摘要：目的 全面评价中药疗法治疗男性不育非随机研究的偏倚风险。 方法 电子检索Cochrane图书馆临床对照试验资料库(CCRCT)、PubMed、Embase、中国生物医学文献数据库(Sinomed)，截至2013年2月，限定为中文与英语文献，采取主题词检索与自由词检索相结合，以“中医、中药、补充疗法”及“男性不育”等为检索词。纳入中药疗法治疗男性不育的非随机对照试验(nonRCT)。由2名评价者评估纳入研究偏倚风险，采用Cochrane非随机干预研究偏倚风险评估工具(ACROBAT-NRSI)进行评价，同时对应用中的问题展开讨论。 结果 ①共纳入53个非随机研究(non-randomized studies，NRS)(CCRCT 8个、PubMed 4 个、Sinomed 41个)，包括7 159例男性不育患者，试验平均样本量为135例(15～208例)。不育诊断主要为少、弱精子症及免疫性不育。对照组干预包括：西药疗法29个研究(54.72%)、手术3个研究(5.66%)、空白对照1个研究，其余为中医药疗法20个研究(37.74%)。② 53个研究总体偏倚风险均高，其中20个研究总体偏倚风险“极高”(37.74%)。3个偏倚域(混杂偏倚、结局测量偏倚、选择性报告结果偏倚)偏倚风险较高。③单个研究偏倚风险比例：“混杂偏倚”高风险占73.58%、极高风险占16.98%；“选择受试者纳入研究偏倚”中风险86.8%；“干预测量偏倚”低风险67.92%、中风险30.19%；“意向治疗退出偏倚”低风险88.68%、高风险11.32%；“丢失数据偏倚”低风险86.79%；“结局测量偏倚”高风险100%；“选择性报告结果偏倚”高风险83.02%、中风险11.32%；总体偏倚高风险63.26%、极高风险37.74%。所有研究偏倚风险比例：低偏倚风险35.58%、中偏倚风险19.41%、高偏倚风险12.94%、极高偏倚风险4.85%。④14个研究(26.42%)未有证据支持其组间均衡性良好(偏倚风险“极高”)，而未对关键的混杂因素予以报道。仅12个研究(22.64%)陈述了明确依据的中医辨证，大多数研究缺乏中医理论的基本运用。6个研究(“高”偏倚风险，11.32%)发生病例脱落或退出，研究者仅采用意向治疗分析而未对丢失数据采用措施，导致结局效益发生偏倚。41个研究(77.36%)在不育干预的关键指标结局方面缺乏相关信息，从而限制其研究结论的临床价值。 结论 ACROBAT-NRSI可用于评估中药治疗男性不育非随机研究偏倚风险，但会遭遇中医药学固有特征的限制，研究总体偏倚风险高。建议在选择受试者纳入、干预测量及结局测量3个偏倚域优化。此外，评价中医药临床研究偏倚风险过程中，建议纳入具有中医学背景学者。

Abstract: Objective As a severe public health problem, male infertility has always been the point at issue on how to diagnose and treat effectively. Traditional Chinese medicines have been used to treat male infertility with a long history and rich experiences, however, their effects have not yet been evaluated rigorously up to now. This paper aims to systematically evaluate the risks of bias of non-randomized studies(NRS) of traditional Chinese medicines in the management of male infertility, so as to provide invaluable information for further clinical and scientific references. Methods We searched the data about male infertility and traditional Chinese medicines from the Cochrane central register of controlled trials (CCRCT), PubMed, Embase and Chinese biomedical literature database (Sinomed)until February 30, 2013. Literature search strategies took the form of combination of subject headings and freewords such as “traditional Chinese medicine”, “Chinese medicine”, “complementary therapies” and “male infertility” as search words. Only NRS of traditional Chinese medicine therapies for male infertility were included. Two reviewers (Xiong, Li) independently screened the studies for eligibility and evaluated the risks of bias of the included studies based on Cochrane risk of bias assessment tool for non-randomized studies of interventions (ACROBAT-NRSI), and meanwhile, the discussion on related problems were conducted. Results Fifty-three NRS (8 studies from CCRCT,4 studies from PubMed,and 41 studies from Sinomed), which were published in Chinese and English and involved 7 159 men (average 135 patients per study) with infertility, were included in the research. The diagnosis of the included studies mainly included oligospermia, asthenozoospermia and immunological infertility. The control interventions included Western medicine therapies in 29 studies (54.72%), surgery in 3 studies (5.66%), blank control in 1 study and traditional Chinese medicine therapies in 20 studies (37.74%). The risks of bias for all the 53 studies were serious, and especially, the risks of bias of 20 studies were critical (37.74%).The bias risk domains of single study showed higher proportions, and the serious and critical risks of bias due to confounding were 73.58% and 16.98%, respectively. The moderate risk of bias in selection of participants into the study was 86.8%, and the low and moderate risks of bias in measurement of interventions were 67.92% and 30.19%, respectively. The low and serious risks of bias due to departures from intended interventions were 88.68% and 11.32%, respectively. The low risk of bias due to missing data was 86.79%, and serious risk of bias in measurement of outcomes was 100%. The serious and moderate risks of bias in selection of the reported result were 83.02% and 11.32%, respectively.The serious and critical risks of bias in the overall judgment about risk of bias were 63.26% and 37.74%, respectively. The low, moderate, serious and critical risks of bias for all 512 studies were 35.58%, 19.41%, 12.94% and 4.85%, respectively. The research suggested that fourteen studies (26.42%) were lack of evidence for supporting the proportionality between the groups with critical bias risks, and the key confounding factors had not been reported. Only 12 studies (22.64%) stated the traditional Chinese medicine syndrome differentiation definitely, and most did not use the Chinese medicine theory. Six studies reported shedding and withdrawal cases, butonlythe intention-to-treat analysis was used not considering the missing data and this resulted in bias. Forty-one studies (77.36%) had no information of key outcome data for interventions, so the clinical applicability was limited. Conclusion ACROBAT-NRSI is suitable for evaluating the bias risks of NRS of traditional Chinese medicines for male infertility,but there are some limitations of inherent characteristics of traditional Chinese medicines, and the bias risks are serious. It is advised to improve and optimize the ACROBAT-NRSI in the bias domains of selection of participants into the study, measurement of interventions and measurement of outcomes. Moreover, the experts with subject-matter knowledge of traditional Chinese medicines should be included into the appraisal process.