Expected estimating equations via EM for proportional hazards regression with covariate misclassification

Area of Science:

• Epidemiology
• Biostatistics
• Medical Statistics

Background:

• Covariate misclassification is a common issue in epidemiological and medical studies, leading to biased estimations in Cox regression.
• Adjusting for misclassification is challenging, especially when gold standard data are unavailable.
• Statistical modeling for misclassification differs significantly from measurement error modeling.

Purpose of the Study:

• To develop statistical methods to accommodate covariate misclassification in Cox regression when gold standard data are absent.
• To propose an expected estimating equation estimator using an expectation-maximization algorithm.
• To evaluate the performance of the proposed method using simulation studies and a real-world clinical trial.

Main Methods:

• Investigated an approximate induced hazard estimator.
• Proposed an expected estimating equation estimator utilizing an expectation-maximization algorithm.
• Employed multiple surrogate variables to address the unobserved latent predictor.

Main Results:

• The proposed method demonstrated effectiveness in accommodating covariate misclassification.
• Simulation studies examined the finite sample performance of the estimator.
• The method was applied to a human immunodeficiency virus clinical trial, using questionnaire data as surrogates.

Conclusions:

• The developed statistical approach provides a viable solution for covariate misclassification in the absence of gold standard data.
• The expectation-maximization algorithm effectively handles latent variables and surrogate data.
• The findings have implications for improving the accuracy of epidemiological and clinical trial analyses.