Naumann Ken
Departments of Biology and Health Sciences, Langara College , Vancouver, Canada .
J Altern Complement Med. 2018 Feb;24(2):168-181. doi: 10.1089/acm.2017.0085. Epub 2017 Sep 15.
To quantify different aspects of the quality of reporting of herbal medicine clinical trials, to determine how that quality is affecting the conclusions of meta-analyses, and to target areas for improvement in future herbal medicine research reporting.
The Electronic databases PubMed, Academic Search Premier, ScienceDirect, and Alt HealthWatch were searched for meta-analyses of herbal medicines in refereed journals and Cochrane Reviews in the years 2000-2004 and 2010-2014. The search was limited to meta-analyses of randomized controlled trials involving humans and published in English. Judgments and descriptions within the meta-analyses were used to report on risks of bias in the included clinical trials and the meta-analyses themselves.
Out of 3264 citations, 9 journal-published meta-analyses were selected from 2000 to 2004, 116 from 2010 to 2014, and 44 Cochrane Reviews from 2010 to 2014. Across both time frames and categories of publication, <42% of the trials included in the meta-analyses described adequate randomization; <19% described concealment methods; <26% described double blinding; <29% described outcome assessment blinding, ≤53% discussed incomplete data, and <36% were nonselective in their reporting. Less than 54% of trials reported on adverse events and 64% of meta-analyses did not include a single trial with a low risk of bias. Taxonomic verification and chemical characterization of test products were infrequent in trials. Only 40% of meta-analyses considered publication bias and, of those that did, 90% found evidence for it. Cochrane Reviews were more likely than other sources to make negative conclusions of efficacy or to defer conclusions because of the absence of high quality trials.
Meta-analyses of herbal medicines include a significant number of clinical trials that do not meet the recommended standards for clinical trial reporting. This quantitative assessment identified significant publication bias and other bias risks that may be due to inadequate trial design or incomplete reporting of outcomes. Suggested improvements to herbal medicine clinical trial reporting are discussed.
量化草药临床试验报告质量的不同方面，确定该质量如何影响荟萃分析的结论，并找出未来草药研究报告中有待改进的领域。
在电子数据库PubMed、Academic Search Premier、ScienceDirect和Alt HealthWatch中检索2000 - 2004年以及2010 - 2014年发表在同行评审期刊上的草药荟萃分析和Cochrane系统评价。检索限于涉及人类的随机对照试验的荟萃分析且为英文发表。荟萃分析中的判断和描述用于报告纳入的临床试验及荟萃分析本身的偏倚风险。
在3264条引文中，2000年至2004年选出9篇期刊发表的荟萃分析，2010年至2014年选出116篇，2010年至2014年选出44篇Cochrane系统评价。在两个时间框架和各类出版物中，荟萃分析纳入的试验中<42%描述了充分随机化；<19%描述了隐匿方法；<26%描述了双盲；<29%描述了结果评估盲法，≤53%讨论了不完整数据，且<36%在报告中无选择性。少于54%的试验报告了不良事件，64%的荟萃分析未纳入任何偏倚风险低的试验。试验中对受试产品的分类验证和化学特性描述很少见。只有40%的荟萃分析考虑了发表偏倚，其中90%发现了发表偏倚的证据。Cochrane系统评价比其他来源更有可能因缺乏高质量试验而对疗效得出否定结论或推迟得出结论。
草药的荟萃分析纳入了大量不符合临床试验报告推荐标准的临床试验。这种定量评估确定了显著的发表偏倚和其他可能由于试验设计不足或结果报告不完整导致的偏倚风险。讨论了对草药临床试验报告的建议改进措施。
