Analysing multicentre competing risks data with a mixed proportional hazards model for the subdistribution

Area of Science:

• Biostatistics
• Survival Analysis
• Epidemiology

Background:

• The Fine and Gray model analyzes competing risks but assumes data independence.
• Clustered data, common in registries and trials, violates this independence assumption.
• Frailty models handle clustered data in single-risk survival but haven't been extended to subdistribution hazards.

Purpose of the Study:

• To propose and evaluate a frailty model for the subdistribution hazard in competing risks.
• To assess and incorporate cluster heterogeneity when testing covariate effects.
• To address limitations of existing models for dependent competing risks data.

Main Methods:

• Development of a frailty model for the subdistribution hazard.
• Simulation studies to investigate the impact of heterogeneity on covariate effect estimation.
• Application of the proposed model to a real-world registry cohort dataset.

Main Results:

• The proposed frailty model effectively assesses and incorporates cluster heterogeneity.
• Simulation results demonstrate the influence of heterogeneity on testing covariate effects.
• The model provides a robust method for analyzing dependent competing risks data.

Conclusions:

• The novel frailty model for subdistribution hazards is a valuable tool for analyzing clustered competing risks data.
• It allows for accurate assessment of covariate effects in the presence of unobserved cluster-level heterogeneity.
• This approach enhances the analysis of complex survival data from sources like patient registries.