MRC Integrative Epidemiology Unit (IEU), University of Bristol, Bristol, UK
School of Psychological Science, University of Bristol, Bristol, UK.
BMJ Evid Based Med. 2023 Apr;28(2):103-110. doi: 10.1136/bmjebm-2022-112006. Epub 2022 Dec 8.
To identify whether Mendelian randomisation (MR) studies are appropriately conducted and reported in enough detail for other researchers to accurately replicate and interpret them.
Cross-sectional meta-epidemiological study.
Web of Science, EMBASE, PubMed and PsycINFO were searched on 15 July 2022 for literature.
Full research articles that conducted an MR analysis exclusively using individual-level UK Biobank data to obtain a causal estimate of the exposure-outcome relationship (for no more than ten exposures or outcomes).
Data were extracted using a 25-item checklist relating to reporting and methodological quality (based on the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE)-MR reporting guidelines and the guidelines for performing MR investigations). Article characteristics, such as 2021 Journal Impact Factor, publication year, journal word limit/recommendation, whether the MR analysis was the primary analysis, open access status and whether reporting guidelines were followed, were also extracted. Descriptive statistics were calculated for each item, and whether article characteristics predicted overall article completeness was investigated with linear regression.
116 articles were included in this review. The proportion of articles which reported complete information/adequate methodology ranged from 3% to 100% across the different items. Palindromic variants, variant replication, missing data, associations of the instrumental variable with the exposure or outcome and bias introduced by two-sample methods used on a single sample were often not completely addressed (<11%). There was no clear evidence that article characteristics predicted overall completeness except for primary analysis status.
The results identify areas in which the reporting and conducting of MR studies needs to be improved and also suggest researchers do not make use of supplementary materials to sufficiently report secondary analyses. Future research should focus on the quality of code and analyses, attempt direct replications and investigate the impact of the STROBE-MR specifically.
确定孟德尔随机化（MR）研究是否以足够详细的方式进行和报告，以便其他研究人员能够准确复制和解释这些研究。
横断面荟萃流行病学研究。
2022 年 7 月 15 日，在 Web of Science、EMBASE、PubMed 和 PsycINFO 上搜索文献。
仅使用英国生物库个体水平数据进行 MR 分析的全研究文章，以获得暴露-结局关系的因果估计（不超过 10 个暴露或结局）。
使用与报告和方法学质量相关的 25 项清单提取数据（基于加强观察性研究的报告标准（STROBE）-MR 报告指南和执行 MR 研究的指南）。还提取了文章特征，如 2021 年期刊影响因子、出版年份、期刊字数限制/建议、MR 分析是否为主要分析、开放获取状态以及是否遵循报告指南。用线性回归分析文章特征是否预测文章整体完整性。
本综述共纳入 116 篇文章。不同项目中，报告完整信息/充分方法的文章比例从 3%到 100%不等。回文变异体、变异体复制、缺失数据、工具变量与暴露或结局的关联以及在单个样本上使用两样本方法引入的偏倚通常未得到充分解决（<11%）。除主要分析状态外，没有明确证据表明文章特征预测整体完整性。
结果确定了需要改进 MR 研究报告和实施的领域，并表明研究人员没有利用补充材料充分报告二次分析。未来的研究应侧重于代码和分析的质量，尝试直接复制，并研究 STROBE-MR 的具体影响。
