Cancer Biostatistics and Bioinformatics Monthly Seminar

Dose-finding designs for an agent with a non-monotonic dose-efficacy profile

Mark Conaway, PhD
Professor, Division of Translational Research and Applied Statistics
Department of Public Health Sciences, The University of Virginia Health System

Join the Zoom meeting here

ABSTRACT: This talk will present a fully adaptive dose-finding design based on safety and efficacy, with the goal of identifying a safe dose likely to show efficacy in subsequent studies. The design is tailored for the particular features of the agent, which based on pre-clinical studies, is expected to have toxicity that increases with dose, but an efficacy curve that increases to a peak then decreases as a function of dose. The design uses both parametric models and methods based on estimating parameters subject to complete and partial orders. In addition to presenting the design that was proposed for the trial, along with its operating characteristics under different scenarios, the talk will address a number of open statistical questions and design issues that would arise in subsequent studies of this agent. These include incorporating patient heterogeneity, using short-term outcomes to predict long-term outcomes in order to facilitate adaptive dose allocations, and the extension of the design to combinations of agents. 

The meeting agenda includes a presentation, programmatic announcements, and updates from Cancer Center program leaders.

This group meets once a month from noon – 1 p.m.