Abstract 1674: Prediction of survival based on tumor size dynamics and new lesions in EGFR mutation-positive non-small cell lung cancer patients treated with gefitinib or carboplatin and paclitaxel

Abstract

Abstract Objectives: Dynamics of tumor size have long been used for the clinical diagnosis, staging, prognosis and treatment of non-small cell lung cancer (NSCLC). The aim of this work was to develop a joint statistical framework to interrogate the longitudinal tumor burden, the appearance of new lesions, and changes in therapy following progression in a population of NSCLC patients treated with an EGFR inhibitor gefinitib to predict, given early results, time to progression (PFS) and overall survival at both population and individual levels. Methods: The model builds on a previously developed joint model for tumor size and survival. The tumor size model was extended to estimate the probability of a new lesion. For the survival submodel, two additional biomarkers were evaluated: the derivative of the sum of longest diameters (SLD) and the probability of new lesions. The model was fitted in a Bayesian framework using STAN. Model was developed from 434 EGFR+, NSCLC patients treated with gefitinib or carboplatin+paclitaxel (‘IPASS’, NCT00322452). Predictive performance was evaluated on a cohort of 102 EGFR+ NSCLC patients (‘IFUM’, NCT01203917). Results: Overall, the model fit well to the observed data and the predictive performance was strong in both within- and out-of-sample evaluations. Surprisingly, SLD derivative and new-lesion association parameters appear to negatively impact model performance, indicating further improvement of these submodels might be necessary. The model recapitulates response evaluation criteria in solid tumors (RECIST) outcomes with nearly 90% accuracy. When comparing progression-specific hazards, it appears the drug effect is mediated by tumor size dynamics rather than new lesion incidence. Conclusion: This Bayesian joint model of tumor size and survival accurately recapitulates the RECIST-based outcomes and generates well calibrated predictions of survival. It offers insights regarding the relative predictive value of the components of RESICT in NSCLC. Citation Format: James Dunyak, Jacqueline Buros-Novik, Eric Novik, Hong Yan, Nidal Al-Huniti. Prediction of survival based on tumor size dynamics and new lesions in EGFR mutation-positive non-small cell lung cancer patients treated with gefitinib or carboplatin and paclitaxel [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1674.