In a cohort study examining the association between occupational dust exposure and chronic obstructive pulmonary disease (COPD) among Indian textile workers, all of the following are valid strategies to control for confounding EXCEPT:

Explanation

## Controlling for Confounding in Epidemiological Studies ### Definition of Confounding **Key Point:** A confounder is a variable that: 1. Is independently associated with the exposure 2. Is independently associated with the outcome 3. Is not in the causal pathway between exposure and outcome 4. Distorts the true exposure-outcome relationship ### Valid Methods to Control Confounding | Method | Mechanism | Timing | Advantage | Limitation | |--------|-----------|--------|-----------|------------| | **Restriction** | Limit study to homogeneous subgroup (e.g., non-smokers only) | Study design phase | Simple, eliminates confounder entirely | Reduces generalizability, may lose statistical power | | **Matching** | Select controls/unexposed similar to cases/exposed on confounder | Study design phase | Increases efficiency, balances confounder distribution | Cannot match on unknown confounders, increases cost | | **Stratification** | Analyze association separately within strata of confounder | Analysis phase | Reveals effect modification, allows confounder-specific estimates | Reduces sample size in each stratum, reduces power | | **Multivariable Regression** | Adjust for confounder statistically in regression model | Analysis phase | Can control multiple confounders simultaneously | Assumes linear relationship, requires adequate sample size | ### Why Sample Size Increase Does NOT Control Confounding **High-Yield:** Increasing sample size **increases statistical power** to detect an association, but it does **NOT eliminate or control confounding bias**. A larger study with confounding will simply estimate the biased association with greater precision. **Clinical Pearl:** If a confounder creates a spurious association (or masks a true association), enrolling 10,000 workers instead of 100 will not remove that bias—it will only narrow the confidence interval around the biased estimate. **Mnemonic: RMS** — **R**estriction, **M**atching, **S**tratification are the three design/analysis methods to control confounding. Sample size is not one of them. **Warning:** Students often confuse "increasing power" with "reducing bias." These are distinct: - **Bias** = systematic error (confounding, selection bias, information bias) - **Random error** = imprecision (reduced by larger sample size) Confounding is a form of bias, not random error, so it cannot be reduced by sample size alone. [cite:Park 26e Ch 8]