Abstract
OBJECTIVE: Patients undergo a drug-induced sleep endoscopy before unilateral hypoglossal nerve stimulation. This study aims to reduce the need for preoperative drug-induced sleep endoscopy by predicting favorable collapse patterns at the velum based on clinical examination and sleep study data. STUDY DESIGN: A retrospective cohort study. SETTING: A single-center trial. METHODS: Drug-induced sleep endoscopy recordings were scored by a blinded rater, and the palatal collapse pattern was classified as favorable (no or anterior-posterior) or unfavorable (lateral or concentric), irrespective of collapse degree. Logistic regression models were constructed using clinical examination and sleep data to predict palatal collapse patterns. Model performance was assessed using receiver-operating characteristic curves and validated internally using bootstrapping and externally using a separate cohort. RESULTS: Recordings of 250 patients were analyzed, of which 118 showed a favorable palatal collapse pattern. Lower body mass index (BMI), lower tonsil grade, and less intense bed-partner-reported snoring were significant predictors of favorable collapse patterns (area under the curve [AUC] 0.70). A further reduction in tonsil grade and BMI resulted in similar model performance (AUC 0.68). In patients with no tonsils and BMI < 28 kg/m(2) or tonsil grade 1 and BMI < 24 kg/m(2), a favorable collapse was correctly predicted in 81%. In external validation, favorable collapse was correctly identified in 91%. CONCLUSION: Favorable palatal collapse can accurately be predicted in patients with small or absent tonsils and low BMI. Before unilateral hypoglossal nerve stimulation implantation, a drug-induced sleep endoscopy might not be necessary in this subset of patients.