临床试验注册并非低偏倚风险的指标:一项方法学研究
研究于2015年11月5日检索Cochrane妇科与生育组专业注册库,纳入2010至2014年发表的693项生育治疗随机对照试验。仅44%的试验注册,未注册试验样本量较小(P < 0.001)。在125项注册与125项未注册试验的子样本中,两组在随机序列生成(P = 0.001)、分配隐藏(P = 0.003)和选择性报告(P < 0.001)领域的偏倚风险存在差异,但在盲法和不完整结局数据领域无显著差异(P > 0.05)。注册试验中仅54项(43.2%)为前瞻性注册。
要点
- 693项生育治疗随机对照试验中仅44%注册,未注册试验样本量较小(P < 0.001)。
- 注册与未注册试验在随机序列生成、分配隐藏和选择性报告领域的偏倚风险存在差异。
- 125项注册试验中仅54项(43.2%)为前瞻性注册。
核心结论
- 系统评价应纳入所有可获得的试验并评估偏倚风险,因为既有高偏倚风险的注册试验,也有低偏倚风险的未注册试验,排除未注册试验将损失超过一半的可用证据。
研究设计与样本
- 于2015年11月5日检索Cochrane妇科与生育组专业注册库,纳入2010至2014年发表的生育治疗随机对照试验,要求全文发表且为英文。
- 由两名评价者独立评估试验注册状态,方法包括检索发表文献、试验注册库并联系原作者。
- 从693项符合标准的随机对照试验中随机抽取125项注册试验和125项未注册试验,使用Cochrane偏倚风险工具五个领域评估偏倚风险。
主要发现
- 693项符合标准的随机对照试验中,仅44%被注册。
- 未注册临床试验的样本量小于注册试验(P < 0.001)。
- 注册与未注册试验在随机序列生成(P = 0.001)、分配隐藏(P = 0.003)和选择性报告(P < 0.001)领域的偏倚风险存在差异。
- 注册与未注册试验在盲法和不完整结局数据领域的偏倚风险无显著差异(P > 0.05)。
- 125项注册试验中仅54项(43.2%)为前瞻性注册。
关键数值与推荐
- 693项符合标准的随机对照试验中注册比例为44%。
- 125项注册试验中前瞻性注册比例为43.2%(54/125)。
- 注册与未注册试验在随机序列生成领域的偏倚风险差异P = 0.001。
- 注册与未注册试验在分配隐藏领域的偏倚风险差异P = 0.003。
- 注册与未注册试验在选择性报告领域的偏倚风险差异P < 0.001。
临床怎么用
- 系统评价应纳入所有可获得的试验并评估其偏倚风险,不应仅依据注册状态排除未注册试验。
- 排除未注册试验将损失超过一半的可用证据。
局限性
- 本综述仅纳入英文发表的试验。
- 无法获取未注册试验的研究方案,因此无法评估选择性报告领域的偏倚风险。
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: To determine the prevalence of registered trials and to evaluate the risk of bias between registered and unregistered clinical trials. STUDY DESIGN AND SETTING: The Cochrane Gynecology and Fertility Group's specialized register was searched on November 5, 2015, for randomized controlled trials published from 2010 to 2014. Studies were selected if they had randomized women or men for fertility treatments, were published in full text and written in English. Two reviewers then independently assessed trial registration status for each trial, by searching the publication, trial registries, and by contacting the original authors. RESULTS: Of 693 eligible randomized controlled trials, only 44% were found to be registered. Unregistered clinical trials had smaller sample sizes than registered trials (P < 0.001). A random subsample of 125 registered and 125 unregistered trials was assessed for risk of bias using five of the Cochrane Risk of Bias "domains." Registered and unregistered trials differed in their risk of bias for random sequence generation (P = 0.001), allocation concealment (P = 0.003), and selective reporting (P < 0.001) but not blinding or incomplete outcome data (P > 0.05) domains. Only 54 (43.2%) of the 125 registered trials were registered prospectively. This study has the following limitations. Only English language trials were included in this review. We were unable to obtain protocols for the unregistered trials and therefore were unable to assess the risk of bias in the selective reporting domain. CONCLUSIONS: All available trials should be included in systematic reviews and assessed for risk of bias as there are both registered trials with high risk of bias and unregistered trials with low risk of bias and by excluding unregistered trials more than half of the available evidence will be lost.
引用
- Farquhar CM, Showell MG, Showell EAE, et al. Clinical trial registration was not an indicator for low risk of bias. J Clin Epidemiol. 2017. PMID: 28088595. doi:10.1016/j.jclinepi.2016.11.011
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。