Flexible joint model for time-to-event and non-Gaussian longitudinal outcomes.
Stat Methods Med Res
; : 9622802241269010, 2024 Sep 09.
Article
en En
| MEDLINE
| ID: mdl-39248224
ABSTRACT
In medical studies, repeated measurements of biomarkers and time-to-event data are often collected during the follow-up period. To assess the association between these two outcomes, joint models are frequently considered. The most common approach uses a linear mixed model for the longitudinal part and a proportional hazard model for the survival part. The latter assumes a linear relationship between the survival covariates and the log hazard. In this work, we propose an extension allowing the inclusion of nonlinear covariate effects in the survival model using Bayesian penalized B-splines. Our model is valid for non-Gaussian longitudinal responses since we use a generalized linear mixed model for the longitudinal process. A simulation study shows that our method gives good statistical performance and highlights the importance of taking into account the possible nonlinear effects of certain survival covariates. Data from patients with a first progression of glioblastoma are analysed to illustrate the method.
Texto completo:
1
Colección:
01-internacional
Base de datos:
MEDLINE
Idioma:
En
Revista:
Stat Methods Med Res
Año:
2024
Tipo del documento:
Article
País de afiliación:
Bélgica
Pais de publicación:
Reino Unido