An epidemiologist in Mumbai designs a cohort study to examine the relationship between oral contraceptive (OC) use and venous thromboembolism (VTE) in women aged 20–45 years. Over 5 years, 10,000 OC users and 10,000 non-users are followed. The crude relative risk (RR) for VTE in OC users is 3.5 (95% CI: 2.8–4.4). However, when the analysis is stratified by the presence of Factor V Leiden mutation, the RR in Factor V Leiden carriers is 8.2 and in non-carriers is 2.1. The crude RR of 3.5 lies between these two stratum-specific RRs. What epidemiological phenomenon is being demonstrated?

Explanation

## Effect Modification vs. Confounding **Key Point:** When stratum-specific estimates differ substantially and the crude estimate lies **between** them (not outside them), the stratifying variable is an **effect modifier**, not a confounder. ### Distinguishing Effect Modification from Confounding | Feature | Effect Modification | Confounding | |---------|-------------------|-------------| | **Definition** | The magnitude or direction of association differs by levels of a third variable | A third variable distorts the true association | | **Stratum-specific estimates** | Differ from each other | Similar to each other | | **Crude estimate** | Lies between strata (weighted average) | Differs from stratum-specific estimates | | **Stratification action** | Report separately (do NOT pool) | Pool or adjust (remove the distortion) | | **Biological meaning** | The third variable modulates the effect of the exposure | The third variable is a confounding variable | ### Analysis of This Study ```mermaid flowchart TD A[OC Use]:::action --> B{Factor V Leiden?}:::decision B -->|Carrier| C[RR = 8.2]:::outcome B -->|Non-carrier| D[RR = 2.1]:::outcome E[Crude RR = 3.5]:::outcome F[3.5 lies between 2.1 and 8.2]:::action ``` **High-Yield:** - **Crude RR = 3.5** (weighted average of stratum-specific RRs) - **RR in Factor V Leiden carriers = 8.2** (much higher risk) - **RR in non-carriers = 2.1** (lower risk) - The crude estimate is a **weighted average** of the two stratum-specific estimates ### Why This is Effect Modification 1. **Stratum-specific RRs differ markedly** (8.2 vs. 2.1) — this is the hallmark of effect modification 2. **The crude RR (3.5) lies between the two strata** — indicating it is a weighted average, not a distorted estimate 3. **Factor V Leiden modulates the effect** of OC on VTE risk — the genetic predisposition amplifies the thrombotic risk from OCs 4. **Biological plausibility:** Factor V Leiden is a hypercoagulable state; OCs also increase thrombotic risk; together they have a synergistic (multiplicative) effect **Clinical Pearl:** In effect modification, you report the stratum-specific estimates separately because the exposure effect is genuinely different depending on the modifier. Pooling would mask this heterogeneity and lose clinically important information. **Mnemonic:** **STRATA-DIFFER** = Effect Modification - **S**tratum-specific estimates **DIFFER** → effect modification - Report **SEPARATELY** by strata - Do **NOT** pool or adjust [cite:Park 26e Ch 9; Epidemiology: Effect Modification and Interaction]