An integrated dose-finding tool for phase I trials in oncology

Abstract

In the past 25 years, the 3+3 design has been the most popular approach for planning phase I dose-finding trials in oncology. During the same time period, major development of more efficient model-based designs has been made by statistical researchers aiming to improve the clinical practice of dose finding in oncology. Despite the effort, 3+3 is still the most frequently used designs in practice. Part of the reason is due to the lack of software tools that allow comparison of different designs, including 3+3 and other model-based methods, in a head-to-head and easy-to-use fashion. To this end, we introduce NextGen-DF, a next-generation tool for designing oncology dose-finding trials that allows for construction, comparison, and calibration of multiple designs via internet, in real time, and independent of computer operating systems. Through NextGen-DF, we present massive and user-generated comparison results based on over 4 million simulated trials, which clearly indicate the inferiority of 3+3. To our knowledge, the reported crowd-sourcing results are the largest and most objective comparison across major dose-finding methods to date. NextGen-DF is expected to improve patient care and drug development by providing safer and more efficient designs for phase I oncology trials. NextGen-DF is available at www.compgenome.org/NGDF.

Keywords: 3+3; Bayesian design; CRM; Next-generation dose finding; Webtool; mTPI.