A new automated method to define clinically relevant pediatric sleep apnea phenotype

<bold>Background.</bold> Pediatric patients suffering from obstructive sleep apnea (OSA) show different responses to treatment, being the reasons for this heterogeneity not completely understood. The phenotypic characterization of this disease could shed new light to personalized treatments by identifying predictive factors of the patient's evolution. <bold>Aims and Objectives.</bold> To identify subgroups of pediatric OSA patients, as well as the most relevant variables to differentiate them. <bold>Methods.</bold> 26 sociodemographic and clinical variables of 464 pediatric patients (5-10 years old) from the Childhood Adenotonsillectomy Trial dataset were used. A subject-based association network was obtained by measuring the similarity between the data from each pair of pediatric subjects. A modularity analysis was then automatically conducted on the network to find a non-pre-fixed number of OSA subgroups with shared characteristics. <bold>Results.</bold> 3 subgroups of pediatric OSA patients with similar intragroup phenotypic characteristics, but significantly different intergroup features, were identified. The main defining characteristics of each group were age, sex, presence of hypertension, and weight-related characteristics (Figure 1). <bold>Conclusions.</bold> Our findings show the ability of the proposed methodology to find subgroups of OSA pediatric subjects with well-differentiated phenotypical characteristics.