A researcher studying the effect of occupational pesticide exposure on Parkinson's disease controls for age, gender, and smoking in a multivariable logistic regression. After adjustment, the OR changes from 3.8 (crude) to 2.1 (adjusted). This change is BEST explained by:
- A Effect modification by age, gender, or smoking
- B Confounding by age, gender, and/or smoking that was removed by adjustment ✓
- C Overfitting of the regression model
- D Regression dilution bias from measurement error
Explanation
When the crude OR is substantially larger than the adjusted OR after controlling for age, gender, and smoking, confounding by these variables is the explanation — they were positively associated with both pesticide exposure and Parkinson's risk, inflating the crude estimate. Effect modification (interaction) would manifest as different ORs across subgroups of the confounders, not a uniform attenuation after adjustment. Overfitting occurs with too many predictors relative to events. Regression dilution bias reduces estimates toward the null but is caused by measurement error in the exposure, not covariate adjustment.
Reference: Park's Textbook of Preventive and Social Medicine, 27th ed.
High-yield for: NEET PGINI-CETNExTFMGEUSMLEPLABMRCP
Written and medically reviewed by the StethoPrep medical team.