavy drinking always precedes the
occurrence of HCC. Some cases of HCC had several years or even several decades
of heavy drinking history.

9.4.3.2 Strength of Association

Many case-control studies have indicated that the risk (OR) of HCC among habitual
heavy drinkers is 2–4-fold greater than that among nonhabitual drinkers or non-
drinkers. Furthermore, a few cohort studies found that the relative risk (RR) of heavy
drinking was 2–5-fold higher than that of nondrinking groups.

9.4.3.3 Consistency

The association between habitual heavy drinking and HCC has been repetitively
studied in many different countries by different study designs at different times. The
results from different studies consistently indicated that heavy drinking is associated
with the occurrence of HCC.

9.4.3.4 Dose-Response Relationship

Previous studies have found that more alcohol consumption and a longer heavy
drinking history result in a higher incidence of HCC. This is an obvious dose-
response relationship.

9.4.3.5 Biologic Plausibility

The relationship between alcohol abuse and HCC is consistent with the current
understanding of alcohol metabolism in the liver. Alcohol is metabolized in the liver,
and its metabolism produces free radicals that cause lipid peroxidation, damage
mitochondria in liver cells, and lead to alcoholic liver injury.

9.4.3.6 Experimental Evidence

A number of animal studies have shown a similar relationship between alcohol

administration and hepatic injury.
Taken together, the evidence is strong for the conclusion that habitual heavy
drinking is a risk factor for HCC.
Chapter 10
Disease Prevention and Surveillance

Chunhua Song

Key Points
• The three levels of prevention of disease: (1) primary prevention: known as
causation prevention, two complementary strategies are used: high-risk strategy
and population-based strategy; (2) secondary prevention: it refers to early detec-
tion, early diagnosis, and early treatment; (3) tertiary prevention: known as
clinical prevention or disease management.
• With the development of healthy connotations, the goal of public health is no
longer just to prevent disease but also to actively maintain and promote health.
The main strategies for achieving this include health protection and promotion.
• Public health surveillance is a persistent, continuous, and systematic process of
information collection about health events and health problems.
• In addition to active surveillance and passive surveillance, routine reporting, and
sentinel surveillance, the commonly used surveillance methods and analytical
techniques are also included.

10.1 Prevention Strategies and Measures

10.1.1 Strategy and Implementation for Prevention

A strategy is a basic principle guiding the overall work under specific circumstances.
It focuses on the overall situation and considers the issue from a macro perspective.
The measures include the application of actual methods, steps, and plans that are
used to achieve the desired objectives. The two parts are closely related. Only by
taking the correct prevention strategies and under the guidance of reasonable

C. Song (✉)
College of Public Health, Zhengzhou University, Zhengzhou, China
e-mail: [email protected]

© Zhengzhou University Press 2023 173

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_10
174 C. Song

measures, can we achieve the desired preventive effect efficiently. However, without
considering the feasibility of the measures, the strategy will not be implemented, and
it is impossible to achieve the desired objectives. The effect of measures is often
small without the guidance of strategies. Therefore, in order to minimize the
influence and achieve the maximum effect, we have to consider the measures and
strategy at the same time, which is the meaning of the strategy.
The successful eradication of smallpox is one of the best example that demon-
strates the importance of the correct use of prevention strategy and measure. In 1796,
Jenner found that Chickenpox vaccines could effectively prevent smallpox, and the
rest of the world had used the same approach. However, in 1960, smallpox
reoccurred and persisted in some countries and regions. There were still deficiencies
in the strategy for preventing smallpox, therefore, it is necessary to modify and
improve them. In addition to the vaccination rate, monitoring work should be carried
out in order to detect, isolate, treat, and report the new cases, and the people who
were in close contact with the cases should acquire artificial immunization (ring
vaccination), which thoroughly controlled the spread of the disease. This method can
not only save manpower and resources, but also achieve the same effect. Our
ultimate goal is to eliminate smallpox worldwide, so we are not just emphasizing
on large-scale vaccination, but also enhancing the monitoring process. This example
shows that the strategies and measures to prevent disease complement each other. On
May 8, 1980, WHO announced that smallpox had been eliminated in the world, and
the strategy was modified again. As a result, countries around the world can stop
vaccinating against smallpox and pay more attention to the risks to smallpox
researchers. Especially, the detection of suspected smallpox cases strengthened
laboratory testing and reservation of sufficient vaccines to meet any emergency
needs.

10.1.2 Disease Prevention

10.1.2.1 The Definition of Disease Prevention

Disease prevention is a series of activities that can reduce the occurrence of disease
(or injury) and disability and prohibit or delay its development. The main purpose of
prevention is to maintain high quality of life for patients by eliminating disease
(or injury) and to minimize the impact of disease (or injury) and disability. If it is
hard to do so, we must delay the occurrence of the disease or delay the development
of disease and disability.

10.1.2.2 The Development of Disease

The natural history of disease can be divided into four stages, including the stage of
susceptibility, the stage of subclinical disease, the stage of clinical disease, and the
10 Disease Prevention and Surveillance 175

stage of recovery. For infectious diseases, the stage of subclinical disease commonly
refers to the incubation period in which pathogen invades the body before the
emergence of clinical symptoms. In chronic non-communicable diseases, the induc-
tion period indicates the time from exposure to the etiology factor to the onset of the
disease, and the latency period indicates the time from the exposure factors to the
occurrence of the disease manifestation. In this definition, the latency period con-
tains the induction period. In another definition, the latency period refers to the time
from the onset of the disease to the presentation of the disease, when the latency
period follows after the induction period and the two periods do not overlap.

10.1.2.3 The Three Levels of Prevention of Disease

Primary Prevention

Primary prevention, also known as causation prevention, refers to taking measures

against etiology or risk factors to reduce the level of harmful exposure, enhance the
ability of the individual to combat harmful exposure, prevent the occurrence of a
disease (or injury), or at least delay the occurrence of the disease when exposed to
the disease (or injury). Primary prevention is a fundamental measure to eliminate a
disease (or injury).
A variety of measures are needed for primary prevention, such as preventing
harmful exposure in the environment, improving the body’s resistance (e.g., immu-
nization), or protecting the individual from harmful exposures, such as the elderly
with osteoporosis wearing hip protection.
① High-risk strategy: high-risk strategy is based on clinical medicine aiming to
achieve the first-level prevention strategy. High-risk strategy refers to taking
targeted measures to reduce the level of risk exposure and its risk of future
disease for small group of individuals with a risk of high incidence in the future.
For example, adults are regularly assessed for risk factors for cardiovascular
disease and appropriate measures are taken for those high-risk individuals, such
as stopping smoking, controlling salt intake, eating more fruits and vegetables,
and so on.
The limited availability of medical information means that healthcare is a
system of limited supply and that there is a need to give priority to groups that are
most likely to benefit or may benefit most. The use of resources by high-risk
strategies may be more cost-effective. However, most lifestyles, such as eating,
smoking, and exercising, are largely influenced and limited by the code of
conduct and the behavior of the people around us in our society. The high-risk
strategy, in essence, requires that the minorities behave differently; therefore, this
undoubtedly limits the effectiveness of this strategy. If the vast majority of cases
of a disease occur in a small group of easily identifiable populations, high-risk
strategies are enough to control the disease and interventions for this group are
effective and affordable. However, when the most fundamental cause of the
176 C. Song

problem, also known as the cause mentioned above, reaches the entire population,
treating only those patients and the most vulnerable individuals, the tip of the
iceberg, is a palliative solution.
② The population strategy: targeting the entire population is based on public
health thinking to achieve the first-level prevention strategy. The population-
wide strategy does not need to determine that individuals are at high or low risk of
future disease; only by eliminating harmful exposures, especially those that are
unobservable or uncontrolled by individuals or determinants of deleterious expo-
sures in the population, namely reasons for the etiology, taking measures to
reduce the level of harmful exposure throughout the population, ultimately
reducing the overall burden of disease in the population.
Both the high-risk strategy and the population strategy have their respective
strengths and weaknesses. In resolving many problems, the two strategies comple-
ment each other and work together.

Secondary Prevention

Secondary prevention refers to early detection, early diagnosis, and early treatment.
The secondary prevention aims at early stages of the disease that symptoms and
signs have not yet appeared or are difficult to detect. It aims at having a greater
chance of achieving cure by early detection, a timely and appropriate treatment, or at
least slowing down the development process of disease and reducing the need for
more complex treatment if the disease cannot be cured.
Diseases can be found early via screening, case finding, regular physical exam-
ination, and so on. For example, fecal occult blood tests and colorectal screening for
colorectal cancer are performed in adults over 50 years of age. In addition, there is
periodic screening for HIV in high-risk populations and other routine physical
examinations.
At present, most of the etiology of chronic diseases has not yet been established,
so full and effective primary prevention cannot be realized. However, the occurrence
of chronic diseases is due to the long-term effects of pathogenic/etiologic factors, so
it is feasible to do early diagnosis and early treatment.

Tertiary Prevention

Tertiary prevention, also known as clinical prevention or disease management, is

implemented after the symptoms and signs of the disease are manifested. At the early
stage of the disease, appropriate treatments are used to relieve symptoms, prohibit
further deterioration of the disease, reduce the occurrence and recurrence of acute
events, and prevent complications and disability. At the late stage of the disease, the
greatest recovery treatments are applied in order to restore body and social function
and improve the quality of life and prolong life. Tertiary prevention mainly consists
10 Disease Prevention and Surveillance 177

of symptomatic treatment and rehabilitation treatment. The prevention is aimed at

reducing the burden of disease and disability on individuals, families, and society.
Symptomatic treatment can mitigate the adverse effects of diseases with fewer
symptoms and prevent the occurrence of complications and disability. Rehabilitation
treatment for loss of labor or disability can promote physical and mental rehabilita-
tion, recovery of labor, and make sure they are able to create economic and social
value.

10.1.3 Health Protection and Promotion

With the development of healthy connotations, the goal of public health is no longer
just to prevent disease but also to actively maintain and promote health. The main
strategies for achieving this include health protection and promotion.

10.1.3.1 Health Protection

Concept of Health Protection

Health protection is to take targeted measures to protect individuals or people from

the external environment of harmful substances (such as biological, physical, and
chemical harmful substances) on human health threat. Health care covers a wide
range of health-related areas: infectious diseases, occupational health, sanitation,
radiation hygiene, food hygiene, school health, medicines, medical devices and
cosmetics safety, accidental injuries, and emergency public health emergency equip-
ment and treatment.

Health Protection Measures

Health protection measures include medical measures, such as immunization and

preventive medication as well as environmental engineering measures, economic
measures, legal measures, etc. Some scholars, of course, refer to health protection as
the latter. Many health protection measures are not available to individuals, and the
non-medical measures can be implemented on their own, requiring the joint efforts
of government and society.
① Eliminate harmful substances in the external environment or control them to the
levels that do not adversely affect human health. Such as pasteurization and other
processes on raw milk disinfection; the construction industry uses non-hazardous
or less hazardous building materials and uses construction techniques, construc-
tion equipment, and tools that do not produce or produce less dust. Frequent hand
washing is one of the measures of personal hygiene and interference control.
178 C. Song

② Provide a barrier for individuals. For example, the cab or operation room of
construction machinery is closed and isolated, and the air inlet is fitted with a filter
device; use of personal protective equipment such as protective clothes, protec-
tive gloves, and protective glasses.
③ Enhance the individual’s ability to fight harmful substances or take measures to
prevent the onset or reduce the symptoms of the disease when exposed, such as
vaccination, immune serum, or immunoglobulin. Rabies or animals suspected of
carrying rabies virus bites generally require 24 h after exposure for the first
vaccination. If the bite is in the upper limbs or head, or injury is heavier, anti-
rabies virus serum or specific immunoglobulin should also be injected for passive
immunization. When people in health care, public security, and other personnel
are inadvertently exposed to human immunodeficiency virus (HIV) due to occu-
pational reasons, the short-term anti-retroviral treatment can be used to reduce the
possibility of HIV infection, i.e., preventive medication.

10.1.3.2 Health Education

Health education is used to help individuals and groups master health care knowl-
edge and to establish a healthy concept through information dissemination and
behavior intervention. Under the premise of the access to information and enhanced
awareness, voluntary adoption of educational activities and processes that are
conducive to healthy behavior and lifestyles.
Health education pays more attention to the process of internalizing the objects of
education, highlighting the voluntary nature of individuals to change their behavior.
Health education can play a role in tertiary prevention. For example, to inform the
public of the common symptoms of TB and encourage timely treatment in the event
of suspicious symptoms. For TB patients, it is important to provide them with
information on the treatment management, standardize the benefits of treatment
under the policy of national free treatment, and improve compliance with standard
treatment. With the popularization of health education concept in the world, a lot of
health education practice shows that behavior change is a long-term and complicated
process. Though some education methods are simple, they have an effect on people’s
awareness and skills, and then change their way of life. However, environmental
constraints and lack of policy may hinder the adoption of healthy behavioral
intentions.

10.1.3.3 Health Management

Health management is a process of comprehensive supervision of the health-risk

factors of individuals or groups; the aim is to get the maximum health for the purpose
of minimizing the input. Unlike general health education, health management is
based on individual health status evaluation, i.e., according to individual risk factors,
individual guidance, dynamic tracking of risk factors, and timely intervention. At
10 Disease Prevention and Surveillance 179

present, health management is mainly used for the prevention of chronic

non-communicable diseases such as hypertension, hyperlipidemia, coronary heart
disease, stroke, diabetes, obesity, osteoporosis, and cancer.
Health management is the main trend of medical development today. It integrates
medical science, management science, and information science and focuses on the
concept, connotation, and evaluation of the standard of health, health risk factors
monitoring and controlling, health intervention methods and means, health manage-
ment service model and implementation path, health information technology and
standards, etc. The main contents of health management services include: health
examination, health monitoring, health risk assessment and intervention, health
education and counseling services and medical Internet services, chronic disease
risk screening, and tracking management.

10.1.3.4 Health Promotion

Health promotion is the process of enhancing people’s ability to control health and
improve their health. It is a comprehensive social and political process that includes
not only health education that directly strengthens individual behavior and life skills,
making people aware of how to stay healthy; but also through policies, legislation,
economic means, and other forms of environmental engineering to improve social,
economic, and environmental conditions to reduce their social impact on the public
and individual health, thereby creating a socially supportive environment that pro-
motes the maintenance and improvement of health. The WHO indicates that health
promotion mainly involves five domains: (1) establishing policies that are beneficial
to health; (2) changing the direction of health services; (3) improving individuals’
and populations’ health knowledge and skills; (4) creating a good physical and
natural environment; (5) developing communities’ abilities to promote health.

10.1.3.5 Global Health Strategies and Practice

The development of national health strategy has evolution strategies. In 1948, the
WHO put forward that health is a fundamental right of mankind. At the 30th World
Health Assembly in 1977, WHO members unanimously adopted a global
strategic goal: “Health for all by the year 2000.” This state of health allows
individuals to live a productive life in both social and economic terms. This goal
does not mean that the medical staff provides medical services for all diseases or
nobody is sick or develops a disease. Its connotation lies in: (1) Whether at home,
school or in other units, people can stay healthy during the period of life and work;
(2) People will use more effective methods to prevent disease, reduce the pain caused
by unavoidable diseases and disability, and grow healthier, grow old, and finally die
happily; (3) All health resources are equally distributed among all members of
society; (4) All individuals and families, through their own active participation,
enjoy basic health care in an affordable manner; (5) People will realize that they are
180 C. Song

capable of getting rid of the burden of disease that can be avoided, shaping
themselves and their families’ lives, keeping healthy, and knowing that disease is
not inevitable.
In 1978, the Declaration of Alma-Ata was adopted by the WHO and the United
Nations Children’s Fund at the International Conference on Primary Health Care,
organized in Almaty, which reaffirmed WHO’s 1948 definition of health. It also
formally put forward the concept of “primary health care,” and clearly stated that
primary health care is the key and fundamental way to achieve the goal of “health
care for all in 2000.” The meeting was recognized as a milestone in modern public
health.
Primary health care refers to basic health care services that can be affordable in
countries and regions, and the methods and techniques used in these services are
viable, scientifically sound, and socially acceptable. Each individual and family in
the community can access these basic services. The primary health care system
should be able to provide appropriate health promotion, disease prevention, diagno-
sis, treatment, and rehabilitation services for major health problems in different
countries and regions. The system is designed to achieve early protection and
prevention. In this process, the health sector also covers all aspects of the relevant
departments, countries, and social development, especially agriculture, livestock,
food, industry, education, housing, public works, transportation, and other sectors,
and requires all of these inter-departmental collaborations.
The basic content of primary health care can vary from country to country, but the
following eight items should be included at least: (1) carrying out publicity and
education for current important health problems and prevention and control
methods; (2) promoting food supply and proper nutrition; (3) promoting the provi-
sion of sufficient safe drinking water and basic sanitation facilities; (4) providing
women and children health care, including family planning; (5) vaccination against
major infectious diseases; (6) preventing and controlling endemic diseases; (7) pro-
viding proper treatment and management of common diseases and injuries; (8) Pro-
viding essential medicine. The 34th World Health Assembly in 1981 adopted the
“Global Strategy of health care for all in 2000.” It also complements “Use all
possible methods to prevent and control noncommunicable diseases and promote
mental health through the impact of lifestyles, controlled substances and psychoso-
cial environments.”
In 1986, WHO adopted the Ottawa Charter for Health Promotion at the First
International Conference on Health Promotion, which for the first time fully
expounded the concept of “health promotion,” principles, and the future develop-
ment direction, directly promoted the health state; the city’s strategy is put forward
and practices in the global widespread.
In 2000, the United Nations Millennium Summit adopted the United Nations
Millennium Declaration, which places health at the heart of the global agenda as the
Millennium Development Goals (MDG), the eradication of extreme poverty and
hunger, universal primary education, the promotion of gender equality, and empow-
erment of women, the reduction of child mortality, maternal health, AIDS, malaria,
and other diseases’ morbidity, in order to ensure environmental sustainability and
10 Disease Prevention and Surveillance 181

global cooperation for development. Countries have raised their health plan in .., line
with national strategies.
The 2030 Agenda for Sustainable Development adopted by all United Nations
Member States in 2015 provides a shared blueprint for peace and prosperity for
people and the planet now and into the future. As its core are the 17 Sustainable
Development Goals (SDGs), which constitute an urgent call to action for all devel-
oped and developing countries in a global partnership; SDGs represent the successor
to the MDGs, which aim to protect people's health and welfare worldwide.

10.2 Public Health Monitoring

10.2.1 Introduction of Public Health Surveillance

10.2.1.1 The Basic Concept of Public Health Surveillance

Definition of Public Health Surveillance

Public health surveillance is a persistent, continuous, and systematic process of

information collection about health events and problems. After scientific analysis
and explanations of important public health information, we should provide timely
feedback to people or institutions that need this information to establish a perfect
process and to evaluate public health interventions and strategies. Its purpose is to
provide decision-makers with basic decisions and to evaluate the effectiveness of
those decisions. In a word, public health surveillance is a process of systematic
collection, analysis, interpretation, management, and use of public health informa-
tion which is persistent and ongoing.
Public health surveillance has three basic characteristics or consists of three
phases of work:
① Health-related information should be collected continuously and systematically
in order to discover the distributions and trends of public health issues.
② The raw materials should be scientifically sorted, analyzed, and interpreted,
then transformed into valuable and important public health information.
③ Public health information should be fed back to the relevant departments and
staff. In order to achieve the ultimate goals of surveillance, this information
should be used timely and fully.

Related Basic Concepts and Terminology

① Passive surveillance and active surveillance

(a) Passive surveillance means that subordinate units routinely report surveil-
lance data to the higher authorities, while higher units passively accept it,
182 C. Song

such as the surveillance information system of statutory infectious disease

and the surveillance spontaneous reporting system of adverse drug reaction.
(b) Active surveillance is based on special needs. Higher units require collecting
information, such as the omission or concealment of infectious diseases,
surveillance of certain behavioral factors (such as smoking and drug abuse)
and the US CDC established active surveillance system of foodborne disease
(Food Net).

② Routine report and sentinel surveillance

(a) The reporting is mainly carried out by the statutory responsible reporting
agencies and professional staff, and the coverage is nationwide.
(b) Sentinel surveillance means choosing a number of representative regions
and/or groups according to the prevailing characteristics of the disease mon-
itored, then monitoring continuously for the purpose of a better understanding
of the distributions of certain diseases in different regions, different
populations, and corresponding influencing factors based on a unified mon-
itoring program. The most typical project is the sentinel surveillance of AIDS,
which refers to selecting the representative areas and AIDS-related high-risk
groups and then continuously carrying out the tests for fixed-point, regular,
and quantitative HIV antibody according to a unified monitoring program and
testing reagents. Meanwhile, workers collect monitoring information of the
high-risk behaviors associated with the spread of HIV/AIDS in populations,
so as to obtain the information on HIV infection status and behavioral risk
factors and trends among different regions and different populations. In
addition, monitoring of influenza-like illness (ILI), mainly choosing a number
of selected hospital clinics as a monitoring sentinel, and weekly reporting of
the number of ILI cases.

10.2.1.2 The Purpose and Application of Public Health Surveillance

The information for public health surveillance may come from a variety of sources,
including demographic information and disease information, health and hygiene
data, many types of environmental surveillance data, animal-related data, and other
relevant information.
10 Disease Prevention and Surveillance 183

The Purpose of Public Health Surveillance

① Describe the characteristics of distribution and trends of health-related events:

through continuous and systematic public health surveillance, we can fully
understand the characteristics of distribution and trends of health-related events
among certain areas and people, which can help to solve the following problems:
(I). Quantitatively assess the seriousness of public health issues and identify
major public health issues. Important information on health issues needs to
be recognized to develop correct and targeted public health policies, plans,
or measures for the current or future periods.
(II). Find abnormal distribution of health-related events, promptly investigate
the causes, and take interventions to effectively curb the development and
spread of adverse health events. Long-term and continuous surveillance can
help us identify abnormal changes in the distribution of health-related
events and then quickly issue an early warning to the health agencies and
related units and timely organize and carry out the necessary epidemiolog-
ical surveys. Once there is an outbreak or epidemic of the disease, we can
take appropriate interventions to control further spread of the epidemic.
(III). Predict the development trends of health-related events and correctly assess
the needs of health service. Through dynamic surveillance and data analy-
sis, it can help to predict the trends and scales of the relevant events and
correctly estimate the needs of future health service.
(IV). Investigate the influencing factors of the diseases and determine the high-
risk populations. Besides, the contents of public health surveillance also
include surveillance of behavioral risk factors, environmental pollutants,
food safety, and nutritional deficiencies or excesses, and the analysis of this
information can contribute to a variety of factors that affect the develop-
ment of the diseases and to determine the high-risk population of the
corresponding diseases, which can provide a scientific basis for developing
targeted interventions and reasonable, effective strategies.

② Evaluate the effectiveness of public health intervention strategies and measures:

public health surveillance is conducted continuously and systematically, the
trends of diseases or related events can provide the most direct and reliable
basis for assessing the effectiveness of intervention strategies and measures.

Application of Public Health Surveillance

According to the purposes of public health surveillance, the application of public

health surveillance is divided into the following six major types by the WHO in
2002:
① Identify one or more cases and intervene to prevent infection or reduce mor-
bidity and mortality.
184 C. Song

② Assess the impacts of health events on public health or judge and measure its
trends.
③ Demonstrate the need for public health intervention projects and resources and
allocate resources rationally in the development of public health programs.
④ Scrutinize the effectiveness of prevention and control methods and
interventions.
⑤ Identify high-risk populations and geographic areas for intervention and the
research of guidance analysis.
⑥ Establish the hypotheses and the analytical research of the risk factors that lead
to the cause and progression of the disease.

10.2.2 Categories of Public Health Surveillance

With the development of public health activities, the types and contents of public
health surveillance are constantly enriched. At present, the public health surveillance
includes surveillance of diseases, cause of death surveillance, surveillance of hospi-
tal infections, symptom-based surveillance, behavior and risk factor surveillance,
and other public health surveillance.

10.2.2.1 Surveillance of Disease

In the perspective of health issues as epidemiological studies, disease surveillance is

a surveillance of outcomes, and surveillance requires an unequivocal diagnosis of
the corresponding diseases and death.

Surveillance of Communicable Diseases

In 2005, the World Health Assembly approved the international health regulations
{IHR (2005)} and began its implementation on June 15, 2007. According to IHR
(2005), WHO defines four kinds of diseases that must be notified in any case and
defines their corresponding cases. It includes smallpox, polio caused by wild strains,
human influenza caused by a new subtype virus, and severe acute respiratory distress
syndrome (SARS). It also provides 20 kinds of infectious diseases that received
global alert and response, including the 2009 pandemic H1N1 influenza, Ebola
hemorrhagic fever, dengue fever, hepatitis, monkeypox, Hendra virus, yellow
fever, gram ramea—Congo hemorrhagic fever, Lassa fever, Rift Valley fever,
influenza, Marburg, meningococcal disease, Nipah virus infection, avian influenza,
plague, smallpox, anthrax, tularemia, severe acute respiratory syndrome (SARS).
10 Disease Prevention and Surveillance 185

According to the Law of the People’s Republic of China on Prevention and

Control of Communicable Diseases, legally reported communicable diseases are
divided into three categories and 40 specific conditions. Any kind of communicable
disease occurring and causing death must be reported to local and national Centers
for Disease Control and Prevention and should be acted upon by persons responsible
for preventing and controlling disease in China.
The main contents and purposes of communicable disease surveillance are as
follows:
① Timely detect and diagnose cases in order to track and control; discover
emerging infectious diseases or new public health problems.
② Learn about the distribution of diseases and determine the epidemic or outbreak
in time so as to initiate an outbreak survey and control the epidemic situation.
③ Monitor the community immunity, serotypes, and genotypes of the pathogen,
the species and distribution of the host and vector insects, and the carrying status
of pathogens. Know about the changing trends of diseases, identify groups or
regions with higher risk factors, and provide information for the formulation and
adjustment of strategies and measures to intervene.
④ Monitoring the progress and effect of public health intervention projects (strat-
egies and measures).

Surveillance of Non-communicable Diseases

With the change of diseases spectrum, the scope of disease surveillance has
expanded to non-communicable diseases. According to the main health problems
or monitoring purposes in different countries or regions, the monitoring content
varies, including malignant tumors, cardiovascular diseases, cerebrovascular dis-
eases, diabetes, mental diseases, occupational diseases, birth defects, etc.
The U.S. National Cancer Institute (NCI) has been monitoring cancer since the
1970s. Centers for Disease Control and Prevention have been promoting health
promotion activities for chronic diseases since the 1980s, which is aimed at ten kinds
of preventable chronic diseases that seriously affect the quality of life.
Non-communicable diseases such as malignant tumors, cardiovascular diseases,
cerebrovascular disease, and birth defects have also been surveyed in some areas of
China. For example, the Beijing cardiovascular and pulmonary vascular Medical
Research Center organized 16 provinces and cities, 19 monitoring areas in China to
survey the trend of cardiovascular disease and its influencing factors (i.e., MONICA
project 1984–1993). In 2014, the National Cancer Registry collected data on cancer
registration from 234 registries in 2011, including 220 million people. The Chinese
cancer registration annual report for 2015 shows that in 2011, the number of China’s
new cancer cases was about 3,370,000, an increase of 270,000 compared to 2010.
186 C. Song

Surveillance of Hospital Infection

Surveillance of hospital infection refers to a long-term, continuous, and systematic

process of observation, collection, and analysis of the occurrence, distribution, and
determinates of hospital infection. A good program disseminates data and provides
scientific evidence for prevention and control of hospital infection. China’s current
hospital infection monitoring standard (WS/T312-2009) was issued by the former
Ministry of health in April 2009 and began to be implemented in December of the
same year. Nosocomial infection surveillance includes comprehensive surveillance,
target surveillance, surveillance of bacterial resistance, and surveillance of antimi-
crobial use. The standard requirement is as follows: establishing the effective
hospital infection surveillance and notification system, identifying hospital infec-
tious cases in time; analyzing the risk factors of hospital infection and taking targeted
prevention and control measures; training hospital infection control full-time per-
sonnel and clinical medical staff on the awareness and ability to identify an outbreak
of hospital infection. When the outbreak occurs, the source of infection and the route
of infection should be analyzed by epidemiological methods and effective control
measures should be taken. According to the different circumstances of the outbreak
of nosocomial infection, reports should be given to the health administrative depart-
ment within the prescribed time.

Surveillance of Death

The purpose of death surveillance is to understand the mortality and distribution of

death causes. Through the statistical analysis of death causes, the health level of the
monitoring population can be assessed, and the main causes of death and disease
control in different periods can be determined. “National Maternal Death Monitoring
Network” and “National Death Monitoring Network for Children under 5 years of
age” were established, respectively, in 1989 and 1992. Monitoring information is
used to reflect the health status of women and children in China. China CDC
formulated and issued the “National Death Surveillance System for Disease Sur-
veillance System (Trial)” and the “National Death Registration Information Network
Report Specification (Trial)” respectively in 2005 and 2007, which made the death
monitoring work more standardized. The medical certificate of death is an important
evidence for the death report and statistical analysis, and the correct judgment of
death causes is the most important basis for monitoring the cause of death.

10.2.2.2 Symptom Surveillance

Symptom-based surveillance is a continuous and systematic process for collecting

and analyzing data on certain groups of symptoms so authorities can mount a quick
response to rapidly detect and predict the occurrence of a disease. Symptom sur-
veillance is especially suitable for some new diseases with unknown etiology, and
10 Disease Prevention and Surveillance 187

the case has no definite diagnosis method. The surveillance of influenza is a

component of syndrome surveillance and played an important role in the surveil-
lance of influenza A (H1N1) pandemic in 2009–2010.
The common symptom surveillance includes influenza symptoms (cough, sneez-
ing, surveillance, fever surveillance, diarrhea surveillance, etc.). Symptom surveil-
lance does not depend on specific diagnosis; its emphasis is on the surveillance of
diseases with specific symptoms. The content of the surveillance is not only the
clinical symptoms (such as fever, diarrhea, respiratory symptoms, etc.) it also
includes many disease-related phenomena.
There are mainly about:
1. The hospitalized features of hospital emergency room or outpatient.
2. Sales situation of OTC drugs (such as vitamin C, cold medicine, antidiarrheal
drugs, etc.).
3. Sales volume of medical related supplies (such as medical masks, sanitary
napkins, etc.).
4. Absenteeism rate of school or unit.
5. Disease or death in animals.
6. Changes in biological vectors.
The classification of symptoms and the diagnosis of symptoms are the basic
components of the symptom surveillance system. More respiratory symptoms,
gastrointestinal symptoms, skin symptoms, and neurological symptoms have been
continuously used for symptom surveillance. Syndrome surveillance does not rely
on the clinician’s ability to consider and detect a specific disease or on the avail-
ability of local laboratory or other diagnostic resources. Since syndrome surveillance
focuses on syndromes rather than the diagnoses and suspicious diagnoses, it is less
specific and more likely to identify multiple individuals without the disease of
interest. As a result, more data have to be handled, and the analysis tends to be
more complex. So far, the symptom surveillance in China has been tested and
evaluated in public health surveillance of many important social activities, such as
the Beijing Olympic Games, Shanghai World Expo, Guangzhou Asian Games, and
so on.

10.2.3 Methods of Public Health Surveillance

10.2.3.1 Surveillance Methods

Public health surveillance is the core function of public health practice. Public health
surveillance requires the establishment of specialized monitoring organizations,
which should have the appropriate administrative and technical conditions and the
funding required ensuring the operation. The surveillance system is an organized and
planned surveillance system for a particular disease or a public health problem that
can be used separately or simultaneously on a population-based, hospital-based, and
188 C. Song

laboratory-based surveillance method. In addition, there are case-based surveillance,

indicator-based, and event-based surveillance. Monitoring-related activities include
the collection of data, analysis of data, dissemination of information, and the use of
information.

Population-Based Surveillance

Population-based surveillance is where the specific populations and the dynamics of

the particular diseases are monitored. It can not only be a regular report monitoring
that covers the entire population but also a monitoring point or sentinel surveillance,
and with good representative sentinel surveillance, which can obtain more accurate,
reliable, and timely information that is less costly and efficient. Many behavioral risk
factors are monitored by population-based surveillance.

Hospital-Based Surveillance

Hospital-based surveillance refers to the hospital as the scene and the patient as the
monitoring object, mainly on hospital infection, pathogen resistance, and birth
defects. The monitoring of statutory infectious disease reporting systems and passive
surveillance of adverse drug reactions fall into this surveillance.

Laboratory-Based Surveillance

Laboratory-based surveillance is mainly the use of laboratory methods to monitor

pathogens or other pathogenic factors. Such as WHO and China’s influenza labora-
tory testing system, carried out routine influenza virus isolation and classification for
identification work, i.e., laboratory-based influenza virus surveillance. Rapidly
developed in many countries, pathogen molecular subtyping pulse net is a
laboratory-based pathogen detection that covers almost all of the world’s major
countries. Laboratory-based surveillance is beneficial in increasing the use of mul-
tiple reverse transcription polymerase chain reaction and faster pathogen identifica-
tion in the twenty-first century.

Case-Based Surveillance

Case-based surveillance is mainly referring to the disease prevention and control

systems as the main body, jointly by clinical care institutions and other health care
units on the special case cases and aggregation of cases of monitoring. Analyzing the
number of episodes of disease outbreaks is often easier and more practical than
investigating individual cases, especially for diseases that are potentially at risk of
failure, poor quality reporting, or clinical type of disease. China’s public health
10 Disease Prevention and Surveillance 189

emergency surveillance, food safety incident surveillance, and so on, belong to case-
based surveillance.

Indicator-Based Surveillance

A variety of surveillance systems can collect quantitative data, such as statutory

infectious disease reporting systems, symptom surveillance systems, behavioral risk
factor surveillance systems, etc., which can provide quantitative data for epidemic/
outbreak intelligence mechanism(EIM).

Event-Based Surveillance

Collecting information from media and web search, news analysis, domestic and
foreign newsletters, public complaints and reporting, health advice, etc., can also
provide clues and basis for EIM, a public health incident reporting system is an
event-based surveillance system. However, there is no uniform standard for the
method of verification of an incident and procedures for reporting incidents, cur-
rently, and event-based surveillance has not yet been established in most countries.

10.2.3.2 Surveillance Methods and Techniques

The correct use of monitoring methods and techniques in the monitoring process
helps to improve the quality and efficiency of surveillance. With the development of
modern information technology, computer networks, geographic information sys-
tems, and other technologies that are more and more used in public health surveil-
lance, so that surveillance information collection, collation, analysis, transmission,
and feedback become more convenient, greatly improving the surveillance system’s
work efficiency, and also making the development of public health strategies and the
implementation of interventions more timely. In addition to active surveillance and
passive surveillance, routine reporting, and sentinel surveillance, the commonly used
surveillance methods and analytical techniques are also included.

Case Registration

Case registration is the registration, detection, diagnosis, and information registra-

tion of daily work surveillance-related cases, and so on. It is the basis of surveillance
work, and it is also an essential part of disease and death surveillance.
190 C. Song

Unrelated Surveillance

When the purpose of surveillance is to understand the prevalence of a disease in the

population, rather than to identify specific cases, the information collected by other
studies can be used to monitor without identifying the individual, known as
unrelated anonymous surveillance. Such surveillance can waive ethical issues to a
certain extent. For example, to collect blood samples from the hospital laboratory
and to identify HIV status by HIV antibody testing without identifying individual
identities.

Record Iinkages

The data from two different sources are connected to form a new database for
statistical analysis in order to obtain more valuable surveillance information,
which is called a record connection. For example, in the absence of information on
the future incidence or death and no record of birth weight in the infant death data,
you can get different birth weight infant mortality information through the two data
link analysis.

Collect Surveillance Information Online

Using computer-assisted telephone interviewing system (CATI)and network sur-

veys, respondents can use short time and less cost to get more quality access to data,
and the data can be used directly by various statistical software, automated data
management, and automatic statistical analysis, which quickly and easily complete
the surveillance information collection and analysis. CATI applies high-speed com-
munication technology and computer information processing technology to tradi-
tional telephone interviews, more than half of the developed countries in Europe and
the United States adopt CATI technology for surveys. There are also examples of the
application of CATI in China, such as the use of influenza-like symptoms, the
situation of seeing a doctor, and the diagnosis of influenza A H1N1 in public in
Beijing. Network survey, also known as online survey, refers to the internet, and its
investigation system will be the traditional paper survey and analysis methods
online, intelligently. These two surveys can greatly expand the number of surveys
and geographical scope, save survey costs and improve the efficiency of data
management and statistical analysis.

Network Direct Reporting System

With the rapid development and popularization of computer network technology, an

increasing number of network direct reporting systems have been established and
adopted in public health surveillance systems. Chinese public health emergency
10 Disease Prevention and Surveillance 191

surveillance system and the statutory infectious disease reporting information system
have achieved a direct network, at the county and township level, greatly reduced the
time of information transmission, and laid the foundation for the rapid processing
of data.

Automatic Warning Technology

Early warning is the response based on abnormal information from the surveillance,
such as an abnormal increase in a particular disease case so that the relevant
departments and those who may be affected by the incident respond in a timely
manner. Automatic warning technology is the use of mathematical models and
computer information technology through a specific algorithm to determine the
warning threshold, and an automatic detection of possible abnormal information
(above the threshold), as well as an early warning signal is then issued. It is worth
noting that the level of early warning threshold directly affects the sensitivity and
specificity of the warning signal, the warning signal prompted by the suspicious
events need further analysis and verification.

The Use of Geographic Information System

The use of geographic information system (GIS) makes public health surveillance
data more visible in the regional distribution, which helps to analyze the geograph-
ical environment and climate factors on the impact of public health problems.

10.2.3.3 Attention in Public Health Surveillance

Case Definition and Disease Surveillance

In large-scale surveillance work, strictly following the clinical diagnostic criteria to

determine a disease case is often limited by working conditions and difficult to
operate. So in order to determine a unified, highly operational monitoring standard is
extremely important, cases identified using monitoring criteria are called surveil-
lance cases. For example, the diagnosis of many infectious diseases is mainly based
on clinical diagnosis and characteristics, and not necessarily needing a pathogen test.
Many of the cases were reported in Chinese legal infectious diseases belong to
surveillance cases. Proportion of the actual cases in monitored cases should be
increased as far as possible, and this proportion can be estimated to a certain extent.
192 C. Song

Static and Dynamic Populations

In the process of monitoring, the population in which no people move out and
displace is called a fixed population. If there are only a small number of births,
deaths, immigrations and emigrations, and relocations in a region with more popu-
lation, it can also be regarded as a fixed population. For the static population to
calculate the sample rate, the average population of the observation period can be
used as the denominator. If people frequently move in or out of the population during
the monitoring process, it is a dynamic population. For the calculation of the
dynamic population rate, it needs to use the total person-hours observed as a
denominator.

Surveillance Information in-Depth Analysis, Exchange, and Sharing

The use of automatic analysis technology can be more efficient and clearer when
analyzing the data and obtain more meaningful information; according to the
monitoring data and analysis information, a timely and effective early warning can
be released through automatic early warning technology. Different surveillance
systems can achieve a certain amount of information through sharing and exchange
of information and greatly improve the effectiveness of monitoring work.

Confidentiality System/Surveillance Ethics

Many diseases involve personal privacy, and some patients or people infected with
the disease may be subjected to social discrimination. The monitoring of these
diseases should strictly comply with the confidentiality system. The risk of supervi-
sion may be at the individual or collective level. If information about health is not
properly released, people may feel embarrassed and discriminated against. On the
one hand, we need to maintain the monitoring of the dignity and rights and interests
of the subject. On the one hand, we need to maintain the monitoring of the dignity
and rights and interests of the subject. On the other hand, we need to enhance the
public knowledge of the activities of the monitoring and the awareness of
participation.

10.2.4 Procedures and Assessment of Public Health

Surveillance

10.2.4.1 Basic Procedures of Public Health Surveillance

Core activities of public health surveillance systems include collecting, analyzing,

disseminating, and utilizing the information of monitored public health events.
10 Disease Prevention and Surveillance 193

Systematically Collecting Relevant Data

Relevant data about public health surveillance are collected systematically and
comprehensively according to the specific purpose of the different supervising
systems. The data collection process should observe the uniform criteria and scien-
tific methods, and normative schedule. In conducting surveillance, there are various
approaches to gather data about public health problems, including demographic
information, disease incidence or death certificates about population, laboratory
reporting about pathogen and serology, investigations about risk factors, records of
the intervention (e.g., distribution of vaccines, supply of iodized salt) measures,
special surveys, and other relevant data.

Managing and Analyzing Collected Data

Data management refers to checking and classifying gathered data to ensure data
integrity and accuracy.
Data analysis refers to the transformation of various data into relevant indicators
using statistics technology. Analysis of surveillance data is usually conducted to
characterize the distribution, trend, and influences about the public health problem
under surveillance. In the process of data analysis, on the one hand, it is necessary to
select the correct statistical methods according to the nature of the data to fully
appraise and apply the data. On the other hand, the influence of various factors on the
results of disease surveillance must be considered in order to arrive at correct and
reasonable explanations.

Dissemination and Feedback of Information

These data and surveillance reports must be shared with those who supplied the data
and those who are responsible for controlling public health problems. The timely,
regular dissemination of basic data and their interpretations is very important for
surveillance. For example, the World Health Organization collects and analyzes all
aspects of monitoring data regularly, disseminating information on public health
surveillance by “Weekly Epidemiological Report” and other publications around the
world. The use of the Internet to distribute information is a new development for
public health surveillance. The Public Health Information Network initiative spon-
sored by the Centers for Disease Control and Prevention, is heavily standards-based
and based on the HL7 Reference Information Model.
The role of information feedback cannot be overlooked. Information feedback is a
bridge linking public health surveillance and intervention. The channels of informa-
tion feedback must be established to disseminate the information from disease
surveillance in a timely fashion to all relevant organizations and individuals so
that they are able to rapidly respond to health problems. The monitoring information
can be disseminated in both vertical and horizontal directions. The vertical direction
194 C. Song

refers to the monitoring information reported to higher-level health administrative

departments and managers. The horizontal direction refers to the monitoring infor-
mation transmitted to the lower-level monitoring institutes and experts and commu-
nities as well as residents. The content and manner of feedback should be different
depending on the objectives.

Information Utilization

Information obtained from monitoring can be used to describe the distribution

characteristics of public health problems, to identify whether the prevalence exists,
to predict the trend of prevalence, to evaluate the effect of interventions, and to
provide a basis for making public health activities. Data and interpretations derived
from surveillance activities are useful in setting priorities, planning and conducting
disease control programs, and assessing the effectiveness of control efforts. It is the
ultimate goal of public health monitoring to make full use of monitoring information,
to develop strategies of public health in a timely manner, and to take effective
interventions.

10.2.4.2 Evaluation of Public Health Surveillance System

The quality and effectiveness of public health surveillance systems need to be

regularly evaluated in order to improve the effectiveness of public health monitoring
systems and to better serve public health activities.

Quality Evaluation of Monitoring System

Evaluation of public health surveillance programs can be based on a number of

important characteristics.
Completeness It refers to the diversity of monitoring content or indicators
contained in the monitoring system, which include the integrity of the monitoring
forms, the integrity of the case reports, and the monitoring data.
Sensitivity It refers to the ability of the monitoring system to identify public health
problems. This assessment mainly involves comparing the proportion of cases
reported by the monitoring system to the number of actual cases and the ability of
the monitoring system to detect when an outbreak or epidemic of a disease (other
public health events) has occurred.
Specificity It is used to reflect the ability of the monitoring system to exclude
non-public health problems. Such as the ability of the monitoring system to correctly
identify the random fluctuations in the phenomenon of the disease population,
thereby avoiding or reducing the false alarms in public health monitoring.
10 Disease Prevention and Surveillance 195

Timeliness This attribute refers to the time interval from the occurrence of a public
health event to the relevant department receiving a report. It reflects the speed of
information dissemination and feedback of the monitoring system. It is very impor-
tant for acute communicable disease surveillance, which may directly affect the
efficiency of the intervention. Measurement of timeliness in surveillance systems
needs to be planned for in system design.
Representativeness It refers to the extent to which the public health problem
identified by the monitoring system can represent the actual occurrence of the target
population. Lack of representativeness may lead to a waste of health resources.
Simplicity Are forms easy to complete? Is data collection kept to a necessary
minimum? The method for performing surveillance typically should be as simple
as possible and achieve the purpose of monitoring.
Flexibility It refers to the ability of the monitoring system to respond to new public
health issues and to adjust the operational procedures or technical requirements in a
timely manner to meet new needs.

Benefit Evaluation of Monitoring System

Benefit evaluation of monitoring system consists of important characteristics.

Health Economics Evaluation The establishment and operation of any monitoring
system requires costs investment, sometimes even costly. The effectiveness of the
monitoring system is mainly reflected in the early warning and timely responses to
the disease or event, the prevention, and control of the disease, and the improvement
of the people’s health levels. Therefore, the evaluation of health economics is
indispensable.
Positive Predictive Value This refers to the proportion of reported or identified
cases accounted for are real cases. To what extent are the reported cases real cases?
To what extent are measured changes in trends truly reflective of events in the
community?
Acceptability It reflects the enthusiasm of individuals and organizations to partic-
ipate in a surveillance system. Acceptability is influenced substantially by the time
and effort required to complete and submit reports or perform other surveillance
tasks.
Interconnection and Sharing Ability in Monitoring Systems Most of the mon-
itoring systems are established for specific purposes, so there may be some limita-
tions in collecting and utilizing information. The establishment of the
interconnection and sharing between monitoring systems is able to greatly improve
the working efficiency, information utilization, and reduce the waste of resources. It
is quite important for the monitoring system to realize the importance of intercon-
nection and sharing between monitoring systems.
196 C. Song

In addition to the evaluation of quality and benefit, the function of the monitoring
system should be evaluated. The function of the monitoring systems consists of core
functions and support functions. The core functions include case monitoring, case
registration, case confirmation, reporting process of pathology, data analysis and
interpretation, epidemic warning, and information feedback. The support functions
refer to those conditions that can facilitate the achievement of core functions,
including the implementation standards and guidelines of monitoring system, the
training and supervision for relevant personnel and organizations, the necessary
communication equipment, the necessary resources of human and material as well
as the monitoring, assessment, and coordination of monitoring systems.
Chapter 11
Communicable Diseases Epidemiology

Rongguang Zhang

Key Points
• Epidemics of communicable diseases have become crucial public health prob-
lems in recent decades.
• Epidemic process is the process in which a communicable disease is transmitted
among individuals in a specific population. Investigation of the epidemic process
forms fundamental basis for prevention of the communicable disease.
• In most cases, both population strategy and high-risk strategy are taken for
obtaining a high efficacy of prevention and full use of the public health resources.
• Surveillance of communicable diseases play an important role in prevention and
control of these diseases and evaluation of efficacy of the strategies and
measures used.
In history, communicable diseases had ever been the most serious threaten to human
life and health until the end of World War II. Then the situation ameliorated
accompanying with the advancement of biologic theory and techniques as well as
the enhancement of human living conditions. In recent decades, however, epidemics
of communicable diseases have become crucial public health problems once again.
Certain newly identified infectious diseases and previously controlled communica-
ble diseases have been emerging or recurring in part due to the variation of
pathogens, changes in social and environmental conditions as well as human living
modes.

R. Zhang (✉)
International School of Public Health and One Health, Hainan Medical University, Haikou,
China
e-mail: [email protected]

© Zhengzhou University Press 2023 197

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_11
198 R. Zhang

11.1 Infection Process

11.1.1 Infection Process

Infection process refers to the process of a pathogen’s invasion into and interaction
with human body. By infection process, it is also meant the entire process from onset
and development to the end of an infectious disease in individuals. Through the
process, an infection may result in various outcomes in individuals.

11.1.2 Spectrum of Infection

Spectrum of infection is the composition of all the various outcomes caused by an

infection in population (Fig. 11.1). Invasion by a species of pathogen may cause a
variety of outcomes in an infected population, such as latent infection, incubatory
infection, apparent infection and death, although only one outcome is exhibited in an
infected individual. Spectrum of infection is an important epidemiologic character-
istic of a communicable disease, defined on outcomes of infected populations. For
instances, most persons infected with poliomyelitis or encephalitis B pathogen have
only unapparent infection, majority of those with measles or crystalli become
clinical patients while suffering from rabies or AIDS commonly results in death.

Fig. 11.1 Spectrum of an assumed infection

11 Communicable Diseases Epidemiology 199

11.2 Epidemic Process

11.2.1 Definition

The epidemic process is the process in which an infectious disease is transmitted

among individuals in a specific population. Insight into epidemic processes of
communicable diseases enables us to make appropriate strategies and measures for
prevention and control of infectious disease transmission.

11.2.2 Three Links in the Epidemic Process

An epidemic process is constituted by three links, namely, reservoir of infection,

route of transmission and susceptible host.

11.2.2.1 Sources of Infection

Humans and animals infected with pathogens can act as sources of infection. The
process of apparent infectious diseases can be divided into such three periods as
incubation period, clinical stage and recovery (convalescent) period. The infectious
stage consists of the later stage of the incubation period, clinical stage and the early
stage of convalescent period, in which the infected persons discharge pathogens and
play the roles of infection sources (Fig. 11.2). In the early stages of the incubation
periods and the later stages of the convalescent periods of most infectious diseases,
the infected persons contribute no or little to the transmission of the diseases. The
carriers of certain infectious diseases like hepatitis B, which have no clinical
manifestations after infection, can also be important sources of infection. For

Fig. 11.2 The infectious stage in an infection process

200 R. Zhang

zoonoses like schistosomiasis, clonorchiasis, paragonimiasis and rabies, the animals

suffering from these diseases are also critical sources of infection.

11.2.2.2 Routes of Transmission

Routes of transmission are the ways by which pathogens are transmitted in the
environment outside their hosts from being discharged to invading new susceptive
hosts. Various communicable diseases are transported via different manners, which
are of significance for formulating strategies and measures to block the transmission.
1. Airborne transmission: Respiratory infectious diseases are mainly transmitted by
respiration. Droplets, produced by respiration and especially sneezes, may carry a
large number of pathogens and can cause infection when in contact with the
respiratory tracts of susceptible masses. Droplets will become droplet nuclei after
floating in air for a time with water evaporated. Droplet nuclei are smaller than
droplets, capable of floating in air for longer time, and what is more serious,
droplet nuclei can reach the deep part of respiratory tracts, and thus easier to cause
infection. Pathogens can also be carried by dust floating in air and infect human
respiratory system.
2. Waterborne transmission: Many communicable diseases are transmitted by drink-
ing water or contacting polluted water, involving gastrointestinal infectious
diseases like cholera, giardiasis, entamebiasis and parenteral parasitic infections
like schistosomiasis.
3. Foodborne transmission: Susceptive masses can catch certain gastroenteral infec-
tious diseases and few respiratory diseases through ingestion of food containing
pathogens. The epidemiological characteristics of food transmissions are as
follows: (1) the patients share the same history of food taking, (2) the persons
without ingestion of the food are not infected, (3) the patients have short incuba-
tion periods, (4) the outbreak may take place if a large population are exposed to
the food and (5) outbreak or epidemic of the communicable diseases will subside
following stopping the food supply.
4. Transmission by contact: Communicable diseases can spread via contact with
sources of infection or pathogen-polluted environments. Contact transmission
includes direct contact transmission and indirect contact transmission. The former
means that infectious diseases like sexually transmitted diseases and rabies are
transmitted by direct contact with infection sources without the participation of
any other environmental media. The later refers to that susceptive masses are
infected by exposure to some environmental factors like polluted public equip-
ment and places (e.g., doorknobs).
5. Arthropod-borne transmission: Arthropods can carry or, as hosts, harbor certain
pathogens and cause infection, so arthropod-borne transmission can be divided
into mechanical transmission and biological transmission. During mechanical
transmission, pathogens are only transported by arthropods without any changes
in morphology and quantity. In biological transmission, the arthropods harbor
11 Communicable Diseases Epidemiology 201

pathogens as their hosts, and the pathogens might develop or multiply during the
transmission process. For instance, malaria and filariasis are biologically trans-
mitted by mosquitoes, while giardiasis and cholera are mechanically spread by
flies’ transportation.
6. Soilborne transmission: Soilborne transmission refers to the routes that infectious
diseases are transmitted by contacting soil polluted with pathogens, such as
parasites’ infective eggs and larvae and certain bacterial spores. Hookworm
infection, ascariasis, trichuriasis, anthrax and tetanus are common soilborne
communicable diseases. The epidemiological importance of this transmission
depends on the vitality of the pathogens in soil, possibility for human contacting
soil, personal hygiene habit and working conditions.
7. Iatrogenic transmission: Iatrogenic transmission is caused by medical and health
care workers when performing diagnosis and treatment for patients without
complying with the regulations and operating procedures. For instance, hepatitis
B and AIDS can be transmitted by blood transfusion and invasive body
examination.
8. Vertical transmission: Vertical transmission refers to the routes by which off-
spring are infected from their mothers by placenta, upstream infection or delivery.
For example, infants can catch toxoplasmosis, hookworm diseases, hepatitis B,
gonorrhea and herpesvirus infection by vertical transmission.
9. Multiple transmissions: In most cases, a communicable disease can be transmitted
by more than one route. For instance, human can get infection with hookworm by
contacting soil or drinking water polluted with the filariform larvae.

11.2.2.3 Herd Susceptibility

Herd susceptibility is the mean level of susceptibility of a population to an infectious

disease. Herd susceptibility is closely and negatively relevant to the population’s
immunity against a pathogen’s infection. The factors for an increase in herd suscep-
tibility includes enhanced quantity of newborn infants, immigration of easily
infected masses, decreased quantity of individuals with immunity and deaths of
immunized masses. Whereas, planned immunization and epidemic of infectious
diseases can contribute to decrease in herd susceptibility.

11.2.3 Two Factors Affecting the Epidemic Process

Epidemic outbreak and intensity of an infectious disease depend on the coexistence

and association of three links, and can vary from changes in any of them. However,
both natural and social factors can influence the epidemic of infectious diseases
through interaction with the three links.
202 R. Zhang

11.2.3.1 Natural Factors

Natural factors include those related to geographic characters, climatic conditions,

soil, animals and plants, and contributing to epidemics of infectious diseases through
complicated mechanism. Certain geographic and climatic conditions have great
influence on growth, reproduction and habits of the vectors, and thereby affect the
transmission of communicable diseases such as malaria, filariasis and epidemic
encephalitis B. Climatic factors can also play key roles in epidemics of infectious
diseases by taking effects on human behaviors. In recent decades, accompanying the
global warming, remarkable changes in the breeding area, growth and multiplication
of mosquitoes and flies and enhanced virulence of the pathogens inside arthropods
have been observed.

11.2.3.2 Social Factors

Social factors involve almost all sorts of human actions, such as hygienic habits,
epidemic prevention work, health care conditions, living and nutritional conditions,
inhabitant environment, production activities, occupation, culture, custom, religion,
migration of populations, steady social environment and so on. In recent decades,
certain epidemics of newly emerging and reoccurring diseases were affected by
social factors. Social factors can aggravate or alleviate epidemics of communicable
diseases. For instance, unreasonable use of antibiotics provokes resistance against
antibiotics, resulting in epidemics of infections with drug-resistant Mycobacterium
tuberculosis. Abuses of insecticides have accelerated the development of mosqui-
toes’ anti-insecticides resistance, and thus aggravated the epidemic of malaria,
dengue fever and yellow fever. Urbanization and population explosion have caused
increase in epidemics of human infectious diseases, and wars, unrests, congestion of
displaced people, famine and rapidly expanded global tourism and industrialization
can play important roles in epidemic of communicable diseases.

11.2.4 Epidemic Focus and Epidemic Process

11.2.4.1 Epidemic Focus

An epidemic focus consists of a source of infection and a region that the pathogens
discharged from the infection source can reach. Multiple epidemic foci in geographic
fusion form an epidemic district.
An epidemic focus can be developed under two necessary conditions, namely,
existence of a source of infection and a route of transmission. The ranges of epidemic
foci vary from the species of infectious diseases, especially the transmission modes
and arthropod vectors. Commonly, air transmission of infectious diseases can
11 Communicable Diseases Epidemiology 203

produce a wider range of epidemic focus. For an arthropod-transmitted disease, the

range of the arthropod’s activity can become the range of an epidemic focus.
An epidemic focus will be extinguished under the following three conditions:
(a) removal of the infection sources, for example, death or migration of the patient;
(b) disinfection of the environment polluted by the pathogens from the sources of
infection; (c) all the contactees have been proved uninfected or not become new
infection cases in the longest incubation period.

11.2.4.2 Epidemic Process

A series of epidemic foci with contact to each other that appear early or late
constitute an epidemic process. An epidemic process is also a process during
which an epidemic of infectious disease occurs and expands in a population, and
the sources of infection, routes of infection and herd susceptibility of infection link to
each other. Investigation of the epidemic process forms the fundamental basis for the
prevention of a communicable disease.

11.3 Strategy and Implementation

Communicable diseases cause serious damage to human health, and are also the
predominant causes of death in many developing countries. In recent decades,
prevalence rates of global infectious diseases have, to a large extent, been rising,
outbreaks of infectious disease epidemic have been continually reported, certain
ever-controlled communicable diseases have reoccurred and many reemerging dis-
eases have been identified. Therefore, prevention and control of infectious diseases
should still be emphasized in public health care work throughout the world. To
improve the efficacy of these works, governments of all countries have to formulate
practical strategies and measures for prevention and control of the communicable
diseases according to their national conditions.

11.3.1 Strategies for Control of Communicable Diseases

For control of infectious diseases, the following policies have been suggested to
implement:
1. Establishment of thoughts for long-term struggles against communicable
diseases.
2. Prevention of infectious diseases should be carried out as the most predominant
action.
3. Comprehensive measures should be taken according to the epidemic processes.
204 R. Zhang

4. Formulate and implement laws and regulations for control of infectious diseases.
5. Prevention of transmission of certain emerging infectious diseases like
HIV/AIDS from high-risk to general populations.
6. Establishment of rapid response mechanism and organization for managing
emergency public health events.
7. Establishment of modern epidemical working staff to solve the problems in the
present situation on communicable diseases.

11.3.1.1 Population Strategy

Population strategy refers to taking prevention measures on whole populations for

lowering their levels of exposure to risk factors for communicable diseases. For
instance, 15 infectious diseases including hepatitis B, tuberculosis, poliomyelitis,
bronchocephalitis, diphtheritis, tetanus, leprosy, measles, parotitis and epidemic
encephalitis B, hepatitis A and leptospirosis are prevented by planned immunization
of all children in China.

11.3.1.2 High-Risk Strategy

High-risk strategy is meant allocation of limited public health sources to the popu-
lation at high risk for certain communicable diseases, in order to enhance the cost-
benefit of prevention works. In most cases, two pronged strategies are taken for
obtaining a high efficacy of prevention and full use of the public health resources.
Up till now, eradication of human smallpox is thought as the best example for
implementing the two strategies to fight against communicable diseases. In 1958, the
World Health Assembly approved the plan for the global eradication of smallpox,
and made up the vaccination strategy to elevate the inoculation rate of the whole
population. In 1967, when the infection rate decreased apparently, the high vacci-
nation rate was proved to contribute little to preventive efficacy of smallpox by
public health surveillance. From then on, WHO implemented a new strategy involv-
ing enhanced surveillance of smallpox cases and encircling inoculation, namely,
when finding new cases by surveillance, inoculation was performed on all the
contacts of the cases and blocked the transmission of smallpox. At last, the epidemic
of this serious disease was globally extinguished in 1979.
11 Communicable Diseases Epidemiology 205

11.3.2 Measures for Control of Communicable Diseases

11.3.2.1 Surveillance of Communicable Diseases

Surveillance of communicable diseases, as one of the main tasks of public health

surveillance, plays an important role in the prevention and control of these diseases
and evaluation of efficacy of the strategies and measures used.
Communicable disease surveillance can be divided into four levels: the local, the
regional, the national and the international infectious disease surveillance, which
take the responsibility for continually and systematically collecting and applying
information on infectious diseases from cities (or towns or villages), provinces
(or states or counties), countries and global area, respectively, in a long term.
Performance of communicable disease surveillance consists of four steps:
collecting information on infectious diseases, collating and analyzing the data,
giving feedback of the information to the populations who need it and applying
the information in the formulation of strategies and measures and evaluation of their
effectiveness.
Surveillance activities of communicable diseases include timely reporting legally
notifiable infectious disease cases, morbidity and mortality, results of field investi-
gations, laboratory isolation and identifications of pathogens and data on vaccina-
tions and immunity levels in populations.
The legally notifiable infectious diseases consist of a variety of serious commu-
nicable diseases, which are different in different countries or times. In China today,
there are 40 species of legally notifiable infectious diseases, which have been
grouped into three categories: category A involves plague and cholera; category B
includes 27 diseases, such as SARS, AIDS, viral hepatitis, poliomyelitis, highly
pathogenic avian influenza, H7N9 avian influenza, measles, epidemic hemorrhagic
fever, epidemic encephalitis B, bacterial and amoebic dysentery, tuberculosis,
typhoid and paratyphoid, gonorrhea, schistosomiasis, malaria, anthrax and so on;
category C comprises 11 diseases like epidemic influenza, epidemic mumps, rubella,
lepriasis, leishmaniasis, echinococcosis, filariasis, hand, foot and mouth disease and
so on.
The surveillance information should essentially be handled with respecting to
security and confidentiality of individual privacy. Only those who need to use the
individual information for the public health practice can have access to these data.
Presently, although the privacy of personal information has drawn more attention
than ever, more effort should be made to manage these data in a confidential and
secure way.

11.3.2.2 Measures on Sources of Infection

Patients, carriers and reservoir hosts infected are the sources of infection. For
eradication of the infection sources, early discovery, early diagnosis, early reporting,
206 R. Zhang

early isolation and early treatment should be performed for the patients. Once a
person is suspected suffering from one of the category A infectious diseases or one
of the two category B diseases, SARS and pulmonary anthrax, he/she must be
medically treated in isolation. The length of the isolation period should be deter-
mined according to the results of medical examinations. The carriers should also be
given treatment, and whether isolation is carried out depends on the species of
communicable diseases they have been infected with. As for the animal host of
pathogens, some of them like mice, invaluable for economic and environments,
should be killed, while some like cattle and dogs, valuable for economic and
production, can be medically treated or exterminated, according to the harm the
infectious diseases probably present to human health.

11.3.2.3 Measures on Routes of Infection

Measures should be taken for disinfection of the environments polluted from the
sources of infection. Various measures have to be used for blocking different
transmission routes of pathogens, for example, for prevention of intestinal infectious
diseases, which are transmitted mainly by feces to mouth route, disinfection has to be
carried out mainly on patients’ discharges, polluted water, rubbishes, polluted goods
and environments. As for the infectious diseases of the respiratory tract, improving
ventilation, air disinfection and individual protection with a nose mask can be useful
measures for prevention of infection; for blood-borne communicable disease like
AIDS and hepatitis B, taking safe sexual action, avoiding drug taking, avoiding
sharing of public syringes and strengthening management of blood transfusion are
practical measures.

11.3.2.4 Measures on Susceptible Populations

Susceptive masses exposed to sources of infection are at high risk for infection, can
probably become new epidemic foci, contributing to the transmission of communi-
cable diseases in populations. Preventive immunization, drug-based prevention and
individual protection are commonly used measures for protection of susceptive
populations.
1. Preventive immunization: Preventive immunizations are carried out prior to
epidemics of infectious diseases to enhance specific immunity of populations
and thereby prevent outbreaks of the epidemics. Preventive immunizations have
become the most important approaches to prevention of communicable disease
epidemics and classified into active immunization and passive immunization.
Vaccinations that induce long-term immunity via active immunization have
proved the most reasonable and effective routes for prevention and control of
infectious diseases. However, for many serious communicable diseases,
11 Communicable Diseases Epidemiology 207

developments of safe and effective vaccines have been great challenges that
humans face.
2. Drug-based prevention: Drug-based prevention is performed on susceptive
masses as an emergency measure when there is an outbreak of infectious diseases
and there are specifically effective drugs for these diseases. For example, certain
antimalarial drugs can be administrated to susceptive populations when an
epidemic of malaria takes place. Preventive administration of drugs can produce
short-term and unsteady efficacy against infection, with the development of
antidrug resistance.
3. Individual protection: Individual protection of susceptive masses plays an impor-
tant role in the prevention of communicable diseases. For instance, in epidemic
seasons of respiratory infectious diseases, avoidance of staying in densely pop-
ulated places, improved ventilation of inhabiting and working places and wearing
nose masks in contact with patients can significantly reduce the possibility of
infections. Also, infections of sexually transmitted diseases like AIDS can to an
extent be prevented by using condoms.

11.4 Immunization Program and Effectiveness

11.4.1 Immunization

Immunization is the inoculation of susceptible populations with prepared antigens or

antibodies through appropriate routes with the aim to elevate the populations’
immunity and thus prevents the epidemic of communicable diseases. Vaccination
has been a basic public health care service provided by governments and a public
health care work of high sociality.
Immunization includes artificial active immunization and artificial passive immu-
nization. In artificial active immunization, biological agents containing immuno-
genic antigens or toxoid are administered to target populations in order to induce
specific immunity against infectious diseases. Vaccines are the biological agents
used in preventive vaccination, which are prepared using detoxicated but antigenic
microorganisms or their metabolites. Currently, vaccines can be divided into many
categories including attenuated vaccine, inactivated vaccine, toxoid vaccine, subunit
vaccine, synthetic peptide vaccine, conjugative vaccine, gene engineering vaccine
and so on. Among them, gene engineering vaccines have drawn more attention of
researchers, in part due to their high safety, low cost of production and simple
preparation procedures. However, problems on relatively low effectiveness of
gene engineering vaccines have to be addressed for certain communicable diseases.
In artificial passive immunization, immune sera or immune globulin are injected
into human bodies to make the immunized objectives acquire specific immunity
against a communicable disease. By this way, the immunity can be obtained rapidly,
but only last for a short term. Therefore, passive immunization is mainly applied in
emergency prevention and therapeutic treatments.
208 R. Zhang

To take the advantages of both active and passive immunizations, the two modes
of artificial immunization can be used in combination. For instance, both vaccines
and antibodies are employed in immunization for the prevention of hepatitis B
infection in newborn infants and diphtheria infection in susceptive contactees.

11.4.2 Immunization Program

Immunization programs involve formulation of the programs, plans and strategies

for preventive immunization of susceptive masses according to the national pro-
grams for prevention and control of communicable diseases, and performance of
planned vaccination using nationally or provincially determined species of vaccines
and immunization programs, for prevention, control and eradication of infectious
diseases via enhancing population immunity.
In 1974, based on the experiences from eradication of smallpox and control of
measles, WHO proposed expanded program on immunization (EPI), in which it was
required that all the member countries should insist on applying immunization
methods and epidemiological surveillance in combination. Preventing epidemics
of diphtheria, bronchocephalitis, tetanus, measles, tuberculosis, poliomyelitis etc.,
with emphasis on heightening vaccination coverage rates and expanding species of
vaccines inoculated has been the mainstay of this approach. China joined EPI in
1981 and presently is carrying out a basic immunization program in children via
using 13 vaccines for prevention of 15 communicable diseases. In 1988, the National
Coalition for Adult Immunization (NCAI), a public health organization constituted
by 40 countries, put forward an adult immunization proposal on programmed
immunization with vaccines against epidemic influenza, pneumonitis, hepatitis B,
measles, tetanus and diphtheria in adults and the populations at high risks. At
present, the adult immunization program has been implemented gradually in some
developed countries and some districts in China.

11.4.3 Evaluation of Immune Effectiveness

Evaluations of immune effectiveness include appraising immunological effects,

epidemiological effects and immunization program management.
Immunological effects of immunization programs can be evaluated using positive
conversion rate of specific antibodies, averaged titer levels of antibodies and lasting
time of antibody positive status in immunized populations.

No:of positive conversion cases

Positive conversion rate = × 100%
No:of immunized individuals
11 Communicable Diseases Epidemiology 209

Immunological effects of immunization programs can be evaluated through the

calculation of immune protection rates and effect index based on the data from field
experiments using double-blind methods.

IR of control group - IR of vaccinated group

Protection rate = × 100%
IR of control group

IR of control group
Effect index =
IR of vaccinated group

Note: IR, incidence rate.

Immunization program management can be appraised referring to the indexes,
such as qualified inoculation rate of vaccines, coverage rate of programmed immu-
nization and record rate of vaccination.

11.5 Emerging Communicable Diseases

11.5.1 Definition

Emerging communicable diseases (ECD) refer to newly identified and previously

unknown infectious diseases which cause public health problems either locally or
internationally. Accompanying the decrease of traditional communicable diseases,
ECD is gradually becoming the main causes of public health problems.

11.5.2 Main Emerging Communicable Diseases

Since 1970, more than 40 species of ECD have been identified as shown in
Table 11.1, most of them are caused by viruses, and many of them do serious
damage to human health, and become new research foci in medical science.

11.6 Summary

Communicable diseases are caused by a variety of microorganisms including bac-

teria, viruses, protozoan and fungus. The associations between human and the
pathogens occur in two processes, that is, infection process and epidemic process,
the former is ongoing in individuals, while the latter is among individuals in specific
populations. A variety of outcomes resulting from infection with a pathogen in
210 R. Zhang

Table 11.1 Emerging communicable diseases identified since 1970

Year Pathogen Disease
1973 Rotavirus Diarrhea of infants
1975 Hepatitis B virus Hepatitis B
1976 Ebola virus Ebola virus disease
1977 Hantaan virus Hemorrhagic fever with renal syndrome
1977 Hepatitis D virus Hepatitis D
1977 Legionella pneumophila Legionellosis
1977 Campylobacter jejuni Enteritis
1981 Toxin-producing strains of Staphylococ- Toxic shock syndrome
cus aureus
1982 Escherichia coli O157: H7 Hemorrhagic colitis
1982 Borrelia burgdorferi Lyme disease
1983 Helicobacter pylori Gastric ulcer and cancer
1983 Human immunodeficiency virus, HIV Acquired immunodeficiency syndrome
1986 Human herpesvirus 6 Roseola infantum
1986 Cyclospora cayetanensis Persistent diarrhea
1989 Hepatitis C virus Hepatitis C
1989 Hepatitis E virus Hepatitis E
1990 Human herpesvirus 7 Fever, erythra, CNS infection
1992 Vibrio cholerae O139 New cholera
1993 Escherichia coli O12:K1:H7 Urethra infection, abortion, sepsis,
encephalomeningitis
1995 Human herpesvirus 8 Kaposi’s sarcoma
1995 Hepatitis G virus Hepatitis G
1996 Prion New Creutzfeldt-Jacob disease
1997 Rickettsia mongolotimonae Tick-borne lymphadenopathy
1998 Nipah virus Cephalomeningitis
2003 SARS-associated coronavirus SARS
2006 Avian influenza A (H5N1) virus Human avian influenza
2008 Anaplasma phagocytophilum Human granulocytic anaplasmosis
2009 Influenza A (H1N1) virus Influenza A
2013 Middle East respiratory syndrome Middle East respiratory syndrome
coronavirus
2016 Highly pathogenic avian influenza A Human avian influenza
(H5N8) virus
2019 SARS-CoV-2 COVID-19

populations is called spectrum of this infection. Communicable diseases can become

prevalent based on coexistence and association of the three links, namely, sources of
infection, routes of infection and susceptive populations. Epidemics of infectious
diseases are also affected by multiple natural and social factors. Surveillance of
public health forms the critical basis for prevention and control of communicable
diseases. Effectiveness of infectious disease control is, to a large extent, dependent
on reasonable strategies and practical measures. The measures employed should
11 Communicable Diseases Epidemiology 211

vary from the epidemic characters of communicable diseases, with the aim to
eradicate the sources of infection, blocking the transmission routes and protecting
the susceptive masses. Programmed immunizations have been proved as effective
and practical strategies and measures for prevention and control of communicable
diseases, and have been developing from protecting children to now covering adults.
Chapter 12
Epidemiology of Noncommunicable
Diseases

Jie Yang and Man Li

Key Points
• The growing burden of non-communicable diseases (NCDs) is one of the major
public health challenges facing all countries in the twenty-first century.
• The common risk factor for NCDs are tobacco use, alcohol use, unhealthy diet,
physical inactivity, raised blood pressure, overweight/obesity, etc.
• Prevention of NCDs should integrate the strategies of individual-based high-risk
population and population-based all-population.

12.1 Introduction

12.1.1 Definition

Noncommunicable diseases (NCDs) are the diseases characterized by multifactorial

causation, long latent period, indefinite onset, and noncontagious among individuals.
NCDs include broad types of diseases such as cardiovascular disease, renal disease,
nervous and mental disease, musculoskeletal conditions, chronic non-specific respi-
ratory disease, cancer, diabetes, and various other metabolic and degenerative
diseases. In this chapter, we focus on cardiovascular disease, cancer, and diabetes.
The risk factors of NCDs generally include tobacco use, alcohol use, physical
inactivity, unhealthy diet, and some unhealthy conditions such as overweight/obe-
sity, high systolic blood pressure, high fasting plasma glucose, and high cholesterol
levels.

J. Yang (✉) · M. Li (✉)

School of Public Health, Hebei Medical University, Shijiazhuang, China

© Zhengzhou University Press 2023 213

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_12
214 J. Yang and M. Li

12.1.2 The Influence of NCDs on Health and Society

At the broadest level of the global burden of disease (GBD) hierarchy, NCDs
contributed 73.4% or 41.1 million deaths. At Level 2, the largest numbers of deaths
from NCDs were 17.8 million deaths for cardiovascular diseases, 9.56 million deaths
for neoplasms, and 3.91 million deaths for chronic respiratory diseases. Total
disability-adjusted life years (DALY) from NCDs increased by 36.6% from 1.07
billion in 1990 to 1.47 billion in 2016. In China, NCDs are also leading cause of
deaths and account for 80% of all deaths. For Chinese population in 2017, stroke,
ischemic heart disease, chronic obstructive pulmonary disease (COPD), and lung
cancer were the leading four causes of all-age DALYs in 2017.
Treatment, rehabilitation, and taking care of disability from chronic diseases have
formed tremendous pressure on the individuals, family, society, and health system.
The economic burden of NCDs is huge, involving in not only the huge extra health
care spending of personal, family, and society due to chronic disease, but also the
loss of productivity due to illness, disability, and premature death.

12.2 Epidemiological Features

12.2.1 Overall Global NCDs Outlook

For the time distribution of NCDs, deaths from NCDs increased by 22.7% from 33.5
million in 2007 to 41.1 million in 2017 globally, while the death rate decreased by
7.9% from 582.1 deaths per 100,000 in 2007 to 536.1 deaths per 100,000 in 2017.
Declines in cardiovascular disease and neoplasms are slowing in many high-income
countries. For the place distribution of NCDs, there are significant difference in
incidence, mortality from NCDs between low-, middle-, and high-income countries.
In low- and middle-income countries, communicable disease and maternal and
infant mortality remain at a high level, but deaths from chronic disease are lower
than in high-income countries. Due to the large population in all low- and middle-
income countries, the absolute numbers of chronic disease patients are still higher
than that of those in high-income countries. About three-quarters of chronic disease
death, three-quarters of death from cardiovascular disease and diabetes, 90% of
chronic respiratory disease deaths, and two-thirds of cancer death overall globally
occur in low- and middle-income countries. The age-standardized mortality is not
affected by population size and age composition of the population. In 2012, the
age-standardized mortality of chronic disease in low-income countries (625/100
thousand) and middle-income countries (673/100 thousand) was higher than that
in high-income countries (397/100 thousand).
The trend of incidence and mortality of NCDs between low-, middle- and high-
income countries is different. In some high-income countries, mortality of cardio-
vascular disease and some cancers (e.g., lung cancer) shows a declining trend. For
12 Epidemiology of Noncommunicable Diseases 215

instance, since the 1950s, the age-standardized mortality of heart disease, stroke, and
cancer declined 70%, 78%, and 17%, respectively, in the United States. From 1980
to 2010, compared to the significant reduction of age-standardized mortality of
ischemic heart disease in high-income countries, Eastern Europe, Central Asia,
South Asia, and East Asia have shown significant upward trends. South Asia has a
larger population, and the average death age of ischemic heart disease is younger,
and the years of life lost (YLLs) due to premature death is greatest. Since early
1990s, ischemic heart disease became popular in Eastern Europe and Central Asia,
the crude mortality and age-standardized mortality in these regions are the highest
overall world. In North Africa, the Middle East, and Southeast Asia, the average
death age of ischemic heart disease is younger, and the age-standardized mortality is
higher, which indicates the death of ischemic heart disease is more likely occur in the
labor population.
The early onset of illness is becoming common. Irrespective of gender, persons
from all age groups will be affected by chronic diseases. Chronic disease is common
in older persons. However, the data from 2012 showed that 42% of NCDs death
occurs in the person less than 70 years old (is regarded as premature deaths). The
proportion of premature deaths in low- and middle-income countries (48%) is higher
than that in high-income countries (28%).
For cardiovascular disease (CVD), most countries experienced four stages: low
stage, rising stage, peak stage, and decline stage. Before the 1950s, the social
economy, living, and medical conditions were at lower status, infectious disease
were the major threat to human health. The incidence of cardiovascular disease was
relatively low, and the number of deaths accounted for only from 5% to 10% of total
deaths. Industrialization improved social economy and living condition, which result
in increased nutrition, diet high in sodium, and insufficient physical activity. The
incidence of CVD in the population was rising, and the death due to CVD accounted
for 10–30% of total death (Stage II). High-fat, high-protein, high-calorie diets and
inactive physical activity led to a rapid increase in CVD, especially coronary heart
disease (CHD)and ischemic stroke. The incidence and death appeared a younger
trend, and the death accounted for 35–65% of total death (Stage III). Due to public
health measures such as health education and community intervention and progress
in medicine, the incidence and mortality of CVD declined gradually and the com-
position of death reduced to less than 40% (Stage IV). Most countries and regions
followed the above four stages in CVD epidemic, but different countries enter
different development period, and the current stage is different. For example,
Western Europe, North America, Australia, Japan, and Korea due to high degree
of industrialization, CVD currently entered Stage IV. While Eastern Europe, Russia,
the Middle East, and some fast-growing countries, the mortality of CVD has
increased 50–100% which accounts for 40–60% of total death within recent
30 years. In Asia, Latin America, and Africa, CVD begins to enter Stage II and
the composition of death accounts for under 30% of total death.
For cancer, from a global perspective, the incidence and mortality of cancer are
increasing gradually, except for cervical cancer, esophageal cancer, and stomach
cancer. The epidemiological feature of cancer among countries and regions is
216 J. Yang and M. Li

different. Lung cancer has a higher age-standardized incidence in North America,

the middle region of Western Europe, South Europe, North Europe, and East Asia,
but lowest in the Middle and West Africa. Breast cancer in developed countries
(except Japan) has higher incidence with 89.7 per 100,000 in Western Europe, but
lower incidence in most underdeveloped countries with less than 40 per 100,000.
Generally speaking, high incidence of cancer in developed countries is lung cancer,
breast cancer, colon cancer, and prostate cancer.
For diabetes mellitus, the incidence between types, countries, and races are
different. For type 1 diabetes mellitus, the difference in age-adjusted incidence is
350 times at the global level. Sardinia of Italy (36.8 per 100,000) and Finland (36.5
per 100,000) own the highest incidence, other European and American countries
have middle incidence (from 5.0 per 100,000 to 19 per 100,000), and the lowest
incidence (from 0.1 per 100,000 to 5.0 per 100,000) happens in some Asia countries
(such as China, Japan, and Korea), American Indians, Mexicans, Chileans, and
Peruvians. Type 1 diabetes show the increased incidence with far away from the
equator. For type 2 diabetes mellitus (T2DM), the incidence is related to lifestyle.
Keeping traditional way of life shows lower incidence, while some westernized
developing countries show higher incidence. The prevalence of type 2 diabetes in
rural Africa in adults is from 1% to 2%; however, in North America and Western
Pacific Region, about from 1/3 to 1/2 of adults have been diagnosed as type
2 diabetes.

12.2.2 Epidemiological Features of the Risk Factors of NCDs

It is accepted that a set of “risk factors” are responsible for morbidity and premature
mortality of NCDs. A large percentage of NCDs are preventable through the changes
in these factors, which include tobacco use, physical inactivity, alcohol use,
unhealthy diet, raised blood pressure, overweight/obesity, high cholesterol, cancer-
associated infections, and environmental risk factors.

12.2.2.1 Tobacco Use

The number of men smokers has steadily increased in the first half of the twentieth
century and up to a peak of 80% within several decades after World War II. The rate
of smoking among men began to decline in English-speaking countries and North
European countries, but female smokers began to ascend in these countries at first
and after the last half of the twentieth century, then spread to Japan, Latin America,
central Europe, and south Europe.
In 2012, the smoking prevalence was 22% and had obvious regional difference.
The smoking prevalence is highest in European countries (30%) and is lowest in
Africa (12%). The smoking prevalence among male (37%) is higher than that among
female (7%). In 2010, the incidence of smoking among adult was 28.1% (male
12 Epidemiology of Noncommunicable Diseases 217

52.9% and female 2.4%). Among men, the highest smoking age is 45–64 years old
(63.0%) and the lowest is 14–24 years old. Smokers in rural areas accounted for
56.1% higher than those in urban region (49.2%). Prevalence of second-hand
smoking is up to 72.4%.

12.2.2.2 Alcohol Use

Alcohol use is undoubtly a risk factor for NVDs. It is reported that the harmful use of
alcohol is one of the four behavioral risk factors (tobacco use, unhealthy diet,
physical inactivity, and alcohol use) for three major NCDs (cardiovascular disease,
cancer, and chronic respiratory disease). About 2.3 million people die from the
harmful use of alcohol each year, contributing about 3.8% of the world’s total
deaths. The attributable DALYs is high for alcohol (85.0 millilion DALYs). Adult
alcohol consumption is highest in Europe and America and lowest in Mediterranean
countries and Southeast Asian countries. The heavy episodic drinking within the past
30 days is highest in Europe and America.

12.2.2.3 Unhealthy Diet

Unhealthy diet is a key modifiable risk factor for NCDs, which include inadequate
consumption of fruits and vegetables, excessive consumption of sugar-sweetened
beverages (SSBs), high sodium intake, and high consumption of saturated fats and
trans-fatty acids. Evidence showed that inadequate consumption of fruits and veg-
etables increases the risk of CVD, stomach cancer, and colorectal cancer. High
consumption of SSBs was associated with excess energy intake and was strongly
linked to obesity. People with much higher levels of sodium input than
recommended by WHO are at higher risk for high blood pressure and cardiovascular
disease. High intake of saturated fats and trans-fatty acids has been linked to heart
disease. Unhealthy diet is increasing rapidly in low-resource areas.

12.2.2.4 Physical Inactivity

Physical inactivity means the inability to achieve the recommended levels (at least
30 min of regular, moderate-intensity physical activity on most days) of physical
activity for health. It is the fourth leading cause of death worldwide and is the major
risk factor for NCDs. About 9% of all deaths globally are attributed to physical
inactivity. People who lack physical activity have a 20–30% increased risk of
all-cause death. Physical inactivity is most severe in high-income countries, but it
is significant in some middle-income countries, particularly among women.
218 J. Yang and M. Li

12.2.2.5 Raised Blood Pressure

It is estimated that raised blood pressure cause 7.5 million deaths, about 12.8% of all
deaths. The percentage of populations with raised blood pressure was higher in
regions with lower income level. In low-income and middle-income countries, such
as eastern, western, middle, and southern Africa and Mongolia in Asia, about 30% of
the population had raised blood pressure, but other countries had a lower population
of raised blood pressure.

12.2.2.6 Overweight/Obesity

Since 1980, prevalence of overweight/obesity has a steady rise with the fastest speed
in the United States, with the second fastest is in China, Brazil, and Mexico. In 2014,
the prevalence of overweight (BMI (body mass index) ≥ 25 kg/m2) of adult was 38%
and 40% in male and female, and the prevalence of obesity (BMI ≥ 30 kg/m2)
accounts for 11% and 15%. The highest prevalence of overweight/obesity is in the
American countries (overweight 61% and obesity 27%), and the lowest is in
Southeast Asia (22% and 5%).

12.3 Risk Factors of Several Common NCDs

12.3.1 Cardiovascular and Cerebrovascular Disease

12.3.1.1 Stroke

Stroke, also known as cerebrovascular accident, is an acute cerebrovascular disease,

which including ischemic stroke and hemorrhagic stroke. Damage to brain tissue
occurs when a blood vessel in the brain suddenly bursts or becomes blocked, then
preventing blood from flowing into the brain. Stroke does not occur by chance, and
there are factors that occur several years before stroke. The risk factors are as
follows:
1. Hypertension
Hypertension is the main risk factor for cerebral thrombosis as well as cerebral
hemorrhage. Data from prospective studies showed that the risk of stroke is
increased by 49% with each increase of 10 mmHg in systolic pressure, and is
increased by 46% with each increase of 5 mmHg of diastolic pressure. The
geographic distribution of stroke is consistent with that of hypertension in
morbidity and mortality.
2. Heart disease
12 Epidemiology of Noncommunicable Diseases 219

Heart damage is the second highest risk factor for stroke. In Framingham heart
study, the majority of stroke patients had coronary heart disease, congestive heart
failure, and atrial fibrillation.
3. Diabetes
Diabetes is also an independent risk factor for stroke. The incidence of stroke
in individuals with diabetes is 2.5–3.5 times higher than those without diabetes.
Men with type 2 diabetes are three times of risk in having a stroke than
nondiabetic patients, but women with type 2 diabetes have five times risk than
nondiabetic patients.
4. Dyslipidemia
The incidence of stroke is increased by 25% for every 1 mmol/L increase in
serum total cholesterol. The incidence of ischemic stroke is reduced by 47% for
every 1 mmol/L increase in high-density lipoprotein (HDL).
5. Other factors
Additional factors include obesity, smoking, glucose intolerance, blood
clotting and viscosity, and oral contraceptives.

12.3.1.2 Coronary Heart Disease

1. Hypertension
The prevention of hypertension and the improvement of blood pressure are
essential and fundamental steps for CHD prevention. The famous Framingham
heart study showed that prehypertension and Stage 1, Stage 2, and higher
hypertension increased the risk of CHD in both men and women. In the past,
emphasis was placed on the importance of diastolic blood pressure (DBP). Many
investigators feel that systolic blood pressure (SBP) is a better predictor of CHD
than diastolic pressure. However, both components are significant risk factors. A
meta-analysis showed that the slope of the association between CHD mortality
and normal SBP levels was almost constant across each age range throughout the
normal SBP values drop to lower than 115 mmHg. Furthermore, for the
age-specific harzard ratio between CHD mortality and DBP values drop to
lower than 75 mmHg, it was equivalent to that associated with a 20 mmHg
difference in normal SBP values.
2. Dyslipidemia
It is well known that hyperlipidemia with elevated serum total cholesterol,
low-density lipoprotein (LDL) cholesterol, and triglycerides is a major risk factor
for CHD. The increased serum total cholesterol and low-density lipoprotein and
declined high-density lipoprotein are associated with increased risk of CHD. The
risk of CHD is decreased by 2% for every 1% decrease in serum TG. The
reduction of every 0.03 mmol/L in HDL-C will increase the risk for CHD by
2–3%. The prevalence of dyslipidemia was 75–85% in patients with early onset
CHD, compared with 40–48% in age-matched controls without CHD. Clinical
studies have shown that lowering total or LDL cholesterol can play a better role in
primary or secondary prevention.
220 J. Yang and M. Li

3. Diabetes
CHD is common in patients with diabetes, and its prevalence increases with
worsening glycemic status due to a higher risk of accelerated atherosclerosis and
other lipotoxic and glycotoxic effects. The risk of CHD in people with diabetes is
2–3 times higher than in those without. In industrialized countries, 30–50% of
diabetic among people over the age of 40 years die from CHD.
4. Overweight/obesity
The overall obesity rate among adults was 12.0% in 2015, with higher rates in
women across all age groups. Globally, elevated BMI causes more than four
million deaths and 120 million DALYs each year, most of which are directly
attributable to the subsequent development of cardiovascular disease. In fact,
between 1980 and 2000, higher BMI resulted in approximately 25,905 additional
deaths due to CHD. Compared with normal weight, the relative risk for those
overweight/obesity developing to CHD and death is 1.5–2.0.
5. Tobacco use
Tobacco use has been identified as a major CHD risk factor. Nearly, six
million people die each year from tobacco use, including direct smoking and
second-hand smoking. Data from several studies suggest that the relative risk of
CHD is 2–3 times higher among smokers. These risks have age gradient, with
higher relative risk in the younger age group (5–6 times).
6. Physical inactivity
A sedentary lifestyle is associated with the risk of early CHD development.
There is evidence that regular physical activity reduces body weight and blood
pressure and increases the HDL level, which are beneficial for cardiovascular
health. In a meta-analysis of 43 prospective cohort studies, compared with
600–3999 MET -min/week of total physical activity across all domains, the RR
among people <600 MET-min/week increased by 19%. A total of 5.0% CHD
deaths can be attributed to physical inactivity.

12.3.2 T2DM

12.3.2.1 Genetic Factor

T2DM has strong family aggregation. The prevalence of diabetic relatives is 4–8
times higher than that of nondiabetic relatives. Twin studies have shown a consis-
tency of about 90% in identical twins with T2DM, thus demonstrating a strong
genetic component. The heritability of T2DM in China is 51.2–73.8%. In addition, a
person’s risk of developing diabetes may also depend on the genetic susceptibility to
diabetes. Many genome-wide association studies have investigated genetic variants
in different populations that influence disease susceptibility through rare alleles and
common variants. For example, Chauhan et al. showed the association of eight gene
variants (PPARγ, KCNJ11, TCF7L2, SLC30A8, HHEX, CDKN2A, IGF2BP2, and
CDKAL1) with diabetes in Asian Indians.
12 Epidemiology of Noncommunicable Diseases 221

12.3.2.2 Overweight/Obesity

Overweight/obesity, particularly central adiposity, has long been accepted as a risk

factor for prediabetes or type 2 diabetes (T2MD). Overweight/obesity leads to
inflammation, endoplasmic reticulum stress, and fat factor, all of which occur in
the pathogenesis of insulin resistance of the liver and skeletal muscle in the work,
and increase the risk of diabetes. In a meta-analysis of 84 articles involving more
than 2.69 million participants, the combined prediabetes risk of overweight/obesity
versus normal weight was 1.24. Based on the race-specific BMI classification, the
combined risk for type 2 diabetes relative to normal weight was 0.93 for under-
weight, 2.24 for overweight, 4.56 for obese, and 22.97 for severely obese. The RR of
T2DM in overweight/obesity decreased with age. Another meta-analysis showed
that obesity in children and adolescents was positively associated with the preva-
lence of T2DM and prediabetes, in which obese subjects 13 times higher than
normal-weight subjects. The prevalence of prediabetes was three times higher in
obese subjects (17.0% vs 6.0%, respectively).

12.3.2.3 Physical Inactivity

Sedentary lifestyle is an important risk factor for the development of T2DM. Lack of
exercise may alter the interaction between insulin and its receptors, which can lead to
T2DM. Those who watched TV for 4 h a day had a 46% higher risk of developing
diabetes less than 1 h a day. Meta-analysis of 55 prospective cohort studies <600
MET-min/week versus 600–3999 MET-min/week of total physical activity across
all domains: 1.17 (1.11–1.23) 4.5 (3.1–6.0) type 2 diabetes 2.7 (1.9–3.5) 4.2
(2.9–5.7) 5.9 (4.2–7.7).

12.3.2.4 Unhealthy Diet

Both food calories and the quality of diet components affect the risk of diabetes.
Excessive calorie intake can increase the overweight, with the passage of time, the
metabolism of liver glucose control and steady state would be destroyed. Poor
dietary quality, such as low intakes of dietary fiber, low-sugar carbohydrates, or
whole grain grains, increases the risk of diabetes, as does high intakes of saturated
fatty acids and trans fats. A diet containing high-quality fats and carbohydrates rather
than low-quality fats and carbohydrates is more important than the relative amounts
of these nutrients in preventing type 2 diabetes.
222 J. Yang and M. Li

12.3.2.5 Malnutrition

Malnutrition in early infancy and childhood or undernutrition early in life (e.g.,

exposure to famine) can lead to partial beta cell failure, impaired carbohydrate
tolerance, and increases the risk of developing type 2 diabetes later in life.

12.3.2.6 Impaired Glucose Tolerance (IGT)

IGT is an intermediate state between normal and diabetes. The IGT patients have a
higher prevalence of diabetes. When IGT patient was followed up to 5–10 years after
the first diagnosis, about one-third of the individuals have developed into diabetes,
one-third were converted to normal blood glucose and one-third remained IGT
status. IGT is easily converted to diabetes when accompanied by the following
factors: fasting blood glucose, 2-h blood glucose, and BMI more than 5.0 mmol/L,
9.4 mmol/L, and 25, respectively. Improved diet and increased physical activity are
beneficial in reducing the chance of IGT conversion to diabetes.

12.3.2.7 Insulin Resistance

Clinical studies found that insulin resistance occurred in obesity, type 2 diabetes,
hyperlipidemia, hypertension, coronary heart disease, stroke, etc. Blood insulin
plays a vital role in the process of diabetes development from normal or IGT. Insulin
resistance is a common pathophysiological mechanism in above pathological
processes.

12.3.2.8 Maternal Diabetes

Offspring of diabetic pregnancies, including gestational diabetes, tend to be large

and heavy at birth, tend to develop obesity in childhood, and have a high risk of
developing T2DM in early life. The risk of diabetes was three times higher in
children born to mothers with diabetes than in children born before their mothers.
Maternal diabetes, which is associated with intrauterine growth retardation and low
birth weight, appears to increase the risk of later diabetes in children if it is associated
with subsequent rapid growth catch-up.
12 Epidemiology of Noncommunicable Diseases 223

12.3.3 Cancer

12.3.3.1 Physical Factors

Ionizing radiation (X, γ, α, β-ray, etc.) can cause a variety of human cancer including
lung, breast cancer, leukemia, multiple myeloma, thyroid cancer, skin cancer, etc. In
occupational factors, other physical factors such as asbestos fiber, coal dust, and
quartz dust can result in lung cancer and mesothelioma.

12.3.3.2 Tobacco Use

Many large prospective studies have provided that tobacco use increases the risk of
cancer mortality, especially lung cancer. Tobacco use also leads to larynx, oral, head
and neck, pharynx, esophagus, bladder, pancreas, cervical, breast, and probably
kidney cancer. A prospective studies reported that the age-adjusted incidence rate
of cancer is highest in current smoker, with the RR is up to 12.0 for lung cancer. The
attributable risk for oral-bladder cancers, other cancers, and all cancers were 46%,
16%, and 29%, respectively. Another systematic review showed that men who are
current smokers have a moderately increased risk of total cancer compared to those
never smoked. In women, the risk is increased but less than in men. The overall
relative risk was estimated at 1.53.

12.3.3.3 Alcohol Use

Excessive intake of alcoholic beverages is associated with oral, pharyngeal, esoph-

ageal, liver, colon, rectal, and breast cancer. With the exception of the American
Cancer Society and the Canadian Cancer Society, all organizations state that alcohol
is a class carcinogen and that even small amounts of alcohol can increase the risk of
certain cancers. However, some studies showed that light or very light alcohol use
was not associated with the risk of common tumors, except for mild increases in
breast cancer in women and colorectal cancer in men.

12.3.3.4 Dietary Factors

A lot of studies suggested the potential role of diet in certain cancers. However, there
is no guarantee of cancer prevention. The study of diet and cancer risk reduction is
complicated not only by the multistage, multifactor nature of the disease, but also by
the inherent complexity of any diet. Prospective cohort findings support an associ-
ation between unhealthy eating patterns and increased risk of colon and breast
cancer, particularly in postmenopausal hormone-receptor negative women. The
limited evidence of an association between unhealthy dietary patterns and the risk
224 J. Yang and M. Li

of upper digestive tract, pancreatic, ovarian, endometrial, and prostate cancers relies
only on case-control studies.

12.3.3.5 Occupational Exposures

Occupational exposure includes exposure to benzene, arsenic, cadmium, chromium,

vinyl chloride, and asbestos polycyclic hydrocarbons. The risk of occupational
exposure is greatly increased if individuals also smoke. It has been reported that
occupational exposures typically account for 1–5% of all human cancers.

12.3.3.6 Biological Factors

Biological factors are one of the main causes of human tumor. By now, it is
identified that at least eight viruses have been linked to human tumors. For example,
the increased risk of hepatitis B, C virus for hepatocellular, human immunodefi-
ciency virus for Kaposi’s sarcoma, the Epstein–Barr virus for Burkitt’s lymphoma
and nasopharyngeal carcinoma, and human papilloma virus for cervical cancer.

12.3.3.7 Genetic Factors

It is now becoming clear that individual differences in the incidence of tumors are
related to genetic background, which means that the occurrence of tumors is also
related to the individual’s own genetic susceptibility. Although there is probably a
complex interrelationship between hereditary susceptibility and environmental car-
cinogenic stimuli in the causation of a number of cancers. With the completion of the
Human Genome Project and the rapid development of high-throughput gene varia-
tion detection methods, genome-wide association studies (GWAS) have become a
major strategy for revealing tumor susceptibility genes. In recent years, tumor
researchers around the world have used GWAS strategy to conduct a series of
studies on nasopharyngeal carcinoma, liver cancer, esophageal cancer, lung cancer,
pancreatic cancer and other tumors in people around the world, and a large number
of genetic variations and genetic loci of tumor-related chromosome regions have
been discovered, which is of great significance for fully revealing the causes of
tumor occurrence.

12.3.3.8 Other Factors

Other factors include the immune, endocrine, and psychosocial factors. The immune
system is closely related to the incidence of cancer. Tumor cells can evade immune
system attacks by one or more mechanism or cannot activate specific antitumor
immunity and induce the tumor development. Endocrine-related tumors include
12 Epidemiology of Noncommunicable Diseases 225

breast, ovarian, and testicular cancer. The risk factors for breast cancer include
non-procreation, early onset, late menopause, and non-lactation. Social psycholog-
ical factors are also one of the important risk factors for cancer. Major adverse events
and depression can cause psychological stress, which lead to the disturbance of the
nervous system and the decline of immunity.

12.4 Prevention and Control of NCDs

12.4.1 Prevention Strategy

The prevention and control of NCDs emphasize the primordial prevention, which
controls the risk factors at the population level. Some of the risks of adult chronic
disease begins with adverse exposure in pregnancy. Many unhealthy lifestyles are
formed from childhood. Once formed, it is difficult to change. Therefore, the
prevention of NCDs takes the life-course approach which is from early life through
the whole life period.
Prevention of NCDs should integrate the strategies of individual-based high-risk
population and population-based all-population. When the risk factors exist among
the whole population, the all-population strategy is particularly important. Smoking
ban in public places and workplaces is a successful strategy for the all-population
strategy.
Member States in the WHO Western Pacific Region endorsed the “For the
Future” Vision at the Regional Committee Meeting in 2019. The paper set out
four thematic priorities for making the Western Pacific the healthiest and safest
region in the world, one of which is NCDs and aging. The burden of NCDs and
related risk factors is a major barrier to the development and achievement of the
sustainable development goals (SDGs), the WHO’s Global Action Plan for the
Prevention and Control of NCDs 2013–2020 and the “For the Future” Vision.
Based on the Global Action Plan for the Prevention and Control of NCDs
2013–2020, WHO provides a list of “Best Buys” and other recommended interven-
tions in 2017 for the four key risk factors for NCDs (tobacco, harmful use of alcohol,
unhealthy diet, and physical inactivity) and for four diseases (cardiovascular disease,
diabetes, cancer, and chronic respiratory disease) to tackle global NCDs problems.
In order to maintain people’s health and build a healthy China in a well-rounded
way, China has recently issued a number of policies, plans, and national actions for
the prevention and control of NCDs, which include the National Basic Public Health
Service Projects (NBPHSP) to manage hypertension and diabetes in primary health
facilities since 2009, community-based and comprehensive intervention projects for
NCD, the China Healthy Lifestyle for All (Phases I and II) launched in 2007, the
Medium- and Long-Term Plan for the Prevention and Treatment of NCDs
(2017–2025), the Outline of the Plan for “Healthy China 2030,” and the Healthy
China Action (2019–2030) promulgated by the Chinese State Council. Early in
March 2021, China issued the 14th Five-Year Plan (FYP), a high-level development
226 J. Yang and M. Li

blueprint for the next 5 years. The 14th FYP calls for “fully implementing the
Healthy China Actions,” “strengthening prevention, early screening and compre-
hensive intervention of chronic diseases,” etc.

12.4.2 Prevention Measures

There are a variety of measures for the prevention of NCDs. Urgent action is needed
to reduce the growing burden of NCDs and prevent the annual toll burden that dying
prematurely before the age of 70 from heat and lung disease, stroke, cancer, and
diabetes. There is not only a growing awareness and concern about the burden of
NCDs on families, individuals, and public health, but also the social and economic
burdens associated with the NCDs. Many interventions for prevention and control of
NCDs exist. Even in the richest countries, it is essential to choose which interven-
tions to prioritize because resources are limited, and this is especially true in most
countries. In 2017, WHO launched Best Buys, which recommended three types of
intervention measures, including “the most cost-effective measures,” “effective
interventions with cost-effectiveness ratio of more than $100,” and “other interven-
tions (without cost-effectiveness analysis),”, for four chronic behavioral risk factors,
tobacco use, unhealthy diet, insufficient physical activity, and harmful use of
alcohol.
For tobacco use, the interventions including increased tobacco excise taxes to
reduce the affordability of tobacco products, implement plain/standardized packag-
ing and/or large graphic health warnings on all tobacco packages, eliminate second-
hand smoke exposure in all indoor workplaces, public places, and public transport,
ban cross-border advertising, and offer smoking cessation services to all who want
to quit through mobile phone apps. For unhealthy diet, the interventions include
reducing salt input, limiting food package sizes and portion sizes, reducing sugar
consumption through effective taxes on sugary beverages, and promoting unsatu-
rated fats instead of trans and saturated fats through formulation, labeling, fiscal or
agricultural policies, etc.
For physical inactivity, the interventions include reducing physical inactivity,
promoting physical activity and national fitness, promoting travel and domestic
physical activity, reducing static behavior, providing physical activity counselling
and referrals within routine primary health care services using short-term interven-
tions, etc.
For alcohol use, the interventions including increased excise taxes on alcoholic
beverages, complete bans or restrictions on alcohol advertising, restrict alcohol use
and promote health education, prevention, treatment, and care of alcohol use disor-
ders and their comorbidities, etc.
Chapter 13
Epidemiology of Public Health Emergencies

Hong Zhu

Key Points
• Public health emergencies have enormous impact on population health.
• Public health emergencies can be divided into natural disasters, man-made
disasters and disease outbreaks.
• Epidemiology plays a crucial role in the management of public health
emergencies.
• Epidemiologic investigation is usually the first step when disease outbreak and
disaster occurs.
• Response to public health emergencies is inextricably linked to advance
preparedness.
The health impacts of public health threats, such as emerging infectious diseases
(e.g., 2003 SARS epidemic, 2009 influenza A pandemic, and 2016 Zika outbreak),
terrorism (e.g., 2001 World Trade Center bombing), environmental catastrophes
(e.g., 2011 Fukushima Daiichi nuclear disaster), and natural disasters (e.g., earth-
quake), have demonstrated the importance of strengthening the public health sys-
tems and improving the community’s ability to respond effectively. Epidemiology is
critical for the management of public health emergencies. This chapter introduces
the application of epidemiology in the investigation of, preparation for, and response
to public health emergencies.

H. Zhu (✉)
School of Public Health, Tianjin Medical University, Tianjin, China
e-mail: [email protected]

© Zhengzhou University Press 2023 227

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_13
228 H. Zhu

13.1 Basic Conception of Public Health Emergencies

Throughout 2018, altogether 484 public health events were recorded in WHO’s
event management system (a 16% increase from 2017), of which 352 (73%) were
attributed to infectious diseases, 47 (10%) to disasters, and 19 (4%) to food safety.

13.1.1 Definition of Public Health Emergencies

A “public health emergency” refers to a sudden-onset natural or man-made event,

which poses a risk to public health, like infectious disease outbreak, bioterrorist
attack, severe food poisoning or industrial poisoning, or other significant or cata-
strophic events. A public health emergency, being an important part of all kinds of
emergent events, seriously affects the health of a certain population and needs multi-
sectoral cooperation co-assistance to cope.
Two terms “disaster” and “accident” are easily confused with “public health
emergency.” The United Nations Disaster Relief Organization (UNDRO) defines
disaster as “a serious disruption of the functioning of a society, causing widespread
human, material, or environmental losses which exceed the ability of the affected
society to cope using its own resources.” This definition shows that disaster has
broader influences not only on health, but also on social stability and economic
development. Many experts agree that the disasters which pose a threaten to human
life can be called public health emergencies, while those posing danger to environ-
ment or material, rather than human health, are not public health emergencies, such
as volcanic eruption in remote areas or the 1998 Asian financial crisis.
The other term “accident” means an unpleasant event that happens unexpectedly
and causes injury or damage. Accidents can usually affect individuals or groups,
while public health emergencies always affect groups. Besides, the accident cannot
be anticipated or predicted. However, different from accidents, some types of public
health emergencies can be predicted, for example, flood, and the activities on risk
assessment, early warning, and preparation will mitigate the consequences of pre-
dictable hazards.

13.1.2 Characteristics of Public Health Emergencies

1. Public health emergency is a sudden-onset event. A public health emergency, as a

kind of emergencies, has the characteristic of emergency, that is, it occurs
suddenly and unpredictably, and its onset and development are hard to predict.
However, now, development in science has made it possible for humans to
predict more and more disasters, like early warning of floods, forest fires, and
infectious diseases.
13 Epidemiology of Public Health Emergencies 229

2. Public health emergency has a great impact on human health. The victims of
public health emergencies are not limited to some certain individuals, but rather a
large population or substantial proportion of people. The term also includes
events that may pose a potential health threat, such as exposure to infectious
agents, contaminated water, or food that may cause harm to humans.
3. Public health emergencies caused by different reasons have different character-
istics, and correspondingly have different treatment and management strategies.
For example, for the outbreak of an infectious disease, the main aim of the
investigation is to identify the pathogen and transmission route; while for a
natural disaster, the main aim of the investigation is to assess the situation rapidly.
4. Immediate action and unconventional measures should be taken to deal with
public health emergencies. Multi-sectoral and multinational cooperation is usu-
ally needed. Health-care services, together with administrative institutes, media,
military, traffic agencies, academic institutes, etc., may deal with emergencies
more effectively and efficiently.

13.1.3 Classification of Public Health Emergencies

There are many different classifications of public health emergencies. The most
commonly used method is based on the cause(s) of emergencies.
1. According to the nature of these events, public health emergencies can be divided
into as follows:
Biological emergencies: These include communicable diseases, biological
agent-related terrorisms, and vaccine inoculation-related events.
Chemical emergencies: These include leaks of hazardous chemicals, inten-
tional or unintentional chemical food poisoning, and the use of chemical agents in
terrorist incidents.
Radiological emergencies: These include the release of harmful radiation
caused by the explosion of a nuclear weapon or improvised nuclear device.
Weather and home emergencies: These include the threats caused by abnormal
meteorological conditions, like thunderstorms, flooding, tornado, and extremely
hot or cold weather, and the threats from home, like kitchen fire, gas leak or
explosion, and carbon monoxide poisoning.
2. According to the originating source of the disaster, public health emergencies can
be divided into as follows:
Natural disasters: Natural disasters include weather phenomena (such as
tropical storms, tsunamis, avalanches, extreme temperatures, winds/typhoons/
hurricanes, and floods) and geologically related disasters (like earthquakes,
landslides, and volcanic eruptions).
Man-made disasters: Man-made disasters include industrial accidents, traffic
accidents, pollution incidents, terrorism, and armed wars. Natural disasters have
long been considered the ones that cause the most deaths and economic losses,
but man-made disasters are becoming more prominent.
230 H. Zhu

Epidemic diseases: Disease epidemics or outbreaks would threaten the health

of a certain population. These diseases are usually infectious or communicable,
such as cholera, measles, hepatitis, influenza, malaria, SARS, H1N1, and HIV.
They can spread among population through different routes of transmission,
including air, food, water, direct or indirect contact, and insect or animal vectors.
It is worth noting that the risk of disease outbreak usually increases following
natural or man-made disaster (i.e., a flood), mainly due to poor sanitation and
overly dense populations.

13.1.4 Phases of Public Health Emergencies

Emergency situations always change dynamically and require changes in response

accordingly at different phases. The whole process of a public health emergency is
often thought of as a cycle, which consists of six phases including preparedness
phase (pre-emergency phase), warning phase, impact phase (emergency phase),
response phase, reconstruction (rehabilitation) phase, and mitigating (preventing)
phase. While the stages of dynamic change appear to be continuous, clearly identi-
fying the end of each phase facilitates the adjustment of response strategies and
better accommodation of new demands.
The whole process of an emergency may be divided into four phases: impact
phase, response phase, recovery phase, and mitigation phase. Sometimes, it also
includes a warning phase. Impact phase is the stage when the disaster causes real
harm to the population. The duration of this phase depends on the number of people
affected, and the type of incident. It may last for a few minutes (e.g., an earthquake),
or several days (e.g., a flood), or several months (e.g., a disease outbreak). A
response is made following the impact phase of an emergency. Preparedness actions
taken in a timely manner prior to an emergency can be extremely helpful when
facing an emergency. Relief activities (such as patients’ treatment or victims rescue)
occur during the response phase, which is followed by the reconstruction (rehabil-
itation) phase. During this phase, the focus is no longer on relief, but rather on
development, which aims to help the affected people self-reliant. Besides, the
lessons learned from the emergency are applied to prevent the recurrence of such
disasters or to reduce the harm caused by such disasters, and to adequately prepare
for such disasters should it recur.
The above process helps to formulate emergency response strategies and pro-
tocols. This concept suggests that four phases occur successively and unidirection-
ally. Actually, however, many things may occur at same time. The cycle concept
ignores the reality that the serious consequence caused by emergencies may extend
well beyond efforts at the reconstruction phase. The weakness of the efforts at
reconstruction and mitigation phases can easily be magnified by subsequent emer-
gencies (Fig. 13.1).
13 Epidemiology of Public Health Emergencies 231

MITIGATION PREPAREDNESS
(Prevent disaster or reduce its harmful (Planning, training, exercises,
impact, surveillance ) resource acquisition, and researching)

RECOVERY RESPONSE
(Restore vital structures, systems, (Activate plans, rescue injured,
services; restore public confidence deploy assets during event)
and health)

Fig. 13.1 Phases cycle of public health emergencies

13.1.5 Harm Caused by Public Health Emergencies

Public health emergencies, especially those occurring in large scales not only result
in human mortality, but also cause physical, psychological, and social disabilities.
1. Physical harm: Public health emergencies may cause deaths, injuries, and some-
times malnutrition. The impact phase is the primary phase of an emergency
resulting in serious injury or death. When a public health emergency occurs,
the most important and urgent task is the treatment of the wounded or patients.
2. Psychological harm: Psychological trauma is a unique individual mental experi-
ence caused by a serious injury event, especially when enormous pressure derived
from the event is beyond one’s ability to cope. Psychological effects of fear,
helplessness, anxiety, depression, and terror caused by emergencies might linger
for years, even decades, hence, psychological rehabilitation must be done on a
long-term basis.
3. Economic and social disabilities: Natural disasters or terrorist attacks may lead to
social disruption and infrastructure damage, such as transportation, residential
building, enterprise assets, and electricity. It will directly or indirectly reduce the
profits of the economy and reduce the speed of economic development, and
ultimately affects social stability. Usually, developing countries are more vulner-
able to emergent events than developed ones.
4. Environmental and ecological harm: Some kinds of emergencies, like volcanic
eruptions, chemical or radiological releases, may affect the environment and
disrupt the ecological balance in a certain geographic area, which in turn leads
to harmful effect on human.
232 H. Zhu

13.2 Basic Principles and Application of Epidemiology

in Public Health Emergencies

13.2.1 Role of Epidemiology in Public Health Emergencies

Epidemiology of public health emergencies may be defined as the application of

epidemiology in public health emergency preparedness and response, with the aim
of determining the nature of the events, exploring the causes and risk factors of the
events, identifying the high-risk population, developing and assessing the health
programs to prevent and control harmful effects caused by emergencies.
From an epidemiological perspective, policymakers and practitioners can focus
on the main issues of the entire affected population rather than on individuals, and to
further develop measures to improve the health of the entire community. For public
health personnel, epidemiological investigation is usually the first step when the
disease or the disaster occurs.
Table 13.1 summarizes the epidemiological tools and principles applicable to
responding to public health emergencies, with the aim of containing the progress of
events, and reducing the harmful impacts on the whole populations.

13.2.2 Key Epidemiological Indicators

In epidemiology, the occurrence of disease or disaster is not in a random way;

instead, follows a specific pattern which can be studied and predicted with respect to
“what, who, where, when, how, why, whom, and what next.” Hence, some indictors
can be used to describe this pattern, evaluate the impact of disease or disaster, and to
assess the effectiveness and efficiencies of intervention programs. Table 13.2

Table 13.1 The application and goals of epidemiology in public health emergencies
Application of
epidemiology Goals of epidemiology
Demand assessment Evaluating the size and structure of the affect and potentially affected
population, identifying the priority health issues and the health-related
demands (medical devices, staffs, etc.) in the community
Population surveys Determining the mortality and morbidity of disease (death rates and
incidence/prevalence) or the number of injured; determining health
status (nutrition and immunization status)
Public health Monitoring health trends of the community; early predicting and
surveillance warning
Disease outbreak Identifying the causes or risk factors of the disease; determining the
investigation source of infection, the route of transmission, and the susceptible
population
Program evaluation Assessing the coverage and the impact of health programs; economic
evaluation
13 Epidemiology of Public Health Emergencies 233

Table 13.2 Epidemiological indicators for evaluating emergency intervention programs

Indicator Examples
Public Degree of protocol commitment
strategies 1. Follow the diagnostic criteria for the case and treatment protocols
2. Degree of public and community involvement
3. Degree of cooperation and coordination between inter-sectors
4. Equity and accessibility of resource allocation
Demographics Estimated number and structure of the affected population
1. Age and sex proportion
2. Population mobility and migration
3. Proportions and characteristics of high-risk and vulnerable groups
4. Ratio of urban and rural population
Health status Rate of disease and death
1. Incidence and prevalence of common diseases or the infectious disease to be
studied, secondary attack rate
2. Death rate (crude, age- or sex-specific, infant, under-fives, maternal) and
fatality rate
3. nutritional status especially among under-fives
Program Input of the following resources:
inputs 1. Government financial guarantee funds
2. Facilities and equipment (health centers, beds, medicine, sanitize devices)
3. Staff (professional staff, volunteers, army, firefighters)
4. Basic supplies (food, water, shelter material, daily necessities)
5. Ancillary resources (fuel, charcoal, transport, communication)
Program Access, coverage, and quality of the following services:
process 1. Daily necessities sanitation
2. Environmental sanitation (feces/garbage disposal, disinfection)
3. Coverage of vaccination and prophylaxis
4. Demand for and utilization of health services (two-week attendance rate,
hospitalization rate)

summarizes the epidemiological indicators that can be used to evaluate an emer-

gency intervention program’s process and outcome.

13.2.3 Outbreak Investigation

13.2.3.1 Purpose of Outbreak Investigation

A disease outbreak (or epidemic) refers to the sudden occurrence of similar disease
among many people in a region or a community due to the same source or carrier,
with the incidence rate higher than normal expectations. Epidemiological investiga-
tions can be used to ascertain the nature of disease epidemics, for example, the cause
of diseases (why), transmission routes and vectors (how), disease distribution by
place, time and population (what, where, when, who), susceptible population
(whom), and disease trend (what next). Determination of these natures of disease
outbreak is critical to identify effective and proper clinical and public health
234 H. Zhu

interventions. The objectives of outbreak investigation are to determine the existence

and the severity of the outbreak, to explore the cause(s) and risk factors of the
disease, to identify high-risk groups who are at higher risk of being affected and who
would benefit most from interventions, and to identify and evaluate effective inter-
ventions to slow down the spread of the disease and reduce the harm caused by the
outbreak.

13.2.3.2 Three Elemental Epidemiological Designs in an Outbreak

Investigation

In the epidemiological investigation of disease outbreaks, three elemental epidemi-

ological designs can be used:
1. Descriptive epidemiology, especially cross-sectional design, can be used to
describe the characteristics of disease distribution and to identify the difference
in disease frequency among different regions and populations, which provide
clues to the establishment of etiological hypotheses. Multiple cross-sectional
studies can also provide information on temporal trends of disease outbreak.
2. Analytical epidemiology, mainly including case-control design and cohort
design, can be used to explore the risk factor (e.g., environmental and behavioral
factors) of disease by comparing the exposure proportion between cases and
non-cases (case-control design) or by comparing the incidence rate between
those with or without a certain exposure (cohort design).
3. Experimental epidemiology, including randomized controlled trial, filed trial, and
community intervention trial, can be used to assess the safety and efficacy of a
certain intervention, such as a new medicine, a new vaccine, or a new public
health program. For example, evaluating a cholera prevention program by com-
paring the incidence of cholera between two communities with or without
initiating this program.

13.2.3.3 Key Steps in Carrying Out Outbreak Investigation

1. Confirm the outbreak: All reports of the suspected outbreak by health-care

workers or victims themselves require an immediate response from the relevant
health authorities. The first step is to confirm the existence of the outbreak by
local public health teams (e.g., staffs in CDC) through field investigation. Some-
times, besides epidemiologist, other specialists, such as microbiologists, zoolo-
gist, internists, and environmentalist, are needed to cooperate with the field
investigation.
Initial investigations provide first-hand information on disease outbreaks, which
is helpful for the follow-up investigation. Besides, the findings can also be used to
develop diagnostic criteria for the disease, which is essential for the further
investigation and clinical treatment.
13 Epidemiology of Public Health Emergencies 235

2. Define a “case”: The main work at this step is to define a “case.” Health workers
use this standard case definition to determine whether a person has a certain
disease. Based on this case definition, investigators can identify how many people
are cases (the numerator), and how many people are at risk of the occurrence of
this disease (the denominator). Subsequently, attack rates (for disease outbreak)
or incidence rate (for disease epidemic) can be calculated and compared with
previous average rate level. Please note that data on the whole population is
always needed as the denominators in the calculation of incidence rate or
death rate.
The definition of the case includes suspected case, clinically diagnosed case,
and a laboratory-confirmed case. In the early stage of the investigation, due to the
lack of laboratory equipment for diagnosis, most cases are diagnosed based on
epidemiological and clinical information. At this stage, more sensitive case
definition (e.g., suspected case) is helpful to find more cases. In the middle
stage of investigation, more specific definition (e.g., clinically diagnosed or
laboratory-confirmed case) should be used in order to identify real patients and
explore risk factors of the disease through case-control study or cohort study. In
the end stage of the investigation, the definition used should be feasible and
proper for disease surveillance, with the aim of assessing the effect of outbreak
control activities.
3. Describe the outbreak by place, time, and person: Epidemic curve, which shows
the number of diseases (y-axis) in an outbreak over time (x-axis), provides key
information about an outbreak, including time trend (how quickly it is growing,
and whether it is ongoing), what are the potential sources of disease (single or
multiple sources), and how long the incubation period is. Dotted map which
graphs the location of all reported cases can help to identify regional distribution
of the outbreak and disease spread direction. Rates calculated by age and sex
provide information on the most susceptible subpopulations and causal clue.
Investigators can access to public health surveillance system to get data for rate
calculation.
4. Analyze what caused the outbreak: Explore key differences between the
non-cases and cases to determine which groups or individuals are more suscep-
tible to disease, what are the potential risk factors, and what are the possible
sources of disease and routes of transmission. Biological specimens are also
collected from two groups for experimental test. If there has been some evidence
that certain exposure is the potential cause of the disease outbreak, historical or
prospective cohort design can be used to identify whether the persons with the
exposure has the higher incidence rate than those without the exposure.
5. Assess environment: If disease outbreak is speculated to be related to environ-
mental factors, such as animals or vectors, fecal contamination, or toxic
chemicals, assess environment becomes necessary. Animal hosts or vectors
should be investigated and some abnormal phenomenon should be observed
and reported, especially for animal-borne diseases.
6. Initiate and improve prevention and control strategies: The ultimate goal of
epidemiological investigation is to prevent and control disease outbreak. Hence,
236 H. Zhu

intervention strategies and programs should be developed and launched based on

the available information on disease outbreak. A clear understanding of the
outbreak characteristics is needed for effective prevention and control. This
involves three aspects: morbidity and mortality of the diseases, three elements
of disease epidemic (source of infection, route of transmission, susceptible
population), and context information (the amount and availability of medical
resources, administrative divisions, etc.).
The main countermeasures against disease outbreak include controlling the
source of infection (isolation and treatment of confirmed cases, management of
asymptomatic carriers and animal hosts), curb disease spread (personal hygiene,
environmental disinfection, health education, avoid gatherings, vector control,
and entry-exit health quarantine) and protecting susceptible individuals (vaccina-
tion, nutrition, personal protection, and chemoprophylaxis). All these control
measures should be established in accordance with national disease control
regulations and policies.
Note: Taking control measures as early as possible is the most effective way to
prevent the spread of disease. Hence, the process of outbreak investigation and
intervention must be conducted rapidly and simultaneously. When little is known
about the cause of disease, nonspecific measures for control of communicable
diseases can be taken, such as isolation of the suspected patients, environmental
disinfection, and personal protection. With the further understanding of the
disease, prevention and control strategies will be improved correspondingly,
and more targeted measures can be introduced (e.g., immunization).
7. Summarize an outbreak investigation and write a report:
Summarizing the whole process of the investigation and sharing the experi-
ence with all involved is the last step when the outbreak ends. The report should
include the source of infection and possible routes of transmission, possible
causes of the outbreak, case characteristics and clinical symptoms, geographical
distribution, time trends, and lessons learned from epidemic control.

13.2.4 Disaster Investigation

13.2.4.1 Purpose of Disaster Investigation

In contrast to disease outbreaks, the cause and the harmful effects of most disasters
(natural or man-made) are relatively easy to identify and diagnose. Hence, the
objectives of investigation in disease outbreak and disaster are different. In the
event of disasters, investigation is conducted mainly for identifying the amount
and the trend of harmful impact, and corresponding demands for staff, materials,
and services. For example, it is important to understand the need for intervention and
to determine the size and type of the intervention as soon as possible. Unfortunately,
several weeks are usually needed to collect and organize precise and reliable data.
Hence, the urgent work for disaster investigation is to conduct a rapid needs
13 Epidemiology of Public Health Emergencies 237

assessment to get less-precise and less-reliable data, based on which crucial deci-
sions are made. Besides rapid needs assessment, further investigation is also needed
to collect more detailed, precise, and complete information, which is essential for the
reconstruction of the community.
A rapid needs evaluation should be initiated as soon as possible, preferably within
the first 3 days after the event. The objectives of rapid needs assessment are to assess
the amount of affected population, disaster situation, local rapid response capacity,
secondary disaster risk, resources needed, and the recommended actions.

13.2.4.2 Key Steps in Carrying Out Disaster Investigation

1. Preparing for rapid health needs assessment

Prior to rapid assessment, adequate preparation is required, such as collect
background information about the emergency location, inspect safety conditions
at the site of disaster, contact with local authorities and relevant organizations,
and prepare essential equipment and supplies. Besides, a plan for the field
assessment should be made to determine the order and manner of information
collection, the forms for recording and analyzing the collected information, time
schedule, and tasks assignment to each team member.
2. Making checklists for rapid needs assessment
Making a list of information to be collected during the preparation phase
ensures the quality and comprehensiveness of needs assessment. A number of
rapid assessments checklists are available. The checklist chosen should be
adapted to the specific culture and background of the emergency.
3. Identifying methods and sources for collecting data
The main sources of information include health records at local health facil-
ities; satellite maps; local pictures and videos; news media and social media
information; vital statistics; medical insurance records; and field interviews with
local health workers, health officials, or affected population. Please note that the
interviewees should be selected from different demographic characteristics, such
as gender, age, and race.
The method of data collection is determined by the type of the emergency,
financial, material, and staff resources for the assessment as well as time require-
ments. Both qualitative and quantitative methods can be used, such as reviewing
past records, observation, focus group discussions, key informant interviews,
patient narratives, and questionnaire.
4. Conducting rapid health needs assessment
The following steps may provide a logical approach to rapid health needs
assessment:(a) Conduct preliminary observation when approaching the site to
evaluate the extent of population displacement and the damage caused by the
disaster. (b) Interview local authorities and personnel, such as local government
leaders, public health workers, volunteers, as well as people in affected commu-
nities to identify water and food supplies, medical services, and demographic
characteristics. (c) Review existing records, such as maps, census data, and
238 H. Zhu

surveillance system data (d) Conduct detailed visual inspection, like field surveys
around affected communities to collect information on the structure of camps,
living conditions and sanitation, population movement and migration, social
influences and reactions, damage to transportation and communication systems,
epidemic risk of infectious diseases, and priority health issues.
Rapid needs assessment is usually conducted through cluster or convenience
sampling methods. The results are useful for planning follow-up interventions;
hence, the main findings of needs assessment and corresponding recommenda-
tions should be reported to policy-makers or related agencies as soon as possible.
5. Conducting thorough investigation
When the efforts of disaster rescue are more focused on the reconstruction or
rehabilitation of the community, more precise and complete information or data
are needed, such as detailed data on death and injury, nutritional status of the
residents, the damage of health-care services and other public facilities, and long-
term effect of the disaster on physical and psychological health. Such information
plays an important role in guiding after-disaster reconstruction. Besides, the
information can also be used as evidence for determining legal responsibility in
man-made accidents.

13.3 Public Health Emergency Preparedness

It is generally believed that managing public health emergencies is about how to

respond to the occurrence of it. However, this is but a corner of the picture. The
management of public health emergencies must cover all the four phases of emer-
gency: pre-emergency phase, response phase, recovery phase, and mitigating phase.
In every phase, some certain measures should be taken, that is, preparedness,
response, and recovery. Efforts conducted in all these phases are essential and
valuable. For example, in preparedness phase, we take measurements to prevent
the occurrence of emergencies or carry out researches to achieve precise prediction
and early alarming of harmful events. When these events occur unavoidably, we can
minimize the harm they cause. Previous experience has illustrated the important role
of adequate preparation in advance on the successful management of public health
emergencies, no matter natural or man-made. Hence, without adequate preparation,
it is difficult to respond effectively and successfully, even with sufficient resources.

13.3.1 Definition of Public Health Emergency Preparedness

Public health emergency preparedness (PHEP) is the ability to prevent, prepare for,
rapidly respond to, and recover from public health emergencies in coordination with
local government, health-care system, health-related institutes, communities, and
individuals, especially when big threatens occur that exceed the normal scale.
13 Epidemiology of Public Health Emergencies 239

Preparedness involves a continuous and long-term process of planning and imple-

mentation, and its success depends on proper planning, corrective actions, coordi-
nation, and selfless dedication of many people and institutes concerned. Public
health preparedness requires the collaboration of government and community
leaders to ensure that the community are adequately prepared to respond to a
possible emergency, do their best to mitigate its damage and recover from the
emergency as quickly as possible if they cannot prevent it.

13.3.2 Significance of Public Health Emergency

Preparedness

Organizing responses to emergencies ensures adequate access to mental, medical

and all other health-related services for the affected population. Emergency pre-
paredness cannot completely eliminate all hazards, but it is a strong prerequisite for
ensuring rapid and effective response to emergencies, thereby minimizing morbid-
ity, mortality, and other damages. For example, governments can develop food
security programs to ensure that food is protected from pests to avoid food crises,
meteorological warnings and material reserves help residents to withstand extreme
heat and extreme cold weather, and wide-coverage immunization procedure and
systematic surveillance system help communities prevent from outbreaks of infec-
tious diseases.

13.3.3 The Main Activities of Public Health Emergency

Preparedness

Preparedness is reflected both in the ability to prevent possible public health

emergencies and in the ability to respond adequately when emergencies occur.
The main activities of public health preparedness are as follows:
1. Establishing an emergency management agency
Preparedness activities must be undertaken jointly by the many relevant health
sectors in order to respond more quickly and effectively to an emergency or
disaster. Hence, an effective national emergency management system is needed to
organize and coordinate various departments to deal with emergencies effectively
and efficiently. However, there is still no such comprehensive national emergency
management system in many developing countries. Instead, the military is in
charge of the relief effort by default. In China, a three-level emergency manage-
ment system has been set up since 2003, which include national-, provincial-, and
city-level agencies, with Prime Minister as the top leader. In 2006, Emergency
Management Office of the State Council was set up, which is a significant
milestone representing the building of complex emergency response system in
240 H. Zhu

China. In April 2018, Ministry of Emergency Management of the People’s

Republic of China was set up. As the national-level agency for emergencies
management, it coordinates emergency response in China.
2. Developing laws and regulations
In the event of the SARS outbreak (2003), the Chinese government realized
the necessity of legalization in emergency management, and then issued Regula-
tions on Preparedness for and Response to Emergent Public Health Hazards on
May 7, 2003. After that, another law was issued on August 30, 2007, which was
Law of the PRC on Response to Emergencies. Besides, there are some other
related laws or regulations, such as Law of the PRC on the Prevention and
Treatment of Infectious Diseases; Law of the PRC on Prevention and Control
of Occupational Disease; Hospital Infection Management Measures; Manage-
ment of Information Report on Monitoring of Public Health Emergencies and
Infectious Diseases; etc. These laws and regulations identify the specific obliga-
tions and responsibilities of each related institute and individual.
3. Establishing action protocols
The purpose of establishing action protocols is to facilitate a prompt, efficient,
coordinated response in the case of an emergency. Detailed protocols should be
set up for all kinds of emergencies. Given the wide scope of emergencies, there
are a corresponding number of different types of emergencies. Plans and pro-
tocols are developed and established to mitigate adverse impacts of emergency
events and train the team to keep them ready.
4. Simulating emergency events to improve readiness
Planning, preparation, and practice are the keys to achieving success in the
case of an actual emergency. One kind of practice is a simulation program, like a
fire drill, or earthquake drill, which is focused on training departments and
residents in the timely recognition and appropriate intervention for critical emer-
gency events.
5. Training public and professional personnel
On the one hand, professional training on the knowledge and skills related to
emergency preparedness and response should be given to emergency-related
personnel, such as public health professionals, emergency physicians, clinicians,
other health-care workers, and even firefighters or military soldiers. On the other
hand, appropriate training and education to community residents in advance is
also relevant, which helps public know basic knowledge on common infectious
diseases, emergency rescue skills, and proper response to a public health
emergency.
Various ways of training can be used, including lectures, courses, and network
training provided by universities, Centre for Disease Control and Prevention, and
academic health centers. Moreover, emergency drills and exercises are important
part of training, familiarizing the personnel with the practical application of
emergency procedures, systems, and facilities.
6. Reserving supplies and staffs
Disaster relief goods and materials should be prepared in advance and spe-
cially used for the rescue, transfer, and arrangement of the affected population in
13 Epidemiology of Public Health Emergencies 241

the emergency. These materials may be provided by governments at all levels or

donated by public or private institutions/organizations and individuals. Till 2010,
the Chinese government has set up 17 central-level lifesaving disaster relief
materials storage warehouse, including foods, drinking water, medical goods,
and daily necessities. Besides, some specific warehouses are building for certain
kinds of emergencies, like fire, flood, earthquake. In addition, prophylactic
medicines, such as antitoxins, antibiotics, chemical antidotes, and vaccines
should be reserved. Professional staffs should also be cultivated by school
education or on-the-job training.
7. Monitoring and early warning
Continuous and systematic monitoring, also called “surveillance,” public
health data, as well as analyzing and interpreting these data, provides policy-
makers critical information for planning, implementation, and evaluation of
public health interventions. As the foundation of emergency preparedness, sur-
veillance is to monitoring any health-related changes or patterns. Based on
monitoring system, abnormal signs and changes which may adversely affect
communities can be detected promptly, thereby, there is enough response time
for emergency systems to adequately prepare for possible disasters and minimize
their harmful effects. Especially, for certain events, like disease outbreaks, local
armed conflicts, and floods, the authorities may issue warnings to alert people to
the impending dangers. This can reduce the material and economic losses and
prevent loss of life. Different colors can be used to represent different threat alert
levels. For example, in American green means low threat, blue means general
threat, yellow means significant threat, orange means high threat, and red means
severe threat. The main responsibilities of a comprehensive monitoring and early
warning system includes data monitoring, risk analysis, event detection, warning
release, communication, and feedback, with the aim of realizing complete sur-
veillance, accurate prediction, and timely warning.
8. Risk assessment
Risk assessment is a risk analysis of predefined vulnerabilities and hazards
based on systematically collected data. In a broader sense, risk assessment
includes daily risk assessment and specific risk assessment. The former is
conducted in the pre-emergency phase to assess the hazard vulnerability in an
area, and the latter is carried out after the onset of an emergency to assess the
potential further risk caused by this event. In a narrower sense, risk assessment
only refers to daily risk assessment.
Daily risk assessment involves gathering information about the most common
risk factors affecting the area, the likelihood and risk of emergencies, the extent of
damage to infrastructure, emergency response capacity and weaknesses of public
facilities, and the amount and characteristics of vulnerable populations.
Specific risk assessment would be conducted once an emergency is detected or
notified, and then confirmed. Specific risk assessment takes into account the
actual site conditions and address only the relevant hazards, and then results in
grading the event, which determines the level of response that needs to be taken
and activating the appropriate emergency response. Event grading is based on
242 H. Zhu

five criteria: scale, severity, urgency, respond ability of local or national govern-
ment, and government’s reputational risk.

13.4 Public Health Emergency Response

13.4.1 Definition of Public Health Emergency Response

Public health emergency response refers to the actions, which are taken immediately
following an emergency, to save lives, provide assistance, minimize economic
damage, and accelerate post-disaster reconstruction, so as to mitigate the adverse
impact on the public and society.

13.4.2 Significance of Public Health Emergency Response

Rapid and proper response to an emergency is important with respect to life safety
(which is usually the Number 1 Goal), stabilizing the emergency and protecting the
environment and property.

13.4.3 The Main Activities of Public Health Emergency

Response

Emergency response is a dynamic process. The response actions should be initiated

during the first 24 h of an incident. Specialized rapid-response teams are needed to
respond quickly to new disasters and disease outbreaks.

13.4.3.1 Ensuring Availability of Preventive and Emergency Medical

Treatment

When disease outbreak or disaster occurs, the top priority of any public health
response is to save lives and control the spread of disease. Many resources are
required, including isolation treatment hospital (for communicable diseases), med-
ical supplies (antitoxins, antibiotics and chemical antidotes, laboratory agents, and
equipment), staff (medical workers, health-care workers, public health personnel,
volunteers, and psychological counselor), guidelines for diagnostic and treatment,
transport, and stationery. Countermeasures for communicable diseases must be
administered to patients, prophylactics must be provided to high-risk groups when
necessary, like vaccine or prophylactic antibiotics, and psychological canceling must
13 Epidemiology of Public Health Emergencies 243

be provided alongside treatment. Please note that medical workers must attach
enough importance to personal protection, patients isolation, and environment and
items disinfection to avoid nosocomial infections.

13.4.3.2 Preventing Secondary Public Health Emergencies After

Disaster

The primary cause of mass casualties from terrorist attacks or natural disasters may
not be chemical or biological in nature, but be the subsequent secondary hazards,
such as infectious diseases (such as cholera and plague) epidemic, leakage and
spread of toxic gas, destruction of lifeline supplies (communications, transportation,
water supply, power supply, etc.), and social unrest (robbery). For instance, World
Trade Center explosion, which caused severe damage to urban facilities and high
levels of social panic may further lead to secondary infectious diseases and poison-
ing. Another example is that on March 11, 2011, the earthquake in Japan triggered a
tsunami, causing a nuclear leak at the Fukushima nuclear power plant. Hence, in
emergency situations, priority should be given to preventive measures to avoid
further deaths or more health threatens. Local public health managers protect the
affected and surrounding areas from secondary health threats caused by pest or
rodent infestations, ensure adequate water, food, and living space, dispose garbage
and human feces, promote hygiene practices, handle dead bodies appropriately, etc.

13.4.3.3 Interrupting the Route of Transmission

Public health authorities should take measures to prevent further spread of the
disease as soon as the source of the disease is identified. Measures taken include
isolating sources of outbreak, such as closing contaminated restaurants or water
supply, isolating and treating communicable disease patients, and enclosing build-
ings. When a disaster is severe enough, the government has the right to declare a
lockdown in the affected areas. In addition, there must also be an adequate contact
tracing workforce to track and trace cases and contacts to prevent the spread of
outbreaks.

13.4.3.4 Remediating of Environmental Health Conditions

The role of public health authorities also includes decontamination and disinfection
of affected site and facilities. Decontaminating is the neutralization, removal or
destruction of toxic or hazardous substances, such as toxic substances, radioactive
materials, or disease pathogens, in order to avoid harm to other patients and health-
care providers, and prevent secondary pollution and nosocomial infection. The
manner and extent of decontamination mainly depends on the nature of the hazard-
ous substance and its viability (microorganisms) and degradation rate (radiation).
244 H. Zhu

13.4.3.5 Performing Laboratory Analyses to Support Epidemiology

and Surveillance

Laboratory tests can strongly support epidemiological survey and surveillance. In

many instances, laboratory work is inseparable from the discovery of pathogens or
chemical poisons, the diagnosis of patients, and the determination of harmful sub-
stances in environment. In some special cases, new disease pathogens must be
detected through more sophisticated laboratory analysis such as RT-PCR and
high-throughput DNA sequencing. In addition, laboratory techniques can be used
for the development of vaccines or new drugs, as well as for the study of pathogen
resistance or homology.

13.4.3.6 Communicating with Media and Delivering Message

to the Public

Information about the emergency must be communicated to the public through the
media, telling them the good, the bad, the ugly, and what we do not know yet. This
communication can also be called risk communication. As one of the key counter-
measures for emergencies, risk communication is the timely exchange of informa-
tion, knowledge, attitude, and advice between government or health officials or
relevant experts and the affected people. Effective risk communication reduces
mortality and morbidity by promoting good personal and home hygiene and improv-
ing self-rescue and self-care ability, it also helps government maintain political and
economic stability of the country. Therefore, all countries should regard risk com-
munication as a core part of their efforts to respond to emergencies. The ultimate
goal of risk communication is to enable people at risk to respond correctly when a
risk occurs to mitigate the effects of a hazard. The message delivered in risk
communication must be simple, timely, accurate, relevant, credible, and consistent.
In addition, more emphasis should be placed on effective public education and
two-way conversation, as well as timely communication of risks to the public.

13.5 Summary

In the management of public health emergencies, epidemiology has the ability to

monitor, detect, and investigate potential hazards, and to maintain and improve the
systems necessary to support this capability. Epidemiological functions would be
especially important to address hazards that are environmental, radiological, toxic, or
infectious in nature. Epidemiological functions are one of several core public health
capabilities as being critical for public health emergency preparedness.
For public health emergencies, preparedness and response are inextricably linked.
Preparedness is based on lessons learned from both actual and simulated response
13 Epidemiology of Public Health Emergencies 245

situations. An effective response is all but impossible without extensive planning

and thoughtful preparation. Public health emergency management conveys the
important idea that protecting populations and property involves the estimation of
risks, preparation, and activities which will mitigate the consequences of predictable
hazards and post-disaster reconstruction in a way that will decrease vulnerabilities.
An important goal is building a culture of awareness that preparation is not only
possible but also will greatly reduce the consequences of disasters in terms of human
and economic loss. In these, public health is an important partner with engineers,
planners, elected leaders, and community organizations.
Chapter 14
Molecular Epidemiology

Hui Wang

Key Points
• Describe the concept of molecular epidemiology
• Familiar with the concept and classification of biomarkers and know how to select
biomarkers.
• Understand the relationships between molecular epidemiology and traditional
epidemiological methods
• Discuss, apply and interpret application of molecular epidemiology in disease
control and prevention.
In molecular epidemiology, the study of the determinants of disease will pay
attention to the causative, protective, and predisposing factors (including infectious
agents and various environmental exposures, e.g., chemical or physical agents and
lifestyle habits) and host characteristics such as genetic susceptibility. Most of these
studies are performed via molecular techniques within the molecular biology.

14.1 Introduction

14.1.1 Concept

Molecular epidemiology is defined as the study of the epidemiology of human

diseases by application of the techniques of molecular biology at the population
level. Molecular investigations can contribute to the elucidation of diseases’
etiology.

H. Wang (✉)
School of Public Health, Peking University, Beijing, China
e-mail: [email protected]

© Zhengzhou University Press 2023 247

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_14
248 H. Wang

Proposing the clear definition of biomarkers is the vital step in the study of
molecular epidemiology. Biomarkers are referred to any markers which could be
detected and represented the changes from the exposure to the onset of disease on a
population scale. The range of biomarkers is quite broad, including cellular, bio-
chemical, and immunological elements. Generally, all biomarkers (e.g., nucleic acid,
protein, lipids, and antibody) can be investigated in the molecular epidemiology.
According to the process of disease, biomarkers applied in the molecular epidemi-
ology are divided into three categories: markers of exposure, markers of biological
effects, and markers of susceptibility.

14.1.2 Characteristic

Compared with conventional epidemiology, molecular epidemiology lay emphasize

on the knowledge of the pathogenesis of diseases by elucidating specific molecular
pathways and pointing out specific molecules or genes that influence the risk of
developing diseases. For instance, genetic biomarkers rather than family history
might be more precise in characterizing host susceptibility. Molecular epidemiology
can enhance the validity and reduce bias in the assessment of environmental
exposures. It can predicate the onset of disease at the subclinical level and provide
tools to discern heterogeneity within a disease, such as the development of breast
cancer subtypes (i.e., basal, luminal A, luminal B, normal breast-like, and ERBB2+).
Not all biomarkers are suitable for molecular epidemiological studies due to
expensive cost or intensive labor. It is necessary to examine those laboratory
techniques used in the studies of molecular epidemiology, for their sensitivity,
validity, specificity, and variability within and between laboratories before using
them in any epidemiological research. Meanwhile, the acquisition of appropriate
biological specimens, costs, and ethical issues need to take into considerations in
molecular epidemiological research design as well.
This chapter shows the introduction of main characteristics of molecular epide-
miology, the description of three major categories of biomarkers, most commonly
used research methods and application and prospective of molecular epidemiology.
This chapter will present an overview of molecular epidemiology. The authors refer
the readers to more articles published recently and update the knowledge of molec-
ular epidemiology.

14.2 Classes of Biomarkers

14.2.1 Biomarkers of Exposure

Molecular epidemiological studies intend to establish the causal and biological

associations between exposures and diseases. The exposure is defined as any contact
14 Molecular Epidemiology 249

with physical, chemical, or biological agents by the International Programme on

Chemical Safety. Exposure assessment should be continuous which requires that
biomarkers of exposure should be continuous as well. Thus, this assessment will
provide more exact knowledge with regard to the exposure-disease association.
Molecular epidemiology implements biomarkers to improve exposure assess-
ments in the complexity of distinguishing respect between the effect of individual
and environmental factors to diseases etiology. Validated biomarkers could measure
the disease process at the individual level which leads to the causal inference or
biological plausibility of an exposure-disease association. For instance, in the area of
virology, antibodies have been used to identify what kind of virus which a person
has been infected with. Furthermore, by measuring the accumulation of chemical
agents or metals in biospecimen, such as arsenic in hair or mercury in fingernails or
toenails, it can directly measure an exposure at the individual level.
Using sensitive laboratory techniques, low levels of exposure can be detected by
trace analysis of biomarkers. Most biomarkers usually represent the exposure of
environmental toxicants, nonetheless, they could identify the crude amount of
ingested dietary components as well, such as bacterial or viral infections. Moreover,
they also serve as terminuses for a determination of the success of interventional
strategies.
Biomarkers of exposure are classified relying on what they measure, either an
internal dose (i.e., serum vitamin D) or a biological effective dose (i.e., a dose that
causes DNA damages). Biomarkers of internal dose estimate the presence of envi-
ronmental chemicals and their metabolites in human tissues, excretions, and/or
exhaled air. Further, measurements of dietary biomarkers either as “recovery” bio-
markers, for example, measurements of sucrose and fructose in 24-hour urine
samples for direct assessment of sugar consumption, or as “concentration” bio-
markers, such as serum carotenoids, which indirectly indicates dietary intake since
they are the results of complex metabolic processes. Exposure biomarkers also serve
as evaluation indicators for the effects of interventional strategies. The utility of such
biomarkers is restricted to the availability of detectable levels of the compound.
Although biomarkers of exposure are often described separately from biomarkers of
effect, many actually overlaps exist, which can be partially attributed to the fact that
the biomarkers provide information related to both of the exposure and the effect.
For example, lymphocytes could be a surrogate for exposure and also a target for the
exposure’s effect.

14.2.2 Biomarkers of Effects

Biomarkers of effects measure the interaction between an agent and/or its metabo-
lites and target cell(s) or molecule(s); they are defined as “measurable changes in the
organism.” This definition consists of markers of effects that can indicate a preclin-
ical response and is not always detected by using traditional clinical diagnostic
techniques. The early effects of an individual can be used as informative markers
250 H. Wang

of disease risk. Several molecular-based assays have been developed to identify

cellular response(s) activated by such exposures.
The amount of the agent that reaches a crucial cellular target can be detected by
biomarkers of a biologically effective dose, for instance, DNA adducts or the amount
of a chemical agent bound to a cellular receptor. Other biomarkers of effect measure
the damage caused by the agent. For instance, under situation of a biological
effective of UV dose, DNA damage induced by UV is measured and UV length
can be further used to classify the type of damage. UVA exposure induces base
excision repair, while UVB exposure induces nucleotide excision repair. Biomarkers
of DNA damage include mutations, DNA strand break, adducts, micronuclei, sister
chromatid exchanges, and chromosomal aberration.

14.2.3 Biomarkers of Susceptibility

Molecular epidemiology provides tools to identify the genetic and acquired suscep-
tibility (such as DNA repair capacity). At present, association studies are the most
common genetic epidemiological studies.
Abundant studies have been carried out on candidate genes based on biochemical
hypotheses in terms of DNA repair, carcinogen metabolism, or cell cycle. Multiple
GWAS and meta-analyses (see below) are currently evaluating large numbers of
SNPs for an overview of methods and genetic loci that seems to correlate with
diseases. Although these studies include thousands of cases and controls, which tend
to be huge, some smaller studies with very well-designed selection of subjects have
contributed to the understanding of genetic susceptibility as well. For instance, in the
study by Klein et al., a genome-wide screen of 96 cases and 50 controls shown that
an intronic and common genetic variant in the complement factor H gene (CFH)
played a crucial role in age-related macular degeneration (OR = 7.4, 95%CI =
2.9~19).

14.2.4 Biomarker Selection

Before starting the experiments, several issues should be considered when identify-
ing candidate biomarkers. For example, it should be known the prevalence of
biomarkers of interest in the population. The ability of the biomarker representing
the agent of interest and the sensitivity and specificity of the biomarker measuring
low dose of exposure should be considered as well. Additionally, the validity of the
biomarker and reliability must be determined. Epidemiological studies
implementing biomarkers must take into consideration that environmental exposures
will vary qualitatively and quantitatively over time. It also should be taken into
account that part of biomarkers decay over their lifetime, thus, when selecting an
appropriate design of a molecular epidemiological study, the half-life of biomarkers
14 Molecular Epidemiology 251

must be considered. Most biomarkers are transient with a relatively short half-life
period. In conventional epidemiology, case-control studies have great advantages in
research of which the disease of interest is rare and the exposure is frequent and easy
to identify. For instance, when investigate cancers or other chronic diseases, studies
are usually focused on events that happened many years before the disease onset and
often involve chronic exposures. However, implementing molecular epidemiology
in chronic diseases, the method of case-control is restricted if the biomarker has a
short half-life time. Therefore, it indicates an acute exposure that occurred in a short
time before disease onset. However, such studies are often confined in the sense that
they are using a “one-time” biological sample, which cannot certainly represent the
common exposure or of changing exposures. For those biomarkers, prospective
molecular epidemiological studies are more suitable.
In molecular epidemiological study, biomarker validation contains several issues
that need to be considered when assessing the utility of a biomarker. Analytical
validity, clinical validity, clinical utility, and ethical, legal, and social implications
and safeguards are often regarded as the ACCE evaluation of a biomarker.
The analytical validity points at the ability of a test to reliably and accurately
measure the genotypes/markers of interest which includes its sensitivity and speci-
ficity. The ability of a genetic test to detect or predict the phenotype is the clinical
validity or the positive predictive value. The clinical utility component of this
assessment considers the risks and advantages related to the incorporation of the
test into routine clinical practice. Recently, the legal, ethical, social implications and
safeguards are other issues that need to consider when assessing the utility of a
biomarker.
Betsou and colleagues recommended several methods to evaluate the vulnerabil-
ity of a biomarker to pre-analytical variation. These evaluations can be conducted to
ensure that association with clinical end points is not because of uncontrolled
pre-analytical variation. For example, the characteristics could change rapidly
(e.g., vitamin C is light-sensitive) for serum is not processed rightly. Other reasons
of pre-analytical variations include fasting conditions, specimen collection’s time,
the position of the patient when collecting, the patient’s diet, or other life habits, all
of which are necessary consideration when choosing appropriate biomarkers. The
use of inappropriate biomarkers may partly due to the publication bias which causes
false-positive associations. Many biomarkers were tested but never published
because of the unfavorable results. Publication bias might be a result of time
consuming and costly assays, such that the positive findings of manuscripts are
more likely to be published than negative findings, although they may have been
acquired by chance.
252 H. Wang

14.3 Main Research Methods Used in Molecular

Epidemiology

14.3.1 Study Design in Molecular Epidemiology

In the past several years, it might be the most revolutionary changes in molecular
epidemiology for the emerging of discovery technologies that can been put into use
in many study designs, such as genome-wide scans of common genetic variants,
messenger RNA (mRNA) and microRNA expression arrays, proteomics, and
metabolomics (also referred to as metabonomics). These approaches are helping
investigators to explore biological responses to exogenous and endogenous expo-
sures, to evaluate potential modification of those responses by variants in essentially
the entire genome, and to define tumors at the chromosomal, DNA, RNA, and
protein levels. Biomarkers of genetic and environmental factors referred to human
disease have been applied to cross-sectional studies, case-control studies, and cohort
studies.

14.3.1.1 Cross-Sectional Studies

Cross-sectional studies can be used to assess allele and genotype frequencies,

exposure levels in the population, and the relationships among genotypes, expo-
sures, and phenotypes. Although cross-sectional studies cannot infer causality
between incidence and natural history, they can provide information on genetic
variants and environmental exposures that may help to guide research and health
policy at population level. For example, a population-based prevalence study ana-
lyzes two common mutations in the hemochromatosis gene (HFE) (C282Y and
H63D variants of HFE) in the U.S. population. Steinberg et al. genotyped 5,171
samples from the Third National Health and Nutrition Examination Survey
(NHANES III) of Center of Disease Control and Prevention (CDC), a nationally
representative survey conducted in the United States from 1992 to 1994. Genotype
and allele frequency data were cross-classified by sex, age, and race/ethnicity. The
CDC provides an ongoing assessment of the U.S. population's exposure to environ-
mental chemicals by the analysis of NHANES surveys. The first National Report on
Human Exposure to Environmental Chemicals was issued in 2001 and presented
exposure data for 27 chemicals from NHANES 1999–2001. In 2003, The second
report presented exposure data for 116 environmental chemicals stratified by age,
gender, and race/ethnicity. Furthermore, by cooperating with the National Cancer
Institute (NCI), the CDC applied the NHANES III survey to measure prevalence of
variants in 57 genes and correlate the resulting genotypes with clinical, medical
history, and laboratory data. When completed, such studies will provide valuable
information on the association between genetic variations and numerous health end
points.
14 Molecular Epidemiology 253

14.3.1.2 Case-Control Studies

The case-control approach is especially well suitable to study genetic variants in that
(a) unlike other biologic markers of exposures such as DNA adducts and hormonal
levels, genetic markers are stable indicators of host susceptibility; (b) case-control
studies can implement an all-sided search for the effects of several genes, along with
other risk factors, and look for gene-environment interactions; and (c) case-control
studies are suited for plentiful unusual disease end points (e.g., specific cancers and
birth defects). Furthermore, because the environmental exposures change over time,
cohort studies with repeated biomarkers of exposures and intermediate outcomes
may be preferable to case-control studies, unless case-control studies are nested
within an underlying cohort of a well-defined population for which biological
samples stored at the start of the study are later analyzed for exposures. Case-
control studies can synchronously support gene discovery and population-based
risk characterization. For instance, registries of population-based incident disease
cases and their families offer a platform to conduct family-based linkage and
association studies. The reflection of this philosophy is the NCI sponsors Coopera-
tive Family Registries for Breast and Colorectal Cancer Research. Population-based
case registries can support many study designs, including extended family studies,
case-parent trios, and case-control family designs. One type of family-based associ-
ation study is the kin-cohort design in which researchers access the genotype-
specific risk of disease occurrence in first-degree relatives of study participants
(probands), inferring genotypes of relatives from genotypes measured in probands.

14.3.1.3 Cohort Studies

Efforts are now being done to integrate genomics into cohort studies started in the
pregenomic era to study disease incidence and prevalence, natural history, and risk
factors. Well-known cohort studies include the Framingham study, the Atheroscle-
rosis Research in Communities study, the European Prospective Investigation on
Cancer, and the newly designed National Children Study, a planned U.S. cohort
study of 100,000 pregnant women and their offspring to be followed from before
birth to age 21 years. In addition, the genomics era is enlightening the development
of very large longitudinal cohort studies and even studies of entire populations to set
up repositories of biologic materials (“biobanks”) for discovery and characterization
of genes relevant to common diseases. There are adequate number of studies could
be listed, which range from large random samples of adult populations such as the
UK Biobank (N = 500,000) and the CartaGene project in Quebec (N = 60,000) to
populations of entire countries such as Iceland (N = 100,000) and Estonia (N =
1,000,000; Estonian Genome Project), to a cohort of twins in multiple countries
(GenomeEUtwin). It is worth mentioning that the China Kadoorie Biobank was
launched in 2004, which recruited 0.5 million people with blood data and then
collect their health information for at least two decades. These biobanks can help
254 H. Wang

epidemiologists to quantify the occurrence of diseases in multifarious populations

and to understand their natural histories and risk factors, including gene-
environment interactions.
Longitudinal cohort studies allow for repeated phenotypic and outcome measures
of individuals over time, including intermediate biochemical, physiologic, and other
precursors and sequels of disease. Cohort studies can also be applied to nested case-
control studies or even as an initial screening method for case-only studies
(as explained before). Such studies will generate abundant data on disease risk
factors, lifestyles, and environmental exposures, and make preparations for data
standardization, sharing, and joint analyses. An example of data standardization
across international boundaries is the global P3G (Public Population Project in
Genomics), which, to date, includes three international studies from Europe and
North America. “Harmonization” is vital for creating comparability across sites on
measures of genetic variation, environmental exposures, personal characteristics and
behaviors, and long-term health outcomes.

14.3.2 Main Molecular Methods Used in Molecular

Epidemiology

Numerous molecular biological techniques are implemented in the epidemiological

studies. This section listed main methods used in the molecular epidemiological
studies.

14.3.2.1 Electrophoretic Mobility Shift Assay (EMSA)

The EMSA or mobility shift electrophoresis referred as a gel mobility shift assay, gel
shift assay, gel retardation assay, or band shift assay as well, a usual affinity
electrophoresis techniques, is used to study protein-DNA or protein-RNA interac-
tions. This procedure can confirm if a protein or mixture of proteins is able to
combine with a given DNA or RNA sequence. Sometimes, it can be used to indicate
if more than one protein molecule take part in the binding complex. Gel shift assays
are often performed in vitro concurrently with DNase footprinting, primer extension
and promoter-probe experiments when studying transcription initiation, DNA repli-
cation, DNA repair or RNA processing and maturation. Precursors can be found in
earlier literature, but most present assays are based on methods described by Garner
and Revzinand Fried and Crothers.
The EMSA technique is based on the observation that protein-DNA complexes
migrate more slowly than free linear DNA fragments when subjected to
non-denaturing polyacrylamide or agarose gel electrophoresis. Because the rate of
DNA migration is shifted or retarded when bound to protein, the assay is also defined
as a gel shift or gel retardation assay. The ability to resolve protein-DNA complexes
14 Molecular Epidemiology 255

depends greatly on the stability of the complex during each step of the procedure.
During electrophoresis, the protein-DNA complexes are quickly resolved from free
DNA, providing a “snapshot” of the equilibrium between bound and free DNA in the
original sample. The gel matrix provides a “caging” effect that contribute to stabilize
the interaction complexes: even if the components of the interaction complex
dissociate, their localized concentrations remain high, promoting positive
reassociation. Additionally, the relatively low ionic strength of the electrophoresis
buffer helps to stabilize transient interactions, permitting even labile complexes to be
resolved and analyzed by this method.
Protein-DNA complexes formed on linear DNA fragments lead to the character-
istic retarded mobility in the gel. However, if circular DNA is used (e.g., mini-circles
of 200–400 bp), the protein-DNA complex may actually migrate faster than the free
DNA, analogous to what is observed when supercoiled DNA is compared to nicked
or linear plasmid DNA during electrophoresis. Gel shift assays also help to resolve
altered or bent DNA conformations that induce by the binding of certain protein
factors. Also, gel shift assays are suited for protein-RNA and protein-peptide
interactions by using the same electrophoretic principle as well.

14.3.2.2 Dual Luciferase Reporter Assay

The wide applications of genetic reporter systems help to study eukaryotic gene
expression and cellular physiology, including the study of receptor activity, intra-
cellular signaling, transcription factors, mRNA processing and protein folding, and
so on. Dual reporters are usually applied to enhance experimental accuracy. The term
“dual reporter” refers to the simultaneous expression and measurement of two
individual reporter enzymes within a single system. Generally, the “experimental”
reporter has relation to the effect of specific conditions of experiment, while the
activity of the co-transfected “control” reporter offers an internal control that act as
the baseline response. Normalizing the activity of the experimental reporter to the
activity of the internal control minimizes experimental variability caused by differ-
ences in cell viability or transfection efficiency, which also can effectively eliminate
other sources of variability, including differences in pipetting volumes, assay effi-
ciency and cell lysis efficiency, and so on. Hence, dual-reporter assays often permit
more reliable interpretation of the experimental data by reducing extraneous influ-
ences. The Dual-Luciferase Reporter (DLR™) Assay System offers an efficient
method of performing dual-reporter assays. In the DLR™ Assay, the activities of
firefly (Photinuspyralis) and Renilla (Renillareniformis, also known as sea pansy)
luciferases are measured sequentially from a single sample. The firefly luciferase
reporter is measured first by adding Luciferase Assay Reagent II (LAR II) to
generate a stabilized luminescent signal. After quantifying the firefly luminescence,
this reaction is quenched, and the Renilla luciferase reaction is simultaneously
initiated by adding Stop & Glo Reagent to the same tube. The Stop & Glo Reagent
also produces a stabilized signal from the Renilla luciferase, which decays slowly
over the course of the measurement. In the DLR™ Assay System, both reporters
256 H. Wang

yield linear assays with subattomole sensitivities and no endogenous activity of

either reporter in the experimental host cells. Furthermore, the integrated format of
the DLR™ Assay provides rapid quantitation of both reporters either in transfected
cells or in cell-free transcription/translation reactions. Promega provides the pGL4
series of firefly and Renilla luciferase vectors designed for use with the DLR™
Assay Systems. These vectors may be used to co-transfect mammalian cells with
experimental and control reporter genes.

14.3.2.3 The Comet Assay

The “comet” assay was developed in the late 1980s/early 1990s and used only some
lymphocytes. The lymphocytes are frozen at a very low temperature to ensure their
viability, and then treated and run out on a gel that was spread on a glass slide. DNA
from the cell “migrates” to form a “tail.” If DNA is “broken” (i.e., single-strand
breaks), then the length of the tail is relative to the amount of breakage. This assay
tends to measure DNA single-strand breaks, cross-links, base damage, and apoptotic
nuclei. Cells could be subject to damaging agents first, then allowed to repair, and
placed on the gel on the glass slides. In this situation, this assay measures DNA
repair “capacity” by the length of the comet tail. The comet assay is commonly used
in assessment environmental toxicant-induced DNA damage. The application of this
assay exponentially increased based on its high sensitivity and specificity. This
method also enables researchers to detect increased risk for different health out-
comes. Massive validation efforts have been taken on optimizing standardization
and reliability of the comet assay by the European Standards Committee on Oxida-
tive DNA Damage.

14.3.2.4 Micronucleus (MN) Assay

MN assay, which is used to detect MN, extracellular bodies, after the cells go
through first cell cycle, has the ability to discern chromosome breaks from aneu-
ploidy (abnormal number of chromosomes) and can detect chromosome loss. Since
MN are formed from acentric chromosomal fragments or chromosomes that are not
involved in either daughter nuclei, they are classified relying on whether they include
chromosomal fragments or whole chromosomes.
This assay is suited for use in molecular epidemiological studies for the relative
ease of scoring, limited costs and personnel requirements, and the precision that
scoring larger numbers of cells provides. The MN assay can be proceeded in
peripheral blood lymphocytes, alveolar macrophages, erythrocytes, epithelial cells,
and fibroblasts. In this assay, the cells under investigation must survive at least one
round of nuclear division, so some of the damaged cells are lost before the analysis
begins, and the survivability of the damaged cells is not known with this assay.
A review of published evaluated the occurrence of MN and the influence of
genotoxic exposures on MN frequency in children and adolescents. This review
14 Molecular Epidemiology 257

indicated that this cytogenetic assay is a helpful and sensitive tool which is suitable
for biomonitoring studies of children including those with low-dose exposures to
environmental agents. The confounding effects of age, sex, and chronic and infec-
tious diseases on MN levels were evaluated in these studies, and the only variable
irrelevant to MN frequency was sex.

14.3.3 Genome-Wide Association Studies (GWAS)

GWAS are designed to identify the entire human genetic associations with detect-
able traits or the presence or absence of a disease of interest. The precondition is the
entire genome can be assessed for variation and a few SNPs would stand out as key
risk factors of disease. The comparison is carried out between individuals with and
without the disease of interest. Since the genome is large and the number of SNPs is
countless, participants by the thousands are required to suitably investigate the
associations. The method of GWAS takes advantages over candidate gene studies
and it enlarges the potential of exploration of genetic analyses. GWAS recruit
numerous study subjects with a disease or phenotypic trait of interest. The study
subjects usually originate from ongoing collaborative scientific work including
different institutions or over all the continents. These studies take benefits of high-
throughput genotyping technologies, DNA isolation, automated collection of
biospecimen, and high-quality-control practices, and then employ statistical analyses
to determine associations between qualified SNPs and diseases or phenotypic trait of
interest. Great efforts from the laboratory and biostatistical have contributed to
thousands of GWAS so far, which, no doubt, conduce to the knowledge base of
molecular epidemiology around the world. Accurate GWAS would replicate their
results in different populations or in experimental animals, when the biological
pathways have mechanistic modeling. Regarding the “common disease, common
variant” hypothesis, GWAS depending on SNPs as markers of allelic variants that
indicates over 1–5% of each human genome. By genetic characterization, and then
fine mapping and analyses, researchers are capable of determining common genetic
variations of chronic diseases.
Generally, genome-wide scanning is conducted on an initial group of cases and
controls, and then a smaller standout SNPs are assessed to replicate findings in a
second and a third set of cases and controls. The possibility of false-positive or false-
negative findings will be reduced by the performance of such multistage study
design. Furthermore, it reduces the genotyping costs as well. Additionally, with
employed quality controls, the replicated genotyping provides the essential valida-
tion, particularly for SNPs of intron or unknown functional region.
Biases are inclined to happen in GWAS. Especially, population stratification is
one of the most crucial confounders. For instance, a potential population structure
leads to false-positive associations when the detected SNPs are also linked with
unknown factors which reflect geographical origin or ethnicity of study individuals.
The vast data produced by GWAS is prone to false-positive associations. Effective
258 H. Wang

statistical skills must be used to decrease the possibility of false positives raised by a
lot of multiple comparisons.
Another common limitation is that current statistical methods used in GWAS
capture a large number of common variants, which derived from the concept of
linkage disequilibrium or other statistical algorithms validated according to the
Human Genome International HapMap databases. These methods establish on the
theory of human genome is constituted by blocks of nucleotides named haplotypes.
Haplotypes are inherited together. Some SNPs within a given block define and
explain or “tag” within block variability. These tagging SNPs get popular in
GWAS. However, certain variants may not be captured by the current genotyping
chips while they potentially represent crucial but unknown function. Besides, it is
necessary to consider that some genetic variants might be influenced only when
combining with exposures that initiate or modify expression of that gene. Without
considering exposures to assess risks of chronic disease, we cannot successfully
reveal the complicated patterns of gene-environment or gene-gene interactions
which contribute to a great degree chronic disease risk. “Next-generation GWAS”
probably should combine with more detailed analyses of common exposures (e.g.,
smoking, alcohol, dietary patterns, air pollutants, over-the-counter medications (like
common non-steroidal anti-inflammatory drugs (NSAIDs)), and recreational drugs),
which influenced the chronic disease etiology and pathogenesis.

14.3.4 Mendelian Randomization (MR)

Confounding, selection bias, and reverse causation are major problems in building
causal relationships between exposures and diseases, which may lead to spurious
associations. MR is a method by using genetic variations of known function to detect
the causal effect of a modifiable exposure on disease in nonexperimental situation. A
vital characteristic of observational epidemiology is to identify the causes of com-
mon diseases which public health takes interest. For the purpose of confirming the
favorite effects of a recommended public health intervention, the association of
observation between the certain risk factor and a disease must prove that the risk
factor indeed causes the disease. Well-known successful examples are that causal
relationships are identified between smoking and lung cancer, and between blood
pressure and stroke. However, there are failures when identified exposures were later
demonstrated by randomized controlled trials (RCTs) to be noncausal. For example,
hormone replacement therapy (HRT) was previously thought prevent cardiovascular
disease. However, it did not and may even have other adverse effects in health. In
observational epidemiological studies, the confounders such as social, behavioral, or
physiological factors result commonly in such spurious findings. They are easy to
uncontrol and especially difficult to measure accurately. Furthermore, many findings
repeat unlikely by RCTs for ethical reasons.
MR allows one to test for a causal effect from observational data in the presence
of confounders by taking common genetic polymorphisms with well-understood
14 Molecular Epidemiology 259

effects on exposure patterns. Necessarily, the genotype must only affect the disease
process directly through its effect on the exposure. Since genotypes are assigned
randomly when inherit from parents to offspring during meiosis, if we hypothesized
that option of mate is unrelated with genotype (panmixia), the genotype distribution
among population should be irrelevant to confounders that commonly trouble
observational epidemiological studies. Therefore, MR can be considered as a “nat-
ural” RCT. From a statistical perspective, it is a use of instrumental variables, with
genotype serving as an instrument/proxy for the exposure.
The same with all studies of genetic epidemiology, trouble exists in the require-
ment for large sample sizes, the non-replicable results, and the lack of functional
proof on genetic variants. In addition to these limitations, genetic findings could be
confounded by other genetic variants by linkage disequilibrium with the variant
under study or by population stratification. Moreover, pleiotropy of a genetic variant
may contribute to null associations on account of canalization of genetic effects. If
correctly performed and carefully interpreted, MR studies can offer valuable evi-
dence to identify causal hypotheses between environmental exposures and common
diseases.

14.4 Application and Prospection

14.4.1 Control and Prevention of Infectious Diseases

The aim of molecular epidemiology of infectious diseases is to apply molecular

(amino acid or nucleotide) sequences to study the ecology and dynamics of patho-
gens. For infectious diseases, it includes the transmission system (source of infec-
tion, transmission route, and susceptible population), pathogenesis and virulence of
the microbe, the interaction between microbe and the human (or other) host(s), and
the microbiota of the host (the area microbes usually live on and in the human body).

14.4.1.1 Outbreak Investigation

In all outbreak investigations, setting the definition of a case is a key step. Molecular
techniques are the standard tool in an outbreak investigation for clarifying case
definitions, enhancing specificity, and decreasing misclassification. During an out-
break of disease, it is commonly assumed that a single microbe causes the clinical
symptoms. A microbe of the same genus and species but different strains is possible
cause of disease during the same period. Case definitions can be refined by including
the molecular typing which would increase the specificity, reduce misclassification
of non-outbreak cases with outbreak cases, and potentially increase the possibility to
identify the outbreak source. Only based on clinical symptoms, we are hard to
distinguish between diseases. This could make outbreak investigations complicated,
especially if the symptoms are not very typical. For instance, lots of viruses could
260 H. Wang

cause flulike symptoms; however, classification of influenza based on clinical

symptoms is specific only during an epidemic when a large number of flulike
patients suffer from influenza. Even during an epidemic, the confirmation from
laboratory is required as well, since there may be not only one strain of influenza
in transmission. In 2008, there were two predominant influenza A strains in circu-
lation: H1N1 and H3N2. Laboratory test is particularly helpful for identifying
individuals with mild or atypical symptoms, and determining the specific type. A
variety of methods of laboratory tests could provide a molecular fingerprint based on
the microbial genotype. For example, pulsed-field gel electrophoresis (PFGE) is
applied as the standard method for foodborne outbreaks investigations.

14.4.1.2 Trace Dissemination of a Specific Subtype of Pathogen Across

Time and Space

Microbes that cause human disease are constantly emerging and reemerging. In
order to prevent and control the spread of infection, we must be capable to trace the
origin and source of entry of pathogens into the population. By comparing strains,
we can determine if there have been single or multiple points of entry, and if
emerging resistance is from multiple spontaneous mutations or from dissemination
of a single clone. For example, Streptococcus pneumonia (S. pneumoniae), a major
human pathogen and one of the most common indications for antibiotic use, results
primarily in pneumonia, but also gives rise to meningitis and otitis media. However,
resistance to penicillin emerges relatively slowly, once it emerged it was widely
disseminated in relatively few clones as defined by multilocus sequence typing
(MLST). By contrast, the recent emergence of S. pneumoniae resistant to
fluoroquinolones has been due to various genetic mutations, suggesting spontaneous
appearance after treatment. Because the resistance of S. pneumoniae to
fluoroquinolones rapidly followed the introduction of fluoroquinolones, alternative
antibiotics will be needed in relatively short order to treat S. pneumoniae infections.

14.4.1.3 Determine the Origin of an Epidemic

Molecular tools help us to trace an outbreak or epidemic return to its origin in time,
and return to its reservoir in space. Knowing the origin in time is critical to predict
future spread, and identification of the reservoir for infection is the key to control
disease spread. For instance, the prevalence of methicillin-resistant staphylococcus
aureus (MRSA) has been a steady increase in America’s hospitals. In 2004, among
some intensive care units, the prevalence was as high as 68%. Nevertheless, in the
early 2000s, the emerging of new strains of MRSA in population from community
could not be traced back to hospitals. Genetic typing of the strains verified that
strains isolated from those who had no linkage with hospitals on epidemiology were
genotypically different from hospital strains. More recently, community-acquired
MRSA has been transmitted into hospitals. When comparing with hospital-acquired
14 Molecular Epidemiology 261

MRSA, community-acquired MRSA has different virulence factors and different

patterns of antibiotic resistance so there is a clinical benefit in enabling us to
distinguish between the two.

14.4.1.4 Follow the Emergence and Spread of New Infections

Severe acute respiratory syndrome (SARS) is anew infectious disease emerging

firstly this century. Before its identification, coronaviruses were not regarded as
primary pathogens in that only 12 known coronaviruses can be able to infect humans
or other animals. The identification of SARS resulted in a search for other corona-
virus pathogens, and horseshoe bats were identified as the reservoir and civets as the
amplification hosts at last. The time from the initial observation to the sequencing of
the virus and development of a diagnostic test was 5 months. The story of the rapid
isolation, identification, and sequencing of the coronavirus causing SARS is illus-
trative of the synergistic effects of the combination of molecular methods with
epidemiology. This effective combination enables scientists to follow the emergence
and spread, and to identify ways to prevent transmission and further introductions of
the virus into human populations.

14.4.1.5 Identify Previously Unknown or Uncultivable Infectious

Microbes

The most microbes could not be cultured using standard laboratory techniques. The
ability of replication of genetic material and determination of genetic sequence,
which can then be compared to known genetic sequence, has brought about a
fundamental revaluation of vast life around, in, and on us. Noncultural techniques
have enabled us to describe the microbial communities living in the mouth, vagina,
gut, and other body sites, and the body sites thought to be sterile by previous
detection, such as the blood. Epidemiological data may suggest an infectious origin
for a disease. Previously, if an organism cannot be cultured, it remained only a
suggestion. Molecular tools have altered this by the achievement of detecting
uncultivable microbes. It is now known that human papillomavirus (HPV) types
16 and 18 can cause cervical and other cancers, and vaccines are licensed to prevent
acquisition. HPV 16 was first identified in 1983 before the virus could be grown.
When discovering HPV 16, we realize that papillomavirus could give rise to cancers
in cows, rabbits, and sheep, but it was unclear whether the HPV can lead to human
cancers. HPV was a suspected cause of genital cancer for the similarity to Kaposi’s
sarcoma, and the epidemiology suggested that an infectious agent was involved. But
other genital infections, especially herpes simplex virus, were also suspects. HPV
had been excluded by many, but a new molecular technique, the hybridization assay,
detected in cancerous tissue a new subtype, HPV 16, which was specifically
262 H. Wang

associated with cervical and other cancers. The correlation of HPV 16 with cancers
was verified by comparing presence of HPV 16 between cancer patients and
controls. Notwithstanding this evidence can be very suggestive, it does not differ-
entiate temporal order, because the cancer might happen before the infection of HPV
16. Demonstrating temporal order required large-scale prospective cohort studies.
These studies also offered crucial perspectives supporting the possibility that vacci-
nation could protect against HPV because of rare occurrence of reinfection with the
same HPV subtype, and antibody could prevent reinfection and persistence of
low-grade lesions.

14.4.2 Control and Prevention of Chronic Diseases

With the development of economics and the implementation of vaccines, most

infectious diseases were controlled, while the incidence and mortality of chronic
diseases were increased dramatically, such as cancer, cardiovascular disease, and
type 2 diabetes mellitus. Molecular epidemiology played an important role in the
discovery of the cause, mechanism of pathogenesis, and individual susceptibility.

14.4.2.1 Improving the Understanding of Mechanism of Pathogenesis

Previously, most cancer epidemiological studies were restricted to evaluating pos-

sible causal relationships between two types of events: exposure to potential caus-
ative “environmental” agents (cigarette smoking, dietary factors, specific chemicals
from workplace, etc.) and disease outcome (i.e., clinical cancers incidence or cancers
mortality). However, the specific mechanism was unknown. Increasingly, molecular
epidemiological studies are combining panels of biomarkers related to exposure,
preclinical effects, and susceptibility using samples of exfoliated cells, blood cells,
body fluids, or tissues. These biomarkers are now being widely used in cross-
sectional, retrospective, prospective, and nested case-control epidemiological stud-
ies, for the purpose of improving our cognition to the causes of specific human
cancers. For example, the cotinine in serum or urine represented cigarette smoke
exposure, which was a valuable supplement to traditional means of evaluating
exposure. Moreover, assays have been implemented to measure “biologically effec-
tive dose” of a compound, for example, the amount that has reacted with key cellular
macromolecules. The metabolite of cotinine could form the carcinogen-DNA
adducts which related with lung cancer. Other molecular epidemiological studies
in Chinese populations have prospectively linked DNA damage induced by aflatoxin
B1 (AFB1) to liver cancer risk.
14 Molecular Epidemiology 263

14.4.2.2 Evaluating the Susceptibility of Individual and Defining

the Risk Population

Human beings evidently differ from one another in physical characteristics, person-
ality, and other factors. They are also different in genetically determined suscepti-
bility to disease. When we investigate the etiology of a disease, we cannot help
asking the question: How much of the incidence of the disease is due to genetic
factors, how much is due to environmental factors, and how do these types of factors
interact with each other to increase or decrease the risk of disease? Obviously, not
everyone who exposed to an environmental risk factor will necessarily develop
disease. Even though the relative risk for exposed to a specific factor is very high,
the notion of attributable risk implies that not all occurrence of a disease is due only
to the specific exposure in question such as the relationship between cigarette
smoking and lung cancer. It is demonstrated that lung cancer does not develop in
every smoker, and it does develop in someone who does not smoke.
People often accept a fatalistic approach when they are told that a disease is
primarily genetic in origin. But even in diseases originate primarily from gene, a
good deal of environmental interaction often occurs. For example, phenylketonuria
is characterized by a deficiency of phenylalanine hydroxylase for genetic reason; the
child who affected cannot metabolize phenylalanine, an essential amino acid, and the
excessive phenylalanine accumulation causes irreversible mental retardation. Can
we prevent the genetic abnormality? No, we cannot. Can we decrease the likelihood
that a child manifest mental retardation because of this genetic abnormality? Yes, we
can do so by providing a diet with low phenylalanine to reduce or eliminate the
child’s exposure to phenylalanine. As shown in this example, we can prevent the
adverse effects of a genetic disease by controlling the affected person’s environment
so that the manifestations are not expressed. Hence, in viewpoints of both public
health and clinical medicine, it is crucial that bear in mind the interrelationships
between genetic and environmental factors in disease causation and expression.

14.4.3 Conclusions

Traditional epidemiology has achieved greatly vital goals by means of simple tools
such as interviews and questionnaires. Even a difficult issue, for example, the
relationship between air pollution and chronic disease, has been successfully dis-
posed by time-series analysis and other means not depended on the laboratory.
Hence, it needs to be evaluated carefully for the application of molecular techniques
combined with epidemiological designs.
As the examples above demonstrated, molecular epidemiology is not different
with conventional epidemiology, but represents an endeavor that commence to
achieve specific scientific goals: (1) a better description of exposures, especially
when exposure doses are fairly low or different sources of exposure should be
264 H. Wang

integrated in a single measure; (2) the study of gene-environment interactions;

(3) the application of markers of early response, for the purpose of overcoming the
main limitations of chronic disease epidemiology, that is, the relatively low fre-
quency of specific forms of disease and the long latency period between exposure
and the onset of disease. Also limitations of molecular epidemiology should be
acknowledged: the complicacy of various laboratory methods, with scanty knowl-
edge of measurement error or interlaboratory variability; the lacking recognition of
some sources of bias and confounding; in some situations, the lower degree of
accuracy (such as urinary cotinine compared to questionnaires on smoking habits);
and the indefinite biological meaning of markers, like some circumstances of some
types of adducts or some early response markers.
Chapter 15
Pharmacoepidemiology

Xiaotian Liu and Jian Hou

Key Points
• Pharmacoepidemiology is the process of the application of the principles and
methods of epidemiology to study the uses and effects of drugs in human
population.
• The post-marketing pharmacoepidemiology study not only can supplement the
information available from pre-marketing studies, but also provide the new types
of information that cannot be obtained from pre-marketing studies.
• Pharmacoepidemiology has taken advantage of the principles and methods of
epidemiology and developed sophisticated methods to deal with problems in the
field.
In the past decades, there was an enormous progress in the medical sciences, which
has contributed to developing a great number of new powerful pharmaceuticals to
provide better medical care for the patients. However, the new pharmaceuticals
caused harm and led to the increase of serious adverse reactions that were unex-
pected in preclinical studies or premarketing clinical trials occasionally. Therefore,
pharmacoepidemiology was developed as a scientific discipline at the interface
between clinical pharmacology and epidemiology against this background.

15.1 A Brief History and Definition

15.1.1 A Brief History of Pharmacoepidemiology

More than 2000 years ago, there was a record of drug poison in Chinese medical
literature, “Shen Nong tasted 100 herbs and encountered 72 poisons one day. The

X. Liu (✉) · J. Hou (✉)

College of Public Health, Zhengzhou University, Zhengzhou, China

© Zhengzhou University Press 2023 265

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_15
266 X. Liu and J. Hou

international community really paid attention to the safety of drugs about 70–80
years ago. Since the beginning of the twentieth century, there were already adverse
drug events that occurred occasionally, caused illness, disability, and death, and
even led to the deformity and death of offspring. In 1935, pharmacists found the
effect of sulfanilamide on the antibiosis, then various types of sulfanilamide (such as
tablet and capsule) came out one after another. To improve the taste, diethylene
glycol was used to replace ethanol as solvent by the pharmacist of the Massengill
company of the United States in 1937, then sulfanilamide oral liquid agent was put
into market to treat the infectious diseases without the premarketing clinical trials.
As a result, a total of 107 people, including more than 30 children, died from renal
failure. In response, the US Congress drafted and passed the Food, Drug, and
Cosmetic Act in 1938, which stipulated that preclinical toxicity testing was required
for both the marketing and clinical trial. In addition, manufacturers needed to collect
clinical data on drug safety and submitted the data to the Food and Drug Adminis-
tration (FDA) before drug marketing. There were 60 days for the FDA to audit and
object to an application of marketing, otherwise, it would be proceeded. However,
the proof of efficacy was not required in this process.
Until chloramphenicol was found to cause aplastic anemia in the early 1950s,
more attention was paid to adverse drug reaction (ADR). In 1952, the American
Medical Association (AMA) Council on Pharmacy and Chemistry established the
first official registry of ADRs, which was used specifically to collect the information
about cases of blood dyscrasias caused by drugs. Then in 1960, the FDA began to
collect reports of ADR and sponsored the hospital-based drug monitoring programs
for new drugs. In addition, the Johns Hopkins Hospital and the Boston Collaborative
Drug Surveillance Program developed the combination application of in-hospital
monitor and cohort study to assess the short-term effect of drug used in hospital. The
approach was transported to the University of Florida-Shands Teaching Hospital
later.
In 1961, the former Federal Republic of Germany and other European countries
witnessed the infamous “thalidomide disaster.” Thalidomide was taken to treat
pregnancy vomiting in the first 3 months of pregnancy. Shortly its post-marketing,
a dramatic increase was observed in the prevalence of phocomelia, which was a
previously rare birth defect with the characteristics of the parts, or even absence of
limbs, sometimes with the presence of flippers. The epidemiological study was used
to establish the cause, and it was discovered that exposure to thalidomide in utero
was a risk factor of phocomelia. After the event that shocked the world, the United
Kingdom, the United States, and other countries in western Europe all began to
strictly check and examine the drug qualification before marketing. The United
Kingdom set up the Committee on Safety of Medicines in 1968. These facilitated
the establishment of bureau to collect and collate information from the national drug
monitoring organizations in the World Health Organization (WHO) in 1970. How-
ever, there were still the epidemic of subacute myelo-optic neuropathy (SMON) in
Japan in the late 1960s. A collaborative study by Japanese epidemiologists and
clinicians confirmed that the SMON was caused by clioquinol taken to prevent
traveler’s diarrhea. In 1971, Herbst et al. found that the use of diethylstilbestrol in the
15 Pharmacoepidemiology 267

early stages of pregnancy to preserve the fetus could cause vaginal adenocarcinoma
in their daughters. By the 1980s, the occurrence of ADR after post-marketing
prompted governments and drug administration to strengthen the management of
new drugs, which facilitated the development of pharmacoepidemiology.
With the increase in variety and quantity of drugs, the evaluation and manage-
ment of drug use have become essential. Sweden was the first country to establish
the major of clinical pharmacology in 1956. In 1964, WHO fully affirmed the
necessity of the clinical pharmacology specialty. After more than 20 years of effort,
clinical pharmacology had become a mature profession with the main function to
monitor the ADR in developed countries by the 1980s. However, in the early 1980s,
the United Kingdom medical profession showed that the existing medicine manage-
ment methods, clinical pharmacology, and other specialties still could not meet the
needs of ensuring the safety of drug users, and then drug surveillance was put
forward. Against this background, pharmacoepidemiology was developed as a
scientific discipline at the interface between clinical pharmacology and epidemiol-
ogy. In 1984, the word “pharmacoepidemiology” was first appeared in the British
Medical Journal, and then well-known by the public.

15.1.2 Definition of Pharmacoepidemiology

Pharmacoepidemiology is defined as the application of the principles and methods of

epidemiology to study the uses and effects of drugs in human population, and to
optimize the benefit risk ratio of drugs, vaccines, and medical devices through the
development and evaluation of risk management strategies, so as to improve the
medical care. The study subjects of pharmacoepidemiology are human populations,
the research contents are the distribution of drug use and drug effect in the popula-
tion, and the aims are to provide information on the safety and efficacy of drug use in
population, as well as to form a scientific basis for clinical rational drug use and
policy-making. Pharmacoepidemiology studies both beneficial and adverse effects
of a drug. Its focus is to assess the risk of uncommon, at times latent, and usually
unexpected ADR that occurs for the first time after post-marketing.

15.1.3 Drug-Related Concepts

1. Drugs
Drug is defined as a substance used for the prevention, diagnosis, and treatment of
the diseases, and to purposefully regulate human physiological functions, with
specified indications, usage, and dosage, including traditional Chinese medicinal
materials, radioactive drugs, serum, vaccines, blood products, and diagnostic
drugs, excluding drugs that are under the premarketing clinical trials.
2. Adverse drug reaction
268 X. Liu and J. Hou

Adverse drug reaction (ADR) is defined as a harmful reaction that occurs

under normal dosage of the qualified drug and is not related to the purpose of drug
use. The definition excludes ADR caused by intentional or accidental overdose or
medication errors. ADR has traditionally been separated into Type A reaction,
Type B reaction, and Type C reaction according to the pharmacological effect.
Type A reaction is the result of an exaggerated pharmacological effect under the
normal drug dosage. It is characterized by be predictable, common, dose-related,
and can be treated by reducing the dose or cessation of the drug. The common
reasons are the recipients receive overdose of a drug, or they cannot metabolize or
excrete the drug normally resulting in high level of the drug, or they are sensitive
to the drug for some reasons despite normal drug level. In contrast, Type B
reaction is an aberrant effect, with the characteristics of be unpredictable, uncom-
mon, not related to dose, and potentially more serious requiring cessation of the
drug. Type B reaction represents the major focus of pharmacoepidemiologic
study. The reasons of Type B reaction may be attributed to idiosyncratically
inherited reactions to the drug or hypersensitivity reactions to the drug or some
other mechanisms. Type C reaction generally occurs after long-term use of the
drug and is characterized by a long incubation period, unpredictability, and no
clear time relationship. The mechanism of Type C reaction is not clear and may
be related to the teratogenesis, carcinogenesis, and changes in the cardiovascular
system and fibrinolytic system after long-term usage of the drug.
3. Adverse drug event
Adverse drug event (ADE) is any adverse clinical event that occurs during the
period of drug use, but it is not necessarily causally related to drug use. Although
ADE occurs during the period of drug use, the causal relationship between drug
use and ADE needs to be verified through investigation and evaluation of
pharmacoepidemiology.

15.2 Main Research Contents

15.2.1 Drug Safety Evaluation

To explore the incidence and risk factors of ADE and ADR as well as to provide
scientific basis for drug risk management, to quickly find the adverse reactions to
ensure the safety of drug users through the data mining techniques of the observa-
tional database and the analysis of safety signals, to standardize the monitoring
methods of drugs after post-marketing and improve their practicality, and to develop
the flow chart of establishment of a causal association of ADR.
15 Pharmacoepidemiology 269

15.2.2 Drug Effectiveness Evaluation

Comparative effectiveness research (CER) aims to study the effects of interventions

and strategies on prevention, diagnosis, treatment, and health monitoring, and to
compare the health-related outcomes of different disease groups through the data
mining techniques, then to provide the evidence on which kind of intervention was
the safest, easiest, and most effective for the patients, medical staff, consumers, and
policy-makers.

15.2.3 Drug Utilization Study

The definition of drug utilization is that “marketing, distribution, prescription and

use of drugs in a society, with special emphasis on the resulting medical, social and
economic consequences” by WHO in 1977. Thereby, the aims of drug utilization
study are to examine drug utilization, study the effects of drug utilization on the
population, identify problems of drug utilization in relation to its importance, causes,
and consequences, provide a scientific basis for decision-making, and assess the
effects of actions taken. It involves pharmacy, pharmacology, pharmaceutical man-
agement, social anthropology, behavior, economics, and other fields.

15.2.4 Pharmacoeconomic Study

Pharmacoeconomics includes two levels: broad and narrow. In the broad sense, the
pharmacoeconomic study is defined as applying the principles, methods, and ana-
lytical techniques of economics to study the economic behavior of drug supply and
requisitioning parties, drug market price under the interaction between supply and
requisitioning parties, as well as various intervention policies and measures in the
field of drugs. In the narrow sense, the pharmacoeconomic study is the economic
evaluation of drug utilization based on the comprehensive analysis of drug efficacy,
safety, and utilization, and provides the theoretical basis for clinical drug use, the
prevention and therapy of disease and medical insurance payment decision-making.
By collecting and comparing economic data related to drug utilization,
pharmacoeconomics is to carry out the cost-effect analysis, cost-benefit analysis,
cost-utility analysis, or minimal cost analysis from the cost and benefit
considerations.
270 X. Liu and J. Hou

15.3 Aims and Significances

15.3.1 Aims of Pharmacoepidemiology

According to different purposes, different organizations and individuals decide

whether to conduct pharmacoepidemiological study. Generally, one study is
conducted for multiple purposes and can be refined in four respects: regulatory,
marketing, legal, and clinical needs.
(1) Regulatory
(a) Requirements of the pharmaceutical administration. (b) Manufacturers
want the drug to be approved for marketing as soon as possible. (c) Answer
questions from the pharmaceutical administration. (d) Producers want to the
drug apply for marketing in other countries.
(2) Marketing
(a) Assist in entering and occupying markets by verifying the safety of
medicines. (b) Raise the profile of the drug. (c) Assist in repositioning of
marketing. For example, adopt different outcomes, such as life quality evalua-
tion and economic evaluation; for different patients, such as children or the
elderly; discovery of new therapeutic indications; or to reduce the restrictions on
drug labels. (d) Protect developed and tested drugs from adverse reactions.
(3) Legal demands
Prepare for possible lawsuits of drug liability.
(4) The clinical need
1. To generate hypotheses. Whether or not the hypotheses need to be generated
depends on the following factors: is it a new chemical monomer? Safety of
similar drugs, relative safety of the drug among similar drugs, and drug
formulation.
2. To test hypotheses. Conducting hypotheses testing is to solve the following
problems: problems based on drug structure, questions raised by preclinical
animal tests or premarket human studies, questions raised by voluntary
reports on adverse drug reactions, and to better quantify the frequency of
adverse reactions.
3. The disease to be cured. In addition, the characteristics of disease to be
cured, such as course, prevalence, severity, availability of alternative thera-
pies, and so on, determine whether pharmacoepidemiology study should be
conducted [11].
15 Pharmacoepidemiology 271

15.3.2 Significances of pharmacoepidemiology

15.3.2.1 Improve the Quality of Premarketing Clinical Trials

To ensure the efficacy and safety, new drug must undergo human clinical trials no
matter how many vitro and animal trials each drug has undergone before
premarketing. Premarketing clinical trial of new drugs is one of the main types of
experimental epidemiology. Mastering the theoretical bases and methods of epide-
miology can help to design clinical trials in a standardized way, collect and analyze
the experimental data, identify and control bias, thereby improve the quality of
premarket clinical trials.

15.3.2.2 Post-Marketing Study of Drug

Premarketing study of drug effect is necessarily limited in size, time, and sample
(e.g., elderly, pregnant women, and children were not included). Moreover, the
disease or drug used is single during clinical trial, some low incidence of adverse
reactions, delayed reactions or drug interactions caused by the combination use of a
variety of drug, are difficult to find. Thus, the nonexperimental epidemiological
study is needed to evaluate the effect of drug administered as part of ongoing
medical care after marketing. Apart from verifying the results information in the
clinical trials of pre-marketing, the post-marketing pharmacoepidemiology study can
not only supplement the information available from premarketing studies, but also
provide the new types of information that cannot be obtained from premarketing
studies.
The potential contributions of pharmacoepidemiological study are as follows:
1. Supplement information available from premarketing studies (to better quantify
the incidence of known adverse and beneficial effects).
① Study the incidence of adverse reactions or the frequency of effective
effects during the prevention or treatment under use of drugs through the epide-
miological survey in a large number of population. ② Understand the effects of
drugs on special groups, such as the elderly, pregnant women, and children. ③
Explore the modified effects by other drugs and other illnesses. ④ Evaluate
whether the new drug is better than other drugs used for the same indications.
⑤ Evaluate the drug safety, effectiveness, quality standard, etc.
2. Provide the new types of information that cannot be obtained from premarketing
studies.
① Discover previously undetected delayed adverse and beneficial effects, and
verify them by epidemiological methods and reasoning. ② Study the character-
istics and influencing factors of drug utilization. ③ Assess the effects of drug
overdoses. ④ Evaluation of the economic benefits of drug utilization.
In addition, pharmacoepidemiology study is also conducted to fulfill the
ethical and legal obligations as the general contributions [11].
272 X. Liu and J. Hou

15.4 Methods of Pharmacoepidemiology

Pharmacoepidemiology is a scientific discipline at the interface between clinical

pharmacology and epidemiology. Therefore, the methodological approaches in
epidemiology can be used to generate and test hypotheses on drug risks or benefits
according to the purpose of the study (Fig. 15.1). Thus, both the primary study (such
as descriptive study, analytical study, and experimental study) and secondary study
(e.g., systematic review and meta-analysis) can be applied in the
pharmacoepidemiology study. Meanwhile, the multiple epidemiological methods
might be conducted to explore the associations between the drug and ADR/ADE in
one study, especially studying in the post-marketing monitoring and major drug
harm events.

15.4.1 Case Report and Case Series Study

Case report is the simplest report of events observed in single patient. As used in
pharmacoepidemiology, a case report describes a single patient who was exposed to
a drug and experienced the effect, especially an adverse outcome. For example, a
published case report about a young woman suffered a pulmonary embolism who
was taking oral contraceptives. Case report could be used to generate hypotheses
about ADR, but more rigorous study designs are needed to test the hypotheses.
Nevertheless, unreliable conclusion might be resulted in the information bias by
patients or medical staff in case report. After drug marketing, case series study is the
most useful for two related purposes. Firstly, it can be used to explore the incidence
of ADR/ADE. Secondly, it can be beneficial to finding some special or delayed
adverse reactions. However, causal correlation cannot be inferred due to the absence
of control group. Thus, case series study is useful in providing clinical descriptions
of a disease or patients who receive an exposure, but not in determining causation.

15.4.2 Ecological Study

Ecological study is used to explore the relationship between drug use (exposure) and
outcome (both in terms of beneficial and adverse effects) in different populations.
Ecological study includes two types: ecological comparative study and ecological
trend study. The study on the association between thalidomide and phocomelia is a
typical ecological trend study. The sales curve of thalidomide was consistent with
the incidence of phocomelia, and there was about one pregnancy interval between
them. All these proofs suggested that thalidomide was the cause of phocomelia.
However, the results of ecological study should be carefully discussed to avoid
ecological fallacies. For lack of data of individuals, only group data is utilized in the
 15

Case report

Case series study

Descriptive study
Ecological study Develop hypotheses
Pharmacoepidemiology

Observational study Cross-sectional study

Case-control study
Primary study Analytical study
Cohort study

Randomized controlled trial,

Experimental study Test hypotheses
Methods of pharmaco- Non-randomized controlled trial
epidemiology study
Systematic review,

Meta-analysis,

Nonsystematic reviews,
Secondary study
Reviews, Guidelines,

Decision analysis,

Economic analysis

Fig. 15.1 Hierarchy of pharmacoepidemiological study design

273
274 X. Liu and J. Hou

ecological study. Moreover, the confounding variables could not be adjusted in this
study. Therefore, the ecological study is unable to distinguish which factor is the real
cause among exposures coinciding with the outcome.

15.4.3 Cross-Sectional Study

Cross-sectional study is to describe the distribution of drug use-related states or

events among a specific group of population at a specific time and explore the
influencing factors of ADR/ADE, which would provide clues and scientific bases
for further disease study and the rational drug use.

15.4.4 Case-Control Study

Case-control study is to look for the exposure differences between cases with a
disease of interest and controls without the disease in antecedent exposures. This
design can be extremely useful when the disease is relatively rare, or when time or
resources are limited. As a classic example, information was collected from only
eight cases with vaginal adenocarcinoma, and each case was matched with four
patients without vaginal adenocarcinoma. The information on the cases, controls,
and their mothers were collected. Through comparing the data between case and
control groups, it was found that the use of diethylstilbestrol to preserve the fetus in
the early stages of pregnancy caused vaginal adenocarcinoma in their daughters.
Nevertheless, the case-control study generally collects information on exposures
retrospectively by referring to medical records or by questionnaires or interviews.
Therefore, the exposure information retrospectively collected is one of limitations in
the case-control study. As such, the proper selection of controls is a crucial task to
reduce selection bias. If the case-control study is done well, the subsequent cohort
study or randomized clinical trial (RCT) will generally verify the results of the case-
control study.

15.4.5 Cohort Study

Cohort study is used mainly to test etiological hypotheses. It generally is used to

compare the incidence of certain outcome (both beneficial and adverse effects)
between the drug (exposed) group and no drug (unexposed) group. Cohort study
might be performed either prospectively or retrospectively. In prospective cohort
study, the participants are divided into two groups according to whether they take
drugs at baseline. The outcomes are not available at the beginning of the cohort, and
should be collected during the follow-up among a period of time. For example, two
15 Pharmacoepidemiology 275

groups of women of childbearing age who took oral contraceptives or other contra-
ceptives were followed up to collect and compare the incidence of venous throm-
bosis. However, the main disadvantage of the prospective cohort study is that it
needs a relatively long time until a sufficiently large number of events occur. For rare
outcomes or delayed drug effects, the follow-up period may span one, or even
several decades. In this circumstance, the prospective cohort might be not suitable,
especially when some drug might be harmful which contraries to the ethics. In the
retrospective cohort study, the outcomes under study had already occurred, and the
exposure history of drug is obtained using medical records, questionnaires, or
interviews. Although there is a long period between drug use and outcome, the
collection and analysis of data can be completed in a short period. Moreover, there is
no ethical issues. Thus, the retrospective cohort study may be a better choice for
ADR study. Significantly, the information on the exposure history of drug and
outcome should be complete and reliable.
Cohort study is useful in post-marketing drug surveillance study, which looks at
any possible effect of a newly marketed drug. For example, cimetidine was marketed
in 1976, and post-marketing monitoring began in the United Kingdom since 1978.
During the follow-up period, there were 9928 patients using cimetidine and 9351
controls without using cimetidine with complete hospitalization and death records in
four regions. With the improvement of drug post-marketing monitoring and database
sharing, a “computerized” cohort study will play a significant role in ADR study.
The exposure is determined before the outcome occurs in cohort study. Therefore,
the causal association is more convincing compared with case-control study. How-
ever, there is still confounding bias because there is no good comparability between
the drug group and no drug group. Thus, the stratified or logistic regression analysis
are necessary to control confounding bias in cohort study.

15.4.6 Experimental Study

Experimental study, especially randomized controlled trial (RCT), is the gold stan-
dard for evaluating the efficacy of drug. Nevertheless, it cannot be used to verify the
causal association in pharmacoepidemiology due to ethical issues. Sometimes, under
certain conditions, the reverse verification of causal association can be conducted in
the population. Reverse validation based on experimental study in population is that
remove the hypothesized etiology (drug), and follow-up to observe the trend of
incidence of related outcomes of drug use. For example, in 1982, the Ministry of
Health in China eliminated a batch of drugs (containing tetramisole). Then, the
number of encephalitis syndrome in Wenzhou city decreased in 1983–1984 after the
elimination of these drugs.
276 X. Liu and J. Hou

15.4.7 Systematic Review and Meta-Analysis

Secondary use of published data has become an important means of active monitor-
ing of drug safety in many countries. Systematic review and meta-analysis have been
widely used in the medical research in the past 20 years, especially when there are
doubts about the efficacy or safety of drugs, and when there are few studies with
large samples.

15.4.8 Real-world Study

With the widespread use of computers, mobile devices, wearables, and other bio-
sensors to gather and store huge amounts of health-related data, real-world study has
been paid more and more attention to the real-world data evidence. Under the real
medical conditions, the clinical effects between drugs and drugs, vaccines and
vaccines, surgical treatment and drug treatment, inpatient and outpatient treatments,
etc. can be observed and compared in real-world study in which no other interven-
tion factors are added other than the test factor. See Chap. 19 for details.

15.4.9 Newly Derived Study Design

1. Derivative study of case-control study: Nested case-control study, case-cohort

study, case-crossover design (to explore the effect of short drug use on the acute
adverse events), and case-time-control design (to solve the effect of drug use on
outcome when there are mixed indications caused by the disease severity or time
of taking medicine changes).
2. Pharmacogenetics and pharmacogenomics: They aim to explore the relationship
between genetic factors and drugs at the molecular level, and estimate the effect
of genetic factors on drug efficacy. They can not only promote the development
of new drugs, but also provide scientific bases for effective individualized
treatment plan in the clinical practice.
3. Propensity score and instrument variable: To solve the incompatibility between
exposure group and unexposed group in observational study, especially in real-
world study, statistical techniques (such as propensity score and instrument
variable) are developed to adjust for confounding factors in
pharmacoepidemiology.
15 Pharmacoepidemiology 277

15.5 Data Collection and Analysis

15.5.1 Data Collection

Data collection in the pharmacoepidemiology is mainly divided into primary data

collection, existing (secondary) data collection, as well as the combination of two
types of data collection. The sources of data mainly include routine data, literature
data, and surveillance system of ADR.

15.5.1.1 Routine Data

1. Vital Statistics
(1) Demographic data: Demographic data can be obtained through population
census, sampling survey, and household registration system. These data are
mainly used for ① the denominator, used to calculate some relative numbers,
such as per capita consumption of drugs, drug expenditure, the proportion of drug
users in the population, etc.; ② the standardized rate or the standard population
composition, so as to compare the results of pharmacoepidemiology among
different regions; ③ demographic characteristics, such as age and gender, are
important factors that influence drug use in quantitative study on the relationship
between these factors and drug use.
(2) Death data. The death data provides the distribution of the causes, sex, age
and other information of deaths. Exploring the relationship between mortality and
drug use or sales could provide clues for the future study.
(3) Disease data: The disease data is an important source of data frequently
used in pharmacoepidemiological study. Such data can be obtained from litera-
ture published by medical institutions or professional prevention and treatment
institutions. The main applications of disease data include the following: ① to
provide the background data, ② to evaluate the effect of drug, ③ to provide clues
for the future study, and ④ to assess the causal association between drug use and
outcome (both beneficial and adverse effects).
2. Data collected by relevant agencies
This data comes from medical and drug administration and institutions for
academic research, such as FDA and State Administration of Traditional Chinese
Medicine collect and preserve data on the pharmaceutical production and sales,
customs preserve data of import and export of pharmaceutical products, National
Center for ADR Monitoring has data of ADR monitoring, the medical insurance
agencies have data of prevalence, medication use, expenses, and other related
information of insured population.
3. Data from pharmaceutical companies and manufacturers
Pharmaceutical manufacturers generally obtain materials related to their own
products through the accumulation of daily working materials, special investiga-
tion and literature search, etc. Pharmaceutical manufacturers often have the data
278 X. Liu and J. Hou

of drug purchase and sale. The integrated pharmaceutical data is of great signif-
icance in the study of pharmacoepidemiology. However, the protection of com-
mercial intelligence makes it very difficult to obtain such data.
4. Hospital information
Because hospitals mainly carry out the diagnosis and treatment of disease,
most of the data obtained from hospitals can be used for pharmacoepidemiology
study. The data mainly includes drug warehousing records, prescription, outpa-
tient and inpatient medical records, and drug expenses. But the data of hospital
does not represent the entire population, and the information of each hospital has
its own characteristics. All these circumstances may lead to the selection bias.
Simultaneously, when analyzing data, more attention should be paid to the
hospital level, nursing quality, hospital facilities, medical expenses, and so on.

15.5.1.2 The Literature

The literature in medical journals play an important role in discovering ADR and
preliminarily assessing the causal relationship. Published literatures are important
sources of systematic review and meta-analysis. Based on mathematical and statis-
tical methods, bibliometrics analysis is used to quantitatively analyze the data of
literature.

15.5.1.3 ADR Monitoring and Reporting System

ADR monitoring and reporting system refers to the process of discovery, reporting,
evaluation, and control of ADR, with the purpose to effectively control ADR, to
prevent the occurrence of ADE, as well as to ensure the safety of drug use. The
common methods of monitoring ADR in the world include spontaneous reporting
system (SRS), intensive hospital monitoring, intensive medicines monitoring, and
expedited reporting. The main international monitoring agencies for ADR include
the Uppsala Monitoring Centre (UMC), International Society of Pharmacovigilance
(ISOP), the US Food and Drug Administration (FDA), and others, such as the
Council for International Organization of Medical Sciences (CIOMS), the
European Medicines Agency (EMEA), and National Center for ADR Monitoring,
China.

15.5.2 Data Processing and Analysis

After data sorting, checking and processing of missing data in

pharmacoepidemiology study, appropriate statistical analysis method should be
adopted for data analysis according to different study designs, research purposes,
15 Pharmacoepidemiology 279

and data types. However, data analysis methods of pharmacoepidemiology study

also have their own characteristics.

15.5.2.1 Mining and Analysis of ADR Monitoring Database

The measure of disproportionality is used to analyze ADR monitoring data. This

method is based on the classical 2 × 2 fourfold table (Table 15.1). The basic idea is to
estimate the ratio of the actual amounts of ADR related to a certain drug in SRS to
the expected amounts or the number of adverse reactions caused by other drugs. For
example, the center for pharmacovigilance in the Netherlands uses the ratio as the
reporting odd ratio (ROR), which is calculated as ROR = AD/BC. If the ratio is large
enough (“out of balance”), there is likely to be certain association between the
suspected drug and the suspected ADR, and the association is not caused by an
opportunistic factor or “noisy background” of the monitoring database.

15.5.2.2 Mining and Analysis of Prescription Database

The prescription database is also a resource that can be fully mined and analyzed.
Prescription sequence analysis (PSA) is a method to monitor ADR based on reliable
and complete drug prescription records. When the adverse reaction of one certain
drug is the therapeutic indicator of other drugs, the prescription record will show a
specific sequence of drug use, and a specific frequency distribution in a large
prescription database. For example, drug A and drug B, drug A is the original
prescribed drug. If drug A leads to some adverse reactions which require drug B
to treat. In this way, the frequency distribution of the two drugs in the prescription
database will change.
Prescription sequence symmetry analysis (PSSA) develops based on PSA. The
method is to assess whether a drug is associated with an event by evaluating the
symmetry of the event distribution before and after taking a specific drug. For
example, drug A may cause some adverse reactions which need to be treated by
drug B. Firstly, if there is no causal correlation, the patients who took drug A and B
are equally ranked in the database within a certain period of time, in other words, the
number of patients who first took drug A and then took drug B is the same as the
number of patients who first took drug B and then took drug A. However, if drug A
really can cause adverse reactions requiring drug B to treat, then the prescription of
drug A will lead to an increase in drug B, which will result in an asymmetric
sequence distribution.

Table 15.1 Measure of Suspected event Other events

disproportionality
Suspected drug A B
Other drugs C D
Shen and Qi [12]
280 X. Liu and J. Hou

In summary, pharmacoepidemiology is still a relatively new scientific discipline

although there has been tremendous progress in the development of the methods.
Pharmacoepidemiology has taken advantage of the principles and methods of
epidemiology to study the drug safety and effectiveness in human population. It is
of great interest to explore why the reactions to the same drug use vary from
individual to individual. The genome study will be greatly improved with the
development of the pharmacogenomics and molecular biology. Besides, new study
designs and statistical approaches may emerge as consequences of these
developments.
Chapter 16
Evidence-Based Medicine and Systematic
Review

Qi Gao and Huiping Zhu

Key Points
• EBM is conscientious, explicit and judicious use of current best evidence in
decision-making of the care of individual patients.
• Systematic review aims to answer a specific research question through retrieving
the relevant evidence satisfying the pre-defined eligibility criteria.
• Meta-analysis is a quantitative analysis that combines the results of two or more
studies on a given research issue.

16.1 Evidence-Based Medicine

16.1.1 Concept

Evidence-based medicine (EBM) is a clinical discipline that bridges the gap between
research and clinical practice. EBM is dedicated to make decision-making more
objective and structured by better reflecting the evidence from researches, especially
from studies on clinical epidemiology. It facilitates a transformation of clinical
practice and medical education by introducing the research evidence in clinical
decision-making. Epidemiologists and clinicians have been intensively involved in
developing evidence-based practice. It has been implemented in almost all fields of
medicine including general practice, pathology, surgery, pharmacotherapy, den-
tistry, and nursing.
EBM is defined as “the conscientious, explicit and judicious use of current best
evidence in decision-making of the care of individual patients.” As its application
expanded from individual patients to health-care services and health professions,

Q. Gao (✉) · H. Zhu (✉)

School of Public Health, Capital Medical University, Beijing, China
e-mail: [email protected]; [email protected]

© Zhengzhou University Press 2023 281

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_16
282 Q. Gao and H. Zhu

EBM is also named as evidence-based health care (EBHC) or evidence-informed

health care (EIH) or evidence-based practice (EBP). EBM involves two essential
principles: (1) it delineates that scientific evidence alone is not sufficient in making a
clinical decision, and decision-makers need to take into account the patient’s values
when assessing the benefits and risks of any treatment strategy; (2) EBM displays a
hierarchy of evidence to guide clinical decision-making. Whereas, the hierarchy is
not absolute, and should reflect what different levels of evidence refers to, and
describe the context and agents.
EBM suggests that a valid set of rules can be a complement to medical training
and common sense for clinicians to interpret the results of clinical research correctly.
EBM requires careful, systematical, and specific application of knowledge acquired
through the combination of individual clinical expertise with the best available
external evidence obtained from systematic research. Clinicians should be clearly
aware of the systematic evidence and make pragmatic and ethical decisions in terms
of patient care. These bring about a question of whether evidence can only be taken
as proof, say, from randomized clinical trials (RCT), conducted by a particular
population in a particular country, or from a particular magazine. Actually, the
evidence is obtained not only from the evidence derived from RCTs conducted by
academic medical centers or publications from two or three admitted top journals,
but as well as a large amount of evidence, including the patient’s preference for
treatment and acceptable resources. This contradicts the question of making a
decision just by means of proof. EBM underlines the need to actively seek out all
valid and relevant data and to continually evaluate such data to ensure its accuracy
and applicability.
EBP is decision-making process by which someone makes clinical decisions
using the best available research evidence, his/her clinical expertise and patient
preferences, in the context of available resources. The concept of EBM has been
increasingly accepted in the field of health care. There is no doubt that the practice of
EBM by using the best available scientific evidence in medical research can solve the
specific problems in clinical practice. EBM provides a sound scientific basis so as to
achieve efficiency, consistency, high quality, and safety in medical care. Corre-
spondingly, EBM experts set a focus on a clinical guideline; retrieve, evaluate, and
synthesize the evidence; summarize the benefits and risks, and determine the fitness
of interventions.

16.1.2 Development of EBM

The term “evidence-based medicine” has existed for a long time. Clinicians who
received formal medical education make decisions during clinical practices based on
the clinical features of patients, combined with their clinical expertise. Thus, to some
extent, the clinical procedures for diagnosis and treatment are certainly evidence-
based, although there may be some shortages when adopting the latest and best
16 Evidence-Based Medicine and Systematic Review 283

evidence. The present clinical decision-making process adopted by clinicians should

not be considered as “empirical medicine” simply.
In essence, clinical medicine is a branch of practical science, and constantly
develops along with the development of natural science and clinical science. Thus,
it is necessary for the clinicians to continually update their knowledge, learn, master,
and apply advanced skills and theories to guide their clinical practices in order to
improve their clinical work. Just as Dr. Sydney Burwell, dean of Harvard Medical
School put it: “Half of what you are taught as medical students will in 10 years have
been shown to be wrong. Moreover, the trouble is, none of your teachers knows
which half.” That also explains the importance of constant learning and knowledge
updating.
Since the late 1970s, with the increasing development of clinical epidemiology
and advanced clinical research methodology, and the emphasis on scientific design,
measurement, and evaluation, clinical research has improved dramatically, which
produced abundant high-quality clinical outcomes and elicited a series of methods
and standards of critical appraisal. All of these have been accepted and applied in the
international medical field, which greatly improve the development of clinical
medicine and the practice of EBM.
In the early 1980s in McMaster University – one of the places where the clinical
epidemiology originated, Dr. David Sackett and his colleagues from the department
of clinical epidemiology and internal medicine held a workshop on “How to read
clinical literature” for young resident doctors. The residents incorporated the clinical
problems of patients, medical literature retrieval and evaluation, and application of
the current best evidences being generated by medical researchers worldwide into
their clinical practices. They received the training of EBM and achieved great
success on the basis of studying the principle and methods of clinical epidemiology.
Since 1992, JAMA and other journals published a series of review articles, in which
Sackett et al. named the new method as “evidence-based medicine.” Afterward,
initiated by Haynes and Sackett, the American College of Physicians established a
journal club, namely, ACPJC. In order to enhance the development of EBM, experts
in clinical epidemiology and clinical medicine selectively and systematically ana-
lyzed and evaluated the articles that were published in 30 famous medical journals
around the world since 1991. They refined the selected articles, rewrote the abstracts
and comments, and then published them in Annals of Internal Medicine as supple-
ments. All these were recommended to clinicians so as to use the best evidence
during the practice of evidence-based medicine. In 1995, Sackett in Oxford
established the Evidence-Based Medicine Center of the United Kingdom. Sackett
and his colleagues successively published EBM monographs, and some EBM
journals which are cohosted by BMJ and American College of Physicians. What is
more, the Cochrane Collaboration was established in 1993. Cochrane Collaboration
extensively collects the study results of RCT, and then conducts systematic review or
meta-analysis based on a strict evaluation of the quality of the RCTs. It also
recommends valuable study findings to clinicians and other professional practi-
tioners to help them to practice EBM.
284 Q. Gao and H. Zhu

In 1996, the Evidence-Based Medicine Center of China and the Chinese

Cochrane Centre were established under the support of the Ministry of Health of
China. The two organizations provide training programs for medical professionals,
develop extensive national and international cooperation, and publish two journals
on evidence-based medicine. In addition, monographs on EBM and national
evidence-based medicine teaching materials have been published. All of these
accelerated the practice of clinical medicine and preventive medicine, and improved
the quality of medical care in China.

16.1.3 Categories of EBM Practice

EBM practitioners can be grouped into two types: one is the producer of the best
evidence, and the other is the user of the best evidence. Producers of the best
evidence include clinical epidemiologists, clinical experts, health statisticians, med-
ical sociologists, and scientific medical information workers. They collect, analyze,
evaluate, and integrate the best evidence from more than two million articles of
biomedical literature around the world. They aim to provide evidence for clinicians
to practice EBM. For the time being, the best resources of clinical evidence are
Clinical Evidence, ACPJC, EBMJ, and Cochrane Library, which are published by
the BMJ. Evidence producers are the important components of EBM, and EBM
practice would not be processed without their hard work.
These experts will not finish their work until they push this best evidence to be
applied to EBM practice. They dedicated to provide EBM education for medical
students and clinicians. The only way to achieve the real purpose of EBM is to
transform the best research evidence into health prevention and treatment services
for patients at the highest level, and to enable the clinicians to learn and apply these
theories and methods of EBM practice.
Users of the best evidence are the medical personnel engaged in clinical medi-
cine, including policy-makers. In order to make a diagnosis or treatment decision for
patients, or to make health management and policy decisions, they should consider
their practical issues, and seek, identify, understand, and apply the best of the latest
scientific evidence.
Both the producers and users should not only have clinical expertise, but also
possess the knowledge of the related subjects. The difference between them is
simply based on the level of the requirements. Of course, evidence producer can
also be an evidence user, meanwhile, an evidence user can also become an evidence
producer.
16 Evidence-Based Medicine and Systematic Review 285

16.1.4 Procedures of Practicing EBM

The practice of EBM is composed of five steps which can broadly be categorized as
follows:
Step 1: Translate the indetermination (causation, diagnosis, therapy, prognosis,
prevention, etc.) into an answerable question.
What a doctor needs to do first is to identify the problem of his/her patients. It is
necessary for him/her to deal with the problem using different useful knowledge.
Doctors could track down published evidence and contemporaneous research
review as the basis for clinical decisions. The primary task is transforming clinical
problems into questions. Without an answerable question, no more exploration
and research can be done. Also, a good question can help clinicians to make a
good strategy in collecting evidence to resolve clinical problems. However, since
the practitioners of EBM are highly qualified clinicians who have varying degrees
of expertise and backgrounds, clinical questions should be different in clinical
practice. And even when different doctors face the same patient, the questions
they raise will be different as well.
“PICOS” Model
“PICOS” model is usually used in building a specific clinical question. “P” means
patients or population; “I” refers to intervention or exposure; “C” means the
control group or other interventions; “O” refers to the outcome; and “S” means
study type.

Patients or Population
A clinical question must be used to identify a problem of patient. When defining the
“P,” it is essential to ask the important characteristics of the patient, including
(a) primary problem; (b) patient’s main health concern; (c) health status; (d) age,
sex, and race; and (e) current medications.
For example, ascites with or without infection is an important clinical problem for a
patient with liver cirrhosis. If it is not determined whether the condition is
complicated with spontaneous bacterial peritonitis, then a timely and appropriate
medical treatment cannot be performed.

Intervention or Exposure
The second step in the PICO model is to identify “I.” What intervention that the
doctor chooses for the patient is the main consideration. The interventions include
the use of diagnostic test, treatment, medication, and so on.
And how to make an appropriate intervention? There are plenty of factors affecting
the impact of intervention, including exposures, etiology, treatment, prognostic
factors, the patient’s understanding, and compliance. For instance, when treating
a patient with peptic ulcer, doctors must consider the cause firstly, since the
treatment plan varies depending on whether the patient’s stomach ulcer is due to
286 Q. Gao and H. Zhu

H. pylori (HP) infection, or due to the use of nonsteroidal anti-inflammatory

drugs, or due to stress.

Control
The third step is to identify the “C.” Mostly, there will be a control, which is the
contrast used to compare with the intervention. The control is the only optional
component in a “well-built” question. Then, how to make a choice? For example,
cancer can be treated by surgery, chemotherapy, radiation therapy, or other types
of therapy like intervention. The choice of therapy should be considered
according to the disease condition and the economic status of patients, as well
as the views of their family members.

Outcome
The last step is to identify the outcome that a doctor desires to achieve. The outcome
should be measurable. Outcomes can be defined as an improved sign of symptom
or function, survival, mortality, and disability. Different “types” of outcomes
refer to the different clinical questions.

Study Type
What is the best study design to find the evidence to answer a clinical question?
Systematic review of double-blind, randomized controlled trials, cohort studies,
case-control studies, or case series? Which is it depending on what type of clinical
question that a doctor is asking.

Step 2: Systematically retrieving the available best evidence to answer the clinical
question
Quality of Evidence
Users of clinical evidence need to know how much confidence they can place in the
evidences. EBM classifies clinical evidence as several types and rates them in
order from the strongest to the weakest levels according to the strength of their
freedom from the various biases that exist in medical research (Fig. 16.1). For
instance, the systematic review of randomized, placebo-controlled, triple-blind
trials with allocation concealment and complete follow-up involving a homoge-
neous patient population and medical condition provides the strongest evidence
for therapeutic interventions. By contrast, expert opinions and case reports have
little value due to the biases inherent in observation and reporting of cases, the
placebo effect, etc.

The 5S Model
The first “S” refers to original studies which is at the bottom of the “5S” model; the
second “S” syntheses include systematic reviews and meta-analysis; synopses
(brief comments of original research articles and reviews) is at the third level; then
summaries (concise descriptions of an individual study or a systematic review) is
16 Evidence-Based Medicine and Systematic Review 287

Fig. 16.1 Levels of

evidence

at the fourth level; and the top of “5S” model, systems like computer decision-
making system which links individualized patient characteristics to the current
evidence. Original studies, syntheses, and synopses often evaluate one aspect of
health-care problems, but summaries integrate the best evidence available from
the lower layers, and form a complete chain of evidence relating management
options for a given health issue. Summaries can be made universally available,
and is more feasible to keep up with the latest evidence.

Step 3: Critical appraisal of evidence for its validity that can be categorized into the
following aspects:
Systematic mistakes stem from information bias, selection bias, and confounding
variables;
Aspects of diagnosis and treatment that are quantitative;
Scale of the effect;
Clinical importance of the result;
External validity or generalizability.
Step 4: Application of the critically appraised evidence into clinical practice while
taking into account of doctors’ clinical expertise, the patient’s unique biology,
and values.
Step 5: Evaluation of performance (effectiveness and efficiency in taking 1–4 steps
and seeking ways to improve for the next time)

16.2 Systematic Review and Meta-Analysis

16.2.1 Systematic Review

Systematic review is a method to synthesize literatures. It aims to answer a specific

research question by collecting all relevant evidence that satisfies the predefined
288 Q. Gao and H. Zhu

criteria. It begins with a specific question and predefined criteria for studies, and then
uses systematic and reproducible methods to search and select eligible studies. All
studies searched are evaluated for possible bias before they are synthesized. The
entire process of systematic review is clear and can be repeatedly performed.
Although systematic review can provide an array of information in a given area, it
is limited by the quality of original literatures, the method of conducting systematic
review, and the reviewers’ levels of background knowledge. Therefore, it is neces-
sary to assess the reliability of systematic reviews before applying them into practice.

16.2.1.1 Cochrane Systematic Review

A Cochrane systematic review is done by reviewers from the Cochrane Collabora-

tion according to the standard Cochrane Handbook under the guidance of Cochrane
review groups. Cochrane review is of higher quality since it adheres to a strict
process and quality control system. Cochrane reviews are updated every 2 years or
when new evidence becomes available. Thus, Cochrane review is considered to be
the single best source of evidence of effectiveness of interventions. For now,
Cochrane systematic reviews focus on the prevention and treatment of diseases.

16.2.1.2 Importance of Systematic Review

Meeting the Challenge of Increasing Information Supply in Medicine

Researchers, clinicians, or decision-makers need a large amount of information to

make scientific decisions; however, there are too many messages for them. Every
year, over two million medicine-related literatures are published in more than 20,000
biomedical journals, and the growth rate is 6.7% annually. A clinician has to read
19 professional papers every day to keep up with the latest development in their
fields. Systematic review can select the essential information and discard the dross
using strict selection and evaluation method. Also, it combines the true and reliable
information with clinical application value, which can provide scientific basis for
decision-making.

Timely Transforming and Applying Study Results

Evaluation of the treatment of malignant tumor, cardiovascular and cerebrovascular

diseases, and various other chronic diseases needs to conduct large-sample clinical
trials as far as possible, especially RCT. However, large-scale RCT will take a lot of
manpower and time, material resources and financial resources, which is usually
beyond the capacity of an institution. Although there are numerous clinical studies
now, most of them do have small sample sizes, and the results of them are unreliable.
Hence, using systematic review to synthesize results of several homogeneous
16 Evidence-Based Medicine and Systematic Review 289

clinical trials which are of higher quality can produce relatively more reliable results
which can be applied for clinical practice and decision-making.

16.2.1.3 The Difference Between Systematic Review and Traditional

Review

Systematic review differs from traditional review in several ways. Traditional review
usually tends to be descriptive, and they are liable to bias. Systematic review, on the
other hand, needs a comprehensive protocol and search strategy to obtain all eligible
studies on a specific topic. Also, systematic review often includes a meta-analysis,
and can be updated continuously when new research becomes available, but this is
not the case of traditional review.

16.2.1.4 How to Do a Systematic Review?

Systematic review can assess and synthesize several clinical studies with contradic-
tory results in a strict way so as to resolve disputes and make a more reliable
conclusion, which provides appropriate guide for clinical practice and decision-
making. Nevertheless, the poor quality of included primary studies or inappropriate
methods of conducting systematic reviews can introduce bias in the review. There-
fore, the methods and process of conducting a systematic review are crucial to ensure
the accuracy and reliability.
Systematic review is just a research method and is not limited to RCT or
evaluating the efficacy of interventions. A systematic review can be done to inves-
tigate causes, diagnosis, treatment, prognosis, or health economics of disease, and it
can be divided into systematic review of controlled trials and of observational study
according to different study design of the included primary studies. In addition,
systematic reviews can be qualitative or quantitative according to whether reviewers
have used statistical method (meta-analysis) to analyze data.

Determining a Title and Formulating a Protocol

A systematic review usually stems from important but controversial clinical ques-
tions about the treatment and prevention of diseases encountered in clinical practice.
For instance, whether using low doses of aspirin among high-risk population can
prevent the occurrence of cardiovascular disease? What are the differences on the
efficacy and safety between early laparoscopic cholecystectomy (within 7 days after
the onset) and delayed laparoscopic cholecystectomy (6 weeks after hospitalization)
for patients with acute cholecystitis?
To avoid duplicate work, it is necessary to conduct a comprehensive search first
to find out whether there is an existing or ongoing systematic review or meta-
analysis addressing the same clinical issue. If yes, then what is the quality of the
290 Q. Gao and H. Zhu

review? If the existing systematic review is out of date or poor quality, then it is
useful to update the existing review or conduct a new one.
Studies included in a systematic review should have similar design and interven-
tions within the similar population. Therefore, when determining the title, four
factors have to be confirmed around the research question: (1) participants, the
type of disease, diagnostic criteria, characteristics, and locations of the study;
(2) intervention and comparator of the study; (3) the key findings (primary and
secondary results) and serious adverse effects; and (4) study design. These factors
are of great importance to guide the search, screening, and assessment of each study,
to collect and analyze data, and to explain the application value of results.
After determining the title, a protocol should be developed that includes title,
background, objectives, methods and literature search strategy, eligibility for inclu-
sion criteria, assessment of bias in included studies, as well as methods to collect and
analyze data.
Overall, the objective and the clinical question about a systematic review has to
be determined before protocol formulating and literatures collecting, which can
prevent the reviewer from manipulating the title and contents according to the
collected data and results in analysis.

Retrieving Literatures

Systematic and comprehensive collection of all relevant evidence is one of the

distinctions between traditional literature review and systematic review. Reviewers
need to use various ways and systematic searching methods to avoid publication and
language bias, which is based on the established search strategy in the project plan.
Both published papers and unpublished materials in several languages such as
graduation thesis, academic reports should be collected. For those published papers,
professional evaluation groups from the Cochrane Central Register of Controlled
Trials and the Register of Controlled Trials use computer search and manual search,
which can not only compensate the shortcoming of searching tool that is unable to
completely label RCT, but also help systematic reviewers to retrieve relevant
original literatures in a rapid and comprehensive way.

Selecting Literatures

Selecting literature refers to identifying literatures that can answer the constructed
question from all collected literatures in line with the established inclusion and
exclusion criteria. Thus, the selecting criteria should be drawn up based on the
established research problem and the four factors including the research problem
(subjects of study), interventions, main study results, and the study design.
There are three steps to carry out literature selection: (1) Preliminary screening:
Screen for literatures which are obviously ineligible according to citation informa-
tion. (2) Reading the full text: Read literatures that are probably eligible one by one.
16 Evidence-Based Medicine and Systematic Review 291

(3) Contacting the author: For literatures providing information in the text is obscure,
then the reviewers can get in touch with the author and retrieve relevant information
for further assessment to decide whether or not to include the literatures.

Evaluating the Risk of Bias for Included Studies

Bias is a phenomenon that study results deviate from the true values, and it is
essential to avoid bias. It can occur in every single step from allocating subjects to
intervention groups, following up the subjects, measuring, and reporting results.
Evaluating the bias risk in the included studies means to assess the degree to which
an individual study can eliminate or minimize bias. Literatures are supposed to
include three aspects contents: (1) Internal validity: It refers to how close the results
of study are to the true value or the influences of various bias such as selection bias,
performance bias, and measurement bias; (2) External validity: It means whether the
study results can be applied to other study populations; (3) Factors affecting the
results: Factors such as the drug dosage, period of treatment, and compliance in
therapeutic trials.
There are five major types of bias: (1) Selection bias: It occurs in the process of
selecting and allocating participants when randomization is not perfectly
implemented. (2) Performance bias: It happens when one group of subjects in an
experiment gets more attention from investigators than another group, and it also
refers to the fact that participants can change their behavior or responses if they know
which group they are allocated in. This type of bias can be minimized or eliminated
by using blinding, which prevents the investigators from knowing who is in the
treatment or control group. (3) Attrition bias: It refers to bias arose from systematic
differences in the way that participants are lost from a study (differences between
people who leave a study and those who continue, particularly between study
groups). Over-recruitment can prevent important attrition bias. Also, tailored replen-
ishment samples and sampling weights can compensate for the effects of attrition
bias. (4) Measurement bias: It is caused by measuring exposure or disease different
between participants in the intervention and control groups, and it might be avoided
by adopting standardized measuring methods and blinding the participants as well as
result measurers. (5) Reporting bias: It occurs when chance or selective outcome
reporting rather than the intervention contributes to group differences. The
prevailing concern about this type of bias is the possibility of results being modified
toward specific conclusions.
There are many ways to appraise the quality of literatures, such as the list or
checklist (there are many items that are not given a score) and scale (each item has a
score and weight according to its importance), but is still no consensus. Since these
assessment methods can be easily influenced by the quality of literature in combi-
nation with some information irrelevant to internal validity as well as the fact that the
score of scales is limited by some subjective factors, Cochrane handbook 5.0 does
not recommend any checklist or scale, and recommends to use a new risk-of-bias
assessing tool created jointly by the methodologists, editors, and systematic
292 Q. Gao and H. Zhu

reviewers from the Cochrane Collaboration. The new tool includes six aspects:
(1) random allocation methods; (2) allocation concealment; (3) blinding to the
participants, implementers of treatment regimen, and results measurers; (4) integrity
of the data; (5) selective outcome reporting; and (6) other sources of bias. Result of
each single study has to be assessed from the six aspects above according to the
standards of “yes” (low bias), “no” (high bias), and “unclear” (lacking relevant
information or the bias condition is uncertain). The first, second, and fifth items are
used to evaluate risk of bias in the included studies while the other three items are
used to assess the different results of the included studies and reveal how biases
influences different results from the same study. The result of risk-of-bias assess-
ment can not only be described by words and tables, but also by graphs which can
display the bias more plainly.
To avoid the selection bias when reviewers select and assess the quality of
literatures, it is useful to adopt blinding methods or considering more researchers
(professional and nonprofessional personnel) to do these works. With regard to the
disagreements existing in selecting and assessing literatures, then the reviewers can
discuss them together or the third party can be asked for help. Also, the consistency
(Kappa value) can be calculated when there are several reviewers to select
literatures.

Extracting the Data

When conducting a systematic review, data is not only the statistical figure, but also
the collection of information such as the basic information about the study, inter-
vention, outcomes, and results. Extracting data should be comprehensive and accu-
rate, and avoid bias, mistakes, and duplication. The extracting process steps are as
follows: determining the data type, designing data collection form, carrying out a
pretest to modify and perfect the data collection form, extracting data, checking data,
and handling disagreements.
The extracting data include four aspects:
① General information: such as author, title, date, original literature number, and
source;
② Characteristics of study: including the research method, characteristics of partic-
ipants, settings of study, the type of design, intervention, and preventive and
control measures for bias;
③ Results: outcome measurement and the results, loss to follow-up and dropout, and
adverse effects.

Analyzing Data and Reporting Results

Results can be obtained using nonquantitative synthesis or quantitative synthesis

methods.
16 Evidence-Based Medicine and Systematic Review 293

① Nonquantitative synthesis (NQS)

NQS adopts description method including the participants, intervention strategy,
result of the study, the quality of the study, and design technique. And all these
presents using forms so that readers can know the process of the study, whether
the study method is rigorous, the difference among set of single studies, how to do
a quantitative synthesis, and the explanation of the results.
② Quantitative synthesis
Quantitative syntheses include heterogeneity tests, meta-analysis, and sensitivity
analysis.
Heterogeneity tests: Heterogeneity in the results is common. Heterogeneity among a
number of different studies may be produced by differences in study design,
study populations, and chance, and the extent of heterogeneity might have an
impact on the conclusions of a meta-analysis. Heterogeneity can be divided into
three categories: Clinical heterogeneity refers to the difference in a set of single
studies like the patient factors (such as disease severity, age, and so on), inter-
vention (such as drugs, dosage, and so on), and so on. Methodological heteroge-
neity means differences in different studies (clinical trials), such as blindness in
trials, methods of measurement, etc. Statistical heterogeneity refers to the differ-
ence in the effective value of intervention strategies, which is the result of the first
two kinds of differences. Heterogeneity test is used to check the degree of
variation on the results of the original studies. If the test result is statistically
significant, people should explain the reason and consider whether it is appropri-
ate to combine results/findings. There are two methods to make sure whether the
respective results have the same property. Firstly, it is to inspect by chart to check
whether the effect value of the results overlap between credibility interval. If the
credibility gap is very significant, then synthetic analysis cannot be conducted
and stochastic effect model should be taken. Q test is another method that can be
taken directly. Based on this, the quantitative method can be used to estimate
heterogeneity. For example, 0–40% indicates that heterogeneity is not significant,
30–60% means moderate heterogeneity, 50–90% refers to a notable heterogene-
ity, and 75–100% means a major heterogeneity.
Meta-analysis: Quantitative analysis of synthesis can be done according to the type
of data and the objective assessment of the amount of selection effects. For
instance, odds ratio, risk difference, relative risk, and number needed to treat as
the effect size to represent the result of synthesis that can be chosen for categorical
variables. And for continuous variables, mean difference can be chosen when
measuring the result using the same unit of measurement, while standardized
mean difference should be selected when using the different units of measure-
ment. When conducting a meta-analysis, it is always to select the fixed effect
model or random effect model, and choose the forest plots to display results.
Sensitivity analysis: Sensitivity analysis is to test those important factors (such as
inclusion criteria, quality of study, attrition, and statistical methods), which could
affect the integration results. Then which factors might influence the integration
results and how to do sensitivity analysis based on actual conditions should be
considered according to the various studies included in the integration.
294 Q. Gao and H. Zhu

Results Interpretation

The interpretation of the results of systematic reviews must be based on results of

reviewer’s conclusions. It should address the following issues:
① The strength of evidence
The strength of evidence depends on the study design and the quality of included
studies, whether there are important methodological limitations, whether there is
a dose-response relationship, how large and significant the observed effects
are, etc.
② The applicability of the results
Value of the systematic review results is determined by considering the relationship
between pros and cons of interventions in patients. In addition, researchers should
consider whether the participants included in the systematic reviews are similar to
the conditions of his/her patients. Is there any discrepancy in biological, social, or
cultural variations, variation in baseline risk, variation in compliance, etc.?
③ Clarification of important trade-offs between benefits and potential harms as well
as costs of the interventions.

Updating Systematic Review

Carrying out systematic reviews is time- and resource consuming, and provide a
snapshot of knowledge at the time of data incorporation from studies identified
during the latest search. Newly identified studies can change the conclusions of
reviews. The validity of reviews can be threatened if they have not been included,
and even the reviews could mislead. Thus, there are clear benefits to updating
reviews when new evidence emerges or new methods develop. An update of a
systematic review refers to a new edition of a published systematic review with
changes that can include new data, new methods, or new analyses to the previous
edition. Updating a systematic review requires assessment and revision of the
question, background, inclusion criteria, methods of the existing review, and the
existing certainty in the evidence. In particular, methods need to be updated, and
search strategies might be reconsidered.

16.2.1.5 Evaluation and Application of Systematic Review

Systematic reviews have played an increasingly important role in healthcare in

recent years, with about 2500 systematic reviews published around the world each
year. They are usually used as a starting point for developing clinical practice
guidelines. Clinicians, nurses, health-care workers, health policy-makers, and even
the general patients read the systematic reviews that were rigorously appraised and
scientifically integrated, which can not only greatly save time, but also improve the
efficiency of using the related useful information. However, the existing
16 Evidence-Based Medicine and Systematic Review 295

evidence-based medicine research literatures have good and bad quality. And only
high-quality systematic reviews and/or meta-analyses can provide a scientific basis
of clinicians, patients, and other decision-makers. Thus, a key stage in a systematic
review is to assess the quality of studies to ensure the application and conclusions are
based on sound evidence.
It is a basic skill for medical workers and/or researchers to be able to determine
whether a systematic review is a high quality. This section describes the basic
principles and methods of assessing the quality of a systematic review.

Assessing the Quality of Systematic Reviews

A high-quality system review should enable the practitioners to understand and

critically judge the authenticity and clinical application value of its results. The
principles of assessment of the quality are slightly different for different types of
systematic reviews. The following section describes the basic principles of assess-
ment of quality for therapy systematic reviews and meta-analysis.

Principles of Quality Assessment for Therapy Systematic Reviews

Assessment of therapy systematic reviews mainly focuses on three areas, that is,
whether the results of a systematic review are true, whether the results of a system-
atic review are important, and whether the results of a systematic review can be
applied to an individual patient.
① The authenticity evaluation on a system review
The evaluation of authenticity of a systematic review result covers four aspects:
(i) Was the systematic review based on randomized controlled trials? Random-
ized controlled trials can control significant sources of bias very well, and
have high homogeneity. The systematic review conducted on randomized
controlled trials with high homogeneity is identified as the highest level of
evidence, while it conducted on nonhomogeneous randomized controlled
trials or non-randomized controlled trials is indicated as lower level of
evidence due to biases.
(ii) Did the systematic review conduct a thorough literature search and describe
retrieval strategy clearly? A systematic search for research studies (system-
atic reviews) means the more comprehensive literature collected, the less
affected by publication bias, and the higher credibility. By reading the
description of search strategy, readers can judge whether the literature
collection of the systematic review is comprehensive. Inadequate literature
search, especially in the case of publication bias, may lead to false-positive
results and affect the conclusions of the systematic review. The search
296 Q. Gao and H. Zhu

strategy should follow Preferred Reporting Items for Systematic Reviews

and Meta-Analyses (PRISMA) Statement, and present the flow diagram
according to it. The flow diagram displays the process for selecting studies,
and helps the reader to judge whether the literature search of a systematic
review is complete and appropriate.
(iii) Did the review appraise the authenticity of individual studies? The system-
atic review is a type of study that analyzes and summarizes the results of the
original studies. Thus, the quality of systematic review is not only affected
by the methods used, but also affected by the included studies. Systematic
reviews need to clearly state the evaluation methods of the studies included,
and describe whether it uses multiplayer independent evaluation and
appraises the consistency among them.
(iv) Did a meta-analysis of individual patient data is conducted? Individual
participant data from multiple studies are used to conduct a comprehensive
meta-analysis. Meta-analysis has more advantages than other various stud-
ies. For instance, it can examine heterogeneity in included studies at
patient’s level by consistently defining the results of different studies and
unifying the threshold, test hypotheses by using subgroup analysis, deter-
mine the randomization method clearly and judge the quality of assessment
through contacting the authors, and reduce the effect of system bias and
opportunity errors using existing medical record information.

② Are the results of systematic reviews important?

Applying the results of a systematic review to clinical practice should not only
emphasize the importance of using the systematic review findings in clinical
practice, but also make sure whether the systematic reviews include high quality
of individual studies? What main outcomes are reported in the systematic review?
Is the precision of results reported? Specifically, the assessment covers the
following:
(i) Are the results consistent across the included studies? System reviews
usually include high-quality original articles, with sufficient individual
study and better homogeneity of the study results. The authors should test
the heterogeneity first. If the results of a single study are homogenous, then
the findings of different studies can be integrated. On the contrary, if result of
homogeneity test is significantly different, the author needs to explain the
differences and consider whether it is appropriate to combine the individual
study results.
(ii) How accurate is the systematic review? In systematic reviews, more accurate
conclusions are considered mostly. Given the impact that quality assessment
can have on the findings of systematic reviews, it would be reasonable to give
different weights according to the quality of the studies when synthesizing
the results of the included studies. The application of a systematic review is
associated with the outcome indicators. If a systematic review uses low
related intermediate outcomes, it may limit the findings of the systematic
16 Evidence-Based Medicine and Systematic Review 297

review being applied to guide clinical practice. If the outcome indicators are
the event rates, such as mortality and the incidence of serious adverse events,
it may bear important clinical significance even if the combined effect did not
generate statistically significant difference.
Indicators used in systematic reviews include risk difference (RD), odds
ratio (OR), relative risk (RR), weighted mean difference (WMD), number
needed to treat (NNT), and number needed to harm (NNH). Among them,
NNT and NNH are easy to calculate and are among the most clinically useful
statistics. NNT is the number of patients which need to treat to prevent one
additional bad outcome (death, stroke, etc.). When NNT is smaller, preven-
tive effect is better. NNH is an epidemiological measure that indicates how
many patients need to be exposed to a risk factor to cause harm in one patient
that would not otherwise have been harmed. Intuitively, the lower the
number needed to harm, the worse the risk factor. The NNH is a crucial
EBM metric that aids doctors in determining if it is wise to proceed with a
specific treatment that can endanger the patient while yet having therapeutic
advantages. Drugs with a low NNH may still be indicated in specific
circumstances if the number needed to treat (the opposite of side effects, or
the advantages of the medicine) is less than the NNH if a clinical end point is
devastating enough without the drug (e.g., death and heart attack).

③ Can the results of systematic review be applied to an individual patient?

The findings of systematic review are the average effect among all participants. For a
particular patient, to answer whether the findings of a systematic review could be
applied, the following four aspects should be assessed:
(i) Is the patient similar to those in the studies included in the systematic
review? It can compare the differences between the patient and the cases
in the systematic review, focusing on the age, gender, race, comorbidities,
severity of disease, duration, socioeconomic status, cultural background,
complications, compliance, etc.
(ii) Is the therapy feasible and safe in the clinical trial setting? Since there is
difference in socioeconomic conditions, technology, and equipment condi-
tions, the conclusions of systematic reviews sometimes can hardly be
applied to a particular patient, even if the intervention effects are remarkable.
(iii) What are the potential benefits and harms of therapy for the patient? Only
when the advantages outweigh the disadvantages, results of the system
review have applicable value.
(iv) What are the patient’s values and preferences? Incorporating patient values
and preferences as an essential input for decision-making has its potential
merits. EBM stresses that any medical decision-making should take into
account of the combination of personal experience and expertise, current
best research evidence, and patient choice.
298 Q. Gao and H. Zhu

Assessing the Quality of a Meta-Analysis

Assessment of the quality of meta-analysis should consider the following questions:

① Does the question conform to the principle of DOE?
② Does the question state clearly?
③ Is the search strategy clearly described?
④ Are appropriate inclusion and exclusion criteria used to select articles?
⑤ Is there worrying homogeneity? Has it been appropriately addressed?
⑥ Are the statistical methods used correctly? Has a sensitivity analysis been
undertaken? If not, should it have been?
⑦ Does the combined analysis clarify the difference between the test group and the
control group?
⑧ Are the conclusions appropriate?
⑨ Do you put the recommendations of the systematic review in your clinical
practice?

16.2.1.6 Methods of Evaluating System Review

Effective quality assessment is very important for properly using the conclusion of a
system review/meta-analysis. Quality assessment instruments of systematic reviews
mainly include two types: one is to assess the methodological quality of systematic
reviews and the other is to appraise the quality of reporting of meta-analyses. The
poor quality of reporting will affect the applicability of the results of systematic
reviews. Thus, it is recommended to use both of the quality assessment tools when
assessing the quality of systematic reviews/meta-analysis.

Tools for Methodological Quality Assessment

Methodological quality assessment aims to understand whether the review follows

the scientific standards and effectively controls the biases in the process so as to get
true and reliable results. Biases are the main challenge for the quality of systematic
reviews. When assessing the authenticity of systematic reviews, the focus should be
put on examining the quality of the methods used in systematic reviews, and how to
control biases. Usually, methodological quality assessment instrument of systematic
reviews includes the following: A Measurement Tool to Assess Systematic Reviews
(AMSTAR), Critical Appraisal Skills Programme (CASP), Sacks Quality Assess-
ment Checklist (SQAC), and Overview Quality Assessment Questionnaire (OQAQ).
16 Evidence-Based Medicine and Systematic Review 299

Tools for Reporting Quality Assessment

Reporting quality assessment aims to ensure complete, accurate, and transparent

reporting of research studies. The main content of the process of assessment is to
assess the integrity and comprehensiveness of the systematic review. Report spec-
ification is one of the quality assessment tools used to appraise the reporting quality
of the systematic review. It includes asking a research question, review methods,
presentation of the results, discussion, and conclusion. The reporting specification of
the systematic review is mainly used for assessing the quality of a system review
report, and cannot be used to assess methodological quality of systematic reviews.
The Quality of Reporting of Meta-analysis (QUOROM) is the first report spec-
ification used to assess the quality of reporting of meta-analyses of clinical random-
ized controlled trials. QUOROM covers 6 aspects with a total of 18 items, that is,
title, abstract, introduction, methods, results, discussion, and conclusion section.
QUOROM has been considered as a “gold standard” to assess the quality of
systematic reviews/meta-analysis. In 2006, QUOROM was updated as PRISMA.
The PRISMA Statement was released in 2009, and an official update of it is currently
under development. It consists of a 27-item checklist and a four-phase flow diagram
(Fig. 16.2). PRISMA Statement focuses on randomized trials, but can also be used as
a basis for reporting systematic reviews of other types of research, particularly
evaluations of interventions. PRISMA may also be useful for critical assessment
of published systematic reviews. The full score of PRISMA scale is 27 points: “full
report” gets 1 point, “part of the report” gets 0.5 points, and “not reported” gets
0 points. When the score is between 21 and 27, the report is considered relatively
complete; when the score is between 15 and 21, the report is considered to have
some flaws; and when the score is equal to or less than 15, it is considered that there
may be serious missing of information.

16.2.1.7 Application of Systematic Review

Sound and reliable evidence providing by a high-quality systematic review is of

great importance for clinical decision-making. It is necessary for decision-makers to
concern whether the results of a systematic review can be applied to clinical practice.
A high-quality systematic review should enable readers to judge whether the results
of systematic reviews can be used to solve a specific clinical problem. For example,
in a systematic review of disease prevention study, the following aspects are the
main concerns:
1. The clinical problems (the review questions);
2. The basic information such as age, gender, disease, and diagnosis;
3. The inclusion criteria and exclusion criteria;
4. The interventions, such as the intervention method of experimental group and the
control group;
5. The information on outcome indicators.
300 Q. Gao and H. Zhu

Identification

Records identified through Additional records identified

database searching ↓ through other sources ↓
(n = )., (n = ).,

Records after duplicates removed ↓

(n = ).,
Screening

Records screened ↓ Records excluded ↓

(n = )., (n = ).,

Full-text articles assessed Full-text articles excluded

Eligibility

for eligibility ↓ with reasons ↓

(n = )., (n = ).,

Studies included in
qualitative synthesis ↓
(n = ).,
Included

Studies included in
quantitative synthesis
(meta-analysis) ↓
(n = ).,

Fig. 16.2 PRISMA 2009 flow diagram

Systematic reviews encompass all fields of clinical medicine, including the

studies on disease etiology and risk factors, disease diagnosis, treatment, disease
prognosis, rehabilitation, and prevention. Results of systematic reviews have been
widely used to develop clinical practice guidelines, facilitate health policy decision-
making, guide clinical practice, and enlighten scientific research. In addition, there is
a series of guidelines to help interpret and use the systematic reviews.
The application of systematic review should meet the needs of clinical practice,
medical education, scientific research, and health policy-making. Results of system-
atic reviews should reflect the latest evidences and developments in a given area. In
clinical practice, when a clinician needs to know whether a systematic review
conclusion can be applied to solve clinical problems, firstly, he/she must consider
the following questions by checking the results of the systematic review: Are the
benefits important and necessary for clinical decision-making? Are there less poten-
tial side effects than other interventions? Are the costs less than other methods? Is
my patient similar to the participants in the systematic review?
16 Evidence-Based Medicine and Systematic Review 301

16.3 Meta-Analysis

16.3.1 Introduction

A meta-analysis is defined as a systematic review that uses statistical methods to

aggregate quantitative data from at least two and ideally several studies on a given
research issue to produce an overall quantitative estimate of effect. Narrowly
speaking, the meta-analysis refers to the quantitative analysis of a system review.
However, only a few systematic reviews can be quantitatively analyzed because of
the differences in quality of research, design, methodology, etc. The general steps of
a meta-analysis are as follows: formulation of the research questions, literatures
retrieve, making inclusion and exclusion criteria, describing the basic information,
doing comprehensive quantitative analysis, and a series of other steps.
At present, meta-analysis has been widely used in clinical medicine. In particular,
it has been used for small effect size or controversial research (mainly for RCTs),
etiology study, the dose-response relationship research, diagnostic trials, and prog-
nosis research.

16.3.1.1 Basic Concepts

1. Effect size is related to different variables or values caused by treatment effect.

2. Heterogeneity test refers to a test that evaluates whether variation between the
results stemmed from a random error.
3. Bias means a system error in study designs, trial implementation, data analysis,
and results interpretation. It brings about difference between the results and the
real situation, and thus wrongly describes the relationship between exposure and
disease.
4. Publication bias refers to studies which present negative results or insignificant
results are less likely to be published.
5. Sensitivity analysis is that researchers can eliminate lower quality studies or use
different statistical methods to analyze the same data by changing the inclusion
criteria so that to observe the changes of the meta-analysis results.

16.3.2 Steps to Perform a Meta-Analysis

The general steps of a meta-analysis are as follows:

16.3.2.1 Data Extraction

Accuracy and reliability of the data is the key of a meta-analysis. Thus, when
conducting the data extraction, a research should collect data onto many channels
302 Q. Gao and H. Zhu

to ensure complete data included. At the same time, effective quality control
measures should be used to prevent selective bias.

16.3.2.2 Data Types and Effect Size

Data used for a meta-analysis mainly include the following five types: (1) Continuous
numerical variable data: often have the units and can accurately be measured, such as
height and weight. (2) Binary variable data: have two incompatible categories, such
as survival or death. (3) Data in hierarchical classification of variables: can be
divided into multiple classes, and have degree or level differences, like never/rarely,
sometimes and often, (4) Count data: individuals in a certain observation time
experiences many adverse events, such as myocardial infarction and fracture.
(5) Survival data: observation of two types of data at the same time, the occurrence
of adverse events, and the time of adverse events occurring.
Different data types determine different effect size expressions. When the out-
come is a binary variable, the effect size commonly used is odds ratio (OR), relative
risk (RR), absolute risk (AR) or NNT, etc. When outcome is continuous variables,
the effect size is the mean differences (MD) or standardized mean differences
(SMD). For hierarchical data or count data, it can be converted to binary variables
or continuous variables in light of the actual conditions. For survival data, the effect
size used is a hazard ratio (HR).
In addition, the important information, such as sample size, analysis methods,
design, main outcome variables, publication year, and quality control measures
should be included in the study.

16.3.2.3 Heterogeneity Test

When carrying out a meta-analysis, strict literature inclusion and exclusion criteria
should be used to control the heterogeneity source maximally. However, because of
the differences in the research participants, design and statistical analysis model,
heterogeneity will occur inevitably. In such case, there will be a mistake if the results
are combined. Thus, heterogeneity test should be done before conducting a meta-
analysis, and determine whether to estimate the combined effect size according to its
results. If the heterogeneity is obvious, the source of the heterogeneity should be
addressed. The heterogeneity test includes Q test and visual graphic method. Also,
there are some other methods to show the heterogeneity, such as standardized Z
score, radial figure, forest figure, and L’Abbé figure. The most commonly used
method to determine the heterogeneity is to observe confidence interval overlapping
degree in forest figure. If most confidence intervals overlap and there is no obvious
abnormal value, then it can be recognized as a high homogeneity.
16 Evidence-Based Medicine and Systematic Review 303

16.3.2.4 Combining Effect Size Estimates and Hypothesis Testing

Next, appropriate statistical methods should be chosen based on the results of

heterogeneity test. If the heterogeneity is not obvious, namely, assuming that the
theoretical effect size is a fixed value and differences between effect variables caused
by opportunity, then the fixed effect model can be used to estimate the combined
effect size. If there is heterogeneity, and the assumption theoretical effect size is not
fixed, but follows a certain distribution pattern, such as the normal distribution, then
the random effect model can be adopted to estimate the effect. If the heterogeneity
is too great, subgroup analysis and meta regression analysis can be considered, or
only describe the results.
There are many methods to estimate combining effect size, like Mantel–Haenszel
method, Peto method, and variance inversion method. According to the character-
istics of data, different methods can be used. For binary variable, Mantel–Haenszel
method can be used, and for continuous variables, variance inversion method can be
adopted. Z test is used to test whether there is a statistical significance of combined
effect variables.

16.3.3 Fixed Effect Model and Random Effect Model

Model selection depends on the results of heterogeneity test and the theoretical effect
hypothesis. If the heterogeneity test is not statistically significant, then it can be
deemed that theoretical effect size is fixed. Then, a fixed effect model can be selected
to estimate it. On the other hand, if the heterogeneity is larger, and assuming the
theoretical effect size follows a normal distribution, then a random effect model need
be chosen. The random effect model takes variation factor τ2 of the study as the
correct weight, so the result is more robust than the fixed effect model.

16.3.3.1 Fixed Effect Model

For binary variable data, MH method fixed effect model can be chosen to estimate
the combined effect size. If it is continuous variable data, and there is no statistically
significant heterogeneity, fixed effects model can be used for meta-analysis, and
the process is the same as that for binary variable data. The variance inversion
method should be used for combining effect size estimation. For effect size expres-
sion of continuous variable data, mean difference (MD) or standardized mean
difference (SMD) need to be used. When all studies use the same way to measure
outcome variables, MD can be used as the effect size. If those outcome variables
have the same definition, but not have the same measuring scale, then SMD should
be selected, and authors need to explain these results carefully.
304 Q. Gao and H. Zhu

16.3.3.2 Random Effect Model

When the heterogeneity test is statistically significant, and assuming the real effects
size is not fixed, but follows a normal distribution pattern, then the random effect
model can be selected to estimate and combine the effect variables. The random
effect model adds DerSimonian–Laird correction to fix effect model on the basis of
the variance inversion method or MH method. Weight between the two models is
different. The fixed effect model takes the reciprocal of variance as the weight in
individual studies. The random effect model takes the reciprocal of the sum of the
variance in the study and variance between studies as the weight, and the adjustment
results give less weight for larger sample size study.

16.3.4 Evaluating the Result of a Meta-Analysis

16.3.4.1 Heterogeneity Test

If the research has adequate homogeneity, then a fixed effect model, random effect
model, or both can be used to estimate the combined effect size. If the research has
adequate heterogeneity and the source of heterogeneity is known, then meta regres-
sion model or subgroup analysis can be selected. If the heterogeneity test is statis-
tically significant, but heterogeneous source is unknown, then random effect model
estimation is more conservative. When assuming that the effect size is not fixed, but
obey a normal distribution, then random effects model can be adopted. If heteroge-
neity is too great, then meta-analysis cannot be conducted.

16.3.4.2 Robustness of Meta-Analysis Results

Sensitivity analysis is often used to check the robustness of meta-analysis results. In

sensitivity analysis, by changing the inclusion criteria to eliminate lower quality
study, or using different statistical methods to analyze the same data, which can be
used to observe the changes of the meta-analysis results so as to evaluate the
resulting stability. For instance, after excluding some lower quality studies, the
combined effect size is estimated again, then to compare the results with original
meta-analysis results and explores the influence of these lower studies on the
combined effect size and the stability of the results. If the result changes a little, it
has lower sensitivity. On the contrary, after excluding some lower quality studies,
the difference become bigger, or even the meta-analysis gets the opposite conclu-
sions. It is of high sensitivity and poor robustness, and the results need to be
explained carefully.
16 Evidence-Based Medicine and Systematic Review 305

16.3.4.3 Applicability of the Meta-Analysis Results

The application of meta-analysis avoids the limitations of individual clinical trials

with small samples, and makes the results of analysis more comprehensive and
reliable. Therefore, it can provide a good basis for medical decision-making.
Chapter 17
Disease Prognosis

Fang Wang

Key Points
• Prognosis is a prediction of the outcome and influencing factors of the disease
following its onset.
• Risk factors and prognostic factors are often considerably different. The most
common design of prognosis is cohort study.
• Case-fatality and five-year survival are often used to express prognosis. The life
table approach and the Kaplan-Meier method can be used to calculate observed
survival over time.
• Assembly bias, migration, zero bias, and survival cohort bias are major sources of
bias in prognosis study. Randomization, matching, COX proportional hazards
model and other methods may help control bias in prognostic studies.
When a person developed a disease, doctor, patient, and even the patient’s family
members may have lots of questions about the disease. Will it go further to be worse?
Could it be cured? How about the possibility of a worse outcome? How long do the
patients have to continue with their normal activities? All those questions are
discussed about the prognosis of a disease. In this chapter, we will introduce
qualitative and quantitative ways that prognosis can be described.

F. Wang (✉)
School of Public Health, Shanxi Medical University, Taiyuan, China
e-mail: [email protected]

© Zhengzhou University Press 2023 307

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_17
308 F. Wang

17.1 Basic Concepts

17.1.1 Concept of Prognosis

Prognosis is a prediction of the outcome and influencing factors of the disease

following its onset. The outcomes can be recovery, relapse, disability, deterioration,
complications, and death. Prognostic studies are to identify the probability and
possible influencing factors of those outcomes. Understanding of prognosis helps
to predict the future of the patients better. Clinicians may choose more appropriate
decisions on the following aspects: (1) What kind of treatment guidelines should
clinicians follow? and (2) What kind of treatment should be adopted?
A better understanding of influencing factors of disease outcomes can alter the
outcome of a certain disease. If there are several types of medical treatments,
clinicians can compare the effectiveness of different therapy.

17.1.2 Natural History and Clinical Course of Disease

The natural history of disease refers to the prognosis of disease without medical
intervention which can be divided into the following four periods: the biological
stage, subclinical stage, clinical stage, and outcome. Changes like DNA alternation
are subcellular in the biological stage, and thus often cannot be defined due to the
sensitivity of the clinical test. In the subclinical stage, pathologic evidence develops
and could be obtained. Later, when noticeable signs and symptoms like pain,
disfigurement, or fever occur in the clinical stage, patients may come to seek help
from the clinician and then a diagnosis may be made. After treatment, the outcomes
may be cure, disease controlled, or even death.
The natural history of different diseases varies greatly. Some diseases with a short
latency may present obvious symptoms and outcomes in a short period of time, such
as acute infectious diseases. Some chronic noncommunicable diseases may have a
relatively long natural history such as cardiovascular diseases and diabetes. Different
strategies can be taken in different stages of natural history to improve prognosis.
The clinical course is the progression of disease following medical interventions.
Patients receive a variety of treatments that may affect subsequent course of the
disease. The clinical course may be altered by medical intervention; however, there
is no medical intervention in natural history. The earlier the effective treatment, the
better the prognosis. Prognostic researches are about the clinical course and medical
treatment that can improve prognosis and alter the outcome.
17 Disease Prognosis 309

17.1.3 Prognostic Factors

Prognostic factors are conditions that are associated with the outcomes of disease.
Prognostic factors help to identify groups of patients with different outcomes.
Prognostic factors are different from risk factors in two ways. Firstly, risk factors
are those associated with increased risk of a disease in healthy people, whereas
studies of prognostic factors deal with sick people. Secondly, risk and prognosis
describe different phenomena of disease. Risk describes the onset of disease, usually
predicting low-probability events. The incidence of disease varies from 1/1000 to
1/100,000 or even less. The study of risk factors requires a relatively large amount of
population to evaluate or confirm the relationship between exposure and disease.
Prognosis describes a variety of disease consequences following its onset. The
consequences, including recovery, disability, complications, and death are relatively
frequent events.
1. For a given disease, risk factors and prognostic factors are not necessarily the
same and are often considerably different in the following three important points.
Factors associated with an increased risk of disease have little to do with
prognosis, in other words, risk factors do not necessarily make a worse prognosis.
For example, high blood pressure increases the risk of acute myocardial infarc-
tion, but it is not related to a worse outcome of the acute event.
Factors associated with a certain outcome of disease do not have an association
with increased risk of disease. Those prognostic factors are not risk factors of a
certain disease. Infarction location and arrhythmia are prognostic factors of acute
myocardial infarction, but they do not increase the chance of having the disease.
Some factors do have a similar effect on both risk and prognosis. Those factors
can not only increase the risk of a certain disease, but also lead to a worse
outcome. For example, with the increase of age, both the risk of an acute
myocardial infarction attack and the risk of death from it may increase.
2. Prognostic factors are complex and variable, which can usually be described by
several categories.
Timing of diagnosis and treatment: Early diagnosis and proper treatment for
any diseases are important prognostic factors. The 5-year survival rate of gastric
cancer can be up to 100% if discovered early, but may fall to less than 20% for
advanced gastric cancer discovered through normal diagnosis.
Characteristics of the disease: The spectrum of disease – from mild to severe –
is related to outcomes. Patients with mild illness can have a better prognosis than
the severe ones. The duration and pathologic types also vary. Different diseases
have different natural histories.
Pathogenic factors: The quantity, quality, and invasive manner of pathogen
can affect the consequences.
Characteristics of the patients: The demographic characteristics, genetic back-
ground, nutrition, immune system function, and psychological state of the
patients all have some influence on prognosis.
310 F. Wang

Social and family aspects: Economic development level, social insurance

system, local medical condition, family economic situation, the relationship
between family members, and the patient’s religious/belief can affect the prog-
nosis of disease.

17.2 Design of a Prognosis Study

In the beginning, qualified patients with a complete description are selected as the
study population. Then the patients are observed and followed up at a pointed time.
Finally, all the concerned outcomes are measured to describe the prognosis of
disease.

17.2.1 Research Methods

Studies of prognosis are like those of risk. Generally, relevant prognostic factors are
identified through descriptive study, and verified through case-control and prospec-
tive cohort studies. Any method can be chosen, depending on the study purpose,
resources, and time.
In descriptive study, patients with diabetes were randomly selected as subjects.
And then they were asked about habitual, diet, and treatment to identify whether
those factors were possible prognostic factors of diabetes complications.
In case-control study, newly diagnosed diabetes patients with complication, that
is, diabetic nephropathy, were selected as patients; diabetes patients without com-
plications were selected as controls. Cases and controls must both meet the inclusion
criteria to ensure they were from the same base population. Controls may be matched
to cases on age and gender. The frequency of smoke, cyto-factors and other
interesting factors were measured and compared within the two groups. It is efficient
and indispensable as it does not need to collect data from a large number of people.
But selection bias and recall bias are difficult to manage.
The most common design is cohort study. The incidence and relative risk are
measured directly. Prognostic factors are measured before the outcome of the
disease. It is discussed in detail in the following section.

17.2.2 The Patient Sample

It is best for a prognosis study when the patients are selected based on the population
of all people with the disease in a certain region. Under this condition, the patients’
sample is representative and there will be no bias in patient selection. However, most
17 Disease Prognosis 311

studies are hospital-based and the outcomes for patients with the same disease may
be different when selected from different levels of health facilities.
It is important for a prognostic study to thoroughly describe the patients’ charac-
teristics, the sampled and assigned method, the criteria of diagnosis, inclusion, and
exclusion. Thus, it will be good for others to reference the study results and decide
whether the conclusions can be applied to their patients.

17.2.3 Determine the Starting Time Point

Cohorts are observed from a pointed time in the course of the disease, which is called
zero time. The point can be the time when patients are enrolled in the cohort, such as
the onset of symptoms, date of diagnosis, or the beginning of medical interventions.
Whatever, it is especially important to make a clear definition of zero time. If the
patients are recruited near the onset of the disease, the cohort is called the inception
cohort.
What if the patients are observed at different points of the disease? When zero
time changes for patients, precise description of subsequent events would be much
more difficult. Interpretation of the timing of recovery, death, and others would be
hard or even misleading. For example, a cohort of women with breast cancer is
assembled. However, women in the cohort are at different stages of breast cancer.
Those in the early stages can have better survival than those beginning surgical
treatment. Moreover, these results of cancer prognosis would be hard to interpret.

17.2.4 Determine the Outcomes

Outcomes should be clearly defined and a full range of clinical events of the disease
should be included. While clinicians tend to focus on the clinical effects, such as
normalization of blood chemistries or reduction of tumor size, patients are more
concerned about the remission of symptoms. To guide patient care, outcomes should
be related to something that patients can perceive.
Outcomes vary a lot, and death is the easiest to determine. Some outcomes, like
myocardial infarction, usually need to be valued precisely or technically measured.
Others liked the health-related quality of life, are difficult to determine directly, and
often need a set of variables to measure. The value of the prognosis study is
strengthened when blinding is applied to outcome determination, and composite
outcomes are reported.
312 F. Wang

17.2.5 Sample Size

There are no special points in the estimation of the sample size in prognosis study.
Experience and formulas both can be applied to determine sample size. In clinical
studies, each prognostic factor requires at least 10 individuals. If there is more than
one prognostic factor, the maximum sample size required by the factor is taken.

17.2.6 Follow-Up

Follow-up is important, and all patients should be followed for an enough long time
to observe concerning events, including some rare adverse outcomes. The length of
follow-up depends on the course of the disease. For some communicable diseases,
the period is several weeks and decades years for the onset of hepatitis B. The
interval should be reasonable to obtain various dynamic changes of the disease. The
outcome of diseases with a short course changes rapidly, and the interval can be
shorter than those with a longer course.

17.3 Describing Prognosis

17.3.1 Case Fatality

Case fatality, discussed in Chap. 2, is often the first way to describe prognosis. It is
defined as the percent of people who die of a disease in people with the disease. Case
fatality is usually used to express the prognosis of short-term diseases, such as acute
infectious diseases, acute phase of cardio-cerebrovascular diseases, and cancers with
short survival. Case fatality is not suitable for chronic diseases as death may occur
many years after diagnosis and competitive events occur more likely.

17.3.2 Remission Rate

Remission rate refers to the percent of patients in clinical undetectable state after
treatment.
17 Disease Prognosis 313

17.3.3 Recurrence Rate

Recurrence rate is the percent of patients who return to disease after a period of
remission or recovery.

17.3.4 Disability Rate

Disability rate is the percent of patients who are unable to function normally as a
result of the disease.

17.3.5 Quality of Life

Describing prognosis as a single rate is relatively simple. However, much informa-

tion is hidden, and in most cases, survival can be quite different with similar rates.
Figure 17.1 shows the survival curve of two populations. The 5-year survival is
about 15% in both populations, but clinical courses are quite different. In
population A, most deaths occurred in the fifth year, while most deaths occurred
during the first 2 years in group B. Although 5-year survival is the same in the two
groups, survival in group A is clearly better than group B.

Fig. 17.1 Five-year survival curves in two hypothetical populations

314 F. Wang

17.4 Analysis for Prognosis Study Data

When we summarize prognosis as a single rate, it does not include the likelihood the
patients will experience an outcome at any point during follow-up. Survival analysis
is a simple statistical method for estimating the survival of a group over time. It is
performed to describe the survival distribution of participants in the cohort. The
curves can help to figure out information about the survival event at any point in the
course of the disease.
To learn about survival, patients in the cohort are at the same starting point in
the disease course and all followed up. Thus, completely data can be achieved when
the outcome of interest occurs during follow-up. However, when patients die of the
disease other than the outcome event, dropout at any point of time or loss to follow-
up, only incomplete data can be obtained which is called censored data. Survival
analysis can make efficient use of all data and can be applied to any outcomes from
all subjects in the cohort.
Survival time is defined as the period between the starting event and the terminal
event of the disease. In survival analysis, the most basic thing is to calculate survival
time. Time intervals can be made of interest. In general, life table and Kaplan–Meier
analysis are adopted to analyze survival data. Survival curves can be compared by
log-rank test. COX proportional hazard model is used to estimate the hazard risk of
multiple prognosis factors.

17.4.1 Calculating Survival Rate

17.4.1.1 Five-Year Survival

Five-year survival is the percent of patients who are alive 5 years from a certain point
(usually diagnosis or treatment) in the clinical course of the disease. Actually, 5-year
survival is a proportion instead of a rate. It is frequently used to measure the
prognosis of cancer after diagnosis because most deaths occur during this period.

17.4.1.2 Life Tables

Life table is the most commonly adopted approach to measure actual observed
survival over time. It is a little more sophisticated method that tries to predict the
prognosis of patients. The probability of surviving for one interval following the start
of the observation is calculated. Cumulative survival is defined as the proportion
who survived from enrollment to the end of the interval. Cumulative survival for the
entire follow-up period is the product of each surviving probability of each period.
Hazard is usually estimated for a year at a time to estimate survival. Person-years is
calculated sometimes and adopted as the denominator. All the data obtained are used
17 Disease Prognosis 315

including patients who entered the cohort after the beginning of the observation.
Compared to 5-year survival, life table describes survival experience of patients in a
more efficient and economical way.

17.4.1.3 Kaplan–Meier Analysis

In Kaplan–Meier method, the exact point in time when each outcome occurs is
identified. Survival time, including censored data, is arranged from small to large.
Survival probability is the ratio of the number of survivals at each point to the
number of followed-up till the end of observation (including the alive and death of
the disease at that point of time) except for the dropouts. The overall survival is the
product of survival probability at each point. The survival curve is the total survival
experience during follow-up time presented in a graph. Figure 17.2 shows two
simplified survival curves. The horizontal axis is the follow-up period from the
beginning of the observation, and the vertical axis is the estimated survival proba-
bility. At the beginning of the observation, no one dies and the survival probability is
100%. As time goes on, more and more deaths occur and the probability of surviving
decreases. Here, censored data is not used to calculate survival at each point of the
time since it is not related to prognosis. If the sample size is small, the survival curve
is shown in a stepwise fashion because survival is constant and changes only when
the next event happens (Fig. 17.2a). The steps would diminish with the increasing
number of patients and for a large cohort, the curve would become a smoothed slope
(Fig. 17.2b). Although the follow-up intervals between each new death can be
yearly, monthly, or weekly depending on interest or need, the estimated survival is
more accurate when the intervals are shorter.
By plotting survival time rather than calendar time on the horizontal axis,
Kaplan–Meier analysis deals with censored data and makes considerable use of
information on follow-up patients, dropout patients, and deaths. If a patient is
censored at 21 months, it is assumed that he/she survived until the next death
occurred. The survival curve contains more information besides the observed rates
and can be used for any type of time-to-event data. It helps to predict the prognosis of
patients with the information of all available data. Although it is used to be applied to

Fig. 17.2 Survival curves of two cohorts with 20 and 200 patients, respectively
316 F. Wang

studies of small number, more and more large data on survival are now accom-
plished by Kaplan–Meier analysis with appropriate statistic software.

17.4.2 Several Points About Interpreting Survival Curves

When interpreting survival curves, several points need to be noticed. The survival
rate on the survival curve comes from patients of a hypothetical cohort. It is an
estimated probability of the hypothetical population instead of a real population. The
survival probability is the best estimation of survival for patients in the cohort.
However, the precision of estimated survival depends on the number of cases under
observation. On the left side of the curve, in the earlier period of follow-up, there are
more individuals at risk. However, on the right side, at the tail of the curve, fewer and
fewer cases are at risk because of deaths, dropouts, and late entries. As a result, the
estimation would be more reliable on the left side of the curve than the right-
hand side.

17.4.3 Comparison of the Survival Curves

We just discussed the estimation of the survival curve for a single group. If there are
two groups, for example, in the clinical trial, survival is estimated separately. Now
comes the question, how could we learn the differences between the two survival
curves?
In survival analysis, instead of mean survival time, we express average survival
time with median survival time since survival time is usually skewed distribution.
Median survival time is the length of time that 50% of the study population has had
the event. It is the survival experience of one-half individuals and can be easily
estimated from the Kaplan–Meier curve. The confidence intervals may also be
calculated. In this case, we can compare median survival times for the two groups
by easily estimating the ratio of the two medians.
For the comparison of the overall survival experience, Mantel–Haenszel or
log-rank chi-square test is applied. The null assumption is that the two survival
curves are equal. Essentially, it is used to determine the difference between the
observed number of events and the theoretical number at the time of each event.
When the null hypothesis turns out to be false, Mantel–Haenszel statistic weights the
survival experience on the right side of the curve more than the left side. The
log-rank statistic is more powerful when the hazard rates are proportional. When
the hazard rates are not proportional, the difference between the two curves will not
disappear until the curves cross at a point. The conclusion is still valid but may not be
as powerful as when the hazard rates are proportional. If there is a cross in the two
curves, the conclusion should be interpreted cautiously.
17 Disease Prognosis 317

Life table and Kaplan–Meier analysis can not only be applied to calculate
survival, but also useful for any dichotomous or once-occurring outcomes. The
end points other than death can be the development of diabetes, the recurrence of
acute myocardial infarction, or the cure of infection. When we describe such events
other than survival, time-to-event analysis is adopted.

17.4.4 Dealing with Multiple Prognostic Factors

COX proportional hazards model is an appropriate model that can deal with multiple
variables at one time. The dependent variable is the time from the beginning of
observation until the concerning outcome. The independent variables are factors that
might be associated with the outcome. Hazard risk is the ratio of probability derived
from the time-to-event analysis and is a reasonable approximation of relative risk.
COX model can be fitted quickly with suitable statistical software.

17.5 Common Bias and Controlling

17.5.1 Bias in Prognosis Study

17.5.1.1 Assembly Bias

Assembly bias is one form of selection bias that is also called classification bias.
When patients are assembled in the cohort, it is the best that the distribution of the
non-studied variables in the groups is similar. However, some important factors,
such as the extent of disease progression, the existence of complications, the course
of disease, and prior treatment, often differ in the groups. Moreover, those factors
may determine the outcome. Therefore, the susceptibility to the outcome would not
be equal among the groups being compared.

17.5.1.2 Migration

Migration is also one form of selection bias. When study subjects leave the original
group, drop out of the research altogether or move to another group, migration bias
occurs. If the migration in the groups is random, there would be no bias. However,
the number and the characteristics of patients dropping out or moving to another
group are usually not the same in different groups. As a result, when these migrations
are large enough, the groups are no longer comparable, and the validity of the
conclusion will be affected.
318 F. Wang

17.5.1.3 Loss to Follow-Up

Patients in the cohort may drop out at any point due to the long follow-up period of
observation; moving away from the place that they are now living, refusing to
continue the research, becoming ill, encountering a competitive event, or are suffer-
ing from side effects of the treatment. The outcomes of those dropping out of the
study cannot be obtained. When the number of losses to follow-ups takes place on a
large scale, the validity of the study would be affected. It is generally considered that
the proportion of loss to follow-up should be kept under 10%.

17.5.1.4 Survival Cohort Bias

In a cohort study, if patients in the groups are assembled from a hospital because they
are receiving treatment there and are just available for the recruitment, the cohort is
called a survival cohort or available patient cohort. Survival cohorts are not really
cohorts and should be distinguished from true cohorts. Patients in a true inception
cohort are observed from the onset of the disease, while patients in survival cohorts
are recorded at various points in the course of the disease. Also, survival cohorts do
not include patients not in the hospital. Therefore, reports of survival cohorts are lack
of precision and are misleading.

17.5.1.5 Zero Bias

In a cohort study, all patients should be followed-up at the same starting point of the
disease course. This starting point, also called zero spot, can be the time of diagnosis
or the beginning of treatment. Zero spot should be well defined and complied
throughout the research. If patients in the cohort were assembled at different points
of the disease course, zero bias takes place and could have an effect on study
validity.

17.5.1.6 Measurement Bias

Some clear and objective outcomes, such as death, major cancers, and cardiovascu-
lar disorders are rarely misdiagnosed. However, other outcomes without a clear
definition, such as side effects, subclinical diseases, and special course of disease or
disability can be missed due to misclassification or different diagnosis criteria.
Measurement bias occurs when researchers try to record those less-clear-cut out-
comes, especially when patients in one group have more opportunity to detect their
outcomes than patients in another group. The “differences” found among the groups
are due to the inconsistencies in the recording of outcomes rather than possible
prognostic factors.
17 Disease Prognosis 319

17.5.2 Control of Bias

17.5.2.1 Randomization

Randomization is the best way to control confounding. Each subject has an equal
chance to be randomly assigned to the experimental group or control group. Not only
known factors but those unknown variables that might affect prognosis are balanced
in both groups, which means that baseline information in different groups is com-
parable. As a result, the conclusion will be more precise when differences in
prognosis found between groups are caused by certain prognosis factors.
However, the application of randomization is limited. In cohort or case-control
studies, it is not possible to adopt this process. It is only possible to use randomiza-
tion when the study purpose is to evaluate the effects of treatment measures on
prognosis.

17.5.2.2 Matching

Patients in the study group can be matched with one or more patients in the
comparison group with the same characteristics except for the prognostic factor of
interest. The non-studying characters of patients in different groups become similar
through matching. Factors such as age and gender are chosen for matching for their
relatively strong relationship with most diseases. However, other variables such as
race, clinical stage, the severity of the disease, and treatment history may also be
considered for matching. Paired matching and frequency matching are two matching
choices. For paired matching, one may select no more than four controls at a time.
One case and one or two controls are often adopted. In frequency matching, the
distribution of the matching factors should be kept consistent between the two
groups.
Although matching makes patients similar in different groups as randomization
does, it only controls factors that are known to affect the outcome and cannot control
the effect of those unknown factors. Another thing is that only a few variables can be
matched at a time. Controls will be difficult to find with so many criteria to be met,
and overmatching may occur. Finally, the effect of the matched factors on the
outcome cannot be evaluated.

17.5.2.3 Restriction

Patients are limited to a narrow range of characteristics to ensure characteristics of

the included subjects are consistent among groups under comparison. For example,
when evaluating the effect of a new drug on blood pressure, male patients aged
20–60 with a blood pressure of 140–180 mmHg were enrolled. Such restriction
certainly increases the homogeneity of patients, but with the sacrifice of
320 F. Wang

generalizability. The representation of patients in the study may no longer be

possible. Moreover, another method is needed to learn the effect of the drug on
females, ages, and blood pressure out of that range.

17.5.2.4 Stratification

Data can be stratified into several subgroups, and results of patients with similar
characters are presented. Factors of stratification are those possible confounding
factors similar to matching variables such as age, gender, or smoking. The effect of
each category on prognosis is then analyzed to discover whether there is
confounding in the crude effect. Stratification is a common way to recognize and
control confounding. It should be noticed that stratifying many factors at a time may
result in having too few patients in some subgroups.

17.5.2.5 Standardization

Patients from different places or times may have different age or gender distribu-
tions. The comparison of two rates would be affected when there are factors that
have a strong association with the outcome. Here, the standardized rate can be
applied to make the comparison without bias by giving equal weight to the factors.
Standardized mortality ratio (SMR) is usually used in prognosis studies. A detailed
definition of SMR is discussed in previous chapter.

17.5.2.6 Multivariable Analysis

In most cases, the prognosis of disease is influenced by many factors. Those factors
may be related to one another besides the outcome. Moreover, modifications of
effect among factors may also exist. Multivariable analysis is the only way that
provides us a comprehensive way to consider the relationship between variables and
the outcome, as well as the joint effect of the variables. It deals with many variables
simultaneously and can pick up variables that affect prognosis independently from
all those in the model. The contribution of each factor to the outcome can also be
figured out. In prognosis study, COX proportional hazards model is usually applied
in case of time-to-event analysis. Logistic regression is used more often in case-
control design in which outcome is dichotomous.
The limitation is that the process of multivariable analysis is just like a “black
box.” Multiple comparisons, statistical power, and correlation between variables
such as multicollinearity may have an effect on the result. The conclusion may be
misleading and, even worse, it is much more difficult to learn where it happens.
Therefore, multivariable analysis is usually used after matching or stratification.
17 Disease Prognosis 321

17.5.2.7 Other Methods to Control Bias in Prognosis

In a prognosis study, it is effective to enroll patients with better compliance to reduce

the possibility of loss to follow-up. A clear definition of outcome and use of double-
blinding when deciding the occurrence of an outcome will be helpful to minimize
measurement bias.
Chapter 18
Nosocomial Infections

Zhijiang Zhang

Key Points
• Nosocomial infections, also termed as “healthcare associated infections”, are
infections acquired in health-care facilities. For patients, the infections should
not be present or incubating at admission.
• Nosocomial infections can be categorized as endogenous infections and exoge-
nous infections according to types of reservoirs. There are two types of exoge-
nous infections: iatrogenic infections and cross infections.
• The epidemic process of nosocomial infections is usually described with 3 terms:
(1) Source of infection; (2) Route of transmission; and (3) Susceptible population.
• Nosocomial infections are related to both patient factors, such as
immunocompromise, and medical procedures that is implemented by health
professionals at hospitals. It is the responsibility of all health professionals to
reduce nosocomial infections.

18.1 Introduction

Nosocomial infections constitute a serious threat to the global health. Acquiring

infections in health-care settings adds to the patient’s functional disability and
emotional stress and in turn affect the patient’s quality of life. Furthermore, noso-
comial infections can lead to excessive mortality among hospitalized patients. More
than that, nosocomial infections may be disseminated from health-care settings to the
general public if not appropriately controlled, and threaten the health of total
population.
Nosocomial infections may lead to considerable economic costs. The increased
consumption of drugs, the use of additional laboratory and other diagnostic

Z. Zhang (✉)
School of Public Health, Wuhan University, Wuhan, China

© Zhengzhou University Press 2023 323

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_18
324 Z. Zhang

procedures, and the need for isolation contribute significantly to the direct costs. The
increased length of hospital stay also increases the indirect costs due to work hours
lost. Nosocomial infections constitute an economic burden both for the individual
patients and for the public health.

18.2 Definition and Diagnostic Standards

Nosocomial infections, also termed as “health-care-associated infections,” are infec-

tions acquired in health-care facilities, usually during hospital care after admission.
To be considered to be nosocomial, the infections cannot be present or incubating at
admission. Those infections acquired in the hospital but appearing after discharge,
and the occupational infections among staff of the hospital are also considered to be
nosocomial.
According to the “Diagnostic Standard for Nosocomial Infections’ issued by the
National Health Planning Commission (formerly Ministry of Health of China), the
following infections are considered to be nosocomial:
1. For those infections without a clearly defined incubation period, occurrence of
infection beyond 48 h since admission; for those infections with a clearly defined
incubation period, occurrence of infection beyond the average length of incuba-
tion period since admission;
2. The infection is directly related to the last hospital stay;
3. Appearance of new infection in sites other than the original ones (except for
metastatic lesions caused by pyemia) or isolation of new pathogens in addition to
the known pathogens during hospital stay (excluding the possibility of pollution
or previous coexistence of infections);
4. Neonatal infections acquired during or after delivery;
5. Latent infections reactivated by diagnostic or therapeutic procedures, for exam-
ple, herpes virus and Mycobacterium tuberculosis;
6. Infections acquired while working at the hospital as medical staff.
The following cases are not considered to be nosocomial:
1. Bacterial colonization without inflammation in open wounds of skin and mucous
membrane;
2. Inflammations caused by trauma or non-biological factors;
3. Infection of the newborn through the placenta (onset of the disease within 48 h
after birth), for example, herpes simplex virus, toxoplasmosis, or chicken pox;
4. Acute attack of the original chronic infections during hospital stays.
18 Nosocomial Infections 325

18.3 Nosocomial Infection Sites

There are many potential body sites for the occurrence of nosocomial infections. The
following are the most frequent sites for nosocomial infections.

18.3.1 Surgical Sites

Surgical sites are subject to nosocomial infections. The infections are usually
acquired during the surgical operation. The diagnosis is mainly based on clinical
criteria: purulent discharge or spreading cellulitis around the wound or the insertion
site of the drain. There are varieties of possible infecting microorganisms for surgical
sites nosocomial infections, depending on location and aggressiveness of the sur-
gery, patients’ immunity status, and antibiotics use. The level of contamination
during the surgical procedure is one of the most important risk factors for surgical
site infections. Other risk factors for surgical site infections are the surgical pro-
cedures per se, the level of asepsis, as well as the virulence of the infecting
microorganisms and concomitant infections at other body sites.

18.3.2 Respiratory System

Patients with several diseases are at high risk of nosocomial pneumonia while
hospitalized. The most frequently reported nosocomial pneumonia is among patients
on ventilators. Monitoring of clinical manifestation and using radiological imaging
support the diagnosis of pneumonia. Specimen investigation can improve specificity
for the diagnosis. Infecting microorganisms may be either endogenous, for example,
from nose, throat, or stomach, or exogenous, for example, from contaminated
equipment. Patients with decreased level of consciousness are also at higher risk
for nosocomial pneumonia. Children are susceptible to viral bronchiolitis, while the
elderly are vulnerable to influenza and secondary bacterial pneumonia.

18.3.3 Bacteremia

The incidence of nosocomial bacteremia is low, but its case fatality rate is high –
over 50% for some microorganisms. When infection occurs at the entry site of the
device, it may be visible. If bacteremia is caused by the microorganisms colonizing
the device within the vessel, it may be invisible. In the case of catheterization, the
length of catheter, level of asepsis for the insertion procedures, and duration of
catheter care are important factors influencing the risk of nosocomial bacteremia.
326 Z. Zhang

18.3.4 Urinary Tract

Urinary infections are the most frequently reported nosocomial infections. Com-
pared with nosocomial infections in surgical sites, pneumonia or bacteremia, urinary
infections cause less morbidity but occasionally lead to bacteremia and death.
Urinary infections can be diagnosed through quantitative urine culture (>105/mL).
The microorganisms responsible may be acquired from the patient’s gut flora or from
the health-care facilities.

18.4 Microorganisms

The infecting microorganisms in nosocomial infections may be bacteria, virus,

parasites, or fungi, dependent on the patient populations, medical and surgical
interventions, implemented nosocomial infection control programs, and health-
care settings.

18.4.1 Normal Microorganisms in Nosocomial Infections

18.4.1.1 Bacteria

Bacteria are among the most frequently reported pathogens in nosocomial infections.
These can be commensal bacteria. Infection occurs when immunity of the host is
compromised. These can also be pathogenic bacteria, which lead to nosocomial
infections when introduced regardless of the immunity status of the host. Staphylo-
cocci, pseudomonads, and Escherichia coli are the three pathogens of great concern
for nosocomial infections.

18.4.1.2 Viruses

Many viruses can cause nosocomial infections. For example, the hepatitis B virus
can be transmitted through invasive medical procedures, for example, transfusions,
dialysis, and injections. Enteroviruses may be transmitted by the fecal-oral route.
SARS-CoV-2 may be transmitted by respiratory droplet and aerosol.

18.4.1.3 Parasites and Fungi

Some fungi and parasites may cause infections among hospitalized patients with
compromised immunity or undergoing extended antibiotic treatment. Risks of fungi
infection increase for hospitalized patients when renovating aging hospitals. The
18 Nosocomial Infections 327

infecting pathogens can be Aspergillus spp., Candida albicans, or Cryptococcus

neoformans.

18.4.2 Antimicrobial Resistance and Nosocomial Infections

Due to the inappropriate and uncontrolled use of antimicrobial agents, varieties of

microorganisms, including bacteria, viruses, fungi, and parasites that can cause
infections in humans, animals, or plants, no longer respond to antimicrobial agents
that used to be effective. Antimicrobial resistance constitutes a global concern and is
especially a problem for nosocomial infections. In health-care settings, resistant
microorganisms have larger capability to spread. Patients undergoing surgery,
cancer chemotherapy, and transplantation are at high risk of infections with resistant
microorganisms. Genetic mutations of the antimicrobial-resistant microorganisms
are thus more likely to spread between hospitalized patients in health-care settings.
Restriction on antimicrobial consumption plays a role in the control of nosocomial
infections.

18.5 Categories of Nosocomial Infections

Nosocomial infection can be categorized as endogenous infection and exogenous

infection according to types of reservoirs.

18.5.1 Endogenous Infections

Endogenous infections occur when microorganisms that cause nosocomial infec-

tions are already present within the body. For example, a patient undergoing
chemotherapy has a compromised immunity and the dormant tuberculosis becomes
reactivated and infects the patient.

18.5.2 Exogenous Infections

Microorganisms that cause nosocomial infections are transmitted from outside the
patient. There are two types of exogenous infections.
Iatrogenic infections: The infections are caused by the contamination of medical
instruments, equipment, supplies, and sanitary materials used in health care or by the
poor sterilization, for example, microorganisms in water, damp environment, and
contaminated devices.
328 Z. Zhang

Cross infections: Microorganisms are transmitted between patients, member of

staff, or visitors through direct or indirect contact.

18.6 Epidemic Process of Nosocomial Infection

The epidemic process refers to the development and spread of nosocomial infections
within the health facilities. The three terms frequently used to describe the epidemic
process for infectious diseases, that is, source of infection, route of transmission, and
susceptible population, are used here to describe the epidemic process of nosocomial
infections. In view of the significant difference in epidemic process between endog-
enous infection and exogenous infection, the following details pertain primarily to
the latter.

18.6.1 Source of Infection

Source of infection refers to the natural habitat of microorganisms which may cause
nosocomial infection. Patients are one of the most important sources of nosocomial
infections. Other important sources of nosocomial infection may be carriers, wet
environment, mouse, arthropods, mosquito, and others. For more details, refer to
Chap. 11.

18.6.2 Route of Transmission

Route of transmission refers to the spread of infecting microorganisms directly or

through the environment to another person. The important routes of transmission for
nosocomial infections may be direct contact, transfusion or infusion of other medical
products, intramuscular injection or other invasive medical procedures, airborne
transmission, waterborne transmission, arthropod-borne transmission, vertical trans-
mission, and others. For more details, refer to Chap. 11.

18.6.3 Susceptible Population

The susceptibility varies for patients according to their age, gender, immunity, as
well as medical procedures they are undergoing. Patients with compromised immu-
nity are among the most susceptible populations. For more details, refer to Chap. 11.
18 Nosocomial Infections 329

18.7 Prevention of Nosocomial Infections

The risk of nosocomial infections is affected by both patient factors, such as

immunity status, and health-care settings and medical procedures that elevate the
likelihood of infection. It is the duty of all health professionals to prevent nosocomial
infections.

18.7.1 Preventing Human-to-Human Transmission

18.7.1.1 Hand Decontamination

Maintaining hand hygiene is important for reducing nosocomial infections. For

handwashing in hospitals, it requires running water, soap, and drying facilities. For
hand disinfection, proper antiseptic is required. The procedures of handwashing/
disinfection vary for medical procedures that patients will receive or have
undergone.
Adherence to hand decontamination is frequently suboptimal. There may be a
variety of reasons, including high frequency of patient contact, allergies to hand
decontamination products, low perceived risk of infection, low awareness of hand
decontamination procedures, lack of time required to complete the hand decontam-
ination procedures, and lack of accessible equipment. The facilities must have
policies to evaluate and manage this problem.

18.7.1.2 Clothing

An outfit, usually a white coat, is needed for staff. In special areas, uniform trousers
and gown are required. An outfit must be changed in the case of being exposed to
blood or other body fluid. In aseptic units and operating rooms, dedicated shoes
should be used as well as caps or hoods that can cover the hair.

18.7.1.3 Masks

In operating room, staff wear masks to protect patients. When caring for immune-
compromised patients, staff must wear masks. For infections which can be trans-
mitted by the air, patients must wear masks when not isolated. When approaching
patients with airborne infections, staff must wear masks to avoid being infected.

18.7.1.4 Gloves

Staff must wear sterile gloves in surgery or other invasive procedures. To protect the
immune-compromised patients, staff must wear sterile gloves. Whenever hands are
330 Z. Zhang

likely to be contaminated, non-sterile gloves should be worn before patient contacts.

When caring for patients with infections that can be transmitted by direct contact or
respiratory droplets, non-sterile gloves should be worn to protect the staff. Wash
hands with running water after removing or changing gloves.

18.7.1.5 Safe Injection and Other Skin-Piercing Practice

Injection or other skin-piercing procedures increases the risk of infection transmis-

sion between patients. It is required to use sterile needle and syringe, prevent
contamination of medications, and eliminate unnecessary injections.

18.7.2 Preventing Transmission from Environment

The hospital environment can be classified into five types of zones according to the
possibility of contamination, required level of asepsis, and risk of infection:
Zone A: It is clean areas without patient contact, for example, administrative office
and library;
Zone B: It is areas possibly contaminated by microorganisms, for example, passage-
way and lab;
Zone C: It is areas with patient contact and microbial contamination, for example
patients’ room and bathroom;
Zone D: It is passageways for clinicians and patients in the designated zone for the
diagnosis of infectious respiratory diseases. The entrance of clinicians’ passage-
way connects to clean zones, while the entrance of patients’ passageway connects
to contaminated zones.
Zone E: It is buffer areas between clean and contaminated zones.
To minimize the microorganisms from environment, cleaning, disinfecting, and
sterilizing must be used appropriately for each type of zone.

18.7.2.1 Routine Cleaning

Routine cleaning is scheduled to make the environment visibly clean. The frequency
of routine cleaning and cleaning agents need to be specified for all types of reused
equipment/devices used in the health-care settings and all areas in the hospital.

18.7.2.2 Disinfection of Equipment

The purpose of disinfection is to remove microorganisms without complete sterili-

zation. The disinfectants must be nonvolatile, free from irritating smells, and not
18 Nosocomial Infections 331

harmful to equipment or persons. The disinfection procedures must kill or remove

the targeted microorganisms.

18.7.2.3 Sterilization

Sterilization is to destruct all microorganisms on the medical devices. Sterilization is

performed for those medical devices used to penetrate sterile body surface, as well as
medications and parenteral fluids.

18.8 Surveillance of Nosocomial Infections

Surveillance is a program designed to monitor nosocomial infections in a continu-

ous, systematic, and long-term manner. It plays an important role on identifying the
early signs of local problems and the evaluation of the effectiveness of nosocomial
infection control policy.

18.8.1 Objectives of Surveillance Programs

The ultimate aim of surveillance is to reduce the burden of nosocomial infections in

the local area and alleviate the costs.
The specific objectives of a nosocomial infections surveillance program usually
include the following:
1. To monitor trends in nosocomial infections, including incidence, prevalence, and
distribution of nosocomial infections;
2. To evaluate the effectiveness of prevention programs, and to adjust the currently
ongoing prevention programs accordingly;
3. To recognize sources of nosocomial infections, particularly in situations of an
outbreak, and to take immediate actions to control transmission;
4. To find aspects for improvement in the local nosocomial infections control
programs.

18.8.2 Implementation of Surveillance Programs

Surveillance programs can be implemented at the hospital level. Involving partners

include the infection control practitioner, physician, nurse, lab staff, director, and
administrator. Before implementing a surveillance program, the partners need to
decide the following:
332 Z. Zhang

1. Which patients and units to be monitored;

2. What type of infections and relevant information to collect;
3. The time period of the surveillance and frequency of monitoring;
4. Data collection and information retrieve methods;
5. Methods for data management and analysis;
6. Methods for information dissemination and feedback collection;
7. Methods for maintaining confidentiality.
Besides at the hospital level, surveillance of nosocomial infections may also be
implemented at the levels of local, regional, national, or international networks. On a
confidential basis, hospitals may share data with other facilities in the local, national,
or international network for the purpose of improving nosocomial infection control
programs.

18.8.3 Evaluation of Surveillance Program

Evaluation of the surveillance programs is necessary. Maintaining contacts with the

surveillance program staff can also help maintain their compliance with the guidance
of the surveillance program. An evaluation usually includes the following:

18.8.3.1 Strategy Evaluation

Evaluate whether the surveillance program has the following quality: simplicity,
flexibility, acceptance, sensitiveness, effectiveness, and efficiency.
Evaluation can be undertaken by means of field survey, focus group, or interview.

18.8.3.2 Feedback Evaluation

Feedback evaluation aims to address the following specific issues:

1. Is confidentiality respected during the implementation of the program? Is
maintaining confidentiality compatible with data dissemination required for the
purpose of infection control?
2. Are the results of the evaluation widely shared internally within the units and
externally between facilities in the network?
3. Is the population under surveillance well representative of the target population?
4. Is risk adjustment/stratification appropriately used?
5. Is the length of the surveillance period sufficient to draw a conclusion?
18 Nosocomial Infections 333

18.8.3.3 Evaluation of Data Quality

The denominator and nominator used in calculating the incidence or prevalence of

nosocomial infections need to be periodically evaluated in terms of exhaustiveness
(missing patients), completeness (missing data), and correctness (data error).
Chapter 19
Epidemiology Design in Clinical Research

Yi Wang

Key Points
• The process of clinical research include forming research questions, selecting
proper epidemiology design, collecting clinical data and doing statistical analysis,
preparing reports to publish. The PICO process could help investigators to form
research question. For different types of questions, there are different appropriate
epidemiological designs could be selected.
• Some checklist items should be included in clinical research reports. The check-
lists composed reporting guidelines. The reporting guidelines for clinical research
include STROBE for observational studies, STARD for diagnostic/prognostic
studies, CONSORT for clinical trials, PRISMA for systematic reviews/Meta-
analysis, et al.
• Real-world studies are used widely in clinical research now. Real-world studies
are different from randomized control trials in many aspects.RCT provides
evidence for clinical practice guideline recommendation and real-world study
tests if guideline recommendation is practicable.
In previous chapters, we introduced different types of epidemiological study designs
and discussed their strengths and weaknesses. The overall strategy of clinical
research is the same as that utilized in other areas of epidemiology: observation of
incidences between groups and then extrapolation based on any differences. In
clinical research studies, the defining characteristics of groups can be symptoms,
signs, diseases, diagnostic procedures, or disease treatment. The discussion that
follows in this chapter will consequently summarize and integrate the core epidemi-
ological topics involved in the previous chapters. We will concentrate mainly on
observational studies, diagnostic/prognostic studies, clinical trials, and systematic
reviews looking for the general principles frequently applied in clinical research.

Y. Wang (✉)
School of Public Health and Management, Wenzhou Medical University, Wenzhou, China
e-mail: [email protected]

© Zhengzhou University Press 2023 335

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9_19
336 Y. Wang

Clinical epidemiological studies prefer randomized groups to epidemiological stud-

ies. Firstly, the “exposure” in clinical research is usually a treatment approach that
tends to be more randomized than the exposures considered in most epidemiological
studies (e.g., tobacco or alcohol consumption, diet, or personal or environmental
characteristics). Secondly, the results uncovered in clinical epidemiological studies,
such as disease progression, complications, or mortality, are comparatively fre-
quently found in the patient groups being compared, making randomized studies
more feasible. Thirdly, the potential for confounding is particularly high in clinical
epidemiological studies where there is no randomized grouping. In a large number of
nonrandomized treatment studies in which a correlation has been detected, it is
unclear whether changes in patients’ risk of disease progression, complications, or
death are related to the type of treatment they receive.

19.1 Design and Implementation of Clinical Research

Good epidemiological studies are complicated to design and conduct, and the
interpretation of their consequences and findings is not as straightforward as
researchers would like it to be. So, what can we do to make the best research design?
How can we make the most of the clinical practice information available to us?
When we read or write clinical research papers, the central question we need to
answer is “Are the findings valid?”. If a relationship between predicted values and
results is reported by researchers, is this true? If they come up empty, can we
trust them? Or could there be another interpretation of the findings, namely, chance,
bias, and/or confusion? When investigators perform the clinical study, they almost
certainly must read individual articles and reports, especially the guidelines for
clinical research published in professional journals. They may produce some of
their scientific papers when they are engaged in clinical research.
The first stage in establishing clinical research is to design the study issue that you
aim to answer. Then, you would utilize several epidemiological designs to try to
uncover the explanation. Therefore, first of all, investigators need to focus on what
clinical questions should be answered. Secondly, researchers should also consider
whether the research design was suitable for replying to the questions raised. A
highly practical approach is very necessary, which we will outline in the parts that
follow.

19.1.1 Forming Research Questions

Usually, clinical problems could be divided into two categories: background ques-
tion and foreground question. The background question is about the general knowl-
edge of disease such as “what is tuberculous pericarditis?” and “what are the
antituberculosis drugs?” The foreground question is the actual problems that
19 Epidemiology Design in Clinical Research 337

physicians or surgeons encounter in the process of diagnosis and treatment of

patients. For instance, physicians want to know “how the utility of the ascites
adenosine deaminase (ADA) in the diagnosis of the tuberculous peritonitis?” and
“does tuberculous pericarditis require glucocorticoid treatment?”. The foreground
question is the main problem in clinical practice. According to different process of
clinical practice, there are four types of foreground questions: treatment question,
diagnosis question, etiology question, and prognosis question. When physicians are
confronted with clinical foreground questions, they want to design a study to solve
these problems, they could use “PICO” process to decompose the research problems
into specific research content.
In “PICO” process, “P” is an abbreviation for patients or population. It refers to
the clinical features of research patients or population. “I” is an abbreviation for
intervention or exposure. It means treatment measures or exposure issues that are
concerned. “C” is an abbreviation for comparison. It means the control measure and
usually means the “gold standard” if it is a diagnostic study. “O” is an abbreviation
for outcome. It is the outcome indicators that the research focused on. In Table 19.1,
it listed some examples of how to use “PICO” framework to form research question
in four different question types.

19.1.2 Commonly Used Epidemiological Design in Clinical

Research

The most important point in clinical research is to identify the research question. If
clinicians have proposed a research question, there are different epidemiological
designs that could be selected to help answer these questions. Commonly used
epidemiological design in clinical research includes cross-sectional study, case-
control study, cohort study, nonrandomized controlled trials, and randomized con-
trolled clinical trials. In general, prospective study design has the most content, the
most complex methods, and the most representative of epidemiological data analy-
sis. Figure 19.1 illustrates how to decide which epidemiological design to select.
Previously, we have introduced four types of foreground questions. For different
types of questions, different epidemiological designs could be selected. For the
evaluation of treatment efficacy, the most appropriate study design is a randomized
control trial (RCT). However, it is very difficult to carry out an RCT in real clinical
practice, especially conducted it in a multicenter study. Besides RCT, a cohort study,
case-control study, case report could be selected for the treatment questions. For
prognosis question, the most appropriate study design is a cohort study. In
Table 19.2, it listed best design could select for each type of foreground questions.
338 Y. Wang

Table 19.1 Examples of application of the PICO process in clinical research

Question
type Clinical question PICO Research content
Treatment Do patients with tuberculous P: adult patients Can glucocorticoids reduce
question pericarditis need to be treated with tuberculous the risk of death in adult
with glucocorticoids? pericarditis patients with tuberculous
I: antituberculosis pericarditis?
+ glucocorticoid
C: antituberculosis
O: death
Diagnosis What is the utility of the P: patients with How about the sensitivity
question ascites adenosine deaminase celiac effusion and specificity of the ascites
(ADA) in the diagnosis of the I: ascites adeno- adenosine deaminase exami-
tuberculous peritonitis? sine deaminase nation for the diagnosis of
examination tuberculous peritonitis?
C: gold standard
diagnostic method
for tuberculous
peritonitis
O: validity of
diagnosis for
tuberculous
peritonitis
Etiology How about the risk of a veg- P: adults Are vegetarians at more risk
question etarian suffering from I: vegetarian diet to develop tuberculosis than
tuberculosis? C: common diet nonvegetarians?
O: tuberculosis
Prognosis Do patients with tuberculous P: tuberculous What is the probability of
question pericarditis could develop pericarditis patients with tuberculous
into constrictive pericarditis? patients pericarditis to develop con-
O: constrictive strictive in the future? What
pericarditis are the prognostic factors to
(there usually have predict patients with
no “I” and “C” in coarctation?
the prognosis
question)

19.1.3 Collection and Analysis of Clinical Research Data

Routine clinical epidemiological data are primarily those with health, illness, and
clinical services that are routinely collected in the population for other uses, such as
patient data routinely collected in hospitals. In addition, some are disposable,
irregularly collected data, but others are data from specific epidemiological studies
(such as prospective studies), such as the purpose of the analysis is not to answer the
original questions of the study, but to use the data to explore new non-primary
research questions. They are collectively referred to as routine clinical epidemiolog-
ical data. Routine clinical epidemiological data analysis steps: (1) Analyze the time
frame of the data and the characteristics of the variables; (2) Ask questions that can
be explored to determine the final research question; (3) Compare with the best
19 Epidemiology Design in Clinical Research 339

Fig. 19.1 The flow chart of selection of epidemiological design in clinical research

Table 19.2 Best study design to select for different clinical questions
Question type Best study design
Treatment question RCT
Adverse effect of treatment question RCT
Diagnosis question Cross-sectional study
Prognosis question Cohort study
Etiology question Cohort study, case-control study

research design, check data for “research design” defects; (4) Estimate the necessary
indicators and their confidence intervals; (5) Analyze other possible biases in the
data (selection bias, information bias, and confounding bias); (6) Integrated design
flaws, biases, and results, and draw conclusions on research issues.

19.1.3.1 Data Collection

The collection and management of clinical research data is the main content in the
design and implementation phase of clinical research. It involves management
techniques and skills and requires researchers to invest a great deal of time and
effort. The collection and management of clinical data is a process. Understanding
340 Y. Wang

Fig. 19.2 Process of clinical research data collection

the content of each link in the process and the relationship between the various links
can be a good job in clinical research design and implementation. The process of
collecting and organizing clinical data is characterized by a linear process, multi-
stage, and multi-link. Figure 19.2 shows this process.
Clinical research is the process of collecting, sorting, storage, analysis, and
evaluation of clinical data. It is a linear process and can only be carried out in a
sequential manner. The starting point of clinical research is the research object. The
researchers have to use various technical methods to obtain clinical data from the
research object, then transfer the clinical data to the case report form (CRF), and then
transfer the clinical data from the CRF to the database, and prepare for the later
statistical analysis and evaluation work. In the process of clinical data collection, the
completion of CRF filling and the establishment of the database are treated as a
two-phased landmark in the collection of clinical research data. The completion of
the CRF design marks important progress in the design of the clinical research
implementation plan. The establishment of a database is a key link between the
collation and storage of clinical data. There are sophisticated methods and tech-
niques, and the workload is large. The quantity and quality of the input data are
guaranteed, and it is organized for later data analysis. The data completed by the
CRF, the quality, and the completion of the database are the main evaluation
indicators for evaluating the implementation phase of the clinical research
organization.
Besides collecting clinical data from practice clinics or hospitals, clinicians could
collect clinical data from some open access databases, like SEER (surveillance,
epidemiology, and end results). Here, we will introduce an open access database
commonly used in clinical oncology research – TCGA. The Cancer Genome Atlas
(TCGA) program was launched in 2005 to apply the latest genomic analysis
technology. In particular, the whole genome sequencing technology, in-depth under-
standing of cancer gene changes, and promoting the discovery of new cancer
treatment programs, diagnostic methods and prevention strategies, plans to draw a
wide range of tumor types and tumor subtypes, multidimensional map of the key
genome changes. Moreover, all the data can be shared for free in scientific practice.
The TCGA plans to collect sample data for 11,000 patients and 33 cancers
(Table 19.3). In 2015, the amount of data collected and generated by the TCGA
program had reached 20PB, including 10 million mutations. Investigators could
choose interested cancers to download the gene and clinical information and analyze
them for particular purpose.
19 Epidemiology Design in Clinical Research 341

Table 19.3 TCGA plan cancer sample distribution (33 cancers, 11,000 patients)
Number Number
Cancer of Cancer of
symbol Type of cancer samples symbol Type of cancer samples
BRCA Breast invasive 1097 THYM Thymoma 124
carcinoma
KIRC Kidney renal clear 536 SKCM Skin cutaneous melanoma 470
cell carcinoma
LUAD Lung 521 ACC Adrenocortical carcinoma 80
adenocarcinoma
THCA Thyroid 507 DLBC Lymphoid neoplasm diffuse 48
carcinoma Large B-cell lymphoma
PRAD Prostate 498 LGG Brain lower-grade glioma 516
adenocarcinoma
LIHC Liver hepatocellu- 377 LAML Acute myeloid leukemia 200
lar carcinoma
LUSC Lung squamous 504 MESO Mesothelioma 87
cell carcinoma
HNSC Head and neck 528 OV Ovarian serous 586
squamous cell Cystadenocarcinoma
carcinoma
COAD Colon 461 TGCT Testicular germ cell tumors 150
adenocarcinoma
UCEC Uterine corpus 548 UCS Uterine carcinosarcoma 57
endometrial
carcinoma
KIRP Kidney renal pap- 291 UVM Uveal melanoma 80
illary cell
carcinoma
STAD Stomach 443 CESC Cervical squamous cell Carci- 307
adenocarcinoma noma and endocervical
adenocarcinoma
KICH Kidney 66 PCPG Pheochromocytoma and 179
chromophobe paraganglioma
BLCA Bladder urothelial 373 SARC Sarcoma 261
carcinoma
ESCA Esophageal 185 CHOL Cholangiocarcinoma 36
multiforme
READ Rectum 171 GBM Glioblastoma multiforme 528
adenocarcinoma
PAAD Pancreatic 185
adenocarcinoma

The TCGA research team has collected and generated various types of histolog-
ical and genetic data for these cancers, including gene expression, exon expression,
small RNA expression, copy number changes (CNV), single nucleotide polymor-
phism (SNP), loss of heterozygosity (LOH), gene mutations, DNA methylation, and
342 Y. Wang

protein expression. The clinical information includes patient’s basic geographic

information, treatment method, historical or clinical stage, survival status, and so on.
By analyzing the cancer genome information to understand the mechanism of
cancer development and discover cancer markers and drug effect on gene targets, it
can provide support for the accurate diagnosis and treatment of cancer. The TCGA
plans to collect data on a large number of cancer genomes and clinical phenotypes.
There are potential molecular markers and drug targets for cancer that need to be
tapped. Scientific data management programs provide protection for cancer genome
research. The practical exploration of cancer genomic map planned in data manage-
ment can provide a reference to the development and implementation of large-scale
scientific programs such as precision medicine and data-driven collaborative
research models.

19.1.3.2 Data Analysis

Unlike basic medical research, clinical epidemiological research is an applied

research conducted in the population to quantitatively explore the general rule of
disease, health, and clinical practice, and the results can be directly applied to clinical
practice. Clinical epidemiological studies need to be based on specific research
questions, selecting designs, controlling bias, collecting data, and then analyzing
the data to quantitatively answer research questions. Therefore, data analysis is an
important and indispensable part of clinical epidemiology research. Clinical ques-
tions generally include etiology questions, diagnosis questions, treatment questions,
and prognosis questions. The purpose of data analysis is to scientifically and
quantitatively answer these practical questions. Data analysis must have a clear
purpose for analysis. The common purpose of clinical epidemiology is shown in
Table 19.4. Clearly studied issues are the premise of data analysis. After the question
is clarified, it is necessary to put forward a specific and clear analysis purpose. Its
content generally includes the following: (1) describe the change in the number of
subjects, (2) variable classification and data sorting, (3) describe and compare
baseline data between groups, (4) estimate the frequency of outcome events, (5) esti-
mate the magnitude of the effect, (6) the confidence interval of the estimated effect,
(7) identify and control the confounding, (8) identify and measure effect modifica-
tion effects, (9) identify and measure dose-response relationships, (10) other

Table 19.4 The purpose of clinical epidemiological data analysis

Research purpose
1. Estimating relevant statistical indicators such as relative risk and sensitivity
2. Estimating the confidence interval for the statistical indicator
3. Controlling for possible confounders
4. Analysis of dose-response relationships
5. Analysis of possible effect modification factors
6. Analysis of other possible biases
19 Epidemiology Design in Clinical Research 343

analysis. Although the design principles of different studies could variate and the
clinical problems, the purpose, contents, and methods of analysis are also different,
the analysis of other research data can be regarded as one or more components of the
data analysis of prospective research.
In addition, the estimation of this indicator must simultaneously control possible
confounding factors. In a randomized control trial, investigators could thoroughly
control confounding factors through randomized assign research objects to different
groups. However, in observational studies (such as nonrandomized allocation trial
studies, cohort studies, and case-control studies), the most effective and feasible
method for controlling confounding factors is multivariate regression analysis. The
premise of controlling confounding is to recognize possible confounding factors,
and the baseline data with confounding factors were collected at the beginning of the
study. Other analytical purposes may include identifying and measuring effect
modifiers, identifying and describing dose-response relationships, and analyzing
and controlling other possible biases.

19.1.4 Preparing Papers for Publication

The following recommendations provide a guide for investigators to prepare clinical

research reports:

19.1.4.1 Choose Target Journal(s)

Selecting the intended journal category for publication is always the first step for the
researchers while drafting the report. When selecting target journals, investigators
should take the following issues into account.

How High to Aim

A question that researchers will face is what height your paper may reach. Many
investigators believe that five top general medical journals are particularly attractive
carriers for their article: The Lancet, The New England Journal of Medicine (NEJM),
The Journal of the American Medical Association (JAMA), Annals of Internal
Medicine, and British Medicine Journal (BMJ). Investigators often have the question
of whether their research is suitable for these or other famous journals. It is also
challenging to predict success (or lack of success) for experienced researchers,
which makes it very difficult to select the most appropriate target journals. In
general, if it is adequate for you to seriously consider contributing to a famous
journal, it means that your research has been well designed and implemented, and
beyond that, you are so courageous. More commonly, the internal debate (within you
or your survey team) may be whether you first submitted to a secondary journal (e.g.,
344 Y. Wang

possibly a top journal in your subspecialty field) or are more likely to accept a journal
with a lower reputation for your manuscript. One advantage of foresight is that sharp
comments may help you improve your article. It is unusual to make substantial
improvements to your manuscript based on the opinions of reviewers, but it can
happen in some cases. Therefore, if multiple submissions do not make you tired, and
receiving too many rejection letters from magazines does not hurt your self-esteem,
set a higher goal. If your mental state is irritable and fragile, then choosing a journal
with a less high impact factor is more likely to accept your manuscript at the first
submission.

Selecting a Journal with a Fondness for Researcher Topic

Some certain topics or fields are often favored by certain journals. If research in an
area that is closely related to your study has previously been published by a journal,
it is eligible to be one of your chosen targets. In the meantime, the lack of articles in
your field or using your methodology provides the information you should search for
elsewhere.

Tailoring Content to the Target Journal

The majority of the manuscripts you write will be reporting on the clinical investi-
gations you conduct. However, in some cases, investigators may write a paper that
focuses more on research methods. These papers explored some issues, such as the
best research design, measurement methods, or results interpretation. Researchers
can consider publishing their manuscripts in these three types of journals: general
medical journals, subspecialty journals, and methodologically oriented journals.
Many articles on clinical research are likely to be published in multiple target
journals.

Tailoring Format to the Target Journal

Almost every journal has its own format requirements. Most of these requirements
are relatively trivial (e.g., section titles or reference citation styles), and when you are
ready to submit your manuscript, you must modify it as required. Of course, there are
other more important issues that researchers should address them early on.

19.1.4.2 Choose a Clear Message

The work may be exceedingly complicated, and a definite result might not be
obvious. Until a clear message is determined, investigators must continue to evaluate
the essence of the results. Just considering what the reader is going to take away from
19 Epidemiology Design in Clinical Research 345

a single point, what is that point? After determining the information, the investigator
needs to craft the introduction so that readers will be convinced of the significance of
the research. Your research should be presented to readers as a narrative. The
reader’s curiosity should be piqued by the introduction, satisfied by the outcome,
and reinforced by the discussion, which should highlight how significant the
finding is.

19.1.4.3 Achieve High Quality in Writing

Here are some tips for creating a high-quality manuscript. (1) Use the active voice:
Passive writing is a well-established medical practice. Although the passive voice
makes writing more awkward and difficult to understand, adds extra words, and
makes the work lose some power, this tradition still exists. The use of active voice is
advised in all current publications on writing quality from a variety of nonmedical
professions as well as writing suggestions offered by the top medical magazines.
(2) Delete unnecessary words: Unnecessary words are utilized by medical writers.
Eliminating these terms makes the writing more direct and clearer to read. Journal
articles must adhere to strict space restrictions as well. Use as few adverbs and
adjectival phrases as you possibly can. (3) Avoid using the verb “to be”: The verb “to
be” and the passive voice frequently has the same impact. It robs the writing of vigor
and energy. (4) Keep paragraphs short: Each paragraph in the article should not
exceed five sentences. Clarity will be considered a priority.

19.1.5 Common Problems in Clinical Research Design

According to the analysis of clinical papers published, there are six major problems
in the design of clinical research programs.
1. Researchers are unclear about the design scheme adopted by their institute
After investigating the research questions that are of interest to the researcher,
the researchers must determine the research design plan based on the results they
expect, the strength of the causal connection, and the feasibility.
2. Unclear definitions of primary and secondary study end points
A study generally has only one primary end point, but there can be several
secondary end points. In some studies, it is not correct to write only what the
study end point is, but not to distinguish between the primary and secondary end
points. In the design plan, the primary and secondary end points of the study need
to be clearly written out, which is conducive to the establishment of hypotheses
and the calculation of sample size. It is not possible to write all the indicators in
parallel and regardless of primary and secondary levels.
3. Have no scientific hypothesis
346 Y. Wang

The entire research process is the process of testing the hypothesis. The
hypothesis is based on scientific research problems. Based on the hypothesis,
the researcher can determine the sample size, follow-up time, and determine the
type of quantitative collection, and statistical methods. Researchers need to
establish reasonable assumptions based on the primary end point after the design
of the study.
4. Have no controls or unreasonable controls
The four principles of clinical trial design are “random,” “control,” “blinding
method,” and “repetition.” The establishment and selection of appropriate con-
trols for the control group is an important part of the research design. Researchers
can design blanks, placebo controls, positive standard controls, and other controls
based on the purpose of the study. Parallel control is best for the same period.
5. Nominally a randomized control study but not an actual randomized grouping
The stochastic method includes two layers of meanings: One is the generation
of random distribution sequences and the other is the concealment of random
distribution sequence schemes. If the scheme is not hidden, randomization may
be disrupted, resulting in selection bias and measurement bias. The purpose of
blinding is to make the research executor not know the specific stochastic method
and do not know whether the research object in accordance with the random
sequence belongs to the experimental group or belongs to the control group so
that complete randomization can be achieved. Researchers should be trained in
systematic clinical epidemiology or clinical research methodologies. The signif-
icance of clinical research method training is similar to that of standardized
training for clinicians. It is an important foundational work and requires the
support and efforts of all parties.
6. No sample size was calculated
In addition to exploratory research, because no basic data cannot calculate the
sample size, a general clinical study needs to estimate the sample size in advance.
A too small sample size will result in large sampling errors, resulting in poor
representativeness and poor reproducibility. Researchers should pay attention to
the significance of sample size and know the concept of power. As long as there is
a consciousness of calculating sample size, the calculation process is not a
problem. Now there are many statistical software applications for calculating
sample size.

19.2 Reporting Guidelines for Clinical Research Reports

19.2.1 Observational Studies Reporting Guidelines

In September 2004, the STROBE (Strengthening the Reporting of Observational

Studies in Epidemiology) working group was founded and convened in the United
Kingdom to draft the normative meeting of the observational research report. After
many revisions, a list containing 22 items (STROBE statement) was released in
19 Epidemiology Design in Clinical Research 347

2007, which is divided into 6 major aspects including the title, abstract, introduction,
method, result, and discussion. Eighteen of these items are applicable to all three
major observational research designs (cross-sectional, case-control, and cohort
design), and the remaining four are specially used for cohort, case-control, or
cross-sectional design, respectively. A new STROBE statement extension was
released in 2014 by the Lancet Infectious Diseases. Enhance Molecular Epidemiol-
ogy Reporting for Infectious Diseases (STROME-ID). The goal is to provide
guidelines for effective scientific reporting of molecular epidemiology research to
urge authors to take particular hazards to reliable inference into account. The official
website (http://www.strobe-statement.org) offers free downloads of the STROBE
statement and STROME-ID statement.

19.2.2 Diagnostic/Prognostic Studies Reporting Guidelines

In 2003, Bossuyt PM, an authoritative expert in the field of diagnostic tests,

convened a group of experts to establish the STARD group to develop a report on
the diagnostic accuracy study – Standards for Reporting of Diagnostic Accuracy
(STARD), which was used to standardize diagnostic test studies. In order to solve
new problems in diagnostic tests, streamline the reporting process, increase its
applicability, and align STARD with CONSORT-2010, Bossuyt PM again con-
vened a group of experts in 2015, including epidemiologists, statisticians, evidence-
based medicine experts, doctors, editors, and journalists, and 85 people, based on the
STARD 2003, developed a STARD 2015 guide using document research, drafting
entries, expert surveys, and group discussions. The STARD statement could be
downloaded from http://www.stard-statement.org.

19.2.3 Clinical Trials Reporting Guidelines

The CONSORT (Consolidated Standards of Reporting Trials) declaration was

created by a team of scientists and editors to enhance the caliber of RCT reporting.
It was revised in 2001 after being initially published in 1996. The statement includes
a flow diagram and checklist that researchers can employ to report an RCT. The
CONSORT declaration has received support from several top medical publications
and influential international editorial organizations. The claim makes it easier to
evaluate and understand RCTs critically. The ideas underpinning the CONSORT
statement were clarified and expanded upon during the 2001 CONSORT revision to
assist researchers and others in writing or evaluating trial reports. In 2001, the
CONSORT declaration and an essay explaining and expanding upon it were both
published. The CONSORT statement was further amended following an expert
meeting in January 2007 and is now available as the CONSORT 2010 Statement.
This revision clarifies and updates the prior checklist’s language and includes
348 Y. Wang

suggestions for subjects like selective outcome reporting bias which have just
recently gained attention. This explanation and elaboration paper, which has also
undergone substantial revision, aims to improve the use, comprehension, and diffu-
sion of the CONSORT declaration. Each newly added and revised checklist item is
explained along with its purpose, with illustrations of effective reporting and, if
available, references to pertinent empirical research. There are several flow diagram
examples provided. Resources to aid with randomized trial reporting include the
CONSORT 2010 Statement, the updated explanatory and elaboration paper, and the
related website (CONSORT, http://www.consort-statement.org).

19.2.4 Systematic Reviews Reporting Guidelines

In 1996, the Quality of Reporting of Meta-Analyses (QUROM) guide was

published, which focused on the quality of reporting of randomized controlled trial
meta-analysis, which was the earliest reporting specification for systematic review/
meta-analysis quality. In the classic monograph “Systematic reviews in health care:
meta-analysis in context,” QUROM was recommended as the “gold standard” for
evaluating the quality of systematic reviews/meta-analysis reports. The items
involved in the QUROM report specification are divided into 6 parts and 18 items,
including the title, abstract, introduction, method, result, and discussion. The results
section includes the search process and gives the reasons for identifying, including,
and excluding randomized controlled trials and exclusions. In 2009, QUROM
updated Systematic Reviews and Meta-Analyses for Protocols (PRISMA) in order
to improve the quality of systematic review, and meta-analysis article reports.
PRISMA is more comprehensive and complete than the QUROM developed in the
past. It has a wide range of applications, not only for meta-analysis, but also for
systemic evaluation; not only for systematic evaluation of randomized controlled
trials but also as a basic specification for evaluation reports of other types of research
systems. The PRISMA Reporting Guide consists of a list of 27 items, a four-phase
flow chart, and detailed explanations and explanations of relevant items. All these
materials could be downloaded from http://www.prisma-statement.org.

19.3 Real-World Study

The collection and storage of enormous volumes of health-related data via com-
puters, mobile devices, wearables, and other biosensors have been expanding
quickly. This information has the potential to help us design and carry out clinical
research in the health-care sector more effectively to provide answers to previously
19 Epidemiology Design in Clinical Research 349

unanswerable problems. Additionally, we are better equipped to examine these data

and apply the findings to the development and approval of medical products as a
consequence of the development of sophisticated, new analytical skills. As a result, a
growing number of clinical trial designs are being designed which have been derived
from real-world data and evidence.

19.3.1 Definition of Real-World Study

Real-world studies (RWS) originate from effective clinical trials and refer to the
nonrandom choice of interventions based on the patient’s actual condition and
willingness to perform long-term evaluations based on the larger sample size
(covering a representatively larger number of subjects). Focus on meaningful out-
come indicators to further evaluate the external effectiveness and safety of interven-
tions. The RWS covers a wide range of areas and can be used for diagnosis,
prognosis, etiology, in addition to curative studies. The RWS focuses on the
effectiveness of research, namely, the size of the evaluation interventions in the
real clinical environment. RWS can also be used to evaluate the cost-effectiveness of
different health interventions.

19.3.2 The Difference Between RWS and RCT

Although RWS is very different from RCT (Table 19.5), RCT and RWS are not
contradictory or alternative relations of opposition but are complementary and
forming a connecting link between the preceding and the following. RCT is the
highest level of evidence-based medicine; is the “gold standard” of clinical trial
design. It is a recommendation to formulate corresponding treatments guidelines
based on RCT, which tells doctors that they can do and should do, rather than have to
do it. Therefore, the guideline cannot replace clinical practice. It needs RWS as an
effective supplement, and RWS can be used to determine the true benefits, risks, and
therapeutic value in clinical practice, so that clinical research conclusions will return
to the real world after RCT. Therefore, RWS and RCT are not antagonistic, but
complementary to each other.
Overall, RCT provides evidence for clinical practice guideline recommendation.
RWS tests if guideline recommendation is practicable, whether answers clinical
questions and summarizes treatment recommendations, then returns to clinical
practice. Among them, RWS plays a more and more important role nowadays.
350 Y. Wang

Table 19.5 Differences between RCT and RWS

RCT RWS
Research The outcome of an ideal situation The outcome of the real situation
purposes
Research Strictly controlled conditions Actual clinical conditions
environment
Research design Randomized controlled trials Nonrandomized Control/Effective Ran-
domized Control/Observational Study
Research Cannot be changed after the pro- Can be adjusted according to clinical
scheme gram is fixed practice
Research object Strict inclusion/exclusion criteria Inclusion/exclusion criteria are loose,
and good homogeneity and diversity is good
Sample size Minimum sample size As much as possible
Control group Standard treatment/placebo Effective treatment/no placebo
Research data Designed before the start of the Forward-looking/retrospectively
trial, prospectively collecting data collecting data according to need
Study outcomes Most recent indicators Mostly long-term indicators
Follow-up time Short Long
Follow-up Better Uncertain
completion
Ethical review Need Need
Clinical Need Need
registration
Internal effec- Good Poor
tiveness and
safety
External effec- Poor Good
tiveness and
safety
Difficulty of Relatively small difficulty Very difficult
work
Evaluation angle Evaluating effectiveness from a Evaluate the effect from the patient
medical perspective (efficacy) (effectiveness)
References

1. Torre LA, Bray F, Siegel RL, Ferlay J, Lortet-Tieulent J, Jemal A (2015) Global cancer
statistics, 2012 [J]. CA Cancer J Clin 65(2):87–108
2. Ministry of Health of the People’s Republic of China (2003) China health statistics yearbook
2003 [M]. Peking Union Medical College Press, Beijing
3. National Health Commission of the People’s Republic of China (2021) China health statistics
yearbook 2021 [M]. Peking Union Medical College Press, Beijing
4. American Cancer Society (2017) Cancer facts and figures 2017 [M]. American Cancer Society,
Atlanta
5. Zheng X, Wu K, Song M et al (2019) Yogurt consumption and risk of conventional and serrated
precursors of colorectal cancer [J]. Gut. pii: gutjnl-2019-318374
6. Monson R (1980) Occupational epidemiology [M]. CRC Press, Boca Ration, FL
7. Zheng T, Boffetta P, Boyle P (2011) Epidemiology and biostatistics [M]. International Preven-
tion Research Institute, Lyon
8. Du H, Li L, Bennett D, Guo Y et al (2016) Fresh fruit consumption and major cardiovascular
disease in China [J]. N Engl J Med. 374(14):1332–1343
9. Klemetti A, Saxén L (1967) Prospective versus retrospective approach in the search for
environmental causes of malformations. Am J Public Health Nations Health 57(12):2071–2075
10. Rothman KJ, Greenland S (2005) Causation and causal inference in epidemiology [J]. Am J
Public Health 95(Suppl 1):S144–S150
11. Shen HB, Qi XY (eds) (2018) Epidemiology [M], 9th edn. People’ Medical Publishing House,
Beijing
12. Strom BL, Kimmel SE, Hennessy S (eds) (2013) Textbook of pharmacoepidemiology [M], 2nd
edn. Wiley-Blackwell, Hoboken, NJ

© Zhengzhou University Press 2023 351

C. Wang, F. Liu (eds.), Textbook of Clinical Epidemiology,
https://doi.org/10.1007/978-981-99-3622-9