261 Clustering in Central Disorders of Hypersomnolence biomarker [325]. It is important to keep in mind that chronic sleep deprivation and sometimes circadian disturbances also may show a similar difference. These disorders are, however, currently not included in our database. Extended 32-hour polysomnography recordings [164, 339] or actigraphy assessment of weekend-week sleep length differences during consecutive weekends may help find all potential causes for hypersomnolence without cataplexy [338, 340]. Our analyses highlight the potential of adding quality of awakening variables and weekend-week sleep length difference as sensitive new clinical biomarkers for future diagnostic criteria because they are potentially more reliable than overall mean sleep duration exclusively. Clinical interviews and questionnaires are normally used to assess these variables [341], but optimal quantification practices need to be validated through future studies. Because the number of clusters was determined on the subgrouping of people without cataplexy, the differences between clusters 1 through 4 and 7 should not be overinterpreted. Clusters 1 through 4 should instead be treated as validation of the algorithm to separate people with and without cataplexy and to identify subtypes rather than strict subgroups within the narcolepsy type 1 population that likely reflect different severities. Among clusters 1 through 4, the dendrogram (Appendix E) indicates that cluster 2 (mild severity with virtually absent hypnagogic hallucinations and sleep paralysis) is the closest to clusters 5 and 6 (clusters without cataplexy). The striking female preponderance within cluster 4, in combination with mild cataplexy, hypnagogic hallucinations, and sleep paralysis, suggests the existence of a female narcolepsy type 1 subtype. Possible sex-specific narcolepsy and the influence of hormone levels should be investigated in future studies because such mild cataplexy with fewer triggers may be easily overlooked by clinicians. The influence of sex on narcolepsy symptomatology remains essentially understudied, even though longer diagnostic delay in females individuals has previously been reported [342]. Interpretation of cluster 7 remains difficult without evident differential variables, apart from the presence of hypnagogic hallucinations and sleep paralysis. Cluster 7 is the smallest cluster among all with poor distinctness and clustering reproducibility (Appendix A) and could therefore reflect an inhomogeneous group of borderline phenotypes with limited clinical relevance. The strong differences between clusters in the significances barcodes (Figure 2) and the substantially better goodness of fit in the original clustering compared with the randomly generated datasets (Appendix A) emphasize the possibility of identifying distinct subgroups within the larger central hypersomnolence population. Lack of a clear clustering evaluation metrics peak suggests that, independently of the number of subgroups, there will 9
RkJQdWJsaXNoZXIy MjY0ODMw