How can effect modification be detected in analyses?

• A. By checking for balance of baseline characteristics.
• B. By performing a crude analysis.
• C. By observing that the effect of exposure on outcome differs across strata.
• D. By planning longer follow-up.

Answer: (C)

Detecting effect modification means looking for a change in how an exposure affects an outcome across levels of another variable. The telltale sign is that the measure of association (like a risk ratio, odds ratio, or hazard ratio) differs when you examine subgroups defined by that third variable. In other words, the exposure’s impact isn’t uniform across strata, which indicates the modifier changes the effect.

You detect this by estimating the exposure effect separately within each stratum (stratified analysis) or by including an interaction term in a regression model and testing whether it’s significant. If the effect is stronger in one stratum and weaker or reversed in another, that’s evidence of effect modification. If the effect stays roughly the same across all strata, there isn’t evidence of modification.

Why the other ideas don’t fit: checking baseline balance is about confounding control, not about whether the exposure effect varies by another variable; a crude analysis averages effects across all groups and can mask any modification; and planning longer follow-up changes timing or precision but not the existence of a different effect by a modifier.