A hospital-based study reports a crude odds ratio of 2.5 for coffee intake and pancreatic cancer. After adjusting for smoking in a multivariable model, the odds ratio falls to 1.4. What is the most appropriate interpretation?

• A Coffee is an independent risk factor; the association strengthened on adjustment
• B Smoking is an effect modifier, requiring stratified reporting
• C Smoking is a strong confounder of the coffee-pancreatic cancer association ✓
• D The change indicates random error rather than any systematic distortion

Explanation

Because smokers drink more coffee and smoking itself causes pancreatic cancer, the crude association mixed a real coffee effect with smoking's effect. When adjustment moves the estimate substantially, conventionally by more than about 10 to 20 percent, confounding is demonstrated. Effect modification would appear as stratum-specific estimates that differ importantly from each other, which is a different phenomenon from a uniform shift toward the null across strata. Random error is addressed by confidence intervals, not by movement on adjustment.

Reference: Gordis Epidemiology, 6th ed.

High-yield for: NEET PGINI-CETNExTFMGEUSMLEPLABMRCP

Written and medically reviewed by the StethoPrep medical team.