Joint modeling of longitudinal endpoints and its applications to trial planning, monitoring and analysis

In the context of clinical trial practices, the study power and sample size are typically determined based on the expected treatment effects on the primary endpoint collected over time. The utilization of longitudinal modeling for the primary endpoint offers a flexible approach that has the potential to reduce the sample size and duration of the trial, thereby improving operational efficiency and costs. Joint modeling of multiple endpoints presents a unique opportunity to understand how the primary endpoint evolves over time with other clinically important endpoints, and has the potential to increase precision of estimates and therefore increase study power when designing a study at planning stage and enhance understanding and interpretation of the data at a multi-dimensional level at the analysis stage. This approach enables a comprehensive evaluation of clinical evidence from various perspectives, rather than relying solely on isolated pieces of information. Joint modeling of multiple longitudinal endpoints would also help trial monitoring process as the trial accumulates clinical evidence of efficacy data, and there is a high demand in developing tools for statistical learning the treatment benefits on the go especially when the endpoint(s) is not well-established yet in some therapeutic indications. In this article, we will illustrate the use of joint modeling of longitudinal endpoints and its applications to study design, analysis, and trial monitoring practices. Simulation studies suggest that the potential efficiency gain would be achieved via leveraging information within endpoint over time and/or between endpoints. We developed an R shiny application to aid in and support identifying promising efficacy signals from endpoints under investigation during the trial monitoring. The implementation of the joint models and the added values will be discussed through case studies and/or simulation studies.