Ineffective inspiratory efforts are a hallmark of obstructive apneas. The resulting exaggerated negative intrathoracic pressure swings increase left ventricular (LV) transmural pressure by increasing the difference between extracardiac and intracardiac pressures, and hence afterload, but without increasing BP (25-27). It also increases venous return to the right ventricle, leading to its distension. The resulting leftward shift of the interventricular septum can impede LV diastolic filling (28). There is also evidence that exaggerated negative intrathoracic pressure during apnea can impair LV relaxation, which could further impede LV filling (29). The combination of increased LV afterload and reduced LV preload leads to a reduction in stroke volume during obstructive apneas that is proportional to the negative intrathoracic pressure generated (1, 30-33). However, it appears that exaggerated negative intrathoracic pressure does not acutely impair LV contractility when underlying LV function is normal (34). Following release of obstructive apneas, stroke volume abruptly increases (1, 33, 34).

Vagal afferent feedback from lung inflation inhibits sympathetic outflow, whereas apnea disinhibits it (35). Thus, during a breath hold, sympathetic vasoconstrictor outflow to muscle (MSNA) increases progressively from its onset until its termination owing to progressive hypoxic stimulation (36). However, during the initial phase of an obstructive apnea or Mueller maneuver, MSNA is suppressed (36, 37). This is due to the effects of obstructive apneas on baroreceptor activity. Negative intrathoracic pressure causes transmural intrathoracic aortic pressure to increase, which activates aortic baroreceptors and inhibits sympathetic outflow (36, 37). On the other hand, there is a simultaneous fall in BP due to the reduction in stroke volume, which suppresses carotid sinus baroreceptor activity and tends to reflexively augment sympathetic outflow. Because the influence of the aortic baroreceptors predominates, the net effect is suppression of MSNA. Toward the end of these events, however, MSNA rises in response to hypoxia. Depending on the strength of the hypoxic stimulus and the sympathetic vasoconstrictor response, BP can, but does not invariably increase toward the end of apneas (24, 26, 34, 38).

During obstructive apneas, the sympathoexcitatory effect of hypoxia is amplified by apnea and CO₂ retention. This results in increased sympathetic vasoconstrictor tone (39). However, as discussed above, these sympathoexcitatory effects are not engaged until several seconds into the apnea. Owing to the circulatory delay between the lung and the peripheral chemoreceptors, the sensing of nadir Sa_O2 that occurs in the lungs at the end of apnea is not detected at the carotid bodies until several seconds later (40). As a result, the maximum vasoconstrictor and chronotropic effects of apnea-related hypoxia occur during the postapneic ventilatory phase and are associated with surges in BP and HR (41) (Figure 1). These effects increase the metabolic demands of the myocardium in the face of reduced O₂ supply. Moreover, intermittent hypoxia during obstructive apneas may directly depress cardiac contractility (42), or reduce cardiac performance indirectly by causing pulmonary vasoconstriction and increasing pulmonary arterial pressure (33). The degree of desaturation during each obstructive apnea has been directly related to the magnitude of the increase in BP following the apnea (8, 43). Although these effects can be partly inhibited by administration of O₂ during voluntary Mueller maneuvers (36), supplemental O₂ has little effect on BP surges following apneas in patients with OSA (24, 44). This observation indicates that factors other than hypoxia, such as hypercapnia and arousals from sleep, must also be contributing to surges in postapneic BP (16, 18, 39).

Fig. 1. Repetitive elevations in muscle sympathetic nerve activity and blood pressure associated with recurrent obstructive apneas and hypoxia. From Hedner and coworkers (53), with permission.

Download Figure | Download Powerpoint

[More] [Minimize]

Hypoxia has varying influences on HR according to the presence or absence of airflow, and the balance of its parasympathetic and sympathetic stimulatory effects. In the absence of airflow, hypoxic stimulation of the carotid body is vagotonic, and causes bradycardia (45, 46). Conversely, in the presence of airflow (e.g., hypoxic rebreathing), hypoxia causes tachycardia because stretching of the lungs inhibits vagal outflow to the heart and permits unopposed cardiac sympathetic discharge. Nevertheless, HR responses to obstructive apneas vary greatly among individuals. This variability in HR responses is probably due to differences in severity of hypoxia, intrinsic hypoxic chemosensitivity, and the relative influence of hypoxia on vagal and sympathetic input to the sinoatrial node (47, 48). Accordingly, where parasympathetic influence predominates, HR may slow, where sympathetic influence predominates, HR may rise, and where vagal and sympathetic influences are relatively equal, HR may remain unchanged (49). However, upon resumption of airflow at termination of apnea, HR invariably rises owing to disengagement of hypoxia-mediated cardiac vagal outflow and unopposed cardiac sympathetic discharge (16, 39, 50). When bradycardia occurs during obstructive apneas, administration of supplemental O₂ or atropine attenuates it (51).

Arousal is a critical defense mechanism that activates upper airway dilator muscles and prevents asphyxiation in OSA (21). However, it also contributes to the abrupt surges in HR and BP following termination of apnea (16), but the degree to which it does so remains controversial. O'Donnell and colleagues (26) induced obstructive apneas of similar length in dogs with and without arousal. Apneas terminated prior to arousal caused increases in BP, but termination by an arousal caused a further increase. Similarly, Trinder and coworkers (18) described augmentation in HR and BP during the ventilatory phase of Cheyne–Stokes respiration, both awake and during sleep in the absence of arousals. Arousals caused only a small further increment in HR and BP. One confounding factor is that arousals are accompanied by an abrupt increase in ventilation that precedes increases in HR and BP. This suggests that the increased ventilatory drive at termination of apnea coactivates cardiovascular sympathetic neurons that are closely linked with respiratory neurons in the brainstem (13). Indeed, voluntary periodic breathing during wakefulness causes postapneic surges in HR and BP even in the absence of hypoxia or arousals from sleep (52). Matters are further complicated by the observation that sudden lung inflation at termination of apnea counteracts asphyxia-induced sympathetic activation causing an abrupt decrease in MSNA, which plays a role in the fall in BP prior to the onset of the next apnea (41). Taken together, these observations indicate that although arousals from sleep can contribute, they are not critical to the development of postapneic surges in HR and BP.

OSA is associated with chronic abnormalities of cardiovascular autonomic regulation both during sleep and wakefulness. These are characterized by increased SNA, reduced baroreflex sensitivity and HR variability, and increased BP variability (41, 53-56). A number of studies have shown that patients with OSA have higher SNA during sleep and wakefulness than control subjects (41, 53, 54) (Figure 2). Treatment of OSA either by tracheostomy (57) or continuous positive airway pressure (CPAP) leads to a reduction in nocturnal and daytime SNA. The latter effect appears to require several months of CPAP therapy (58-60). This beneficial effect is presumably related to elimination of OSA, nocturnal hypoxia, and arousals from sleep.

Fig. 2. Increased muscle sympathetic nerve activity during wakefulness in patients with obstructive sleep apnea (OSA) compared with normal subjects. From Somers and coworkers (41), with permission.

Download Figure | Download Powerpoint

[More] [Minimize]

The mechanisms by which OSA leads to persistent sympathetic activation are incompletely understood. There are several lines of evidence favoring an important role for hypoxia in this phenomenon. Urinary norepinephrine concentrations are inversely proportional to minimal nocturnal Sa_O2 in patients with OSA (61). In addition, both short-term sustained and intermittent challenges with combined hypoxia and hypercapnia cause elevations in MSNA that persist for at least 20 min following withdrawal of the exposure (62, 63). This prolonged sympathetic effect appears to be due to hypoxia rather than hypercapnia (64).

The carotid bodies may be an important intermediate step in the pathway between exposure to intermittent hypoxia and development of sustained elevations in SNA. For example, it has been reported that patients with OSA have increased peripheral chemoreflex sensitivity (65) and pressor responses to hypoxia (66). A mechanism for this is suggested by the observation that daily short-term exposure to hypoxia increases peripheral hypoxic chemosensitivity (67). However, other investigators have reported either normal (68) or depressed hypoxic ventilatory responses in patients and animals with OSA (69, 70). Nevertheless, carotid body denervation prevents the development of hypertension in rats exposed to intermittent hypoxia (71), and desensitization of the peripheral chemoreceptors by administration of 100% O₂ in awake normoxic patients with OSA reduces MSNA, HR, and BP (72). These findings suggest that the peripheral chemoreceptors stimulate cardiovascular sympathetic outflow in patients with OSA even when they are normoxic and awake.

Under normal conditions, activation of the carotid sinus and aortic arch baroreceptors by an increase in BP reflexively inhibits sympathetic outflow, increases cardiac vagal outflow, and reduces HR. A decrease in BP does the opposite. Therefore, depression of baroreceptor sensitivity could contribute to sympathetic activation and parasympathetic withdrawal (73). In patients with OSA, repetitive nocturnal surges in BP may downregulate baroreceptors and blunt their sensitivity. However, studies of baroreflex sensitivity in patients with OSA have yielded inconsistent results; some investigators report that it is depressed (73, 74), and others report that it is normal (75). Brooks and coworkers (76) found that in concert with an increase in BP, induction of OSA in dogs caused a parallel shift of the baroreflex curve to the right, indicating an increase in the set point to a higher BP, but without any change in sensitivity. Thus, the baroreflex appeared no longer capable of counteracting the higher BP, and may have been contributing to it by indirectly maintaining a higher sympathetic outflow (77). However, because dogs are quadrupeds that do not undergo as extreme postural changes as humans, it is unclear what implications these findings have for patients with OSA. On the other hand, Tkacova and colleagues (78) demonstrated that acute elimination of OSA by CPAP in patients with congestive heart failure (CHF) caused both an immediate increase in baroreflex sensitivity and a decrease in the set point in association with a reduction in BP. These findings have two important implications. First, they suggest that OSA may both increase the set point and depress the sensitivity of the baroreflex. Second, such abnormalities are to some extent reversible.

Cardiovascular autonomic function has also been assessed in patients with OSA by examination of daytime heart rate variability using power spectral analysis. Normally, HR varies at high frequency (HF) as a function of respiration. This respiratory sinus arrhythmia is modulated primarily by cardiac vagal activity. Heart rate variability at lower frequencies is thought to be modulated by sympathetic activity, although there is controversy on this point (79). In general, patients with OSA have a decrease in the HF component of HR variability and an increase in the low-frequency component (55, 80, 81). This abnormal pattern is thought to reflect decreased parasympathetic and increased sympathetic modulation of HR. Treatment of OSA with CPAP has been shown to restore these indices toward normal, both acutely (82) and chronically (83).

In addition to catecholamines, levels of other circulating hormones involved in the regulation of BP and fluid volume, such as renin, aldosterone, and vasopressin, have been studied in relation to OSA. All of these hormones have pressor or fluid- and sodium-retaining effects that would predispose to hypertension. However, there is no clear indication what, if any, effect OSA has on the levels of these hormones (84-86).

Results have been more consistent in the case of atrial natriuretic peptide, a vasoactive hormone secreted primarily in response to right atrial distension (87). In patients with OSA, its concentration is elevated in proportion to the degree of hypoxemia-induced increases in pulmonary artery pressure and negative intrathoracic pressure swings (85, 88-90). Atrial natriuretic peptide promotes diuresis, natriuresis, and vasodilation and thereby counteracts the pressor and fluid-retaining effects of the hormones mentioned above (91). Consequently, high nocturnal atrial natriuretic peptide levels probably contribute to nocturia, a common feature of OSA. In uncontrolled studies, abolition of OSA by CPAP was associated with reductions in nocturnal urinary atrial natriuretic peptide and urine excretion (90, 91).

Vascular and endothelial effects, and atherosclerosis . . Abnormalities of vascular responsiveness could contribute to chronic elevations in BP in patients with OSA. However, experiments on vascular responses to hypoxia, and α- and β-adrenergic stimulation in OSA have yielded inconsistent results (66, 68, 92). Studies of the potential role of the potent endogenous vasoconstrictor, endothelin-1, in the pathogenesis of systemic hypertension in patients with OSA have also been inconclusive (93, 94). Similarly, studies of responses to endogenous vasodilators in patients with OSA have yielded conflicting results: some investigators found depressed circulating nitric oxide levels and a selective defect in endothelium-dependent vascular relaxation (95, 96), whereas others did not (97). Taken together, the above findings do not provide persuasive evidence of either a primary vascular lesion or abnormal vascular responsiveness to hypoxia or to endothelially dependent vasoconstrictors or dilators.

It has been hypothesized that intermittent hypoxia with abrupt decreases and increases in cardiac output in association with OSA could provoke elaboration of oxygen free radicals and ischemia–reperfusion injury to the vascular wall (98). Such a process could precipitate or accelerate the formation of atherosclerotic plaques and vascular smooth muscle proliferation (99, 100). A recent study demonstrated increased superoxide anion production in patients with OSA that was markedly diminished following CPAP therapy (101). However, there is no direct evidence that OSA can precipitate or accelerate atherosclerosis.

Platelet aggregability is enhanced by catecholamines and is associated with an increased risk of cardiovascular events (102, 103). Shortly after arising in the morning, there is a surge in plasma catecholamines and a simultaneous increase in platelet aggregability to peak levels, which corresponds to the peak incidence of cardiovascular and cerebrovascular events (103-107). The hypothesis that OSA might predispose to such ischemic events is plausible because the longest REM episode with the most severe apneas generally occurs just before awakening in the morning. Moreover, in contrast to healthy subjects, in patients with OSA, platelet aggregability increases significantly overnight in association with elevated nocturnal catecholamine levels (108-111). Abolition of OSA by CPAP has been reported to reduce platelet aggregability in association with reductions in overnight catecholamine levels (109, 110, 112).

OSA is also associated with laboratory evidence of a predisposition to clot formation. For example, hematocrit is increased, probably due to nocturnal hypoxemia (113). Overnight and daytime fibrinogen levels (114), as well as whole blood viscosity (115), are also elevated. The mechanisms for these associations remain to be determined. Nevertheless, the observations that CPAP therapy can alleviate some of these abnormalities and can reduce Factor VII clotting activity (116) suggest that OSA is contributing to them. However, definitive evidence that OSA predisposes to clot formation awaits further research.

Potential adverse cardiovascular physiological effects of sleep apnea are listed in Table 1.

Hypertension, which affects 20% of the adult population, is one of the commonest diseases in North America (118). It is a major risk factor for the development of coronary artery disease, CHF, and strokes. Moreover, effective therapy of hypertension reduces the risk of developing these disorders. However, in only 5–10% of cases of hypertension is an underlying secondary cause identified. It is in this context that the potential importance of OSA as a secondary and possibly treatable cause of hypertension must be viewed.

Methodologically rigorous population-based studies have yielded convincing evidence in favor of a modest, but definite association between OSA and systemic hypertension, independent of age, obesity, or other confounding factors (120-122). The Sleep Heart Health study employed in-home polysomnography (120) in a cross-sectional analysis of 6,132 subjects. A clear independent association between OSA and hypertension was observed in which the prevalence of hypertension increased with increasing AHI. However, the association was quite modest with an adjusted odds ratio for the most severe OSA category, AHI > 30 (versus AHI < 1.5) of only 1.37 (CI 1.03–1.83). Similarly, in a cross-sectional analysis of 1,069 subjects who underwent in-laboratory polysomnography in the Wisconsin Sleep Cohort Study (121) a significant linear increase in daytime BP with increasing AHI was observed. The association between OSA and hypertension was stronger in this study with an adjusted odds ratio for hypertension associated with AHI > 30 (versus < 1) of 3.1 (CI 1.7–5.7). Furthermore, results have recently been reported from prospective follow-up of 893 of those subjects over 4 to 8 yr (122). The odds ratio for the new onset of hypertension at follow-up assessments, associated with the presence of OSA at baseline, was 2.89 (CI 1.46–5.64 for an AHI > 15 versus 0). These are the strongest epidemiological data yet demonstrating an association between OSA and daytime hypertension. Indeed, they support the notion that OSA contributes to the development of hypertension.

The most compelling experimental evidence that OSA can cause hypertension comes from studies in animals. Brooks and coworkers (119) demonstrated that exposure of dogs to obstructive apneas during sleep for 1–3 mo caused chronic elevations in BP both during sleep and wakefulness. In addition, reversal of OSA caused nocturnal and daytime BP to fall back to baseline levels within 1–4 wk. However, exposure of the same dogs to acoustic stimuli during sleep, which provoked arousals and acute BP elevations equal to those observed during OSA, did not lead to diurnal hypertension. Similar results have been described in rats (123). These findings indicated that arousals from sleep were insufficient to cause diurnal hypertension, but that additional stimuli related to OSA were necessary for its development.

The mechanisms through which OSA promotes hypertension have not been fully elucidated. However, converging evidence from physiological studies in humans and animals points toward intermittent hypoxia and sympathetic nervous system activation playing central roles. As described above, elevated sympathetic vasoconstrictor nerve traffic in OSA appears to be sustained into wakefulness through chronic alterations in chemoreflex sensitivity and baroreflex set point. Together, these may promote upregulation of sympathetic vasoconstrictor tone. Conversely, when OSA is eliminated by CPAP, sympathetic vasoconstrictor activity falls both acutely at night (41) and chronically during wakefulness (58, 59) in association with reductions in BP.

Consistent with these observations, Fletcher and colleagues demonstrated that intermittent exposure of rats to hypoxia for 8 h/d over 35 d induced sustained arterial hypertension (124). Elevations in BP were prevented by both carotid body and sympathetic nervous system denervation (125, 126). These observations indicated that episodic hypoxia exerted its long-term effects on BP through stimulation of the peripheral chemoreceptors, which are potent stimulators of brainstem sympathetic vasoconstrictor outflow. Additional studies demonstrated that exposure of rats to intermittent hypoxia provoked increased activity of the adrenal glands, renal sympathetic nerves, and the renin angiotensin system, all of which could promote increases in BP (127, 128). Recent work also suggests that patients with OSA have an increased pressor response to hypoxia (66). However, although abnormal vascular endothelial-mediated vasodilatory and constrictor responsiveness has been postulated as contributing to hypertension in OSA, data on these points are inconclusive (66, 68, 93-97).

Another interesting observation arising from these experiments was that sustained elevations in BP in response to intermittent hypoxia occurred only in certain spontaneously hypertensive strains of rats. These observations therefore suggest that vascular responses to intermittent hypoxia are at least partially under genetic control. They further imply that there may be a subset of humans and animals with a genetic predisposition to the development of systemic hypertension in response to OSA. If so, this might explain why, among patients with OSA of equal severity, some develop hypertension and others do not.

In summary, animal models of OSA strongly support the notion that sustained diurnal hypertension can arise from chronic exposure to recurrent obstructive apneas. However, findings from these experiments in animals must be interpreted with a few caveats. First, there may be important differences in response to apneas between species. Second, OSA in humans probably evolves over many years, which could allow for the development of compensatory protective mechanisms. Findings derived from animal models in which severe sleep apnea is rapidly introduced may not be generalizable to the human disease state. Notwithstanding these reservations, animal models have provided powerful evidence for a direct causative link between OSA and daytime hypertension.

In the controversy about whether OSA can lead to daytime hypertension, one point that is frequently overlooked is that OSA undoubtedly causes elevations in BP during sleep. Considering that humans typically spend one-third of their lives sleeping, it is necessary to consider the possibility that these nocturnal increases in BP might in themselves contribute to hypertensive cardiovascular complications (129-132).

The normal nocturnal decrease in BP averages about 15% below daytime levels and is generally preserved in patients with essential hypertension (i.e., “dippers”). In contrast, many OSA patients are “nondippers” because they have an attenuated or absent fall in nocturnal BP (133-135). It is therefore possible that many patients with essential hypertension who have a “nondipping” pattern of nocturnal BP have undiagnosed OSA. For example, Portaluppi and coworkers (135) reported the presence of unsuspected OSA in 10 of 11 “nondipping” patients with hypertension but in none of 10 “dipping” patients with hypertension. These results imply that the “nondipping” condition is probably related to sleep apnea in many cases of essential hypertension. Because “nondipping” is associated with a higher risk of cardiovascular complications independent of daytime BP levels (129-132), the debate about whether OSA leads to daytime hypertension may be less important than previously supposed.

A number of studies have suggested that hypertension in patients with OSA is more difficult to control by conventional means than is hypertension in nonapneic patients (136). Conversely, OSA is common in patients with difficult to control hypertension (137). Furthermore, it has been recently reported that in patients whose hypertension was refractory to maximal medical therapy, 87% had OSA (138). Taken together, these findings suggest that hypertensive patients with OSA may be relatively resistant to the antihypertensive medications.

Among the commonly prescribed antihypertensive agents, β-blockers have been reported to be the most effective in lowering daytime BP in hypertensive patients with OSA (139). This observation is in keeping with the hypothesis that sympathetic overactivity is involved in the pathogenesis of OSA-induced hypertension. However, antihypertensive medications have little effect on nocturnal BP in OSA, possibly because they do not alleviate OSA (139, 140).

A number of uncontrolled studies have reported that reversal of OSA by tracheostomy or CPAP in patients who were normotensive while awake was associated with a reduction in BP and MSNA during sleep (41, 57). Similar results have been reported in patients with heart failure with OSA who were on BP lowering medications (38).

Several randomized (141-143) and nonrandomized (144– 147) clinical trials have addressed the effects of CPAP on daytime BP in patients with OSA. However, these studies suffered from a serious flaw in design; the great majority of patients studied were normotensive while awake. Not surprisingly, the effects of CPAP on diurnal BP were trivial or nonexistent. Only one nonrandomized trial has examined the effects of CPAP in patients with OSA all of whom were hypertensive while awake despite aggressive antihypertensive therapy. Logan and coworkers (148) found that treatment of these patients with CPAP over a 2-mo period led to clinically significant reductions in both nocturnal and daytime BP. Although not definitive, these results do emphasize the need for well-designed randomized trials to ascertain with certainty whether treating OSA will lower daytime BP in patients with hypertension. The findings of such studies will be very important because OSA may well prove to be the commonest, potentially treatable secondary cause of hypertension.

In North America, stroke is the third leading cause of death and the leading cause of long-term disability (149, 150). The first large-scale population-based study to examine the potential relationship between sleep apnea and stroke was in 6,424 subjects in the Sleep Heart Health Study. The presence of OSA was associated with a clear, but modestly increased prevalence of stroke; the odds ratio for the highest AHI quartile (AHI > 11) was 1.58 times (CI 1.02–2.46) that of the lowest quartile (AHI < 1.4) (151).

Among subjects who have suffered a stroke, sleep apnea is extremely common and is reported to occur in 43–91% of patients (152-156). These prevalences are significantly higher than in control subjects without stroke. In all these studies, OSA predominated, whereas central sleep apnea (CSA) was observed in only a small minority of patients (less than 10%). It remains unclear, however, to what extent sleep apnea detected following a stroke might be a consequence of the stroke, rather than a preexisting condition. Bassetti and Aldrich (154) argued that OSA most likely preceded stroke, based on the observation that the frequency and severity of sleep apnea did not differ between patients with stroke and those with transient ischemic attacks. In addition, Parra and colleagues (156) found that the frequency of obstructive apneas did not decline between the period immediately after the stroke to 3 mo later. However, there was a decline in frequency of central apneas. These data suggested that acute stroke predisposed to development of CSA, but that OSA was more likely already present at the time of the stroke.

A number of mechanisms through which OSA could predispose to cerebrovascular events are discussed above. These include systemic hypertension, increased platelet aggregability, and blood coagulability (109, 110, 112). In addition, during obstructive apneas there is a significant decline in cerebral blood flow due mainly to a reduction in cardiac output (157, 158). The decline in cerebral perfusion is also tightly linked with reductions in BP, which could predispose to cerebrovascular accidents in subjects with flow-limiting lesions of the carotid or vertebral circulations. This would be of particular relevance during REM sleep when cerebral blood flow and oxygen demands are normally highest, but when apneas are accompanied by the greatest degrees of hypoxia (159). Such an imbalance could put the cerebral circulation at risk, especially during the early morning hours during, or immediately following, the longest episode of REM sleep (106, 153, 160). In addition, abrupt and marked alterations in blood flow velocity associated with alternating obstructive apneas and hyperpneas could lead to abrupt alterations in vascular shear forces, and acceleration of atherosclerosis (159, 161, 162).

Central apneas have been reported, in a number of older studies, to be associated with bilateral hemispheric and brainstem infarctions (163-167). However, in all except one of these studies (167), sleep monitoring was not performed and potential behavioral influences related to wakefulness could not be ruled out. More recent data have not detected an association between stroke location and apnea type (156). The observation that central apneas dissipate over time following stroke suggests that strokes may promote central apneas in the immediate poststroke period, but as brain inflammation and edema resolve, so do central apneas. However, there are as yet insufficient physiological data to determine the type (hypercapnic or nonhypercapnic) or etiology of central apneas in the poststroke period (168).

What has become clear, however, is that Cheyne–Stokes respiration is decidedly uncommon following stroke. Thus the attribution of Cheyne–Stokes respiration to cerebral damage in the older literature (169) must be reevaluated in light of more recent findings. Because Cheyne–Stokes respiration is essentially periodic breathing with a long cycle length, and because cycle length is inversely proportional to cardiac output (40), it is possible that this breathing disorder is a sign of underlying cardiac dysfunction (see below). In fact, coronary artery disease is present in approximately one-third of patients with stroke and, over the long term, cardiac events are the leading cause of death following stroke (170). It is therefore possible that in the older studies of Cheyne–Stokes respiration in poststroke patients, underlying cardiac dysfunction may have been present, but undetected (165, 166, 169). Consequently, the pathophysiology and clinical significance of CSA following stroke remain to be elucidated.

When sleep apnea is present following a stroke, it could further compromise cognitive and physical functioning. For example, OSA is associated with excessive daytime sleepiness and impaired cognitive function, reaction times, and simulated driving performance (171, 172). In addition, apnea-related hypoxia can lead to the elaboration of neuroinhibitory peptides, such as γ-aminobutyric acid, which could also further compromise cerebral function (173). Indeed, following a stroke, patients with OSA have worse functional capacity than patients without OSA. This difference could not be attributed to differences in the severity of the stroke (152, 153).

Well-designed randomized clinical trials in which the effects of treating sleep apnea on neurological outcomes in patients with strokes have yet to be performed. Such studies need to be undertaken, for they will provide the evidence upon which a rational approach to the diagnosis and treatment of sleep apnea in patients with strokes will be based.

In a cross-sectional analysis of the Sleep Heart Health Study cohort, OSA was found to be an independent risk factor for coronary artery disease (CAD) (151). However, the association was modest; the odds ratio for CAD of the highest AHI quartile (AHI > 11) was 1.27 times (CI 0.99–1.62) that of the lowest quartile (AHI < 1.4). Prospective follow-up of the cohort should provide additional insights into the relationship between OSA and CAD. Because it has been established that OSA increases the risk for hypertension, it seems likely that it would also increase the risk for disorders, such as CAD, that are associated with hypertension.

As described above, OSA exerts several acute physiological effects that could predispose to myocardial ischemia during sleep. In dogs with experimentally induced coronary artery stenosis, obstructive apneas can lead to myocardial ischemia even in the absence of hypoxia (174, 175). However, in the absence of a coronary stenosis, myocardial ischemia was not observed. Others have reported that electrocardiographic signs of ischemia in patients with OSA with CAD were more closely linked to increases in HR and BP related to apneas than to O₂ desaturation. These observations suggested that the main trigger of ischemia was an increase in O₂ demand rather than a reduction in supply (176). In another study, Mueller maneuvers, which simulated the effects of obstructive apneas, caused more pronounced reductions in LV ejection fraction (LVEF) in humans with, than in those without CAD (177). These findings emphasize that the diseased myocardium is more susceptible to the adverse effects of obstructive apneas than is the normal myocardium.

Nocturnal ST-segment changes consistent with myocardial ischemia are quite common among patients with OSA and coexisting CAD. Various studies have reported prevalences of such ischemic changes ranging from 20 to 100% (176, 178-182). ST-segment depressions are more frequent in those with more severe OSA or prior complaints of nocturnal angina (179, 180). Ischemic episodes have been related both to O₂ desaturation and to the postapneic surges in HR and BP (176, 179, 180), and can provoke awakening with complaints of angina (179-181) (Figure 3). Myocardial ischemia during sleep in patients with OSA may also be asymptomatic (i.e., silent). However, to date, there are no reports of the prevalence of OSA among patients with CAD with silent nocturnal ischemia.

Fig. 3. ST segment depressions, expressed as ST segment change vector magnitude (STC-VM) in a patient with coronary artery disease during recurrent obstructive apneas. Increased STC-VM indicates ST segment depression. The most pronounced fall in oxyhemoglobin saturation (Sa_O2 ) is accompanied by the greatest degree of ST segment depression during which the patient woke up with angina. From Franklin and coworkers (179), with permission.

Download Figure | Download Powerpoint

[More] [Minimize]

Studies in patients with OSA without coexisting CAD have been less consistent. In some, no evidence of nocturnal ischemic episodes in such patients was reported (176, 183, 184), whereas in another, ST-segment depressions during the night occurred in 30% of patients (185). Acute application of CPAP to these patients significantly reduced the total duration of ST-segment depression. The reason for discrepancies between these studies remains unclear. Accordingly, there is uncertainty as to whether OSA causes nocturnal myocardial ischemia in the absence of CAD.

In patients with CAD, OSA may be a poor prognostic indicator. Among 62 patients with CAD followed prospectively for 5 yr, Peker and coworkers (186) found that those with OSA had a significantly higher mortality rate (38%) than those without OSA (9%, p = 0.018), even after controlling for potentially confounding factors.

There are no published randomized trials on the effects of treating OSA on nocturnal myocardial ischemia and angina. However, in uncontrolled studies, treatment of OSA with CPAP in patients with nocturnal angina was associated acutely with reduced frequency of ST-segment depression and chronically with relief of nocturnal angina (176, 179, 180). Nevertheless, larger, longer term randomized trials are required to more clearly delineate the role of diagnosing and treating OSA in patients with CAD.

Observations from epidemiological studies indicate an association between OSA and CHF. In the Sleep Heart Health Study, the presence of OSA (i.e., AHI ⩾ 11) was associated with a 2.38 relative odds for CHF independent of other known risk factors (151). This risk exceeded that for all other cardiovascular diseases examined including hypertension, CAD, and stroke. In the two largest series of patients with CHF with systolic dysfunction evaluated for sleep-disordered breathing, 11% of 81 patients (187) and 37% of 450 patients (188) had OSA. This prevalence exceeds the approximate 5–10% prevalence of OSA reported in otherwise healthy adults (189). Risk factors for OSA among patients with CHF differed by sex: among men only increasing body mass index, and among women only increasing age were significant (188).

In addition, as many as one-third of patients with clinical CHF have normal systolic function, but are thought to have diastolic dysfunction (190). Diastolic dysfunction results from inadequate diastolic filling due to reduced LV compliance. There has been only one study examining the prevalence of sleep-disordered breathing in diastolic heart failure. Chan and coworkers (191) found that 55% of 20 patients with this condition had an AHI > 10. Of these, approximately two-thirds had predominantly obstructive apneas, suggesting that OSA may play a role in the development of diastolic dysfunction. Taken together, these epidemiological data raise the possibility that OSA can contribute to the development of both systolic and diastolic LV dysfunction.

On the other hand, there remains the possibility that CHF itself can contribute to the development of OSA. Two lines of evidence support this view. First, it has long been hypothesized that an underlying tendency to periodic breathing could destabilize the upper airway and predispose to collapse. Patients with CHF are predisposed to periodic breathing because of increased chemosensitivity, hyperventilation, and, possibly, prolonged circulation time (192-194). According to this view, as respiratory drive declines during the waning phase of periodic breathing, drive to the pharyngeal dilator muscles declines as well (195, 196). Should the relative reduction in pharyngeal dilator muscle tone exceed that of the diaphragm, upper airway collapse could ensue. Indeed, in contrast to the abrupt rise and rapid fall in tidal volume that characterizes OSA in patients with normal cardiac function, in patients with CHF with OSA, tidal volumes during hyperpneas often have a waxing–waning appearance typical of Cheyne–Stokes respiration (38). This suggests the presence of an underlying periodic breathing disorder. Second, since patients with CHF suffer from fluid retention and dependent edema, it is possible that upon reclining, accumulation of edema in the soft tissues of the neck and pharynx could narrow the upper airway and make it more collapsible (197). Nevertheless, regardless of the underlying cause of upper airway collapse in patients with CHF with OSA, it is rapidly reversed by acute application of CPAP (38, 198).

As discussed above, and illustrated in Figure 4, OSA exposes the left ventricle to several factors that could impair its function. It is conceivable that over months to years, these factors could have cumulative adverse effects on LV structure and diastolic and systolic function in some individuals.

Fig. 4. Schematic representation of the pathophysiological effects of OSA on the cardiovascular system. Modified from Hall and Bradley (275), with permission.

Download Figure | Download Powerpoint

[More] [Minimize]

There are several case reports of patients presenting with acute nocturnal pulmonary edema who were subsequently found to have OSA, but normal LV systolic function (199, 200). Treatment of OSA with CPAP was associated with a reduction in episodes of pulmonary edema. These cases suggest that the adverse effects of OSA can be sufficient to cause acute LV failure during sleep in susceptible individuals.

Hypertension is the single most important predisposing factor for the development of LV hypertrophy and systolic and diastolic LV failure (201, 202). Accordingly, the most obvious mechanism through which OSA could lead to the development or progression of LV failure is systemic hypertension. However, ischemia and reduced contractility due to hypoxia (42), as well as cardiac myocyte injury or necrosis due to increased catecholamine stimulation (203), could also contribute. Regardless of the precise mechanisms involved, there is now direct experimental evidence that OSA can lead to interstitial pulmonary edema as well as LV hypertrophy and dysfunction. In anesthetized dogs, Fletcher and colleagues (204) demonstrated the development of subtle degrees of interstitial pulmonary edema after 8 h of exposure to recurrent obstructive apneas. In addition, Parker and coworkers (34) showed that exposure of dogs to experimental obstructive apneas during sleep for several weeks to months led to the development of increased LV mass and reduced ejection fraction in association with systemic hypertension. However, it is not clear whether OSA can lead to LV hypertrophy in humans. Whereas one study reported greater LV wall thickness in normotensive OSA patients than in normotensive control subjects (205), another did not (206).

Because the diseased left ventricle is more susceptible to the negative hemodynamic effects of increases in afterload than the normal left ventricle (207), OSA may have particularly deleterious effects in patients with coexisting heart disease. For example, when humans with CHF are exposed to exaggerated negative intrathoracic pressure during Mueller maneuvers, they experience greater reductions in stroke volume and cardiac output that persist longer into the postapneic period than in subjects with normal ventricular function (208). In addition, survival in CHF is inversely proportional to cardiac noradrenergic drive (209). Therefore, elevated SNA associated with OSA (41, 53, 54) would have particularly adverse prognostic implications in patients with coexisting heart failure.

Only one study has examined the effects of treating OSA on LV systolic function in patients with CHF. Malone and colleagues (198) demonstrated, in eight patients with idiopathic dilated cardiomyopathy, that treatment of coexisting OSA by CPAP for 1 mo caused dramatic improvements in LVEF (from 37% to 49%) and cardiac functional status. Although these data are encouraging, larger randomized trials will be required to determine whether treatment of OSA in patients with systolic and diastolic heart failure improves cardiac function and other cardiovascular outcomes. The findings of the above mentioned study also suggested that OSA can contribute to the development of LV systolic dysfunction in patients with CHF of otherwise unknown etiology.

Patients with CHF with CSR–CSA have lower Pa_CO2 both during wakefulness and sleep than those without CSR–CSA (194, 214, 215). Moreover, episodes of CSR–CSA are usually triggered by abrupt increases in ventilation and reductions in Pa_CO2 below the apneic threshold (194). Inhalation of a CO₂-enriched gas mixture sufficient to raise Pa_CO2 by just 1–3 mm Hg eliminates central apneas and hypopneas both in patients with CSR–CSA and idiopathic CSA (216, 217). These findings also emphasize that the prevailing P_CO2 in patients with CSR–CSA is closer to their sleeping apneic threshold (i.e., within 1–3 mm Hg) than it is in healthy subjects whose prevailing Pa_CO2 is 3–5 mm Hg above the sleeping apneic threshold (218). Under this condition, even a modest increase in ventilation can drive Pa_CO2 below the apneic threshold (194, 219).

The mechanisms responsible for chronic hypocapnia in patients with CSR–CSA have not been fully elucidated. One possible explanation is hypoxia. However, both awake and mean nocturnal Sa_O2 in patients with CHF with CSR–CSA are usually within normal limits and do not differ from those in patients with CHF without CSR–CSA (194, 214, 220). Another possibility is that hyperventilation arises from stimulation of pulmonary vagal irritant receptors by increased LV volumes, filling pressures, and pulmonary congestion (220-222). In favor of this hypothesis, Solin and coworkers (223) demonstrated significantly higher pulmonary capillary wedge pressures and lower Pa_CO2 in patients with CHF than in those without CSR– CSA. Others have demonstrated an inverse relationship between Pa_CO2 and LV filling pressure in patients with CHF (224). Accordingly, CSR–CSA may be a respiratory manifestation of elevated LV filling pressure (220).

Elevated chemoreceptor responsiveness (gain) could destabilize the respiratory control system, both by decreasing the prevailing Pa_CO2 and by increasing the tendency to hyperventilate, thus precipitating ventilatory overshoot (193, 225, 226). Several studies have demonstrated higher central (227, 228) and peripheral chemoresponsiveness (229, 230) in patients with CHF with either periodic breathing while awake or CSR–CSA than among those either with OSA or without any sleep apnea. It remains to be determined whether such increases in central and peripheral chemoresponsiveness predates or is a consequence of the development of CHF (231). Evidence for the latter hypothesis comes from experiments in rabbits in whom induction of CHF led to increased peripheral but not central chemosensitivity (193). These results support the notion that CHF itself can sensitize the peripheral chemoreceptors, possibly by impairing the ventilatory inhibitory effect of carotid body nitric oxide production (193).

The role of hypoxia in the pathogenesis of CSR– CSA in heart failure is uncertain. Studies of patients with CHF and CSR–CSA have consistently shown them to be normoxic while awake (194, 214). Nevertheless, hypoxic dips during apneas may propagate CSR–CSA by provoking arousals and amplifying the ventilatory response to CO₂ at the termination of central apneas (225). A number of studies have investigated the effects of supplemental O₂ in patients with CHF and CSR– CSA, with varying results. In general, where O₂ was administered at high flow rates, the frequency of central respiratory events was significantly reduced (232-234). In one study, reductions in central apneas were associated with increases in P_CO2 (235). This observation suggested that increases in Pa_O2 suppressed respiratory drive allowing Pa_CO2 to rise above the apnea threshold (235). In contrast, during studies in which O₂ was administered to patients with CSR–CSA at lower flow rates just sufficient to raise Sa_O2 into the normoxic range, little or no effect was observed on the frequency of central events or on P_CO2 (216, 236). It is therefore likely that where supplemental O₂ reduces the severity of CSR–CSA, it does so partly by raising Pa_CO2 .

Shifts in the state of consciousness are likely to destabilize breathing. With the transition from wakefulness to NREM sleep, the waking neural drive to breathe is lost, and the threshold for a ventilatory response to CO₂ is increased (4). Therefore, if the ambient Pa_CO2 during wakefulness is below this higher sleeping threshold, the transition to NREM sleep will be accompanied by a transient loss of respiratory drive resulting in a central apnea. During the apnea, Pa_CO2 rises until it reaches the new higher threshold level and initiates breathing. If sleep becomes firmly established, regular breathing resumes (237). However, if an arousal should occur, the increased Pa_CO2 level associated with sleep now represents a state of relative hypercapnia for wakefulness and will stimulate hyperpnea. In addition, other studies have demonstrated that arousals cause ventilation to rise above the waking level independent of Pa_CO2 (17). Accordingly, arousal contributes independently to ventilatory control system instability.

The role of arousal in CSR–CSA contrasts with its role in OSA where it is a critical defense mechanism that terminates apneas by triggering opening of the upper airway. In CSR– CSA, however, arousals often occur after airflow has resumed. It therefore appears that resumption of airflow following central apneas is not always dependent on an arousal (194, 214, 219). Paradoxically, arousals actually trigger and propagate central apneas and therefore contribute to their genesis. Indeed, the critical role of arousals in sustaining ventilatory overshoot during periodic breathing has been demonstrated both in patients with CSR–CSA and those with idiopathic CSA, a disorder that shares many features of CSR–CSA (194, 237, 238).

In contrast to OSA, CSR–CSA is more pronounced during NREM sleep, where chemical–metabolic factors are the predominant influence on ventilatory control, than during REM sleep where behavioral nonmetabolic factors predominate (21, 194, 219). In addition, during REM sleep arousability to respiratory stimuli is diminished compared with the lighter stages of NREM sleep. As a result, the tendency to ventilatory overshoot and hypocapnia is diminished in REM sleep, which dampens CSR–CSA. Similarly, in the deeper (i.e., slow wave) stages of NREM sleep where arousability is also decreased, CSR–CSA tends to dissipate (194). Although Cheyne–Stokes respiration can be observed in wakefulness (18, 219), it is much less common here than in NREM sleep probably because of nonmetabolic and behavioral influences.

Upper airway obstruction has been described at the onset and at the end of central apneas in some patients with CHF (195) and could promote ventilatory overshoot upon the abrupt decrease in upper airways resistance at the termination of apneas (196). It is also possible that upper airway collapse itself can reflexively precipitate central apneas (239). A more likely possibility in the setting of CHF, however, is that adverse effects of OSA on cardiac function can predispose to central apneas. In patients with CHF having approximately equal numbers of obstructive and central apneas, obstructive apneas predominate at the beginning of the night, whereas central apneas predominate at the end. The shift from obstructive to central events occurred in association with a decrease in Pco ₂ and prolongation of circulation time (240). Thus, obstructive events likely converted to central events as Pco ₂ fell below the apneic threshold due to an overnight deterioration in cardiac function. Such deterioration probably arose from the combined effects of OSA and of increasing venous return in the recumbent position that raised LV filling pressure and lowered Pco ₂ (240, 241). These observations raise the question as to whether OSA could, over time, predispose to CSR–CSA after the onset of cardiac failure.

Prolonged circulation time causing delays in transmitting changes in arterial blood gas tensions within the lungs to the chemoreceptors could theoretically destabilize the respiratory control system. It could do so by changing a negative feedback into a positive feedback system (226). In support of this theory, Guyton (242) induced periodic breathing in dogs by inserting a length of tubing between the heart and cerebral circulation to prolong transit time between the lungs and the chemoreceptors. However, CSR–CSA could be generated only when the circulatory delay was a few minutes in duration, far exceeding that seen in patients with CHF. In humans, lung to peripheral chemoreceptor circulation time is inversely proportional to stroke volume and cardiac output (40). Mortara and coworkers (243) found, in patients with CHF, that prolonged circulation time was a major determinant of the presence of CSR–CSA. In contrast, other investigators have found no significant differences in LVEF, circulation time, or cardiac output between patients with CHF with and without CSR–CSA (194, 210, 223). Taken together, these data suggest that prolonged circulatory delay is probably not the critical factor predisposing to central apneas in CHF. However, circulatory delay influences periodic breathing cycle length such that cycle length is proportional to lung to carotid body circulation time (40). This accounts for the greater length of the periodic breathing cycle in patients with CHF than in subjects with idiopathic CSA whose cardiac function is normal.