Which method involves estimating effects within levels of a confounder to assess consistency of the exposure effect?

• A. Randomization.
• B. Stratification by confounders to estimate stratum-specific effects.
• C. Pooled regression without stratification.
• D. Multivariate adjustment only.

Answer: (B)

Estimating effects within levels of a confounder checks whether the exposure effect remains similar across different subgroups. By stratifying the data and calculating the exposure effect in each stratum (for example, separating by age, smoking status, or another confounder), you can see if the association holds consistently or varies by group. If the stratum-specific estimates are similar, it supports that the observed relationship isn’t due to that confounder and suggests a robust, credible effect. If the estimates differ across strata, it may indicate effect modification or residual confounding, signaling the need for stratified reporting or further adjustment.

This approach is preferred here because it directly reveals consistency of the exposure effect across confounder levels. Randomization helps prevent confounding at the design stage but isn’t about estimating within confounder levels in observational data. Pooled regression without stratification can mask differences between strata, and multivariate adjustment, while controlling for confounding, doesn't show stratum-specific effects to assess consistency.