Cohort Studies
Afshin Ostovar
Bushehr University of Medical Sciences
Bushehr, 2011
7/13/2017
1
Epidemiological
Studies
Analytical
Descriptive
7/13/2017
Case Report
Observational
Interventional
Case-series
CrossSectional
Clinical
trials
Ecologic
CaseControl
Field Trials
Crosssectional
Cohort
Community
Trials
2
Cohort studies

Also called follow-up or incidence studies, begin
with a group of people who are free of disease, and
who are classified into subgroups according to
exposure to a potential cause of disease or outcome.

Variables of interest are specified and measured and
the whole cohort is followed up to see how the
subsequent development of new cases of the disease
(or other outcome) differs between the groups with
and without exposure.
7/13/2017
3
7/13/2017
4
Conceptual Framework

The central feature of a cohort study is the collection of exposure
data in a defined population and the subsequent surveillance of
possible outcome events regarding health, morbidity, and mortality.

For this purpose, healthy members of a defined population (the
cohort) are classified according to their exposure status (e.g.
exposed vs. unexposed) and followed over a longer period with
respect to their health status.

Then, the question can be answered if incidence of outcome events
is associated with former presence or absence of exposure, which
would indicate a possible causal relationship.
7/13/2017
5
Concurrent Vs Non-concurrent cohort

Cohort studies can be classified in two major
categories depending on the timing of follow-up
period relative to the time of study conduct:
7/13/2017

Concurrent cohort studies (prospective cohorts)

Non-concurrent studies (historical cohort)
6
7/13/2017
7
Concurrent Vs Non-concurrent cohort

In concurrent studies, the methods for cohort
assembly and data collection can more easily be
controlled.

In non-concurrent studies, the investigators must rely
on data recorded in historical records almost always
for reasons other than medical research.
7/13/2017
8
Non-concurrent cohort
This notable disadvantage of the non-concurrent approach is
compensated by the ability to study exposures, such as occupational
exposures, that meet one or more of the following key conditions:
1)
The exposure can be attributed to selected employed populations
based on individual records of job descriptions or other
employment data,
2)
The exposure is relatively rare in the general population outside the
occupations of interest,
3)
The induction period is long,
4)
The health concern is substantial, making the continued exposure
required for a concurrent study undesirable from a public health
perspective.
7/13/2017
9

In contrast to case-control studies, cohort studies
with their straightforward design allow direct
comparisons of exposed and unexposed persons and
can provide measures of effects for various outcome
events.

Analysis of cohort data requires reasonable care
especially in the steps of data preprocessing for
description and analysis.
7/13/2017
10
Description of Outcome Events in the Cohort
Cumulative Incidence
7/13/2017
11
Risk Vs Rate

The cumulative incidence or risk is unit-free and
represents an individual risk of developing the
disease.

Risk: cohort of acute diseases with short induction
periods and a short time of follow-up, given a fixed
cohort with fixed period of follow-up and a low
fraction of drop-outs.

Rate: cohort of chronic diseases with their long
follow-up periods
7/13/2017
12
Description of Outcome Events in the Cohort
Incidence Density
7/13/2017
13
Data from a fictitious cohort
at risk?
7/13/2017
lost to follow-up?
14
Follow-up time of cohort member No. 4
7/13/2017
15
Crude Vs Age-specific incidence rates for fictitious cohort data
Crude Incidence Rate?
Age-specific (40-49) incidence rate?
7/13/2017
16
Age-specific incidence rates for fictitious cohort data
7/13/2017
17
External Comparisons
7/13/2017

Direct standardisation

Indirect standardisation
18
7/13/2017
Summary Effects of Exposure
19
Key Concerns in Cohort Studies
7/13/2017

Selection of the Study Population

Exposure and Confounders in Cohort Studies

Determining Outcome Events
20
Selection of the Study Population

When a truly representative cohort cannot be obtained,
comparisons with the general population in terms of
mortality and disease rates may not be valid. Thus the
cohort may lack external validity.

However, provided that the recruitment mechanism is
unbiased with regard to the exposure of interest, and
the data obtained on exposure enables the investigators
to stratify their population into exposed and unexposed
subgroups, the estimation of the association between
the exposure and the outcome will be valid (internal
validity).
7/13/2017
21
External Vs Internal Validity
7/13/2017
22
External Vs Internal Validity
7/13/2017
23
Misclassification of exposure

Differential: the degree of
misclassification of the exposure differ
by outcome status.

Non-differential: the degree of
misclassification of the exposure does
not differ by outcome status.
7/13/2017
24
Misclassification of exposure. Cont.

In cohort studies, non-differential
misclassification is the more typical form of
misclassification due to the customary exclusion
of persons with prevalent disease at baseline.

It is unlikely that the measurement of exposure at
baseline will be influenced by the development of
an outcome sometime in the future.
7/13/2017
25
Misclassification of exposure. Cont.

Differential misclassification is potentially a
much greater problem in case-control and crosssectional studies.

Non-differential misclassification always
introduces a bias toward a null finding (a finding
of no association) if the exposure status is
dichotomized; whereas, differential
misclassification can introduce a less predictable
bias.
7/13/2017
26
Sources of data on exposure
7/13/2017

Questionnaires

Type of occupation

Biological material

…
27
Dose-response relationship

The findings of cohort studies regarding the effects of
exposure can be strengthened if it is possible to evaluate a
dose-response relationship.

This requires the assessment of intensity of exposure that
can be quantified as peak, average, or cumulative
exposure.

Sometimes duration of exposure is used as a surrogate for
cumulative exposure. However, using duration in this
way is problematic if the exposure is associated with an
early, perhaps toxic effect.
28
7/13/2017
Non-differential Misclassification of exposure
7/13/2017
29
Change in exposure with time

The concurrent cohort study with its prospective data
collecting does offer the possibility of assessing changes in
exposure while they happen.

To assess changes in exposure patterns, a mechanism has,
however, to be set up specifically e.g. by re-administering
the questionnaire on a regular basis.

This could be done as part of the follow-up mechanism
adopted, though some loss to follow-up will be inevitable.
7/13/2017
30
Determining Outcome Events

The most diseases are relatively rare,

The size of the cohort has to be large,

The follow-up time has to be long and,

The induction period are long
7/13/2017
31
Sources for endpoints





Special surveillance mechanism
Medical records of physicians,
Health maintenance organizations and
hospitals,
Vital statistics systems,
Disease registries
7/13/2017
32
Loss to follow-up

Leads to a loss of power due to the resultant
loss of sample size

Can introduce bias
7/13/2017
33
Loss to follow-up, Cont.

Losses that do not differ by either exposure or
disease status result in no bias, but a loss of power.

Losses that differ by exposure (but not outcome)
status, Internal validity remain intact

losses that differ by outcome status

losses that differ by both exposure and outcome
status
34
7/13/2017
7/13/2017
35
7/13/2017
36
Effects of losses to follow-up that differ by outcome status
7/13/2017
37
Effects of losses to follow-up that differ by both exposure
and outcome status
7/13/2017
38