Correlated gamma frailty models for bivariate survival time data

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Frailty models are essential for analyzing multivariate time-to-event data, quantifying heterogeneity and association.
• Existing correlated gamma frailty models have limitations, particularly in handling differing frailty variances.
• Previous additive decomposition models offer closed-form solutions but impose upper bounds on correlation.

Purpose of the Study:

• To review and propose novel correlated gamma frailty models based on bivariate gamma frailty distributions.
• To address the limitations of existing models, especially regarding correlation bounds.
• To provide a more flexible framework for analyzing bivariate survival data.

Main Methods:

• Review of existing correlated gamma frailty models.
• Development of novel bivariate gamma frailty models.
• Application of frailty methodology to right-censored and left-truncated Danish twins mortality data.
• Analysis of serological survey data on viral infections using current status data.

Main Results:

• Proposed bivariate gamma frailty models offer greater flexibility in association and correlation structures.
• These novel models outperform existing frailty models, including those with additive decomposition.
• The methodology is effective for various censoring types in bivariate survival analysis.

Conclusions:

• Flexible correlated gamma frailty models are crucial for accurate bivariate survival data analysis.
• The proposed models overcome limitations of previous approaches, particularly with heterogeneous frailty variances.
• The enhanced frailty methodology provides a robust tool for epidemiological and biostatistical research.