Patterns of Muscle Health in Single- and Multi-Site Chronic Pain: A UK Biobank Normative Modeling Study
Merve Kaptan1, Yiyu Wang1, Augustijn de Boer2,3
1Division of Pain Medicine, Department of Anesthesiology, Perioperative and Pain Medicine, Stanford University School of Medicine, Palo Alto, CA, USA.
Background:
Chronic pain is associated with impaired muscle health, but whether these changes reflect site-specific factors, broader systemic factors, or both remains unclear. The purpose of this study is to determine whether normative markers of muscle health derived from MRI show site-specific patterns in chronic pain.
Methods:
UK Biobank participants who underwent whole-body MRI from 2006 to 2010 were included in this retrospective cross-sectional study. The MuscleMap Toolbox quantified volume and intramuscular fat (IMF) in 42 muscles of the abdomen, pelvis, and thigh. Normative models trained on a no pain group generated muscle-specific deviations from normal (i.e., Z-scores) for single- and multi-site chronic and acute pain.
Results:
Of 17,843 participants, the primary site-specific analysis included 9,704 no pain, 885 single-site chronic back pain (CBP), 438 single-site chronic hip pain (CHP), and 1,315 single-site chronic knee pain (CKP) participants (n=12,342; mean age 63.7±7.5 years; 52.7% female). Additional analyses included single-site chronic neck/shoulder pain, acute pain, and multi-site chronic pain groups. In CBP, deviations were localized to abdominal muscles, with decreased volume in 6/8 and increased IMF in 6/8. In CHP, deviations were broad, with decreased volume in 3/8 of the abdominal and 14/26 of the thigh muscles, and increased IMF in 6/8 of the abdominal, 5/8 of the pelvic, and 4/26 of the thigh muscles. In CKP, deviations were localized to thigh muscles, with decreased volume in 8/26 and increased IMF in 6/26. Acute pain groups showed no significant differences except for decreased volume in one thigh muscle in acute knee pain. With each additional chronic pain site, volume decreased (β=-.078;IQR:-0.100-0.051), and IMF increased (β=.085;IQR:0.066-0.101). Combined Z-scores classified chronic pain groups better than chance (accuracy: 48.6%;p<.001), but not acute pain groups (accuracy: 39.0%;p=.20).
Conclusions:
Whole-body MRI combined with AI-driven muscle segmentation and normative modeling revealed site-specific patterns of muscle health in single-site chronic pain.
