Introduction: Vertebral endplate bone marrow lesions (BML), also referred to as ‘Modic changes’ (MC), are associated with endplate damage, neoinnervation, and chronic low back pain (cLBP). Histology studies show that BMLs represent a continuum of marrow changes reflecting a complex pathological process. Yet, diagnostic classification of BML is binary, based on MC presence/absence on T₁- and T₂-weighted MRI. Advances in quantitative MRI enable measurement of BML composition. The premise of this work is that quantitative assessment of BML composition using water-fat MRI can reveal characteristics of painful BMLs, providing insight into BML pathology and helping to identify phenotypes of cLBP.

Methods: Eighty four patients with cLBP (>3 month duration) were prospectively recruited for this study. 3T MRI of the lumbar spine included sagittal T₁- and T₂-weighted, and chemical shift encoding-based water-fat sequences. MC type was graded on sagittal T₁- and T₂-weighted images for L1–S1 levels. In those vertebrae with MC, BML severity was quantified by calculating the difference in bone marrow fat fraction (BMFF) between regions affected and unaffected by the BML. To do this, BMFF was averaged for five circular ROIs (10 mm² area each) manually placed inside the trabecular bone encompassed by the BML (Fig 1). Five equivalently-sized ROIs were placed in the unaffected adjacent bone marrow. The mean difference in BMFF between affected minus unaffected marrow was calculated for each BML. In the subset of patients with multiple BMLs, the lesion with the largest absolute difference in BMFF between affected and unaffected marrow was used in the analysis. Linear regression was used to test associations between ODI and BML severity, adjusted for age, sex, and BMI. An additional model was tested replacing BML severity with MC type (I, II, or III).

Results: MCs of any type were present in 48% of patients (40/84; 31 male, 9 female; mean±SD age=48.4±13.7 years; ODI=22.7±15.0; VAS=6.4±2.3) and 10% of lumbar endplates (91/924). ODI was significantly associated with the mean difference in BMFF between affected and unaffected marrow (p=0.001, whole-model R²=0.36, Fig 2a): large negative differences were associated with higher ODI scores, whereas smaller negative changes transitioning to positive changes were associated with lower ODI scores. MC type was also significantly associated with ODI (mean ODI=27.6 [95% CI: 21.2, 34.1] versus 16.1 [9.0, 23.2] for MC type I versus II, respectively, p=0.011, whole-model R²=0.28, Fig 2b).

Discussion: These results indicate that disability in patients with endplate BMLs is related to BML composition assessed along a continuum of marrow changes. BMLs with more severe fibrovascular changes—presenting as greater negative differences in BMFF between affected and unaffected regions—were associated with greater ODI scores. Binary classification of MC type explained less variance (lower R²) with less statistical significance (higher p-value) than the continuous measures from water-fat MRI. This new assessment methodology may help identify characteristics of painful BML, visualize BMLs to their maximum extent (Fig 1 c–d), and improve cLBP phenotyping over the legacy clinical classification system.