Abstract
Background
Studies investigating the relationship between sleep-disordered breathing (SDB) and atrial fibrillation (AF) have largely assessed SDB-severity by the apnea–hypopnea index (AHI). However, the AHI does not incorporate nocturnal hypoxemic burden, which may increase the risk of non-paroxysmal AF (nPAF) as the clinical manifestation of more progressed AF substrates. This investigation sought to systematically characterize and compare the composition of AHI and hypoxemic burden with the aim to defining a disease-orientated metric for SDB-severity best associated with prevalent nPAF.
Methods
Polysomnography including overnight oximetry data were obtained in 435 consecutive ambulatory AF patients to determine the composition of AHI (apneas vs. hypopneas), the number of acute episodic desaturations per hour (oxygen desaturation index, ODI) and the composition of total time spent below 90% oxygen saturation (T90Total) attributed to acute desaturations (T90Desaturation). Logistic regression analysis was used to characterize the association with prevalent nPAF.
Results
One hundred sixty-nine AF patients (38%) had nPAF and one third (n=149, 34%) had moderate-to-severe SDB (AHI>15). 82% of the median total AHI (9.4 [3.6–20.1]) could be attributed to hypopneas. Only 29% of events were associated with episodic desaturations, which contributed to 96% (T90Desaturation) of the variation in T90Total. The high variability in durations and nadirs of distinct desaturation events can expose patients to long T90Total, even if the AHI is low. Not AHI, but T90Total and ODI were associated with nPAF independent of gender and age. However, diabetes, hypertension and body mass index contributed more significantly to the overall risk of nPAF.
Conclusions
In AF patients, hypopneas constitute a majority of respiratory events during sleep. Patients with low AHI can still be exposed to high nocturnal hypoxemic burden, which is mainly a cumulative consequence of episodic desaturations. T90Total and ODI, but not AHI, were associated with nPAF independent of gender and age, but concomitant modifiable risk factors made a more significant contribution to the overall risk of nPAF versus PAF.
Machine learning for nocturnal mass diagnosis of atrial fibrillation in a population at risk of sleep-disordered breathing