指南与共识医学教育与科研方法流行病学与生物统计

为证据合成与指南中的健康结局设定决策阈值:开发流程及在瑞典指南中的应用

Setting decision thresholds for health outcomes in evidence synthesis and guidelines: development process and application to magnitude of effects, net effect, and certainty of evidence in a Swedish guideline.
指南 / 共识 · J Clin Epidemiol · 2026-07-20

该指南类文章通过迭代共识流程,将GRADE指导应用于决策阈值的设定,使用注册数据、系统评价和专家调查等多种证据来源,在瑞典早产儿咖啡因指南制定中实践。文章提出三阶段共15个步骤的决策阈值开发与应用方法,涵盖结局重要性评估、效应量阈值精炼及证据确定性评级。结论认为明确阈值有助于判断效应大小、净效应和形成推荐。

要点

  • 决策阈值开发分为三阶段15步,涵盖证据准备、阈值精炼与应用。
  • 阈值设定结合注册数据、系统评价、专家调查及工作组共识。
  • 明确阈值有助于判断效应大小、估计净效应并形成推荐。

核心结论

  • 提供了GRADE指导用于开发、报告和使用明确决策阈值的实践应用,可推广至其他场景。
  • 使用明确阈值促进了GRADE在证据到决策过程中的多个步骤。
  • 指南小组和方法学家认为阈值有助于确定效应大小、估计效应平衡并达成推荐。

研究设计与样本

  • 通过迭代共识流程,应用GRADE指导设定决策阈值。
  • 使用多种证据来源,包括注册数据、结局价值系统评价和内容专家调查。
  • 在瑞典早产儿咖啡因指南制定中应用该方法。

主要发现

  • 决策阈值开发分为三阶段共15个明确步骤。
  • 第一阶段包含5步,涉及识别结局、开发健康结局描述、评估结局重要性、确定结局发生率及开展或识别相对重要性系统评价。
  • 第二阶段包含6步,涉及与方法学家起草初始阈值、在工作组内精炼、确定负效用值、与指南小组和决策者达成共识及获取利益相关者反馈。
  • 第三阶段包含4步,涉及将阈值应用于单个结局效应量、估计净效应、评级证据确定性及探索利益群体或亚组特定阈值的影响。

临床怎么用

  • 在指南制定中可采用三阶段15步流程系统开发决策阈值。
  • 利用注册数据、系统评价和专家调查等多源证据设定阈值。
  • 通过工作组和指南小组共识精炼并应用阈值以支持推荐形成。

局限性

  • 原文指出关于决策阈值程序步骤的证据有限。

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

BACKGROUND AND OBJECTIVES: Key tasks in the development of guideline recommendations include determining the magnitude of desirable and undesirable health effects, the overall net effect, and assessing the certainty of evidence. Developing and using decision thresholds can support these tasks and increase transparency, however evidence of the procedural steps is limited. We describe a systematic, step-by-step approach on how to develop and use decision thresholds for dichotomous and continous outcomes and apply this approach to guide decision-making during the development of the Swedish Guideline of Caffeine in Preterm Infants. METHODS: Through an iterative consensus-based process, we applied the GRADE Guidance to set decision thresholds by using a variety of different sources of evidence, including register data, systematic reviews of values of outcomes, and surveys to content experts, and discussion, feedback and consensus within working groups and the guideline panel. RESULTS: We used 15 distinct steps across three phases to derive decision thresholds. Phase 1 included five steps related to preparing the evidence: Identify relevant outcomes; develop health outcome descriptors; assess the importance of outcomes; determine the incidence of outcomes-both positive outcomes and outcomes regarded as harms/side effects; and conduct or identify systematic reviews of relative importance of outcomes. Phase 2 included six steps relating to refining decision thresholds for the magnitude of effects: Draft initial thresholds with methodologists; refine thresholds within a working group including content experts; determine disutility-value; refine thresholds within a working group; obtain consensus with guideline panel and decision makers; and receiving further interest-holder feedback. Phase 3 include four steps related to applying the decision thresholds to; magnitude of effects of individual outcomes; estimate the net effect; rate the certainty of the evidence; explore the impact of interest-group- or subgroup-specific thresholds. CONCLUSION: We provide a practical application of GRADE guidance for developing, reporting, and using explicit decision thresholds that can be applied in other scenarios. We found that the use of explicit thresholds facilitated a number of steps in applying GRADE in evidence to decision processes. Overall, thresholds were considered useful by the panel and methodologists for determining magnitude of effects, estimating the balance of effects, and for reaching a recommendation.

来源:J Clin Epidemiol,查看原文。版权归原作者所有。

引用

  1. Ringsten M, Reinap M, Petersson IF, et al. Setting decision thresholds for health outcomes in evidence synthesis and guidelines: development process and application to magnitude of effects, net effect, and certainty of evidence in a Swedish guideline. J Clin Epidemiol. 2026. PMID: 42476474. doi:10.1016/j.jclinepi.2026.112422

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。