Lower Bone Mineral Can Be a Risk for an Enlarging Bone Marrow Lesion: A Longitudinal Cohort Study of Japanese Women Without Radiographic Knee Osteoarthritis

Abstract Background: Although previous cross-sectional studies have clarified that lower bone mineral density (BMD) is associated with more severe bone marrow lesions (BMLs), its longitudinal relationship remains unclear. The aim of this study was to elucidate the longitudinal relationship between BMD at baseline (BL) and BML change during a two-year follow-up period (2YFU). Methods: A total of 78 female participants (Mean age: 54.9 ± 9.6 and BMI: 21.6 ± 3.0 kg/m2) were eligible. All the participants underwent 1.5T magnetic resonance imaging, and fat-suppressed T2-weighted images were acquired. Based on coronal and sagittal images, the maximum BML area (BMLa) was calculated by tracing the BML border. The mean value of coronal and sagittal BMLa was also calculated. The change in mean BMLa was defined using the following formula: [BMLa of 2YFU] – [BMLa of BL] = ΔBMLa. If ΔBMLa was positive, it was defined as enlarged; if ΔBMLa was negative, it was defined as regressed. Dual-energy X-ray absorptiometry was performed to measure the non-dominant side BMD at one-third of the distal radius. Young adult mean [YAM (%)] of the BMD was used for statistical analysis. Linear regression analysis was conducted with ΔBMLa as the dependent variable and YAM as the independent variable, adjusted for age, BMI, pain scale, and BMLa at BL. A receiver operating characteristic curve was drawn for the YAM to predict the prevalence of BML enlargement or regression. Logistic regression analysis was conducted with BML enlargement prevalence or regression as the dependent variable, and with the cut-off of YAM as the independent variable. Statistical significance was defined as P≤0.05. Results: During the 2YFU, 12 (15.4%) patients had enlarged BMLa, 26 (33.3%) participants showed regressing BMLa, and 40 (51.3%) patients remained stable. The YAM was significantly associated with ΔBMLa, according to the linear regression model (B: − 1.459, b: − 0.375, P=0.046). The best predictor of the BML enlargement risk was 85% YAM; this cut-off was significant in predicting the prevalence of BML enlargement (B: 2.126, odds ratio: 8.383, P=0.025), according to the adjusted logistic regression model.Conclusion: Lower BMD could predict BML enlargement during a two-year follow-up period.