1、使用使用GRADEGRADEpropro评价证据质量评价证据质量GRADEGRADE系统简介系统简介GRADEGRADE（Grading of Recommendations Assessment, Grading of Recommendations Assessment, Development and EvaluationDevelopment and Evaluation）是由）是由20002000年建立的年建立的GRADEGRADE工作组提出的一套评级系统。工作组提出的一套评级系统。GRADEGRADE系统使用易于理解的方式评价证据质量和推荐等系统使用易于理解的方式评价证据质量和推荐
2、等级，目前已被世界卫生组织（级，目前已被世界卫生组织（WHOWHO）、）、CochraneCochrane协作网协作网等一批著名机构所采用。等一批著名机构所采用。GRADEGRADE系统较之其他系统的优势系统较之其他系统的优势由一个具有广泛代表性的国际指南制定小组制定由一个具有广泛代表性的国际指南制定小组制定明确界定了证据质量和推荐强度明确界定了证据质量和推荐强度清楚评价了不同治疗方案的重要结局清楚评价了不同治疗方案的重要结局对不同级别证据的升级与降级有明确、综合的标准对不同级别证据的升级与降级有明确、综合的标准从证据到推荐全过程透明从证据到推荐全过程透明明确承认价值观和意愿明确承认价值观和意
3、愿就推荐意见的强弱，分别从临床医生、患者、政策制定者就推荐意见的强弱，分别从临床医生、患者、政策制定者角度做了明确实用的诠释角度做了明确实用的诠释适用于制作系统评价、卫生技术评估及指南适用于制作系统评价、卫生技术评估及指南证据质量证据质量（GRADEproGRADEpro）为达到透明和简化的目标，为达到透明和简化的目标， GRADEGRADE系统将证据质量分为系统将证据质量分为高、高、 中、中、 低、低、 极低极低 4 4 级。一些使用级。一些使用GRADEGRADE系统的组织系统的组织甚至把低和极低归为一级。甚至把低和极低归为一级。推荐等级推荐等级使用使用GRADEGRADE系统时，指南小组
4、系统时，指南小组 用“强推荐”表示他们确信相关的干预措施利大于弊。 用“弱推荐”表示干预措施有可能利大于弊，但他们把握不大。GRADEGRADE中的证据质量和推荐等级中的证据质量和推荐等级 符号符号字母字母/ /数字数字证据质量证据质量 高质量高质量 A 中等质量中等质量 B 低质量低质量 C 极低质量极低质量 D推荐等级推荐等级 支持使用某疗法的强推荐支持使用某疗法的强推荐1 支持使用某疗法的弱推荐支持使用某疗法的弱推荐？2 反对使用某疗法的强推荐反对使用某疗法的强推荐1 反对使用某疗法的弱推荐反对使用某疗法的弱推荐？2关键步骤一关键步骤一 导入导入RevManRevMan文件文件关键步骤二
5、关键步骤二 判断结局的重要程度判断结局的重要程度GRADE GRADE 系统建议采用系统建议采用 9 9 级分级判断结局的重要程度级分级判断结局的重要程度: : 13：不重要结局（NOT IMPORTANT） 46：重要结局（IMPORTANT） 79：关键结局（CRITICAL）关键步骤三关键步骤三 选择研究设计类型选择研究设计类型按研究设计方法划分按研究设计方法划分 随机试验质量等级最高，为高； 观察性研究 （包括队列研究、病例 - 对照研究、病例 - 历史对照研究）质量等级低，为低； 病例报告、病例分析等其他研究质量等级非常低，为极低。关键步骤四关键步骤四 评价证据质量评价证据质量基于研
6、究设计可因基于研究设计可因5 5 条降级因素和条降级因素和 4 4 条升级因素上升或下条升级因素上升或下降级别。降级别。决定证据质量的因素决定证据质量的因素可能降低证据质量的因素可能降低证据质量的因素 研究的局限性 结果不一致 间接证据 精确度不够 发表偏倚可能增加证据质量的因素可能增加证据质量的因素 效应值很大 可能的混杂因素会降低疗效 剂量-效应关系研究设计与证据质量研究设计与证据质量高：为不降级的随机试验和升高：为不降级的随机试验和升 2 2 级的观察性研究；级的观察性研究； 中：为降级的随机试验和升中：为降级的随机试验和升 1 1 级的观察性研究；级的观察性研究； 低：降低：降 2 2
7、 级的随机试验和没有升降级的观察性研究；级的随机试验和没有升降级的观察性研究； 非常低：降非常低：降3 3级的随机试验、级的随机试验、 降降1 1级的观察性研究、级的观察性研究、 病例病例分析分析/ /病例报告。病例报告。帮助教程要点帮助教程要点研究的局限性（研究的局限性（Risk of biasRisk of bias）这些局限性包括（这些局限性包括（RCTsRCTs） 隐藏分组缺失 盲法缺失（特别是结局指标为主观性指标且对其的评估极易受偏倚影响时） 失访过多 未进行意向性分析、观 察到疗效就过早终止试验 或未报道结果（通常是未观察到疗效的一些研究）To rate study limitat
8、ions: To rate study limitations: If you think any limitations were negligible choose no If you think there were serious limitations choose serious this will downgrade the quality of evidence for this outcome by 1 level If you think there were very serious limitations choose very serious this will do
9、wngrade the quality of evidence for this outcome by 2 levels 结果不一致（结果不一致（ Inconsistency Inconsistency）Inconsistency may arise from differences in:Inconsistency may arise from differences in: populations (e.g. drugs may have larger relative effects in sicker populations) interventions (e.g. larger ef
10、fects with higher drug doses) outcomes (e.g. diminishing treatment effect with time)差异可能源于人群（如药物对重症人群的疗效可能相对显差异可能源于人群（如药物对重症人群的疗效可能相对显著）、干预措施（如较高药物剂量会使疗效更显著）或结著）、干预措施（如较高药物剂量会使疗效更显著）或结局指标（如随时间推移疗效降低）。局指标（如随时间推移疗效降低）。I I2 2 One set of criteria would say that an I One set of criteria would say that an
11、 I2 2 of of less than 40% is low 3060% may be moderate 5090% may be substantial 75100% is considerableTo rate inconsistency: To rate inconsistency: If you think any inconsistency was negligible choose no If you think there was serious inconsistency choose serious this will downgrade the quality of e
12、vidence for this outcome by 1 level If you think there was very serious inconsistency choose very serious this will downgrade the quality of evidence for this outcome by 2 levels 间接证据（间接证据（ Indirectness) Indirectness)There are two types of indirectness:There are two types of indirectness: 1. Indirec
13、t comparisonIndirect comparison 2. Indirect Indirect populationpopulationinterventioninterventioncomparatorcomparatoroutcomeoutcomeTo rate indirectness: To rate indirectness: If you think the evidence is direct choose no If you have serious doubts about directness choose serious this will downgrade
14、the evidence for this outcome by 1 level If you have very serious doubts about directness choose very serious this will downgrade the evidence for this outcome by 2 levels 精确度不够（精确度不够（Imprecision Imprecision ）当研究纳入的患者和观察事件相对较少而致可信区间较宽当研究纳入的患者和观察事件相对较少而致可信区间较宽时，指南小组将降低该研究的证据质量。时，指南小组将降低该研究的证据质量。A thr
15、eshold rule-of-thumb valueA threshold rule-of-thumb value Total number of events is less than 300（dichotomous） Total population size is less than 400（continuous）To rate imprecision: To rate imprecision: If you think the results were precise choose no If there was serious imprecision choose serious t
16、his will downgrade the quality of evidence for this outcome by 1 level If there was very serious imprecision choose very serious this will downgrade the quality of evidence for this outcome by 2 levels 发表偏倚（发表偏倚（ Publication bias Publication bias ）Publication bias arises when investigators fail to r
17、eport studies they have undertaken (typically those that show no effect). Methods to detect the possibility of publication bias in systematic reviews exist, although authors of the reviews and guideline panels must often guess about the likelihood of publication bias. A prototypical situation that s
18、hould elicit suspicion of publication bias occurs when published evidence is limited to a small number of trials, all of which are showing benefits of the studied intervention.若研究者未能发表研究（通常是阴性结果的研究）时，证若研究者未能发表研究（通常是阴性结果的研究）时，证据质量亦会减弱。典型情况是当公开的证据仅局限于少数据质量亦会减弱。典型情况是当公开的证据仅局限于少数试验而这些试验全部由企业赞助，此时不得不质疑存在
19、发试验而这些试验全部由企业赞助，此时不得不质疑存在发表偏倚。表偏倚。Funnel PlotFunnel PlotTo rate probability of the publication bias: To rate probability of the publication bias: If you think there is no evidence of publication bias choose unlikely If there is high probability of publication bias choose likely this will downgrade t
20、he quality of evidence for this outcome by 1 level If there is very high probability of publication bias choose very likely this will downgrade the quality of evidence for this outcome by 2 levels 效应值很大（效应值很大（Large effectLarge effect）疗效越显著，证据越有力。疗效越显著，证据越有力。To rate magnitude of the effect: To rate m
21、agnitude of the effect: If the effect was If the effect was not large not large (RR between 0.5 and 2.0) (RR between 0.5 and 2.0) choose choose no no If the effect was If the effect was large large (RR either 2.0 or 2.0 or 2 or 2 or 5.0 or 5.0 or 5 or 5 or 0.2this will this will upgradeupgrade the q
22、uality of evidence for this outcome the quality of evidence for this outcome by by 2 levels2 levels 可能的混杂因素会降低疗效可能的混杂因素会降低疗效（Plausible confounding would change the effectPlausible confounding would change the effect）On occasion, all plausible confounding from observational studies or randomised tria
23、ls may be working to reduce the demonstrated effect or increase the effect if no effect was observed.For example, if only sicker patients receive an experimental intervention or exposure, yet they still fare better, it is likely that the actual intervention or exposure effect is larger than the data
24、 suggest.To rate the effect of all plausible To rate the effect of all plausible residual confounding: residual confounding: If there is If there is no evidence no evidence that the influence of all plausible that the influence of all plausible confounding would reduce a demonstrated effect or confo
25、unding would reduce a demonstrated effect or suggest a spurious effect when results show no effect suggest a spurious effect when results show no effect choose choose nono If there is If there is evidenceevidence that the influence of all plausible that the influence of all plausible confounding wou
26、ld reduce a demonstrated effect or confounding would reduce a demonstrated effect or suggest a spurious effect when results show no effect suggest a spurious effect when results show no effect choose choose yesyesthis will this will upgrade upgrade the quality of evidence for this outcome the qualit
27、y of evidence for this outcome by by 1 level1 level 剂量剂量- -效应关系（效应关系（Dose-response gradiDose-response gradi）The presence of a dose-response gradient may increase our confidence in the findings of observational studies and thereby increase the quality of evidence.To rate the presence of dose-response
28、 To rate the presence of dose-response gradient: gradient: If there is If there is no evidenceno evidence of dose-response gradient choose of dose-response gradient choose nono If there is If there is evidenceevidence of dose-response gradient choose of dose-response gradient choose yesyesthis will
29、upgrade the quality of evidence for this outcome this will upgrade the quality of evidence for this outcome by by 1 level1 level 评级流程评级流程关键步骤五关键步骤五 完善数据完善数据关键步骤六关键步骤六 生成表格生成表格Evidence Profile vs. Evidence Profile vs. Summary of Findings TableSummary of Findings TableGRADE Evidence profileGRADE Evide
30、nce profile Particularly useful for guideline developers Presents information about the body of evidence (e.g. number of studies), the judgments about the underlying quality of evidence, key statistical results, and a grade for the quality of evidence for each outcomeSummary of Findings TableSummary
31、 of Findings Table Most relevant to (and designed for use in) Cochrane reviews Present the main findings of a systematic review in a transparent and simple tabular format Provides key information concerning the quality of evidence, the magnitude of effect of the interventions examined, and the sum o
32、f available data on most important outcomes.SoF TableSoF TableThere are 3 key components of a SoF table: There are 3 key components of a SoF table: Information about the review Summary of the statistical results Grade of the quality of evidence. 关键步骤七关键步骤七 导出文件到导出文件到RevManRevMan MetaMeta分析分析（RevManRevMan）证据等级证据等级（GRADEproGRADEpro）推荐等级推荐等级系统综述流程图系统综述流程图
