Computed tomography improves the diagnostic accuracy but not the interobserver reliability of the Boileau classification of proximal humerus fracture sequelae

Background: The aim of this study was to investigate the impact of computed tomography on the reliability of Boileau classification for proximal humerus fracture sequelae. Methods: A prospective study was designed using STARD guidelines. We included all patients diagnosed with proximal humerus fracture sequelae who underwent surgery at our institution between 2017 and 2021. Preoperative radiographs and computed tomography scans were reviewed by three independent observers. Intra- and inter-observer reliability and the diagnostic accuracy of radiographs and computed tomography scans in detecting chronic dislocation, nonunion, and severe greater tuberosity dislocation were assessed. Results: Fifty-two patients were included in the study. The overall interobserver agreement was low on both radiographs and computed tomography scans. On radiographic images, we found a sensitivity of 97%, 88.9%, and 84.1%, and a specificity of 58.3%, 40%, and 53.3% to detect chronic dislocation, nonunion, and greater tuberosity dislocation, respectively. On computed tomography scans we reported a sensitivity of 100%, 96.8%, and 93.7%, and a specificity of 91.7%, 86.7%, and 93.3% to detect chronic dislocation, nonunion, and greater tuberosity dislocation, respectively. Discussion: Computed tomography scan was more specific than radiographs in the assessment of proximal humerus fracture sequelae. However, even using a three-dimensional evaluation of the deformity, the Boileau classification had a poor interobserver reliability