Hengg 2019.

Study characteristics
Methods	Method of randomisation: stratified block randomisation and opaque sealed envelopes Assessor blinding: no mention Loss to follow‐up at 1 year: 10/67 (8 withdrawals and 2 lost to follow‐up) (see Notes)
Participants	8 European centres: Innsbruck, Austria; Leuven, Belgium; Aachen, Freiburg, Homburg, Ludwigshafen and Tubingen, Germany; Lucerne and Zurich, Switzerland Period of study recruitment: January 2014 to April 2016 67 participants aged 65 years and older, diagnosed radiographically with an acute (≤ 10 days), closed, displaced or unstable 3‐ or 4‐part proximal humeral fracture sustained after low‐energy trauma, and scheduled for primary fracture treatment with a PHILOS plate. (See Notes.) Informed consent. Exclusion criteria: bilateral or previous proximal humeral fracture, cuff‐arthropathy on either side, head‐splitting or impression fracture of the humeral head, or associated nerve or vessel injuries. Known clotting disorders, severe cardiac or pulmonary insufficiencies, severe systemic diseases (American Society of Anesthesiologists (ASA) class IV to VI), or not medically managed severe systemic diseases classified as ASA class III. Known hypersensitivity or allergy to any of the components of the Traumacem V+ cement Kit. Prisoners or recent history of substance abuse. Participation in any other medical device or medicinal product study within the previous month. Report stated that "patients were excluded before randomization if they received implants other than PHILOS or PHILOS screw augmentation"; however, in 2 cases this appears to be post‐randomisation. 55 female, 12 male; mean age 77 years, range 65 to 92 years
Interventions	Interventions and randomisation at surgery (timing of surgery was meant to be within 10 days) 1. Screw tip augmentation with high viscous polymethylmethacrylate (PMMA) cement. PHILOS plate applied according to manufacturer’s manual and screw augmentation using Traumacem V+ cement kit. Leakage tests were performed with an injection of cement (≤ 0.5 mL) under image intensifier control. Each participant in the augmented group must have 2 to 4 screws augmented. 2. PHILOS plate applied according to manufacturer’s manual. All the postoperative treatments were done according to the standard of care at the investigating sites. Shoulder sling use. Assigned: 33/34 (see Notes regarding protocol violations) Completed (at 1 year): 30/27 but analysed according to intention‐to‐treat analysis at 31/27 (or according to data availability for individual outcomes). Per‐protocol analysis: 23/27; not reported in the review
Outcomes	Length of follow‐up: 1 year (also 6 weeks, and 3 and 6 months) QuickDASH score Shoulder Pain and Disability Index (SPADI) Constant score (absolute and relative to contralateral shoulder) EQ‐5D index and health state Mortality ("sudden death" = 0) Mechanical failures (broadly included loss of reduction (varus malposition and a relative displacement of the greater or lesser tuberosity), humeral head impaction, screw/plate loosening, and secondary screw perforation) Complications: overall number of participants with any adverse event (local and general); these included implant‐related problems, cement leakage, nerve injury, systemic events such as stroke, wound‐related problems such as haematoma; impingement, etc.) Reoperation (data not split by treatment group) Radiographic outcomes, including loss of reduction of humeral head, cement leakage into joint, humeral head necrosis
Funding and conflict of interest statements	Funding: The study was supported by the AO Foundation "via the AOTK Trauma and the AOTrauma networks". The AO Clinical Investigation and Documentation (AOCID) was thanked for "conducting the study, including site management, medical and scientific support, statistical analysis, and medical writing services". Open access for the report publication was from the University of Innsbruck and Medical University of Innsbruck. The conflict of interest statement listed one author as a consultant with DePuy‐Synthes who "did not receive personal benefits for the current study"; one author as the chairman of the AO TK‐System during the time of study; and one author as "a member of the AO UEEG".
Notes	The trial registration document states there were 69 rather than 67 participants; originally 128 was the planned enrolment. There were 14 protocol violations: 11 were deemed ineligible post‐randomisation (6 had 2‐part fractures, 2 had fractures > 10 days old, 1 had a nerve/vessel injury, 2 did not receive PHILOS plates); and 3 were protocol violations due to having more than four screws augmented, not having a leakage test performed before augmentation, and/or receiving screw augmentation despite joint perforation. An additional 3 participants from the augmented group crossed over to the control group due to positive leakage tests. The report presented an intention‐to‐treat analysis and a per‐protocol analysis for most outcomes. Adverse events were reported according to actual intervention received; thus the 3 cross‐overs were reported according to treatment rather than allocation.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Randomization was stratified for each participating center ... Three block sizes were used, chosen at random."
Allocation concealment (selection bias)	Low risk	"Randomization was stratified for each participating center and took place during surgery via opaque sealed envelopes after the fracture reduction was achieved and cannulated locking screws were inserted. Three block sizes were used, chosen at random. To maintain allocation concealment, the pattern of the blocks was kept confidential."
Blinding (performance bias and detection bias) Functional outcomes, pain, clinical outcomes, complications	High risk	No mention of blinding. Assessment of mechanical failure was done "by two experienced independent reviewers". It is unclear whether the participants knew of their group but the procedures were different in terms of testing for suitability when allocated to the augmented group.
Blinding (performance bias and detection bias) Death, reoperation	Unclear risk	Outcome less susceptible and so considered at unclear risk of bias despite clear absence of blinding.
Incomplete outcome data (attrition bias) Functional outcomes, pain, clinical outcomes, complications	Unclear risk	Participant flow provided and denominators provided for patient‐reported outcomes (intention‐to‐treat analysis). Some imbalances in losses: e.g. losses 4 (12% of 33) versus 8 (23.5% of 34) for QuickDASH at 1 year. The trial registration document states there were 69 enrolled rather than 67 as reported in the trial report.
Incomplete outcome data (attrition bias) Death, reoperation	Unclear risk	Although participant flow is provided, data for reoperation were not separated by group. The trial registration document states there were 69 enrolled rather than 67 as reported in the trial report.
Selective reporting (reporting bias)	Unclear risk	Trial registration available. Most important outcomes were reported. However, reoperations were not specified according to groups. Trial was stopped prematurely.
Balance in baseline characteristics?	High risk	Although there was balance in many characteristics, there were more 4‐part fractures in augmentation group (15 versus 9), and fewer 2‐part fractures (5 versus 1). This is likely to have been a source of bias.
Free from performance bias?	Unclear risk	Insufficient information to judge this