Cramer Holger, Langhorst Jost, Dobos Gustav, Lauche Romy
Department of Internal and Integrative Medicine, Kliniken Essen-Mitte, Faculty of Medicine, University of Duisburg-Essen, Essen, Germany.
Australian Research Centre in Complementary and Integrative Medicine (ARCCIM), University of Technology Sydney, Sydney, Australia.
PLoS One. 2015 Dec 2;10(12):e0144125. doi: 10.1371/journal.pone.0144125. eCollection 2015.
Bias in randomized controlled trials (RCTs) of complementary therapy interventions seems to be associated with specific factors and to potentially distort the studies' conclusions. This systematic review assessed associated factors of risk of bias and consequences for the studies' conclusions in RCTs of yoga as one of the most commonly used complementary therapies.
Medline/PubMed, Scopus, IndMED and the Cochrane Library were searched through February 2014 for yoga RCTs. Risk of selection bias was assessed using the Cochrane tool and regressed to a) publication year; b) country of origin; c) journal type; and d) impact factor using multiple logistic regression analysis. Likewise, the authors' conclusions were regressed to risk of bias.
A total of 312 RCTs were included. Impact factor ranged from 0.0 to 39.2 (median = 1.3); 60 RCT (19.2%) had a low risk of selection bias, and 252 (80.8%) had a high or unclear risk of selection bias. Only publication year and impact factor significantly predicted low risk of bias; RCTs published after 2001 (adjusted odds ratio (OR) = 12.6; 95% confidence interval (CI) = 1.7, 94.0; p<0.001) and those published in journals with impact factor (adjusted OR = 2.6; 95%CI = 1.4, 4.9; p = 0.004) were more likely to have low risk of bias. The authors' conclusions were not associated with risk of bias.
Risk of selection bias was generally high in RCTs of yoga; although the situation has improved since the publication of the revised CONSORT statement 2001. Pre-CONSORT RCTs and those published in journals without impact factor should be handled with increased care; although risk of bias is unlikely to distort the RCTs' conclusions.
补充疗法干预的随机对照试验（RCT）中的偏倚似乎与特定因素相关，并可能扭曲研究结论。作为最常用的补充疗法之一，本系统评价评估了瑜伽RCT中偏倚风险的相关因素及其对研究结论的影响。
检索Medline/PubMed、Scopus、IndMED和Cochrane图书馆截至2014年2月的瑜伽RCT。使用Cochrane工具评估选择偏倚风险，并通过多因素逻辑回归分析将其与以下因素进行回归分析：a）发表年份；b）原产国；c）期刊类型；d）影响因子。同样，将作者的结论与偏倚风险进行回归分析。
共纳入312项RCT。影响因子范围为0.0至39.2（中位数=1.3）；60项RCT（19.2%）选择偏倚风险较低，252项（80.8%）选择偏倚风险较高或不明确。只有发表年份和影响因子能显著预测低偏倚风险；2001年后发表的RCT（调整比值比（OR）=12.6；95%置信区间（CI）=1.7,94.0；p<0.001）以及发表在有影响因子期刊上的RCT（调整OR=2.6；95%CI=1.4,4.9；p=0.004）更有可能具有低偏倚风险。作者的结论与偏倚风险无关。
瑜伽RCT的选择偏倚风险普遍较高；尽管自2001年修订的CONSORT声明发表以来情况有所改善。CONSORT声明发布前的RCT以及发表在无影响因子期刊上的RCT应格外谨慎对待；尽管偏倚风险不太可能扭曲RCT的结论。
