Cai 2012.

Study characteristics
Methods	Method of randomisation: no details Assessor blinding: not blinded (independent observer) Loss to follow‐up at 24 months: 5/32 (4 lost to follow‐up; 1 died)
Participants	Shanghai Tenth People’s Hospital of Tongji University, Shanghai, China Period of study recruitment: April 2005 to March 2010 32 participants with acute displaced 4‐part proximal humeral fracture of the surgical neck (Neer classification). At least one tubercle needed to be displaced more than 10 mm in relation to the head fragment but the other did not need to meet this criterion (thus 3‐part fractures were also acceptable); see Notes. Age 67 or older with low‐energy trauma. Independent living conditions (not institutionalised), and no severe cognitive dysfunction (3 or more correct answers on a 10‐item Short Portable Mental Status Questionnaire (SPMSQ)) Exclusion criteria: completely displaced shaft in relation to the head fragment, such as a fracture without bony contact; valgus impacted fracture, previous shoulder problems 27 female, 5 male; mean age 72 years, range 67 to 86
Interventions	Randomisation was performed after clearance by an anaesthetist prior to surgery; timing not stated. All participants received a single dose of antibiotic preoperatively. 1. Hemiarthroplasty using the DuPuy prosthesis with suturing of tuberosities. Cemented stem. Bone graft from removed humeral head used to restore the humeral offset 2. Open reduction and internal fixation with PHILOS plate. Suturing of tuberosities Postoperative arm sling for 4 weeks (optional thereafter). All participants referred to physiotherapy. Pendulum exercises and passive elevation/abduction up to 90° were started on postoperative day 1. After 4 weeks, the participants were allowed free active range of motion. Assigned: 19/13 Completed (at 2 years): 15/12
Outcomes	Length of follow‐up: 2 years; also assessed at 4 and 12 months DASH Constant score Pain (VAS) Complications (relating to reoperations): non‐union, fixation failure, dislocation, infection, prosthesis loosening Reoperations (for complications) Length of surgery
Funding and conflict of interest statements	The study was supported by the National Science Foundation for Distinguished Young Scholars of China (30901529) and Research Fund for the Doctoral Program of Higher Education (20090072120021) and the Bureau of Public Health, China. There was no statement on conflicts of interest.
Notes	One participant initially had a 3‐part greater tuberosity fracture but at surgery, the lesser tuberosity was also found to be displaced > 1 cm. Hence all had 4‐part fractures. Three of 32 participants had fracture dislocations. Sent email to Dr Li requesting details of the randomisation method and clarification on deaths on 24 May 2015: no response.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"[T]he patients were randomized". No other details
Allocation concealment (selection bias)	Unclear risk	"[T]he patients were randomized". No other details
Blinding (performance bias and detection bias) Functional outcomes, pain, clinical outcomes, complications	High risk	Not blinded even though there was some independent assessment at final follow‐up: "Final 24‐month follow‐up was performed by an independent orthopedic surgeon (K.T.) not involved in treatment."
Blinding (performance bias and detection bias) Death, reoperation	Unclear risk	Lack of blinding less likely to affect assessment of these outcomes. Standardisation of assessment
Incomplete outcome data (attrition bias) Functional outcomes, pain, clinical outcomes, complications	Unclear risk	Active and systematic surveillance and clear participant flow diagram. However, more participants lost to follow‐up in the hemiarthroplasty group (4 (21%) versus 1 (8%)). There are also some incorrect percentages that give rise to concern.
Incomplete outcome data (attrition bias) Death, reoperation	Low risk	Active and systematic surveillance and clear participant flow diagram. It is likely that participants with complications would have returned.
Selective reporting (reporting bias)	Low risk	No protocol. However, systematic data collection and reporting of all outcomes.
Balance in baseline characteristics?	Unclear risk	Where reported, the baseline characteristics were balanced in the two groups. However, the baseline distribution of the fracture types, which included three 4‐part fracture dislocations, was not reported.
Free from performance bias?	Low risk	"All patients underwent surgery performed by 1 of 2 orthopedic surgeons (M.C., S.L.), both experienced in shoulder surgery." Same rehabilitation.