Probability intervals of toxicity and efficacy design for dose-finding clinical trials in oncology

Comparative review of novel model-assisted designs for phase I/II clinical trials

In recent years, both model-based and model-assisted designs have emerged to efficiently determine the optimal biological dose (OBD) in phase I/II trials for immunotherapy and targeted cellular agents. Model-based designs necessitate repeated model fitting and computationally intensive posterior sampling for each dose-assignment decision, limiting their practical application in real trials. On the other hand, model-assisted designs employ simple statistical models and facilitate the precalculation of a decision table for use throughout the trial, eliminating the need for repeated model fitting. Due to their simplicity and transparency, model-assisted designs are often preferred in phase I/II trials. In this paper, we systematically evaluate and compare the operating characteristics of several recent model-assisted phase I/II designs, including TEPI, PRINTE, Joint i3+3, BOIN-ET, STEIN, uTPI, and BOIN12, in addition to the well-known model-based EffTox design, using comprehensive numerical simulations. To ensure an unbiased comparison, we generated 10,000 dosing scenarios using a random scenario generation algorithm for each predetermined OBD location. We thoroughly assess various performance metrics, such as the selection percentages, average patient allocation to OBD, and overdose percentages across the eight designs. Based on these assessments, we offer design recommendations tailored to different objectives, sample sizes, and starting dose locations.

Optimal biological dose selection in dose-finding trials with model-assisted designs based on efficacy and toxicity: a simulation study

With the emergence of molecular targeted agents and immunotherapies in anti-cancer treatment, a concept of optimal biological dose (OBD), accounting for efficacy and toxicity in the framework of dose-finding, has been widely introduced into phase I oncology clinical trials. Various model-assisted designs with dose-escalation rules based jointly on toxicity and efficacy are now available to establish the OBD, where the OBD is generally selected at the end of the trial using all toxicity and efficacy data obtained from the entire cohort. Several measures to select the OBD and multiple methods to estimate the efficacy probability have been developed for the OBD selection, leading to many options in practice; however, their comparative performance is still uncertain, and practitioners need to take special care of which approaches would be the best for their applications. Therefore, we conducted a comprehensive simulation study to demonstrate the operating characteristics of the OBD selection approaches. The simulation study revealed key features of utility functions measuring the toxicity-efficacy trade-off and suggested that the measure used to select the OBD could vary depending on the choice of the dose-escalation procedure. Modelling the efficacy probability might lead to limited gains in OBD selection.

Bayesian Sample Size Planning Tool for Phase I Dose-Finding Trials

PURPOSE

Through Bayesian inference, we propose a method called BayeSize as a reference tool for investigators to assess the sample size and its associated scientific property for phase I clinical trials.

METHODS

BayeSize applies the concept of effect size in dose finding, assuming that the maximum tolerated dose can be identified on the basis of an interval surrounding its true value because of statistical uncertainty. Leveraging a decision framework that involves composite hypotheses, BayeSize uses two types of priors, the fitting prior (for model fitting) and sampling prior (for data generation), to conduct sample size calculation under the constraints of statistical power and type I error.

RESULTS

Simulation results showed that BayeSize can provide reliable sample size estimation under the constraints of type I/II error rates.

CONCLUSION

BayeSize could facilitate phase I trial planning by providing appropriate sample size estimation. Look-up tables and R Shiny app are provided for practical applications.