在GRADE方法中，若多数相关证据来自高偏倚风险的研究，则起初被定为高质量证据的随机试验和低质量证据的观察性研究均有可能被降低质量等级。随机试验已确定的局限性包括：未进行分配隐藏、未实施盲法、未报告失访情况及未恰当考虑意向性治疗原则。最近提出的局限性包括：因明显获益而早期终止试验和基于结果选择性报告结局。观察性研究的主要局限性包括使用不合适的对照及未能充分调整预后的不平衡。偏倚风险可因不同结果而异（如全死因死亡率的失访远少于生命质量的失访），许多系统评价都容易忽略这一点。在决定是否因偏倚风险而降低质量等级时，不管是随机试验还是观察性研究，作者不应采用对各个研究取平均值的方法。相反，对任何单个结果，当同时存在高、低偏倚风险的研究时，则应考虑只纳入较低偏倚风险的研究。
The original meaning of “allocation concealment” is that the statistician and the trial designer who generated the random sequence and allocation sequence should not recruit and allocate the participants for the purpose of avoiding selection bias. In the process of generating allocation sequence, combined blinding measures are so called “setting blinding”, for instance, by using “double simulation”. We use a example to describe how to perform an adequate allocation and conceal the sequence, and setting blinding. Another example is used to describe how to report the methods of randomization, allocation concealment and the effect of blinding, etc.
ObjectiveTo analyze the current status of the application of randomization methods in randomized controlled trials (RCTs) which have been published in New England Journal of Medicine (NEJM).MethodsRCTs published by NEJM in 2018 were searched and collected. The characteristics of clinical trial design elements and the application status of randomization methods were comprehensively analyzed to distinguish blind trial and non-blind trial, and characteristics of application of randomization methods and selection of allocation concealment mechanisms of non-blind trials were summarized.ResultsA total of 151 RCTs were published in NEJM in 2018, in which blinded trials and non-blinded trials accounted for 75 (49.67%) and 76 (50.33%), respectively. 34 (22.52%) RCTs did not report specific randomization methods, and the remaining 117 (77.48%) reported. Among the latter, stratified block randomization accounted for the main body (72.65%), followed by block randomization (11.11%), minimization method (9.40%), simple randomization method (4.27%) and the others. There was no significant difference in the proportion of reporting or using randomization methods between blind and non-blind trials (P>0.05). In 76 non-blind trials, 38 (50.00%) clearly reported the concealment method of random allocation, among which 37 used the central randomization (97.37%) and 1 used envelope method (2.63%).ConclusionsThe current RCTs published in NEJM still have problems in the selection of randomization methods to be optimized and the transparency of randomization reporting to be improved.