Clinical and Polysomnographic Heterogeneity in Obstructive Sleep Apnea: a Cluster Analysis of the Berlin Cohort from the European Sleep Apnea Database (ESADA)

INTRODUCTION:Obstructive sleep apnea (OSA) is a clinically heterogeneous disorder. Patients with similar disease severity differ in symptoms and nocturnal oxygenation. However, OSA is still primarily classified according to AHI thresholds, which may not fully reflect the diversity of clinical presentations in OSA. In this study, we applied cluster analysis using clinical and PSG variables to explore heterogeneity in OSA. METHODS:Patients with OSA were included in this monocenter observational study. Unsupervised clustering (k-means clustering) based on clinical characteristics and polysomnographic parameters was performed to explore patterns of heterogeneity in OSA. RESULTS:A total of 248 patients were included in our study. After clustering analysis, three clusters were identified. Cluster 1 (n = 29, 11.7%, Obese hypoxic OSA) showed the highest AHI and oxygen desaturation index (ODI), the lowest mean saturation oxygen(SaO2)and lowest minimal SaO2, the highest body mass index (BMI) and Epworth Sleepiness Scale scores (ESS), and the youngest mean age. Cluster 2 (n = 157, 63.3%, Mild OSA) had the lowest AHI and ODI, the highest mean SaO2, the highest minimal SaO2, and the lowest ESS. Cluster 3 (n = 62, 25%, Older hypoxic OSA) included predominantly patients with severe OSA and showed high AHI and ODI, reduced oxygenation, intermediate BMI and ESS, and the highest mean age. By conventional AHI severity classification, differences were observed in objective measures of disease severity, whereas ESS did not differ between groups. CONCLUSIONS:Three clusters with differing clinical and polysomnographic characteristics were identified among patients with OSA. These findings may provide additional insight into the heterogeneity of OSA.