Three-dimensional printing models increase inter-rater agreement for classification and treatment of proximal humerus fractures

BACKGROUND: Proximal humerus fractures (PHF) are frequent, however, several studies show low inter-rater agreement in the diagnosis and treatment of these injuries. Differences are usually related to the experience of the evaluators and/or the diagnostic methods used. This study was designed to investigate the hypothesis that shoulder surgeons and diagnostic imaging specialists using 3D printing models and shoulder CT scans in assessing proximal humerus fractures. METHODS: We obtained 75 tomographic exams of PHF to print three-dimensional models. After, two shoulder surgeons and two specialists in musculoskeletal imaging diagnostics analyzed CT scans and 3D models according to the Neer and AO/OTA group classification and suggested a treatment recommendation for each fracture based on the two diagnostic methods. RESULTS: The classification agreement for PHF using 3D printing models among the 4 specialists was moderate (global k = 0.470 and 0.544, respectively for AO/OTA and Neer classification) and higher than the CT classification agreement (global k = 0.436 and 0.464, respectively for AO/OTA and Neer). The inter-rater agreement between the two shoulder surgeons were substantial. For the AO/OTA classification, the inter-rater agreement using 3D printing models was higher (k = 0.700) than observed for CT (k = 0.631). For Neer classification, inter-rater agreement with 3D models was similarly higher (k = 0.784) than CT images (k = 0.620). On the other hand, the inter-rater agreement between the two specialists in diagnostic imaging was moderate. In the AO/OTA classification, the agreement using CT was higher (k = 0.532) than using 3D printing models (k = 0.443), while for Neer classification, the agreement was similar for both 3D models (k = 0.478) and CT images (k = 0.421). Finally, the inter-rater agreement in the treatment of PHF by the 2 surgeons was higher for both classifications using 3D printing models (AO/OTA—k = 0.818 for 3D models and k = 0.537 for CT images). For Neer classification, we saw k = 0.727 for 3D printing models and k = 0.651 for CT images. CONCLUSION: The insights from this diagnostic pilot study imply that for shoulder surgeons, 3D printing models improved the diagnostic agreement, especially the treatment indication for PHF compared to CT for both AO/OTA and Neer classifications On the other hand, for specialists in diagnostic imaging, the use of 3D printing models was similar to CT scans for diagnostic agreement using both classifications. TRIAL REGISTRATION: Brazil Platform under no. CAAE 12273519.7.0000.5505.