Case-cohort 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 sourceA random subcohort serves as the comparison group for every outcome studied. It sits where the cohort it is drawn from sits.
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 | II | A prospective cohort examines an exposure that cannot be experimentally allocated. |
| Aetiology | III-2 | A retrospective cohort examines an exposure that cannot be experimentally allocated. |
What question does this help answer?
The same cohort question, with one important practical advantage: because the subcohort is sampled at baseline and not tied to any particular outcome, the SAME subcohort can serve several different outcomes.
When to use it
Use it when several outcomes will be studied in one cohort with one expensive exposure measurement, or when a baseline-sampled comparison group is administratively simpler than risk-set sampling.
When not to use it
Prefer a nested case-control design when only one outcome is of interest and the analysis needs to be simple, because the case-cohort variance is more complicated.
How it works in plain language
A random subcohort is drawn at baseline, and every case in the whole cohort is included whether or not it fell in the subcohort. Some subcohort members later become cases, so the two groups OVERLAP, and the analysis must use weights or a method built for the design rather than treating it as an ordinary case-control study.
What the subcohort and the cases look like
Why Mallard may recommend it
Mallard may recommend it when the investigator describes one costly measurement and several planned outcomes in a single cohort.
Key assumptions and design requirements
- The subcohort is a genuine random sample of the cohort at baseline.
- The overlap between cases and subcohort is handled with an appropriate weighted analysis.
- The variance estimator accounts for the sampling, not just the model.
- Follow-up and the risk set are defined the same way for both groups.
- If several outcomes reuse the subcohort, the multiplicity that creates is acknowledged.
What to watch out for
The subcohort contains some future cases, so the analysis needs weights or a method that accounts for the overlap; treating it as an ordinary case-control biases the result. Its advantage over a nested design is that one subcohort serves several outcomes, and its cost is a more complicated variance.
Clinical example
Question: Which of three inflammatory markers predicts three different complications? Design: Assay a random 10 percent subcohort plus every case of each complication, and analyse each outcome with a weighted Cox model using the same subcohort.
How to report it
No reporting guideline dedicated to case-cohort designs was found in a PubMed search on 2026-09-16. STROBE is the guideline to use. State separately how the subcohort was sampled, its size relative to the full cohort, and the weighting the analysis applied, none of which a STROBE item asks for by name. 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
- Overview of clinical study designs (opens in a new tab) · Clin Exp Emerg Med · doi 10.15441/ceem.23.036
- 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