In a study of coffee intake and pancreatic cancer, the crude odds ratio is 2.0. When the data are stratified by smoking status, the odds ratio is 3.0 among smokers and 1.0 among non-smokers, and both stratum-specific estimates differ clearly from the crude value. The correct conclusion is:
- A Smoking is an effect modifier, and the association should be reported separately for each stratum ✓
- B Smoking is a confounder and should be controlled by pooling
- C The difference is due to random error, and the crude estimate should be reported
- D This represents overmatching of the study subjects
Correct answer: A. Smoking is an effect modifier, and the association should be reported separately for each stratum
Explanation
When stratum-specific measures of association differ materially from each other and from the crude estimate, the third variable is an effect modifier rather than a confounder. Pooling into a single adjusted estimate would obscure genuinely different effects in smokers versus non-smokers, so results should be reported stratified with an interaction term tested formally. In pure confounding, the stratum-specific estimates are similar to each other but differ from the crude estimate, making option B incorrect here.
Reference: Gordis Epidemiology, 6th ed.
High-yield for: NEET PGINI-CETNExTFMGEUSMLEPLABMRCP
Written and medically reviewed by the StethoPrep medical team.