Background: The magnetic resonance imaging (MRI) parameters used to diagnose rotator cuff tears are weakly correlated to shoulder function. Our hypothesis was that adding 3-dimensional morphologic parameters resulting from biplanar radiographs (3DXR parameters) to the MRI parameters would improve this correlation. Methods: We assessed 52 patients with rotator cuff tears with an EOS Imaging radiographic examination, MRI study, and clinical evaluation of the shoulder, as well as the Constant score. The bones of the 52 shoulders were reconstructed 3-dimensionally, and eleven 3DXR parameters were automatically extracted. First, the trueness and reliability of these parameters were evaluated. Then, bivariate correlations between each parameter and the Constant score were made. A linear regression model was subsequently built to correlate the 11 parameters and 5 MRI findings with shoulder function at diagnosis, as assessed by the Constant score. Results: The parameters showed good trueness and reliability of most 3DXR parameters. Supraspinatus tear extension, muscle atrophy, and the distance between the greater and deltoid tuberosities were the only parameters with a statistically significant correlation to a lower Constant score (P < .05) in the bivariate study. These correlations were either weak or negligible. A regression model was successfully built with one MRI parameter and four 3DXR parameters. Correlation to function increased from 16.7% to 43.3% with this model. Conclusion: For patients with rotator cuff tears, the combination of MRI and 3DXR parameters of the shoulder in a linear regression model improves the correlation with the Constant score (shoulder function) at diagnosis.