Data-driven phenotypic clustering of idiopathic restless legs syndrome and its longitudinal clinical correlates

Abstract

Objectives Restless legs syndrome (RLS) shows substantial heterogeneity in symptoms, comorbidity, and treatment response. This study aimed to identify clinically meaningful phenotypic subgroups of idiopathic RLS using an unsupervised, data-driven clustering and to evaluate associations with longitudinal clinical outcomes. Methods We retrospectively analyzed 360 patients with idiopathic RLS who had ≥1 year of follow-up. Hierarchical agglomerative clustering based on Gower distance was performed using seven baseline variables selected a priori: onset age, sex, family history of RLS, International RLS rating scale (IRLS) score, painful RLS sensations, psychiatric comorbidity, and serum ferritin. Results Four distinct phenotypic clusters were identified: (1) late-onset RLS in women without family history, (2) painful RLS phenotype, (3) late-onset RLS in men without family history, and (4) early-onset familial RLS. Overall RLS severity, as measured by IRLS, did not differ significantly across clusters. During the follow-up, the painful RLS phenotype showed more frequent daytime symptoms, greater opioid use, and a higher proportion of refractory course than the late-onset female phenotype. Among late-onset patients, women reported more psychiatric symptoms and poorer sleep-related quality of life than men despite comparable RLS severity. The early-onset familial phenotype also demonstrated relatively frequent opioid use, despite the absence of baseline painful symptoms. Conclusions Unsupervised clustering identified four clinically interpretable phenotypes of idiopathic RLS distinguished primarily by onset age, family history, painful sensations, and sex. These phenotypes, defined by baseline clinical characteristics, were associated with differences in longitudinal treatment patterns and disease course, supporting a phenotype-informed RLS management. © 2026 Elsevier B.V.