All Times EDT
Keywords: Dose finding; Dose response; Efficacy and toxicity; Interval designs; Rule-based designs; Software
In this talk, I will introduce some newly developed statistical designs based on probability intervals for dose-finding trials with joint binary efficacy and toxicity endpoints. The type of trials is common in gene or cell therapies in which clinical response may not be monotone with dose levels. The new designs aim to strike a balance of practical convenience and statistical performance, leveraging the interval ideas originally proposed in dose-finding designs for DLT-based dose-finding trials. As a result, some new designs, such as the Ji3+3 design (Lin and Ji, 2021) are based on prespecified rules with no probability modeling or inference. Interestingly, the rule-based design compares well with more complex designs using model-based inference, regardless how complex or simple the models are. I will demonstrate the performance of the new designs in a commercial software.