BACKGROUND: Mutations in the gene ATP1A3 have recently been identified to be prevalent in patients with alternating hemiplegia of childhood (AHC2). Based on a large series of patients with AHC, we set out to identify the spectrum of different mutations within the ATP1A3 gene and further establish any correlation with phenotype. METHODS: Clinical data from an international cohort of 155 AHC patients (84 females, 71 males; between 3 months and 52 years) were gathered using a specifically formulated questionnaire and analysed relative to the mutational ATP1A3 gene data for each patient. RESULTS: In total, 34 different ATP1A3 mutations were detected in 85 % (132/155) patients, seven of which were novel. In general, mutations were found to cluster into five different regions. The most frequent mutations included: p.Asp801Asn (43 %; 57/132), p.Glu815Lys (16 %; 22/132), and p.Gly947Arg (11 %; 15/132). Of these, p.Glu815Lys was associated with a severe phenotype, with more severe intellectual and motor disability. p.Asp801Asn appeared to confer a milder phenotypic expression, and p.Gly947Arg appeared to correlate with the most favourable prognosis, compared to the other two frequent mutations. Overall, the comparison of the clinical profiles suggested a gradient of severity between the three major mutations with differences in intellectual (p = 0.029) and motor (p = 0.039) disabilities being statistically significant. For patients with epilepsy, age at onset of seizures was earlier for patients with either p.Glu815Lys or p.Gly947Arg mutation, compared to those with p.Asp801Asn mutation (p < 0.001). With regards to the five mutation clusters, some clusters appeared to correlate with certain clinical phenotypes. No statistically significant clinical correlations were found between patients with and without ATP1A3 mutations. CONCLUSIONS: Our results, demonstrate a highly variable clinical phenotype in patients with AHC2 that correlates with certain mutations and possibly clusters within the ATP1A3 gene. Our description of the clinical profile of patients with the most frequent mutations and the clinical picture of those with less common mutations confirms the results from previous studies, and further expands the spectrum of genotype-phenotype correlations. Our results may be useful to confirm diagnosis and may influence decisions to ensure appropriate early medical intervention in patients with AHC. They provide a stronger basis for the constitution of more homogeneous groups to be included in clinical trials.