A retrospective cohort study examines the association between oral contraceptive (OC) use and venous thromboembolism (VTE) in 2,000 Indian women aged 20–45 years. The crude relative risk is 3.5 (95% CI 2.1–5.8). Researchers then stratify the analysis by age group and find: in women <30 years, RR = 2.1 (95% CI 1.2–3.7); in women ≥30 years, RR = 5.8 (95% CI 3.2–10.5). The association is consistently positive in both strata, but the magnitude differs substantially. When the researchers calculate the Mantel-Haenszel adjusted RR controlling for age, it is 3.2 (95% CI 2.0–5.1), which is very close to the crude RR. Which of the following best explains these findings?

Explanation

## Identification of Effect Modification **Key Point:** Effect modification (interaction) occurs when the magnitude or direction of an association between an exposure and outcome differs across strata of a third variable. The stratified RRs are substantially different, while the adjusted RR remains close to the crude RR. ### Analysis of the Findings **Crude RR:** 3.5 (95% CI 2.1–5.8) **Stratified RRs:** - Women <30 years: RR = 2.1 - Women ≥30 years: RR = 5.8 **Mantel-Haenszel adjusted RR:** 3.2 (95% CI 2.0–5.1) ### Why This Is Effect Modification, Not Confounding | Feature | Confounding | Effect Modification | |---------|-------------|---------------------| | **Definition** | Third variable distorts the true association | Third variable changes the strength/direction of association across strata | | **Stratified RRs** | Similar across strata | Substantially different across strata | | **Adjusted RR vs. Crude RR** | Substantially different | Similar to crude RR | | **Statistical test** | Mantel-Haenszel test | Test for homogeneity (Breslow-Day test) | | **Action** | Control/adjust for the variable | Report stratified results; do NOT pool | **High-Yield:** The key diagnostic feature here is that stratified RRs (2.1 and 5.8) are very different from each other, but the adjusted RR (3.2) is close to the crude RR (3.5). This pattern is pathognomonic for effect modification. ### Interpretation OC use carries a lower VTE risk in younger women (RR = 2.1) but a much higher risk in older women (RR = 5.8). Age modifies the effect of OC on VTE risk, likely because: - Older women have more baseline thrombotic risk factors (hypercoagulability, reduced fibrinolysis) - OC-induced thrombotic risk is amplified in this higher-risk group **Clinical Pearl:** In pharmacoepidemiology, age is a common effect modifier. The same drug may have different safety profiles in different age groups, which is why stratified safety analyses are essential in post-marketing surveillance. ### Why NOT the Other Options **Confounding** would show: - Stratified RRs similar to each other - Adjusted RR substantially different from crude RR (>10% change) - Here, the adjusted RR (3.2) is very close to the crude RR (3.5), ruling out confounding **Mediation** would require age to lie in the causal pathway (OC → age → VTE), which is biologically implausible; age is a fixed characteristic, not caused by OC use. **Collider bias** occurs when you control for a variable that is caused by both the exposure and outcome. Age is not caused by either OC use or VTE, so it cannot be a collider.