Purpose of the study: Optimal management of proximal fractures of the humerus remains a subject of debate. We conducted a prospective epidemiological study to identify injuries encountered, determine the reproducibility of reference classifications and their pertinence for therapeutic decision making.
Material and methods: All patients presenting a proximal fracture of the humerus admitted to a teaching hospital from November 2007 to November 2008 were included using a standardised computer form. A CT-scan was obtained if necessary. Fractures were classified by three senior observers (CP, GM, AR) according to the Neer and AO classifications.
Results: Two hundred forty-seven fractures were collected in 75 men (30%) and 172 women (70%), mean age 66 years (18–97). There were 112 fractures on the dominant side (45%). Two patients had vessel injury and one an associated injury of the brachial plexus. One patient had an isolated injury to the axillary nerve. According to the Neer classification which describes 15 types of fractures, there was little or no displacement or 38% of the fractures and 97.5% of the fractures were classified within six groups: little or no displacement, surgical neck, trochiter fracture alone or with anterior dislocation, 3 or 4 fragment fractures. Using the nine subtypes of the AO classification, there were 58 A1, 55 A2, 42 A3, 43 B1, 9 B2, 5 B3, 14 C1, 18 C2 and 3 C3. Groups A and B included 88% of the fractures. Regarding the CT-scan, obtained in 40% of patients, changed the radiographic interpretation in six cases. Interobserver reproducibility was good. Orthopaedic treatment was given for 203 patients (82%). Operations were: fixation of the tuberosities (n=7), anterograde nailing (n=29), hemiarthroplasty (n=6), reversed prosthesis (n=2).
Discussion: This distribution of fractures of the proximal humerus corresponds well with data in the literature. Good quality x-rays can provide adequate classification without a CT-scan for the majority of patients. The classification systems currently used are quite exhaustive although the distribution in the subgroups is not homogeneous.
