A flexible-hazards cure model with application to patients with soft tissue sarcoma

Area of Science:

• Biostatistics
• Medical Research Methodology
• Survival Analysis

Background:

• Accurate estimation of cure rates is crucial in medical research for evaluating treatments and patient groups.
• Traditional methods like the Kaplan-Meier estimator face limitations with sample size and dimensionality, especially when proportional hazards assumptions are violated.
• Existing regression models for cure rates often rely on the proportional hazards (PH) assumption, leading to biased estimations when violated.

Purpose of the Study:

• To develop a novel cure model that can simultaneously handle both proportional hazards (PH) and non-PH scenarios for various covariates.
• To provide a more flexible and robust approach for estimating cure rates in medical research, particularly when PH assumptions do not hold.
• To offer trustworthy estimations of cure rates for different treatment and demographic subgroups.

Main Methods:

• A new cure model is proposed to incorporate both PH and non-PH scenarios.
• A stable and implementable iterative procedure is developed for parameter estimation via nonparametric likelihood function maximization.
• Covariance matrix estimation is enhanced by incorporating perturbation weights into the procedure.

Main Results:

• Simulation studies demonstrate that the proposed method yields unbiased estimations for regression coefficients, survival curves, and covariate-specific cure rates.
• Existing models showed bias in simulation studies, highlighting the limitations of current approaches.
• The model was successfully applied to stage III soft tissue sarcoma data, providing reliable cure rate estimations.

Conclusions:

• The proposed flexible cure model offers significant advantages over existing methods, especially when the proportional hazards assumption is violated.
• The method provides unbiased and trustworthy estimations of cure rates, enhancing the evaluation of treatment efficacy and patient outcomes.
• This approach is valuable for medical research requiring accurate cure rate estimations across diverse patient subgroups and treatment strategies.