Grading of Recommendations, Assessment, Development, and Evaluation Notes 6: three strategies to determine the clinically important thresholds for outcomes in evidence-based guideline development

Objectives: In evidence-based clinical practice guidelines (CPGs), the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) approach is commonly used to evaluate the certainty of evidence to develop recommendations. The recently published GRADE guidance papers suggested using a confidence interval approach as the primary criterion to assess the imprecision and inconsistency domains. This involves predetermining three clinically important thresholds, ie, small, moderate, and large effects. Establishing thresholds for outcomes is helpful when judging the strength of the corresponding recommendation. Recognizing that setting these thresholds poses a significant challenge for many CPG developers, we aim to delineate three methods for determining clinically important thresholds. Study Design and Setting: During the development of a diabetes-related CPG, we identified three practical strategies to increase the likelihood of finding clinically significant thresholds in the literature. Results: The three strategies are as follows: First, the minimally important difference or minimal clinically important difference found within the literature can be regarded as a small clinically important threshold. Second, the effect size used to calculate the sample size in a randomized controlled trial can serve as a small threshold for the same outcome, and occasionally may serve as moderate or even large threshold. Third, the different magnitudes of effect between two groups in trials conducted for government drug approval can guide the determination of three thresholds for any outcomes, especially for safety-related outcomes in a CPG. Conclusion: This study presents three approaches that can be utilized to guide the establishment of three clinically important thresholds in the evaluation of the imprecision and inconsistency domains in certainty assessments of evidence and to inform judgement about the strength of the corresponding recommendation.