Copula模型在两药物剂量选择研究中的应用

Application of Copula Model in Two-Agent Dose-Finding Studies

摘要: 本文应用Copula模型对两药物剂量组合的毒性概率关系进行建模，基于极大似然方法研究第I阶段临床试验中两药物组合的剂量选择问题。在有限样本的前提下，为了确保模型参数的极大似然估计存在，本文提出了两阶段的试验设计策略，为了缩短试验周期和节约试验成本，本文对进入试验的患者采用按组自然增长方式。本文根据目标最大耐受剂量组合的数量和位置不同设置了6个毒性场景，模拟研究了本文提出的方法在不同毒性场景中的操作特性和表现。本文提出的方法在6种不同的毒性场景下，基于Copula模型的极大似然方法以较高的概率选择目标剂量组合或低于目标剂量的剂量组合作为最大耐受剂量组合，大部分的患者都被分配在目标剂量组合或低于目标剂量的剂量组合下接受治疗，从而保证了患者的安全性。

Abstract: This study applies the Copula model to characterize the toxicity probability relationship between two drug dose combinations, employing the maximum likelihood method to investigate dose selection in Phase I clinical trials. Under limited sample size conditions, to ensure the existence of maximum likelihood estimates for model parameters, a two-stage trial design strategy is proposed. To shorten the trial duration and reduce costs, patients are enrolled in a group-wise natural increment manner. Six toxicity scenarios are established based on the quantity and location of target maximum tolerated combinations. The simulation studies are conducted to evaluate the operational characteristics and performance of the proposed method across these scenarios. The proposed approach, based on the Copula model and maximum likelihood method, demonstrates a high probability of selecting the target dose combination or a sub-target dose combination as the maximum tolerated dose combination across all six toxicity scenarios. Most patients are allocated to the target dose combination or sub-target dose combinations, thereby ensuring patient safety.