Cohort Studies

Edric D. Estrella
Principles of Epidemiology for Public Health
Department of Public Health
KFU College of Applied Medical Sciences
Cohort
• a group of people who share a common experience during a
defined time period (e.g., birth cohort, school batch)

• Cohors → the tenth part of a legion of warriors

Cohort Study
• The investigator starts with a group of individuals apparently free of the
disease(s) of interest.

• This group is divided into those who are exposed and those not exposed to
a suspected risk factor.

• The group members are followed during a period of time to determine

and compare the occurrence of outcome (morbidity or mortality rate)
among the ‘exposed’ and those ‘not exposed’.
Cohort Study
• The general feature of a cohort study is that it proceeds from a
suspected cause or etiological agent to the disease outcome.

Cause/ Effect/ Disease

Exposure Outcome

• Other names: follow-up studies, incidence studies, prospective studies,

longitudinal studies, panel studies
Schema of a Cohort Study

Source: Basic epidemiology, R. Bonita, R. Beaglehole, T. Kjellström. 2nd edition (WHO, 2006), p. 47.
Uses of Cohort Studies
1. To identify risk factors for diseases;
2. To identify protective factors against diseases;
3. To identify prognostic factors for outcome of disease;
4. To describe the natural history of disease;

5. To project the incidence of new cases of disease in a population

over a period of time;
6. To assess the effectiveness of preventive programs/measures.
Strengths of Cohort Studies
• Efficient for rare exposure
• Good information on exposure (prospective)
• Can evaluate multiple effects of an exposure
• Efficient for disease with long duration and latent periods
(retrospective)
• Less vulnerable to bias (prospective)
• Can directly measure risk
• Clear temporal relationship between exposure and outcome
(prospective)
Weaknesses of Cohort Studies
• Inefficient for rare outcomes
• Poor information on exposures and other key variables (retrospective)
• Expensive and time consuming (particularly prospective)
• Inefficient for disease with long duration and latent periods
(prospective)
• More vulnerable to bias (retrospective)
Cohort versus Experimental Studies
Similarities
• Both make comparisons across 2 or
more groups.
• Both follow participants to monitor
outcome rates.
• Both usually monitor more than one
outcome.
• Relative proportions of subjects in
compared groups does not reflect
that of the general population.
Differences
• Experimental study investigator
allocates the exposure while cohort
study investigator merely observes.
• Baseline comparability is achieved in
an experimental study through
randomization. Cohort study
investigator must carefully select
groups to achieve comparability.
• Experimental study is prospective
while cohort study may be
retrospective, prospective or
ambidirectional.
Types of Cohort Studies
• based on the time relationship between initiation of study and
occurrence of disease

• Prospective cohort versus Retrospective cohort

Prospective Cohort Study
• At the time the study is initiated,
• exposure has or has not yet occurred; and
• outcome has not yet occurred.

• Data on both exposure status and disease outcome are not yet
available at the beginning of the study and must still be ascertained
by the investigator in the future.
2018

2028

2038
Retrospective Cohort Study
• Both exposure and outcome has already occurred before the start of
the investigation.
• Data on both exposure and outcome were collected and recorded in
the past. These are generally available from existing records.
1998

2008

2018
Distinguishing Features of the Types of Cohort
Studies
Ambidirectional Cohort Study
• has both retrospective and prospective phase

Ambidirectional
2008

2018

2028
Selection of Cohort (exposed group)
• depends on a variety of scientific and feasibility considerations
including the frequency of exposure, and the quality of the exposure
and outcome data
• Sources:
• representative of the general population or geographically defined groups
(e.g. Framingham Study)
• special exposure groups (e.g. Hiroshima bombing survivors)
• special resource groups (e.g. nurses, doctors, factory workers)
Selection of Comparison group (non-exposed
group)
• should be similar to the study group in all respects except the
exposure
• Types:
• internal comparison group → members of the same sample who do not have
the exposure or may have lower levels of exposure
• external comparison group → often the general population of the area from
which the exposed group was drawn
Information on exposure
• sources vary from one investigation to another and from one risk
factor to another
• Potential sources of exposure data:
• pre-existing records (e.g., hospitals, clinics, employment)
• direct physical examination and/or laboratory tests of cohort members
• self reports of cohort members through interviews
• environmental measurements
Information on outcome
• There should be complete and comparable ascertainment of outcome
events on both groups.
• Potential sources of outcome data:
• obituaries and death certificates
• periodic exam of the participants
• records (clinic, hospital, employment, etc.)
Analysis
• Incidence measures (cumulative incidence, incidence density, or
mortality rates) are computed separately for the exposed and
unexposed groups or for each level of exposure level, if applicable.

• Incidence measures are compared between exposed and unexposed

groups in terms of ratios or differences.
Types of Population Studied
Types Defined by Follow-up Appropriate
measure of disease
frequency
Open or Dynamic Changeable Members come Incidence density
characteristic and go; losses may
occur
Fixed Irrevocable event Does not gain new Incidence density
members; losses
may occur
Closed Irrevocable event Does not gain new Cumulative
members; no incidence
losses occur
When follow-up for each member is complete and
the outcome is known for each participant in the
study. (Closed population)

Exposure status Outcome status Total

With (cases) Without (non-
cases)
Exposed a b n1
Unexposed c d n0
Total m1 m0 n
When the duration of follow-up for each member is
not the same and the outcome is not known for each
participant in the study. (Open/Dynamic or Fixed
population with censored data)

Exposure status Total

Exposed Unexposed
Cases a b m1
Total person- L1 L0 L
time at risk
Measure of Association
• reflect the strength of association between the exposure and the
outcome under investigation

• 2 Types:
• Relative measures → Ratio measures
• Absolute measures → Difference measures
Ratio measures: collectively referred as
“Relative Risk” (RR)
• Cumulative Incidence Ratio (CIR) or Risk Ratio

𝐶𝐼1 𝑎/𝑛1
𝐶𝐼𝑅 = =
𝐶𝐼0 𝑐/𝑛0

• Incidence Density Ratio (IDR) or Rate Ratio

𝐼𝐷1 𝑎/𝐿1
𝐼𝐷𝑅 = =
𝐼𝐷0 𝑏/𝐿0
Interpretation of Risk/Rate Ratio (Relative
Risk)
• Ratio measures of association can vary from zero (strong inverse
association) to infinity (strong positive association).

• Under the hypothesis of no association between exposure and

outcome, the null value is equal to 1.
• Risk/Rate in exposed = Risk/Rate in unexposed

• If RR = 0 → no association between exposure and outcome

• If RR < 1 → indicates protective effect (protective factor)
• If RR > 1 → indicates harmful effect (risk factor)
Difference measures: collectively referred as
“Attributable Risk” (AR)
• Cumulative Incidence Difference (CID) or Risk Difference or excess risk

𝐶𝐼𝐷 = 𝐶𝐼1 − 𝐶𝐼0

• Incidence Density Difference (IDD) or Rate Difference or excess rate

𝐼𝐷𝐷 = 𝐼𝐷1 − 𝐼𝐷0

Interpretation of Risk/Rate Difference
(Attributable Risk)
• Absolute measures of association can vary from negative infinity to
positive infinity.

• Under the hypothesis of no association between exposure and

outcome, the null value is equal to 0.
• Risk/Rate in exposed = Risk/Rate in unexposed
Example:
Cohort Study of Oral Contraceptive Use and
Bacteriuria Among Women Age 16-49 Years Old
followed for 3 years
Bacteuria
OC use Total
Yes No
Yes 27 455 482
No 77 1831 1908
Total 104 2286 2390
Risk Ratio
CI OC users = 27/482
= 0.0560
CI non-OC users =77/1908
= 0.0403
CIR = 0.0560/0.0403
= 1.40

The 3-year risk of bacteriuria among OC users is almost 1.5 times

higher than that of non-OC users.
Risk Difference
CI OC users = 27/482
= 0.0560
CI non-OC users =77/1908
= 0.0403
CID = 0.0560 - 0.0403
= 0.0157 or 16/1000

There was an occurrence of 16 new bacteriuria cases per 1000 women

16-49 years old because of OC use.