Background: Fractures of the proximal humerus are common injuries and account for 4-5 percent of all fractures, second only to hip and wrist fractures. The incidence is positively correlated with age and osteoporosis, and is likely to increase. Displaced four-part fractures are among the most severe injuries, accounting for 2-10 percent of proximal humeral fractures. The optimal intervention is disputed. Two previous randomised trials were very small and involved a noticeable risk of bias, and systematic reviews consequently conclude that there is inadequate basis for evidence-based treatment decisions. We aim to compare the effect of osteosynthesis with angle-stable plate with non-surgical management, and the effect of primary hemiarthroplasty with both osteosynthesis and non-surgical management. Methods/Design: We will conduct a randomised, multi-centre, clinical trial including patients from ten national shoulder units within a two-year period. We plan to include 162 patients. A central randomisation unit will allocate patients. All patients will receive a standardised three-month rehabilitation program of supervised physiotherapy regardless of treatment allocation. Patients will be followed at least one year. The primary outcomes will be the overall score on the Constant Disability Scale, and its pain subscale, measured at 12 months. A blinded physiotherapist will carry out the assessments. Other secondary outcomes are Oxford Shoulder Score, and general health status (Short Form-36).