cochrane纳入的RCT文献质量评价(风险偏倚评估工具)中英文对照版

1、偏倚类型判断指标评价员的判断选择偏倚随机序列的产生足够详细的描述用于生成分配序 列的方法，以评估产生的分组是否 具有可比性。生成随机序列不充分， 发生选择偏倚分配隐藏足够详细的描述隐藏分配序列的 方法，以决定干预的分配在纳入之 前或纳入过程中是否可见分配前分配隐臧不充分 发生选择偏倚实施偏倚头施者和参与者 双盲应对每个主 要结局进行评估或分类结局如果有，描述对参与者和头施者仃 盲法，防止其了解干预信息的所有 措施。提供任何与所实施的盲法是 否有效地相关信息。参与者和实施者了解干 预的相关信息导致实施 偏倚测量偏倚结局评估中的盲 法每个主要结局 均应评估或分类 结局如果有，描述对结局者行盲法，防

2、 止其了解自己所接受的干预信息 的所有措施。提供任何与所实施的 盲法是否有效地相关信息。结局评估者了解分配的 干预措施将导致测量偏 倚失访偏倚不全结局数据每 个主要结局均应 评估或分类结 局描述每个主要结局数据的完整性， 包括分析中的自然缺失和排除。这 些缺失数据是否报告，在各个干预 组的数目并与总样本量比拟， 数据缺失以及重新纳入分析的原 因不全结局数据的数量， 性质，处理方式导致失 访偏倚发表偏倚Selective report ing.说明如何审查选择性报道结局的 可能性，以及审查结果选择性报道结局导致发 表偏倚其它偏倚其它偏倚来源说明不包括在上述偏倚中的其它 重要偏倚如果特定的问题或条

3、目事先在计 划书中指出，应对每一项说明不包括在上述各项中的 偏倚中文：Table 8.5.a: The Cochra ne Collaborati ons tool for assess ing risk of biasRisk of biasTable 8.5.d: Criteria for judgi ng risk of bias in the assessme nt tool随机序列的产生随机序列产生不充分导致选择偏倚判断为低风险的标准研究者描述随机序列产生过程譬如：?参考随机数字表?使用电脑随机数字生成器?扔硬币?洗牌的卡片和信圭寸?掷骰子? 抽签? 最小化*最小化，可实现无随机元素

4、，被认为相当于是随机的。判断为高风险的标准研究者描述序列的产生使用的是非随机的方法。通常是 系统的非随机方法，例如：?通过奇偶或出生日期产生序列?通过入院日期产生序列?通过类似住院号或门诊号产生序列相对于上面提到的系统方法，其它非随机的方法少见的 多，也更明显。通常包括对参与者进行判断或非随机的 方法，例如：?临床医生判断如何分配?参与者判断如何分配?基于实验室检查或系列测试的结果分配?基于干预的可获取性进行分配偏倚风险不清楚的判断标准没有足够的信息判断随机序列的产生存在高风险或低风 险分配隐藏分配前不充足的分配隐藏导致选择偏倚低风险判断标准参与者以及纳入参与者的研究者因以下掩盖分配的方法 或

5、相当的方法，事先不了解分配情况?中心分配包括，网络，药房控制随机?相同外形的顺序编号的药物容器；?顺序编号、不透明、密封的信封高风险判断标准参与者以及纳入参与者的研究者可能事先知道分配，因 而引入选择偏倚，譬如基于如下方法的分配：?使用摊开的随机分配表如随机序列清单?分发信封但没有适宜的平安保障如透明、非密 封、非顺序编号?交替或循环?出生日期? 病历号?其它明确的非隐藏过程风险未知没有足够信息判断为低风险或高风险。通常因分配隐藏 的方法未描述或描述不充分。例如描述为使用信封分配， 但为描述信封是否透明？密封？顺序编号？对参与者和实施者的盲法因参与者和实施者了解干预情况而导致实施偏倚偏倚低风险

6、标准任何如下标准：? 无盲法或盲法不充分，但系统评价员判断结局不 太可能受到缺乏盲法的影响?参与者和主要实施者均实施可靠的盲法，且盲法 不太可能被打破偏倚咼风险标准任何如下标准：?无盲法或盲法不充分，但系统评价员判断结局很 可能受到缺乏盲法的影响?尝试对关键的参与者和实施者行盲法，但盲法很 可能被打破，结局很可能受到缺乏盲法的影响风险未知任何如下标准：?没有足够信息判断为低风险或高风险?研究未描述此情况对结局评价实施盲法结局评价者了解干预分配信息将导致测量偏倚偏倚低风险标准任何如下标准：? 无盲法或盲法不充分，但系统评价员判断结局不 太可能受到缺乏盲法的影响?参与者和主要实施者均实施可靠的盲法

7、，且盲法 不太可能被打破高风险判断标准任何如下标准：?无盲法或盲法不充分，但系统评价员判断结局很 可能受到缺乏盲法的影响?尝试对关键的参与者和实施者行盲法，但盲法很 可能被打破，结局很可能受到缺乏盲法的影响风险未知任何如下标准：?没有足够信息判断为低风险或高风险?研究未描述此情况结局数据不完整不全结局数据的数量，性质，处理方式导致失访偏倚偏倚低风险标准任何如下标准：?无缺失数据?缺失数据的产生不大可能与真实结局相关对于 生存数据，删失不大可能引入偏倚?缺失数据的数目在各干预组相当，且各组缺失原 因类似?对二分类变量，与观察事件的发生风险相比，缺 失比例缺乏以影响预估的干预效应?对连续性结局数据

8、，缺失数据的合理效应规模均 数差或标准均数差不会大到影响观察的效应规 模；?缺失的数据用适宜的方法进行估算高风险判断标准任何如下标准：?缺失数据的产生很大可能与真实结局相关，缺失 数据的数目及缺失原因在各干预组相差较大?对二分类变量，与观察事件的发生风险相比，缺 失比例足以影响预估的干预效应?对连续性结局数据，缺失数据的合理效应规模均 数差或标准均数差足以影响观察的效应规模；? 意向治疗分析中存在实际干预措施与随机分配 的干预相违背的情况?对缺失数据进行简单的不适宜的估算风险未知任何如下标准：?没有报道缺失或排除的情况，无法判断高风险或 低风险如未说明随机的数量，未提供数据缺失 的原因?研究未

9、描述此情况选择性发表选择性发表导致发表偏倚偏倚低风险标准任何如下标准：?实验的方案书可获取，系统评价感兴趣的所有首 要或次要结局均按方案书预先说明的方式报道?实验方案书不可得，但很明显发表的报告包括所 有的结局，包括预先说明的结局这种性质的有 说服力的文字可能少见高风险判断标准任何如下标准：?不是所有的预先说明的首要结局均被报道? 一个或多个首要结局为采用预先说明的测量方 法、分析方法或数据子集来报道?系统评价感兴趣的一个或多个首要结局报道不 全，以至于不能纳入 meta分析?研究未报道此研究应当包含的主要关键结局风险未知没有足够信息判断高风险或低风险，貌似大局部研究会被分为此类OTHER B

10、IAS不包括在以上五种的其它偏倚偏倚低风险标准研究应未引入其它来源的偏倚高风险判断标准至少有一种重要的偏倚风险，例如：?具有与特殊试验设计相关的潜在偏倚来源?或被指欺诈?或其它问题风险未知可能存在偏倚风险，但存在以卜两种中的一种?没有足够信息评估是否存在其它重要的偏倚风险?没有足够的证据认为发现的问题会引入偏倚Table 8.7.a: Possible approach for summaryassessments of the risk of bias for each importa nt outcome (across doma ins) with in and across studi

11、esRisk of bias解释对单个研究对多个研究整体Low risk of bias.合理的偏倚不太 可能严重改变结 果每一类偏倚均为 低风险绝大多数信息均来 自偏倚低风险的研 究Un clear risk of bias.合理的偏倚会对 结果产生一定的 疑心一类或多类偏倚 风险未知绝大多数信息均来 自偏倚低风险或风 险未知的研究High risk of bias.偏倚严重削弱结 果的可信度一类或多类偏倚 为高风险来自高偏倚风险研 究的信息比例足以 影响结果的解释英文:Table 8.5.a: The Cochrane Collaboration' s tool for asses

12、sing risk of biasDomai nSupport for judgeme ntReview authors ' judgemen tSelecti on bias.Random sequenee gen erati on.Describe the method used to gen erate the allocati on seque nee in sufficie nt detail to allow an assessme nt of whether t should produce comparable groups.Selecti on bias (biase

13、d allocati on to in terve nti ons) due to in adequate gen erati on of a ran domised seque nee.Allocati on con cealme nt.Describe the method used to con ceal the allocati on seque nee in sufficie nt detail to determ ine whether in terve nti on allocati ons could have bee n foresee n in adva nee of, o

14、r duri ng, en rolme nt.Selecti on bias (biased allocati on to in terve nti ons) due to in adequate eon cealme nt of allocati ons prior to assig nment.Performa nee bias.Bli nding of participa nts and pers onnel Assessme nts should be made for each main outcome (or class of outcomes).Describe all meas

15、ures used, if any, to bli nd study participa nts and pers onnel from kno wledge of which in terve nti on a participa nt received. Provide any nformation relating to whether the nten ded bli nding was effective.Performa nee bias due to kno wledge of the allocated in terve nti ons by participa nts and

16、 pers onnel duri ng the study.Detecti on bias.Bli nding of outcome assessme nt Assessme nts should be made for eachDescribe all measures used, if any, to bli nd outcome assessors from kno wledge of which in terve nti on a participa nt received. Provide anyDetecti on bias due to kno wledge of the all

17、ocated in terve nti ons by outcomemain outcome (or class of outcomes).nformation relating to whether the nten ded bli nding was effective.assessors.Attrition bias.n complete outcome data Assessments should be made for each main outcome (or class of outcomes).Describe the complete ness of outcome dat

18、a for each main outcome, including attriti on and exclusi ons from the an alysis. State whether attriti on and exclusi ons were reported, the n umbers n each in terve nti on group (compared with total ran domized participa nts), eas ons for attriti on /exclusi ons where reported, and any re-i nclusi

19、 ons in an alyses performed by the review authors.Attrition bias due to amount, n ature or han dli ng of in complete outcome data.Reporting bias.Selective report ing.State how the possibility of selective outcome report ing was exam ined by the review authors, and what was found.Report ing bias due

20、to selective outcome report ing.Other bias.Other sources of bias.Bias due to problems not covered elsewhere in the table.State any importa nt concerns about bias not addressed in the other doma ins in:he tool.If particular questi on s/e ntries wereRisk of bias ' assessme nt toolpre- specified in

21、 the review ' s protoco esp on ses should be provided for each questi on/en try.RANDOM SEQUENCE GENERATIONSelecti on bias (biased allocati on to in terve nti ons) due to in adequate gen erati on of a an domised seque nee.Criteria for a judgeme nt of ' Low risk ' of bias.The in vestigator

22、s describe a ran dom comp onent in the seque nee gen erati on process such as:?Referri ng to a ran dom n umber table;? Using a computer ran dom n umber gen erator;Table 8.5.d: Criteria for judgi ng risk of bias in the?Coi n toss ing;?Shuffli ng cards or en velopes;?Throw ing dice;? Drawi ng of lots;

23、?Min imizatio n*Mi ni mizatio n may be impleme nted without a ran dom eleme nt, and this is con sidered to be equivale nt to being ran dom.Criteria for the udgement of' High riskof bias.The in vestigators describe a non-ran dom comp onent in the seque nee gerierati on process. Usually, the descr

24、ipti on would in volve some systematic, non-ra ndom approach, for example:?Seque nee gen erated by odd or eve n date of birth;?Seque nee gen erated by some rule based on date (or day) ofadmissi on;?Seque nee gen erated by some rule based on hospital or cli nicrecord n umber.Other non-ran dom approac

25、hes happe n much less freque ntly tha n the systematic approaches men ti oned above and tend to be obvious. They usually in volve judgeme nt or some method of non-random categorization of participants, for example:?Allocati on by judgeme nt of the cli nicia n;?Allocati on by prefere nee of the parti

26、cipa nt;?Allocati on based on the results of a laboratory test or a seriesof tests;?Allocati on by availability of the in terve nti on.Criteria for theudgeme nt of' Un clearisk ' of bias.n sufficie nt in formati on about the seque nee gen erati on process to permit judgement of' Low risk

27、 ' or ' High risk '.ALLOCATION CONCEALMENTSelecti on bias (biased allocati on to in terve nti ons) due to in adequate con cealme nt of allocati ons prior to assig nment.Criteria for a judgeme nt of ' Low risk ' of bias.Participa nts and in vestigators en rolli ng participa nts co

28、uld not foresee assig nment because one of the follow ing, or an equivale nt method, was used to eon ceal allocati on:?Cen tral allocati on (in clud ing teleph one, web-based andpharmacy-c on trolled ran domizati on);?Seque ntially n umbered drug containers of ide nticalappeara nee;?Seque ntially n

29、umbered, opaque, sealed en velopes.Criteria for theudgement of' High riskof bias.Participants or investigators enrolling participants could possibly foresee assig nments and thus in troduce selecti on bias, such as allocati on based on:? Using an ope n ran dom allocati on schedule (e.g. a list o

30、f ran dom n umbers);? Assig nment en velopes were used without appropriate safeguards (e.g. if en velopes were un sealed or nono paque or not seque ntially n umbered);?Alter nati on or rotati on;?Date of birth;?Case record n umber;? Any other explicitly uncon cealed procedure.Criteria for theudgeme

31、nt of' Un clearisk ' of bias.nsufficient information to permit judgement of' Low risk ' crThis is usually the case if the method of concealment is not described or not described in sufficie nt detail to allow a defi nite judgeme nt-forexample if the use of assignment envelopes is des

32、cribed, but itemai ns un clear whether en velopes were seque ntially n umbered, opaque and sealed.BLINDING OF PARTICIPANTS AND PERSONNELPerforma nee bias due to kno wledge of the allocated in terve nti ons by participa nts and pers onnel duri ng the study.Criteria for a judgeme nt of ' Low risk

33、' of bias.Any one of the following:?No bli nding or in complete bli nding, but the review authorsjudge that the outcome is not likely to be in flue need by lack of bli nding;?Bli nding of participa nts and key study pers onnel en sured, andun likely that the bli nding could have bee n broke n.Cr

34、iteria for theudgement of' High riskof bias.Any one of the following:?No bli nding or in complete bli nding, and the outcome is likelyto be in flue need by lack of bli nding;?Bli nding of key study participa nts and pers onnel attempted,but likely that the bli nding could have bee n broke n, and

35、 theHigh riskoutcome is likely to be in flue need by lack of bli nding.Criteria for the udgeme nt of risk ' of bias.Un clearAny one of the following:In sufficie nt in formati on to permit judgeme nt of 'High risk '；w riskorThe study did not address this outcome.BLINDING OF OUTCOME ASSESS

36、MENTDetection bias due to kno wledge of the allocated in terve nti ons by outcome assessors.Criteria for a judgeme nt of Low risk ' of bias.Any one of the following:No bli nding of outcome assessme nt, but the review authors judge that the outcome measureme nt is not likely to be in flue need by

37、 lack of bli nding;Bli nding of outcome assessme nt en sured, and un likely that the bli nding could have bee n broke n.Criteria for theAny one of the following:udgement of High riskof bias.?No bli nding of outcome assessme nt, and the outcomemeasureme nt is likely to be in flue need by lack of bli

38、nding;Bli nding of outcome assessme nt, but likely that the bli nding could have bee n broke n, and the outcome measureme nt islikely to be in flue need by lack of bli nding.Criteria for theAny one of the following:udgeme nt of Un clear risk' of bias.In sufficie nt in formati on to permit judgem

39、e nt of 'High risk '；w riskorThe study did not address this outcome.NCOMPLETE OUTCOME DATAAttrition bias due to amou nt, n ature or han dli ng of in complete outcome data.Criteria for a judgeme ntAny one of the following:of ' Low risk ' of bias.?No miss ing outcome data;?Reasons for

40、missing outcome data unlikely to be related to true outcome (for survival data, censoring unlikely to be in troduc ing bias);?Miss ing outcome data bala need in n umbers across in terve nti on groups, with similar reas ons for miss ing data across groups;For dichotomous outcome data, the proporti on

41、 of miss ing outcomes compared with observed eve nt risk not en ough to have a cli ni cally releva nt impact on the in terve nti on effect estimate;For con ti nuous outcome data, plausible effect size (differe nee in means or sta ndardized differe nee in means) among miss ing outcomes not eno ugh to

42、 have a cli ni cally releva nt impact on observed effect size;Miss ing data have bee n imputed using appropriate methods.Criteria for theudgement of' High riskof bias.Any one of the following:?Reas on for miss ing outcome data likely to be related to trueoutcome, with either imbala nee in n umbe

43、rs or reas ons for miss ing data across in terve nti on groups;?For dichotomous outcome data, the proporti on of miss ingoutcomes compared with observed eve nt risk eno ugh to in duce cli ni cally releva nt bias in in terve nti on effect estimate;? For contin uous outcome data, plausible effect size

44、 (differe nee in means or sta ndardized differe nee in means) among missi ng outcomes en ough to in duce cli ni cally releva nt bias in observed effect size;?' A-sreated ' analysis done with substantial departure of thein terve nti on received from that assig ned at ran domizati on;?Pote nti

45、ally in appropriate applicati on of simple imputati on.Criteria for theudgeme nt of' Un clearisk ' of bias.Any one of the following:?Insufficient reporting of attrition/exclusions to permitjudgement of' Low risk ' or ' High risk ' (e.g. ndumtcieedanot stated, no reas ons for

46、miss ing data provided);? The study did not address this outcome.SELECTIVE REPORTINGReporting bias due to selective outcome reporting.Criteria for a judgeme nt of ' Low risk ' of bias.Any of the following:? The study p rotocol is available and all of the study pre-specified (primary and sec

47、on dary) outcomes that are of in terest in the review have bee n reported in the pre-specified way;? The study protocol is not available but it is clear that the published reports in elude all expected outcomes, in clud ingthose that were pre-specified (convincing text of this n ature may be un com

48、mon).Criteria for theudgement of' High riskof bias.Any one of the following:? Not all of the study-spepifed primary outcomes havebee n reported;? One or more primary outcomes is reported using measureme nts, an alysis methods or subsets of the data (e.g. subscales) that were not pre-specified;?

49、One or more reported primary outcomes were not pre-specified (unless clear justification for their reporting is provided, such as an un expected adverse effect);? One or more outcomes of in terest in the review are reported in completely so that they cannot be en tered in a meta-a nalysis;? The study report fails to in clude results for a key outcome that would be expected to have bee n reported for such a study.Criteria for theudgeme nt of' Un clearisk ' of bias.nsufficient information to permit judgeme