The major finding of this study is that hypercapnic vasodilation in the cerebral circulation, an exquisitely sensitive physiologic mechanism responsible for minimizing changes in brain tissue Pco2 during fluctuations in arterial CO2, is diminished, in graded fashion, across the continuum of mild to severe SDB. We observed a significant positive correlation between the mean level of SaO2 during sleep and cerebrovascular CO2 reactivity. In contrast, AHI, regardless of whether it was represented as a categorical or continuous variable, was not statistically significantly associated with cerebrovascular CO2 reactivity. Although we cannot infer causation from our observational data, we interpret these findings as indirect evidence that nocturnal hypoxemia contributes importantly to impaired cerebrovascular function in individuals with SDB.
This conclusion is predicated on the assumption that reductions in SaO2 during sleep were caused by SDB. Alternatively, SaO2 may have been reduced secondary to lung disease. However, we believe this possibility was minimized because participants with a history of lung disease and those taking medications used to treat lung disease were excluded from this analysis. Also, our assumption that the reduction in CCR we observed in participants in the lower quartiles of nocturnal SaO2 represents impaired hypercapnic vasodilation is valid only if the rebreathing test evoked comparable increases in MAP across all subject groups. This is a reasonable assumption because there was no correlation between nocturnal SaO2 and MAP during rebreathing and because the SaO2–CCR relationship was statistically significant both before and after adjustment for the increase in MAP. In contrast, the AHI–CCR relationship was confounded by between-group differences in MAP: a negative correlation of borderline significance was observed between AHI and MAP during rebreathing. Thus, it is not possible to discern whether CCR was lower in participants with higher AHI because of blunted hypercapnic vasodilation or whether CCR was lower secondary to smaller increases in cerebral perfusion pressure during rebreathing. The finding that the AHI–CCR relationship became statistically nonsignificant after adjustment for the increase in MAP during rebreathing points to the latter possibility. Our data are consistent in demonstrating that, across a wide spectrum of SDB, the mean level of SaO2 during sleep better predicts the blunting of CO2 reactivity than does the frequency of events (i.e., AHI).
In contrast to the significant negative correlation between SDB and CCR in our subjects, ventilatory responses during rebreathing did not vary according to SaO2
or AHI category. In one previous study, enhanced ventilatory responses to CO2
were reported in OSA patients versus control subjects (40
); however, several other studies have found no between-group differences (16
). We speculate that much of the variability in ventilatory responses to CO2
is attributable to performance of these studies during wakefulness, when nonchemoreceptor, behavioral inputs have a substantial influence on respiratory output.
The present findings in a population-based sample free of clinical selection biases are consistent with our previous observations of diminished CCR in patients with moderate to severe OSA (16
). The magnitude of CCR decrement in the present subjects who fell within the lowest versus the highest quartiles of nocturnal SaO2
values was similar to that observed in our clinic-based sample (−17 versus −22%). Because the average CCR in CPAP users was enhanced relative to participants with untreated SDB and AHI greater than 30, the present findings suggest that SDB-related impairment in cerebrovascular function is reversible, at least in part, with treatment. This finding also agrees with our previous observations of patients with moderate to severe OSA (16
). Nevertheless, because the average CCR in CPAP users was still somewhat diminished relative to that of participants with no SDB (AHI <5), the present data also suggest that CPAP does not provide full protection against this vascular consequence of SDB. A potential reason is that many individuals are not fully compliant with this treatment. Patients who are considered “CPAP compliant” use the device for as few as 4 hours on most nights (43
). Thus, many of them remain exposed, for varying amounts of time, to the adverse effects of SDB and resultant intermittent hypoxemia.
The present findings also parallel two previous population-based studies that used flow-mediated dilation in the forearm to assess vascular function (21
). All three studies found associations between vascular dysfunction and SDB severity. Consistent with the present findings, one previous study found that nocturnal SaO2
was a more important predictor of vascular function than was AHI (21
). The other previous study observed that SDB was associated with impaired vascular function in females, but not males (22
). In the present study, the interaction between sex and SDB-associated vascular dysfunction was not statistically significant. Interestingly, a significant positive correlation was observed between vascular reactivity and alcohol consumption in one previous report (22
). A similar correlation was not present in our study.
Mechanisms of SDB-induced Impairment in Cerebrovascular Reactivity
Hypercapnic vasodilation in the brain is a complex, endothelium-dependent process: nitric oxide, prostacyclin, and cytochrome P-450 metabolites have all been implicated (46
). Endothelium-dependent dilation in the forearm is blunted in individuals with SDB (21
). The causes of this impairment are not well understood; however, oxidative stress (i.e., imbalance between production of reactive oxygen species and antioxidant defenses) and inflammation are putative contributors. Both processes have been observed in patients with OSA (4
), are accompanied by decreased expression of endothelial nitric oxide synthase and increased expression of both nitrotyrosine and inducible nitric oxide synthase in venous endothelial cells (51
), and are ameliorated by CPAP treatment (51
). In addition, allopurinol treatment has been shown to normalize impaired flow-mediated dilation in patients with OSA (52
), which suggests an important role for xanthine oxidase–derived superoxide. Excess superoxide would be expected to reduce the availability of nitric oxide by combining with it to form peroxynitrite. Peroxynitrite, in turn, could further limit nitric oxide via oxidation of tetrahydrobiopterin, a critical cofactor for endothelial nitric oxide synthase (53
). Peroxynitrite could also limit prostacyclin production by suppression of prostacyclin synthase (55
). We speculate that the pathologic processes that interfere with endothelium-dependent vasodilation in the forearm also contribute to the observed blunting of hypercapnic vasodilation in the cerebral circulation.
Our conclusions regarding cerebrovascular responses are predicated on the assumption that Doppler measurements of flow velocity are reflective of volume flow, an assumption that is satisfied only when the cross-sectional area of the artery remains constant. We did not measure diameter; however, previous investigators have shown that middle cerebral artery diameter varies by less than or equal to 4% during changes in arterial pressure, CO2
), or gravitational stress (57
). In addition, velocity and volume flow through the middle cerebral artery are highly correlated (58
). Also, one of our measures of SDB severity, AHI, has limited reliability, especially over a single night of observation. Nevertheless, with 373 subjects, we believe our study is adequately powered considering the known night-to-night variability (59
In the absence of SaO2 measurements during wakefulness and concurrent measurements of pulmonary function, individuals with lung disease were identified based on self-reported medical history and use of medications used to treat lung disease. We recognize that this imprecise method is a limitation of our study.
Finally, in this cross-sectional study, we cannot uncover the temporal nature of the relationship between SDB and blunted CCR. On one hand, SDB could cause impairment in reactivity via intermittent hypoxemia and attendant insults to vascular structure and function. However, diminished cerebrovascular CO2 reactivity could cause or exacerbate SDB. Answers to these questions await longitudinal observations in population-based studies.
Clinical Significance of the Present Findings
Several previous reports suggest that the observed impairments in CCR may be clinically relevant. In patients with essential hypertension and diabetes mellitus, impaired CCR was correlated with endothelial dysfunction in the forearm (17
), a strong predictor of cardiovascular disease risk (18
). Because hypercapnic vasodilation in the cerebral circulation is evoked by substances produced in endothelial cells (46
), we believe that CCR is a proxy measure of endothelial function and therefore may also be a predictor of cardiovascular risk. Diminished cerebrovascular responsiveness to CO2
has been observed in patients with ischemic stroke (60
) and those with multiple subcortical infarctions (61
). Therefore, this functional impairment may play a pathogenetic role in cerebrovascular disease. Further, in patients with congestive heart failure, CCR was correlated with ejection fraction, prompting the speculation that depressed cerebrovascular reactivity may be responsible for cognitive impairments in patients with severe ventricular dysfunction (62
). We have shown that CCR is reduced in patients with heart failure and central sleep apnea relative to patients with similar cardiac dysfunction but without central sleep apnea (63
). Because CO2
reactivity in the cerebral circulation minimizes changes in brain Pco2
during fluctuations in arterial Pco2
, the observed compromise in cerebrovascular regulation may disturb breathing stability during sleep. Reductions in CCR could exacerbate breathing instability during sleep by exaggerating the accumulation and also the washout of CO2
from central chemoreceptors during fluctuations in ventilation and arterial Pco2