0701 Perceived Exercise Response Heterogeneity in Adults with Restless Legs Syndrome

Abstract Introduction Restless legs syndrome (RLS) is a prevalent, sensorimotor sleep disorder that is temporarily relieved by movement. There is evidence of inter-individual variation in the response to exercise (i.e., response heterogeneity). We examined factors that may account for inter-individual differences in perceived response to exercise as a mode of treatment for RLS symptoms. Methods Participants (N=528) completed a mixed-methods survey including items assessing RLS outcomes, physical activity levels, and personal experiences with exercise. Participants were classified as “perceived responders” (i.e., exercise impacts symptoms) and “perceived non-responders” (i.e., exercise does not impact symptoms). Perceived responders were further classified as “positive responders” (i.e., exercise improves symptoms) or “negative responders” (i.e., exercise exacerbates symptoms). In this pilot study, an XGboost classifier was used to predict responsiveness to exercise based on clinical and demographic characteristics. Two binary models were implemented, one for non-responders vs responders (M1) and the other for positive responders vs. negative responders (M2). Each model datasets were split into training and test sets. Results There were no significant differences in demographic or clinical characteristics between perceived responders (n=450) and perceived non-responders (n=76). Perceived positive responders (n=172) reported significantly lower BMI, overall RLS severity, and had a lower proportion of people with periodic limb movements than perceived negative responders (n=72). M1 classifier resulted in weighted F1 score, sensitivity, and specificity of 0.92, 0.92 and 1 for training and 0.83, 0.8, 0 for testing, respectively. Whereas the M2 classifier resulted in a weighted F1 score, sensitivity, and specificity of 0.9, 0.90, 0.95 for training and 0.6, 0.79, 0 for testing, respectively. XGBoost feature importance ranked RLS pregnancy incidence, RLS time onset (morning, afternoon), total RLS severity score, physical activity level, BMI, age, and sedentary time as most important. Conclusion The present study reveals factors that correlate with individual differences in response to exercise for people with RLS. If confirmed, such individual differences and specific attributes may be useful to optimize individual treatment plans that include exercise as a recommendation. Support (if any) This work was supported, in part, by the National Heart, lung, and Blood institute [T32HL110952].