更新与动态系统评价中Meta分析方法的范围综述
该系统评价范围综述检索主要数据库至2024年3月6日,纳入51项研究及1份配套报告,梳理更新与动态系统评价中成对Meta分析和网状Meta分析的统计方法,归为9类方法学类别。多数研究针对更新系统评价(72.5%)和成对Meta分析(98%),试验序贯分析最常见,仅1项研究涉及网状Meta分析扩展。
要点
- 更新与动态Meta分析方法可归为9类,试验序贯分析最常用。
- 多数方法关注I类错误,较少考虑II类错误或异质性。
- 网状Meta分析的更新方法证据极少,仅1项方法学研究。
核心结论
- 该范围综述首次全面梳理更新与动态Meta分析方法,现有方法多样但无单一方法在各情境下均优。
研究设计与样本
- 遵循2020年Joanna Briggs Institute方法学开展范围综述,检索MEDLINE、Embase、Cochrane Library等数据库自建库至2024年3月6日,并补充灰色文献与专家推荐,由两名评价者独立筛选与提取。
- 共纳入51项研究和1份配套报告,研究分组为9类方法学类别。
主要发现
- 纳入研究中72.5%针对更新系统评价,98%针对成对Meta分析;试验序贯分析最常见(46/51),仅1项研究描述网状Meta分析扩展。
- 各方法在错误控制和异质性调整能力上差异较大,多数方法处理I类错误,较少考虑II类错误或异质性。
关键数值与推荐
- 9类方法学类别为:传统/累积Meta分析、迭代对数律、Shuster序贯法、序贯Meta分析、试验序贯分析、随机截尾、适应性设计、似然比检验和贝叶斯Meta分析。
临床怎么用
- 选择更新或动态Meta分析方法时需关注其对I类错误、II类错误及异质性的处理能力,并根据系统评价类型匹配方法。
局限性
- 网状Meta分析更新方法的证据仅来自1项方法学研究,相关结论有限。
- 现有方法学比较多为描述或方法学对比,缺乏跨情境的实证比较。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
原文
OBJECTIVES: Systematic reviews (SRs) provide the highest level of evidence for clinical and policy decision-making but quickly become outdated as new studies emerge. Updated systematic reviews (USRs) and living systematic reviews (LSRs) offer 2 frameworks to maintain currency; however, statistical approaches for repeated updates of pairwise meta-analyses (PMAs) and network meta-analyses (NMA) remain heterogeneous. This scoping review aimed to identify, categorize, and summarize existing methods for conducting updated meta-analyses (primarily PMAs and, subject to available evidence, NMAs) in the context of USRs and LSRs, highlighting their properties, as well as their reported advantages and limitations. METHODS: A scoping review was conducted to identify studies that described or compared methods for updating PMAs or NMAs within USRs or LSRs. We followed the 2020 Joanna Briggs Institute methodology and searched across major bibliographic databases (MEDLINE, Embase, the Cochrane Library, Cochrane CENTRAL, the Cochrane Methodology Register, ERIC, and PsycINFO) from inception to March 6, 2024. Gray literature and expert recommendations supplemented our literature search. Two reviewers independently screened records, extracted data, and categorized methods. RESULTS: Of 2710 records identified, 51 studies and 1 companion report met the inclusion criteria. The studies were grouped into 9 methodological categories: Traditional/Cumulative Meta-analysis, Law of Iterated Logarithm, Shuster Sequential Method, Sequential Meta-Analysis, Trial Sequential Analysis, Stochastic Curtailment, Adaptive Design, Likelihood Ratio Test, and Bayesian meta-analysis. Most studies focused on USRs (72.5%) and PMAs (98%). Trial Sequential Analysis was the most frequently encountered approach (46/51), while only 1 study described an extension to NMA. Methods varied widely in error control and ability to adjust for heterogeneity. While most approaches addressed type I error, fewer considered type II error or heterogeneity. CONCLUSION: This scoping review provides the first comprehensive mapping of methods for updated and living meta-analyses. Despite a range of available approaches, no single method demonstrated superiority across contexts. Further research should compare methods empirically, develop accessible software to support reliable, continuously updated evidence synthesis, and extend frameworks to NMA, especially as only a single methodological study was found for NMA.
引用
- Konstantinidis M, Kontouli KM, Zhelnov P, et al. Characterizing updated and living meta-analyses part 1: a methodological scoping review on methods for conducting meta-analyses in updated and living systematic reviews. J Clin Epidemiol. 2026. PMID: 42331113. doi:10.1016/j.jclinepi.2026.112389
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。