BackgroundPrevious studies evaluating the reproducibility of glenoid version measurement rely on the medial border of the scapula. Despite this common protocol, there are little data demonstrating the necessity of the full scapular body. The purpose of this study was to evaluate the reproducibility of glenoid version measurement using CT and MRI with and without visualization of the medial border of the scapula. MethodsPatients were retrospectively identified with both preoperative CT and MRI studies before planned shoulder arthroplasty surgery. Exclusion criteria included patients with studies lacking an axial view and CT studies that did not include the medial border of the scapula. A single image was captured for each study. One copy of the CT images was then made with the medial border of the scapula digitally subtracted, leaving 4 cm of the scapular body in view. A survey was created using full CT images, corresponding CT images with 4 cm of the scapular body and MRI images (with 4 cm of the scapular body) for ten shoulder and elbow surgeons to provide glenoid version measurements. Participants remeasured the glenoid version in all the images after the first measurement to allow for intraobserver reliability. Variation was compared within each category. Interobserver and intraobserver reliability (ICC: intraclass correlation coefficient) were determined using weighted intraclass correlation coefficient values for glenoid version measurements. ResultsA total of 68 patients were available for analysis. The 68 patients then had 204 corresponding images (Full CT, 4 cm CT & MRI). The mean glenoid version for the full CT images, the 4 cm CT, and the 4 cm MRI were −12.0° (±8.4°), −15.2 (±8.1), & −13.6° (±8.2°) respectively. These were statistically different (P < .001). The inter-rater reliability demonstrated moderate agreement (ICC = 0.692-0.709) for the CT, the 4 cm CT (ICC = 0.554-0.580) and MRI (ICC = 0.664-0.715). Intra-rater reliability had excellent agreement (ICC = 0.933) for the CT and good agreement for the 4 cm CT (ICC = 0.841) and MRI (ICC = 0.861). ConclusionAlthough the MRI demonstrated significantly greater retroversion compared with the full CT, this was not considered to be clinically significant (>3°) whereas a truncated CT may provide less accurate version measurements. All imaging modalities yielded moderate agreement when evaluated by shoulder surgeons. Based on the results of this study, it appears accurate to measure glenoid version on MRI, even when the entire scapula is not well visualized.