Epidemiological studies part III

CLS 351
Experimental studies
Outline
Experimental studies:
• Randomized controlled trials (RCT).
• Field trials.
• Community trials.
 Recap: Types of epidemiological studies

 Descriptive studies
•Individuals
• Case reports
• Case series
• Cross-sectional surveys
•Populations (ecological studies) Observational
 Analytical studies
• Cross-sectional studies 
• Cohort 
• Case-control 

• Intervention studies
• Randomized Trials Experimental
Hierarchy of evidence
• Systematic reviews and meta-analysis high
• (Randomised controlled) trials
• Cohort studies
• Case-control studies
• Cross-sectional studies
• Ecological studies
• Case reports/series low
• Expert opinion
Experimental (intervention) study
• Experimental studies resemble cohort studies in that they
require follow up of the subjects to determine outcome.

• The essential difference in experiments is that they

involve some action or manipulation or intervention
done by the investigators. This contrasts with
observational studies, where the investigators take NO
action, but they only observe.

• Experimental (intervention) studies can be more difficult

to design and conduct than observational studies due to
their unique problems of ethics, feasibility, and cost.
Experimental (intervention) study

• In order for the experiment to be informative, it must be

controlled (using a control group);
• comparing outcome in those receiving the intervention
with another similar group who did not receive the
intervention.

• Experimental studies are now done on humans rather

than experimental animals.
• This type of study can take one of three forms:
1. Randomized controlled (clinical) trial;
2. Field trial;
3. Community trial.
Clinical trials

• A clinical trial is a planned experiment on humans,

designed to measure the efficacy or effectiveness of
an intervention.
• The intervention is usually a new drug, surgical
procedure, vaccine, etc...
Why conduct Inappropriate
to change
a trial? practice
without an
evidence base
The design of the Therefore
trial is dictated by essential to
the research validate a new
question intervention

A trial is the
‘gold standard’
way to do this
Features of clinical trials
• Must contain a control group.
• Participants are followed over time.
• Participants should be randomized to intervention
or control group.
• Intervention AND control groups are enrolled,
treated and followed up over the SAME period of
time.
• Ideally, participants and/or researchers are not
aware as to whether a person is in the intervention
or control group (blinding).
What sort of questions can a trial
answer?
A study comparing and assessing the effectiveness
of a new intervention or treatment.

Possible uses:
1. Is treatment better than no treatment?
2. Is new treatment better than old treatment?
3. What/how common are side effects?
4. ‘Proof’ of causation?
Why a control group?
• A control group is composed of those study
participants who did not receive the intervention
under study. They may receive a placebo or
standard clinical practice.
• A control group must be included in any trial,
otherwise you cannot be sure why the outcome
happened; it may be due to the treatment or it may
have happened anyway.
Placebo
• A pill or a substance given to a patient like a drug
but without any physical effects on the patient.
• For example, some people in a study might be given a
new drug to lower cholesterol. Others would get a
placebo (dummy pill). None of the people in the study
will know if they got the real treatment or the placebo.

• Researchers then compare the effects of the drug

and the placebo on the people in the study. That
way, they can determine the effectiveness of the
new drug and check for side effects.
Placebo effect
• Sometimes a person can have a response to a
placebo. The response can be positive or
negative.
• For instance, the person's symptoms may improve, or
the person may have what appears to be side effects
from the treatment. These responses are known as
the "placebo effect“.
• There are some conditions in which a placebo
can produce results. Studies show that placebos
can have an effect on conditions such as:
• Depression.
• Pain.
• Sleep disorders.
Confounding
The effect of an extraneous variable that wholly or
partially accounts for the apparent effect of the study
exposure or that masks an underlying true
association.
How to Identify Potential Confounders
(1) Knowledge of subject matter
(2) At data analysis:
(a) Examine if an extraneous variable
satisfies 3 conditions:
- It is associated with the study
exposure in the control group (source
population).
- It is associated with the study
disease (may or may not be causal) in the
absence of study exposure.
- It is not a consequence of exposure
(i.e., it is not in the causal pathway between
the exposure and the disease)*.
Confounding

Example

Alcohol Drinking Oral Cancer?

Definition of a Confounder ALCOHOL
& ORAL
CANCER

Cases Controls

Alc + 142 104

Alc – 58 96

ORcrude = 2.26
Alcohol drinking is associated with oral cancer ?
How to Identify Potential Confounders

Cases Controls

Alc + 142 104 246

Alc – 58 96 154

200 200

ORcrude = 2.26
Could smoking be confounding this association?
 How to Identify Potential Confounders
Could smoking be a confounder of the association?
Among Controls (n=200) Among ALC – (n=154)
Smk + Smk – Cases Controls
72 32 Smk +
14 8
Alc+ (90%) (27%) (24%) (8%)

Alc– 8 88 Smk – 44 88

OR = 24.8 OR = 3.5

Smoking alcohol use Smoking risk of Oral

Cancer
 Definition of a Confounder

Alcohol use (Exposure variable)

Smoking
(Confounder)
Oral Cancer (Outcome variable)
Confounding: relationship between
coffee drinking (exposure), heart
disease (outcome), and a third
variable (tobacco use)
Control for Confounding
Design Stage
•Randomization (RCT only).
•Matching.
Randomization
• A randomized clinical trial is one ‘that involves the
formation of the treatment groups by the process of
random allocation’.
• Confounders are on average distributed equally
among the study groups.

• Randomization is done to remove bias in treatment

allocation as well as control for confounding factors.
Randomization (example)
In a famous early study on BCG vaccine for TB in
children, deaths from TB were 5x higher in the
control group than the vaccinated children > further
investigation showed that doctors had tended to give
the vaccination to children with cooperative parents
and left the rest as controls > those cooperative
parents were more likely to be educated, health
conscious and therefore have lower mortality from
TB regardless of the vaccine.
Blinding
Some of those taking part in the trial are prevented
from knowing information about allocation to
treatment group.

Single blind

Double blind

Triple blind
Blinding
• Single blinding means that that the patient does
not know whether they are getting the treatment
or not.
• In a double blind trial, neither the patient nor the
investigator knows which treatment they are
getting. This is done to prevent bias in
measurement or reporting of outcome.
• Triple blinding: it is a trials where the patient and
the investigator does not knows which treatment
they are getting, in addition, the identities are also
withheld from the statistician(s) who conduct the
analysis of the data.
Blinding: pros & cons

Cons Pros
No
participant
bias
No Dr/local
investigator
bias
Complexity
and cost No research
team bias
Example of clinical trials
Field trials
In contrast to clinical trials, it involve people who are
disease-free but are at risk:
data collection takes place "in the field",
among non-institutionalized people in general population.
Example,
Salk vaccine for the prevention of poliomyelitis,
involved over one million children.
Community trials
• In this form of experiment, the treatment groups are
communities rather than individuals.
• This is particularly appropriate for diseases that are
influenced by social conditions, and for which prevention
efforts target group behavior.
For example: Water is fluoridated in a community while in
another is not > the two communities are followed up
• if the community with non fluoridated water starts
to complain teeth problem, the effects of fluoride
are confirmed.
Systematic error
• Systematic error (or bias) occurs in epidemiology
when results differ in a systematic manner from the
true values. A study with a small systematic error is
said to have a high accuracy.
• The possible sources of systematic error in
epidemiology are many and varied; the principal
biases are:
• Selection bias.
• Measurement (or classification) bias.
Selection bias
• Selection bias occurs when there is a systematic
difference between the characteristics of the
people selected for a study and the characteristics
of those who are not.

 An obvious source of selection bias occurs when

participants select themselves for a study, either
because they are unwell or because they are
particularly worried about an exposure.
Selection bias

The relation between exposure and disease is

different for those who participate and those
theoretically eligible for the study (source
population).
To Minimize Selection Bias

• Controls from the same source population where

cases arose (e.g., same hospital or community)
both cases and controls are affected by similar
selection factors.
• Controls represent the source population in terms
of prevalence of exposure.
• Keep nonresponse to minimum. Characterize non-
responders.
Classification Bias
• Errors in determining an individual’s exposure and
outcomes status, and they occur unevenly among
cases and controls.
• Errors in exposure and outcome classification
depend on case-control status.
• This bias occurs when the individual
measurements or classifications of disease or
exposure are inaccurate – that is, they do not
measure correctly what they are supposed to
measure.
Classification Bias

Prevention:
• Try to avoid recall and interviewer biases.
• Use validated reliable assessment for biomarkers;
assess cases and controls together.
• Use state-of-the-art accurate diagnostic
techniques for outcome assessment.
Essential reading

Bonita et al. Basic Epidemiology. (2006) pages 49-60.