Tong Lu (1), Yuying Gao (2), Lichuan Liu (1), Alex Huang (1), Amita Joshi (1), Jin Yan Jin (1)
(1) Genentech, Inc.; (2) Quantitative Solutions, Inc.
Objectives: For rational dose selection of targeted anticancer agents, it is critical to obtain early assessment of clinical activity and target modulation. However, in Phase 1 dose escalation, biomarker dose-response (D-R) assessment based on tumor biopsy data is challenging, considering the uncertainty around efficacious dose, small sample size, high inter-subject variability, and high biopsy failure rate. The objective of this work is to assess the power of detecting biomarker D-R relationship by clinical trial simulations.
Methods: 1) Simulation: Based on available biomarker D-R relationship (inhibitory Emax model), 3 scenarios were simulated with ED50 within, below, or above Phase 1 dose range. 1000 sets of parameters were generated per scenario, incorporating typical value and uncertainty for E0 (baseline) and ED50, inter-trial variability for ED50 (assume 75%), and residual error variability for E0. Two biopsy data (pre- and post-treatment) per subject were simulated, with 1000 subjects per dose for 6 dose levels in each scenario. 2) Bootstrapping: 1000 Ph1 trials with 6 dose levels and typical 3×3 design (n=3/dose) for each scenario were generated by resampling from the virtual subjects. 50% failure rate was applied independent of dose (n=9/trial). Biopsy data for each trial was fitted by Emax model to get 1000 sets of bootstrapped parameters per scenario. 3) Power evaluation: a) Power of estimating D-R relationship was assessed by % of bootstrapped parameters falling into predefined interval ([0.7-1.3]) of true value; b) Power of detecting D-R was assessed by % of subject achieving target inhibition at given dose (ie. success rate).
Results: For base scenario where ED50 was in the middle of the dose range, 52% of the bootstrapped ED50 fall into desired range of true value, indicating a decent power of detecting true D-R relationship in spite of small sample size and high failure rate. In scenarios where ED50 was below or above the dose range, the values were down to 43% or 28%. Simulation also suggested strong likelihood of ED50 underestimation when ED50 was above the dose range. The success rate of detecting D-R was high if ED50 was below or within the dose range, and was <50% if above the dose range.
Conclusions: Theoretical simulations of biomarker D-R were used to inform oncology biomarker strategy. Under different scenarios, simulation can be conducted with varying assumptions to assess the probability to reveal the biomarker D-R in Ph1 trial.
References:
[1] Overman MJ, Modak J, Kopetz S, Murthy R, Yao JC, Hicks ME, Abbruzzese JL, Tam AL. Use of Research Biopsies in Clinical Trials: Are Risks and Benefits Adequately Discussed? J Clin Oncol. 2013;31(1):17-22
[2] Lemech C, Dua D, Newmark J, Saggese M, Simmons E, Spiliopoulou P, Arkenau HT. Patients’ perceptions of research biopsies in phase I oncology trials. Oncology. 2015;88(2):95-102
Reference: PAGE 24 () Abstr 3434 [www.page-meeting.org/?abstract=3434]
Poster: Methodology - Study Design