Curve-free phase I/II clinical trial designs for molecularly targeted agents and immunotherapy

项目总结/摘要 该提案的主要目标是开发透明、灵活和有效的I/II期临床试验设计 确定分子靶向药物（MTA）和免疫疗法（IT）的最佳剂量。常规第一/第二阶段 临床试验设计通常使用复杂的参数模型来表征联合毒效分布， 并进行审判。然而，参数模型在实践中很难证明是合理的，并且存在错误指定 参数模型的使用可能导致I/II期试验的实质上不期望的性能。此外， 进行试验的医生在临床上解释这些复杂模型的参数， 这种巨大的学习成本阻碍了将新颖的统计设计转化为现实世界的试验实施。 为了解决这些问题，在本提案中，我们将提出透明的无曲线I/II期临床试验设计。 拟定设计未对剂量-效应关系或毒性-疗效进行参数假设 相关性，因此在任何有临床意义的剂量-反应曲线下均表现稳健。简明 临床上可解释的模型表达和剂量发现算法使所提出的设计具有高度的transla， 从统计界到临床界。所提出的设计也是高度灵活的 因为它们适用于单药试验和药物联合试验， 结果或延迟结果。初步的仿真研究表明，所提出的设计是高度 在最佳剂量选择和患者分配方面有效。此外，我们会开发方便用户使用的网页应用程序， 促进所提出的设计在临床实践中的广泛应用。

Project Summary/Abstract The primary goal of this proposal is to develop transparent, flexible and efficient phase I/II clinical trial designs identifying optimal doses for molecularly targeted agents (MTA) and immunotherapy (IT). Conventional phase I/II clinical trial designs often use sophisticated parametric models to characterize the joint toxicity-efficacy distribu- tions and to conduct the trials. However, the parametric models are hard to justify in practice, and misspecification of parametric models can lead to substantially undesirable performances of phase I/II trials. Moreover, it is difficult for the physicians conducting the trials to clinically interpret the parameters of these sophisticated models, and such great learning costs impede the translation of novel statistical designs into real-world trial implementation. To solve these issues, in this proposal we will propose transparent curve-free phase I/II clinical trial designs. The proposed designs make no parametric assumptions on either dose-response relationship or toxicity-efficacy correlation and therefore performs robustly under any clinically meaningful dose-response curves. The concise clinically interpretable model expression and dose-finding algorithm make the proposed designs highly transla- tional from the statistical community to the clinical community. The proposed designs are also highly flexible because they are applicable for both single-agent trial and drug-combination trial with either quickly observable outcomes or delayed outcomes. The preliminary simulation studies show that the proposed designs are highly efficient in optimal dose selection and patient allocation. In addition, we will develop user-friendly web apps to facilitate the widespread application of the proposed designs in clinical practice.