Joint Models of Longitudinal and Time-to-Event Data with More Than One Event Time Outcome: A Review

Abstract

Methodological development and clinical application of joint models of longitudinal and time-to-event outcomes have grown substantially over the past two decades. However, much of this research has concentrated on a single longitudinal outcome and a single event time outcome. In clinical and public health research, patients who are followed up over time may often experience multiple, recurrent, or a succession of clinical events. Models that utilise such multivariate event time outcomes are quite valuable in clinical decision-making. We comprehensively review the literature for implementation of joint models involving more than a single event time per subject. We consider the distributional and modelling assumptions, including the association structure, estimation approaches, software implementations, and clinical applications. Research into this area is proving highly promising, but to-date remains in its infancy.

The present and future scope of real‐world evidence research in diabetes: What questions can and cannot be answered and what might be possible in the future?

“…Methods are also available to model the transitions of clinical conditions over time (multistate models), which allow to clarify what risk factors play a role in determining a specific transition (eg, from CVD to cancer) rather than another (eg, from cancer to CVD). Lastly, multiple longitudinal measures of risk factors and multiple, successive transitions of complications may be simultaneously modelled, thus giving greater insights into the natural history of diabetes and facilitating the implementation of personalized preventive strategies. Notably, not only is possible to obtain dynamic predictions based on “static” risk models, but models themselves may be regularly and automatically updated (eg, recalibrated) using “real‐time” population data …”

Dynamic monitoring of the effects of adherence to medication on survival in heart failure patients: A joint modeling approach exploiting time‐varying covariates

“…We apply the joint model approach proposed by Rizopoulos (2012) for dealing with time‐to‐event and endogenous longitudinal covariates. The choice of a joint model is driven by the fact that, when the outcome processes are correlated, joint modeling has empirically demonstrated to reduce biases, improve efficiency and prediction, and can be applicable to outcome surrogacy (Hickey et al., 2018). All the analyses were carried out using the free software R (R Core Team, 2018), in particular JMbayes package (Rizopoulos, 2016).…”

“…Since the data we came up with in our procedure were jointly determined with the responses of interest and may be intended as endogenous covariates, this framework enables their proper treatment. The flexibility and wide range applicability of joint models to clinical setting (Hickey, Philipson, Jorgensen, & Kolamunnage‐Dona, 2018) allow for subject‐specific predictions and construction of personalized medicine tools. In fact, the value added by our approach consists in performing an ongoing analysis and a quantification of adherence effect on patient's outcome that allow to carry out a real‐time monitoring and profiling of patients as well as a personalized prediction about long‐term prognosis.…”

Characterization of direct and/or indirect genetic associations for multiple traits in longitudinal studies of disease progression

“…Extensions of joint models for multiple longitudinal and/or multiple time-to-event traits can further improve inference by borrowing information among related traits. Extensions of joint models have been reviewed for multiple longitudinal traits 6,7 or multiple time-to-event traits 8 . A few extensions for both multiple longitudinal and multiple time-to-event traits have been developed ( [9][10][11] , among others), but these models are often formulated for a specific study question, and thus can lack generalizability.…”