HIGHER INTRA-MUSCULAR FAT WITHIN KNEE EXTENSOR AND ADDUCTOR MUSCLES PREDICTS MORE PAINFUL AND PROGRESSIVE OSTEOARTHRITIS

Abstract

Previous studies have shown that inter- and intramuscular fat, together, relate with knee osteoarthritis symptoms and disease features. However, the manual methods of fat and muscle segmentation from MR images are tedious and limit the feasibility of examining intra-muscular fat within separate muscle groups of the thigh. For this reason, it remains unclear whether fat within specific knee flexor, extensor, adductors or abductors contribute differentially to knee symptoms and trajectories of change in symptoms. To investigate how intramuscular fat within separate muscle groups of the thigh differentially predict knee symptoms and faster progression of knee symptoms over 5 years. This was an individual-level longitudinal study of 134 participants of the Osteoarthritis Initiative randomly selected among those who had thigh MR images available at the 24-month visit (v03). A total of 15 transaxial intermediate weighted turbo spin echo image slices were examined. A fully convolutional neural network was trained to segment all thigh muscle groups using manually traced annotations (3D Slicer) verified by two radiologists. Original training (N=134) and CNN-predicted (N=235) muscle group output masks were both subjected through the iterative threshold -seeking algorithm (ITSA) that we previously developed to quantify fat volume and percentages. All algorithms and computations were completed within Python 3.8.8(Jupyter). Group-based trajectory modeling (GBTM) identified and classified participants according to patterns of change in KOOS and WOMAC scores from 36 to 84 months (v05 – v09). Binary logistic regression models examined how fat volume (absolute and percentage, per standard deviation higher) within each muscle group predicted rapidly worsening symptoms versus relatively unchanged symptom levels. General linear models also predicted 36-month (v05) symptom values using single unit exposure contrasts (1cm 3 larger fat or muscle volume). All models adjusted for age, BMI, use of NSAIDs, and 20-m walking pace. Among 134 individuals (mean age: 64±9yrs; BMI: 30.09±4.67kg/m 2 ), GBTM identified 3 classes of pain trajectory patterns with more rapid progression being a distinct class compared to two groups showing unchanging symptom levels (Figure 1). A higher absolute (+7.03cm 3 ) and relative (0.06%) amount of intramuscular fat within the adductor muscles was associated with a higher odds of having rapid pain progression as measured by KOOS knee pain (OR: ranged from 1.57 to 1.65, concordance(C)=0.81) or WOMAC total score (OR: ranged from 1.45 to 1.56, C=0.63). A similar pattern was observed for a lower hamstring muscle volume (OR: 2.54(95%CI: 1.15,5.59)). On a linear scale, each 1 cm 3 larger fat volume within the rectus femoris was associated with 3.02(4.74,1.30) percentage points lower KOOS knee pain (more pain), 1.60(2.94,0.27) percentage points lower KOOS knee symptoms, and 2.51(1.17,3.85) points higher in WOMAC total score (more pain). A greater distribution of fat and smaller muscle volume within the adductors was predictive of a higher rate of pain worsening. However, a similar pattern of higher fat and lower muscle volume within the rectus femoris muscles was primarily responsible for worse absolute knee pain and symptoms within the next 12 months. CIHR PJT156274. Arthritis Society Ken Smith Stars Career Development Award. None of the authors have disclosures. CORRESPONDENCE ADDRESS: andy.wong@uhnresearch.ca