RESUMO
PURPOSE: It is unclear if there is a consistent signature in breath patterns prior to an impending obstructive apneic event in patients with sleep-disordered breathing (SDB). OBJECTIVE: To use continuous recordings of ambient sound in sleep using a smartphone to track auditory signatures of breaths and measure their regularity preceding apneic events. METHODS: We studied 50 patients evaluated for SDB in whom sound was recorded using smartphones concurrently with polysomnography (PSG). Whole-night sound files were analyzed for time and frequency domain analyses of breath periodicity during periods of normal and sleep-disordered breathing. RESULTS: Fifty patients (44% women, 42.0 ± 9.4 years old, BMI 32.8 ± 10.8 kg/m2) recorded sound, of whom 30 were diagnosed with OSA and 20 were not. We analyzed a total of 497 apneic (≥10 s) and 481 non-apneic intervals, confirmed by PSG. Interbreath intervals were 3.75 ± 0.62 s for 1 min in quiet breathing, with SD 1.11 ± 0.48 s that increased to 4.16 ± 3.06 s in successive 60-s epochs up to apnea (p < 0.001). Interbreath SD in the 60 s immediately preceding apnea was higher than the SD in random non-apneic periods (p < 0.01, ANOVA). Interbreath SD ≥1.49 s gave 87.3% sensitivity and 86.5% specificity for predicting apnea in the next minute (c-statistic 0.94). CONCLUSIONS: Breaths increase in variability minutes before proven obstructive apnea in patients with suspected SDB. These results suggest that it may be possible to predict and thus potentially avert apneic events and provide insights into events leading to SDB. TRIAL REGISTRATION: NCT03288376, clinicaltrials.org.