Reverse Shoulder Arthroplasty Versus Non-operative Treatment for Displaced Three- and Four-Part Proximal Humeral Fractures in Older Adults: A Systematic Review and Meta-Analysis

Reverse shoulder arthroplasty (RSA) is increasingly used for displaced three-and four-part proximal humeral fractures in older adults, but comparative evidence has until recently been predominantly observational.Three randomised trials now permit a synthesis restricted to randomised evidence.MEDLINE, Embase, and Cochrane CENTRAL were searched from inception to 15 September 2026 for studies comparing RSA with non-operative treatment in adults aged 60 years and over with an acute displaced Neer three-or four-part fracture.The primary outcome was the unadjusted total Constant-Murley score at 12 months, pooled as a mean difference using a random-effects model with restricted maximum likelihood estimation of τ² and a Hartung-Knapp confidence interval, with the modified (ad hoc) Hartung-Knapp interval reported alongside.Risk of bias was assessed with RoB 2 and certainty with the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach.Three randomised trials (190 analysed participants; 90 RSA, 100 non-operative) and seven observational cohorts, six of which contributed to a narrative synthesis, were included.All three trials favoured RSA at 12 months.The pooled mean difference was 8.49 points (95% CI: 0.53-16.46;p = 0.044; I² = 0%) with the protocol interval and 8.49 points (95% CI: -0.55 to 17.53; p = 0.056) with the modified interval, so statistical significance depended on the variance estimator.The estimate lay below the 9.3-point minimal clinically important difference, which was used as an approximate guide.Sensitivity analyses using the alternative Constant-Murley values published within one trial gave 8.76 and 9.33 points, with confidence intervals including no difference.Complications occurred in 5 of 90 arthroplasty participants and 4 of 100 non-operative participants.Reoperation or revision occurred only after arthroplasty (2 of 90), whereas 5 of 40 participants allocated to non-operative treatment in one trial crossed over to surgery.Certainty was low for the primary outcome and very low for harms.Randomised evidence favours RSA in one-year shoulder function, but the size and clinical importance of the benefit remain uncertain.