Individual patient data meta-analysis of survival data using Poisson regression models

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Survival Analysis

Background:

• Individual Patient Data (IPD) meta-analysis is the gold-standard for survival data synthesis.
• Existing one-stage hierarchical Cox models are computationally intensive and not universally available.
• A novel approach using Poisson-based Generalized Linear Models (GLMs) is proposed as an alternative.

Purpose of the Study:

• To present and evaluate a computationally efficient Poisson GLM approach for IPD meta-analysis.
• To demonstrate the utility of this method in analyzing survival data from clinical trials.
• To extend the framework for modeling treatment-covariate interactions and non-proportional hazards.

Main Methods:

• Application and simulation of Poisson GLM in classical and Bayesian frameworks.
• Comparison of one-stage and two-stage approaches.
• Utilized data from ten hypertension treatment trials with all-cause death as the outcome.

Main Results:

• Poisson approach yielded estimates nearly identical to Cox models.
• The proposed method is computationally efficient and directly estimates the baseline hazard.
• Bayesian approach mitigated downward bias in classical heterogeneity estimates.

Conclusions:

• The Poisson GLM framework is flexible, computationally efficient, and available in standard software.
• Enables investigation of heterogeneity, non-proportional hazards, and treatment effect modifiers.
• Offers a practical alternative for survival data meta-analysis.