Sleep-disordered breathing: how should we judge its severity?

This editorial refers to ‘Nocturnal hypoxaemia is associated with increased mortality in stable heart failure patients’†, by O. Oldenburg et al., on page 1695. Oldenburg and colleagues, in this issue of the journal, confirm that abnormalities of breathing during sleep are very common in patients with heart failure and reduced ejection fraction (HFREF).1 In a cohort of 963 patients with chronic stable HFREF, they report that 58% had moderate or severe sleep-disordered breathing (SDB), as defined by an apnoea–hypopnea index (AHI) of at least 15 episodes per hour of recording time using unattended in-hospital polygraphy. Over a median follow-up of almost 7.5 years, half of the cohort died, and those with moderate to severe SDB (AHI ≥15) had a higher mortality. After correction for baseline covariates, this association was no longer present. However, another measure of SDB, i.e. the time the patients spent with hypoxaemia (arterial oxygen saturation <90%), was highly significantly associated with mortality, even after adjustment for baseline covariates, with an increase in all-cause mortality of 16% [95% confidence interval (CI) 9–24%] per hour spent below 90% oxygen saturation during sleep (P < 0.001). They argue that nocturnal hypoxaemic ‘burden’ may be an important target for the treatment of SDB, and that AHI, the classical descriptor of the severity of SDB, may be less useful for this patient group. Hypoxaemia of ≥22 min duration appeared to be the best discriminator between those who died subsequently and those who did not. That breathing can be disordered during sleep is undisputed. The most common form in the general population, and in those with cardiac disease, is obstructive sleep apnoea (OSA), where there is periodic reduction or absence of airflow during sleep, due to at least partial collapse of the upper airway with ongoing respiratory effort. These episodes can be associated with arousal from sleep and daytime somnolence. International guidelines strongly support the symptomatic treatment of such individuals, typically with positive airway pressure.2 A similar approach is taken to classify the other major form of SDB, central sleep apnoea (CSA), where there is lack of respiratory effort leading to episodes of apnoea or hypopnoea. This phenotype is generally found in patients with heart failure, although it can be idiopathic or associated with narcotic usage. Even if there is a high AHI in heart failure, patients may not have daytime somnolence,3 perhaps because of a high background level of sympathetic activation. Despite less marked swings in intrathoracic pressure, the associations with surges in sympathetic activity, and hypoxaemia due to airflow cessation or reduction are similar to those found in OSA. It is not generally appreciated that scoring systems for the complex phenomenon of SDB are largely consensus based, and have been relatively fluid in recent years, responding to an increasing evidence base regarding the association of abnormalities with symptoms and cardiovascular outcomes, changes in sensor technologies and diagnostic approaches, and changes in reimbursement.4 The AHI averages the number of apnoea and hypopnoea events occurring during a diagnostic study over the total number of hours of sleep time (polysomnography) or recording time (polygraphy). There is consensus about the nature of an apnoea—the absence or near absence of airflow—with the current definition being a reduction in airflow by at least 90% of pre-event baseline for at least 10 s. There has been much more controversy regarding hypopnoea—what severity and duration of reduction in airflow should be classified as abnormal? Originally, the term was introduced to cover the situation where there was an associated drop in arterial oxygen concentration (or an arousal from sleep). Initially, various studies reported that a 4% desaturation was associated with an increased cardiovascular risk in OSA, but more recently it would appear that even a 3% desaturation is associated with increased risk. Currently the recommendation for classification of hypopnoea (in adults) is that airflow reduces by at least 30% of pre-event baseline for at least 10 s in association with either a 3% or greater arterial oxygen desaturation, or an arousal. The classification of respiratory events into ‘obstructive’ or ‘central’ is also generally considered important4 as this reflects different underlying pathophysiological processes and may influence the type of positive airway pressure (PAP) therapy applied. Although the basic differences in mechanism appear clear (respiratory effort being present in OSA but not in CSA) (Figure 1), in practice it can be very difficult to classify hypopnoeas, and some laboratories do not attempt this, providing an overall AHI only and a qualitative assessment as to whether it is predominantly one form or the other. Also, some events may start as central but become obstructive, and the pattern may change during the night. (A) Obstructive sleep apnoea on sleep polygraphy, with thoracoabdominal movement during apnoea. (B) Central Sleep apnoea, with absence of thoraco-abdominal movement during apnoea. In both cases, there are swings in arterial oxygen saturation. One of the major therapeutic objectives of diagnosing SDB is to reduce cardiovascular risk, so a better description of the association of the many possible metrics with cardiovascular events or mortality is vital. Oldenburg et al. suggest that, in HFREF, merely averaging the number of apnoeas and hypopnoeas per hour across the recording time of a diagnostic sleep study may not be the most useful approach.1 AHI does not consider the duration of episodes and does not differentiate between patients who have a number of short episodes of apnoea or hypopnoea from those who have the same number of much longer episodes associated with more marked oxygen desaturation. In heart failure, respiratory cycle lengths become more prolonged as cardiac function reduces, and this can limit the total AHI.5 They show that even after correction of measured baseline covariates, the duration of hypoxaemia in one sleep study is strongly associated with subsequent mortality. Perhaps hypoxaemia is the abnormality that stresses the heart and cardiovascular system, rather than merely the average hourly count of episodes when the airflow is reduced by an arbitrary amount. This would suggest that other simpler screening tests, such as overnight oximetry,6 can clearly identify which patients have SDB to the extent that it may have cardiovascular consequences—and require further study. A bigger question remains. Once the pathophysiology has been identified, what treatment (if any) should be applied? As Oldenburg and colleagues discuss, there are several studies reporting a clear association between the presence of SDB and outcome in patients with HFREF. There are also observational studies that report better outcome if such patients are treated with PAP. Small randomized trials with surrogate endpoints suggest benefit of treating the SDB. Post-hoc analysis of the first larger randomized trial, the CANPAP study, where 258 HFREF patients with predominantly CSA were treated with CPAP therapy, suggested that although the parent study results were neutral on transplant-free survival,7 patients who had their CSA well controlled (AHI <15/h) had a low rate of cardiovascular events.8 However, the much larger SERVE-HF Study was published recently.9 A total of 1325 HFREF patients with predominantly CSA and an AHI ≥15/h were randomized to optimal medical therapy or such therapy plus a specialized from of PAP therapy (ASV). After a median of 31 months follow-up, there was no difference in the primary combined endpoint of all-cause mortality, life-saving cardiovascular intervention, or unplanned hospitalization for worsening heart failure, but an unexpected 34% (9–65%) increase in cardiovascular mortality (P = 0.006). This effect was particularly marked in those with a higher proportion of Cheyne–Stokes respiration (CSR), and those with a lower ejection fraction. The increased mortality appeared to have been largely due to sudden, presumably cardiac, deaths. Various explanations have been proposed: chance, a direct toxic effect of PAP, or that CSR may not be pathophysiological but (at least partially) adaptive.9–11 Oldenburg and colleagues raise the interesting suggestion that although the ASV therapy did effectively suppress the AHI (from 31 to 7/h), and the oxygen desaturation index (from 32 to 9 events/h) the overall time with saturation <90% was only reduced from a mean of 51 min to 19 min, so many patients still had a meaningful hypoxaemic burden on therapy. Further outcome data will come from the ongoing ADVENT randomized study of ASV in patients with HFREF and either OSA or CSA (NCT 01128816). Breathing during sleep is often abnormal, particularly for those with HFREF. How best to describe this phenomenon is uncertain. Those with SDB will do worse than those whose breathing is normal, but when and how to intervene remains unclear. A clear therapeutic objective is paramount. In the absence of daytime somnolence, we need further outcome studies that show that current treatments can improve cardiovascular morbidity and mortality. Oldenburg's work suggests that rather than mere rate the episodes of reduced or absent airflow, we should consider the hypoxaemic burden on the patient: only if we can meaningfully reduce this can we expect to find benefit. This hypothesis requires prospective validation. Conflict of interest: M.R.C. was the co-principal investigator of the SERVE-HF Study and has received research funding from ResMed, the sponsor of that study. He has also received consultancy fees related to sleep-disordered breathing from ResMed, Respicardia, and Sorin.