Validation study methods for estimating odds ratio in 2 × 2 × J tables when exposure is misclassified

Area of Science:

• Epidemiology
• Biostatistics

Background:

• Exposure misclassification in epidemiologic studies can bias results and reduce statistical power.
• The inverse matrix method is an efficient approach to correct odds ratios for binary exposure misclassification.

Purpose of the Study:

• To generalize the inverse matrix method for nondifferential misclassification and 2x2xJ tables.
• To compare the efficiency of the inverse matrix method against other statistical approaches.

Main Methods:

• The inverse matrix method was extended to handle nondifferential misclassification in 2x2xJ tables.
• Estimates for predictive values were used to calculate corrected log odds ratios and their variances.
• A two-step weighted likelihood method was developed.
• Simulation studies compared the inverse matrix method with the matrix and maximum likelihood (MLE) methods.

Main Results:

• The inverse matrix method demonstrated higher efficiency compared to the matrix method across all scenarios.
• Both matrix and inverse matrix methods were more efficient for nondifferential than differential misclassification.
• The inverse matrix method showed acceptable efficiency, especially compared to the computationally intensive MLE method.

Conclusions:

• While maximum likelihood estimation (MLE) is statistically optimal, it poses computational challenges.
• The inverse matrix method provides a practical alternative with a simple closed-form solution and acceptable efficiency.
• The inverse matrix method is a valuable tool for correcting exposure misclassification in complex epidemiologic data.