Estimating the mean hazard ratio parameters for clustered survival data with random clusters

Area of Science:

• Biostatistics
• Clinical Trials Methodology
• Survival Analysis

Background:

• Clustered survival data present unique analytical challenges, particularly in large multi-centre clinical trials.
• Existing models may not adequately capture the complex interplay between random cluster effects and covariates.

Purpose of the Study:

• To develop and validate a latent variable hazard model for clustered survival data.
• To estimate mean hazard ratio parameters and their variances, especially in multi-centre clinical trials.
• To assess the mean treatment effect by treating participating centres as a random sample.

Main Methods:

• Utilizing a maximum pseudo-likelihood estimator for parameter estimation.
• Implementing a bootstrap sampling scheme for variance estimation.
• Applying the model to clustered survival data with random cluster effects and covariate interactions.

Main Results:

• The proposed estimators demonstrate reliable performance in simulation studies.
• The method effectively assesses mean treatment effects in the context of multi-centre trials.
• The model successfully handles interactions between random cluster effects and covariates.

Conclusions:

• The developed latent variable hazard model provides a robust framework for analyzing clustered survival data.
• This approach enhances the assessment of treatment effects in large-scale, multi-centre clinical research.
• The method is validated through simulations and a real-world data example from the SOLVD Prevention Trial.