Department of Biomedical Data Science and Center for Innovative Study Design, School of Medicine, Stanford University, Stanford, CA, USA; Department of Neurology, UMC Utrecht Brain Center, University Medical Center Utrecht, Utrecht, The Netherlands.
Department of Health Evidence, Section Biostatistics, Radboud Medical Center Nijmegen, Nijmegen, The Netherlands.
J Clin Epidemiol. 2022 Jul;147:32-39. doi: 10.1016/j.jclinepi.2022.03.009. Epub 2022 Mar 26.
Correlated longitudinal and time-to-event outcomes, such as the rate of cognitive decline and the onset of Alzheimer's disease, are frequent (co-)primary and key secondary endpoints in randomized clinical trials (RCTs). Despite their biological associations, these types of data are often analyzed separately, leading to loss of information and increases in bias. In this paper, we set out how joint modeling of longitudinal and time-to-event endpoints can be used in RCTs to answer various research questions.
The key concepts of joint models are introduced and illustrated for a completed trial in amyotrophic lateral sclerosis.
The output of a joint model can be used to answer different clinically relevant research questions, where the interpretation of effect estimates and those obtained from conventional methods are similar. Albeit joint models have the potential to overcome the limitations of commonly used alternatives, they require additional assumptions regarding the distributions, as well as the associations between two endpoints.
Improving the uptake of joint models in RCTs may start by outlining the exact research question one seeks to answer, thereby determining how best to prespecify the model and defining the parameter that should be of primary interest.
相关性纵向和事件时间结局，如认知能力下降的速度和阿尔茨海默病的发病，是随机临床试验（RCT）中常见的（共同）主要和关键次要终点。尽管这些数据具有生物学相关性，但通常会分开分析，导致信息丢失和偏倚增加。本文介绍了如何在 RCT 中联合建模纵向和事件时间终点，以回答各种研究问题。
介绍了联合模型的关键概念，并对肌萎缩侧索硬化症的已完成试验进行了说明。
联合模型的输出可用于回答不同的临床相关研究问题，其效应估计值的解释与传统方法相似。尽管联合模型有可能克服常用替代方法的局限性，但它们需要对分布以及两个终点之间的关联做出额外的假设。
通过概述希望回答的确切研究问题，可以提高 RCT 中联合模型的应用，从而确定如何最好地预先指定模型并定义主要关注的参数。
