BMC Med Res Methodol. 2022 Aug 30;22(1):234. doi: 10.1186/s12874-022-01715-5.
BACKGROUND: Systematic reviews that synthesize safety outcomes pose challenges (e.g. rare events), which raise questions for grading the strength of the body of evidence. This is maybe one reason why in many potentially inappropriate medication (PIM) lists the recommendations are not based on formalized systems for assessing the quality of the body of evidence such as GRADE. In this contribution, we describe specifications and suggest adaptions of the GRADE system for grading the quality of evidence on safety outcomes, which were developed in the context of preparing a PIM-list, namely PRISCUS.
METHODS: We systematically assessed each of the five GRADE domains for rating-down (study limitations, imprecision, inconsistency, indirectness, publication bias) and the criteria for rating-up, considering if special considerations or revisions of the original approach were indicated. The result was gathered in a written document and discussed in a group-meeting of five members with various background until consensus. Subsequently, we performed a proof-of-concept application using a convenience sample of systematic reviews and applied the approach to systematic reviews on 19 different clinical questions.
RESULTS: We describe specifications and suggest adaptions for the criteria “study limitations”, imprecision, “publication bias” and “rating-up for large effect”. In addition, we suggest a new criterion to account for data from subgroup-analyses. The proof-of-concept application did not reveal a need for further revision and thus we used the approach for the systematic reviews that were prepared for the PRISCUS-list. We assessed 51 outcomes. Each of the proposed adaptions was applied. There were neither an excessive number of low and very low ratings, nor an excessive number of high ratings, but the different methodological quality of the safety outcomes appeared to be well reflected.
CONCLUSION: The suggestions appear to have the potential to overcome some of the challenges when grading the methodological quality of harms and thus may be helpful for producers of evidence syntheses considering safety.