Abstract

In a recent paper, Larsen and Josiassen (2020) present a new estimand for the analysis of clinical trials with non-compliance. The treatment effect in patients who would comply if treated has the advantage of being less remote from the observed data than other principal stratum estimands (since we know who belongs to this stratum in the active treatment group). Nevertheless, we feel that the key cross-world assumption invoked by the authors to identify this treatment effect is not transparent in the original publication and in many cases implausible. The purpose of this research letter is to highlight this assumption, so that researchers can critically assess it before applying their method. We also clarify how this assumption fundamentally differs from the assumption in Stuart and Jo (2015), which may explain the discrepancy between the two approaches in Larsen and Josiassen's simulation studies.

摘要

在最近的一篇论文中，Larsen和Josiassen（2020）为分析不符合项的临床试验提出了新的估计。如果愿意接受治疗的患者，其治疗效果具有以下优势：与其他主要阶层估计相比，与观察数据的距离更小（因为我们知道谁属于活跃治疗组中的该阶层）。尽管如此，我们认为作者为确定这种治疗效果而调用的关键跨界假设在原始出版物中并不透明，在许多情况下也不可信。这封研究信的目的是强调这一假设，以便研究人员可以在应用其方法之前对其进行严格的评估。我们还阐明了这一假设与Stuart和Jo（2015）的假设有何根本不同，