Simulating survival data with predefined censoring rates for proportional hazards models

Area of Science:

• Biostatistics
• Medical Statistics
• Survival Analysis

Background:

• The proportional hazard model is vital for analyzing time-to-event data in medical research.
• Simulation studies are essential for evaluating statistical models, but complex scenarios with varying censoring rates pose challenges.

Purpose of the Study:

• To propose a general framework for simulating right-censored survival data for proportional hazard models.
• To accommodate various baseline hazard functions, censoring time distributions, and covariate distributions.

Main Methods:

• Developed a framework incorporating baseline hazard (exponential, Weibull), censoring time (uniform, Weibull), and covariates.
• Utilized nested numerical integration and a root-finding algorithm to achieve predefined censoring rates.
• Assessed framework performance through simulation studies, including bias analysis for treatment effects with unmeasured confounding.

Main Results:

• The proposed framework successfully simulates right-censored survival data under diverse conditions.
• Demonstrated the framework's application in a comprehensive simulation study.
• Investigated the impact of censoring rates on bias in estimating conditional treatment effects.

Conclusions:

• The developed framework provides a robust and flexible approach for generating complex survival data.
• This enhances the evaluation of proportional hazard models and alternative methods in medical research.
• The framework aids in understanding the influence of censoring on statistical inference.