## Confounding in the Vitamin Supplement Study **Key Point:** This scenario exemplifies confounding because exercise and socioeconomic status are third variables that independently influence both the exposure (supplement use) and the outcome (cardiovascular disease), distorting the apparent association. ### Criteria for a Valid Confounder A variable must meet ALL three criteria to be a confounder: 1. **Associated with the exposure:** Exercise and socioeconomic status are more common in supplement users 2. **Associated with the outcome:** Both exercise and higher SES independently reduce cardiovascular disease risk 3. **Not in the causal pathway:** Exercise and SES are not caused by supplement use itself; they are independent lifestyle/demographic factors ### Analysis of the Scenario ```mermaid flowchart TD A["Socioeconomic Status<br/>(Confounder)"]:::outcome --> B["Supplement Use<br/>(Exposure)"]:::action A --> C["Lower CVD Risk<br/>(Outcome)"]:::outcome D["Regular Exercise<br/>(Confounder)"]:::outcome --> B D --> C B -.->|Apparent but spurious| C style B stroke:#ff6b6b,stroke-width:2px ``` **High-Yield:** The observed association between supplements and lower CVD is **spurious**—it results from confounding by exercise and SES, not from a true protective effect of supplements. **Clinical Pearl:** In observational studies, health-conscious individuals (who take supplements) naturally engage in other health-promoting behaviors (exercise, better diet, higher SES). These confounders must be measured and controlled to isolate the true effect of the exposure. ### How to Control This Confounding 1. **Stratification:** Compare supplement users and non-users *within* each SES stratum and exercise category 2. **Matching:** Recruit supplement users and non-users matched on exercise level and SES 3. **Multivariable regression:** Adjust for exercise and SES in a logistic regression model 4. **Randomization:** Randomize participants to supplement or placebo (eliminates confounding by design) **Mnemonic:** **"CONFOUND = CON-FOUND a third variable that distorts the true relationship"** — it's a real variable, not a measurement error. [cite:Park 26e Ch 8]
Sign up free to access AI-powered MCQ practice with detailed explanations and adaptive learning.