Case-Control Study
Start with the OUTCOME — take people who have the disease and people who don’t, then look BACKWARD at their past exposures.
An observational study that begins with the outcome: it assembles cases (have the disease) and controls (do not), then compares how often each group was exposed in the past. Because it samples on outcome, it yields an odds ratio, not a relative risk.
Cases vs controls, looking backward at exposure → odds ratio.
Two groups defined by whether they already have the outcome, with their histories compared rather than their futures. The result is the odds of past exposure in cases versus controls — expressed as an odds ratio, which for a rare disease approximates the relative risk.
Read right-to-left in time: pick people by outcome, then trace their exposure backwards. An OR > 1 means cases were exposed more often than controls (exposure associated with the disease); OR < 1 suggests protection; OR = 1 means no association. You cannot read off incidence or absolute risk.
Sampling on outcome makes it brilliantly efficient for rare diseases and conditions with long latency — you do not wait years for cases to accrue, you start with them. The price is that it is retrospective and especially vulnerable to recall and selection bias.
- Direction:
outcome → exposure(looks backward) - Effect measure:
odds ratio(not relative risk) - Best for:
rare diseases& long latency
Doll & Hill smoking and lung cancer (BMJ 1950) — the study that put case-control on the map. They compared ~1,465 lung-cancer patients with matched hospital controls and asked about past smoking; the heaviest smokers had roughly a 25-fold higher association with lung cancer than non-smokers. Starting from the outcome let them probe a long-latency disease without following healthy people for decades. A case-control design gives an odds ratio and a strong signal — the firm causal verdict followed from the prospective British Doctors cohort and the wider body of evidence.
- Recall bias — cases remember and report past exposures more thoroughly than controls.
- Control selection — controls must come from the same source population, or the comparison is biased.
- Quoting incidence or relative risk — sampling on outcome yields an odds ratio only, never true RR or incidence.
Quick check
Why does a case-control study report an odds ratio rather than a relative risk?
Answer: Because participants are sampled on their outcome (you choose how many cases and controls), the underlying population risk — the denominator needed for incidence and relative risk — is unknown. Only the odds of exposure can be compared, giving an odds ratio.
Ready to build your plan? EMF Premium gives you all 40,000+ questions, 20 mocks and 1,215 OSCE stations from £29/month — or a one-off 3- or 6-month pass.