PoD-BIN: A probability of decision Bayesian interval design for time-to-event dose-finding trials with multiple toxicity grades
Meizi Liu, Ji Lin, Gu Mi, Christelle Lorenzato, Xun Chen, Yuan Ji
The Annals of Applied Statistics · 2025 · Volume 19 · Issue 1
Abstract
We introduce a Bayesian framework centered on the “probability of decision” for designing dose-finding trials. The proposed PoD-BIN design evaluates the posterior predictive probabilities of up-and-down decisions. In PoD-BIN, multiple grades of toxicity, categorized as mild toxicity (MT) and dose-limiting toxicity (DLT), are simultaneously modeled, with the primary outcome being time-to-toxicity for both MT and DLT. This approach allows the enrollment of new patients while previously enrolled patients are still being monitored for toxicity, potentially reducing the trial duration. The Bayesian decision rules in PoD-BIN employ the probability of decisions to balance the trade-off between accelerating the trial and the risk of exposing patients to excessively toxic doses. Through numerical examples, we illustrate the trade-off between speed and safety of PoD-BIN and compare it with existing designs. PoD-BIN demonstrates the ability to control the frequency of risky decisions while simultaneously shortening trial duration in simulations.
Keywords
Dose limiting toxicity, equivalence interval, latent probit regression, moderate toxicities, phase I trial, time to toxicity
Enable JavaScript for citation tools and additional paper interactions.