Cohort vs case-control study

Direction is the whole difference

In a cohort, researchers identify people who are exposed and unexposed, then count who develops the outcome. Because the whole group is followed, the proportion who develop it can be calculated, which gives risks, rates and their ratios. In a case control study, researchers choose the number of cases and controls themselves, so the proportion of people with the outcome reflects their sampling decision, not the population. That is why the design can only estimate an odds ratio. Song and Chung's overview of observational studies in Plastic and Reconstructive Surgery sets out the same contrast, including the practical strengths and weaknesses of each design.

When each design is the right choice

Cohorts suit common outcomes, exposures that are rare in the population, and questions about several outcomes of one exposure, such as the effects of shift work on sleep, weight and injury. They also establish that exposure came before outcome, which strengthens causal arguments. Case control studies suit rare outcomes, where a cohort would need to follow enormous numbers of people to see enough cases, and outbreaks, where speed matters. They are also efficient for testing many possible exposures at once. The choice is usually driven by how rare the outcome is and how long it takes to appear. Cost and ethics also shape the decision: an existing registry can turn a cohort question into a retrospective analysis in months, while a new prospective cohort may take a decade to report.

Bias to watch in each

Cohorts are vulnerable to loss to follow up, especially if people who leave differ from those who stay, and to confounding, where exposed and unexposed groups differ in ways that also affect the outcome. Matching and adjustment reduce but do not remove confounding. Case control studies face recall bias, because people with a disease may search their memory harder for possible causes, and selection bias in choosing controls, who should come from the same population that produced the cases. The STROBE statement, which covers cohort, case control and cross sectional designs, asks authors to describe these sources of bias and how they were addressed.

Reporting each design on a poster

The STROBE checklist is the reference for both. For a cohort, report the eligibility criteria, sources and methods of selecting participants, the length of follow up, and the number of outcome events or person time in each group, with crude and adjusted estimates. For a case control study, report how cases were ascertained and controls selected, the matching criteria if any, and exposure counts in cases and controls. On a board, a flow block showing how people were identified, matched and analysed answers the first question a reviewer asks: who exactly was compared with whom.

Common mistakes

Calling a case control study a cohort because data were collected from records is common; the label depends on whether sampling started from exposure or outcome. Reporting an incidence or risk ratio from a case control design is an error that reviewers spot immediately. Describing a matched cohort as if matching removed all confounding overstates the design. Labelling a retrospective cohort as case control because it looked back in time confuses timing with sampling. And omitting the follow up period makes cohort risks impossible to interpret. A last one is choosing controls from a hospital ward for a disease that brings people to hospital for related reasons, which can make an exposure look protective or harmful purely through who was admitted.

Where it shows up in the poster builder

Observational posters use the same building blocks as trials. The protocol block, on 38 layouts, gives the design, the cohort or case and control definitions, exposure, outcome and analysis up to five labelled entries. A flow block of six nodes and one aside, on ten layouts, suits a path from the eligible population through matching to the analysed groups. Key number blocks, on 27 layouts, carry the risk ratio or odds ratio with its interval in the detail line. The criminology example shows a matched cohort on a classic board, with a matching flow and three key numbers traced back to its counts.

Questions people ask