Data-Dependent Early Completion of Dose Finding Trials for Drug-Combination

TL;DR

This paper introduces a data-dependent early stopping rule for dose-finding trials in drug combinations, using beta-binomial probabilities and isotonic regression to improve trial efficiency and decision-making.

Contribution

It proposes a novel early completion method based on dose retainment probabilities adjusted by bivariate isotonic regression, validated through simulation studies.

Findings

Proposed method outperforms existing approaches in simulations.

Early completion reduces patient exposure and trial duration.

Provides program code for practical implementation.

Abstract

We propose a data-dependent early completion of dose finding trials for drug-combination. The early completion is determined when a beta-binomial probability for dose retainment with the trial data and the number of remaining patients is high. This paper also proposes an early completion method that a dose retainment probability is adjusted by a bivariate isotonic regression. We demonstrate the early completion for a virtual trial. We evaluate the performance of early completion method through simulation studies with 12 scenarios. We confirmed the superior performance for our proposed early completion methods. We show the number of patients for determining early completion before a trial starts and a program code for calculating dose retainment probability in this paper.