Certainty may remain the best term for GRADE evidence ratings despite conceptual limitations: a commentary.

Abstract

Background: The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach initially described its ratings as confidence in an effect estimate. Current applications retain this focus whereas specifying the inference of interest, such as whether the true effect lies above or below a decision-relevant threshold. The terminology later changed from "quality of evidence" to "certainty of evidence."

Main argument: GRADE reports four ordered levels-high, moderate, low, and very low-of support for an outcome- and inference-specific conclusion. Ordinary English may give certainty an all-or-none state, but technical attributes permit gradation. We compare certainty, quality, confidence, strength, and trustworthiness through their strengths and limitations. None resolves the tension without adding another concern.

Conclusion: In the absence of a superior alternative, certainty may remain the most practical term because it is established in guidance, teaching, software, and practice. It should be defined as an ordered judgment about support for a specified inference, not as proof.

Plain language summary: GRADE uses four categories to describe how strongly evidence supports a particular inference. Although certainty can imply an all-or-none state in ordinary language, GRADE uses it as an ordered, outcome-specific judgment. Certainty may remain practical if users understand that it does not prove a finding is true.