Abstract Rationale Determining the site of airway collapse is crucial for effective OSA treatment but currently relies on drug-induced sleep endoscopy (DISE). Non-invasive alternatives are needed to aid treatment planning. Because snoring (routinely measured in sleep studies) reflects upper-airway vibrations that may vary by collapse site, we hypothesized that snoring metrics could identify the site of obstruction. Methods First, we analyzed snoring signals from a high-quality microphone collected during DISE in 492 patients with OSA (Boston). Raw snoring signals were used to derive sound intensity (dB) referenced to a background level. For each patient, the snoring frequency was calculated as the percentage of breaths with loudness exceeding 20 dB above the background level. Patients with complete collapse were compared to patients with no collapse at the given site of collapse (partial and multilevel collapse excluded). Logistic regression calculated odds of collapse at each site per SD change in snoring frequency. Second, validation was performed on a clinical cohort (Antwerp) of 184 OSA patients who had recently completed a DISE study and an overnight polysomnography. Results Analysis of patients with simultaneous snoring and DISE (Boston: 70% male, 57 48–64 years, 29 26.5–31.7 kg/m², median 25th–75th percentile revealed that greater snoring frequency (+1SD) was associated with greater odds of lateral wall collapse (OR 95%CI 1.80 1.28–2.52, p = 0.0007) and the absence of tongue-base collapse (OR 1.44 1.14–1.81, p = 0.002), but not significantly with complete concentric collapse of the palate (1.75 0.93–3.31) or absence of epiglottic collapse (0.80 0.53–1.20). In the clinical cohort (Antwerp: 81.5% male, 51 41,59 years, 27.5 25.2–30.3 kg/m², median 25th -75th percentile ), snoring frequency was strongly associated with greater odds of lateral wall collapse: OR 3.55 1.78–7.10 (p = 0.0005) and absence of tongue-base collapse: OR 1.75 1.10–2.85 (p = 0.022). The association between snoring and complete concentric collapse (OR 1.41 0.88–2.26) was not significant, nor was the association with absence of epiglottic collapse (OR 1.64 0.89–2.99). After adjusting for flow-shape–derived prediction of lateral-wall collapse and clinical factors (AHI, age, sex, BMI), snore metrics remained independently associated with lateral-wall collapse (OR 2.70 (95% CI 1.16–6.31, p = 0.022), indicating incremental value beyond these covariates. Conclusion Our study suggests that snoring frequency (defined as the percentage of breaths exceeding a decibel threshold) may help non-invasively identify patients who are likely to exhibit lateral wall collapse and the absence of tongue-base collapse. This abstract is funded by: American Heart Association (salary support to 1st author)
Monteiro et al. (2026) studied this question.