不完善依从性新药临床试验中疗效评价的工具变量估计

Estimating Treatment Effects by Instrumental Variables under Non-Compliance in Randomized Clinical Trials

随机对照临床试验中的意向治疗分析总是假设受试者具有完善的依从性,再来比较治疗组和对照组之间的疗效差异。然而,在临床试验中不依从是普遍存在的,受试者的依从性与试验结果的质量密切相关,不依从或依从性差是导致治疗无效的最常见的原因,在进行药物的疗效评价时,也是造成偏倚的一个重要的因素。本研究考虑在不完善依从性条件下,用工具变量给出药物疗效的评价方法。

In randomied clinical trials, the intention to treat approach has long served as a standard methodological procedure, which just compare response distributions by assignment, ignoring information on compliance. In fact, however, non-compliance is common in many trials. Moreover, compliance is related to the quality of treatment evaluation＇ Many erroneous conclusion is usually attributed to non--compliance, and non-compliance can also lead to a bias on treatment effects evaluating. This paper presents an estimator of treatment effects by instrumental variables under non-compliance.