Levothyroxine Use and Longitudinal Changes in Thigh Muscles in At-Risk Participants for Knee Osteoarthritis: Preliminary Analysis from Osteoarthritis Initiative Cohort

Background: We examined the association between levothyroxine use and longitudinal MRI biomarkers for thigh muscle mass and composition in at-risk participants for knee osteoarthritis (KOA) and their mediatory role in subsequent KOA incidence. Methods Using the Osteoarthritis Initiative (OAI) data, we included the thighs and corresponding knees of participants at risk but without established radiographic KOA (baseline Kellgren-Lawrence grade (KL) < 2). Levothyroxine users were defined as self-reported use at all annual follow-up visits until 4th year and were matched with levothyroxine nonusers for potential confounders (KOA risk factors, comorbidities, and relevant medications covariates) using 1:2/3 propensity-score (PS) matching. Using a previously developed and validated deep learning method for thigh segmentation, we assessed the association between levothyroxine use and 4-year longitudinal changes in muscle mass, including cross-sectional area (CSA) and muscle composition biomarkers including intra-MAT (within-muscle fat), contractile percentage (non-fat muscle CSA/total muscle CSA), and specific force (force per CSA). We further assessed whether levothyroxine use is associated with 8-year risk of standard KOA radiographic (KL ≥ 2) and symptomatic incidence (incidence of radiographic KOA and pain on most of the days in the past 12 months). Finally, using a mediation analysis, we assessed whether the association between levothyroxine use and KOA incidence is mediated via muscle changes. Results We included 1,043 matched thighs/knees (266:777 levothyroxine users:non-users; average ± SD age: 61 ± 9 years, female/male:4). Levothyroxine use was associated with decreased quadriceps CSAs (mean difference, 95%CI: -16.06 mm 2 /year, -26.70 – -5.41) but not thigh muscles’ composition (e.g., intra-MAT). Levothyroxine use was also associated with an increased 8-year risk of radiographic (hazard ratio (HR), 95%CI: 1.78, 1.15–2.75 and symptomatic KOA incidence (HR, 95%CI: 1.93, 1.19–3.13). Mediation analysis showed that a decrease in quadriceps mass (i.e., CSA) partially mediated the increased risk of KOA incidence associated with levothyroxine use. Conclusions Our exploratory analyses suggest that levothyroxine use is associated with loss of quadriceps muscle mass, which partially mediates increased risk of subsequent KOA incidence. Study interpretation should consider underlying thyroid function as a potential confounder or effect modifier; and therefore, future studies are warranted to investigate the underlying thyroid function biomarkers for longitudinal changes of thigh muscles.