HC4: Cohort studies

Study designs

Medical research almost always requires collection of data. Study design described how is the study set up and how the data are collected. There are 2 types of study design:

• Cohort study
• Case-control study

Cohort

A cohort is a group of people with a common characteristic (gender, diabetes, age, etc.). A cohort study follows such a group over time and records certain outcomes in a follow-up study or longitudinal study. Usually, 2 or more groups are compared. For instance, with patients who are hospitalized for Covid-19, there is a follow-up, record of the mortality and comparison of chloroquine treatment versus no chloroquine treatment.

Example:

Cohort studies are often simplified in tables. An example is the association between smoking and lung cancer:

• Cases: individuals who develop of an outcome of interest
• N = size of the group
• T = person time
  • If a person contributes 50 weeks to the study, and another person 2 weeks, the person time is 52 weeks → 1 person year

With a cohort, the risk/rate can be estimated:

• Cumulative incidence = A/N1→ 1600/100 000 = 1,6%
  • The cumulative is also known as the risk → the risk of smokers getting lung cancer is 1,6%
  • Time is not taken into account
• Incidence rate = (A/T1) = 1600/300 000 years = 5,3 per 1000 person years
  • Time is taken into account
  • It is expected that 5,3 per 1000 individuals develops lung cancer in 1 year
• Risk ratio (RR) = (A/N1)/(B/N0) → (1600/100 000)/(400/200 000) = 1,6%/0,2% = 8
  • Can be either the cumulative incidence ratio or the incidence rate ratio
    • Cumulative incidence risk ratio = (A/N1)/(B/N0)
    • Incidence rate ratio = (A/T1)/(B/T0)
  • The risk of developing lung cancer is 8x as large for smokers compared to non-smokers

Cohort versus dynamic population

There are several differences between a cohort and a dynamic population:

• Cohort
  • Membership is established during a certain period in time, defined by a set of characteristics present during that period
  • Becomes “closed for entry”
  • Number of members (alive) can only go down
  • Characteristics (e.g. age) will change over time
  • Either risks or rates can be calculated
  • E.g. all people born in Leiden in 2018
• Dynamic population
  • Remains “open for entry”
  • Participants will be replaced over time
  • Number of participants can go up, down, or be stable
  • Many characteristics are likely to remain stable over time
  • Calculating rates is most useful
  • E.g. student population in Leiden

Experimental versus observational

Studies are either experimental or observational:

• Experimental: exposure status is determined by a scientist → randomization
• Observational: exposure status is determined by the patient/doctor → the scientist only observes

Observational studies are useful because often, randomization is not possible or necessary:

• Not ethical when studying harmful effects
  • E.g. the association between bacon and colorectal cancer
• No possible/necessary
  • E.g. in genetic studies
• Difficult when patient preferences are strong

Prospective and retrospective

Cohort studies can be subclassified related to time → prospective and retrospective cohort studies:

• Prospective: the researcher starts collecting information
  • The researcher can have an impact on the quality of the data
• Retrospective: the researcher uses information that has already been collected in the past
  • The researcher cannot have an impact on the quality of the data

STROBE statement:

The STROBE statement states that authors refrain from simply calling a study “prospective” or “retrospective” because these terms are ill defined. In practice, the terms are used differently by different researchers. Therefore, whenever authors use these words, they should define what they mean. Most importantly, authors should describe exactly how and when data collection took place.