Aims: The aim of this study was to assess displaced proximal humeral fractures using 3D-CT. The hypothesis was that it would be possible to classify these fractures systematically based upon recurring anatomical patterns of injury, and that this would be useful in predicting the outcomes. Methods: CT scanning was performed in 993 patients who sustained a proximal humeral fracture between September 2014 and April 2020, with anonymized measurements of the anatomical features and displacement of the head-shaft fracture, and the extent of the involvement of the tuberosities in 3D reconstructed views. We used exploratory factor analysis and cluster analysis to generate anatomically distinct groups of fractures. The classification was simplified by data reduction techniques and cross-tabulation of initial cluster groupings of head-shaft and tuberosity displacement to produce the most practical system possible. We compared the demographic details, the prevalence of soft-tissue injury, and complications between the groups. Five surgeons independently classified 100 CTs from the series on two occasions six weeks apart, to evaluate observer reliability. Results: The mean age of the patients was 65.1 years (21 to 92), 745 (75%) were female, and 844 fractures (85%) occurred in simple falls. Of the 993 fractures which were evaluated, 928 (93.4%) could be classified using three major criteria of the head-shaft Fracture Anatomy, the degree of head-shaft Separation, and number of fractures of the Tuberosities (FAST) classification. Specific anatomical patterns had a higher risk of early skin compromise and neurovascular complications. Nonoperatively treated patients with head-shaft separation had a higher risk of nonunion (one-year hazard ratio (HR) 6.5, 95% CI 3.9 to 10.9), and those with fractures of one or both tuberosities had a higher risk of developing osteonecrosis (five-year HR 4.1, 95% CI 2.1 to 8.1). Substantial levels of observer agreement were achieved for the final classification. Conclusion: Proximal humeral fractures have recurring anatomical features which can be classified using 3D-CT. The classification is useful in counselling patients about the threat of early soft-tissue injury and the risk of delayed complications after nonoperative treatment.
Robinson et al. (Fri,) studied this question.