Collaborative Inference for Accelerated Failure Time Model Using Clinical Center-Level Summary Statistics

Area of Science:

• Biostatistics
• Clinical Research Methodology
• Survival Analysis

Background:

• Multi-center clinical research yields larger sample sizes and more generalizable findings.
• Existing methods for survival data analysis may have limitations in integrative analyses across multiple sites.
• The Accelerated Failure Time (AFT) model offers an alternative to the Cox proportional hazards model for time-to-event data.

Purpose of the Study:

• To develop a collaborative analytic framework for survival data analysis using summary statistics.
• To implement a distributed inference method based on parametric Accelerated Failure Time (AFT) models.
• To assess the goodness-of-fit for different parametric AFT models using a distributed likelihood ratio test.

Main Methods:

• Developed a collaborative framework utilizing summary statistics for survival data analysis.
• Employed parametric AFT models (Weibull, log-normal, log-logistic) for time-to-event outcomes.
• Established a distributed likelihood ratio test under the generalized gamma distribution for model assessment.

Main Results:

• The proposed distributed inference method demonstrates robust performance.
• Large-sample properties of the distributed method were established.
• The framework was validated through simulations and a real-world kidney transplantation dataset.

Conclusions:

• The developed framework facilitates flexible and robust integrative analyses of multi-center survival data.
• The AFT model and distributed inference approach offer advantages over traditional methods for multi-site studies.
• This methodology enhances the reliability and generalizability of findings from collaborative clinical research.