Smooth semi-nonparametric (SNP) estimation of the cumulative incidence function

Area of Science:

• Biostatistics
• Survival Analysis
• Competing Risks

Background:

• Estimating cumulative incidence in the presence of competing risks is crucial for accurate survival analysis.
• Existing methods may struggle with arbitrary censoring and truncation, limiting their applicability.
• A need exists for robust statistical models that accommodate complex data structures.

Purpose of the Study:

• To present a novel statistical approach for estimating the cumulative incidence function (CIF) under competing risks.
• To develop a flexible model using mixture factorization and semi-nonparametric densities.
• To provide a robust method capable of handling arbitrary censoring and truncation.

Main Methods:

• A statistical model based on mixture factorization of the joint distribution of event type and time-to-event.
• Modeling conditional time-to-event distributions using smooth semi-nonparametric densities.
• A stepwise forward algorithm for model estimation and adaptive selection of polynomial degrees, implemented in R.

Main Results:

• The proposed method demonstrates superior performance compared to traditional parametric and nonparametric alternatives in simulations.
• The approach effectively handles arbitrary censoring and truncation with mild parametric assumptions.
• The simulations support the use of 'ad hoc' asymptotic inference for confidence intervals.

Conclusions:

• The novel semi-nonparametric approach provides a powerful and flexible tool for cumulative incidence estimation in competing risks settings.
• The method offers advantages in handling complex data scenarios often encountered in clinical trials.
• Further extensions to regression modeling show promise for broader applications in survival data analysis.