A study examines the association between coffee consumption and myocardial infarction. The researchers find a positive association, but after stratification by smoking status, the association disappears in both strata. This pattern is most consistent with which type of confounding variable?

Explanation

## Identifying Confounding Through Stratification ### The Scenario - **Crude association:** Coffee consumption → ↑ MI risk (positive association) - **Stratified analysis:** Within each stratum of smoking (smokers and non-smokers), the association disappears - **Interpretation:** The original association was spurious; smoking explained it away **Key Point:** When stratification **eliminates or substantially reduces** an association that was present in the crude analysis, the stratifying variable is a **confounder**. ### Why Smoking is a Confounder 1. **Associated with exposure (coffee):** Smokers may consume more coffee (or the relationship may be reversed) 2. **Associated with outcome (MI):** Smoking is a strong independent risk factor for myocardial infarction 3. **Not on causal pathway:** Smoking does not mediate the coffee–MI relationship; it is an independent risk factor 4. **Stratification removes the association:** The crude association was confounded by smoking ### Mnemonic: CARS for Confounder Criteria - **C**orrelated with the **E**xposure - **A**ssociated with the **O**utcome - **R**ot on the causal pathway (not a mediator) - **S**tratification removes the crude association ### Contrast with Other Variables | Variable Type | Definition | Effect of Stratification | |---------------|-----------|-------------------------| | **Confounder** | Associated with both exposure and outcome; not on causal pathway | Crude association disappears or is substantially reduced | | **Effect Modifier** | Changes the magnitude or direction of the association between exposure and outcome | Stratified estimates differ from each other and from crude estimate | | **Mediator** | Lies on the causal pathway between exposure and outcome | Stratification reduces but does not eliminate the association | | **Collider** | Common effect of both exposure and outcome | Stratification creates or reverses an association (collider bias) | **High-Yield:** The **disappearance of an association upon stratification** is the hallmark of confounding. If the association persists but changes magnitude, think **effect modification**. **Clinical Pearl:** In the coffee–MI example, the positive crude association likely arose because smokers tend to drink more coffee AND have higher MI risk independently. Once you account for smoking, the coffee effect vanishes.