Nocturnal Hypoxemia Burden Correlates With Vestibular Dysfunction in Obstructive Sleep Apnea

Obstructive sleep apnea (OSA) is a potential contributor to vestibular dysfunction, though specific correlating parameters remain unclear. This study aimed to determine whether hypoxemia severity indices correlate quantitatively with caloric slow-phase eye velocity (SPV), a marker of vestibular function.

In a retrospective cohort of 82 adults undergoing both full-night PSG and bithermal caloric testing, the primary outcome was total caloric SPV (°/s; sum of four irrigations). Fifteen prespecified predictors were evaluated, including hypoxemia burden (apnea-hypopnea index [AHI], oxygen desaturation index [ODI], cumulative time with SpO₂ < 90% [CT90], nadir SpO₂), sleep fragmentation (total arousal index), sleep architecture (N1, N2, N3, rapid eye movement [REM] proportions and sleep efficiency) and ancillary measures (snoring index, pulse rate, age, sex and body mass index [BMI]). Unadjusted associations were calculated using Pearson's correlations. For adjusted inference, ordinary least squares with backward Akaike information criterion (AIC) reduction were employed. Feature importance was quantified without prespecifying functional form by training a random-forest regressor and computing TreeSHAP values.

SPV was inversely correlated with AHI (r = -0.295, p = 0.007), ODI (r = -0.261, p = 0.018) and CT90 (r = -0.254, p = 0.021). Sleep efficiency, snoring index, pulse rate, nadir SpO₂, and sleep stage proportions showed no significant associations. In the adjusted linear model, AHI and sex were retained; each 10-event/h AHI increase corresponded to a 3.16°/s lower SPV (95% CI: -6.05 to -0.28; p = 0.032). The random forest model achieved a pooled fivefold R² of 0.065, indicating low explained variance. Regarding mean absolute SHAP values, hypoxemia-domain metrics ranked highest (ODI, AHI, CT90), followed by BMI, while other features had minimal impact.

Nocturnal hypoxemia burden-reflecting the frequency and cumulative duration of oxygen desaturation-emerged as the principal predictor of reduced caloric SPV in patients with OSA. While the low explained variance of our models (R² = 0.065) suggests these associations should be viewed as hypothesis-generating, the findings support prioritising desaturation assessment in clinical settings. Further longitudinal research is essential to verify whether reducing hypoxemia leads to improved vestibular outcomes.