AN ADAPTIVE BAYESIAN APPROACH TO JOINTLY MODELING RESPONSE AND TOXICITY IN PHAS

这个子项目是许多研究子项目中的一个 由NIH/NCRR资助的中心赠款提供的资源。子项目和 研究者（PI）可能从另一个NIH来源获得了主要资金， 因此可在其他CRISP条目中表示。所列机构为 研究中心，而研究中心不一定是研究者所在的机构。 我们提出了一种新的自适应贝叶斯方法，用于I期临床试验中的剂量确定， 毒性，假设反应和毒性阈值共同遵循双变量对数正态分布 分布在癌症试验中很少有反应。但 生物反应 可能是常见的，可能有助于决定 第一阶段的升级应该有多激进在一个理想的决策理论框架中， 连续的病人将结合什么是最好的病人，连同价值的信息， 为审判而获得的。然而，明确地评估后者在计算上将是极其困难的。为 为了简单起见，我们将注意力限制在下一个患者的每个结果的概率上。该模型假设 反应和毒性事件的发生取决于个体的相应剂量阈值，并提供了一个 框架纳入有关人口阈值分布的先验信息，以及积累 数据可以分配下一个剂量以最大化预期效用。一个简单的效用函数将正效用 仅基于反应和无毒性的同时发生。计算机仿真结果表明，该设计 可靠地选择不同场景和不同先验下的优选剂量。

This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. The subproject and investigator (PI) may have received primary funding from another NIH source, and thus could be represented in other CRISP entries. The institution listed is for the Center, which is not necessarily the institution for the investigator. We present a new adaptive Bayesian method for dose-finding in phase I clinical trials based on both response and toxicity under the assumption that the thresholds of response and toxicity jointly follow a bivariate log-normal distribution. Responses are rare in cancer trials. But biological responses may be common, and may help decide how aggressive a phase I escalation should be. In an ideal decision theory framework, the choice of dose for each successive patient would incorporate what is best for the patient, together with the value of the information to be obtained for the trial. However, to evaluate the latter explicitly would be computationally extremely difficult. For simplicity, we restrict attention to the probability of each outcome for the next patient only. The model assumes that response and toxicity events happen depending on the respective dose thresholds for the individual, and provides a framework for incorporating prior information about the population threshold distribution, as well as accumulated data. The next dose can be assigned to maximize expected utility. A simple utility function places positive utility only on the co-occurrence of response and non-toxicity. Computer simulation results show that the proposed design reliably chooses the preferred dose under different scenarios and different priors.