Improving bone mineral density assessment in total shoulder arthroplasty: assessment of phantom-calibrated Hounsfield units in the glenoid and proximal humerus

Abstract

Background

Bone mineral density (BMD) plays a critical role in the success of total shoulder arthroplasty, as poor bone quality is associated with poor implant fixation, particularly in short stem and stemless designs that rely on metaphyseal fixation. However, dual-energy X-ray absorptiometry (DEXA) scans are not routinely obtained pre-operatively. Computed tomography–derived Hounsfield units (HUs) have emerged as a potential surrogate for BMD assessment, but HU values can have high variability, which limits accuracy. Phantom calibration may standardize HU values, but its benefit in the shoulder remains unclear. This study aimed to compare correlations of standard and phantom-calibrated (PC)-HU with DEXA-derived BMD, assess their diagnostic performance with receiver operating characteristic analysis, and identify HU thresholds for osteoporosis detection.

Materials and methods

A retrospective analysis was performed on 84 shoulders from 76 patients, with some patients contributing bilateral shoulders. All patients underwent both shoulder computed tomography and DEXA within 1 year of total shoulder arthroplasty. Demographics, clinical information, and behavioral factors were collected. HU values were measured from the glenoid and proximal humerus and were calibrated to mg/cc using a phantom. Correlations between HU and DEXA T-scores in the femoral neck, total hip, and spine were evaluated using Spearman correlation. Receiver operating characteristic curve analysis was used to identify optimal HU thresholds for osteoporosis classification.

Results

Both standard and PC HU (mg/cc) values of the glenoid and proximal humerus significantly correlated with DEXA T-scores (P < .001), though standard HU generally demonstrated stronger correlations. Optimal thresholds for identifying osteoporosis were 185.38 for the non-PC glenoid (sensitivity = 83.3%, specificity = 65.7%) and 30.90 for the proximal humerus (sensitivity = 78.6%, specificity = 75%). The PC glenoid threshold of 236.44 yielded slightly higher sensitivity (90.9%) but lower specificity (59.0%). PC proximal humerus HU did not demonstrate significant discriminatory power.

Conclusion

The optimal thresholds for identifying osteoporosis were 185.38 HU for the non-PC glenoid, 30.90 HU for the non-PC proximal humerus, and 236.44 mg/cc for the PC glenoid. Both standard and PC HU values of the glenoid and proximal humerus significantly correlated with DEXA T-scores, although standard HU generally showed stronger correlations. These findings suggest that phantom-based calibration at these sites may offer no additional benefit in assessing systemic BMD.