Association Between Central Sleep Apnea And Left Ventricular Structure: the Multi-Ethnic Study of Atherosclerosis

Summary

We assessed whether the presence of central sleep apnea is associated with adverse left ventricular structural changes. We analyzed 1412 participants from the Multi-Ethnic Study of Atherosclerosis who underwent both overnight polysomnography and cardiac MRI. Subjects had been recruited 10 years earlier when free of cardiovascular disease. Our main exposure is the presence of central sleep apnea as defined by central apnea-hypopnea index ≥5 or presence of Cheyne Stokes Breathing. Outcome variables were left ventricular mass/height, left ventricular ejection fraction, and left ventricular mass/volume ratio. Multivariate linear regression models adjusted for age, gender, race, waist circumference, tobacco use, hypertension, and the obstructive apnea hypopnea index were fit for the outcomes. Of the 1412 participants, 27 (2%) individuals had central sleep apnea. After adjusting for covariates, presence of central sleep apnea was significantly associated with elevated left ventricular mass/volume ratio (β= 0.11±0.04 gm/mL, P=0.0071), an adverse cardiac finding signifying concentric remodeling.

Introduction

There is a high prevalence of central sleep apnea (CSA) in patients with comorbid heart failure (HF). CSA is associated with increased morbidity and mortality in HF patients (Khayat, Jarjoura, Porter et al. 2015; Nakamura, Asai, Kubota et al. 2015). What is less understood is whether in a general community sample, the presence of CSA is a marker for individuals with cardiac structural features such as subclinical left ventricular (LV) dysfunction. Further, recent evidence suggests that CSA may be present prior to development of clinically overt HF (Javaheri, Blackwell, Ancoli-Israel et al. 2016). Using rigorously obtained measures from cardiac magnetic resonance imaging (cMRI), we sought to address whether the presence of CSA on overnight polysomnography (PSG) is a marker for reduced LV function and adverse structural changes. We conducted an analysis of cross-sectional data from the Multi Ethnic Study of Atherosclerosis (MESA) to study the association of CSA with left ventricular structure and function in an ethnically diverse, community-based sample. We hypothesized that an incidental finding of CSA would be significantly associated with reduced LVEF and elevated LVMVR and LVHi.

Methods

Study Design and Population

MESA is a prospective multicenter study designed to detect subclinical markers of cardiovascular disease (CVD) in an ethnically diverse, population-based cohort. The study design has been published (Bild, Bluemke, Burke et al. 2002). Briefly, individuals aged 45–84, initially free of CVD, were recruited from 6 different US sites (Baltimore, MD; Chicago, IL, Forsyth County, NC; Los Angeles County, CA; Northern Manhattan, NY; and St. Paul, MN) and underwent five exams at two year intervals beginning in years 2000–2002. At exam 5 (2010–2013) participants underwent standardized anthropometric measurements (height, weight, waist circumference and resting blood pressure) and cMRI. Assessment of cMRI parameters was performed using CIM software (version 6.2, University of Auckland, New Zealand) (Young, Cowan, Thrupp, Hedley and Dell'Italia 2000) and the description of the cMRI methodology has been described (Bild, Bluemke, Burke et al. 2002). Also at Exam 5, participants other than those reporting regular use of oral devices, nocturnal oxygen, or nightly positive airway pressure devices were invited to participate in the MESA Sleep Ancillary Study. Fifteen channel, unattended, single night PSG (Somté PSG, Compumedics Ltd., Abbotsford, Victoria, Australia) and sleep questionnaire data were collected during an in-home examination; details on sleep methodology are provided in prior publications (Chen, Wang, Zee et al. 2014; Javaheri, Sharma, Wang et al. 2015). Our sample consisted of 1412 subjects from MESA exam 5 who had complete cMRI and PSG data. A more detailed description of the sample has been published (Javaheri, Sharma, Wang et al. 2015).

Covariates

Our primary exposure was CSA (defined as having a central apnea index≥5 (all central apneas associated with a 4% desaturation/total sleep time) or presence of Cheyne-Stokes Breathing (CSB)). CSB was defined as at least ten minutes of crescendo decrescendo pattern of breathing during sleep. Our primary cMRI outcome was left ventricular ejection fraction (LVEF). Secondary outcomes were LV mass/volume ratio (LVMVR) and LV mass adjusted for body height in meters (LVHi). LVMVR was derived by dividing the LV mass in grams by end-diastolic volume in milliliters. Covariates included pack years of smoking (by self-report), systolic blood pressure, diastolic blood pressure, or use of anti-hypertensive medications,) and the obstructive AHI (all obstructive apneas plus hypopneas associated with ≥4% desaturation/total sleep time). Blood pressure was obtained in triplicate in the right arm after 5 min rest in a seated position with a Dinamap automated device, and the average of the second and third measurement analyzed. The periodic limb movement arousal index was defined as the sum of periodic limb movements, each associated with an EEG arousal, per hour of sleep.

Statistical Analysis

Between group differences were compared using Fisher’s Exact Test for categorical variables and the 2-sample t-test for continuous variables. If a continuous variable was highly skewed, a Wilcoxon Rank Sum test was used. Multivariable linear regression models adjusted for age, gender, race, waist circumference, smoking pack years, obstructive apnea hypopnea index, and hypertension were used to quantify associations between CSA and MRI outcomes. A p-value of 0.05 was considered significant. SAS 9.3 (SAS Institute, Cary, NC) was used to conduct analyses.

The study protocol was approved by an Institutional Review Board at all participating institutions and written informed consent was obtained from all participants.

Results

Table 1 describes sample characteristics by presence or absence of CSA. Overall sample characteristics are already published (Javaheri, Sharma, Wang et al. 2015). The sample had a mean age of 67 years, approximately 46% were male, and the sample was ethnically diverse with 38% white, 26% black, 22% Hispanic and 13% Chinese. Approximately 7% were smokers, 54% had self-reported hypertension, and 14% had diabetes. Reflecting the community sample and cohort design (free of known CVD at Exam 1), only 23 participants (1.5% of the sample) had an LVEF<45%. CSA was identified in 27 participants (approximately 2%, comparable to other community samples (Donovan and Kapur 2016)). Mean central apnea index was 5.1 ± 5.5 (range: 0 to 19) in those with CSA, with 7 meeting the CSA definition because of the occurrence of CSB with a low CAI. Of the 27 participants with CSA, 20 also had at least moderate obstructive sleep apnea (an obstructive AHI≥15). Compared to those without CSA, those with CSA were significantly older, more likely to be male, had increased pack-years of smoking, lower waist circumference, and a higher obstructive AHI. There were no significant differences in race, hypertension, alcohol use, diabetes, or periodic limb movement index between groups.

Table 1.

Sample Characteristics by Central Sleep Apnea

		CSA (n=27)	No CSA (n=1385)	P-value
Age (years)		77.7±9.7	67±8.7	<.0001
Male gender (n, %)		21, 78%	634, 46%	0.001
Race	White (n, %)	9, 33.33%	527, 38.05%	0.86
	Black (n, %)	5, 18.52%	180, 13%
	Hispanic (n, %)	7, 25.93%	366, 26.4%
	Chinese (n, %)	6, 22.22%	312, 22.53%
BMI (kg/m2)		25.9±4.1	28.13±5.1	0.02
Waist circumference		91.3±11.4	98±13.4	0.01
Pack-years of smoking		14.5±29	9.5±18.6	0.0001
Current alcohol use		13, 48%	603, 43%	0.7
Gross family income	<20K (n, %)	8, 29.6%	263, 19.3%	0.6
	20–40K (n, %)	6, 22.2%	323, 23.8%
	40–75K (n, %)	6, 22.2%	356, 26.2%
	>75K (n, %)	7, 25.9%	418, 30.7%
Central Apnea Index		0.19±0.57	5.09±5.46	<0.0001*
Obstructive AHI		2.87±6.63	11.84±11.1	<0.0001*
PLM Arousal Index		1.52±3.58	3.31±5.8	0.09*
Hypertension (n, %)		17, 63.0%	752, 54.3%	0.47
Diabetes (n, %)		1, 3.7%	111, 7.9%	0.7
Atrial fibrillation (n, %)		12, 0.88%	1, 3.7%	0.22
LVEF		60.2±9.4	61.8±7.2	0.27
LVHi (gm/cm)		79.7±18.0	73.5±18.7	0.08
LVMVR (gm/mL)		1.12±0.28	1.0±0.22	<0.0001
LVEF<45%		1, 3.7%	22, 1.6%	0.36

Data presented as mean ± SD unless otherwise specified. All p-values obtained from T-test for continuous variables and Fisher’s Exact Test for categorical variables unless otherwise noted.

^*p-value obtained from Wilcoxin Rank Sum test

Abbreviations: PLM=Periodic Limb Movement; AHI=Apnea Hypopnea Index; LVEF=Left ventricular ejection fraction; LVHi=Left ventricular mass/height; LVMVR=Left ventricular mass/volume ratio

In unadjusted analysis, LVMVR was significantly associated with presence of CSA (p=0.0007), but not LVEF or LVHi. After adjustment for age, gender, race, waist circumference, smoking pack-years, systolic and diastolic blood pressure, use of anti-hypertensive medications,, and obstructive AHI, CSA remained significantly associated with LVMVR (β= 0.10±0.09 gm/mL, P=0.015) and again there were no significant associations between CSA and LVEF or LVHi. Obstructive AHI, however, was not associated with LVMVR after adjusting for CSA (p=0.08).

Discussion

To our knowledge, this is the first investigation to quantify associations between CSA (defined by an elevated CAI or CSB) and LV structure in a community-based population free of CVD when recruited into the cohort study. Prior work assessing CSA and LV structure has done so in a population with known HF and using echocardiography (Kourouklis, Vagiakis, Paraskevaidis et al. 2013). Our results demonstrate that in an ethnically diverse community-based cohort, after adjusting for potential confounders including OSA, the presence of CSA is significantly associated with increasing LVMVR. A positive association between CSA and LVMVR is of particular interest given higher LVMVR signifies concentric remodeling that usually occurs in response to hemodynamic stress and is a predictor of worse cardiovascular outcomes (Bluemke, Kronmal, Lima et al. 2008). LVMVR is closely associated with functional markers of early LV dysfunction including LV circumferential and global longitudinal strain and dyssynchrony (Choi, Rosen, Fernandes et al. 2013; Sharma, Volpe, Rosen et al. 2014). These changes precede onset of clinical HF and confer significant morbidity and mortality burden. Of particular interest is the reported higher prevalence of HF with preserved ejection fraction (HFpEF), characterized by diastolic dysfunction (Katz, Beussink, Sauer et al. 2013) among those with CSA. As suggested in our analysis, this may be potentially mediated by unfavorable LV remodeling and associated cardiac functional deterioration. Our results also demonstrate that in this community-based cohort, there was no significant association between CSA and LVHi or LVEF. The lack of a significant association between CSA and LVEF likely reflects the low prevalence of both CSA and reduced LVEF in this cohort, a consequence of the study design and low prevalence of clinically overt CVD in the sample. CSA in absence of HF may represent several entities including augmented chemoreflex (Xie, Rutherford, Rankin, Wong and Bradley 1995), subclinical HF, or idiopathic CSA. As the MESA cohort is followed for incident CVD, there will be opportunities to further follow the prognostic value of CSA as an indicator of incident HF, and whether those with CSA and elevated LVMVR have increased risk of HFpEF.

Strengths of this study include the use of a large, community-based sample of ethnically diverse women and men recruited free of baseline CVD as well as use of standardized and objective PSG and cMRI measures. Limitations include the cross-sectional design as well as the relatively low prevalence of both CSA and reduced LVEF, a reflection of a cohort free of CVD at the time of recruitment. However, the prevalence of CSA in this sample is comparable to that of other community samples such as the Sleep Heart Health Study (Donovan and Kapur 2016). Additionally, information to further characterize CSA (such as history of opiate use, duration of Cheyne-Stokes breathing, and cycle length) was not available. The frequent co-occurrence of CSA and OSA in the majority of this sample prevented us from examining a “pure” CSA phenotype and makes the findings more difficult to interpret. Nonetheless, the association with LVMVR was with CSA in models adjusting for the obstructive AHI, and the obstructive AHI was not associated with LVMVR.

Conclusions

In an ethnically diverse community-based sample of adults initially recruited to be free of clinical CVD, participants with CSA have higher LVMVR suggestive of concentric modeling.