Scapular morphology is associated with certain patterns of glenohumeral osteoarthritis but not with full-thickness rotator cuff tears.

BACKGROUND

The role of anatomic factors in rotator cuff (RC) pathology or in the development of shoulder osteoarthritis (OA) is not yet well understood or recognized. While 2-dimensional imaging and measurements are easy to capture and interpret, three-dimensional (3D) methods provide a more thorough description and analysis of the scapula's complex shape. The purpose of this study is to investigate whether scapular anatomy is associated with RC tears or glenohumeral OA using 3D statistical shape modelling.

METHODS

A total of 126 computed tomography scans of complete scapulae were captured from 3 cohorts: RC pathology, OA, and control (C). The RC cohort included patients with posterosuperior RC tears, the OA cohort included patients with glenohumeral OA, and the C cohort included patients without shoulder pathology, imaged for other purposes. The RC and OA cohorts each included 55 patients matched by age and gender, while the C cohort included 16 patients, without a specific matching. A full segmentation of each scapula was manually created and 3D surfaces were reconstructed for input to a statistical shape modelling analysis. Statistical shape modelling-based correspondence particles were analyzed using principal component analysis and linear discriminant analysis. Various landmarks were manually identified on each scapula. From these landmarks, 3D measurements of glenoid version and inclination, critical shoulder angle, acromial angle, among others, were obtained. Measurements were compared between the 3 cohorts using analysis of variance or Student's T-test, as appropriate.

RESULTS

There were no statistically significant differences in mean age (RC: 57 ± 9, OA: 58 ± 8, and C: 61 ± 7 years) or gender proportion (males: RC: 51%, OA: 51%, and C: 75%) between the 3 cohorts. A total of 13 principal component analysis modes captured significant shape variation (83.3% of overall variation) of the scapula. The linear discriminant analysis was significantly different between OA and both RC and C (RC: 1.0 ± 1.4, OA: -1.0 ± 1.9, and C: 1.4 ± 1.2). Compared to C, the OA cohort had smaller critical shoulder angle (24.9 ± 6.1° vs. 32.4 ± 4.3°; P < .001), larger acromion angle (65.4 ± 9.3° vs. 57.4 ± 9.5°; P = .019), among other parameters. Compared to RC, the OA cohort also had less distance between coracoid and acromion (47.2 ± 5.8° vs. 44.2 ± 6.4°; P = .026), as well as similar morphologic differences as to the C cohort. There were no significant differences between RC and C cohort.

CONCLUSIONS

Principal component analysis and linear discriminant analysis applied to manually segmented scapulae revealed statistically significant differences in scapular morphology between OA and control scapulae (a more superior and more horizontal acromion with a lower critical shoulder angle), but only minimal differences between RC and control scapulae (except a more anterior anterolateral corner of the acromion).