The Movement Disorder Spectrum of ATP1A3-Related Disorders: Cross-Sectional Analysis and Video Archive of 88 Patients.

ATP1A3-related disorders are characterized by genetic heterogeneity and phenotypic pleiotropy, posing significant challenges for classification. Although canonical phenotypes have traditionally guided decision-making, increasing evidence highlights their limitations in capturing the clinical complexity. The aims of this study were to characterize movement disorders, paroxysmal features, and genotype-phenotype relationships; to build a curated video archive; and to assess alignment with canonical phenotypes. This is an observational study of 88 individuals with pathogenic or likely pathogenic variants in ATP1A3 who were evaluated in specialized movement disorders programs. Age at last clinical follow-up ranged from 0.1 to 63 years; 80.7% were pediatric patients. Chronic movement disorders were present in 68 of 88 individuals (75%); most had two or more coexisting phenomenologies. Dystonia was most common (47/88, 53%), followed by spasticity (28/88, 32%) and ataxia (28/92, 32%). Paroxysmal events occurred in 78 of 88 (88%) patients, including dystonic spells (45/78, 58%), abnormal eye movements (39/78, 50%), and hemiplegic episodes (37/78, 47%). Common comorbidities included epilepsy (21/88, 24%), cognitive impairment (41/88, 47%), and neuropsychiatric disorders. Only 22 of 88 (25%) fulfilled criteria for a single canonical phenotype; 28 of 88 (32%) met canonical criteria plus additional features, 18 of 88 (20%) satisfied criteria for ≥2 canonical phenotypes, and 20 of 88 (23%) fit no canonical category. We identified 43 distinct ATP1A3 variants; recurrent variants (eg, p.Arg756His, p.Asp801Asn, p.Glu818Lys) showed variable expressivity across categories. The extensive clinical heterogeneity in ATP1A3-related disorders challenges rigid phenotypic classifications. The predominance of patients with overlapping or atypical features supports a shift toward flexible, symptom-based clinical approaches rather than strict reliance on canonical phenotype recognition. © 2026 International Parkinson and Movement Disorder Society.