Institute of Medical Biometry and Informatics, Ruprecht-Karls University Heidelberg, Heidelberg, Germany.
Stat Med. 2019 Jul 30;38(17):3288-3303. doi: 10.1002/sim.8193. Epub 2019 May 9.
The performance of statistical methods is frequently evaluated by means of simulation studies. In case of network meta-analysis of binary data, however, available data-generating models (DGMs) are restricted to either inclusion of two-armed trials or the fixed-effect model. Based on data-generation in the pairwise case, we propose a framework for the simulation of random-effect network meta-analyses including multiarm trials with binary outcome. The only one of the common DGMs used in the pairwise case, which is directly applicable to a random-effects network setting uses strongly restrictive assumptions. To overcome these limitations, we modify this approach and derive a related simulation procedure using odds ratios as effect measure. The performance of this procedure is evaluated with synthetic data and in an empirical example.
统计方法的性能通常通过模拟研究来评估。然而，在网络二项数据荟萃分析中，可用的数据生成模型（DGM）仅限于包括双臂试验或固定效应模型。基于成对数据的生成，我们提出了一种用于模拟包括二分类结局多臂试验的随机效应网络荟萃分析的框架。在成对分析中使用的唯一一种可直接应用于随机效应网络环境的常见 DGM ，使用了非常严格的假设。为了克服这些限制，我们修改了这种方法，并使用优势比作为效应量来推导相关的模拟程序。使用合成数据和实证示例评估了该程序的性能。
