Design of Cancer Phase I/II Clinical Trials Using Drug Combinations of Cytotoxic and Biologic Agents

描述(由申请人提供)：这项拟议研究的动机是需要新的I期和II期癌症临床试验设计，探索两种或两种以上具有连续剂量水平的细胞毒性和生物制剂的药物组合的安全性和有效性。目前关于这一主题的研究依赖于在两种药物的相对较少的剂量组合的固定先验集合中寻找最佳剂量。此外，这些方法中的大多数建议使用单剂组合作为第二阶段剂量。因此，这些设计可能无法确定真正的最优毒性和有效性剂量组合，并且随着剂量组合数量和试剂数量的增加需要更大的样本量。这些方法也缺乏用户友好的软件。我们的建议包括三个目标。在目标1中，我们将开发几种使用贝叶斯自适应设计的方法来估计两种或更多细胞毒性/生物制剂的最大耐受剂量(MTD)曲线或曲面。我们将考虑以下情况：(I)剂量限制毒性(DLT)是二元的，在一个治疗周期内存在或不存在DLT，以及(Ii)DLT是事件发生的时间，也称为迟发性毒性。在目标2中，我们将为早期癌症临床试验开发贝叶斯适应性设计，以估计两种或两种以上细胞毒性/生物制剂的最佳毒性和疗效轮廓。我们将首先开发无缝的I/II阶段设计，用于估计几种药物的MTD曲线或等高线，并确定在两阶段设计中使疗效最大化的最佳剂量组合(S)。当在几个疗程后评估治疗效果时，这样的设计是相关的。由于第一阶段和第二阶段试验的受试者群体可能非常不同，我们将进一步研究替代设计，这些设计可以同时确定在不超过预先指定的毒性阈值的情况下最大化疗效的剂量组合区域。当初步疗效被评估为一个治疗周期内的生物标记物调节时，这些方法也适用。DLT和效率端点将被建模为二进制和事件结果的时间。在目标3中，我们将部署R包和网络应用程序来实施目标1-2中提出的设计，并评估预期试验的操作特性。这项提案的研究人员有丰富的经验来开发使用贝叶斯适应性设计EWOC和CRM的I期临床试验的方法学。他们在与临床医生合作设计和进行癌症剂量发现试验方面拥有数十年的经验。研究人员还开发了独立的用户友好软件，用于使用EWOC和CRM设计单剂癌症I期试验，并为EWOC设计开发了一个完全集成的基于Web的应用程序。

DESCRIPTION (provided by applicant): This proposed research is motivated by the need for novel phase I and II cancer clinical trial designs exploring safety and efficacy of drug combinations of two or more cytotoxic and biologic agents with continuous dose levels. Current research on this subject relies on a search for optimal doses among a fixed a priori set of a relatively small number of dose combinations of two agents. Moreover, most of these methods recommend a single dose combination as the phase II dose. As a result, these designs can fail to identify the true optimal toxicity and efficacy dose combinations and require larger sample sizes as the number of dose combinations and number of agents increase. These methods also lack user friendly software. Our proposal consists of three aims. In aim 1, we will develop several methods using Bayesian adaptive designs known as escalation with overdose control (EWOC) and the continual reassessment method (CRM) to estimate the maximum tolerated dose (MTD) curve or surface of two or more cytotoxic/biologic agents. We will consider cases where (i) the dose limiting toxicity (DLT) is binary, presence or absence of DLT within one cycle of therapy and (ii) DLT is time to event, also known as late onset toxicity. In Aim 2, we will develop Bayesian adaptive designs for early phase cancer clinical trials to estimate the optimal toxicity and efficacy contour of two or more cytotoxic/biologic agents. We will first develop seamless phase I/II designs for estimating the MTD curve or contours of several agents and identify the optimal dose combination(s) that maximize efficacy in a two-stage design. Such designs are relevant when treatment efficacy is assessed after few cycles of treatment. Because the population of subjects in phase I and II trials are likely to be very different, we wil further study alternative designs that simultaneously identify dose combination regions that maximize efficacy while not exceeding a pre-specified threshold of toxicity. These are also applicable when preliminary efficacy is assessed as biomarker modulation within one cycle of treatment. DLT and efficacy endpoints will be modeled as binary and time to event outcomes. In Aim 3, we will deploy R packages and web applications to implement designs proposed in Aim 1-2 and evaluate operating characteristics of prospective trials. The investigators of this proposal have extensive experience developing methodology for phase I clinical trials using the Bayesian adaptive designs EWOC and CRM. They have decades of experience collaborating with clinicians in designing and conducting dose finding trials in cancer. The investigators have also developed stand-alone user friendly software for designing single agent cancer phase I trials using EWOC and CRM and a fully integrated Web based application for the EWOC design.