Nested case-control study
Why this level?
Specify the research question: intervention, diagnostic accuracy, prognosis, aetiology, or screening intervention. Other question types are outside this hierarchy. NHMRC 2009 sourceCases and controls are sampled from inside a defined cohort, so the sampling frame is known rather than assumed. That fixes a real weakness of case-control sampling without changing what the design can support.
Where the NHMRC 2009 hierarchy names this design, it places it here:
| Research question | Level | When |
|---|---|---|
| Intervention | III-2 | An observational comparison uses concurrent controls. Statistical adjustment does not change this design level. |
| Aetiology | III-3 | A case-control study examines the exposure in people selected by outcome status. |
What question does this help answer?
The same question the full cohort would answer, at a fraction of the measurement cost, when the exposure is expensive to assess and the outcome is uncommon.
When to use it
Use it when a cohort already exists with stored samples or records, when measuring the exposure on everyone would be prohibitive, and when the outcome is rare enough that most of the cohort contributes little information.
When not to use it
There is little reason to use it when the exposure is already measured on the whole cohort: analysing everyone is simpler and more efficient. It is also a poor fit when the cohort itself is not well defined.
How it works in plain language
When a case occurs, controls are sampled from the people still at risk AT THAT MOMENT. That is what makes the resulting odds ratio estimate the cohort's rate ratio rather than something else. The design inherits the cohort's definition of the source population, which is the main advantage over an ordinary case-control study.
What risk-set sampling looks like
Why Mallard may recommend it
Mallard may recommend it when the investigator describes an existing cohort or biobank with stored samples and an expensive assay.
Key assumptions and design requirements
- Controls are sampled from the RISK SET at each case's event time, not from those who never became cases.
- The cohort and its follow-up are well defined, since the design inherits both.
- Matching factors, if used, are carried into a conditional analysis.
- The sampling ratio is stated, and the analysis accounts for it.
- Exposure is assessed blind to case status where the measurement involves judgement.
How it is analysed
Each case and the controls sampled from its risk set form a matched set, and the analysis is conditional logistic regression on those sets, which compares the case's exposure with the controls' within each set. Because controls were sampled from people still at risk at the case's event time, the resulting odds ratio estimates the hazard ratio the full cohort would have given, without needing the outcome to be rare. A person can be a control for several cases and can later become a case; under risk-set sampling both are correct, not errors to be removed. Analysing the sets with ordinary logistic regression breaks the time matching and no longer estimates the hazard ratio.
What to watch out for
The efficiency comes from not measuring the expensive exposure on everyone, not from a different question: it estimates what the full cohort would have. Controls must be sampled from those still at risk AT THE TIME each case occurred, and sampling instead from people who never became cases changes the estimand from a rate ratio to an odds ratio of a different quantity.
Clinical example
Question: Does a biomarker measured at admission predict later sepsis? Design: Within an existing admissions cohort, assay stored samples for all sepsis cases and for a sample of patients still sepsis-free at each case's time, rather than assaying every admission.
How to report it
No reporting guideline dedicated to nested case-control designs was found in a PubMed search on 2026-09-16. STROBE is the guideline to use. Its case-control items ask only in general terms where the cases came from and how controls were selected, so name outright what matters most here: the cohort the cases came from, the risk-set sampling rule, and the sampling fraction the analysis relies on. A reporting guideline is not a design manual: use it to make the study transparent, not as a substitute for careful design. Check the target journal's own requirements as well, which this page cannot know.
- Observational studies: cohort and case-control studies (opens in a new tab) · Plast Reconstr Surg · doi 10.1097/PRS.0b013e3181f44abc
- Observational research methods. Research design II: cohort, cross sectional, and case-control studies (opens in a new tab) · Emerg Med J · doi 10.1136/emj.20.1.54
- Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies (opens in a new tab) · BMJ · doi 10.1136/bmj.39335.541782.AD