Statistical Methods for Drug Safety, 1st Edition

Visit the link below to download the full version of this book:

https://medipdf.com/product/statistical-methods-for-drug-safety-1st-edition/

Click Download Now

 Statistical
Methods for
Drug Safety

Robert D. Gibbons
University of Chicago
Illinois, USA

Anup K. Amatya
New Mexico State University
Las Cruces, USA

© 2016 by Taylor & Francis Group, LLC

CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2016 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government works

Version Date: 20150604

International Standard Book Number-13: 978-1-4665-6185-4 (eBook - PDF)

This book contains information obtained from authentic and highly regarded sources. Reasonable
efforts have been made to publish reliable data and information, but the author and publisher cannot
assume responsibility for the validity of all materials or the consequences of their use. The authors and
publishers have attempted to trace the copyright holders of all material reproduced in this publication
and apologize to copyright holders if permission to publish in this form has not been obtained. If any
copyright material has not been acknowledged please write and let us know so we may rectify in any
future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information stor-
age or retrieval system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, please access www.copy-
right.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222
Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that pro-
vides licenses and registration for a variety of users. For organizations that have been granted a photo-
copy license by the CCC, a separate system of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are
used only for identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
 To Carol, Julie, Jason and Michael and the memory
of Donna and Sid
R.D.G.

To my family and friends

A.K.A.

© 2016 by Taylor & Francis Group, LLC

© 2016 by Taylor & Francis Group, LLC
Contents

Preface xv

Acknowledgments xix

1 Introduction 1

1.1 Randomized Clinical Trials . . . . . . . . . . . . . . . . . . . 2

1.2 Observational Studies . . . . . . . . . . . . . . . . . . . . . . 4
1.3 The Problem of Multiple Comparisons . . . . . . . . . . . . . 5
1.4 The Evolution of Available Data Streams . . . . . . . . . . . 6
1.5 The Hierarchy of Scientific Evidence . . . . . . . . . . . . . . 7
1.6 Statistical Significance . . . . . . . . . . . . . . . . . . . . . . 8
1.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

2 Basic Statistical Concepts 13

2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2 Relative Risk . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Odds Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.4 Statistical Power . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.5 Maximum Likelihood Estimation . . . . . . . . . . . . . . . . 17
2.5.1 Example with a Closed Form Solution . . . . . . . . . 19
2.5.2 Example without a Closed Form Solution . . . . . . . 20
2.5.3 Bayesian Statistics . . . . . . . . . . . . . . . . . . . . 21
2.5.4 Example . . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.6 Non-linear Regression Models . . . . . . . . . . . . . . . . . 23
2.7 Causal Inference . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.7.1 Counterfactuals . . . . . . . . . . . . . . . . . . . . . . 25
2.7.2 Average Treatment Effect . . . . . . . . . . . . . . . . 25

3 Multi-level Models 27

3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Issues Inherent in Longitudinal Data . . . . . . . . . . . . . 29
3.2.1 Heterogeneity . . . . . . . . . . . . . . . . . . . . . . . 29
3.2.2 Missing Data . . . . . . . . . . . . . . . . . . . . . . . 29
3.2.3 Irregularly Spaced Measurement Occasions . . . . . . 30

ix

© 2016 by Taylor & Francis Group, LLC

x Contents

3.3 Historical Background . . . . . . . . . . . . . . . . . . . . . . 31

3.4 Statistical Models for the Analysis of Longitudinal and/or Clus-
tered Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.4.1 Mixed-effects Regression Models . . . . . . . . . . . . 32
3.4.1.1 Random Intercept Model . . . . . . . . . . . 34
3.4.1.2 Random Intercept and Trend Model . . . . . 36
3.4.2 Matrix Formulation . . . . . . . . . . . . . . . . . . . 37
3.4.3 Generalized Estimating Equation Models . . . . . . . 39
3.4.4 Models for Categorical Outcomes . . . . . . . . . . . . 40

4 Causal Inference 43

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.2 Propensity Score Matching . . . . . . . . . . . . . . . . . . . 44
4.2.1 Illustration . . . . . . . . . . . . . . . . . . . . . . . . 46
4.2.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 47
4.3 Marginal Structural Models . . . . . . . . . . . . . . . . . . . 50
4.3.1 Illustration . . . . . . . . . . . . . . . . . . . . . . . . 52
4.3.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.4 Instrumental Variables . . . . . . . . . . . . . . . . . . . . . 55
4.4.1 Illustration . . . . . . . . . . . . . . . . . . . . . . . . 59
4.5 Differential Effects . . . . . . . . . . . . . . . . . . . . . . . . 61

5 Analysis of Spontaneous Reports 69

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
5.2 Proportional Reporting Ratio . . . . . . . . . . . . . . . . . 70
5.2.1 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 72
5.3 Bayesian Confidence Propagation Neural Network (BCPNN) 72
5.4 Empirical Bayes Screening . . . . . . . . . . . . . . . . . . . 77
5.5 Multi-item Gamma Poisson Shrinker . . . . . . . . . . . . . 80
5.6 Bayesian Lasso Logistic Regression . . . . . . . . . . . . . . 83
5.7 Random-effect Poisson Regression . . . . . . . . . . . . . . . 87
5.7.1 Rate Multiplier . . . . . . . . . . . . . . . . . . . . . . 88
5.8 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91

6 Meta-analysis 93

6.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
6.2 Fixed-effect Meta-analysis . . . . . . . . . . . . . . . . . . . 94
6.2.1 Correlation Coefficient . . . . . . . . . . . . . . . . . . 94
6.2.2 Mean Difference . . . . . . . . . . . . . . . . . . . . . 95
6.2.3 Relative Risk . . . . . . . . . . . . . . . . . . . . . . . 95
6.2.3.1 Inverse Variance Method . . . . . . . . . . . 97
6.2.3.2 Mantel-Haenszel Method . . . . . . . . . . . 97

© 2016 by Taylor & Francis Group, LLC

Contents xi

6.2.4 Odds Ratio . . . . . . . . . . . . . . . . . . . . . . . . 98

6.2.4.1 Inverse Variance Method . . . . . . . . . . . 98
6.2.4.2 Mantel-Haenszel Method . . . . . . . . . . . 99
6.2.4.3 Peto Method . . . . . . . . . . . . . . . . . . 100
6.3 Random-effect Meta-analysis . . . . . . . . . . . . . . . . . . 100
6.3.1 Sidik-Jonkman Estimator of Heterogeneity . . . . . . 103
6.3.2 DerSimonian-Kacker Estimator of Heterogeneity . . . 103
6.3.3 REML Estimator of Heterogeneity . . . . . . . . . . . 104
6.3.4 Improved PM Estimator of Heterogeneity . . . . . . . 106
6.3.5 Example . . . . . . . . . . . . . . . . . . . . . . . . . . 106
6.3.6 Issues with the Weighted Average in Meta-analysis . . 108
6.4 Maximum Marginal Likelihood/Empirical Bayes Method . . 109
6.4.1 Example: Percutaneous Coronary Intervention Based
Strategy versus Medical Treatment Strategy . . . . . . 110
6.5 Bayesian Meta-analysis . . . . . . . . . . . . . . . . . . . . . 112
6.5.1 WinBugs Example . . . . . . . . . . . . . . . . . . . . 114
6.6 Confidence Distribution Framework for Meta-analysis . . . . 119
6.6.1 The Framework . . . . . . . . . . . . . . . . . . . . . . 120
6.6.1.1 Fixed-effects Model . . . . . . . . . . . . . . 120
6.6.1.2 Random-effects Model . . . . . . . . . . . . . 121
6.6.2 Meta-analysis of Rare Events under the CD Framework 126
6.7 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128

7 Ecological Methods 131

7.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

7.2 Time Series Methods . . . . . . . . . . . . . . . . . . . . . . 132
7.2.1 Generalized Event Count Model . . . . . . . . . . . . 135
7.2.2 Tests of Serial Correlation . . . . . . . . . . . . . . . . 135
7.2.3 Parameter-driven Generalized Linear Model . . . . . . 136
7.2.4 Autoregressive Model . . . . . . . . . . . . . . . . . . 138
7.3 State Space Model . . . . . . . . . . . . . . . . . . . . . . . . 141
7.4 Change-point Analysis . . . . . . . . . . . . . . . . . . . . . 142
7.4.1 The u-chart . . . . . . . . . . . . . . . . . . . . . . . . 143
7.4.2 Estimation of a Change-point . . . . . . . . . . . . . . 145
7.4.3 Change-point Estimator for the INAR(1) Model . . . 146
7.4.3.1 Change-point Estimator for the Rate Param-
eter . . . . . . . . . . . . . . . . . . . . . . . 147
7.4.3.2 Change-point Estimator for the Dependence
Parameter . . . . . . . . . . . . . . . . . . . 149
7.4.4 Change-point of a Poisson Rate Parameter with Linear
Trend Disturbance . . . . . . . . . . . . . . . . . . . . 150
7.4.5 Change-point of a Poisson Rate Parameter with Level
and Linear Trend Disturbance . . . . . . . . . . . . . 153
7.4.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 154

© 2016 by Taylor & Francis Group, LLC

xii Contents

7.5 Mixed-effects Poisson Regression Model . . . . . . . . . . . . 155

8 Discrete-time Survival Models 161

8.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 161

8.2 Discrete-time Ordinal Regression Model . . . . . . . . . . . . 164
8.3 Discrete-time Ordinal Regression Frailty Model . . . . . . . 166
8.4 Illustration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
8.5 Competing Risk Models . . . . . . . . . . . . . . . . . . . . . 170
8.5.1 Multinomial Regression Model . . . . . . . . . . . . . 170
8.5.2 Mixed-Effects Multinomial Regression Model . . . . . 172
8.6 Illustration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174
8.6.1 Model Parameterization . . . . . . . . . . . . . . . . . 175
8.6.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . 176
8.6.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . 185

9 Research Synthesis 187

9.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 187

9.2 Three-level Mixed-effects Regression Models . . . . . . . . . 188
9.2.1 Three-level Linear Mixed Model . . . . . . . . . . . . 188
9.2.1.1 Illustration: Efficacy of Antidepressants . . . 190
9.2.2 Three-level Non-linear Mixed Model . . . . . . . . . . 193
9.2.3 Three-level Logistic Regression Model for Dichotomous
Outcomes . . . . . . . . . . . . . . . . . . . . . . . . . 197
9.2.3.1 Illustration: Safety of Antidepressants . . . . 198

10 Analysis of Medical Claims Data 203

10.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 203

10.2 Administrative Claims . . . . . . . . . . . . . . . . . . . . . 203
10.3 Observational Data . . . . . . . . . . . . . . . . . . . . . . . 206
10.4 Experimental Strategies . . . . . . . . . . . . . . . . . . . . . 206
10.4.1 Case-control Studies . . . . . . . . . . . . . . . . . . . 206
10.4.2 Cohort Studies . . . . . . . . . . . . . . . . . . . . . . 207
10.4.3 Within-subject Designs . . . . . . . . . . . . . . . . . 208
10.4.3.1 Self-controlled Case Series . . . . . . . . . . 209
10.4.4 Between-subject Designs . . . . . . . . . . . . . . . . . 211
10.5 Statistical Strategies . . . . . . . . . . . . . . . . . . . . . . . 212
10.5.1 Fixed-effects Logistic and Poisson Regression . . . . . 212
10.5.2 Mixed-effects Logistic and Poisson Regression . . . . . 212
10.5.3 Sequential Testing . . . . . . . . . . . . . . . . . . . . 213
10.5.4 Discrete-time Survival Models . . . . . . . . . . . . . . 215
10.5.5 Stratified Cox Model . . . . . . . . . . . . . . . . . . . 215
10.5.6 Between and Within Models . . . . . . . . . . . . . . 216

© 2016 by Taylor & Francis Group, LLC

Contents xiii

10.5.7 Fixed-effect versus Random-effect Models . . . . . . . 217

10.6 Illustrations . . . . . . . . . . . . . . . . . . . . . . . . . . . 225
10.6.1 Antiepileptic Drugs and Suicide . . . . . . . . . . . . . 225
10.6.2 Description of the Data, Cohort, and Key Design and
Outcome Variables . . . . . . . . . . . . . . . . . . . . 226
10.6.3 Statistical Methods . . . . . . . . . . . . . . . . . . . . 227
10.6.4 Between-subject Analyses . . . . . . . . . . . . . . . . 228
10.6.5 Within-subject Analysis . . . . . . . . . . . . . . . . . 232
10.6.6 Discrete-time Analysis . . . . . . . . . . . . . . . . . . 233
10.6.7 Propensity Score Matching . . . . . . . . . . . . . . . 233
10.6.8 Self-controlled Case Series and Poisson Hybrid Models 236
10.6.9 Marginal Structural Models . . . . . . . . . . . . . . . 237
10.6.10 Stratified Cox and Random-effect Survival Models . . 237
10.6.11 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . 238
10.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 242

11 Methods to be Avoided 245

11.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . 245

11.2 Spontaneous Reports . . . . . . . . . . . . . . . . . . . . . . 245
11.3 Vote Counting . . . . . . . . . . . . . . . . . . . . . . . . . . 246
11.4 Simple Pooling of Studies . . . . . . . . . . . . . . . . . . . . 247
11.5 Including Randomized and Non-randomized Trials in Meta-
analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 248
11.6 Multiple Comparisons and Biased Reporting of Results . . . 248
11.7 Immortality Time Bias . . . . . . . . . . . . . . . . . . . . . 249

12 Summary and Conclusions 253

12.1 Final Thoughts . . . . . . . . . . . . . . . . . . . . . . . . . . 253

Bibliography 255

Index 275

© 2016 by Taylor & Francis Group, LLC

© 2016 by Taylor & Francis Group, LLC
Preface

“It is a capital mistake to theorize before one has data. Insensibly one begins
to twist facts to suit theories, instead of theories to suit facts.”
(Sir Arthur Conan Doyle, Sherlock Holmes)

My (RDG) first encounter with statistical issues related to drug safety

came on the heels of then President George Bush senior taking the drug Hal-
cion and throwing up on the Japanese ambassador. Sidney Wolfe of Public
Citizen immediately filed suit against the U.S. FDA and the Upjohn Company
for a myriad of adverse effects of the drug. The FDA turned to the Institute
of Medicine (IOM) of the National Academy of Sciences to review the mat-
ter. I received a call from Dr. Andy Pope of the IOM asking if I would be
willing and interested in being a member of the IOM committee that was to
opine on this question. I told him that I would think about it and get back
to him. I then called Dr. Joe Flaherty, Dean of the School of Medicine, and
asked him if the IOM (which I had never really heard of) was some kind of
right-wing religious organization. He told me that it was, but that I should do
whatever they asked. I agreed to be a member of the committee, and it led to
a series of different committees which have been among the most interesting,
challenging, and rewarding scientific experiences of my career.
In terms of the safety and efficacy of the drug Halcion, during the first IOM
committee meeting, I sat quietly and listened while the group of distinguished
scientists discussed the evidence base and voiced their opinions. The late Bill
Brown from Stanford and I were the two statisticians on the committee and
at the end of the meeting we said, this all sounds great, but why don’t we
just get the data and find out what is really going on. This turned out to
be a really bad idea because a truck pulled up to my office with a room full
of paper supplied by Upjohn, and Don Hedeker (now at the University of
Chicago) and I spent about a month digging through it. We ultimately found
the appropriate measures of safety and efficacy for each study which we were
able to synthesize across the 25 or so randomized clinical trials (RCTs) and
show that the drug was indeed efficacious at its currently labeled dosages
and that it had a similar safety profile to the other drugs in its class. Most
importantly we found no evidence that the drug would increase one’s level of
nausea in the presence of foreign dignitaries beyond the already elevated base
rate when in the company of politicians.
The IOM report was published in 1997 (IOM, 1997) and reports in the
medical (Bunney et al. 1999) and statistical literatures (Gibbons et al. 1999)

xv

© 2016 by Taylor & Francis Group, LLC

xvi Statistical Methods for Drug Safety

followed. This early experience led to two more IOM studies that involved
drug safety issues, the IOM Study on the Prevention of Suicide (Goldsmith
et al. 2002) and the IOM Study on the Future of Drug Safety (Burke et al.
2006). It was clear to me at this time that there were several singular statistical
features of interest in the analysis of pharmacoepidemiologic data that had
yet to be explored. The issue of whether drugs, in particular antidepressants,
increased the risk of suicide, a condition they were at least in part designed to
treat, was one of the more interesting problems that I had encountered, and
because of my service to the IOM committee on the prevention of suicide, I
was asked to be on the FDA Scientific Advisory Committee which ultimately
placed a black box warning on all antidepressants with regards to suicidal
thoughts and behavior in children, adolescents and young adults. My former
student now colleague and co-author of this book, Anup Amatya, and I have
spent many years studying both this question and the impact of the black
box warning on the treatment of depression in youth and on its relationship
to suicidal events. Statistically it is a very challenging area because suicidal
thoughts and behavior may lead to antidepressant treatment, but the question
is can we disentangle these selection effects from the possible causal effect
of antidepressant treatment on suicidal thoughts, behavior, and completion.
The public health importance of this question is enormous given the large
number of patients with depression and related mental health disorders and
the frequency with which they receive antidepressant treatment. As such,
many of the illustrations of statistical methods presented in this book are
drawn from this area. If you can solve this problem, you can solve most other
pharmacoepidemiologic problems, because most other drug safety problems
are far less complicated.
I tell my students that applied statistics is a lot like dating; you should
hang around wealthy people and marry for love. The same is true in applied
statistics; it takes just as much effort to provide a rigorous statistical solution
to an unimportant problem as it does to solve important problems that can
change the world and improve our public health. Drug safety is certainly one
of the most important problems in this era. What I don’t tell them is that
working at the interface between public policy and statistics is not for the faint
of heart. For every person that admires your work, there is one, and often far
more, that are not at all pleased with the conclusions that you or others draw
from your statistical work. These individuals can be quite vocal about their
dissatisfaction. My favorite blogger comment is “Dr. Gibbons should stick
to his statistical knitting, he doesn’t know his front end from his rear end
when it comes to clinical judgment.” These are generally not waters where
statisticians have experience treading.
This book covers a wide variety of statistical approaches to pharmacoepi-
demiologic data, some of which are commonly used (e.g., proportional report-
ing ratios for analysis of spontaneous adverse event reports) and others which
are quite new to the field (e.g., use of marginal structural models for control-
ling dynamic selection bias in analysis of large-scale longitudinal observational

© 2016 by Taylor & Francis Group, LLC

Preface xvii

data). Readers of this book will learn about linear and non-linear mixed-effects
models, discrete-time survival models, new approaches to the meta-analysis
of rare binary adverse events and when the traditional approaches can get
you into trouble, research synthesis involving reanalysis of complete longitu-
dinal patient records from RCTs (not to be confused with meta-analysis which
attempts to synthesize effect sizes), causal inference models such as propen-
sity score matching, marginal structural models, differential effects; mixed-
effects Poisson regression models for the analysis of ecological data such as
county-level adverse event rates, and a wide variety of other methods use-
ful for analysis of within-subject and between-subject variation in adverse
events abstracted from large scale medical claims databases, electronic health
records, and other observational data streams. We hope that this book pro-
vides a useful resource for a wide variety of statistical methods that are useful
to pharmacoepidemiologists in their work and motivation for statistical scien-
tists to work in this exciting area and develop new methods that go far beyond
the foundation provided here.

© 2016 by Taylor & Francis Group, LLC

© 2016 by Taylor & Francis Group, LLC
Acknowledgments

We are thankful to Fan Yang and Don Hedeker of the University of Chicago,
and Arvid Sjolander of the Karolinska Institute for their helpful review and
suggestions. Hendricks Brown (Northwestern) was instrumental in conceiving
of many of the original examples related to suicide and antidepressants and
Kwan Hur (Veterans Administration and the University of Chicago) helped
greatly in the preparation of the illustrations. The book would not have been
possible without financial support of the National Institute of Mental Health
R01 MH8012201 (Gibbons and Brown Principal Investigators) and the Center
for Education and Research on Therapeutics (CERT) grant U19HS021093
funded by the Agency for Healthcare Research and Quality (Bruce Lambert
(Northwestern) Principal Investigator) 1 . And of course, to my (RDG) teacher,
R. Darrell Bock, Professor Emeritus at the University of Chicago, who taught
me (and in turn all of my students) how to think as a statistician and ignited
my passion for the development of new statistical methodologies for interesting
applied problems. I (RDG) have been an expert witness on various legal cases
involving problems in drug safety for the U.S. Department of Justice, Wyeth,
Pfizer, GlaxoSmithKline, and Merck pharmaceutical companies

1 This project was supported by grant number U19HS021093 from the Agency for Health-

care Research and Quality. The content is solely the responsibility of the authors and does
not necessarily represent the official views of the Agency for Healthcare Research and Qual-
ity.

xix

© 2016 by Taylor & Francis Group, LLC

1
Introduction

“All drugs are poisons, the benefit depends on the dosage.”

(Philippus Theophrastrus Bombast that of Aureolus Paracelsus - 1493-1541)

As noted in the Institute of Medicine report on the Future of Drug Safety

(Burke et al. 2006):

“Every day the Food and Drug Administration (FDA) works to

balance expeditious access to drugs with concerns for safety, conso-
nant with its mission to protect and advance the public health. The
task is all the more complex given the vast diversity of patients and
how they respond to drugs, the conditions being treated, and the range
of pharmaceutical products and supplements patients use. Reviewers
in the Center for Drug Evaluation and Research (CDER) at the FDA
must weigh the information available about a drug’s risk and benefit,
make decisions in the context of scientific uncertainty, and integrate
emerging information bearing on a drug’s risk-benefit profile through-
out the lifecycle of a drug, from drug discovery to the end of its useful
life. These processes may have life-or-death consequences for individ-
ual patients, and for drugs that are widely used, they may also affect
entire segments of the population. The distinction between individual
and population is important because it reflects complex determinations
that FDA must make when a drug that is life-saving for a specific pa-
tient may pose substantial risk when viewed from a population health
perspective. In a physicians office, the patient and the provider make
decisions about the risk and benefits of a given drug for that patient,
whereas FDA has to assess risks and benefits with a view toward their
effects on the population. The agency has made great efforts to bal-
ance the need for expeditious approvals with great attention to safety,
as reflected in its mission to protect and advance the health of the
public.”

Although pre-marketing clinical trials are required for all new drugs be-
fore they are approved for marketing, with the use of any medication comes
the possibility of adverse drug reactions (ADRs) that may not be detected
in the highly selected populations recruited into randomized clinical trials. A
primary aim in pharmacovigilance is the timely detection of either new ADRs

© 2016 by Taylor & Francis Group, LLC