From the Department of Surgery and Perioperative Care, Dell Medical School at the University of Texas at Austin, Austin, Texas.
Anesth Analg. 2019 Mar;128(3):575-583. doi: 10.1213/ANE.0000000000004014.
Clinicians encounter an ever increasing and frequently overwhelming amount of information, even in a narrow scope or area of interest. Given this enormous amount of scientific information published every year, systematic reviews and meta-analyses have become indispensable methods for the evaluation of medical treatments and the delivery of evidence-based best practice. The present basic statistical tutorial thus focuses on the fundamentals of a systematic review and meta-analysis, against the backdrop of practicing evidence-based medicine. Even if properly performed, a single study is no more than tentative evidence, which needs to be confirmed by additional, independent research. A systematic review summarizes the existing, published research on a particular topic, in a well-described, methodical, rigorous, and reproducible (hence "systematic") manner. A systematic review typically includes a greater range of patients than any single study, thus strengthening the external validity or generalizability of its findings and the utility to the clinician seeking to practice evidence-based medicine. A systematic review often forms the basis for a concomitant meta-analysis, in which the results from the identified series of separate studies are aggregated and statistical pooling is performed. This allows for a single best estimate of the effect or association. A conjoint systematic review and meta-analysis can provide an estimate of therapeutic efficacy, prognosis, or diagnostic test accuracy. By aggregating and pooling the data derived from a systemic review, a well-done meta-analysis essentially increases the precision and the certainty of the statistical inference. The resulting single best estimate of effect or association facilitates clinical decision making and practicing evidence-based medicine. A well-designed systematic review and meta-analysis can provide valuable information for researchers, policymakers, and clinicians. However, there are many critical caveats in performing and interpreting them, and thus, like the individual research studies on which they are based, there are many ways in which meta-analyses can yield misleading information. Creators, reviewers, and consumers alike of systematic reviews and meta-analyses would thus be well-served to observe and mitigate their associated caveats and potential pitfalls.
临床医生即使在其狭隘的专业领域或兴趣范围内，也会遇到呈指数级增长且频繁地令人应接不暇的信息量。鉴于每年发表的大量科学信息，系统评价和荟萃分析已经成为评估医疗干预措施和提供循证最佳实践不可或缺的方法。因此，本基础统计学教程重点介绍循证医学背景下系统评价和荟萃分析的基本原理。即使正确执行，单个研究也不过是初步证据，需要通过额外的独立研究来证实。系统评价以描述性、系统性、严谨性和可重复性（因此是“系统性”）的方式总结特定主题已发表研究的现有研究。系统评价通常包括比任何单一研究更广泛的患者范围，从而增强其研究结果的外部有效性或普遍性以及对寻求循证医学实践的临床医生的实用性。系统评价通常是同时进行荟萃分析的基础，在荟萃分析中，对所确定的一系列单独研究的结果进行汇总和统计合并。这可以得出对效应或关联的单一最佳估计值。联合系统评价和荟萃分析可以提供治疗效果、预后或诊断测试准确性的估计值。通过汇总和合并系统评价中得出的数据，精心设计的荟萃分析实质上可以提高统计推断的准确性和确定性。由此产生的单一最佳效应或关联估计值有助于临床决策和循证医学实践。精心设计的系统评价和荟萃分析可以为研究人员、政策制定者和临床医生提供有价值的信息。然而，在进行和解释它们时存在许多关键的注意事项，因此，与它们所基于的个别研究一样，荟萃分析可能会产生误导性信息。系统评价和荟萃分析的创作者、评论者和使用者都应该观察并减轻它们相关的注意事项和潜在陷阱。
