A cohort study in Delhi investigates the association between occupational dust exposure and chronic obstructive pulmonary disease (COPD) in factory workers. Researchers follow 500 exposed and 500 unexposed workers for 5 years. At baseline, 15% of exposed workers and 5% of unexposed workers already had mild COPD. After 5 years, the incidence of new COPD cases is 12% in the exposed group and 8% in the unexposed group. The crude relative risk is 1.5 (95% CI 1.1–2.0). However, when the researchers exclude workers with baseline COPD and reanalyze, the relative risk becomes 1.8 (95% CI 1.3–2.5). What does this change in relative risk indicate?

Explanation

## Understanding the RR Change on Exclusion of Baseline Cases ### The Key Observation **High-Yield:** When crude RR = 1.5, but RR after excluding baseline cases = 1.8, the relative risk **increased** (became stronger). This means baseline COPD status was **diluting** (weakening) the true association — a hallmark of **negative confounding**. ### What Is Negative Confounding? **Key Point:** Negative confounding occurs when a confounder is: 1. Associated with the exposure in one direction (e.g., more prevalent in exposed) 2. Associated with the outcome in the opposite direction (protective effect) 3. The net effect is to **reduce** (underestimate) the true association ### The Mechanism in This Study | Aspect | Finding | |--------|----------| | **Baseline COPD in exposed** | 15% (higher) | | **Baseline COPD in unexposed** | 5% (lower) | | **Effect of baseline COPD** | Reduces new COPD incidence? (workers already have disease, less room for new cases) | | **Crude RR** | 1.5 (diluted by baseline COPD) | | **RR after excluding baseline** | 1.8 (true effect, unconfounded) | **Clinical Pearl:** Baseline COPD status is a confounder because: - It is associated with exposure (more common in exposed workers, likely due to prior dust exposure) - It is associated with the outcome (workers with baseline COPD have different risk of incident COPD) - It is not on the causal pathway (baseline disease is not caused by current exposure in this analysis) ### Why Exclusion Reveals Negative Confounding When baseline COPD cases are removed: - The denominator shrinks, but the numerator (new cases) shrinks disproportionately in the exposed group - This is because exposed workers with baseline COPD had already "used up" their COPD risk - The remaining exposed workers (without baseline COPD) show a stronger association with incident COPD - Thus, RR increases from 1.5 → 1.8 **Mnemonic:** **NEGATIVE CONFOUNDING = CRUDE < TRUE** — When crude estimate is smaller than the true (unconfounded) estimate, confounding is negative (protective). ### Confounding vs. Bias | Feature | Confounding | Bias | |---------|-------------|------| | **Mechanism** | Third variable distorts association | Systematic error in measurement/selection | | **Effect of stratification** | Association changes | Association does not change | | **Effect of exclusion** | Association changes | Association does not change | | **Example here** | Baseline COPD explains part of crude RR | — | **High-Yield:** Confounding can be **positive** (inflates RR) or **negative** (deflates RR). Bias is always **systematic error** and is not corrected by stratification or exclusion of the confounder.