Cohort Study Design

Dr (Mrs.) I A Adeoye
EMS dept
 Outline
• What is Epidemiology
• Types of Research Designs
• Definition of Cohort Study
• Basic Component
• Design of Cohort Study
• Selection of Exposure
• Measurement of Exposure
• Selection of controls
• Sources of bias
• Strengths and weakness
 What is Epidemiology
• Epidemiology is the study of the frequency,
distribution and determinants of health related
events among human population and the
application of this study to the prevention of
diseases and the promotion of health
 Epidemiological study designs
• Descriptive - describe occurrence of outcome
• Case Report
• Case Series
• Correlational studies
• Cross-sectional studies
• Analytical - describe association between exposure
and outcome. Useful in elucidating etiologic or causal
relationships
• Case Control
• Cohort
• Experimental
 Two broad categories
1. Observational – no manipulation of study
factor by the investigator
2. Experimental –study factor is manipulated by
the investigator
RANDOMIZATION outcome

Intervention
no outcome

Study
population
outcome
Control

no outcome

baseline
future

time
Study begins here (baseline point)
 Definition
• The second major type of observation analytic design.

• In a cohort design group of individuals are defined on

the basis of presence or absence of exposure to a
suspected risk factor for a disease.

• At the time exposure status is defined, all potential

subjects must be free from the disease under
investigation, and eligible participants are then
followed over a period of time to assess the occurrence
of the outcome.
• Also called follow-up studies, incidence
studies, panel studies, longitudinal studies, or
prospective studies
 Basic Components of a Cohort Study
1. Enrolling subjects at risk for developing the
outcome.
2. Measuring exposure status on study
participants
3. Following subjects overtime
4. Recording the outcomes
 Design of a Cohort Study
• The hallmark of a cohort design is to compare disease
outcomes between the exposed a non exposed.

NOT
EXPOSED AND
EXPOSED

DISEASE
DISEASE DISEASE DOES
DOES NOT
DEVELOPS NOT DEVELOP
DISEASE DEVELOP
DEVELOPS
• If a positive association exist between the
exposure and the disease, we would expect
that the proportion of exposed group in
whom the disease develops (incidence in the
exposed group) would be greater than the
proportion of the non-exposed group in whom
the disease develops (incidence in the
unexposed)
 Design of a Cohort Study

TABLE 9- 1. Design of a Cohort Study

Then Follow to See Whether

Incidence Rates
Disease Develops Disease does not Develop Totals
Of Disease

a+b Aa/a+b
Exposed a b
First Select c/c+d
Not exposed c d c+d
 Smoking and Coronary heart Disease (CHD): A
Hypothetical Cohort Study of 3000 Cigarette Smokers
and 5000 Non smokers
CHD+ CHD- Total Incidence per
1,000 per year

Smoke 84 2916 3000 28.0

cigarettes
Do not smoke 87 4913 5000 17.4
cigarettes

Incidence among smokers = 84/3000 = 28.0 per

1000
Incidence among non smokers = 87/5000 = 17.4
per 1000
 Selection of the Exposed Population
• General population (the outcome of interest has a high
incidence rate)
• Special exposure groups (Rare exposures from a high risk
population - More efficient accrual)
– Special exposure groups
• (e.g. certain occupations e.g. asbestos workers with high levels of
specific exposures)
• unusual diet or lifestyle practices
• undergone a particular therapy
– Special resource groups (e.g. physicians, nurses)
– Geographically Exposed to a particular event (e.g. Hiroshima
residents
– facility-defined groups (e.g. hospitals with specialized
maternity care)
 Who do you enrol in the cohort?
• The subjects must be at risk for developing the
outcome and must be free of the outcome at
the time of enrolment
• The Framingham Heart Study is an example of
enrolment from the general population. It
improves the generalizability of the study. It
may not be suitable if the exposure is rare.
• The Nurses Health Study is an example of
enrolment from a specific population
 How do you want to quantify the
exposure?
• If you wish to study the association between
smoking (exposure) and the incidence of heart
attack.
• The various options include
– Smoking versus not smoking (present/absent)
– Extent of smoking (non-smoker vs. light smoker vs
heavy smoker)
– Number of cigarettes smoked per day (intensity)
– Duration of smoking ( no. of years smoked)
– Pack-years of smoking (Duration and intensity)
Selection of the Comparison Group
1. Internal comparison group - General cohort
group classification into exposure categories
is by levels of exposure e.g. levels of cigarette
smoking or quintiles of blood cholesterol
levels
2. External Comparison Group – General
population
 Single Sample Cohort Study Design
TIME

Diseased
Exposed
Not Diseased
Target Disease-Free
Cohort
Population
Diseased
Not Exposed

Not Diseased
Multi-Sample Cohort Study Design
TIME

Diseased
Study Exposed
Cohort Not Diseased

Diseased
Control Not Exposed
Cohort
Not Diseased
 Sources of Data
1. Study subjects – Interviews and
questionnaires
2. Medical examination – blood test
3. Existing records - Medical records
employment records, death certificates
4. Direct measurement from the environment
 Follow-up of Participants
1. Follow-up visit
2. Follow-up mailings
3. Clinical encounters
4. Medical records
 Analysis
• Cohort studies allows for the direct calculation
of incidence rates.
• Calculation of Incidence Rates of the specified
outcome among the cohort under
investigation.
• Incidences are compared among the exposed
and the non-exposed group
• Cumulative incidence OR incidence rates
 How will you measure Incidence?
• Cumulative Incidence
– Requires common period of potential follow-up
– Requires complete follow-up
– No losses due to competing risk
• Incidence Rate
– More general measure in Cohort Studies
– Valid when follow is not complete
– Assumes the reason people are being lost to
follow-up is not related to the OUTCOME
How do we measure association
 Risk Calculation in Cohort Study

TABLE 9- 1. Design of a Cohort Study

Then Follow to See Whether

Incidence Rates
Disease Develops Disease does not Develop Totals
Of Disease

a+b Aa/a+b
Exposed a b
First Select c/c+d
Not exposed c d c+d

a/a+b=Incidence
Incidenceininexposed
exposed c/c+d= Incidence in non exposed
a/a+b=
Example: Calculating the Relative Risk
Disease Status
CHD cases No CHD
(Cases) (Controls) TOTAL

Exposure Smoker 112 176 288

Status Non-
88 224 312
smoker

A/(A+B) 112 / 288

Relative Risk = = = 1.38
B/(C+D) 88 / 312
Example: Interpreting the Relative Risk
Relative Risk = 1.38

The risk of developing CHD is 1.38 times higher for

a smoker than for a nonsmoker.
or
The risk of developing CHD is 38% higher for a
smoker than for a nonsmoker.
Interpretation of Relative Risk
 RR<1 RR=1 RR>1

Risk for disease is

Risk comparison Risk of disease is equal Risk for disease is
lower in the exposed
between exposed for exposed and higher in the exposed
than in the
and unexposed unexposed than in the unexposed
unexposed

Exposure reduces
Exposure as a risk disease risk Exposure increases
Particular exposure is
factor for the disease risk
(Protective not a risk factor
disease? (Risk factor)
factor)
 Types of Cohort Study
• Prospective cohort design - (longitudinal study
or concurrent study)
• Retrospective cohort design – (historical
cohort design)
• Ambi-directional cohort design – (historical
prospective)
 Cohort study data collection

DESIGN PAST PRESENT FUTURE

Prospective E D

Retrospective E D

Historical prospective E E D
 Time frames for hypothetical prospective
cohort study and a hypothetical retrospective cohort study
begun in 2008

Prospective Retrospective
Defined
2008 1988
Population

NON - RANDOMIZED
2018 1998
Exposed Non-Exposed

NO NO
DISEASE DISEASE
DISEASE DISEASE
2028 2008
 Prospective versus Retrospective
Cohort Studies
• The classification depends on the temporal
relationship between the initiation of the
study and the occurrence of the disease.
• In a prospective cohort study, the follow-up
period begins on the date of the study and all
outcomes occur after the beginning of the
study.
• In a retrospective cohort study, all outcomes
occur before the beginning of the study.
 Prospective versus Retrospective
Cohort Studies
PROSPECTIVE RETROSPECTIVE
1. Better control of the quality 1. Poorer control of the
and quantity of the data quantity and quality of the
data
2. More time consuming 2. Less time consuming
3. More expensive 3. Less expensive
 BIAS
• Bias may be defined as any systematic error in
an epidemiologic study that results in an
incorrect estimate of the association between
exposure and the risk of disease.
• Selection Bias – error that arises in the process
of identifying the study population.
• Information Bias – error that arises from the
measurement of variables.
 Bias in Cohort Studies
• Losses to follow up
– Most important form of bias in cohort studies.
Particularly if the reason for the loss is related to
the outcome of the study. E.g. smokers who
develop early signs of developing CHD
• Effects of non participation
– it cannot be assumed that those who chose to
participate had the same prevalence of exposures
nor incidence of disease as those who did not
participate
 Bias in Cohort Studies
• Measurement Bias
– Detection Bias occurs in a cohort study if detection of
the outcome is performed differently for exposed and
non-exposed groups for instance from over
surveillance of the exposure group compared to
control. This can be avoided by carrying out uniform
procedures in both groups.
– The knowledge of exposure status or study hypothesis
may influence the interviewer. Solution is by blinding
of the interviewers to the outcome of the study.
 Confounding
• It involves the possibility that the observed association
is due totally or in part to the effects of differences
between the study group other than the exposure
under study that could affect their risk of developing
the outcome of interest
• It results in the mixing of the effects of the exposure
under study on the disease by a third factor with must
be associated with the exposure and independent of
the disease it is a risk factor of the outcome
• Example in the relationship between smoking and
heart disease alcohol is a possible confounder
 Confounding
Distribution of Potential confounders
Smokers Non Smokers
Mean Age 48.1 51.7
Mean SBP 129.8 135.9
% Male 53.9 34.1
 Advantages
1. The temporal sequence between exposure and
disease is more clearly established because the
participants are disease-free at the time the disease is
defined.

2. Are well suited for assessing the effects of rare

exposures especially those that arise in occupational
settings.

3. It allows for the examination of multiple effect of a

single exposure. Hence it can provide information on
the full range of health effects of a single exposure.
 Disadvantages
1. Very time consuming
2. Expensive
3. Potential for bias from losses to follow-up
4. Not efficient for rare diseases
 Assignment
1. Watch –
❑ The Framingham Heart Study was initiated in 1948 by
enrolling 5,209 study subjects, age 30-60 and following
them for the primary outcome of Cardiovascular
Disease (CVD). Detailed information about this study
can be found at
❑ • [Link]
❑ • [Link]

2. Appraising Cohort Studies in Africa

❑ GROUP WORK – per program/department
• THANK YOU