Transparent and Efficient Designs for Clinical Trials

Abstract

Modern early phase clinical trials are integral in assessing the efficacy and safety of new treatments. Traditional methodologies heavily rely on complex parametric models to determine dose-response relationships. They come with inherent challenges: difficulty in practical validation, potential for poor performances if parametric assumptions are inaccurately defined, and a heavy learning burden for medical practitioners. The need for novel methods that bridge the gap between statistical robustness and clinical applicability is evident. To accommodate those issues, we proposed two transparent and efficient designs. The modified isotonic regression based phase I/II clinical trial design (mISO) and the utility-based model free phase I/II design (UFO) represent innovative strides in identifying optimal doses for clinical trials. The mISO design, eschewing traditional parametric assumptions, offers a transparent and efficient method, adaptable to various dose-response curves and enhanced by the mISO-B extension for delayed outcomes. In parallel, the UFO design, specifically tailored for immunotherapy trials, diverges from complex models to employ a dynamic, utility-based approach. This approach continuously updates with trial data, optimizing dose allocation for each patient cohort. Both designs have demonstrated superior performance in comprehensive simulation studies by comparing them with existing methods. Several sequential methods populate the statistical literature, but there remains a notable gap in addressing secondary objectives without altering the primary aim. Addressing this, a two-stage design for randomized controlled trials sequentially testing superiority and noninferiority introduces a novel two-stage group sequential strategy. This strategy primarily aims to establish the superiority of a treatment, assessed at both interim and final stages. Uniquely, it shifts to test noninferiority only if the superiority criterion is not met at the end of the second stage. This dual-focus approach is particularly appreciated in clinical settings for its practical application. Furthermore, it provides a valuable alternative in scenarios where achieving sufficient power for the superiority objective is hindered by limited participant recruitment, allowing the study to pivot towards demonstrating noninferiority.