Helen Hg Handoll, Joanne Elliott, Theis M Thillemann, Patricia Aluko, Stig Brorson

Fractures of the proximal humerus, often termed shoulder fractures, are common injuries, especially in older people. The management of these fractures varies widely, including in the use of surgery. This is an update of a Cochrane Review first published in 2001 and last updated in 2015. To assess the effects (benefits and harms) of treatment and rehabilitation interventions for proximal humeral fractures in adults. We searched the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, Embase, trial registries, and bibliographies of trial reports and systematic reviews to September 2020. We updated this search in November 2021, but have not yet incorporated these results. We included randomised and quasi-randomised controlled trials that compared non-pharmacological interventions for treating acute proximal humeral fractures in adults.  DATA COLLECTION AND ANALYSIS: Pairs of review authors independently selected studies, assessed risk of bias and extracted data. We pooled data where appropriate and used GRADE for assessing the certainty of evidence for each outcome. We prepared a brief economic commentary for one comparison. We included 47 trials (3179 participants, mostly women and mainly aged 60 years or over) that tested one of 26 comparisons. Six comparisons were tested by 2 to 10 trials, the others by small single-centre trials only. Twelve studies evaluated non-surgical treatments, 10 compared surgical with non-surgical treatments, 23 compared two methods of surgery, and two tested timing of mobilisation after surgery. Most trials were at high risk of bias, due mainly to lack of blinding. We summarise the findings for four key comparisons below. Early (usually one week post injury) versus delayed (after three or more weeks) mobilisation for non-surgically-treated fractures Five trials (350 participants) made this comparison; however, the available data are very limited. Due to very low-certainty evidence from single trials, we are uncertain of the findings of better shoulder function at one year in the early mobilisation group, or the findings of little or no between-group difference in function at 3 or 24 months. Likewise, there is very low-certainty evidence of no important between-group difference in quality of life at one year. There was one reported death and five serious shoulder complications (1.9% of 259 participants), spread between the two groups, that would have required substantive treatment. Surgical versus non-surgical treatment Ten trials (717 participants) evaluated surgical intervention for displaced fractures (66% were three- or four-part fractures). There is high-certainty evidence of no clinically important difference between surgical and non-surgical treatment in patient-reported shoulder function at one year (standardised mean difference (SMD) 0.10, 95% confidence interval (CI) -0.07 to 0.27; 7 studies, 552 participants) and two years (SMD 0.06, 95% CI -0.13 to 0.25; 5 studies, 423 participants). There is moderate-certainty evidence of no clinically important between-group difference in patient-reported shoulder function at six months (SMD 0.17, 95% CI -0.04 to 0.38; 3 studies, 347 participants). There is high-certainty evidence of no clinically important between-group difference in quality of life at one year (EQ-5D (0: dead to 1: best quality): mean difference (MD) 0.01, 95% CI -0.02 to 0.04; 6 studies, 502 participants). There is low-certainty evidence of little between-group difference in mortality: one of the 31 deaths was explicitly linked with surgery (risk ratio (RR) 1.35, 95% CI 0.70 to 2.62; 8 studies, 646 participants). There is low-certainty evidence of a higher risk of additional surgery in the surgery group (RR 2.06, 95% CI 1.21 to 3.51; 9 studies, 667 participants). Based on an illustrative risk of 35 subsequent operations per 1000 non-surgically-treated patients, this indicates an extra 38 subsequent operations per 1000 surgically-treated patients (95% CI 8 to 94 more). Although there was low-certainty evidence of a higher overall risk of adverse events after surgery, the 95% CI also includes a slightly increased risk of adverse events after non-surgical treatment (RR 1.46, 95% CI 0.92 to 2.31; 3 studies, 391 participants). Open reduction and internal fixation with a locking plate versus a locking intramedullary nail Four trials (270 participants) evaluated surgical intervention for displaced fractures (63% were two-part fractures). There is low-certainty evidence of no clinically important between-group difference in shoulder function at one year (SMD 0.15, 95% CI -0.12 to 0.41; 4 studies, 227 participants), six months (Disability of the Arm, Shoulder, and Hand questionnaire (0 to 100: worst disability): MD -0.39, 95% CI -4.14 to 3.36; 3 studies, 174 participants), or two years (American Shoulder and Elbow Surgeons score (ASES) (0 to 100: best outcome): MD 3.06, 95% CI -0.05 to 6.17; 2 studies, 101 participants). There is very low-certainty evidence of no between-group difference in quality of life (1 study), and of little difference in adverse events (4 studies, 250 participants) and additional surgery (3 studies, 193 participants). Reverse total shoulder arthroplasty (RTSA) versus hemiarthroplasty There is very low-certainty evidence from two trials (161 participants with either three- or four-part fractures) of no or minimal between-group differences in self-reported shoulder function at one year (1 study) or at two to three years' follow-up (2 studies); or in quality of life at one year or at two or more years' follow-up (1 study). Function at six months was not reported. Of 10 deaths reported by one trial (99 participants), one appeared to be surgery-related. There is very low-certainty evidence of a lower risk of complications after RTSA (2 studies). Ten people (6.2% of 161 participants) had a reoperation; all eight cases in the hemiarthroplasty group received a RTSA (very low-certainty evidence). There is high- or moderate-certainty evidence that, compared with non-surgical treatment, surgery does not result in a better outcome at one and two years after injury for people with displaced proximal humeral fractures. It may increase the need for subsequent surgery. The evidence is absent or insufficient for people aged under 60 years, high-energy trauma, two-part tuberosity fractures or less common fractures, such as fracture dislocations and articular surface fractures. There is insufficient evidence from randomised trials to inform the choices between different non-surgical, surgical or rehabilitation interventions for these fractures.