Poor reproducibility of classification of proximal humeral fractures. Additional CT of minor value

Fractures of the proximal humerus can be described using the Neer and AO fracture classifications. To assess the reproducibility and reliability of these classifications, we investigated 26 proximal humeral fractures with both plain radiographs and CT. 5 specialists in orthopedic surgery and 5 specialists in radiology independently classified all radiographs on 2 occasions. There was a moderate agreement between the observers when using the Neer classification, but only a fair agreement with the AO classification. The Neer system had a kappa value of 0.42 and the AO had a value of 0.31 in the first assessment. In the second assessment the kappa values were 0.45 and 0.30, respectively. Intraobserver reproducibility was slight to almost perfect agreement with Neer (kappa range 0.20-0.85) and slight to moderate agreement with AO (kappa range 0.16-0.60). The observers most familiar with shoulder fracture radiographs and shoulder fracture treatment were more consistent in their classifications. We conclude that even with CT, the fracture classifications of Neer and AO have a low consistency. Neither classification system is reproducible enough to allow comparisons of different studies.