Regression with incomplete covariates and left-truncated time-to-event data.

Area of Science:

• Biostatistics
• Epidemiology
• Clinical Trials

Background:

• Chronic disease studies often involve left-truncated event time data, where individuals are selected based on surviving to a certain point.
• Incomplete covariate data is common in these studies, but standard methods fail to account for left truncation's effect on covariate distributions.

Purpose of the Study:

• To develop a statistical method to handle incomplete covariate data in the presence of left-truncated event time distributions.
• To extend this method for subgroup analyses in clinical trials with incompletely observed stratification variables.

Main Methods:

• An expectation-maximization (EM) algorithm is proposed to address incomplete covariate data.
• The algorithm utilizes the covariate distribution conditional on the selection criterion (left truncation).
• An extension is described for handling incompletely observed stratification variables in subgroup analyses.

Main Results:

• The developed expectation-maximization algorithm effectively handles incomplete covariate data in left-truncated samples.
• The extension allows for robust subgroup analyses even when stratification variables are incompletely observed.

Conclusions:

• The proposed expectation-maximization algorithm provides a robust solution for analyzing chronic disease data with left-truncated distributions and incomplete covariates.
• This method enhances the reliability of risk factor assessment and subgroup analyses in epidemiological and clinical research.