Nocturnal Hypoxemia Is Associated with White Matter Hyperintensities in Patients with a Minor Stroke or Transient Ischemic Attack

Abstract

Study Objectives:

Obstructive sleep apnea (OSA) is a risk factor for stroke, which is modulated by accompanying nocturnal hypoxemia. White matter hyperintensities (WMH) share many of the same risk factors as stroke. The purpose of this study was to investigate whether OSA and nocturnal hypoxemia are associated with white matter disease in patients with minor stroke and transient ischemic attack.

Methods:

Patients with minor stroke or TIA were recruited. Level 3 diagnostic sleep testing was used to diagnose OSA and quantify nocturnal hypoxemia. Significant OSA was defined as respiratory disturbance index ≥ 15, and nocturnal hypoxemia was defined as oxyhemoglobin saturation < 90% for ≥ 12% of total monitoring time. WMH were assessed and quantified on FLAIR MRI. The volume of WMH was compared between those with and without significant OSA and between those with and without nocturnal hypoxemia.

Results:

One hundred nine patients were included. Thirty-four (31%) had OSA and 37 (34%) had nocturnal hypoxemia. Total WMH volume was significantly greater in the OSA than in the non-OSA groups (p = 0.04). WMH volume was also significantly higher in the hypoxic than the non-hypoxic groups (p = 0.001). Mutivariable analysis with adjustment for age, hypertension, and diabetes showed that nocturnal hypoxemia was independently associated with WMH volume (p = 0.03) but OSA was not (p = 0.29).

Conclusions:

We conclude that nocturnal hypoxemia, predominantly related to OSA, is independently associated with WMH in patients who present with minor ischemic stroke and TIA and may contribute to its pathogenesis.

Citation:

Patel SK, Hanly PJ, Smith EE, Chan W, Coutts SB. Nocturnal hypoxemia is associated with white matter hyperintensities in patients with a minor stroke or transient ischemic attack. J Clin Sleep Med 2015;11(12):1417–1424.

Obstructive sleep apnea (OSA) increases the risk of vascular disease including hypertension, coronary artery disease, and stroke.^1–3 It is characterized by repetitive apneas due to airway closure that are associated with intermittent hypoxemia, which is thought to modulate the increased risk of vascular disease.^4,5 OSA may occur in up to 20% of the general population, 72% of those following an established stroke, and 62% of those who have suffered a transient ischemic attack or minor stroke.^6–10 Furthermore, OSA may be asymptomatic in many stroke patients.¹⁰

White matter hyperintensities (WMH) of presumed vascular origin in the brain represent areas of abnormal white matter with ischemic demyelination^11–13 and are diagnosed by either magnetic resonance imaging (MRI) or computed tomography (CT). Although WMH can be associated with several medical conditions including hypertension, aging, and diabetes mellitus, they are also associated with an increased risk of stroke.^14–18 Since the etiology of WMH is not fully explained by conventional risk factors, we postulated that OSA with associated nocturnal hypoxemia may also contribute to its pathogenesis. Previous publications that have evaluated the relationship between OSA and WMH have yielded conflicting results, with some studies reporting an increased risk of WMH^19,20 while others have not.^4,21 Potential reasons for these discordant results include the subjective visual assessment of WMH volumes and lack of correction for total brain volume, small sample sizes, inaccurate assessment of OSA and nocturnal hypoxemia, and poor adjustment for confounding comorbidities.

BRIEF SUMMARY

Current Knowledge/Study Rationale: The etiology of white matter hyperintensities (WMH), which represent areas of ischemic demyelination in the brain, is not fully understood. Nocturnal hypoxemia associated with obstructive sleep apnea (OSA) increases the risk of stroke and may also contribute to the pathogenesis of WMH.

Study Impact: WMH were associated with nocturnal hypoxemia in patients who presented with a TIA or minor stroke. Treatment of OSA may reduce the risk of recurrent neurological disease in this vulnerable patient population.

This study was designed to address the limitations of previous studies. We hypothesized that patients who present with transient ischemic attack (TIA) or minor stroke and who also have OSA and/or nocturnal hypoxemia have a higher volume of WMH that is independent of other comorbidities.

METHODS

Patient Recruitment

This was a sub-study of the prospective CATCH (CT And MRI in the Triage of TIA and minor Cerebrovascular events to identify High-risk patients) study.²² We recruited minor stroke (National Institutes of Health Stroke Scale (NIHSS) < 4) and high-risk TIA (motor or speech symptoms lasting ≥ 5 min) patients assessed within 24 h of symptom onset and who had no preexisting disability (premorbid modified Rankin score [mRS] 0 or 1). Patients had a brain MRI scan within 48 h of symptom onset and completed nocturnal monitoring for sleep apnea (described below). Patients were excluded if they were receiving treatment with continuous positive airway pressure (CPAP) or oxygen therapy, both of which would have confounded diagnostic sleep testing. All patients received routine medical management of their stroke or TIA. The research study was approved by the University of Calgary Conjoint Health Research Ethics Board, and all patients provided written informed consent.

Sleep Questionnaire and Nocturnal Cardiopulmonary Monitoring for Sleep Apnea

Patients completed a questionnaire which included the Ep-worth Sleepiness Scale,²³ sleep history, past medical history, and current medications. Patients were instructed how to use a level 3 sleep monitor (Remmers Sleep Recorder Model 4.2, Saga Tech Electronics, Calgary, AB, Canada), which was used to diagnose OSA and to quantify the severity of nocturnal hypoxemia. This device consists of an oximeter to record oxyhemoglobin saturation (SpO₂) and heart rate variability, a pressure transducer to record nasal flow, a microphone to record snoring, and a body position sensor. The Remmers Sleep Recorder has been validated by comparison to attended polysomnography.^24,25 The respiratory disturbance index is calculated based on the number of episodes of oxyhemoglobin desaturation > 4% divided by the duration of the recording. We performed nocturnal monitoring on 2 consecutive nights and averaged the results. If only one night was completed, the results from the single night were used. If only one night of monitoring was performed and the study was not technically adequate, the patient was asked to repeat the study and was excluded if they refused. Our criteria for a technically adequate study were: (1) > 5 h of interpretable recordings of airflow and oximetry, and (2) self-reported sleep for > 75% of the monitoring time. The raw data were reviewed by a sleep medicine physician (PJH), who confirmed that the estimated RDI was accurate and determined whether apnea was central (Cheyne-Stokes respiration [CSR]) or OSA, based on the morphology of the airflow recordings. Specifically, nasal pressure recordings with a characteristic crescendo/decrescendo pattern and no evidence of airflow limitation were classified as CSR, whereas recordings without a crescendo/decrescendo pattern and with airflow limitation were classified as OSA. Finally, 2 measurements of SpO₂ were used to quantify the severity of nocturnal hypoxemia: (1) mean SpO₂ (average SpO₂ during the entire recording) and (2) the duration of SpO₂ < 90% during the recording.

Assessment of OSA and Hypoxemia

Patients were stratified based on their RDI and severity of nocturnal hypoxemia. These criteria were used independently of each other. Patients with RDI ≥ 15 were defined as the OSA group (as this is consistent with clinically significant OSA),²⁶ and those with RDI < 15 were defined as the non-OSA group. Patients with oxyhemoglobin saturation < 90% for ≥ 12% of total monitoring time were defined as the hypoxemic group, as this level of hypoxemia has been correlated with cardiovascular outcomes in previous studies,²⁷ and those with oxyhemoglobin saturation < 90% for < 12% of total monitoring time were defined as the non-hypoxemic group. Patients with CSR were excluded from analysis as decided a priori in the study design. The objective of the study was to investigate whether OSA and associated hypoxemia predispose patients with vascular disease to develop WMH as a complication of their sleep disordered breathing. Cheyne-Stokes respiration is more likely to be a consequence than a cause of cardiovascular or cerebrovascular disease, and consequently inclusion of patients with CSR may confound interpretation of the results.

Imaging and White Matter Lesion Volume Measurement

All patients received a multimodal 3T MRI brain scan within 48 h of stroke or TIA symptom onset. The 3T MRI scanner (Signa; GE Medical Systems) consisted of high performance gradients (40 mT/m; 184 μs rise time) and a standard quadrature head coil. Patients received sagittal T1 localizer, axial diffusion- tensor imaging (DTI) (B = 1000), axial 3D pre and post gadolinium time-of-flight MR angiography, axial T2 spoiled gradient-recalled echo (SPGR), and axial fluid-attenuated inversion recovery (FLAIR). The FLAIR sequence (TE/TR = 140/9000 ms, slice thickness = 5.0 mm) was used to investigate for presence of white matter lesions (Figure 1). WMH were defined according to consensus standards²⁸ and measured quantitatively using a computer software tool called Quantomo (Version 1.0; Cybertrials, Inc, Calgary, Canada).²⁹ Quantomo was developed using 3-D threshold-based region growing segmentation algorithm incorporated from the open-source Insight Segmentation and Registration Toolkit (ITK; National Library of Medicine, Bethesda, MD). The user interface was developed with Xcode using Objective-C and Cocoa (Apple Inc., Cupertino, CA). In this method, WMHs are segmented by placing a single seed within each visible WMH, with identification of the WMH borders by the automated computer algorithm. The WMH borders can be adjusted as necessary by the user to correct errors such as misidentification of skull as WMH. WMH volumes were quantified using the spatial dimensions of the voxels in each MRI slice. Acute DWI lesions that were also visible on FLAIR were not included in these measurements. A trained medical student blind to all clinical data made all measurements. A test of inter-rater reliability on 10 scans showed an excellent correlation (intra-class correlation coefficient 0.99) between WMH volumes generated by the medical student compared to a neurologist with 10 years' experience in research in cerebral small vessel disease.

Figure 1. T2 FLAIR sequence shown in Quantomo.

(A) FLAIR before measurement. (B) FLAIR after WMH measurement. Purple regions represent WMH regions that were automatically selected by the computer, red regions represent manual traced regions. WMH regions that were overestimated were blocked out using the blocker tool (green region).

Statistical Analysis

All statistical analyses were performed using SAS version 9.3 (SAS Institute, Cary, NC, USA). The results were expressed as the mean ± standard deviation or as median (interquartile range) where appropriate. Pearson correlation coefficient tests were used for continuous variables, Wilcoxon rank-sum tests for dichotomous variables, and the Kruskal-Wallis test for categorical variables. A p value < 0.05 was considered statistically significant. Percent WMH was first logarithmically transformed to achieve a normal distribution. To account for differences in baseline brain size, WMH were expressed as a percentage of total brain volume. Because the percent WMH was right skewed, we logarithmically transformed it to an approximately normal distribution for univariate analyses and multivariate modeling. Multivariable linear regression was used to determine whether OSA or hypoxemia were independent predictors of percent WMH. Because OSA and nocturnal hypoxemia represent inter-correlated aspects of abnormal sleep physiology after stroke, we built separate models for each risk factor rather than including both OSA and hypoxemia in the same model. Models were also adjusted for other known predictors of WMH including age, hypertension, and diabetes that were available in the dataset.

RESULTS

Study Population

One hundred twenty-four of the 510 patients in the CATCH study were enrolled. We excluded 5 patients who were unable to use the Remmers Sleep recorder and 10 patients with CSR. The remaining 109 patients comprised the final study population.

Of the 109 patients, 78 (72%) were male, mean age was 64.6 ± 12.7 years, 18% were diabetic, 64% hypertensive, and 15% were current smokers. Baseline NIHSS total was 0 in 35 (32%), 1 in 21 (19%), 2 in 39 (31%), and 3 in 19 (17%) patients. Patients underwent their level 3 sleep test a median of 76 days (25th percentile 39 days, 75th percentile 110 days) after their stroke or TIA event. Thirty-three (30%) patients had no OSA (RDI < 5), 42 (39%) had mild OSA (5 ≤ RDI < 15), 18 (17%) had moderate OSA (15 ≤ RDI ≤ 30), and 16 (15%) had severe OSA (RDI > 30). Thirty-seven (34%) patients were found to have nocturnal hypoxemia as previously defined. Higher RDI scores were moderately correlated with lower mean SpO₂ (r = −0.41, p < 0.001) and higher duration of time with SpO₂ < 90% (r = 0.54, p < 0.001). Median WMH volume was 4.23 mL (interquartile range, IQR, 1.77–9.37 mL) and median WMH volume as a percentage of total brain volume was 0.33% (IQR 0.14% to 0.71%). As expected, WMH volume as a percentage of brain volume was correlated with age (r = 0.69, p < 0.001) and was higher in patients with a history of hyper-tension (p < 0.001) and diabetes (p = 0.01)

Comparison between OSA and non-OSA Groups

Patients with OSA were older, heavier, more frequently male, and had more severe nocturnal hypoxemia (Table 1). The prevalence of hypertension was high but not significantly different between the groups. The prevalence of diabetes was higher in those with OSA but did not reach statistical significance. Total WMH volume and WMH volume corrected for brain volume were significantly greater in the OSA group.

Table 1.

Characteristics of patients with and without OSA.

Comparison between Hypoxemic and Non-Hypoxemic Groups

Thirty-seven (34%) patients had nocturnal hypoxemia (Table 2) and were older, heavier, and more frequently male. Despite a higher prevalence of sleep apnea in the hypoxemic group, symptoms of sleep apnea including daytime sleepiness (reflected by the ESS), and snoring were not significantly different. WMH volumes (total volume or total as a percentage of brain volume) were significantly higher in the hypoxemic group but acute stroke lesions on diffusion weighted imaging (DWI) and total brain volume were similar in both groups.

Table 2.

Characteristics of patients with and without hypoxemia.

Comparison between Hypoxemic Patients with and without OSA

There were no differences in patient demographics, features of OSA, comorbidities, or brain volumes between hypoxemic patients with OSA and hypoxemic patients without OSA (Table 3).

Table 3.

Characteristics of hypoxemic patients with and without OSA.

Multivariable Analysis

Multivariable models used to evaluate the association of OSA and nocturnal hypoxemia with WMH volume are shown in Table 4. Although OSA in isolation was significantly associated with WMH volume, this was no longer significant when adjusted for other vascular risk factors (hypertension and diabetes) and age. In contrast, patients with nocturnal hypoxemia had 44% higher WMH volume (95% CI 13% to 90%, p = 0.009) after adjustment for vascular risk factors and age.

Table 4.

Multivariable models of the association between OSA and hypoxemia and MRI white matter hyperintensity (WMH) volume.

DISCUSSION

We investigated the relationship of OSA and nocturnal hypoxemia with WMH by measuring both in an objective and blinded fashion in patients who had suffered a minor stroke or TIA. The median WMH burden in this population (4.23 cm³) was significantly higher than what has been reported in a recent population-based study of the same age (2.2 cm³),³⁰ reflecting the increased prevalence of cerebrovascular disease in our patient cohort. Although we found that OSA was associated with an increased risk of WMH, this association lost statistical significance following adjustment for other vascular risk factors and age. In contrast, the association between nocturnal hypoxemia and WMH was significant and independent of hypertension, diabetes, and age.

Previous investigations of the relationship between OSA and WMH has shown mixed results with some finding a strong association,^20,31 while others have not.^4,32–34 These discrepancies are likely due to poor control for vascular risk factors and age,³⁵ case mix differences (e.g. general population versus cohorts with large strokes), no assessment of hypoxemia, and diversity of methods and rating scales for quantifying WMH. One of the largest studies to find an independent association between OSA and WMH evaluated middle-aged to older aged adults in the general population.³⁶ Our population was older and had overt cerebrovascular disease, which may explain the lack of an association between OSA and WMH when both age and vascular risk factors are controlled. Further, previous studies that showed a significant correlation between OSA and WMH recruited subjects with severe OSA or patients with major stroke.³⁷ In contrast, our study population consisted entirely of patients who had suffered only a minor stroke or TIA and their prevalence of severe OSA was relatively low, which may have hindered our ability to find an association between OSA and WMH.

We found evidence of a significant association between hypoxemia and WMH even after adjustment for vascular risk factors. Most of the hypoxemia was due to OSA since the prevalence of lung disease was low and the nocturnal oximetry profiles did not raise any suspicion of nocturnal hypoventilation. Obstructive sleep apnea is known to increase sympathetic nervous system activity, oxidative stress, and systemic inflammation, ultimately leading to vascular complications including hypertension, coronary artery disease,³⁸ and stroke.³ However, oxidative stress may have a direct effect on the brain independent of these vascular complications, since it has been shown to contribute to the pathogenesis of demyelination and tissue injury in multiple sclerosis.³⁹ Further, an animal model of OSA has demonstrated that chronic intermittent hypoxia leads to cell damage and death in the cortex.⁴⁰ Recently, white matter damage has been correlated with indices of systemic inflammation in patients with OSA, and a pathological study of aging brains has implicated hypoxia in the pathogenesis of WMH.^41,42 Consequently, there are several mechanisms through which OSA may increase the risk of WMH. These possibilities are further supported by a recent study of two elderly patient cohorts, one in the community and the other recruited from patients who had suffered a “non-disabling stroke,” which highlighted that the etiology of WMH was predominantly related to non-vascular factors that remain to be identified.⁴³

What are the potential clinical implications of these results? Although WMH may be an incidental finding in asymptomatic individuals, they increase with age and have been associated with cognitive impairment and dementia.⁴² Cognitive impairment is an important and common complication of OSA that may persist in some patients after effective treatment with CPAP.⁴⁴ Further, OSA may be a risk factor for progression of mild cognitive impairment to dementia and poor sleep been implicated in the pathogenesis of Alzheimer disease.⁴⁵ These findings raise the possibility that WMH may modulate some of the neurological complications of OSA.

It is important to distinguish between WMH and white matter integrity, which has also been evaluated in patients with OSA.⁴⁶ The degree of white matter integrity can be inferred by MRI diffusion tensor imaging (DTI), which measures the degree and directionality of water proton diffusivity.⁴⁷ In the presence of highly organized white matter tracts, water protons are freer to diffuse in parallel to the white matter fibres than perpendicular to the fibres. This directionality of diffusion can be measured mathematically using constructs such as fractional anisotropy. When white matter microarchitecture is disrupted, fractional anisotropy tends to decrease and the total mean diffusivity tends to increase. Fractional anisotropy is decreased in regions of WMH, and furthermore is decreased in regions of normal-appearing white matter that become WMH in the future.⁴⁸ Therefore, decreased white matter integrity precedes WMH and may be a precursor to WMH. In one trial, CPAP therapy restored more normal white matter integrity in OSA patients, suggesting that CPAP might also prevent WMH although this has not yet been evaluated in clinical trials.

We chose to use level 3 diagnostic sleep monitoring rather than polysomnography (PSG) for several reasons. Firstly, we have used level 3 monitoring extensively over the past several years both for clinical purposes and research studies. As such, we have considerable experience with this methodology, including its strengths and limitations. The main limitations of level 3 monitoring include underestimation of sleep disordered breathing and nocturnal hypoxemia due to prolonged wakefulness, loss of respiratory signals due to monitoring equipment being disconnected, and poor signal quality due to the fact that the test is not attended by technical staff. We believe that the criteria we used for technically acceptable monitoring curtailed the potential impact of these limitations on our results. Secondly, we felt that performing an ambulatory sleep study in the home would be more acceptable to patients and would improve our recruitment. Finally, provision of ambulatory monitoring enabled us to avoid the delay we would have encountered with limited access to PSG.

The strengths of our study include the semi-automated volume measurement of WMH, which provided a quantitative and objective measurement of WMH. Previous studies used visual rating scales, which are subjective and generally less reliable than quantitative volumetric measurements. Our study also has limitations, which should be addressed. Firstly, although our sample size is larger than most of the observational studies to date, our relatively small study population limited our analysis and may have precluded additional findings. Secondly, our cohort was not representative of the general population since all patients had a recent TIA or minor stroke. Although this limits the generalizability of our findings, our results are highly relevant for this population, which has a high risk of recurrent cerebrovascular disease.⁴⁹ Finally, we have shown an association between nocturnal hypoxemia and white matter disease but cannot infer causality from our data. We do not believe that OSA was secondary to neurological disease, since the latter was mild and had resolved in the majority of patients by the time the sleep study was performed.

In summary, we found that nocturnal hypoxemia, which was predominantly due to OSA, was associated with WMH independent of other risk factors such as hypertension, diabetes, and age. This implies that nocturnal hypoxemia, which is often clinically silent, contributes to the pathogenesis of WMH, and should be considered as part of the spectrum of cerebrovascular complications of OSA. Future research is required to determine the physiological and cognitive impacts of WMH in patients who present with TIA and minor stroke, and whether treatment of OSA and nocturnal hypoxemia improves clinical outcomes such as cognitive function and stroke recurrence.

DISCLOSURE STATEMENT

This study was funded by Canadian Institute of Health Research (CIHR) and a Pfizer Cardiovascular research award. Dr. Coutts, received salary support from Alberta-Innovates-Health Solutions and the Heart and Stroke Foundation of Canada's Distinguished Clinician Scientist award, supported in partnership with the Canadian Institute of Health Research (CIHR), Institute of Circulatory and Respiratory Health and AstraZeneca Canada Inc. Dr. Patel received funding from a Canadian Stroke Network summer studentship. Dr. Smith received salary support from Alberta-Innovates-Health solutions and the Canadian Institutes of Health Research. The other authors have indicated no financial conflicts of interest.

ABBREVIATIONS

BMI

body mass index

CATCH

CT and MRI in the triage of TIA and minor cerebrovascular events to identify high-risk patients

CPAP

continuous positive airway pressure

CSR

Cheyne-Stokes respiration

CT

computed tomography

DTI

diffusion tensor imaging

DWI

diffusion weighted imaging

ESS

Epworth Sleepiness Scale

FLAIR

fluid-attenuated inversion recovery

MRI

magnetic resonance imaging

mRS

modified Rankin score

NIHSS

National Institutes of Health Stroke Scale

OSA

obstructive sleep apnea

PSG

polysomnography

RDI

respiratory disturbance index

SPGR

spoiled gradient-recalled echo

SpO₂

oxyhemoglobin saturation

TIA

transient ischemic attack

WMH

white matter hyperintensities