A New Bayesian Model For Survival Data With a Surviving Fraction.

Area of Science:

• Biostatistics
• Statistical Modeling
• Survival Analysis

Background:

• Analyzing right-censored survival data in populations with a surviving fraction (cure rate) is crucial for accurate prognostic assessment.
• Standard mixture models for cure rates have limitations, necessitating alternative statistical frameworks.

Purpose of the Study:

• To propose and investigate a novel Bayesian model for right-censored survival data with a cure fraction.
• To explore the properties and interpretability of the proposed model, contrasting it with existing mixture models.

Main Methods:

• Development of a new Bayesian statistical model for survival data incorporating a cure fraction.
• Derivation of the model's proportional hazards structure and properties of its hazard function.
• Detailed discussion on prior elicitation, proposing noninformative and informative priors.

Main Results:

• The proposed model exhibits a proportional hazards structure where covariates naturally influence the cure rate.
• Novel mathematical relationships between the proposed model and standard mixture models for cure rates were established.
• Theoretical properties of proposed priors and resulting posteriors were derived and compared to the standard mixture model.

Conclusions:

• The novel Bayesian model offers a distinct and interpretable framework for survival data with cure fractions.
• The model's proportional hazards property and derived relationships provide valuable insights for statistical inference.
• Application to a melanoma clinical trial dataset demonstrates the model's practical utility.