A new classification for dislocated and displaced proximal humeral fractures

Background: Although the Neer and AO/OTA classifications have been widely accepted, observer reliability studies of these two classifications have questioned their reliability and reproducibility to date. We developed an entirely new classification, the Mitsuzawa classification, for dislocated and displaced proximal humeral fractures and tested all three classifications for their intra- and interobserver reliability.

Methods: Two experienced shoulder surgeons and two orthopedic residents independently evaluated the Xray (xR) values of 100 proximal humeral fractures (PHFs). The inclusion criteria for PHFs were (1) fracture-dislocation of the glenohumeral joint, (2) severely displaced fracture that required arthroplasty, such as hemi-arthroplasty or reverse shoulder arthroplasty, and (3) age > 18 years. Four reviewers classified all 100 fractures according to the Neer, AO/OTA, and Mitsuzawa classifications on two occasions. The intraobserver reliability was calculated using a Cohen κ statistic, while the interobserver reliability was calculated using a Fleiss κ statistic.

Results: The average intraobserver agreements for the Neer, AO/OTA, and Mitsuzawa classifications were 0.57 (moderate), 0.67 (substantial), and 0.77 (substantial), respectively. The average interobserver agreements for the Neer, AO/OTA, and Mitsuzawa classifications were 0.49 (moderate), 0.56 (moderate), and 0.73 (substantial), respectively. The most common fracture type in each classification was an anterior dislocated fracture with a greater tuberosity fragment, which corresponded to A3a (57 cases) in the Mitsuzawa classification.

Conclusions: The Mitsuzawa classification of PHF incorporates different perspectives regarding glenohumeral compatibility, assessment before and after shoulder dislocation reduction, and the degree of displacement of the proximal stump of the humeral shaft. Compared with the Neer and AO/OTA classifications, our new classification system adopted a user-friendly flowchart format and provided satisfactory intra- and interobserver reliability.

Level of evidence: Level IV.