A phase I/II seamless dose escalation/expansion with adaptive randomization scheme (SEARS)

Abstract

Background: Standard drug development conducts phase I dose finding and phase II dose expansion sequentially and separately. Information between the two phases is rarely shared. Administratively, such a sequential process is time-consuming and burdensome.

Purpose: We propose seamless dose escalation/expansion with adaptive randomization scheme (SEARS), a seamless design that combines phase I dose escalation based on toxicity with phase II dose expansion and dose comparison based on efficacy. SEARS allows extension from phase I to phase II under one design with no gap in between and employs a dynamic and parallel procedure involving simultaneous dose escalation, dose graduation, and adaptive randomization.

Methods: SEARS integrates three components into a seamless scheme. Specifically, in phase I, SEARS applies the modified toxicity probability interval (mTPI) method to monitor dose escalation based on toxicity outcome. Doses that show promising efficacy and safety are immediately graduated from phase I and placed to a phase II stage in which patients are adaptively randomized based on efficacy outcome. Phase I dose escalation, dose graduation, and phase II adaptive randomization proceed simultaneously throughout the entire trial.

Results: Examples are given comparing SEARS with two other designs, in which superior performance of SEARS is demonstrated. An important and promising finding is that SEARS reduces sample sizes without losing power. R program and demo slides of SEARS can be obtained at http://health.bsd.uchicago.edu/yji/soft.html

Limitation: We assume that the binary efficacy and toxicity response can be measured in the same time frame. This is often achievable with surrogate efficacy markers in practice.

Figure 1

Schema of the SEARS design. Five doses are compared. In Phase I, dose escalation is based on the mTPI design and toxicity outcome. Doses can graduate from Phase I to Phase II under a graduation rule. Graduated doses are compared with each other as patients are randomized in Phase II.

Figure 2

Comparison of the Conventional, XJT, and SEARS design, when the placebo efficacy rate q₀ = 0.2. Scenarios labels are in the format of EffShape-ToxShape representing the shapes of dose response curves for the efficacy and toxicity outcomes. Trial type I error rate is matched for the Null-Inc scenario for all three designs, and trial power and sample size are compared for the remaining scenarios.

Figure 3

Comparison of the Conventional, XJT, and SEARS design, when the placebo efficacy rate q₀ = 0.5. Scenarios labels are in the format of EffShape-ToxShape representing the shapes of dose response curves for the efficacy and toxicity outcomes. Trial type I error rate is matched for the Null-Inc scenario for all three designs, and trial power and sample size are compared for the remaining scenarios.