Joint model of longitudinal tumor size and overall survival (OS) in non-small cell lung cancer (NSCLC): A Bayesian approach

Objective: The methodology of joint modeling longitudinal tumor size and OS data is very promising and provides the opportunity of survival prediction in individual patients. However, current joint modeling software are restricted in terms of distributions/parameterizations available and non-linear mixed effects models are not supported. The objective of this work was to develop a joint model in Bayesian framework to relax these restrictions and enable investigation of additional (continuous and/or discrete) biomarkers. Methods: Longitudinal tumor size and OS data from gefitinib Phase 3 study in NSCLC (‘IPASS9, NCT00322452) was used in the modeling analysis. Joint model was developed in JM package in R and the Bayesian formulation of the joint modeling was implemented in STAN. Different parametrizations were investigated for both sub-models and the impact of lag-times was explored. The parameter estimates and model fit were compared between models by JM package and STAN. The final Bayesian joint model was validated on an independent dataset from a Phase 4 study (‘IFUM9, NCT01203917, EGFR positive patients only) by simulating individual survival based on their tumor size data and compared with observed survival. Results: The estimated parameter values were similar to those from the fit from JM package. The different parametrizations provided similar fits for the tumor size model, while the survival model was improved by using a b-spline model. The association parameters indicate a strong association between tumor size and survival and introduction of lag-times did not impact the association parameters to a great extent. Posterior predictive simulations of OS in phase 4 study were consistent with observed data. Conclusions: Bayesian formulation further improved the joint model of tumor size and OS which would enable investigation of additional biomarkers (e.g. new lesions). Citation Format: Nidal Al-Huniti, Hongmei Xu, Diansong Zhou, Helena Edlund, Sergey Aksenov, Gabriel Helmlinger, James Dunyak. Joint model of longitudinal tumor size and overall survival (OS) in non-small cell lung cancer (NSCLC): A Bayesian approach [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 4761.