Modeling clustered, discrete, or grouped time survival data with covariates

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Analyzing time-to-event data with correlated observations is challenging.
• Existing methods may not adequately address grouped or clustered survival data.
• Right-censored data requires specialized statistical approaches.

Purpose of the Study:

• To present novel methods for modeling discrete or grouped time, right-censored survival data.
• To provide a framework for analyzing data from correlated groups or clusters.
• To illustrate the application of these methods using clinical trial data.

Main Methods:

• Utilized a linear log odds survival model for marginal hazard.
• Employed a gamma frailty model to capture the dependence structure within clusters.
• Incorporated cluster-level covariates to model dependence.
• Derived likelihood equations for parameter estimation.
• Described generalized estimating equations and pseudolikelihood methods.

Main Results:

• Developed and described robust statistical methods for clustered survival data.
• Demonstrated the estimation of marginal hazard regression and dependence parameters.
• Successfully applied the methods to real-world data from two clinical trials.

Conclusions:

• The proposed methods offer a flexible and effective approach for analyzing discrete, grouped, and right-censored survival data with correlation.
• The gamma frailty model effectively captures within-cluster dependence.
• These methods enhance the statistical rigor in analyzing complex clinical trial data.