Cohort Study

Definition

Distinguishing Features

• The cohorts are identified prior to the appearance of the disease under investigation
• The study groups are observed over a period of time to determine disease frequency
• The study proceeds forward from cause to effect

Concept of "Cohort"

• Birth cohort - all persons born in the same year (e.g., birth cohort of 2010)
• Exposure cohort - persons exposed to a common drug, vaccine, or infection in a defined period
• Marriage cohort - persons married in the same period
• Disease cohort - e.g., all those who survived a myocardial infarction in one year

Framework (Basic Design)

• Incidence in exposed = a/(a+b)
• Incidence in unexposed = c/(c+d)
• If a/(a+b) is significantly higher than c/(c+d), an association is suggested

Types of Cohort Studies

1. Prospective Cohort Study (Current Cohort Study)

• The outcome has not yet occurred when the investigation begins
• Most begin in the present and continue into the future
• Examples: Framingham Heart Study, Doll & Hills study on smoking and lung cancer, Royal College of GPs study on oral contraceptives

2. Retrospective Cohort Study (Historical Cohort Study)

• Outcomes have already occurred before the study starts
• The investigator goes back 10-30 years using existing records (employment, medical), then traces subjects forward to the present
• Also called: historical cohort, prospective study in retrospect, non-concurrent prospective study
• Example: A 1978 study of 17,080 babies born 1969-1975 investigated effects of electronic foetal monitoring - neonatal death rate was 1.7x higher in unmonitored infants
• Generally more economical and produces results more quickly than prospective studies

3. Combination (Retrospective + Prospective)

• The cohort is identified from past records, assessed up to the present for outcome, then followed prospectively into the future
• Example: Court-Brown and Doll (1957) studied 13,352 patients who received radiation therapy for ankylosing spondylitis (1934-1954) and found substantially higher death rates from leukaemia/aplastic anaemia

Indications for Cohort Studies

• (a) There is good evidence of an association between exposure and disease (from clinical observations, descriptive studies, and case-control studies)
• (b) Exposure is rare but incidence of disease is high among exposed (e.g., workers in industries, exposure to X-rays)
• (c) Attrition of study population can be minimized (easy follow-up, stable cohort)
• (d) Ample funds are available

Elements of a Cohort Study

1. Selection of study subjects - from general population or special groups (professional groups, exposure groups)
2. Obtaining data on exposure - documenting exposure status at baseline
3. Selection of comparison groups - unexposed group comparable in all other variables
4. Follow-up - both groups followed under identical conditions over time
5. Analysis - incidence rates calculated and compared

Measures Derived from Cohort Studies

Relative Risk (RR)

• Ratio of incidence among exposed to incidence among unexposed
• RR = 1: no association; RR > 1: positive association; RR < 1: protective effect
• Example: RR = 10 means smokers are 10 times more likely to develop lung cancer than non-smokers

Attributable Risk (AR)

• Difference in incidence between exposed and unexposed groups
• Indicates what proportion of disease is attributable to the exposure
• Formula: (Incidence in exposed - Incidence in unexposed) / Incidence in exposed × 100

Population Attributable Risk

• Incidence in total population minus incidence among unexposed

Advantages

• (a) Incidence can be calculated directly
• (b) Multiple outcomes can be studied simultaneously (e.g., smoking cohorts revealed associations with lung cancer, coronary heart disease, peptic ulcer, oesophageal cancer)
• (c) Provides a direct estimate of relative risk
• (d) Dose-response ratios can be calculated
• (e) Certain forms of bias are minimized (e.g., misclassification of exposed vs. unexposed groups)

Disadvantages

• (a) Requires large numbers of subjects; unsuitable for rare diseases
• (b) Long follow-up often needed (20-30 years in cancer studies) - results are delayed
• (c) Administrative problems: loss of staff, funding, extensive record keeping
• (d) Loss to follow-up (migration, loss of interest, refusal)
• (e) Selection of truly representative comparison groups is difficult
• (f) Changes in diagnostic criteria during prolonged follow-up
• (g) Expensive
• (h) The study itself may alter participants' behaviour (Hawthorne-like effect)
• (i) Ethical problems - as evidence accumulates, investigators are obliged to intervene
• (j) Limited to a finite number of factors that can be studied

Cohort Study vs. Case-Control Study

Classic Examples