2. Cohort studies outline• Big picture
• Basic approach
• History
– Origins
– Landmark studies
– Current era
• Strengths and challenges
• Example: asbestos and mortality – occupational non-
concurrent, prospective cohort study
• Example: estrogen and CHD – NHS concurrent,
prospective cohort study
• Use of Poisson regression in cohort studies
• Summary
3. Big picture
• Fundamental study design in observational
epidemiology
• Other observational designs are best understood
as sampling from a cohort experience
– Some do this literally
– Others theoretically
4. Basic approach
• Group of people identified and followed up over
time to ascertain health outcomes
– A cohort study could involve one population group or
cohort
• Example: population representative cohort of the NYC
metropolitan area followed over the 2 years after 9/11/2001
5. Basic approach
– A cohort study could involve two or more cohorts
(sometimes called a double cohort)
• Example: exposed cohort of premature infants who had
received palivizumab (anti-RSV monoclonal antibody) in the
first year of life compared to unexposed cohort of premature
infants (no palivizumab)
– All called cohort studies
9. Basic approach
• Cohort studies can be conducted looking back in
time, forward in time, or some combination
Szklo Figure 1-16
10. Basic approach
• Terminology on types/characteristics of cohort
studies for this class
– Investigator perspective
• Concurrent: investigator begins study before outcome occurs
• Non-concurrent: investigator begins study after outcome occurs
11. Basic approach
• Terminology cont.
• SER teaching epidemiology symposium – Tim Lash
– Timing of record of exposure with respect to outcome
• Prospective: exposure data recorded before outcome occurs (can
be recorded in records that investigator later abstract, by interview
before outcome occurs etc.)
• Retrospective: exposure data recorded after outcome occurs
• Understand how and when data collection took place and consider
the potential for bias in that process
12. Basic approach
• Group identified by particular exposures,
particular population, occupational group,
convenience (depends on purpose of study)
– Will describe some examples in detail after covering
the basics
13. Basic approach
• One strong advantage of a cohort study over
other designs is that the dynamic nature of many
exposures and their relations in time to disease
occurrence is most comprehensively captured in
the cohort design
14. Basic approach
• However, classification of exposure over time
may be extremely complicated
– Participants may have different exposures for different
time periods during the study
– Effects of exposure may be expected to accumulate
– Participants with similar average or cumulative past
exposures may have different patterns of exposure
that do not have the same effect on disease
• Example: heavy smoker in youth, light smoker currently vs
light smoker in youth, heavy smoker currently
15. Basic approach
– Time between when an exposure would have caused
disease and when disease was detected must be
considered
– There may be little information available to guide
decisions about how to classify exposure or what
latency period is appropriate
16. Basic approach
• Outcomes are ascertained over follow-up time
– Outcome examples: death, disease incidence (single
or multiple occurrence), change in a measure of
health status (e.g., BMI, depression symptoms)
– Outcomes must be common enough in the population
studied or the design will be inefficient
17. Basic approach
– Example: Leukemia incidence is approximately
10/100,000 person-years in the general population; a
massive cohort would be required to detect a
reasonable number of leukemia cases over a
reasonable follow-up period
18. Basic approach
• Possible data sources
• Exposure
– Records
• Occupational
• Medical/pharmacy
– Interviews/questionnaires
– Direct measurements on participants
• Outcome
– Vital records
– Disease registries
– Medical records
– Interviews/questionnaires
– Direct measurements on participants
19. Basic approach
• Closed or fixed cohort study collects information
from a group defined at the start of follow-up
– Szklo Fig 1-13
– Example: study of the effects of exposure to hurricane
Katrina on mental health of children with recruitment
from schools over a short enrollment period and
follow-up over several years
20. Basic approach
• Open cohort study collects information from a
changing roster of individuals
– Example: study of effects of HAART on progression of
HIV disease with participants identified when they first
test positive for HIV with enrollment continuing over
10 years