In a case-control study, investigators individually match each oral cancer case to a control on the basis of tobacco chewing status. After analysis, they find they cannot estimate the independent effect of tobacco chewing on oral cancer risk, and the confidence interval around the adjusted odds ratio for alcohol use has widened compared with unmatched analysis. These consequences are best explained as:

• A Matching on a confounder, which improved precision
• B Recall bias introduced by the matching procedure
• C Overmatching, because the matched factor was strongly linked to the exposure, removing its estimable effect and sacrificing power ✓
• D Simpson's paradox due to unadjusted strata

Explanation

When a matched variable is closely tied to the exposure of interest, matching no longer controls a true confounder but forces similarity on a factor that partly defines exposure, a situation termed overmatching. It renders the matched variable's effect unestimable and can dilute or widen estimates for other factors, reducing statistical efficiency. Correctly matching on a genuine confounder improves precision, recall bias stems from differential memory, and Simpson's paradox describes reversal of associations after pooling strata.

Reference: Modern Epidemiology (Rothman), 4th ed.

High-yield for: NEET PGINI-CETNExTFMGEUSMLEPLABMRCP

Written and medically reviewed by the StethoPrep medical team.