Bayesian shrinkage approach for a joint model of longitudinal and survival outcomes assuming different association structures

Abstract

The joint modeling of longitudinal and survival data has recently received much attention. Several extensions of the standard joint model that consists of one longitudinal and one survival outcome have been proposed including the use of different association structures between the longitudinal and the survival outcomes. However, in general, relatively little attention has been given to the selection of the most appropriate functional form to link the two outcomes. In common practice, it is assumed that the underlying value of the longitudinal outcome is associated with the survival outcome. However, it could be that different characteristics of the patients' longitudinal profiles influence the hazard. For example, not only the current value but also the slope or the area under the curve of the longitudinal outcome. The choice of which functional form to use is an important decision that needs to be investigated because it could influence the results. In this paper, we use a Bayesian shrinkage approach in order to determine the most appropriate functional forms. We propose a joint model that includes different association structures of different biomarkers and assume informative priors for the regression coefficients that correspond to the terms of the longitudinal process. Specifically, we assume Bayesian lasso, Bayesian ridge, Bayesian elastic net, and horseshoe. These methods are applied to a dataset consisting of patients with a chronic liver disease, where it is important to investigate which characteristics of the biomarkers have an influence on survival. Copyright © 2016 John Wiley & Sons, Ltd.

A Comparison of Joint Models for Longitudinal and Competing Risks Data, with Application to an Epilepsy Drug Randomized Controlled Trial

“…Ideally, a pragmatic balance between biologically plausible associations and a parsimonious model should be used; however, this might still be challenging. To date, there is limited research on this particular issue (Andrinopoulou and Rizopoulos, 2016). Although each model can be used for inference, if interest lies in risk prediction for personalized treatments, then model 4 would perhaps be most appropriate.…”

“…Different characteristics of the longitudinal profile may influence the cause-specific hazards submodel. Within the joint model literature, several association structures have been proposed, but little attention has been given to the challenge of choosing the most appropriate structure for a given data set (Andrinopoulou and Rizopoulos, 2016). Among the models that are examined here, various latent association structures have been considered, including several alternatives for some models (Table 1).…”

Bayesian joint modeling of bivariate longitudinal and competing risks data: An application to study patient‐ventilator asynchronies in critical care patients

“…Models M 1 and M 2 are similarly defined, except that both also include the longitudinal information of the AI, M 1 in terms of its mean m 2i (t) with association parameters α (1) 2k , and M 2 through the area under the mean of the longitudinal trajectory up to the time point t, t 0 m 2i (s)ds, with association parameters α (2) 2k . An interesting issue is the choice of the scale of the AI variable included in the joint model, the subject-specific mean scale m 2i (t) or the linear predictor scale logit(m 2i (t)) (Andrinopoulou & Rizopoulos, 2016). We opted for m 2i (t) because it facilitates interpretation even though more computational effort is required.…”

TheRPackageJMbayesfor Fitting Joint Models for Longitudinal and Time-to-Event Data Using MCMC

“…• Implement priors for data-based selection of association structures (Andrinopoulou and Rizopoulos 2016). …”