I have a study in which healthy participants enroll in the study at time T0, a subset of my entire cohort experience a single event (cases; time T1), so we decided to create a nested case-control where each case is matched with 4 controls by age, sex, location (this is how it was designed, I have no control over this) where the selection of controls is made at time T1. Then the study ended at time T3, with the controls not experiencing the event.

(each T0 is different for each individual of course)

My question is the following: when doing a stratified Cox proportional hazard analysis on this cohort, is the time to event in the analysis:

a) for cases: T1-T0; for controls: T1-T0?

b) for cases: T1-T0; for controls: T2-T0?

From what I been told, I understand that this does not compute absolute risk, but is doing stratified cox-proportional hazards valid and which time to the event should I use. I couldn't find a lot about what I should do.

Thanks a lot.


1 Answer 1


Typically in that situation you might construct a case within cohort analysis and use the non-cases as the controls analyzed from T0 to T1. I suppose you could set the study with a 4-1 match of controls to cases and analyze both groups for the period T0-T1. If you included time after T1 you will be subject to the criticism that there may have been a shift in treatment strategies, diagnostic strategies or other conditions that might affect the probability of events during the interval T1-T2. Shortening the time interval in this case will not reduce the power of test because statistical power in survival analysis comes from the numbers of events rather than the duration of the study.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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