Text

Objectives and questions: Accurate quantification of glenoid bone loss is a key determinant in the management of anterior shoulder instability and directly influences surgical decision-making. Three-dimensional computed tomography is the standard for assessing glenoid morphology. Different measurement techniques—including linear, area-based, and height-derived approaches – are commonly used; however, these methods differ conceptually, and their comparative reliability and accuracy remain incompletely defined. The aim of this study was to compare the accuracy, inter-rater reliability, and intra-rater reliability of three commonly used 3D glenoid bone loss measurement methods using bilateral glenoid data as a reference standard.

Material and methods: A retrospective 3D imaging study was conducted to evaluate measurement accuracy and reliability. Three independent raters (two surgeons, one engineer) assessed 41 shoulders with anterior instability and an intact contralateral glenoid. Measurements were performed at two time points using three methods: linear, area-based, and height–width ratio–based. For each glenoid, raters defined the best-fit circle and height–width parameters using standardized digital tools. Custom scripts extracted geometric parameters including bone loss magnitude, circle size and position, and glenoid height and width. A semi-automated bilateral comparison served as the reference standard for linear glenoid bone loss and was independently validated. Accuracy was evaluated using mean differences and root mean square error (RMSE), and inter- and intra-rater reliability was assessed using ICCs and Bland–Altman analysis.

Results: The average inter-rater difference was 4.06 ± 3.46% for linear, 3.85 ± 3.38% for area-based, and 7.56 ± 7.14% for height–width ratio–based measurements, indicating greater variability for the height–width method. Compared with the ground truth, linear measurements demonstrated a mean difference of 3.90 ± 3.46%, with RMSE values ranging from 4.79 to 5.61 %. Intra-rater reliability was good to excellent for best-fit circle–based techniques, with ICC(3,1) ranging from 0.80–0.95 for linear and 0.81–0.95 for area-based measurements. In contrast, height–width ratio–based measurements showed poor to moderate reliability (ICC 0.32–0.49). Height measurements alone showed excellent reproducibility (ICC 0.88–0.99), whereas width measurements were substantially less reliable (ICC 0.45–0.51), accounting for the reduced reliability of the height–width method.

Discussion and conclusions: The choice of measurement method has clinically relevant implications for the assessment of glenoid bone loss in anterior shoulder instability. Best-fit circle–based techniques provide a more robust foundation for surgical planning by reducing measurement variability and minimizing the risk of over- or underestimation of bone loss. Height–width ratio–based approaches should be interpreted with caution when used to guide treatment decisions.