A cohort study examines caffeine intake and pregnancy outcomes. The pooled relative risk for miscarriage is 1.2, but stratified analysis gives a relative risk of 0.6 among women under 30 years and 2.8 among women aged 30 years or above. The most appropriate next step in reporting these results is to:

• A Report the pooled relative risk alone, as stratification has created noise
• B Conclude confounding and report the Mantel-Haenszel adjusted estimate
• C Exclude the younger women as outliers and recompute the pooled estimate
• D Report the stratum-specific estimates separately, since effect modification is present ✓

Explanation

When stratum-specific effect estimates differ meaningfully from each other and from the pooled value, effect modification is present. In this situation the single summary measure is misleading, and the correct practice is to report each stratum separately and describe how the effect varies by the modifying variable. Confounding requires the stratum-specific estimates to be similar to each other, which is not seen here, and excluding subgroups would introduce selection bias.

Reference: Modern Epidemiology (Rothman), as cited in Indian PSM teaching ed.

High-yield for: NEET PGINI-CETNExTFMGEUSMLEPLABMRCP

Written and medically reviewed by the StethoPrep medical team.