Phase 1 oncology clinical trials for related subpopulations: The power walk design

Motivation/Background: This paper addresses designing a Phase 1 trial where the doseresponse curve differs across patient subpopulations. We examine a design which "power walks" through clinically ineffective dose levels, accruing from either subpopulation, until dose limiting toxicity appears. The design then splits into two Phase 1 trials for each subpopulation. Methods/Results: We performed simulation studies of 10000 Phase 1 trials comparing the "power walk" design to the alternatives. The power walk design correctly estimates the maximum tolerated dose (MTD) up to 20% more often. The power walk design performs best with a steep dose-response curve, and overestimates the MTD roughly 10% of the time. Conclusions: The power walk design is appropriate to obtain MTDs for separate patient subpopulations where the initial dose is clinically ineffective or substantially below the MTD. A power walk design study involves fewer patients and achieves the correct MTD more often. The power walk design is not recommended when the initial dose level is near the MTD.