In a cohort study, alcohol consumption shows a crude relative risk of 2.0 for oral cancer. When the data are stratified by smoking status, the relative risk is 5.0 among smokers and 0.8 among non-smokers. The most appropriate conclusion is:
- A Smoking is a confounder and should be controlled by stratification
- B The crude estimate should be reported as the best summary of the association
- C Effect modification is present and the stratum-specific estimates should be reported separately ✓
- D The findings indicate information bias in exposure ascertainment
Correct answer: C. Effect modification is present and the stratum-specific estimates should be reported separately
Explanation
When stratum-specific estimates differ substantially from each other and from the crude estimate, effect modification (interaction) is present. Pooling would obscure the truth that alcohol raises risk sharply in smokers but not in non-smokers, so separate stratum-specific reporting is correct. In pure confounding, stratum-specific estimates are similar to each other but differ from the crude value, which is not the pattern seen here.
Reference: Gordis Epidemiology, 6th ed.
High-yield for: NEET PGINI-CETNExTFMGEUSMLEPLABMRCP
Written and medically reviewed by the StethoPrep medical team.