Nocturnal sleep phenotypes in idiopathic hypersomnia – A data-driven cluster analysis

Abstract

Introduction

The diagnostic process for idiopathic hypersomnia (IH) is complex due to the diverse aetiologies of daytime somnolence, ambiguous pathophysiological understanding, and symptom variability. Current diagnostic instruments, such as the multiple sleep latency test (MSLT), are limited in their ability to fully represent IH's diverse nature. This study endeavours to delineate subgroups among IH patients via cluster analysis of polysomnographic data and to examine the temporal evolution of their symptomatology, aiming to enhance the granularity of understanding and individualized treatment approaches for IH.

Methods

This study included individuals referred to the Uppsala Centre for Sleep Disorders from 2010 to 2019, who were diagnosed with IH based on the International Classification of Sleep Disorders-3 (ICSD-3) criteria, following a thorough diagnostic evaluation. The final cohort, after excluding participants with incomplete data or significant comorbid sleep-related respiratory conditions, comprised 69 subjects, including 49 females and 20 males, with an average age of 40 years. Data were collected through polysomnography (PSG), MSLT, and standardized questionnaires. A two-step cluster analysis was employed to navigate the heterogeneity within IH, focusing on objective time allocation across different sleep stages and sleep efficiency derived from PSG. The study also aimed to track subgroup-specific changes in symptomatology over time, with follow-ups ranging from 21 to 179 months post-diagnosis.

Results

The two-step cluster analysis yielded two distinct groups with a satisfactory silhouette coefficient: Cluster 1 (n = 29; 42 %) and Cluster 2 (n = 40; 58 %). Cluster 1 exhibited increased deep sleep duration, reduced stage 2 sleep, and higher sleep maintenance efficiency compared to Cluster 2. Further analyses of non-clustering variables indicated shorter wake after sleep onset in Cluster 1, but no significant differences in other sleep parameters, MSLT outcomes, body mass index, age, or self-reported measures of sleep inertia or medication usage. Long-term follow-up assessments showed an overall improvement in excessive daytime sleepiness, with no significant inter-cluster differences.

Conclusion

This exploratory two-step cluster analysis of IH-diagnosed patients discerned two subgroups with distinct nocturnal sleep characteristics, aligning with prior findings and endorsing the notion that IH may encompass several phenotypes, each potentially requiring tailored therapeutic strategies. Further research is imperative to substantiate these findings.