Multi-perspective clustering of obstructive sleep apnea towards precision therapeutic decision including craniofacial intervention

• Published in: Sleep & breathing 2021-03, Vol.25 (1), p.85-94
• Main Authors: Kim, Su-Jung, Alnakhli, Waleed Maqbul, Alfaraj, Ali Saeed, Kim, Kyung-A, Kim, Sung-Wan, Liu, Stanley Yung-Chuan
• Format: Article
• Language: English
• Subjects: Apnea; Arousal; Body mass index; Cluster analysis; Comorbidity; Decision making; Dentistry; Diabetes mellitus; Genotype & phenotype; Hypertension; Internal Medicine; Medicine; Medicine & Public Health; Neurology; Obesity; Otorhinolaryngology; Pediatrics; Phenotypes; Phenotyping; Pneumology/Respiratory System; Principal components analysis; Risk factors; Sleep; Sleep and wakefulness; Sleep apnea; Sleep Breathing Physiology and Disorders • Original Article; Sleep disorders
• ISSN: 1520-9512; 1522-1709
• DOI: 10.1007/s11325-020-02062-9

Purpose Previous studies focusing on phenotyping obstructive sleep apnea (OSA) have outlined its heterogeneity in clinical symptoms, comorbidities, and polysomnographic features. However, the role of anatomical or pathophysiological causality including craniofacial skeletal deformity has not been studied. We aimed to identify and characterize phenotypes of OSA based on multi-perspective clustering by incorporating craniofacial risks with obesity, apnea severity, arousability, symptom, and comorbidity. Methods A total of 421 Korean patients with OSA (apnea-hypopnea index, AHI ≥ 5; age ≥ 20 years old) were recruited. A K-means cluster analysis was performed following principal component analysis with sagittal and vertical skeletal variables (ANB and mandibular plane angle), AHI, body mass index, and Epworth sleepiness scale. Inter-cluster comparison was conducted using demographic, cephalometric, and polysomnographic variables in addition to presence of diabetes and hypertension. Risk factors contributing to OSA severity were evaluated in each cluster using multivariable regression analysis with adjustment for age and gender. Results Three phenotypic clusters were identified and characterized as follows: Cluster-1 (noncraniofacial phenotype, 39%), non-obese moderate-to-severe OSA with no skeletal discrepancy representing low arousal threshold (ArTh), little sleepiness, and low comorbidity; Cluster-2 (craniofacial skeletal phenotype, 33%), non-obese moderate OSA with definite skeletal discrepancy showing low ArTh, mild sleepiness, and low comorbidity; and Cluster-3 (complicated phenotype, 28%), obese severe OSA with skeletal discrepancy exhibiting high ArTh, excessive daytime sleepiness, and high incidence of hypertension. Conclusions The three OSA phenotypes from multi-perspective clustering may provide a basis for precise therapeutic decision-making including craniofacial skeletal intervention beyond usual characterization of OSA subgroups.