Sohn 2017.

Study characteristics
Methods	Method of randomisation: computer‐generated block randomisation with sealed envelopes Assessor blinding: no mention in the paper Loss to follow‐up: 17 lost to follow‐up
Participants	Either National Medicalk Center or Ewha Womans University Mokdong Hospital, Yangcheon‐gu, Seoul, Republic of Korea Period of study recruitment: August 2010 to May 2014 107 skeletally‐mature participants with unilateral two‐part surgical neck, three or four‐part proximal humeral fractures, previously uninjured humerus, time to surgery within 3 weeks, and 3 dimensional computed tomography (3D CT) evaluation. Written consent. Exclusion criteria: isolated greater or lesser tuberosity fractures, injuries with more than 3 weeks between the injury and surgery, pathologic fractures, open fractures, combined dislocation or scapular fractures, neurovascular compromise from the initial trauma. Also stated: "less than 1 year follow up". Of 90: gender not stated; mean age 61.8 years, range not stated
Interventions	Interventions started on average 3.8 days after injury. A 3.5 mm proximal humerus anatomical locking plate (PHILOS; Synthes, Paoli, PA) was used in both groups 1. Minimally invasive plate osteosynthesis (MIPO), which reduction is not done under direct view. A deltoid splitting approach was used proximally, starting from the anterolateral corner of the acromion along the anterior raphe of the deltoid muscle. 2. Open reduction and plate fixation was performed through the conventional deltopectoral approach with careful dissection so that the periosteal blood supply was preserved as much as possible. Assigned: 55/52 Completed: 45/45 (12 months)
Outcomes	Follow‐up: mean 14.65 months; also monthly until bone union was achieved, and at 6 and 12 months Constant score University of California, Los Angeles (UCLA) score Patient satisfaction (VAS) Complications, including from radiological assessment: intra‐articular screw penetration; screw loosening; varus collapse; greater tuberosity migration; stiff shoulder; AVN, axillary nerve injury (0), infection (0), non‐union (0) Length of surgery Time to fracture union
Funding and conflict of interest statements	There was no statement on funding. The authors declared that they had no conflicts of interest.
Notes	The sample size was calculated based on the interim results for Constant score after 40 participants had been recruited. Request sent by HH on 23 February 2020 to Sang‐Jin Shin asking where trial was conducted, and for brief details of rehabilitation, how many women and men in each group and on secondary surgery for complications split by treatment group: no response.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"The randomization was performed by permuted block randomization. The block length was 4, and randomization was performed by means of consecutive numbering. The randomization sequence was created using a web‐based service with a 1:1 balanced allocation." The statement is confusing but we considered that sequence generation was very likely to be random.
Allocation concealment (selection bias)	Low risk	"After the patients were transferred to the operating room, an independent assistant oversaw the randomization using the web site. After confirming the allocation of the patient, the independent assistant notified the surgeon about the allocation."
Blinding (performance bias and detection bias) Functional outcomes, pain, clinical outcomes, complications	High risk	No mention of blinding for functional outcomes. "Radiographic outcomes were estimated independently and agreed upon by 2 orthopaedic surgeons who were not involved in this study." Not relevant to the outcomes collected here.
Incomplete outcome data (attrition bias) Functional outcomes, pain, clinical outcomes, complications	Unclear risk	Non‐attendance of 12‐month follow‐up 17/107 = 16%; 10/55 = 18% in MIPO and 13.5% in open group. No information on why these were lost to follow‐up although reasonably balanced losses between groups
Selective reporting (reporting bias)	Unclear risk	No trial registration or published protocol. Sample size calculated after interim analysis. Focused on radiographic outcomes; incomplete account of assessment of functional outcomes
Balance in baseline characteristics?	Unclear risk	Only age and fracture type data provided and only for the 90 (84%) of 107 randomised. The authors stated without data that there “were no statistically significant differences in other demographic data”.
Free from performance bias?	Unclear risk	No information on care provider expertise, on co‐interventions or rehabilitation