Drug‐Induced Sleep Endoscopy in Children Predicts OSA Severity

ABSTRACT Background While adenotonsillectomy (AT) is first line for pediatric obstructive sleep apnea (OSA), treatment failure is common. Drug‐induced sleep endoscopy (DISE) provides dynamic airway assessment, but its role in predicting OSA severity requires clarification. This study correlates DISE findings with polysomnographic severity. Objective To determine the association between DISE‐based obstruction patterns and OSA severity in children. Methods This prospective study included 95 children with adenotonsillar hypertrophy. All underwent preoperative Level III polysomnography followed by DISE before AT. Two blinded otolaryngologists independently reviewed DISE recordings, classifying obstruction via the VOTE system (degree 0–2; pattern: concentric, lateral, or anteroposterior). Statistical analysis employed non‐parametric tests, relative risk calculation, and ROC curves. Results Severe OSA (AHI > 10) was significantly associated with Brodsky grade IV tonsils (40.7% vs. 16.7%) and severe adenoid hypertrophy (44.1% vs. 16.7%). DISE identified severe obstruction at the velopharynx ( p = 0.003) and tongue base ( p = 0.036) as specific predictors of severe OSA, with high specificities (91% and 97.1%) and positive predictive values (87.7% and 90.9%). Concentric collapse was the most frequent velopharyngeal pattern (54.2%). Obesity was significantly linked to severe tongue base obstruction (38.9% vs. 5.4%, p = 0.001) and higher AHI. ROC analysis for tongue base obstruction combined with the Obstructive Apnea Index (OAI) showed high diagnostic accuracy (AUC = 0.811). Discussion and Conclusion DISE is a valuable tool for predicting pediatric OSA severity. Obstruction at the velopharynx and tongue base, particularly concentric collapse and in the context of obesity, are key risk factors. Integrating DISE findings with PSG enhances preoperative risk stratification, aiding in the selection of candidates who may benefit from personalized, multilevel surgical interventions.