Cohort studies

Cohort studies

Earlier we considered John Snow’s 19th century study of a cholera outbreak in London, which is an example of a cohort study. Cohort studies follow one or more groups of individuals over time and are considered to be the archetype of all epidemiologic studies.⁴ Generally, cohort studies will follow a cohort of patients with an exposure and a cohort who are unexposed (or have an alternative exposure) and compare the incidence of a disease or other outcome among the exposed and unexposed. In Snow’s study, he compared the risk of cholera in a cohort of individuals who received water from the Southwark and Vauxhall Company (exposed to contaminated water) and a cohort who received water from the Lambeth Company (unexposed to contaminated water).

In a more recent example of a cohort study, researchers examined the association of different antipsychotic medications with hyperglycemic emergencies. The study compared new users of the atypical antipsychotic risperidone to new users of olanzapine, other atypical antipsychotics (mostly quetiapine), and typical antipsychotics. Patients in each of these cohorts were followed for up to 365 days after starting one of these drugs to assess whether they experienced a hyperglycemic emergency. One of the study findings was that patients in the “other atypical antipsychotics” group had a lower risk of hyperglycemic emergencies than those taking risperidone.²⁰

Those who are studied in a cohort study must be from a population at risk of the disease (or study outcome of interest).⁴ The above study of hyperglycemic emergencies excluded patients who had been recently hospitalized for a hyperglycemic emergency to ensure patients were at risk of the study outcome.²⁰ Similarly, a study examining risk of a chronic disease such as diabetes would exclude patients previously diagnosed with diabetes. If a disease takes time to develop and manifest itself, this needs to be taken into account.⁴ For example, a cohort study analyzing the association between the type 2 diabetes medication pioglitazone and bladder cancer examined incidence of bladder cancer from one year after exposure began in part because patients would not be at risk of the disease from exposure to the drug during the first year of exposure.²¹

When comparing the incidence of disease among exposed and unexposed groups in a cohort study, researchers may use the absolute measures such as the risk difference or relative measures such as the risk ratio discussed earlier or other measures. For example, the study of bladder cancer mentioned above found the incidence rate of bladder cancer among pioglitazone users was 121.0 per 100,000 person years compared to 88.9 per 100,000 person-years for patients who used non-thiazolidinedione antidiabetic drugs, indicating a rate difference of 32.1 per 100,000 person-years.²¹