I noticed from several articles saying that case-controlled study is more bias prone than cohort study (in the context of cancer early detection using molecular marker). I spent sometime to try to rationalize the idea (I guess it has something to do with case-controlled study having more risk of carrying confounder?) and could not find the logics. Can someone some comments on this?
the most important source of bias unique to case referent studies is recall bias when exposure history is obtained by interview. Cases may be more likely to recall an exposure once they have a diagnosis.
this bias does not always occur eg if exposures are obtained from records (which would be your case)
cohort studies have biases of their own in non random follow up loss.
Confounders occur in both case referent and cohort studies and are controlled by similar methods. eg randomisation, restriction,stratification, matching, statistical modelling.
Retrospective studies do not necessarily need to rely on data prone to recall-bias. This may be the case for both case-control studies and retrospective cohort studies. Exposure data can be from surveys completed in the past, or register-based data. What you call a retrospective cohort study, I would call a historical cohort study. It's a bit tricky, because in the historical cohort study you look back in time, but investigate a prospective association between an exposure and outcome.
A historical cohort study has the advantage that it is possibly to do time-to-event analysis, on a given event (e.g. diagnosis or mortality), without waiting many years to have enough events in your analyses. An advantage of all retrospective studies, case-control and cohort studies, is that it is possible to investigate exposures related to rare diseases, by using data from e.g. national registers. Thus, a well-conducted case-control study may be as methodological sound as a cohort study.
- Badges
- Science topic
- Similar topics
- Comment