Bone and Joint Institute
Estimating glenoid width for instability-related bone loss: A CT evaluation of an MRI formula
Document Type
Article
Publication Date
1-1-2015
Journal
American Journal of Sports Medicine
First Page
1726
Last Page
1730
URL with Digital Object Identifier
10.1177/0363546515581468
Abstract
© 2014 The Author(s). Background: Determining the magnitude of glenoid bone loss in cases of shoulder instability is an important step in selecting the optimal reconstructive procedure. Recently, a formula has been proposed that estimates native glenoid width based on magnetic resonance imaging (MRI) measurements of height (1/3 3 glenoid height + 15 mm). This technique, however, has not been validated for use with computed tomography (CT), which is often the preferred imaging modality to assess bone deficiencies. Purpose: The purpose of this project was 2-fold: (1) to determine if the MRI-based formula that predicts glenoid width from height is valid with CT and (2) to determine if a more accurate regression can be resolved for use specifically with CT data. Study Design: Descriptive laboratory study. Methods: Ninety normal shoulder CT scans with preserved osseous anatomy were drawn from an existing database and analyzed. Measurements of glenoid height and width were performed by 2 observers on reconstructed 3-dimensional models. After assessment of reliability, the data were correlated, and regression models were created for male and female shoulders. The accuracy of the MRI-based model's predictions was then compared with that of the CT-based models. Results: Intra- and interrater reliabilities were good to excellent for height and width, with intraclass correlation coefficients of 0.765 to 0.992. The height and width values had a strong correlation of 0.900 (P<.001). Regression analyses for male and female shoulders produced CT-specific formulas: for men, glenoid width = 2/3 3 glenoid height + 5 mm; for women, glenoid width = 2/3 3 glenoid height 1 3 mm. Comparison of predictions from the MRI- and CT-specific formulas demonstrated good agreement (intraclass correlation coefficient = 0.818). The CT-specific formulas produced a root mean squared error of 1.2 mm, whereas application of the MRI-specific formula to CT images resulted in a root mean squared error of 1.5 mm. Conclusion: Use of the MRI-based formula on CT scans to predict glenoid width produced estimates that were nearly as accurate as the CT-specific formulas. The CT-specific formulas, however, are more accurate at predicting native glenoid width when applied to CT data. Clinical Relevance: Imaging-specific (CT and MRI) formulas have been developed to estimate glenoid bone loss in patients with instability. The CT-specific formula can accurately predict native glenoid width, having an error of only 2.2% of average glenoid width.
