Department of Anesthesiology, pain and palliative care, Radboud university medical center, Nijmegen, The Netherlands.
Systematic Review Centre for Laboratory animal Experimentation (SYRCLE), Department for Health Evidence, Radboud Institute for Health Sciences, Radboud university medical center, Nijmegen, The Netherlands.
Res Synth Methods. 2022 Nov;13(6):790-806. doi: 10.1002/jrsm.1578. Epub 2022 Jul 8.
Since the early 1990s the number of systematic reviews (SR) of animal studies has steadily increased. There is, however, little guidance on when and how to conduct a meta-analysis of human-health-related animal studies. To gain insight about the methods that are currently used we created an overview of the key characteristics of published meta-analyses of animal studies, with a focus on the choice of effect size measures. An additional goal was to learn about the rationale behind the meta-analysis methods used by the review authors. We show that important details of the meta-analyses are not fully described, only a fraction of all human-health-related meta-analyses provided rationales for their decision to use specific effect size measures. In addition, our data may suggest that authors make post-hoc decisions to switch to another effect size measure during the course of their meta-analysis, and possibly search for significant effects. Based on analyses in this paper we recommend that review teams: 1) publish a review protocol before starting the conduct of a SR, prespecifying all methodological details (providing special attention to the planned meta-analysis including the effect size measure and the rational behind choosing a specific effect size, prespecifying subgroups and restricting the number of subgroup analyses), 2) always use the Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) checklist to report your SR of animal studies, and 3) use the random effects model (REM) in human-health-related meta-analysis of animal studies, unless the assumptions for using the fixed effect model (FEM) are all met.
自 20 世纪 90 年代初以来，系统评价（SR）动物研究的数量稳步增加。然而，对于何时以及如何对与人类健康相关的动物研究进行荟萃分析，几乎没有指导。为了深入了解当前使用的方法，我们创建了一个已发表的动物研究荟萃分析的关键特征概述，重点是选择效应量度量。另一个目标是了解评论作者使用的荟萃分析方法背后的基本原理。我们表明，荟萃分析的重要细节没有得到充分描述，只有一小部分与人类健康相关的荟萃分析为他们决定使用特定效应量度量提供了理由。此外，我们的数据可能表明，作者在荟萃分析过程中做出了事后决定，转而使用另一种效应量度量，并可能寻找显著效果。基于本文的分析，我们建议审查团队：1）在开始进行 SR 之前发布审查方案，预先规定所有方法细节（特别注意计划中的荟萃分析，包括效应量度量和选择特定效应量度量的基本原理，预先规定亚组并限制亚组分析的数量），2）始终使用系统评价和荟萃分析的首选报告项目（PRISMA）清单报告您的动物研究 SR，3）除非满足使用固定效应模型（FEM）的所有假设，否则在与人类健康相关的动物研究荟萃分析中使用随机效应模型（REM）。
