Parametric analysis for matched pair survival data.

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Matched pairs survival data present unique analytical challenges.
• Censoring in either or both components of a pair requires specialized statistical methods.
• Covariate analysis in paired data necessitates accounting for within-pair correlation.

Purpose of the Study:

• To apply Hougaard's bivariate Weibull distribution with positive stable frailties to matched pairs survival data.
• To analyze the impact of covariates on survival times in the presence of censoring.
• To compare the efficiency of a random effects model against a marginal (independence working) model.

Main Methods:

• Utilizing Hougaard's bivariate Weibull distribution for correlated survival times.
• Incorporating positive stable frailties to model heterogeneity.
• Applying fixed-effects and marginal (independence working) models for comparison.
• Estimating regression coefficients and quantifying Fisher information gain.

Main Results:

• The random effects model offers a simple algebraic form with appropriate parameterization.
• The independence working model performs poorly when within-pair correlation and covariate variability ratios are high.
• The fixed-effects analysis captures more information under high correlation and specific covariate variability conditions.
• The choice of model significantly impacts information capture based on correlation and covariate variability.

Conclusions:

• Hougaard's bivariate Weibull model provides a flexible framework for matched pairs survival data with censoring.
• The random effects approach is advantageous in scenarios with high correlation and specific covariate distributions.
• The study highlights the importance of accounting for correlation in survival analysis.
• Extensions to Generalized Estimation Equation methodology are indicated.