Session VI - Upper Extremity

Understanding Proximal Humerus Fractures: Image Analysis, Classification and Treatment

Michael W. Shrader, MD; Joaquin Sanchez-Sotelo, MD; John W. Sperling, MD; Robert H. Cofield, MD; Mayo Clinic, Rochester, Minnesota, USA

Purpose: It has proven to be difficult to define proximal humerus fractures because of their extreme variability and the potential for complexity. We designed a study to further evaluate why this is true.

Methods: During 1997 and 1998, 113 patients with 113 proximal humerus fractures presented acutely to our institution. We obtained anteroposterior and axillary radiographs, supplemented by scapular Y views for 32 and CT scans for 18. These films were used for decision making by the treating surgeon. Three knowledgeable observers were asked to independently answer nine questions about the fracture: "Is the greater tuberosity fractured or displaced? Is the lesser tuberosity fractured or displaced? Is the shaft fractured or displaced, all relative to the humeral head? Is the head-glenoid relationship preserved? Is there head impaction, or is there head splitting?" They were then asked to classify the fracture according to the Neer classification and to recommend treatment. Two months later, a learning session was held to discuss discrepancies among the observers and to develop "learning points" to improve analysis of the images. Two months later, the three observers again reviewed the films.

Results: Developing ten "learning points" enhanced the ability to interpret images at the second review and provide more consistent fracture classification with statistically significant improvements in five of eight questions about the fracture. However, overall agreement on fracture classification occurred in only 59% of fractures, kappa 0.47. Agreement for one- or two-part fractures was 76%, kappa 0.46, whereas for three- or four-part fractures it was 4%, kappa 0.22. When three of three agreed on interpretation of image questions, agreement on classification was 97%, kappa 0.91. When there was agreement about classification, there was agreement on treatment selection in 97%, kappa 0.77, and also agreement on whether or not surgery was indicated in 97%, kappa 0.77.

Conclusion: Thus, the problem was in understanding the images of the complex fractures not with the classification system nor, apparently, with differences in treatment philosophy. Imaging of complex fractures must be enhanced to improve consistency in understanding these fractures.