Observational designs (cross-sectional, case-control, cohort) describe and associate; the randomised controlled trial tests causation. Each yields its own measure.
How it's asked: Match a scenario to its design, and the design to its measure — odds ratio for case-control, relative risk for cohort — plus the classic biases.
Why this is true
The direction of enquiry decides what you can calculate. A cohort starts with exposure and follows people forward, so you observe new cases and can compute incidence — and therefore relative risk and attributable risk. A case-control study starts with disease and looks back at exposure; because the investigator fixes the number of cases and controls, incidence is unknown, and only the odds of exposure can be compared — the odds ratio, which approximates relative risk when the disease is rare. Randomisation distributes known and unknown confounders equally, which is why only an RCT can establish causation directly.
Key points
Designs at a glance
| Design | Starts from | Measure | Best for |
|---|---|---|---|
| Cross-sectional | Population at one time | Prevalence | Burden, hypothesis generation |
| Case-control | Disease (cases vs controls) | Odds ratio | Rare diseases |
| Cohort | Exposure | Relative & attributable risk | Rare exposures, incidence |
| RCT | Random allocation | Efficacy, NNT | Testing interventions |
Common traps
- Incidence can't be calculated from a case-control study — the investigator chose how many cases to include.
- Odds ratio approximates relative risk only when the disease is rare.
Clinical case
Researchers compare 100 patients with mesothelioma and 200 matched controls, asking both groups about past asbestos exposure.
High-yield
Case-control → odds ratio (rare disease). Cohort → relative risk, attributable risk, incidence (rare exposure). RCT → causation.
Quick check
Q1.Relative risk is directly calculated from:
Q2.Recall bias is most characteristic of: