Guyatt Gordon, Wang Ying, Eachempati Prashanti, Iorio Alfonso, Murad M Hassan, Hultcrantz Monica, Chu Derek K, Florez Ivan D, Hemkens Lars G, Agoritsas Thomas, Yao Liang, Vandvik Per Olav, Montori Victor M, Brignardello-Petersen Romina
Department of Health Research Methods, Evidence and Impact, McMaster University, Hamilton, ON, Canada
Department of Medicine, McMaster University, Hamilton, ON, L8S 4L8, Canada.
BMJ. 2025 May 13;389:e083864. doi: 10.1136/bmj-2024-083864.
This fourth article in a seven part series presents the Core GRADE (Grading of Recommendations Assessment, Development and Evaluation) approach to addressing risk of bias, publication bias, and rating up certainty. In Core GRADE, randomised controlled trials begin as high certainty evidence and non-randomised studies of interventions (NRSI) as low certainty. To assess certainty of evidence for risk of bias, Core GRADE users first classify individual studies as low or high risk of bias. Decisions regarding rating down for risk of bias will depend on the weights of high and low risk of bias studies and similarities or differences between the results of high and low risk of bias studies. For publication bias, a body of evidence comprising small studies funded by industry should raise suspicion. Core GRADE users appraising results from well conducted NSRI can consider rating up certainty of evidence when risk ratios from pooled estimates suggest large or very large effects.
本系列七篇文章中的第四篇介绍了核心GRADE（推荐分级评估、制定与评价）方法，用于处理偏倚风险、发表偏倚以及提升证据确定性。在核心GRADE中，随机对照试验一开始被视为高确定性证据，而干预性非随机研究（NRSI）则为低确定性证据。为评估偏倚风险的证据确定性，核心GRADE使用者首先将个体研究分类为低偏倚风险或高偏倚风险。关于因偏倚风险而降低证据等级的决策将取决于高偏倚风险研究和低偏倚风险研究的权重，以及高偏倚风险研究和低偏倚风险研究结果之间的异同。对于发表偏倚，由行业资助的小型研究组成的证据群体应引发怀疑。评估精心开展的NRSI结果时，若汇总估计的风险比表明存在大或非常大的效应，核心GRADE使用者可考虑提升证据的确定性。
