Distal humerus fractures are difficult to characterise and to classify according to the AO system. In this multicentre study, our objectives were to assess the usefulness of computed tomography (CT) and to measure intra-observer and inter-observer reliability according to observer experience.

An online survey of professional practice was performed using a questionnaire based on a clinical case. Participants were asked to determine the AO classification using radiographs then to reappraise their answers after the addition of CT images. For the reliability study, 16 observers in five centres evaluated radiographs and CT scans of 26 distal humerus fractures. They used the radiographs to determine the AO classification and assess the main fracture characteristics then reappraised their findings after adding the CT images. The radiographs and 2D CT images were read twice at an interval of 2 weeks, and during the second reading, 3D CT images were available also. At least 1 month later, the same observers performed similar readings 2 weeks apart (radiographs and 2D CT images at the first reading and addition of 3D CT images at the second reading).

Correct fracture classification was achieved in 95% of cases with the CT images compared to only 73% with the radiographs. CT led to diagnostic and therapeutic changes in 90% and 25% of cases, respectively. Inter-observer reliability was poor for both AO classification and fracture characteristics, not only with the radiographs and 2D CT images, but also with the added 3D CT images. In contrast, intra-observer reliability improved after the addition of 3D CT images. Assessment accuracy was influenced by image quality and geographic origin of the observer but not by observer experience.

CT improves diagnostic accuracy and, in some cases, changes the surgical strategy. In our study of a large number of observers, CT did not improve inter-observer agreement about the study variables. Intra-observer agreement was improved by 3D CT but not by 2D CT. Accuracy was not influenced by years of observer experience but was dependent on image quality, proficiency with computer-based tools and, above all, image observation and interpretation.

Level III.