Seamless Phase I/II Designs for Precision Medicine Trials

在药物开发中，一个关键属性是选择无效和/或不安全的治疗方法，这些治疗方法应在早期临床试验中排除。离散的传统设计，通常不完全考虑I期的疗效数据和II期的毒性数据，可能导致适当剂量的错误指定。通过将I期和II期合并为一项研究，并对疗效和毒性进行建模，无缝I/II期设计在提高药物开发的效率和准确性方面获得了越来越多的关注。在精准医学时代，随着人们对开发免疫疗法和分子靶向药物（MTA）的兴趣越来越大，无缝I/II期设计的主要目标变成了估计最佳生物剂量（OBD）。这些治疗带来了新的统计学挑战，包括非单调的剂量-反应关系、迟发性结局和临床试验参与者的异质性。该项目旨在研究精确医学试验的无缝I/II期设计：首先，使用贝叶斯分层模型改进所谓的剂量递增随后扩展队列（DE-EC）方法。接下来，考虑评估MTA的试验，其中单调剂量反应假设不成立。常微分方程可以用来拟合曲线。最后，我计划对多项结果进行建模（例如，毒性，无毒但有效），而不是双变量二元毒性和有效性。

In drug development, one key attribute is to select inefficacious and/or unsafe treatments which should be excluded inearly-phase clinical trials. Discrete traditional designs, usually incompletely considering efficacy data in phase I andtoxicity data in phase II, can lead to the misspecification of the appropriate dose(s). By consolidating phases I and II intoa single study and modelling both efficacy and toxicity, seamless phase I/II designs have gained increasing attention toimprove the efficiency and accuracy of drug development. With growing interest in developing immunotherapies andmolecularly targeted agents (MTAs) in the era of precision medicine, the principal objective of seamless phase I/IIdesigns becomes to estimate an optimal biological dose (OBD). Such treatments have raised new statistical challenges,including non-monotonic dose-response relationships, late-onset outcomes, and heterogeneity of participants in clinicaltrials. This project aims to research seamless phase I/II designs for precision medicine trials: First, improve on the so calleddose-escalation followed by expansion cohorts (DE-EC) method using Bayesian hierarchical modelling. Next,consider trials evaluating MTAs, where the monotonic dose-response assumption does not hold. Ordinary differentialequations may be used to fit the curves. Finally, I plan to model multinomial outcomes (e.g., toxicity, no toxicity butefficacy) rather than bivariate binary toxicity and efficacy.