A New Method for Imputing Censored Values in Crossover Designs with Time-to-Event Outcomes Using Median Residual Life

Area of Science:

• Biostatistics
• Clinical Trial Design
• Survival Analysis

Background:

• Crossover designs offer efficiency and subject parsimony over parallel studies.
• Time-to-event outcomes in crossover designs present analytical challenges due to censored data.
• Standard regression models are limited when censoring is ignored, reducing statistical power.

Purpose of the Study:

• To propose and evaluate a novel imputation method for censored time-to-event data within crossover designs.
• To enhance the statistical power of analyses by accurately handling censored observations.
• To provide a practical approach for analyzing complex clinical trial data.

Main Methods:

• Median residual life regression for imputing censored observations.
• Analysis of Covariance (ANCOVA) incorporating period-specific baseline differences as a covariate.
• Simulation studies to compare the proposed method against multiple imputation with model averaging and ANCOVA (MI^MI).

Main Results:

• The proposed imputation method using median residual life regression showed favorable performance in simulations.
• The method effectively handles censored data, leading to improved statistical power.
• Demonstrated advantages over the MI^MI approach in analyzing time-to-event outcomes in crossover designs.

Conclusions:

• The proposed median residual life regression imputation method is a valuable tool for analyzing time-to-event outcomes in crossover designs.
• This approach offers a practical and statistically sound alternative for handling censored data in such studies.
• Further application and validation in real-world clinical trial data are recommended.