A retrospective cohort study basically is classifying people based on exposure status. The disease and exposure have already occurred. So we can also classify people based on cases and look at whether they were exposed or not. Thus can we change a retrospective cohort study into a case-control study?

  • $\begingroup$ Usually retrospective cohort studies are used when you can actually quantify the population at risk. If you don't mind me asking why would you want to reduce your data to case-control data, efficiency? $\endgroup$
    – scottyaz
    Feb 27, 2012 at 17:10
  • $\begingroup$ @scottyaz As mentioned in my answer, the usual reason I've found is an extremely expensive lab assay, exposure measurement, etc. that would be impractical to do on the full cohort. $\endgroup$
    – Fomite
    Feb 27, 2012 at 17:24

2 Answers 2


The short answer is yes. There is a view that essentially all observational epidemiology studies are nested within a hypothetical cohort study - the question is whether or not that cohort was ever recorded.

Given they're nested, that implies that any cohort study could be converted into a case-control study, albeit with the loss of some nice properties of a cohort. The "nested case-control study" is a fairly common study design choice when one wants to look at particularly expensive exposure measurements, and the cost-efficiency of a case-control study outweighs the loss of some information.


If you are interested in the association between the exposure and the outcome, then the sampling design is of a little interest because odds ratio is unbiased under different sampling scheme, whether it is a cohort or case-control study.
However, if you are particularly interested in knowing the risk or a rate of a disease given an exposure status, this is a different story. You cannot obtain these in a case-control study but only in cohort study and I assume this is why you are interested in "converting" a case-control study into a cohort study.
To do so, you need the probability of sampling a case and control. This is based on a simple rule of probability, Bayes rule. As a rough explanation, in case control study, you are oversampling the case compared to the control so, in your calculation of your risk/rate estimate, you need to somehow adjust for some weights such as the probability of sampling a case or a control.
That said, I disagree with Fomite regarding the nested case-control study example because nested case-control is still a case-control study, that is, the case is oversampled compared to the control.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.