A bivariate survival model with compound Poisson frailty

Area of Science:

• Biostatistics
• Epidemiology
• Survival Analysis

Background:

• Traditional survival analysis often assumes universal susceptibility to an event.
• Bivariate time-to-event data requires specialized modeling techniques.
• Existing correlated frailty models may not account for non-susceptible populations.

Purpose of the Study:

• To propose a novel correlated frailty model for bivariate time-to-event data.
• To extend the compound Poisson frailty model to a bivariate setting.
• To incorporate a non-susceptible fraction into the population.

Main Methods:

• Development of a bivariate correlated frailty model.
• Extension of the compound Poisson frailty model.
• Maximum likelihood estimation for model parameters.
• Simulation studies to assess parameter estimation properties.

Main Results:

• The proposed model accommodates a non-susceptible fraction, addressing a limitation in standard survival analysis.
• The model includes correlated gamma and inverse Gaussian frailty models as special cases.
• Application to Swedish twin data estimated a 15% breast cancer susceptibility rate.

Conclusions:

• The new correlated frailty model provides a flexible framework for bivariate survival data with non-susceptible individuals.
• The model offers valuable insights into population susceptibility for diseases like breast cancer.
• This approach enhances the analysis of time-to-event data in epidemiological studies.