Frailty modeling for clustered competing risks data with missing cause of failure

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Survival Analysis

Background:

• Multi-center clinical trials often yield correlated event times within centers due to unobserved factors.
• Competing risks are common, and missing cause of death information can bias traditional analyses.

Purpose of the Study:

• To develop a statistical model for analyzing competing risks data in multi-center trials with correlated event times.
• To address the challenge of missing cause of death information in such settings.
• To propose a robust method that avoids biased inferences from data exclusion.

Main Methods:

• Utilized a cause-specific proportional hazards model with a shared frailty term to account for within-center correlations.
• Employed a hierarchical likelihood approach, circumventing complex integration over frailty terms.
• Integrated multiple imputation techniques to handle missing cause of death data under the missing at random assumption.

Main Results:

• Simulation studies demonstrated the effectiveness of the proposed methods, showing good performance even with imputation model misspecification.
• The hierarchical likelihood approach successfully modeled associations between event times within centers.
• The methods provided reliable inferences despite missing cause of death information.

Conclusions:

• The proposed hierarchical likelihood approach with shared frailty is a viable method for analyzing competing risks data in multi-center trials with correlated event times.
• The approach effectively handles missing cause of death information, preventing biased results.
• This methodology offers a robust solution for complex survival data encountered in clinical research.