INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Sleep apnea syndrome (SAS) has been identified as a risk factor for cardiovascular diseases such as coronary heart disease (CHD) and arterial hypertension (1, 2). The high morbidity and mortality reported in patients with SAS appears to be related to cardiovascular disease (3, 4).

Increasing evidence suggests that snoring and SAS are also associated with cerebrovascular disease (CVD), either as a risk factor (5) or as a consequence of certain neurological locations (9). Several studies have reported a high frequency of snoring and/or a history of sleep-related apneic periods (5, 12), but it is only more recently that the presence of SAS has been tested by means of polysomnography (PSG) (12). This technique has been applied only to small groups, probably owing to the complex arrangements required when working with these neurological patients. Nevertheless, all of these studies have shown a high frequency of sleep apnea (SA) events. It is actually of major concern to what extent SA may be a risk factor or a consequence of the neurological disease. Some case control studies have shown a significantly higher frequency of snoring and/or SA in patients with cerebrovascular disease, suggesting that this condition is acting as a risk factor (5, 13, 14, 17, 18); most of these studies performed PSG on only a few patients (except for the study by Bassetti and Aldrich [18]). Alternatively, there are case reports suggesting that SA could be a consequence of neurological impairment (9).

Various types of sleep respiratory disturbances (SRD) (obstructive apneas [OA], central apneas [CA], and Cheyne- Stokes breathing [CSB] pattern) have been described in CVD, but in general, these descriptions lack topographic specificity (14, 16). From a prognostic point of view, it seems that long-term mortality is higher in patients with sleep-related breathing disorders (SRBDs) and acute CVD (14, 15), but it is not clear if there is a relationship between these phenomena and short-term functional outcome (16). None of these studies has analyzed the time course of the SRBD and its correlation with the neurological outcome.

Therefore, the aims of our study were (1) to analyze the frequency and characteristics of SRBD in patients with a first-ever stroke or transient ischemic attack (TIA), (2) to examine the correlation between neurological location and SRBD, and (3) to show the time course of SRBD after the neurological disease has stabilized. The following discussion tries to establish to what extent SRBD is a consequence of the acute phase or whether it remains when neurological impairment is stable or fully recovered.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients and Study Design

We prospectively studied 161 patients (82 males) with a first-ever stroke or TIA; these patients were admitted consecutively to the stroke unit of Sagrat Cor Hospital (Barcelona, Spain) and are included in the Stroke Registry, the details of which have been reported (19). One patient refused to take part, and is not included in this study. The human ethics committee of Sagrat Cor Hospital approved the protocol and informed consent was obtained from all patients or their families. Complete neurological assessment was performed to determine parenchymatous and vascular localization and the extent of the neurological vascular lesion, as well as the stroke subtype categorization. The protocol consisted of two sleep studies based on unattended portable respiratory recording (PRRs): the first at 48-72 h after admission (acute phase), and the second in a subsequent admission after 3 mo, when the neurologic disease was considered to be stable (stable phase).

Acute Phase

Neurological assessment. All patients were admitted to the hospital within 48 h of onset of symptoms. A complete clinical history was obtained: demographic characteristics (age and sex), cerebrovascular risk factors (history of hypertension, diabetes, myocardial infarction or angina, rheumatic heart disease, congestive heart failure, atrial fibrillation, smoking [> 20 cigarettes/d], alcohol abuse [> 80 g/d], chronic obstructive pulmonary disease), salient features of clinical and neurological examination, and results of routine laboratory tests (blood cell count, biochemical profile, serum electrolytes, urinalysis), chest radiography, and 12-lead electrocardiography were recorded.

Brain computed tomography (CT) was performed in all patients within the first week after admission. Overall, 53 (32.9%) patients were studied by magnetic resonance imaging (MRI). Other investigations performed in these patients included cerebral arterial digital subtraction angiography in 26 (16.2%), Doppler ultrasonography of supraaortic trunks in 93 (57.8%), and echocardiography in 113 (70.2%) to determine cardioembolic origin.

Functional abilities were assessed by the Barthel Index, a multifaceted scale questionnaire that measures morbidity and daily living activities (score ranges from 0 [maximal discapacity] to 100 [no discapacity]). The maximal severity of stroke or neurological impairment was estimated with the Canadian Scale (total score ranges from 0 [maximal impairment] to 10 [no impairment]). Stroke subtypes were also categorized according to the Cerebrovascular Study Group of the Spanish Society of Neurology classification, which is similar to the National Institute of Neurological Disorders and Stroke classification (20): transient ischemic attack (TIA); ischemic stroke (IS), either atherothrombotic, cardioembolic, lacunar, unusual, or undetermined origin; and intraparenchymatous hemorrhagic stroke (HS). Stroke parenchymatous topography was classified, when possible, on the basis of brain CT or MRI as hemispheric (identifying different areas), brainstem, or cerebellum. Stroke vascular location was classified when possible, according to neurological examination and/or complementary tests, as carotid or vertebrobasilar.

Clinical sleep assessment. Within the first 48-72 h we applied an in-house sleep-wake habits and symptoms questionnaire that consisted of 15 items including snoring, observed apnea, and hypersomnia in different situations (see Appendix ). The possible answers were as follows: never, rarely, sometimes, often, and always. Daytime sleepiness was also analyzed by means of the Epworth Sleepiness Scale (ESS) (21). The answers were obtained from the patients, if they were able to cooperate, or from relatives.

In all cases a respiratory sleep study was performed in the ward during the first 48-72 h after admission, using a portable respiratory recording device (Edentec monitor 3711) that had been previously validated for full polysomnography (PSG) (22). This portable device measures nasal and oral airflow (thermistry), chest wall movements (impedance), heart rate and oxygen saturation (finger pulse oximetry), snoring, and body position. In addition, full PSG was recorded simultaneously in 10 patients according to the established standard criteria (23).

SRBDs were classified as usual as obstructive apnea (OA) or central apnea (CA), with apnea considered a cessation of airflow lasting 10 s or more; OA is apnea with maintenance of thoracic motion, and CA is apnea without any thoracic motion. Hypopnea was considered a discernible reduction in airflow or thoracic motion lasting more than 10 s, associated with a cyclical dip of Sa_O2 greater than 2%. The CSB pattern was defined as periodic breathing, with central apnea or hypopnea alternating with hyperpnea in a crescendo-decrescendo pattern over 10% of bedtime (24). The apnea-hypopnea index (AHI) was calculated by taking into account the time spent with the PRR (lights-off marked the beginning of recording, usually between 11:00 and 12:00 P.M., the ending being between 6:00 and 7:00 A.M.). Manual scoring of these variables was performed in all cases. One fully trained, experienced scorer who, in addition, was blind to the neurological clinical data, performed the scoring. The CT₉₀ (i.e., the percentage of nighttime below 90% saturation) was obtained automatically.

Stable Phase

A second PRR was performed when possible (86 cases) after a mean period of 3 mo, when the neurologic disease was considered to be stable. The same Edentec monitor and the same criteria were used for all of the respiratory parameters. The scorer was always the same and was blind to previous sleep respiratory parameters and to neurological clinical data. For the remaining 75 patients a second study was not possible for various reasons: 8 died and 67 were denied readmission for various reasons, mainly difficulties for displacement to the hospital. A new physical examination was performed by the neurologist, and the Barthel Index and Canadian Scale were readministered.

Statitical Analysis

Data are expressed as means ± SD. We used for nominal variables the ² test or Fisher exact test. When comparing demographics and sleep parameters for different stroke subtypes or topographies, parametric and nonparametric tests were used according to normal or abnormal distributions. Analysis of variance (ANOVA) was used for multiple comparisons. The Bland and Altman test of concordance (25) was used to analyze the concordance between sleep events obtained by PRR and full PSG. A multiple logistic regression analysis was used to analyze if any variable of our in-house sleep questionnaire could predict the presence of an AHI > 10 (dependent variable). A simple linear correlation was used to correlate improvement (reduction) of sleep events from baseline and neurological improvement according to the Barthel Index and Canadian Scale. Significance was set at p < 0.05. All analyses were performed with the Statistical Package for the Social Sciences (SPSS, release 8.0 for Windows; SPSS, Chicago, IL).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Baseline data (Table 1) from the 161 patients who met the criteria of first-ever stroke or TIA were as follows: age, 72 ± 9 yr; body mass index (BMI), 26.6 ± 3.9 kg/m²; AHI, 21.2 ± 15.7; ESS, 4.8 ± 3.3. Stroke subtypes were classified as ischemic stroke (IS) in 112 cases (69.6%) (44 lacunar, 34 thrombotic, 25 cardioembolic, 7 essential, 2 unusual), TIA in 39 (24.2%), and hemorrhagic stoke (HS) in 10 (6.2%). Mean baseline data from the various groups did not show significant differences in terms of age, BMI, or sleepiness. The baseline mean Barthel Index was 75.5 ± 25.7; Canadian Scale, 7.9 ± 2.2. 

Acute Phase

Clinical sleep assessment and stroke subtype. Snoring was reported as "never or rarely" in 42 patients (26.1%), "sometimes" in 51 (31.7%), and "often or always" in 61 (37.8%); from 7 patients no answer was obtained. None of the in-house variables could be used to predict the presence of an AHI > 10 by multiple logistic regression analysis.

As shown in Table 1, there were no differences in AHI OAI, and CAI according to the stroke subtype (ANOVA). However, the CAI was lower in TIA when compared with the whole stroke group (3.3 ± 7.9 versus 6.4 ± 10.6, respectively) (t test, p = 0.059). Because a trend toward a difference was observed in the CAI in an ANOVA global comparison (F = 2.589, p = 0.078) we applied Scheffè contrast test, which gave us a p = 0.03 between TIA and hemorrhagic stroke. A CSB pattern was observed in 42 (26.1%) patients and was not significantly associated with any kind of stroke (² test). Only two patients with cardioembolic disease had evidence of cardiac insufficiency, and they did not exhibit a CSB pattern.

Considering different cutoff points for AHI: 116 (72%) patients had an AHI > 10, 76 (47.2%) had an AHI > 20, 45 (28%) had an AHI > 30, 19 (11.2%) had an AHI > 40, and 8 (5%) had an AHI > 50. These respiratory events, considering the whole group, were predominantly obstructive in 84 (52.2%) patients and predominantly central in 62 (38.5%), with the remaining 15 (9.3%) not exhibiting a predominant pattern. Of those with predominantly central events, 37 patients (23% of the whole group) were pure central (all apneic events being central). The mean CT₉₀ was 7.8 ± 15.7%.

According to the Bland and Altman method there was good concordance among the 10 patients evaluated simultaneously for AHI by PRR and full PSG (mean difference, 1.2 ± 6.8 [CI ± 4.5], which is not different from zero).

Clinical sleep assessment according to parenchymatous topography and vascular involvement. According to brain CT scan and/or MRI results, we defined the parenchymatous location in 97 of the 122 patients with established stroke. In the remaining 25 patients neuroimaging was negative (Table 2). Thus, the strokes of 81 patients (50.3% of the whole group) had a hemispheric location, 13 (8.1%) were located in the brainstem, and 3 (1.9%) were in the cerebellum. The carotid artery territory was involved in 84 (52.2%) patients and the vertebrobasilar area in 39 (24.2%) patients when considering the whole group.

The presence of SA, among patients with a cut-off AHI > 10 or more, was similar whether the stroke was characterized by hemispheric or brainstem involvement, or was carotid or vertebrobasilar. There were no significant differences in AHI, obstructive apnea index (OAI), or central apnea index (CAI) according to parenchymatous location or vascular involvement (ANOVA). Since hemispheric neurological lesions usually affect more than one specific area (frontal, parietal, temporal, occipital, internal capsule, base ganglion, ventricle, and/ or semioval center) we failed to show differences for each particular area. The brainstem was not associated with a higher incidence of SA. The presence of a CSB pattern was similar for various parenchymatous locations. The CT₉₀ did not show significant differences depending on parenchymatous or vascular location (ANOVA).

Stable Phase

Table 3 shows mean data of the 86 patients who had a second study. No differences in respiratory acute-phase data were found in comparing those with a second study and those without a second study. The BMI did not change significantly between the studies. Time in bed did not differ significantly between the acute and stable phase studies (398 ± 65.6 and 392 ± 57.2 min, respectively, p = NS). Snoring frequency was lower in the second study, being reported as "never or rarely" in 45 patients (40.7%), "sometimes" in 25 (29.1%), and "often or always" in 15 (17.5%); 1 patient did not give any answer. The AHI significantly decreased to 16.9 ± 13.8 events/h. The frequency of SA was also lower, considering different cutoff points for AHI: 53 patients (61.6%) had an AHI > 10, 29 (33.7%) had an AHI > 20, 17 (19.8%) had an AHI > 30, 5 (5.8%) had an AHI > 40, and 2 (2.3%) had an AHI > 50. These respiratory events were predominantly obstructive in 53 (61.6%) patients and predominantly central in 25 (29.1%) patients; the remaining 9 patients (8%) did not exhibit a predominant pattern. Of the 25 predominantly central patients, 12 (14% of the whole group) were pure central. A CSB pattern was observed in six patients (7.3%).

Table 3 also shows the levels of significance of the differences obtained between the first and second studies, including respiratory and neurological assessment. We found significant (t test for paired data) reductions in the AHI and CAI, but the OAI was not significantly different across both studies. When considering obstructive apneas plus hypopneas, no significant change was observed from baseline (15.7 ± 14.5 to 13.4 ± 11.6, p = 0.68). Therefore differences in AHI were probably due mainly to a reduction in the CAI. The presence of CSB also decreased, from 17 patients in the acute phase to 6 patients in the stable phase. The CT₉₀ was not significantly reduced in the second study (t test for paired data).

When we compared the acute and stable phases for each stroke subtype we obtained a similar tendency (Table 4). Thus, the AHI and CAI showed a significant reduction in the stable phase for IS. This was also true for TIA.

The presence of an AHI > 10 or > 30 did not represent a prognostic factor for a better or worse recovery from neurologic sequelae, since it failed to correlate with improving Barthel Index (BI) and Candian Scale (CS) scores: no correlation was found between improvement in AHI from baseline and improvement in Barthel Index and Canadian Scale (r = 0.34 and r = 0.17, respectively).

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Our results indicate that SRBDs are frequent in patients with first-ever stroke or TIA. However, there is a lack of correlation between neurological topography and the presence or type of SRBD. Obstructive events appear to be a condition prior to the neurological disease, since there are no significant differences between the acute and the stable phase or between different stroke subtypes. However, central events and CSB could be a consequence of the neurological disease, since they decrease significantly in the stable phase.

The identification of a high frequency of snoring and SA events in patients with CVD is not an isolated issue. Several studies have emphasized the high frequency of snoring and/or a history of sleep-related apneic periods in stroke patients but without confirming the presence of SRBD by polysomnographic means (5). Although large sample studies are lacking, Kapen and coworkers (12) observed that 34 of 47 (72% of their patients with stroke) had SAS defined as an AHI  10 and 30% had an AHI > 40. In the study by Bassetti and colleagues (13) 25 of 36 subjects (69%) had an AHI > 10, and 20 of 36 (55%) had an AHI > 30, with similar results in a more recent study from the same authors (18). Our results are similar, as 69.6% of our patients reported snoring, and 71.4% had an AHI > 10. The high number of central events and CSB in our study is also a remarkable feature since these episodes are rarely observed in typical patients with obstructive sleep apnea when the same PRR device is used (22).

Considering that patients with stroke have a high mean age and that the AHI increases with age (26, 27), some concern must be expressed about the actual clinical significance of an AHI between 10 and 20. In addition, daytime symptoms such as hypersomnolence are difficult to explore in these patients. However, given the high percentage of patients with an AHI indicating moderate to severe SAS (in our study, 46.6% of patients with an AHI > 20 and 28.8% with an AHI > 30), we believe this must be considered as clinically relevant. In this sense, epidemiological studies performed outside (28) and inside our country (29) have shown a lower prevalence of sleep-related breathing disorders, although these studies included patients less than 65 or 70 yr old. If we analyze a particular age group (60 to 70 yr old) in the study performed by Durán and co-workers in our country (29), and compare that group with those of our patients in the same age group, we observe remarkable differences in frequency. Thus, considering an AHI > 10, a prevalence of 28.9% was found in the population-based study (29) versus a frequency of 88.9% in our patients with CVD; available figures for an AHI > 20 and AHI > 30 are as follows: 14 versus 59.3% and 8.6 versus 37%, respectively. Ancoli-Israel and colleagues (30), in a study of a large, randomly selected group of elderly patients (65 yr of age or more), found a high frequency of SRBD (62% of patients with an AHI > 10); this frequency is lower than that obtained in our study. Other studies with smaller numbers of patients have obtained a frequency at least two times higher in patients compared with controls (13, 14, 18).

The full PSG set-up could be a limitation, considering the difficulty of applying this complex technique in these kinds of patients. Having a wide experience with PRR, we have used this simplified, easy-to-perform, easy-to-repeat, and validated method (22) in such patients with stroke. In the present study, we have also found a good concordance of PRR with respect to full PSG in the 10 patients in whom both methods were simultaneously performed.

Comparison of acute- and stable-phase parameters suggests that two different respiratory phenomena are interacting in CVD. The first phenomena is snoring and "obstructive events," which are present in a high proportion of our patients and appear independent of the type of stroke and its topography, being similar in both phases (acute and stable). The same considerations apply to the TIA group. If we consider obstructive apneas and hypopneas as a whole, we did not observe significant changes in the stable phase relative to baseline. Because a mild trend toward a reduction in hypopnea was observed in the stable phase, some caution must be introduced in evaluating the contribution of each type of event to the final AHI reduction. However, in support to our hypothesis, the mild percentual change in global obstructive events was clearly smaller than the decrease observed for central events. Overall, these findings strongly suggest that OA is a condition preceding CVD, and probably is acting as a risk factor. In this sense some pathophysiologic studies illustrate how OA could impair cerebral blood flow and thus cause stroke (31, 32). In particular, Netzer and coworkers (32) have demonstrated that a significant decline in blood flow of the middle cerebral artery occurs during obstructive apneas and hypopneas, but not during central events, concluding that these observations provide a pathophysiologic basis on which obstructive events could be considered a risk factor for acute cerebrovascular disease. The second phenomena is "central events" that are significantly lower in TIA and that clearly predominate in established stroke, especially those that are hemorrhagic. CA frequency decreases when the neurological disease is stable. In fact, we observed a reduction in AHI in the stable phase mainly as a result of a reduction in CA, thus suggesting the latter may be, at least in part, a consequence of the neurological lesion. Interestingly, CA reduction is observed not only for IS and HS, but also for TIA, suggesting that even when we have negative conventional neuroimaging, some kinds of injury act to alter the complex automatic breathing control. Although CSB has been described as more frequent in elderly people (23), the lower percentage in the follow-up study suggests that some cases could also be a consequence of the acute phase.

As in the Bassetti and colleagues (16) study, we found a lack of correlation between frequency and type of SRBD and neurological topography. In contrast to previous reports (9- 11), brainstem involvement does not result in a significantly higher frequency of SRBD, either central, obstructive, or CSB, although we have found a trend toward this circumstance.

From a prognostic point of view we did not find a correlation between neurologic outcome measures, according to changes in the Barthel Index and Canadian Scale, and the improvement from baseline in SRBD. Previous studies have suggested a trend toward a poorer short-term neurologic functional outcome (16).

In conclusion, both snoring and SRBD are frequent in patients with CVD, with the prevalence of SRBD two to three times higher than expected from the available epidemiological data in Spain. Although the relationship between both conditions remains intriguing, overall current data suggest that obstructive apnea could act as a risk factor, whereas CA and CSB could be the consequence of CVD. In addition, PRR appears to be an appropriate screening method for detecting SRBD in these patients. Nevertheless, some issues, such as the identification of SA as a long-term prognostic factor or the impact of treatment, remain to be settled.