B20-02 A Novel Method to Quantify the Arousal Threshold Using Survival Analysis in Obstructive Sleep Apnea

Abstract Rationale The role of a low arousal threshold (ArTh) in OSA pathophysiology has remained controversial, partly because lower arousal thresholds are not associated with greater OSA severity using current analytical approaches1. Since ArTh is measured only during events, it overlooks stable breathing when the threshold is not reached and therefore not observed, skewing estimates toward lower values even though it may be higher during those times. We introduce a whole-night arousal threshold (ArThnight) that incorporates stable breathing and hypothesize that it will inversely correlate with AHI and vary systematically across sleep stages. Results from the conventional event-based approach (ArThevent) are presented in parallel for comparison. Methods Ninety-two participants with suspected or diagnosed OSA underwent overnight polysomnography with airflow (pneumotachometer via oronasal mask) and ventilatory drive (intraesophageal diaphragm EMG) measurements. Arousals were scored by a single technician using AASM criteria. Breath-level data included ventilatory drive (normalized to local wake %eupnea) and a binary indicator of arousal on the next breath. To estimate ArTh, we applied a survival framework treating ventilatory drive as the “time” variable, where each breath contributed an observation at its corresponding drive level; arousals were considered events, and non-arousals were right-censored. Within 10-min windows, Kaplan-Meier survivor functions were computed and then interpolated to a common drive grid and averaged to yield a subject-level survival curve. The ArThnight was defined as the drive at 50% arousal probability. For comparison, the event-based ArTh was calculated using the conventional approach, defined as the mean of the drives on breaths immediately preceding arousals terminating respiratory events1. Results ArThnight was higher than the event-based measure (median IQR: 206.7 139.3, 313.9 vs 164.5 133.2 to 194.9 %eupnea) and was significantly associated with lower AHI, decreasing by ‒11.8 (95%CI: ‒17.1, ‒6.5, p = 3 × 10-5) events/h per SD increase (Figure 1A). The ArThevent was not significantly related to AHI and trended toward higher AHI with increasing arousal threshold (+3.6 ‒1.9, 9.1 events/h per SD increase; p = 0.2; Figure 1C). Across sleep stages, relative to N2, ArThnight was lower in N1 and REM and higher in N3, as expected (Figure 1B). ArThevent showed a similar stage-dependent pattern but did not exhibit a lower level in N1 (Figure 1D). Conclusion The whole-night ArTh better explains OSA severity than the conventional event-based method, suggesting it provides a more reliable characterization of an individual’s propensity to awaken from respiratory stimuli and its contribution to OSA pathophysiology. This abstract is funded by: R01HL168067