Baseline DISE Anatomy Predicts Jaw-Thrust Responsiveness in Obstructive Sleep Apnea

Drug-induced sleep endoscopy (DISE) with a jaw-thrust maneuver is used to simulate mandibular advancement in obstructive sleep apnea (OSA), yet determinants of functional airway improvement remain incompletely defined.

To identify clinical, polysomnographic, and baseline DISE anatomic factors associated with jaw-thrust responsiveness.

We conducted a single-center retrospective observational study of adults with polysomnography-confirmed OSA who underwent DISE with paired baseline and jaw-thrust VOTE assessments between 1 January 2015 and 31 December 2025 (n = 355). Jaw-thrust responsiveness was defined a priori as a within-subject reduction in the number of obstructed VOTE sites (grade ≥ 1). Multivariable logistic regression was used to identify independent correlates within a prespecified explanatory modeling framework. The study was approved by the Institutional Review Board of Taipei Tzu Chi Hospital (protocol 14-IRB079), with the need for informed consent being waived.

Jaw thrust reduced overall obstruction burden from two (two to three) to one (one to two) sites (Wilcoxon p < 0.001). Hypopharyngeal levels demonstrated the greatest improvement, particularly at the tongue base (39.2% to 7.6%) and epiglottis (23.9% to 5.4%) (both p < 0.001). Overall, 62.8% met responder criteria and 18.9% achieved complete normalization. In multivariable analysis (n = 272), baseline tongue-base collapse (adjusted odds ratio [aOR] 2.46, 95% CI 1.20-5.04) and greater baseline multilevel obstruction burden (aOR 1.85 per SD, 95% CI 1.19-2.85) were independently associated with responsiveness, whereas conventional PSG severity metrics were not.

In adults with OSA, jaw-thrust responsiveness during DISE is more strongly associated with baseline anatomic phenotype than with global PSG severity. Standardized DISE functional assessment may provide complementary information to support phenotype-informed selection of non-CPAP therapies, pending prospective validation.