|
Abstract:
|
Phase I trials assign patients to different dose levels to determine the safest and most beneficial therapy. Dose-finding algorithms typically assume that the probability of a response to treatment and toxicity increase with the dose level, and focus on the estimation of the maximum tolerated dose (MTD). Most dose-finding designs define allocation rules that assign patients to dose levels that are close to current estimates of the MTD. The hope is that this assignment rules provide sufficient information about the MTD. In this work, we introduce Information-Driven Dose-Finding (IDDF) designs, which make use of utility functions that are monotone functions of toxicity and efficacy rates at a given dose. The optimal dose maximizes this function. Under IDDF, patients are assigned to the different doses such that the information about the optimal dose is maximized, and both toxic and ineffective doses are avoided. We apply the IDDF approach to different dose-finding problems, including single-agent and combination-therapies trials. Through simulation studies, we show that IDDF has very appealing operating characteristics.
|
