On semiparametric accelerated failure time models with time-varying covariates: A maximum penalised likelihood...

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• The Cox proportional hazards model is standard but relies on a proportional hazards assumption.
• Accelerated failure time (AFT) models provide an alternative, especially when this assumption is violated.
• Semiparametric AFT models, particularly with time-varying covariates, present significant computational challenges.

Purpose of the Study:

• To develop a penalized likelihood approach for estimating semiparametric AFT models.
• To accommodate both time-fixed and time-varying covariates in the presence of right-censored data.
• To provide a computationally feasible method for complex survival data analysis.

Main Methods:

• Utilized a penalized likelihood estimation framework.
• Employed Gaussian basis functions for smooth approximation of the nonparametric baseline hazard.
• Implemented a constrained optimization approach for model fitting.
• Incorporated both time-fixed and time-varying covariates.

Main Results:

• The proposed penalized likelihood method effectively estimates semiparametric AFT models.
• The method demonstrates good performance in simulation studies.
• The approach is applicable to real-world survival data, as shown with a motor neuron disease dataset.

Conclusions:

• The penalized likelihood approach offers a viable solution for estimating semiparametric AFT models with complex covariate structures.
• This method provides a valuable tool for survival data analysis where Cox model assumptions may not hold.
• The technique is robust and applicable to various censored survival data scenarios.