Opportunistic Chest CT Paraspinal Muscle Biomarkers to Predict Longitudinal Thoracic Vertebral Bone Mineral Density Loss: Results from MESA.
Authors
Affiliations (9)
Affiliations (9)
- Department of Radiology, University of Pennsylvania, 3400 Spruce St, 1 Silverstein, Philadelphia, PA 19104.
- Department of Radiology and Radiologic Sciences, Johns Hopkins University, Baltimore, Md.
- Department of Radiology, Houston Methodist, Houston, Tex.
- Department of Family Medicine, University of California, San Diego, La Jolla, Calif.
- Departments of Medicine and Epidemiology, Columbia University Medical Center, New York, NY.
- Department of Radiology, University of Wisconsin School of Medicine and Public Health, Madison, Wis.
- Division of Cardiology, Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, Md.
- Division of Cardiology, Department of Medicine, University of California, San Francisco, San Francisco, Calif.
- Lundquist Institute at Harbor-University of California Los Angeles School of Medicine, Torrance, Calif.
Abstract
Background Muscle and bone loss commonly coexist with aging, but the role of paraspinal muscle in predicting longitudinal thoracic vertebral bone mineral density (BMD) remains poorly understood. Purpose To evaluate whether CT-derived paraspinal muscle attenuation and volume independently predict longitudinal thoracic vertebral BMD loss in a prospective, multicenter cohort. Materials and Methods In this secondary analysis of the Multi-Ethnic Study of Atherosclerosis (MESA), noncontrast chest CT scans from examinations 5 (baseline, April 2010) and 6 (September 2016) were analyzed after excluding participants with incomplete Fracture Risk Assessment Tool without BMD (FRAXnb) data or failed segmentations. Multilevel (T1 through T10) deep learning (DL)-based three-dimensional segmentation was used to derive paraspinal muscle volume and attenuation, which were compared with manual segmentation (reference standard). Paraspinal muscle volume and attenuation were dichotomized (quartile 1 vs quartiles 2-4) for multivariable regression models predicting follow-up vertebral BMD and incident vertebral fracture (VF), adjusted for clinical risk factors and compared with baseline BMD. Results A total of 1316 participants (mean age, 67.2 years [IQR, 60-74 years]; 694 female) were included. DL-derived three-dimensional segmentations correlated with manual segmentations (Dice score, 0.94). At baseline, participants with low versus high attenuation (quartile 1 vs quartiles 2-4; 4 HU vs 27 HU) were older (70.7 years vs 66.0 years) and were more often White (49% [162 of 328 participants] vs 30% [295 of 988 participants]) and female (77% [254 of 328 participants] vs 45% [440 of 988 participants]). Compared with quartiles 2-4, low attenuation (176 mg/cm<sup>3</sup> vs 205 mg/cm<sup>3</sup>; <i>P</i> < .001) and volume (180 mg/cm<sup>3</sup> vs 203 mg/cm<sup>3</sup>; <i>P</i> < .001) were associated with lower BMD. Muscle attenuation at baseline was the strongest predictor of follow-up BMD beyond FRAXnb (adjusted <i>R</i><sup>2</sup>, 0.27 vs 0.20; <i>P</i> < .001) and improved prediction of incident VF beyond BMD alone (area under the receiver operating characteristic curve, 0.82 [95% CI: 0.76, 0.88] vs 0.76 [95% CI: 0.67, 0.85]; <i>P</i> = .02). Conclusion CT-derived paraspinal muscle attenuation provided independent and incremental value for predicting longitudinal thoracic vertebral BMD loss. ClinicalTrials.gov NCT00005487 © RSNA, 2026 <i>Supplemental material is available for this article.</i> See also the editorial by Roemer in this issue. See also the editorial by Rai and Janu in this issue.