Skip to content
Scintria
ToolsGrant writerFunding
HomeLiteratureClinical trialsEvidence LabProtocol packData analysisLearnProjectsLibraryReferencesApplications
Start free
ScintriaStart free

Learn / Epidemiology / Study designs

Case-control studies

14 minute read · 6-question quiz

Download this lesson as a colour Word or PDF study guide with the figures and answers. Included in paid plans; reading and quizzes stay free.Sign in

What you will learn

  • Explain the logic and the steps of a case-control study

  • Choose cases and controls properly and recognise matching

  • Calculate and interpret the odds ratio and its confidence interval

In a case-control study, you start with the outcome: you identify people with the disease (cases) and a comparison group without it (controls), then look back to compare their past exposures.

1CasesPeople with thedisease2ControlsPeople without it,from the same sourcepopulation3Look backAsk about pastexposures4CompareOdds of exposure incases vs controls
The case-control design runs backwards from outcome to exposure.

When it is the right design

  • The disease is rare (a cohort would need huge numbers): cancers, birth defects, sudden infant death.
  • The disease has a long latency: waiting decades is impractical.
  • You are investigating many exposures for one outcome, or an outbreak with an undefined exposed population.
  • Speed and cost matter: fewer participants than a cohort for the same power.

It is not suitable for rare exposures (use a cohort of the exposed) or for studying many outcomes.

Choosing cases

Define the disease precisely and use incident (newly diagnosed) cases where possible. Prevalent cases include long survivors, and exposures may have changed after diagnosis. Use clear diagnostic criteria, ideally laboratory-confirmed. Sources: hospital registers, disease registers, surveillance.

Choosing controls: the hardest part

Controls must represent the source population that gave rise to the cases: those who, if they had developed the disease, would have been selected as cases. The control group tells us how common the exposure is in that population.

  • Population controls (random sample of residents; best but costly).
  • Hospital controls (convenient, but patients with other illnesses may share the exposure: smokers fill many wards).
  • Neighbourhood or family controls (match on shared environment).
  • Using more than one control group increases confidence if the answer agrees.
  • Matching (on age and sex, for example) improves efficiency but a matched variable can no longer be studied and must be handled in the analysis (conditional logistic regression, or matched odds ratio).
  • Sample 1 to 4 controls per case; beyond 4 gains little.

Worked example: charcoal stoves and pneumonia

A study enrols 100 children admitted with pneumonia (cases) and 100 community controls of similar age. Mothers are asked whether the child lived in a home using indoor charcoal cooking.

CasesControls
Exposed (indoor charcoal)6030
Not exposed4070
Total100100
  • Odds of exposure in cases = 60/40 = 1.5; in controls = 30/70 = 0.43.
  • Odds ratio = (60 × 70)/(40 × 30) = 3.5.
  • SE(ln OR) = sqrt(1/60 + 1/40 + 1/30 + 1/70) = 0.299; 95% CI 1.95 to 6.29.

Children with pneumonia had 3.5 times the odds of living with indoor charcoal cooking than controls (95% CI 1.9 to 6.3). Because pneumonia is relatively uncommon in the population, this OR is also a fair estimate of the risk ratio.

Why the odds ratio, and why it works

You cannot compute risk, because the researcher fixed the number of cases and controls. But the exposure odds ratio equals the disease odds ratio (a mathematical identity). When the disease is rare, odds approximate risk and so the OR approximates the risk ratio. In a case-control study the OR is the only valid measure.

Main biases

BiasWhat happensDefence
Selection biasControls are not from the same source as cases (e.g. hospital controls with an exposure-related illness)Choose controls carefully; use several groups
Recall biasMothers of ill children search their memories harder than controlsUse records, biomarkers, blinded interviewers, questions on neutral exposures
Interviewer biasProbing cases more than controlsStructured questionnaires, blinding
TemporalityDisease changes exposure (low appetite after early disease)Define exposure window before diagnosis
ConfoundingThird factor linked to bothMatching, restriction, multivariable analysis

Variations

  • Nested case-control: cases and matched controls drawn from within a cohort; exposure stored blood measured only for them: cheaper and avoids recall bias.
  • Case-cohort, case-crossover (each case acts as own control for transient exposures such as a road accident) and test-negative designs (vaccine effectiveness: compare vaccinated proportion among test-positive and test-negative patients).

Reporting

Use the STROBE checklist for case-control studies. State how cases and controls were identified, the response rates in each, the matching and whether adjusted ORs were used.

Summary

  • Case-control studies start with cases and controls and compare past exposure.

  • They suit rare diseases, long latency, outbreaks and many exposures.

  • Controls must come from the same source population as the cases.

  • The odds ratio is the valid measure; it approximates the risk ratio when disease is rare.

  • Recall and selection bias are the main threats; nested and test-negative designs reduce them.

Check your understanding

Question 1
1

Which question favours a case-control design?

Question 2
2

Controls should be:

Question 3
3

Exposure: 80 of 100 cases, 50 of 100 controls. The odds ratio is:

Question 4
4

Mothers of children with birth defects recall exposures more thoroughly than controls. This is:

Question 5
5

A case-control study CANNOT directly estimate:

Question 6
6

A nested case-control study is conducted:

Draft study guide. Spotted an error or something unclear? Tell us at hello@scintria.com. This is educational material, not clinical or regulatory advice.

The landscape of designs, ecological and cross-sectional studiesCohort studies