CFO: Calibration-Free Odds Bayesian Designs for Dose Finding in Clinical Trials.

Abstract

Listen

Translate article

Calibration-free odds type (CFO-type) designs have been demonstrated to be robust, model-free, and practically useful, which have become the state-of-the-art approach for dose finding. However, a key challenge for implementing such designs is a lack of accessible tools. We develop a user-friendly R package and Shiny web-based software to facilitate easy implementation of CFO-type designs. Moreover, we incorporate randomization into the CFO framework. We created the R package CFO and leveraged R Shiny to build an interactive web application, CFO suite, for implementing CFO-type designs. We introduce the randomized CFO (rCFO) design by integrating the exploration-exploitation mechanism into the CFO framework. The CFO package and CFO suite encompass various variants tailored to different clinical settings. Beyond the fundamental CFO design, these include the two-dimensional CFO (2dCFO) for drug-combination trials, accumulative CFO (aCFO) for accommodating all dose information, rCFO for integrating exploration-exploitation via randomization, time-to-event CFO (TITE-CFO), and fractional CFO (fCFO) for addressing late-onset toxicity. Using all information and addressing delayed toxicity outcomes, TITE-aCFO and fractional-aCFO are also included. The package provides functions for determining the subsequent cohort dose, selecting the maximum tolerated dose, and conducting simulations to evaluate performance, with results presented through textual and graphical outputs. The CFO package and CFO suite provide comprehensive and flexible tools for implementing CFO-type designs in phase I clinical trials. This work is highly significant as it integrates all existing CFO-type designs to facilitate novel trial designs with enhanced performance. In addition, this promotes the spread of statistical methods using a user-friendly R package and Shiny software. It strengthens collaborations between biostatisticians and clinicians, further enhancing trial performance in terms of efficiency and accuracy.