Abstract
OBJECTIVE: To determine whether changes in knee-adjacent subcutaneous fat (kaSCF) are associated with changes in cartilage thickness, cartilage T(2) relaxation times, pain, and radiographic Kellgren Lawrence (KL) grade over four years. DESIGN: This study included 4644 participants from the Osteoarthritis Initiative with annual 3.0T MRIs from baseline to four years. Deep learning algorithms quantified kaSCF, cartilage thickness, and cartilage T(2). Associations between standardized annual changes in kaSCF and cartilage thickness, T(2), and WOMAC pain were evaluated using mixed-effects models. Concurrent (Δsame-year) and lagged (Δprior-year kaSCF predicting Δsubsequent-year outcomes) models were adjusted for age, sex, race, baseline BMI, and BMI change. Cox models evaluated whether time-varying kaSCF was associated with KL grade progression. RESULTS: In concurrent models, a 1 SD greater annual increase in kaSCF was associated with a 0.035 SD greater cartilage thickness loss (β = -0.035; 95% CI -0.055 to -0.015; p = 0.001), but was not significantly associated with cartilage T(2) (β = 0.012; 95% CI -0.009 to 0.032; p = 0.276) or WOMAC pain (p = 0.081). In lagged models, prior-year increases in kaSCF were associated with greater subsequent cartilage thickness loss (β = -0.034; 95% CI -0.058 to -0.009; p = 0.007) and greater cartilage T(2) (β = 0.047; 95% CI 0.022 to 0.072; p < 0.001), while WOMAC pain remained non-significant (p = 0.072). Time-varying kaSCF was not significantly associated with KL progression (HR = 1.06 per SD; p = 0.31). CONCLUSIONS: While effect sizes were modest, increases in kaSCF were associated with greater cartilage thinning and, in lagged models, with greater cartilage T(2) over time. These findings suggest that increases in local knee adiposity may contribute to cartilage loss independent of BMI.