Testing for the Functional Form of a Continuous Covariate in the Shared-Parameter Joint Model.

Area of Science:

• Biostatistics
• Survival Analysis
• Statistical Modeling

Background:

• Shared-parameter joint models are used to link longitudinal and time-to-event data.
• Conventional models assume a linear relationship between covariates and the hazard function, which can be overly restrictive.

Purpose of the Study:

• To develop an easy-to-use nonparametric test for checking the linearity assumption of continuous covariates in joint models.
• To assess the impact of non-linear covariate effects on model performance.

Main Methods:

• A penalty-modified Akaike information criterion was adapted to create a nonparametric test criterion.
• Extensive numerical simulations were performed to validate the test within the joint modeling framework.
• The study evaluated the extent of deviation from linearity and the subsequent improvement in predictive performance.

Main Results:

• The proposed test effectively assesses the linearity assumption for continuous covariates in joint models.
• Deviations from linearity were identified, and their impact on predictive performance was quantified.
• The methodology demonstrated its utility in a real-world clinical trial setting.

Conclusions:

• The developed nonparametric test provides a valuable tool for validating linearity assumptions in joint modeling.
• Accurate modeling of covariate effects, including non-linear relationships, enhances the predictive accuracy of joint models.
• This approach is crucial for reliable analysis of complex health data, such as in HIV clinical trials.