Fractures of the proximal humerus have been diagnosed and managed since the earliest known surgical texts. For more than four millennia the preferred treatment was forceful traction, closed reduction, and immobilization with linen soaked in combinations of oil, honey, alum, wine, or cerate. The bandages were further supported by splints made of wood or coarse grass. Healing was expected in forty days. Different fracture patterns have been discussed and classified since Ancient Greece. Current classification of proximal humeral fractures mainly relies on the classifications proposed by Charles Neer and the AO/OTA classification. Since the late 1980's it has been known that intra- and inter-observer variation was high within the two systems. I conducted a series of observer studies to qualify the disagreement further and to study to what extent improvement of agreement could be obtained. No clinically significant differences in observer agreement were found at different levels of clinical experience, by reducing the number of categories, or by adding high quality radiographs, CT or 3D CT scans. A consistently low agreement on the Neer classification within and between untrained orthopaedic doctors was found. However, we also found that inter-observer agreement on treatment recommendation was higher than the agreement on the Neer classification. In a randomized trial we found that agreement could improve significantly by training of doctors, especially among specialists. However, classification of proximal humeral fractures remains a challenge for the conduct, reporting, and interpretation of clinical trials. The evidence for the benefits of surgery in complex fractures of the proximal humerus is weak. In three systematic reviews I studied the outcome after locking plate osteosynthesis or reverse arthroplasty in complex fractures patterns. No randomized trials or well-conducted comparative studies were identified. High failure rates suggest that the use of these implants for complex fractures of the humerus should not be used outside clinical protocols. I recommend the conduct of randomized trials, and a design of such study is proposed.