Flexible Link Continual Reassessment Methods for Trivariate Binary Outcome Phase I/II Trials

Abstract

Even though the primary goal of a Phase I clinical trial is to determine the dosing schedule of a new treatment subject to toxicity or other safety concerns, treatment efficacy often remains an important secondary consideration. In such settings, the trial may collect both information on final efficacy as well as a surrogate efficacy marker that is obtained more easily, quickly, or both.Extended versions of Bayesian continual reassessment methods (CRMs) offer an attractive approach for a principled combination of all three sources of information, but the precise shape of the doseresponse curve may be difficult to specify, especially with the small sample sizes typical of early phase studies. In this paper, we propose flexible semi-and non-parametric link functions for a trivariate binary outcome CRM that allows for differential weighting of the outcomes, or even within outcome (say, higher penalties for over-rather than under-shooting toxicity). Illustrating in the context of a non-Hodgkin lymphoma trial, we show via simulation that our flexible link methods can outperform standard parametric CRM approaches in terms of both the probability of correct dose selection and the proportion of patients treated at that dose.

Dose escalation with over-dose and under-dose controls in Phase I/II clinical trials

“…Another limitation of this design is that it does not consider the common case of missing or incomplete efficacy data, especially at low starting dose levels. [22] have proposed a modified CRM-based Phase I/II clinical trial design in which a surrogate response is employed to represent the ultimate efficacy in order to solve the problem of missing efficacy data. However, the reliability of surrogate responses need to be verified and the availability of reliable surrogates are limited.…”

Discussion of ‘Small-sample behavior of novel phase I cancer trial designs’ by Assaf P Oron and Peter D Hoff

“…Potential enhancements include simultaneous models for toxicity and efficacy (or surrogate efficacy), range parameters to capture lower or upper bounds on response probabilities, differential weighting schemes to place a higher penalty on overdosing than underdosing, early termination rules to control overtoxicity and to enable an early decision regarding the optimal dosage, and so on (see Zhong et al . [6] and [7] for a discussion of some of these issues in the standard BP1 context).…”