A longitudinal model for tumor size measurements in clinical oncology studies
Claret, L. (1), P. Girard (2), K.P. Zuideveld(3), K. Jorga (3), J. Fagerberg (3), R. Bruno (1)
(1) Pharsight Corp., Mountain View, CA; (2) INSERM, EA3738 University Lyon Sud, Oulins, France; (3) F. Hoffmann-La Roche, Basel, Switzerland
Introduction: The analysis of anti-tumor response in clinical studies of anti-cancer drugs remains very empirical (assessments based on response rate). A drug–disease model of observed longitudinal tumor size in phase II and Phase III trials offers a more reliable quantitative assessment.
Methods: We developed a longitudinal exposure-response model of drug effect on tumor growth dynamics based on phase II data on capecitabine (C) and on phase III data on docetaxel (T) in metastatic breast cancer (MBC). The model involves sub-models for tumor growth dynamics, drug effect and drug resistance. The model was validated using a posterior predictive check. It was used retrospectively to predict expected anti-tumor response in a Phase III study of C in combination with T. For C, extrapolation from Phase II to Phase III used Phase III estimated tumor growth dynamics and patients characteristics using T data. Multiple replicates (n=1000) of simulated phase III studies were compared to actual results.
Results: Change of tumor size (week 6 measurement/baseline) in the Phase III study was well predicted. Expected median (90% prediction interval) response for patients treated with T + C was 0.27 (0.18-0.36) vs. 0.21 observed)
Conclusion: The tumor size dynamic model succeeded in predicting response in a combination Phase III trial based on single agent Phase II or Phase III data in MBC. This model will be part of a modeling framework to simulate expected clinical response of new compounds and to support end-of-Phase II decisions and design of Phase III studies.