CRM2DIM: A SAS macro for implementing the dual-agent Bayesian continual reassessment method

Abstract

Background and objective: The continual reassessment method (CRM) is a model-based dose-finding design for single-agent phase I oncology trials. With the advance of targeted therapies in oncology, more and more phase I trials investigate drug combinations rather than a single agent in order to find one or more maximum tolerated dose combinations. Several designs have been proposed for such dose-finding trials but only a few software packages are available to implement them. One of the designs is the two-dimensional Bayesian CRM proposed by Wang and Ivanova. Our goal was to provide an easy-to-use program to implement this design.

Methods: We developed a new SAS macro, CRM2DIM, for implementing this design. This macro can be used to run a phase I dose-finding trial for two-drug combination and to perform simulations.

Results: We describe the program with its different features, including the possibility of running an initial design (start-up rule), the possibility of incorporating historical data, and the choice of using either a power or a logistic regression model with or without interaction term. We illustrate our program by presenting simulation results and by a hypothetical trial example.

Conclusions: The CRM2DIM macro provides a SAS implementation of the two-dimensional Bayesian CRM for dual-agent phase I oncology trials. It is an easy-to-use program that includes many useful features and provides statisticians involved in the early phases of development a new tool for designing dual-agent phase I oncology trials.

Keywords: CRM; Continual reassessment method; Dose-finding; Drug combinations; Phase I clinical trial; SAS MCMC procedure.

Figure 1:

The functional block diagram of the SAS macro CRM2DIM. (1, 1): corresponds to dose level 1 for both the first and second agent, (x, 1): corresponds to dose level x and 1 for the first and second agent, respectively, MTDC: maximum tolerated dose-combination, simulk: simulated dataset for the kth replication

Figure 2:

Example of a trial with 35 patients, 20 patients assigned in cohorts of 2 during the start-up and 15 patients in cohorts of 3 during the part guided by the two-dimensional CRM. Dotted lines represent the path observed during the start-up and dashed lines the path of the two-dimensional design. The stars correspond to observed toxicities, at combinations (5,2) and (4,3) during the start-up and (5,1),(4,2),(4,3) during the two-dimensional design. The circles indicate the maximum tolerated dose combinations recommended after the trial.