Model-based predictions of expected anti-tumor response and survival in phase III studies based on phase II data of an investigational agent

Abstract

6025 Background: Decision making during early oncology drug development remains very empirical. A drug-disease simulation model to predict expected clinical response and survival in phase III studies from observed longitudinal tumor size in phase II trials offers a science-based alternative. Methods: We developed the following: 1) longitudinal exposure-response models of drug effect (and resistance) on tumor growth dynamics based on phase II data on capecitabine (X) in metastatic breast (MBC) and colorectal cancer (MCRC) and on historical phase III data on docetaxel (T) in MBC and 5-FU in MCRC; and 2) a survival model relating change in tumor size and patient characteristics to survival time in MBC and MCRC based on phase III data. The models were validated and used to predict expected anti-tumor response and survival in phase III studies of XT vs. T alone in MBC and X vs. 5-FU in MCRC. Multiple replicates (n = 1000) of simulated phase III studies were compared to actual results. Results: Change of tumor size and survival time distributions in X arms were well predicted. In MBC, expected median (90% prediction interval) survival for patients treated with XT was 412 (range 330 to 526) days (vs. 431 days observed) with an improvement over T of 57 days (range -17 to 148) (vs. 78 days observed). For MCRC patients, X treatment resulted in a potential survival improvement of 39 (range -21 to 110) days (vs. 35 days observed) compared to patients treated with 5-FU. Conclusions: The dose-tumor size-survival model succeeded in predicting survival in phase III trials based on X phase II data in MBC (in combination with T) and as a single agent in MCRC. This model is a useful tool to compare expected clinical response of new compounds to various competitors and to support end-of-phase II decisions and design of phase III studies. [Table: see text]