An investigator conducts a case-control study of coffee intake and pancreatic cancer and matches controls to cases on chronic gastritis status, believing gastritis relates both to coffee drinking and to cancer risk. Subsequent analysis shows the matched odds ratio is far weaker than the crude odds ratio, and gastritis turns out to lie partly on the causal pathway between coffee and pancreatic cancer. This loss of a true association is termed:
- A Overmatching ✓
- B Recall bias
- C Confounding by indication
- D Effect modification
Correct answer: A. Overmatching
Explanation
Matching on a variable that is an intermediate step in the causal chain, or too tightly linked to the exposure, artificially equalises exposure distribution between cases and controls and drives the estimate toward the null. This is overmatching. Effect modification describes genuine variation of effect across strata, which stratified analysis should reveal rather than mask, so it cannot explain the attenuated estimate.
Reference: Gordis Epidemiology, 6th ed.
High-yield for: NEET PGINI-CETNExTFMGEUSMLEPLABMRCP
Written and medically reviewed by the StethoPrep medical team.