Background/Purpose: Osteoporosis (OP) is assessed by dual-energy X-ray absorptiometry (DXA) for bone mineral density (BMD). Computed tomography (CT) offers an alternative for “opportunistic” screening. This study evaluated the performance of CT-derived BMD alone and combined with psoas muscle mass to classify OP based on DXA and predict major osteoporotic fracture (MOF) risk based on FRAX.

Methods: Consecutive abdominal CT scans (2015–2018) of women with DXA available within 1 year of the CT exam were processed to quantify L1 vertebral attenuation and psoas muscle volume. T-scores and EHR data were extracted for age, race, and BMI. We obtained the lowest T score among lumbar spine, femoral neck, or total hip. OP was defined as the lowest T-score ≤ -2.5. CT BMD was defined as L1 attenuation in Hounsfield units. We fit 4 probit models to evaluate the discrimination ability of identifying OP: CT BMD alone (Model 1), CT BMD + psoas volume (Model 2), CT BMD + age + race + BMI (Model 3), and CT BMD + psoas volume + age + race + BMI (Model 4). AUROCs were used to compare discrimination. Clustered bootstrap was used to calculate 95% CI and compare AUROC, accounting for repeated scans. We then built a linear model to predict the lowest T-score from CT BMD, psoas volume, age, race, and BMI by 5-fold cross-validation, with results visualized in a nomogram. A subgroup analysis of women with full FRAX risk factor data available was performed, modeling MOF risk ≥20% as the outcome, using similar analytical approaches.

Results: A total of 1,761 CT scans linked with DXA were analyzed(mean age: 67.5 ± 9.7). ROC curves and AUROCs are shown in Figure 1. Discrimination was modest (AUROC 0.69) using the CT L1 BMD alone and appreciably improved after adding psoas volume (AUROC 0.75). The AUROC for CT L1 BMD, independent of age, BMI, and race, was 0.67 [0.63–0.71], and for psoas volume was 0.62 [0.59–0.66] (p = 0.11). The linear model predicting lowest T-score showed good discrimination (g-index=0.84) and calibration (slope=0.996; see nomogram in Figure 2). However, neither CT BMD alone (AUROC = 0.66 [0.55–0.76]) nor combined with psoas volume (AUROC = 0.66 [0.56–0.76], p = 0.69) effectively identified high MOF risk.

Conclusion: Adding psoas muscle volume to L1 CT BMD from abdominal CT performed for other purposes significantly improved OP classification compared to CT L1 attenuation alone. Combining muscle volume and BMD measures may improve incidental OP diagnosis accuracy for targeted interventions.

Figure 1. Abdominal CT-derived BMD and psoas volume in osteoporosis assessment: ROC curve comparison between models for osteoporosis classification (lowest T-score <=-2.5 vs. > -2.5)

ROC curve comparison between models for osteoporosis classification (*: p < 0.001 vs. Model 1. ^: p=0.18 vs. Model 2. **: p=0.039 vs. Model 3. N&#3f1,442 women)

Figure 2. Regression-based nomogram for linear model predicting lowest T-score among lumbar spine, femoral neck, or total hip based on CT L1 BMD, psoas volume, and clinical variables (age, BMI, race)

« Back to ACR Convergence 2025

ACR Meeting Abstracts - https://acrabstracts.org/abstract/using-abdominal-ct-derived-bone-mineral-density-and-psoas-volume-for-opportunistic-osteoporosis-screening-and-fracture-risk-assessment/