Introduction
Obstructive sleep apnea is found in at least half of individuals who have experienced an acute stroke or transient ischemic attack (TIA) [
1‐
9]. Sleep apnea has been associated with a variety of adverse disease states and outcomes including hypertension, diabetes, ischemic stroke and transient ischemic attacks (TIA), incident coronary events and cardiovascular mortality, and all-cause mortality [
3,
10‐
14]. Estimates suggest, however, that as many as 70–80% of the patients with sleep apnea are neither diagnosed nor treated [
15]. The barriers to diagnosing and treating sleep apnea involve patient, provider, and system issues, including access to sleep laboratory-based testing [
15].
The American Heart Association/American Stroke Association secondary stroke prevention guidelines recommend that patients with ischemic stroke and TIA receive screening for sleep apnea and treatment because of data suggesting that post-cerebrovascular event patient outcomes are improved with treatment of sleep apnea [
16]. However, this guideline recommendation has not been adopted widely. For example, a recent report found that only 6% of the eligible patients received polysomnography in the year after an index cerebrovascular event [
17].
The “Diagnosis and Treatment of Sleep Apnea in Cerebrovascular Disease” (
GoToSleep) study evaluated a strategy to improve the diagnosis and treatment of sleep apnea among veterans with chronic ischemic stroke or TIA who also have hypertension [
18]. Specifically, the
GoToSleep study was designed to overcome some of the barriers that existed within the Veterans Health Administration (VHA) to the timely diagnosis and treatment of sleep apnea (e.g., limited mobility, travel distance) by using home-based ambulatory polysomnography and auto-titrating continuous positive airway pressure (CPAP) to reduce the reliance on laboratory-based sleep studies. The primary objectives of the
GoToSleep study were to determine whether a diagnostic and therapeutic intervention strategy among veterans with chronic cerebrovascular disease and hypertension could improve: (1) the detection of sleep apnea; (2) the treatment for obstructive sleep apnea; and (3) the control of hypertension.
Design and methods
The
GoToSleep study was a randomized controlled strategy trial conducted at two geographically distant Veterans Health Affairs (VHA) medical centers. The methods of this study have been described in detail elsewhere [
18]. Briefly, patients with chronic cerebrovascular disease and hypertension were randomly assigned to an intervention group or a usual care control group. The randomization was stratified by medical center, baseline blood pressure, and risk of sleep apnea from the Berlin Questionnaire [
19]. This study (NCT00984308) received institutional review board approval; patients provided written informed consent.
Chronic cerebrovascular disease and hypertension patient definition
Patients were included if they had a history of ischemic stroke or TIA
and either a diagnosis of hypertension or a blood pressure of ≥140/≥90 mmHg [
20]. This study was designed prior to the revision of the American Heart Association/American Stroke Association definition of TIA, and employed the following definitions: an ischemic stroke is a persistent focal neurological deficit of presumed ischemic origin lasting more than 24 h [
21] and a TIA is a focal neurological deficit of presumed ischemic origin lasting less than 24 h [
22]. Brain imaging was not required for the diagnosis of stroke or TIA. Patients were recruited at least 30 days after their most recent stroke and at any point after their most recent TIA.
Measurements
Patient interviews and medical record reviews were conducted to obtain information about demographics, comorbidities, medication use, and symptoms. Patients were examined to measure heart rate, neck circumference, height, and weight. The Berlin Questionnaire was used to classify patients according to “high” versus “low” risk for sleep apnea [
19]. Daytime sleepiness was assessed using the Epworth Sleepiness Scale (ESS) with a score of ≥10 indicating excessive daytime sleepiness [
23,
24]. Neurological functioning was measured with the National Institutes of Health Stroke Scale (NIHSS) [
25,
26].
Blood pressure and antihypertensive medications
Blood pressure measurements were taken at baseline and at the end of the study with a 24-h ambulatory monitor. Blood pressure data were classified as either sleeping or waking on the basis of actigraphy data or sleep diary information for patients with incomplete actigraphy data. Nocturnal blood pressure dipping patterns were assessed by comparing mean systolic blood pressure during sleep with mean systolic pressure during the wakeful state [
27]. Antihypertensive medications were not changed as part of the study but were recorded at the time of the blood pressure measurements. The defined daily dose (DDD) of all antihypertensive medications was calculated using the World Health Organization (WHO) DDD methodology which allows for comparison across antihypertensive regimens [
28]. The medication-adjusted mean systolic blood pressure was calculated from the 24-h mean systolic blood pressure measurement and the patient’s DDD as follows: medication-adjusted systolic blood pressure (SBP) = [mean systolic blood pressure in mmHg] + [the patient’s DDD × (8.0 mmHg)] [
18].
Polysomnography
Intervention patients received full unattended (type 2) polysomnography at baseline (Safiro, Compumedics, Victoria, Australia). Control patients received polysomnography at the end of the 1-year study period. Nearly all of the patients received the polysomnography in their homes. Full unattended polysomnography provided data on electroencephalogram (EEG), electrooculogram, chin electromyogram, thoracic and abdominal displacement, airflow, finger pulse oximeter, electrocardiogram, body position, leg movement, and audio recording of the participant’s snoring. The polysomnographic data were read centrally and scored according to standard criteria [
29]. An apnea-hypopnea index (AHI) was calculated as the number of respiratory events per hour of sleep [
29‐
31]. Patients with an AHI ≥5 events/h were diagnosed as having sleep apnea and were offered auto-titrating CPAP [
30]. Patients classified as having central sleep apnea or Cheyne-Stokes respiration were continued in the study but were referred for adaptive servoventilation (ASV). Patients with oxygen desaturations by polysomnography received one night of oximetry while using CPAP; patients with persistent desaturations despite auto-titrating CPAP were provided supplemental nocturnal oxygen via the CPAP machine.
CPAP adherence support
Our approach to improving CPAP adherence consisted of early intensive education and support, followed by regular and ongoing contact with patients. During the in-home visit where the results of the baseline polysomnography were reviewed, staff members provided patients with information about sleep apnea and CPAP. During the first month after delivery of the CPAP machine, the patient was visited each of the first 2 days and then weekly for the first month; thereafter, patients received monthly telephone calls for the remainder of the 1-year study period. An in-home visit was scheduled after 6 months of CPAP use and also at the end of the study. During the in-home visits, staff downloaded pressure, residual AHI, air leak, and compliance data from the CPAP machine and reviewed these data with the patients. Staff members were also available to patients for ad hoc issues related to difficulties with mask fit or other technical problems with the CPAP machines.
Usual care
Control patients received usual care as directed by their physicians, including care of their vascular risk factors such as hypertension and referral for usual care sleep apnea testing. Control patients in the GoToSleep study were neither prohibited from nor specifically encouraged to receive polysomnography as part of their usual care. Control patients received unattended polysomnography at the end of the study period.
Outcomes
Sleep apnea diagnosis
The proportion of intervention and control patients with a diagnosis of obstructive or central sleep apnea was calculated from the unattended polysomnography performed in the intervention group and from any polysomnography that was performed as part of usual care among control patients. The primary outcome related to sleep apnea detection was the prevalence of undiagnosed sleep apnea in the control group which was determined using polysomnography data obtained as part of the research protocol at the end of the study.
Obstructive sleep apnea treatment
CPAP adherence was classified as “none” if the patient refused to take receipt of a CPAP machine or did not use the machine at all, “poor” if the nights of usage was ≤10% of total nights available, “some” if the total nights used >10% of total nights available or the cumulative hours of use was <4 h per night times 70% of the total nights available for use, and “excellent” if the total cumulative hours of use was ≥4 h per night times 70% of the total nights available for use [
32]. The primary outcome related to obstructive sleep apnea treatment was the proportion of intervention and control patients with obstructive sleep apnea with excellent CPAP adherence.
Hypertension control
The primary hypertension outcome was the final mean 24-h SBP, adjusted for baseline mean 24-h SBP. Our secondary hypertension analyses included the final medication-adjusted SBP, the final DDD, and the proportion of patients with a nocturnal decline (“dipping”) in systolic blood pressure of ≥10% [
27].
Sample size
The GoToSleep trial was designed to recruit 318 patients to have at least 80% power to detect the following pre-specified differences in the three primary aims: 60% of the intervention patients compared with 10% of the control patients would be diagnosed with sleep apnea; 50% of the intervention patients diagnosed with sleep apnea as part of the study protocol, compared with 15% of the control patients diagnosed with sleep apnea as part of usual care, would use CPAP therapy; and the 1-year mean systolic blood pressure would be 125 mmHg with a standard deviation of 20 mmHg in the intervention group compared with 135 mmHg with a standard deviation of 20 mmHg in the control group. The sample size estimates assumed a 10% loss to follow-up in both the intervention and the usual care arms.
Analyses
The primary analyses were designed to be intention-to-treat comparisons of the outcomes for patients in the intervention versus control groups. For the analysis of diagnosis rates, all patients were analyzed according to the group to which they had been randomly assigned regardless of subject study completion. Ninety five percent confidence intervals were calculated around the undiagnosed sleep apnea rate obtained from the control group. For the analysis of treatment rates, CPAP adherence was assessed among patients with obstructive sleep apnea. The comparison of final mean 24-h SBP was not restricted to patients with sleep apnea, but rather was conducted across all patients with final ambulatory blood pressure data, regardless of sleep study results. A secondary hypertension control analysis was pre-specified to be conducted among sleep apnea patients in the intervention group with or without excellent CPAP adherence, to examine the effect of CPAP use on antihypertensive outcomes.
Baseline characteristics were compared between intervention and control groups using Fisher’s exact test, chi-squared test, 2-sample t test, or Wilcoxon 2-sample tests. Chi-squared tests were used to test for group differences in diagnosis and treatment rates. An analysis of covariance (ANCOVA) model, with terms for site, group, and baseline measure, was used to test for differences in mean final 24-h SBP. Residual plots were examined for possible violation of model assumptions. The final medication-adjusted SBP was analyzed similarly. Final DDD was rank transformed because of skewness and was analyzed similarly, including a squared baseline DDD term. Summary statistics of change in measures are presented but were not modeled or statistically tested, because the analyses of the final measurements included adjustment for baseline measurements. A logistic model with terms for site and group was used to estimate the effect of the intervention on the odds of having nocturnal decline of ≥10%. A cumulative logistic model was used for the ordinal outcome of nocturnal dipping category and modeled the probability of having greater nighttime dipping. The proportional odds assumption was verified. The secondary analyses of hypertension control measures by CPAP adherence used similar models with the term for treatment group replaced by a variable for CPAP adherence (excellent versus not). No corrections for multiple comparisons were made, because we pre-specified a single primary outcome for the three primary aims. SAS® version 9.2 (Cary, NC) was used for all analyses.
Discussion
These results demonstrate that sleep apnea is very common among patients with chronic cerebrovascular disease and hypertension; the observed prevalence of 120/194 (62%) is similar to rates observed in studies that have primarily focused on acute and subacute cerebrovascular disease populations [
1‐
9]. These results also demonstrate that a strategy of diagnosing and treating sleep apnea in the homes of patients with chronic cerebrovascular disease and hypertension can lead to substantial improvements in the detection of sleep apnea and in treatment with CPAP.
Usual care identified less than one in ten patients who actually had sleep apnea, a rate that leaves considerable room for improvement. Moreover, the sleep apnea that was detected in this study population included some patients with very high AHIs indicative of severe disease; therefore, it is incorrect to assume that only mild sleep apnea remained undetected by usual care. Only 64/225 (28%) participants in this study had excessive daytime sleepiness according to the Epworth Sleepiness Scale, indicating that “typical” patient characteristics that might alert primary care providers to the need for performing polysomnography were uncommon. Specifically, neither the Epworth Sleepiness Scale nor the Berlin Questionnaire was useful in identifying patients with sleep apnea in this study. It is unknown whether aging or cerebrovascular disease may contribute to a reduced perception of drowsiness. In the absence of validated and reliable methods of screening for polysomnography in the post-stroke and post-TIA populations [
33,
34], it is reasonable to use polysomnography to identify sleep apnea given the high underlying prevalence of sleep apnea in this population and the known consequences of untreated sleep apnea.
Because testing for sleep apnea was so uncommon in the usual care group in this study, the number of patients who were eligible for treatment as part of usual care was very small; this limited our assessment of CPAP adherence in usual care versus the intervention groups. None of the usual care patients used CPAP for the recommended ≥4 h per night for ≥70% of the nights, whereas 38.6% (22/57) of the intervention patients achieved this degree of adherence. The two VHA medical centers differed in their approach to usual care. One provided CPAP mask fitting and education via a contract with a home health agency, and the other provided these services via an on-site durable medical equipment (DME) CPAP clinic. Both sites provided a visit with a trained sleep staff member after 30 days, and as-needed support related to mask fit or equipment functioning. In general, usual care at these two sites is similar to sleep care received in the non-VHA setting. The intervention patients, in contrast, had both more frequent contact with staff and the content of the contact was different from usual care. In particular, the staff worked with patients to identify salient motivators for using CPAP (e.g., having more energy to play with grandchildren) and salient barriers inhibiting CPAP use (e.g., nasal stuffiness). As the patients began to use the CPAP, their salient positives and negatives might have changed, but the staff consistently sought to identify methods to overcome those barriers and remind patients of their individual reasons for trying the CPAP. In this way, the content of the visits was not only focused on technical issues related to mask fit, leak, and humidification but also on motivators to try CPAP and to persist over time. In this study, CPAP adherence was not associated with physical impairments (data not shown). Even patients with hemiparesis and blindness were able to achieve excellent CPAP adherence with adequate support from study staff and from caregivers.
This study did not identify a difference in blood pressure management for intervention versus control patients. Three key reasons likely contributed to this negative finding. First, the sample size of this trial was designed to detect a 10-mmHg difference in mean 24-h systolic blood pressure and this study did not achieve the anticipated sample size (
N = 225 versus
n = 318). Our inability to achieve the targeted sample size was principally due to delays in hiring and training needed staff within the VHA human resource constraints and because patients who were included in the study lived further away from the medical center than in our pilot work; hence, the resources needed (e.g., staff time, mileage reimbursement) to provide in-home polysomnography and CPAP support were greater than expected. Moreover, several meta-analyses have been published since the time of designing this trial which suggest that the actual difference in systolic blood pressure that can be expected with CPAP ranges from −1.4 to −7.2 mmHg [
35‐
38], with most of the estimates closer to the 1.4 mmHg end of the range. A sample size of 167 patients with sleep apnea in each group would have been required to demonstrate a 5-mmHg difference in mean 24-h systolic blood pressure among excellent versus not excellent CPAP users. A much larger sample size would have been required to identify a statistically significant relationship between intervention versus control group patients with a systolic blood pressure change in the 1.5- to 2.0-mmHg range [
35‐
38]. Second, this population of veterans with hypertension was treated in the VHA primary care setting, and the overwhelming majority was taking antihypertensive medications and their blood pressure was generally very well controlled. The VHA system has placed an emphasis on hypertension care quality, and blood pressure management is an important performance measure within the VHA primary care system nationwide. A recent study of the effect of CPAP on blood pressure among patients with resistant hypertension demonstrated a 3.1-mmHg difference in 24-h systolic blood pressure, but in that study, baseline blood pressures were approximately 10 mmHg higher than in our population (24-h systolic blood pressure of 144.2 mmHg) [
39]. Third, we appreciate that hypertension may be attributed to many causes. Because we did not seek to identify etiologies of hypertension, we were not able to identify specific patients in whom sleep apnea may have been the primary etiology of the hypertension.
The primary limitation of the
GoToSleep trial is its focus on the veteran population, which may limit generalizability to women as there were relatively few female patients in this study. The patients in this trial all had a history of cerebrovascular disease and hypertension, so these results should not be generalized to the acute cerebrovascular event setting—where sleep apnea may be more prevalent and where hemodynamic fluctuations may be more prominent [
1‐
9,
13,
40,
41]. Another potential limitation is that although the NIHSS in this population ranged from 0 to 19, the mean values of approximately 2 suggest that the population had relatively mild stroke severity. Therefore, these results may not be generalizable to a population of chronic cerebrovascular disease patients with more prominent neurological impairments.
Several randomized clinical trials and observational cohort studies have examined the use of CPAP in the cerebrovascular disease population for the purpose of improving post-stroke or post-TIA outcomes; in general, these studies have focused on patients in the acute or subacute event setting. The randomized controlled trials have all been relatively small, with sample sizes insufficient to identify changes in outcomes associated with treatment [
40,
42,
43]. The largest of the randomized trials was conducted by Parra et al. [
44] and followed (
N = 126) acute stroke patients with sleep apnea over a 2-year period. The stroke rate was similar in both groups: 3/57 (5.3%) fixed pressure CPAP and 3/69 (4.3%) control (
p = 1.0); however, the mean time from stroke onset to the first cardiovascular event was longer in the CPAP group: 15 months in the CPAP patients versus 8 months in the control patients (
p = 0.044). Two randomized controlled trials evaluated the use of CPAP in subacute stroke patients and reported mixed results, finding improvements in several outcome domains for the CPAP group (e.g., depression) versus usual care, and also no statistically significant differences in several outcomes [
45,
46].
In conclusion, this study identified both a high prevalence (62%) of sleep apnea in a chronic post-stroke and post-TIA population with hypertension and a very low sleep apnea detection rate (9%) for patients in the usual care group. This study further demonstrated that in-home diagnostic and treatment strategies can be effectively deployed to supplement sleep medicine services that are provided in a sleep laboratory. Although no published data support the use of CPAP among chronic (as opposed to acute) post-cerebrovascular event patients specifically for the reduction of vascular events, there is considerable evidence to support the treatment of sleep apnea in general to improve a variety of patient outcomes [
35,
47]. Our findings add to the growing literature that suggests that the effect of CPAP on blood pressure is modest and likely only present in patients with uncontrolled hypertension. The study results provide a rationale for using polysomnography to identify sleep apnea among all patients post-stroke or post-TIA given its high prevalence, and suggest that support for CPAP adherence should persist over time, because many patients in the intervention group in this study continued to adapt to CPAP well beyond the first month of receiving a CPAP machine.