Reporting of harms in systematic reviews focused on reverse shoulder arthroplasty: a cross-sectional study

Background: Reverse shoulder arthroplasty (RSA) is projected to increase in frequency by over 300% in the next decade due to the recent expansion of indications for this procedure. Therefore, a more thorough understanding of efficacy and harms is relevant for clinicians and patients to make unbiased evaluations of the intervention. Systematic reviews(SRs) are commonly used to guide clinical decision-making in orthopaedics, but they are known to weigh efficacy more heavily than harms in their reporting. Therefore, the objective of this cross-sectional analysis was to investigate completeness of harms reporting in SRs relating to reverse shoulder arthroplasty (RSA).

Methods: We performed a comprehensive search using EMBASE, MEDLINE (Pubmed and Ovid), Epistemonikos, and the Cochrane database for Systematic Reviews for relevant literature. Search returns were screened for inclusion and extracted data using a masked, duplicate method. General study characteristics, harms items, and overall methodological quality for each SR were extracted. Corrected covered area (CCA) was quantified for SR pairs. Regarding data analysis, Stata 16.1 was used to conduct a bivariate analysis between variables.

Results: After screening and full-text review, our sample consisted of 89 SRs. Of the included SRs, 26 (26/89, 29.2%) reported ≤ 50% of harms items. Fifteen included a pre-specified protocol that addressed harms (15/89, 16.9%), specific harms or harms language were included in the search strategy of 13 included SRs (13/89, 14.6%), and harms were listed and separately defined in the methods section of only 38 included SRs (38/89, 42.7%). Seventy-four SRs assessed harms in a quantitative manner only (74/89, 83.2%), 5 assessed harms both qualitatively and quantitatively (5/89, 5.66%), 1 assessed harms in a qualitative manner only (1/89, 1.1%) and the remaining 9 did not perform a harms assessment (9/89, 10.1%). Eighty-four SRs were graded as ‘critically low’ quality and 5 SRs were graded as ‘low’ quality by AMSTAR-2. Six SR dyads had CCAs greater than 50% overlap and were compared for shared harms.

Conclusion: Our findings suggest inadequate harms reporting pertaining to RSA in SRs. To illustrate this, nearly 30% of SRs related to RSA in our sample failed to report at least 50% of harms items. We recommend improvement to reporting guidelines regarding harms reporting and that these improved guidelines be used by future studies. Complete harms reporting may facilitate better patient outcomes and allow for more thorough risk-benefit assessments.