Bertoft 1984.

Study characteristics
Methods	Use of permutation table, single‐blind, independently administered Assessor blinded Loss to follow‐up at 1 year: 7/20 (2 excluded)
Participants	Central hospital, Vasteras, Sweden Period of study recruitment: not stated 20 participants with non or minimally displaced proximal humeral fractures (7 had fracture of the greater tubercle); sling for 10 days. Exclusion criteria: no information 17 female, 3 male; mean age 64 years, range 50 to 75 years
Interventions	Interventions started 10 to 12 days post injury, after removal of sling. 1. Instructed self‐exercise: participants instructed to train 5 to 10 minutes, 4 to 5 times daily. They had three training sessions (day 1, weeks 3 & 8 post injury) 2. Conventional physiotherapy: 9 sessions (average 20 to 30 minutes), 1 to 2 times each week, over 10 to 12 weeks. No thermoelectrotherapy. Assigned: 10/10 Completed ( > 1 year): 7/6
Outcomes	Length of follow‐up: 1 year; also assessed at 3, 8, 16 & 24 weeks Range of motion: forward flexion (graph), abduction, internal & external rotation Functional movements: placing hand on neck, placing hand on back Pain: when placing hand on neck: combing hair (graph) Isometric muscle strength: vertical & horizontal pushing Change of treatment requested
Funding and conflict of interest statements	There was no statement on funding. There was no statement on conflicts of interest.
Notes	The 2 excluded participants were in the control group: 1 died and 1 underwent an operation.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Mention of "permutation table" and "randomized controlled" trial
Allocation concealment (selection bias)	Low risk	"A third person was responsible for the randomization procedure and kept the key to the permutation table"
Blinding (performance bias and detection bias) Functional outcomes, pain, clinical outcomes, complications	Unclear risk	"A second physiotherapist examined the patients. She did not know to which group the patient belonged, and the patients were instructed not to tell her." However, there is no guarantee of blinding and, for practical reasons, neither participants nor care provider were blinded.
Blinding (performance bias and detection bias) Death, reoperation	Low risk	Lack of blinding unlikely to affect assessment of these outcomes
Incomplete outcome data (attrition bias) Functional outcomes, pain, clinical outcomes, complications	Unclear risk	Participant flow provided but large loss to follow‐up (7/20 = 35%).
Incomplete outcome data (attrition bias) Death, reoperation	Low risk	Participant flow provided
Selective reporting (reporting bias)	Unclear risk	Insufficient information to judge this.
Balance in baseline characteristics?	Unclear risk	Incomplete data to back up claims of lack of baseline differences but a 4‐year difference in mean age between groups (66 versus 62 years).
Free from performance bias?	Low risk	No indication of performance bias.