A Randomized Bayesian Phase 1 Design Combining An Mps-1 Inhibitor With Paclitaxel: A Strategy To Improve Determination Of The Incremental Toxicity Of A Novel Compound Over A Known Backbone Therapy.

2537 Background: Here we present a study combining BAY1217389 (BAY), a potent MPS-1 kinase inhibitor with a backbone chemotherapy paclitaxel. Since we expected overlapping toxicities we sought to improve determination of the maximal tolerated dose (MTD) using a randomized phase 1 design with Bayesian dose modeling. We hypothesized that this approach may determine the MTD of BAY more accurately by limiting the effect of variability of dose limiting toxicities (DLTs) related to paclitaxel. Methods: Patients (≥18 years) were randomized to receive oral BAY with intravenous paclitaxel (experimental arm) or paclitaxel monotherapy (standard arm) in cycle 1. Dose escalation was guided by Bayesian modeling targeting a DLT-rate in the experimental arm of 10% over DLT-rate in the standard arm. PK profiles were determined for both BAY and paclitaxel. Simulations were performed to estimate MTD for several scenarios. Results: We were able to establish an MTD of 65 mg BAY using 50 patients in the dose-escalation part. As expected the main DLTs were hematologic toxicity. Grade ≥ 3 neutropenia was predominantly observed in the experimental arm and was related to higher BAY AUC0-12 on D8 (p<0.001) and not to paclitaxel AUC0-24 (p>0.1). To determine whether the randomization adds value to the study design we ran simulations comparing our randomized strategy using variable toxicity rate (5, 10, 20, 40%) for paclitaxel monotherapy with the 3+3 design. These data showed that the randomized design outperformed the 3+3 design. The 3+3 design underestimated the MTD as dose escalation was terminated more frequently at first dose for higher paclitaxel toxicity rate. Conclusions: Randomized Bayesian phase 1 dose escalation design was feasible with BAY plus paclitaxel. A major advantage of this design is the precise determination of an exposure-toxicity relation for the experimental drug. Moreover, simulations support our hypothesis that the randomized design was able to determine the MTD accurately regardless of variable toxicity rate for paclitaxel. This approach may improve dose determination in phase I combination trials. Clinical trial information: NCT02366949.