A Hierarchical Bayesian Design for Phase I Trials of Novel Combinations of Cancer Therapeutic Agents

Abstract

Summary We propose a hierarchical model for the probability of dose-limiting toxicity (DLT) for combinations of doses of two therapeutic agents. We apply this model to an adaptive Bayesian trial algorithm whose goal is to identify combinations with DLT rates close to a pre-specified target rate. We describe methods for generating prior distributions for the parameters in our model from a basic set of information elicited from clinical investigators. We survey the performance of our algorithm in a series of simulations of a hypothetical trial that examines combinations of four doses of two agents. We also compare the performance of our approach to two existing methods and assess the sensitivity of our approach to the chosen prior distribution.

A Two-Dimensional Search Algorithm for Dose-Finding Trials of Two Agents

“…The second scenario is similar to the first, except that an additional dose level of Agent 2 is created to allow for an unequal number of doses for each agent. The third and fourth scenarios were proposed by Braun and Wang (2010 target toxicity level of 30%. In the fourth scenario, the DLT rates increase rapidly with increasing doses of Agent 1 but increase slowly with increasing doses of Agent 2.…”

Effective sample size for computing prior hyperparameters in Bayesian phase I–II dose-finding

“…Consequently, the design's operating characteristics (OCs) are very sensitive to the numerical values of $\tilde{\theta }$ In phase I or phase I-II dose-finding trials, the adaptive decisions include choosing doses for successive patient cohorts and deciding whether to stop the trial early, and the OCs include the selection probability and sample size for each dose under each of several dose-outcome scenarios. Discussions of this problem in particular settings, and various methods for determining $\tilde{\theta }$ are given by Braun and Wang [1], Cheung [2], and Thall et al [3, Section 4.3].…”

Operating Characteristics of Restrictions on Skipping Dose Level for Adaptive Dose-Finding Method in Two-Agent Phase I Trials

“…Conversely, Braun and Wang (2010) proposed the restriction of dose escalation or deescalation to one dose level of change only, but allowed a simultaneous escalation or de-escalation of both agents (Restriction 2). This implies that if the current dose combination is (Aj, B k ), then the next cohort of patients would be allocated to the adjacent dose combination in the set;…”

“…Yuan (2009a, 2009b) and Hirakawa et al (2013) also used the same restriction in their dose-finding methods. Braun and Wang (2010) proposed the restriction of dose escalation or deescalation to one dose level of change only, but allowed a simultaneous escalation or de-escalation of both agents. In contrast, Wages et al (2010aWages et al ( , 2010b incorporated no restriction on skipping dose levels.…”

“…A Bayesian adaptive design based on latent 2 × 2 tables (Yin and Yuan, 2009a) and a copula-type model (Yin and Yuan, 2009b) have also been developed for two agents. Braun and Wang (2010) proposed a hierarchical Bayesian model to determine the toxicity probability of the combination of two agents. Wages et al (2011a) developed a design based on the notion that there are pairs of dose combinations for which the ordering of the probabilities of toxicity cannot be known a priori, resulting in a partial order.…”