Objectively measured sleep characteristics and prevalence of coronary artery calcification: The Multi-Ethnic Study of Atherosclerosis Sleep Study

Abstract

Background

We tested whether objectively measured indices of obstructive sleep apnea (OSA) and sleep quality are associated with coronary artery calcification (CAC) prevalence independent of obesity, a classic confounder.

Methods

A total of 1,465 Multi-Ethnic Study of Atherosclerosis participants [mean age 68 years], who were free of clinical cardiovascular disease, had both coronary CT and in-home polysomnography and actigraphy. OSA categories were defined by apnea-hypopnea index (AHI). Prevalence ratios for CAC >0 and >400 (high burden) were calculated.

Results

Participants with severe OSA (AHI ≥ 30; 14.6%) were more likely to have prevalent CAC, relative to those with no evidence of OSA, after adjustment for demographics and smoking status [1.16 (95% CI: 1.06–1.26)], body mass index [1.11 (1.02–1.21)], and traditional cardiovascular risk factors [1.10 (1.01–1.19)]. Other markers of hypoxemia tended to be associated with higher prevalence of CAC >0. For CAC >400 a higher prevalence was observed with both a higher arousal index and less slow-wave sleep. Overall, associations were somewhat stronger among younger participants, but did not vary by sex or race/ethnicity.

Conclusions

In this population-based multiethnic sample severe OSA was associated with subclinical coronary artery disease (CAC >0), independent of obesity and traditional cardiovascular risk factors. Furthermore, the associations of the arousal index and slow wave sleep with high CAC burden suggest that higher nightly sympathetic nervous system activation is also a risk factor. These findings highlight the potential importance in measuring disturbances in OSA as well as in sleep fragmentation as possible risk factors for coronary artery disease.

INTRODUCTION

Obstructive sleep apnea (OSA) has been associated with elevated risk of incident cardiovascular disease (CVD) in several observational studies^{1, 2}, and among patients with OSA, treatment has been associated with lower CVD morbidity and mortality³. Despite this suggestive evidence, important key questions remain. As noted in an American Heart Association/American College of Cardiology Foundation Scientific Statement, it remains unclear “whether sleep apnea is important in initiating the development of cardiac and vascular disease.”¹ One way to address this question is through examination of the association between OSA and markers of subclinical CVD, such as coronary artery calcification (CAC). MESA has previously shown CAC to be predictive of incident clinical CVD risk⁴.

Information examining associations between objective measurements of sleep and CAC is sparse. Of the five existing studies which objectively measured sleep apnea, all found sleep apnea to be associated with CAC prevalence in basic models^5–9, however in some instances the association was non-significant after adjustment for BMI^{5, 6}. It is important to note that all but one⁹ of these studies were relatively small (N < 260), and two were conducted among individuals with suspected sleep disorders^{7, 8}. In addition to sleep apnea, several other sleep phenotypes, such as abnormal (short or long) sleep duration and insomnia have also been evaluated in relation to CAC^{8, 10} and CVD risk^{11, 12}. Results have been mixed, and interpretation is challenging as many of the existing studies had small sample sizes and several used sleep data based on self-report. Self-reported measures of typical sleep are only modestly correlated with objectively measured sleep characteristics¹³, and no study has yet evaluated the influence of sleep stages and arousal frequency in relationship to CAC. Sleep disturbances may influence atherogenesis through several pathways, such as hypoxemia or sympathetic nervous activation¹, but no prior study has comprehensively assessed the independent contributions of alternative measures of sleep disturbances.

A related question is whether racial/ethnic variation in sleep disturbances may contribute to the well-established racial/ethnic differences in CVD risk¹⁴. Few studies have objectively measured sleep characteristics in racially/ethnically diverse populations. However, existing literature suggests that sleep disorders vary by race/ethnicity¹⁵. African Americans appear to be disproportionately affected by OSA and short sleep duration, relative to Caucasians¹⁶. Much less is known about other racial/ethnic groups, though recent work from the MESA Sleep study suggests that after accounting for BMI, Chinese have a higher prevalence of OSA then do Caucasian¹⁷. Similarly, the prevalence of OSA varies by age and sex⁴, as do CVD incidence rates¹⁸, but it is not clear whether OSA may underlie this variation.

Data from the Multi-Ethnic Study of Atherosclerosis (MESA) were used to test the hypotheses that objectively measured OSA, and adverse levels of other metrics of sleep quality and quantity, are associated with greater CAC prevalence and CAC severity, independent of obesity and possible mediating traditional cardiovascular risk factors. We also explored whether associations between sleep characteristics and CAC prevalence varied by age, sex, or race/ethnicity.

MATERIALS AND METHODS

Design Overview

The MESA Study¹⁹ (http://www.mesa-nhlbi.org/) began when, between 2000 and 2002, a total of 6,814 men and women who were free of clinical CVD and aged 45–85 years were recruited from six U.S. communities: Chicago, IL; Los Angeles County, CA; New York, NY; Forsyth County, NC; St Paul, MN and Baltimore, MD. At baseline, participants self-identified as non-Hispanic African American (28%), Chinese (12%) and Caucasian (38%), and 22% identified as Hispanic. A total of 5 clinical exams have taken place, the most recent of which (Exam 5) was conducted from April 2010-December 2011, and attended by 4,651 individuals (78% of original MESA participants who were alive). CAC was measured, among those with no contraindications, in a subset of 3,305 Exam 5 participants. MESA participants who took part in Exam 5 and were not on current treatment with positive airway pressure, oral devices or oxygen (n=95) were also invited to take part in the MESA Sleep ancillary study, which conducted overnight in-home polysomnography, 7-day actigraphy, and sleep questionnaires after their MESA Exam 5 clinic visit. In total, 2,237 participants underwent polysomnography, of which 1,581 participants also had CAC measured at MESA Exam 5. From this sample, we further excluded participants with prevalent myocardial infarction, stroke, congestive heart failure, or who had undergone interventional cardiology procedures (n = 116). The final analytic sample for most analyses was 1,465. All participants gave informed consent, and local Institutional Review Boards approved the study protocols for the main MESA exams, and for the MESA Sleep ancillary study.

Sleep Ascertainment

In-home polysomnography was conducted using the Compumedics Somte System (Compumedics LTd., Abbostville, Australia) using techniques similar to those previously described²⁰. The sensors and recording montage consisted of central, occipital and frontal electroencephalograms bilateral electrooculograms, chin EMG, thoracic and abdominal respiratory inductance plethysmography; airflow (by nasal-oral thermocouple and pressure recording from a nasal cannula); ECG; leg movements (piezoelectric sensors), and finger pulse oximetry. Nocturnal recordings were transmitted to the centralized reading center at Brigham and Women’s Hospital and data were scored by trained technicians using current guidelines. For our primary analyses we defined OSA according to the apnea hypopnea index (AHI) which includes all apneas (regardless of desaturation or arousal) and all hypopneas with ≥4% oxygen desaturation. Arousals, transient awakenings from sleep lasting <10 seconds, were characterized by the American Academy of Sleep Medicine criteria²¹. Sleep stages were identified for each 30 second epoch using American Academy of Sleep Medicine scoring criteria²² Other polysomnography measures included measures of sleep disordered breathing (average oxygen saturation during sleep; percentage time during sleep with an oxygen saturation <90%; percentage sleep time occupied by apneas or hypopneas); percentage time in stage N3 (slow-wave sleep); and the arousal index in stages REM and NREM. Inter- and intra-scorer intraclass correlation coefficients for the AHI ranged from 0.95 to 0.99; for arousal index 0.84 to 0.99; and for percentage time in stage N3 0.79 to 0.99.

Actigraphy was performed using the Actiwatch Spectrum wrist actigraph (Philips Respironics, Murrysville, PA) worn on the participant’s non-dominant wrist. Output was sent to the Sleep Reading Center at Brigham and Women’s Hospital where records were scored with use of the sleep diary. Summary variables examined included average sleep duration, sleep efficiency (proportion of the sleep period asleep), and time wake after sleep onset (WASO). A minimum of 3 days of data with > 50% reliable data were required to meet minimal standards for analysis. Inter-scorer reliability for average sleep duration, sleep efficiency, and WASO were 0.91, 0.97, and 0.91, respectively.

CAC Measurement

Details of procedures for CAC measurement, scanner quality assurance, and scan reading are provided here: http://www.mesa-nhlbi.org/MesaInternal/documents/Exam5/MESAE5_CT_MOO.pdf. Briefly, CAC was assessed by chest computed tomography with a multi-detector computed tomography system at all 6 sites, using standardized protocols. A cardiologist or radiologist interpreted all scans at the MESA CT reading center (Los Angeles Biomedical Research Institute at Harbor-UCLA Medical Center), blinded to participant data. Agatston scores were quantified. In prior MESA exams, the kappa statistics for intra- and inter-reader reproducibility of CAC prevalence were both 0.92. Intraclass correlation coefficients for intra- and interreader reproducibility of CAC scores exceeded 0.99. For the present analysis CAC was considered prevalent (CAC>0) if the Agatston score was greater than 0. High CAC burden (CAC>400) was defined by Agatston scores >400.

Other Variables

Sex, age, cigarette smoking status (current, former, never), and use of anti-hypercholesterolemics, anti-hypertensives, and diabetes medications were self-reported. Body mass index (BMI) was calculated as weight over height squared (kg)/(m²). Resting blood pressure was measured three times in the seated position using a Dinamap model Pro 100 automated oscillometric sphygmomanometer (Critikon, Tampa, Florida). The average of the last two measurements was used in analyses. Participants were asked to fast for at least 8 hours prior to their visit. Serum glucose was measured by rate reflectance spectrophotometry using thin film adaptation of the glucose oxidase method on the Vitros analyzer (Johnson & Johnson Clinical Diagnostics, Inc., Rochester, NY). HDL cholesterol was assessed in EDTA plasma using the cholesterol oxidase method (Roche Diagnostics) after precipitation of non-HDL-cholesterol with magnesium/dextran, and LDL cholesterol was calculated in plasma specimens having a triglyceride value <400 mg/dL using the Friedewald formula. Serum assays were performed at the Collaborative Studies Clinical Laboratory at Fairview-University Medical Center (Minneapolis, MN).

Data Analysis

Descriptive characteristics of participants are presented as means and proportions, stratified by clinical categories of OSA presence/severity (no OSA, AHI 5–14, AHI 15–30, AHI >30). For the primary analysis, relative risk regression (binomial regression) was used to explore the association between sleep phenotypes and CAC prevalence (CAC >0). OSA prevalence/severity was modeled according to aforementioned clinical categories. Sleep duration was categorized: <399 minutes, 399–444 minutes (reference; corresponds to the 50^th to 75^th percentiles of the MESA distribution), and >444 minutes. All other sleep phenotypes were modeled per 1 standard deviation (SD). Nonlinearity was tested using generalized additive models with any CAC as the endpoint and adjustments of age, gender, race, site, education, income and smoking status. Additional models examined high CAC burden (CAC >400) as the outcome.

A series of models were conducted: Model 1 adjusted for age, race/ethnicity, sex, center, education (< HS, HS or some college, college degree or higher), income (<$20K, 20K to <$50K, ≥$50K) and smoking status (current, former, never). Model 2, our primary model, further adjusted for BMI. Model 3, an “overadjusted model”, additionally adjusted for traditional CVD risk factors, some of which are believed to be on the pathway through which OSA influences CAC prevalence. Specifically we added to Model 3 prevalent diabetes, systolic BP, hypertension medication use, HDL-C, LDL-C, use of lipid-lowering medications, and hs-CRP. Regarding interpretation, if magnitudes of association are attenuated with adjustment for Model 3 covariates, it implies that the association between OSA and CAC is mediated through these factors. Model 4 was also adjusted for sleep duration categories in order to evaluate whether sleep duration was a confounder of the association between OSA and CAC. Effect modification of the OSA and CAC association by age, sex, and race/ethnicity was explored by including cross-product terms with model 1 adjustments.

RESULTS

The 1,465 participants in our analytic sample were on average 68 years old; 54% were female, and 35.9% Caucasian, 12.5% Chinese, 27.4% African American, and 24.2% Hispanic. Of the total sample, 14.6% had severe OSA (AHI >30), 18.0% had moderate OSA (AHI 15–30), 32.6% had mild OSA (AHI 5–14), and 34.8% had no OSA (AHI <5). As shown in Table 1, participants with severe OSA were slightly older, more likely to be male, have greater adiposity (as assessed by BMI or waist circumference), and have worse cardiovascular disease risk factor profiles (i.e. more diabetes, elevated SBP, total cholesterol, low HDL-C, and greater usage of blood pressure and lipid-lowering medications). CAC presence (>0) was found in 64.0% of the cohort; while CAC > 400 was found in 14.9%. Associations of demographic, behavioral, cardiovascular and sleep characteristics stratified by CAC categories are presented in e-Table 1. In general, relative to participants with no CAC, those with a high CAC burden tended to be older, male, and have a more adverse cardiovascular risk factor profile.

Table 1.

Descriptive^* characteristics by obstructive sleep apnea severity: The MESA Study 2010–2013

	OSA Severity Category				p-trend†
	Normal (AHI <5)	Mild (AHI 5–14)	Moderate (AHI 15–29)	Severe (AHI ≥ 30)
N (%)	510 (35)	478 (33)	263 (18)	214 (15)
Demographics & Behaviors
Age (yrs)	66.9 ± 9.0	68.9 ± 9.2	69.1 ± 9.2	67.9 ± 9.0	0.03
Male gender, n (%)	165 (32.3)	214 (44.8)	155 (58.9)	136 (63.6)	<0.001
Race/ethnicity, n (%)
Caucasian	192 (37.6)	181 (37.9)	89 (33.8)	65 (30.2)	0.041
Chinese	68 (13.3)	49 (10.3)	36 (13.7)	30 (14.0)
African American	150 (29.4)	133 (27.8)	64 (24.3)	54 (25.2)
Hispanic	100 (19.6)	115 (24.1)	74 (28.1)	65 (30.4)
Education, n (%)
Less than HS	67 (13.2)	68 (14.3)	40 (15.2)	37 (17.3)	0.26
HS or some college	156 (30.6)	168 (35.4)	84 (31.9)	78 (36.5)
College degree	286 (56.2)	239 (50.3)	139 (52.9)	99 (46.3)
Income‡, n (%)
<$20k	94 (19.1)	92 (19.4)	54 (21.0)	54 (25.7)	0.41
$20k to <$50k	164 (33.3)	171 (36.1)	91 (35.4)	64 (30.5)
≥$50k	235 (47.7)	211 (44.5)	112 (43.6)	92 (43.8)
Smoking Status, n (%)
Never	250 (49.3)	216 (45.3)	127 (48.9)	94 (43.9)	0.16
Former	217 (42.8)	224 (47.0)	125 (47.9)	106 (49.5)
Current	40 (7.9)	37 (7.8)	9 (3.5)	14 (6.5)
Pack-years	8.6 ± 16.7	11.1 ± 21.5	9.7 ± 17.8	9.2 ± 17.9	0.64
Anthropometry
BMI (kg/m2)	26.6 ± 5.0	28.9 ± 5.0	29.6 ± 5.3	31.7 ± 6.0	<0.001
Waist size (cm)	93.6 ± 14.0	100.1 ± 13.5	102.1 ± 12.8	106.5 ± 14.7	<0.001
Cardiovascular Risk Factors
Diabetes, n (%)	71 (14.0)	90 (19.0)	63 (24.1)	51 (23.8)	0.001
Systolic BP (mmHg)	121.1 ± 21.4	123.0 ± 19.4	122.4 ± 19.1	126.0 ± 19.8	0.007
BP medication, n (%)	232 (45.5)	255 (53.4)	137 (52.1)	124 (57.9)	0.009
Total Cholesterol (mg/dl)	190.0 ± 34.9	184.7 ± 34.4	182.7 ± 36.4	180.9 ± 35.8	<0.001
HDL-C (mg/dl)	60.7 ± 17.7	54.9 ± 15.2	52.3 ± 14.4	50.8 ± 13.5	<0.001
LDL-C (mg/dl)	109.4 ± 31.4	107.8 ± 30.5	107.9 ± 31.3	105.2 ± 33.3	0.13
Lipid medication, n (%)	155 (30.4)	177 (37.0)	97 (36.9)	84 (39.3)	0.05
eGFR categories					0.43
≥90	195 (38.6)	159 (33.5)	87 (33.2)	78 (36.5)
60–89	259 (51.3)	259 (54.5)	148 (56.5)	119 (55.6)
<60	51 (10.1)	57 (12.0)	27 (10.3)	17 (7.9)
Agatston score, median
Overall	6.2	32.6	31.8	62.9
Among those with CAC>0	84.3	125.5	123.7	117.4

^*Numbers in table are mean ± standard deviation, unless otherwise noted as n (% col) or median.

^†P-values are based on Chi-squared tests for categorical variables, and trend tests for continuous variables.

^‡Income refers to total gross family income.

Obstructive Sleep Apnea & Prevalent CAC

Participants with severe OSA were 1.16 (95% CI: 1.06, 1.26) times more likely to have prevalent CAC relative to participants with no evidence of OSA, after adjustment for demographics and smoking status (Table 2). The association remained significant after additional adjustment for BMI [1.11 (1.02, 1.21)], and traditional CVD risk factors which may be mediators of the association between OSA and CAC [1.10 (1.01, 1.19)]. Results were essentially unchanged with additional adjustment for eGFR or sleep duration category. There were no interactions by either sex or race/ethnicity. The association between severe OSA and CAC > 0 was slightly stronger among younger participants than older participants (median population age = 67 years): p-interaction <0.01; PR (95% CI) ≤ 67 years = 1.22 (1.03, 1.45) >67 years = 1.16 (1.04, 1.28). No associations were seen between mild or moderate OSA and CAC prevalence, regardless of degree of adjustment.

Table 2.

Prevalence Ratios (95% CI’s) of obstructive sleep apnea and coronary artery calcification (CAC): The Multi-Ethnic Study of Atherosclerosis 2010–2013

	OSA Category				p-trend
	Normal (AHI <5)	Mild (AHI 5–14)	Moderate (AHI 15–29)	Severe (AHI≥30)
N total (%)	510 (34.8)	478 (32.6)	263 (18.0)	214 (14.6)
CAC > 0 (n=937)
N	280	318	177	162
Model 1	1.00	1.05 [0.97,1.14]	1.03 [0.94,1.13]	1.16 [1.06,1.26]	0.001
Model 2	1.00	1.04 [0.96,1.12]	1.00 [0.91,1.10]	1.11 [1.02,1.21]	0.03
Model 3	1.00	1.04 [0.97,1.12]	1.00 [0.91,1.09]	1.10 [1.01,1.19]	0.07
Model 4	1.00	1.04 [0.96, 1.12]	1.00 [0.91, 1.10]	1.10 [1.01, 1.20]	0.05
CAC > 400 (n=218)
N	62	72	44	40
Model 1	1.00	0.95 [0.70,1.28]	1.09 [0.81,1.48]	1.16 [0.85,1.58]	0.44
Model 2	1.00	0.93 [0.68,1.26]	1.05 [0.76,1.45]	1.10 [0.79,1.53]	0.42
Model 3	1.00	1.00 [0.73,1.38]	1.07 [0.75,1.51]	1.20 [0.85,1.69]	0.33
Model 4	1.00	1.00 [0.71,1.39]	1.07 [0.75,1.53]	1.20 [0.86,1.68]	0.33

Model 1: adjusted for age, race/ethnicity, sex, center, education (< HS, HS or some college, college degree or higher), income (<$20K, 20K to <$50K, ≥$50K) and smoking status (current, former, never).

Model 2: Adjusted for Model 1 + BMI

Model 3: Adjusted for Model 2 + traditional CVD risk factors (prevalent diabetes, systolic BP, hypertension medication use, HDL-C, LDL-C, use of lipid-lowering medications)

Model 4: Adjusted for Model 3+sleep duration category

When CAC > 400 was considered as the outcome, although severe OSA was associated with a qualitatively greater prevalence of CAC > 400, the association was not statistically significant in any of the models considered. There was no interaction by age.

Other Sleep Phenotypes & Prevalent CAC

Associations between other sleep phenotypes and prevalent CAC (> 0) are presented in Table 3. With the exception of sleep duration, sleep phenotypes were modeled linearly (per 1 SD). There was no evidence of nonlinearity. In demographic-adjusted models, prevalence of CAC >0 was significantly associated with average oxygen saturation during sleep, percentage of time at oxygen saturation <90%, percentage of sleep time spent in apnea or hypopnea, number of apnea/hypopnea events per night, and AHI (modeled continuously). Associations were slightly attenuated after adjustment for BMI. With additional adjustment for traditional cardiovascular risk factors, only percent time in apnea and hypopnea and the AHI remained significant. There was little evidence that measures of sleep architecture, sleep fragmentation, or sleep duration were associated with CAC >0. There were no interactions between sleep phenotypes and race or sex on CAC >0. However, several age interactions were noted (all p<0.01) when age was modeled continuously, and with adjustment for model 1 covariates. For the following phenotypes, associations were stronger among younger individuals (see e-Table 2): average oxygen saturation, percentage sleep time in SA0₂ <90%, number of apnea/hypopnea events/night, AHI, the overall arousal index, and the non-REM arousal index.

Table 3.

Prevalence Ratios (95% CI’s) ^†‡ of measures of sleep disordered breathing, sleep architecture, sleep fragmentation and sleep duration and coronary artery calcification prevalence (CAC > 0): The Multi-Ethnic Study of Atherosclerosis 2010–2013

CAC > 0
		Model 1		Model 2		Model 3
	1 SD	PR (95% CI)	p-value	PR (95% CI)	p-value	PR (95% CI)	p-value
Hypoxemia & Disordered Breathing
Average Oxygen Saturation In Sleep	1.7	0.96 [0.94,0.99]	0.005	0.98 [0.95,1.01]	0.12	0.98 [0.95,1.01]	0.25
Percentage Sleep time SA02<90%	8.6	1.03 [1.01,1.05]	0.004	1.02 [1.00,1.04]	0.05	1.02 [1.00,1.04]	0.13
Percent Time in apnea+hypopnea	12.4	1.05 [1.02,1.08]	0.001	1.03 [1.00,1.06]	0.03	1.03 [1.00,1.06]	0.04
# Apnea/Hypopnea Events Per Night	109	1.05 [1.02,1.07]	0.001	1.03 [1.00,1.06]	0.02	1.03 [1.00,1.05]	0.07
AHI (continuously measured)	16.5	1.05 [1.02,1.08]	<0.001	1.04 [1.01,1.07]	0.01	1.03 [1.00,1.06]	0.03
Sleep Architecture
Percent Time in Stage N3	9.0	1.00 [0.96,1.04]	0.94	1.00 [0.97,1.04]	0.93	1.00 [0.97,1.04]	0.83
Sleep Fragmentation
Arousal Index	12.0	1.02 [0.99,1.06]	0.10	1.02 [0.99,1.05]	0.26	1.02 [0.99,1.05]	0.25
Arousal Index—REM	11.8	1.03 [1.00,1.06]	0.05	1.02 [0.99,1.06]	0.13	1.02 [1.00,1.06]	0.10
Arousal Index—NREM	12.8	1.02 [0.99,1.05]	0.11	1.02 [0.99,1.05]	0.25	1.02 [0.99,1.04]	0.26
Average Sleep Efficiency %	3.5	0.98 [0.95,1.01]	0.18	0.98 [0.96,1.01]	0.24	0.98 [0.95,1.01]	0.20
Average Sleep WASO	16.7	1.01 [0.98,1.04]	0.44	1.01 [0.98,1.04]	0.38	1.01 [0.98,1.04]	0.47
Sleep Duration
399–444 min		Ref		ref		Ref
<399 min		1.04 [0.96,1.12]	0.37	1.02 [0.94,1.11]	0.59	1.03 [0.95, 1.11]	0.46
>444 min		0.99 [0.91, 1.09]	0.88	0.99 [0.91, 1.08]	0.83	0.98 [0.90, 1.07]	0.71
3 degree of freedom test			0.47		0.71		0.46

^†All continuous variables have PR expressed per SD increment

^‡Only 1,390 observations were present for certain measures based on actigraphy (i.e. sleep efficiency, WASO and sleep duration).

Model 1: adjusted for age, race/ethnicity, sex, center, education (< HS, HS or some college, college degree or higher), income (<$20K, 20K to <$50K, ≥$50K) and smoking status (current, former, never).

Model 2: Adjusted for Model 1 + BMI

Model 3: Adjusted for Model 2 + traditional CVD risk factors (prevalent diabetes, systolic BP, hypertension medication use, HDL-C, LDL-C, use of lipid-lowering medications)

A greater prevalence of CAC >400 was observed with higher scores on the arousal indices (total, REM, and non-REM) (Table 4). In fully-adjusted models, each SD higher score on the total arousal index was associated with a prevalence ratio of 1.14 (1.02, 1.29) for CAC >400. An inverse association between percent time in stage N3 (slow-wave sleep) and prevalence of CAC >400 was also observed in the fully-adjusted model [prevalence ratio per 1 SD higher percent time in Stage N3: 0.77 [0.64,0.92]. No meaningful interactions by age, sex, or race were observed.

Table 4.

Prevalence Ratios (95% CI’s) ^† of measures of sleep disordered breathing, sleep architecture, sleep fragmentation and sleep duration and high coronary artery calcification burden (CAC > 400): The Multi-Ethnic Study of Atherosclerosis 2010–2013

CAC > 400
		Model 1		Model 2		Model 3
	1 SD	PR (95% CI)	p-value	PR (95% CI)	p-value	PR (95% CI)	p-value
Hypoxemia & Disordered Breathing
Average Oxygen Saturation In Sleep	1.7	1.03 [0.83,1.29]	0.77	1.04 [0.81,1.34]	0.77	1.18 [0.89,1.56]	0.25
Percentage Sleep time SA02<90%	8.6	0.93 [0.68,1.29]	0.68	0.93 [0.66,1.31]	0.66	0.78 [0.52,1.19]	0.25
Percent Time in apnea+hypopnea	12.4	1.02 [0.92,1.13]	0.73	1.02 [0.92,1.13]	0.72	1.04 [0.93,1.16]	0.50
# Apnea/Hypopnea Events Per Night	109	1.06 [0.95,1.19]	0.30	1.06 [0.95,1.19]	0.31	1.02 [0.90,1.16]	0.76
AHI (continuously measured)	16.5	1.02 [0.91,1.14]	0.80	1.02 [0.90,1.14]	0.78	1.02 [0.91,1.15]	0.71
Sleep Architecture
Percent Time in Stage N3	9.0	0.86 [0.70,1.06]	0.15	0.86 [0.70,1.06]	0.15	0.77 [0.64,0.92]	0.005
Sleep Fragmentation
Arousal Index	12.0	1.09 [0.97,1.21]	0.15	1.09 [0.97,1.21]	0.15	1.14 [1.02,1.27]	0.02
Arousal Index—REM	11.8	1.14 [1.01,1.30]	0.04	1.15 [1.01,1.30]	0.04	1.15 [1.02,1.29]	0.02
Arousal Index—NREM	12.8	1.08 [0.96,1.22]	0.18	1.08 [0.96,1.21]	0.18	1.14 [1.02,1.28]	0.03
Average Sleep Efficiency %	3.5	1.02 [0.92,1.14]	0.68	1.02 [0.92,1.14]	0.68	1.00 [0.89,1.13]	0.97
Average Sleep WASO	16.7	1.07 [0.97,1.19]	0.19	1.07 [0.97,1.19]	0.19	1.03 [0.91,1.17]	0.62
Sleep Duration
399–444 min		ref		ref		ref
<399 min		0.68 [0.49,0.95]	0.02	0.66 [0.47,0.94]	0.02	0.75 [0.55,1.03]	0.08
>444 min		0.95 [0.68,1.32]	0.74	0.94 [0.68,1.31]	0.72	0.84 [0.56,1.26]	0.40
3 degree of freedom test			0.06		0.05		0.19

^†All continuous variables have PR expressed per SD increment

Model 1: adjusted for age, race/ethnicity, sex, center, education (< HS, HS or some college, college degree or higher), income (<$20K, 20K to <$50K, ≥$50K) and smoking status (current, former, never).

Model 2: Adjusted for Model 1 + BMI

Model 3: Adjusted for Model 2 + traditional CVD risk factors (prevalent diabetes, systolic BP, hypertension medication use, HDL-C, LDL-C, use of lipid-lowering medications)

DISCUSSION

In this large, cross-sectional study of racially/ethnically diverse middle-aged and older individuals, objectively measured indices of sleep disturbances were found to be associated with CAC after considering a number of potential confounders. Measures of OSA, in particular both the AHI and measures of overnight hypoxemia, were positively associated with prevalence of subclinical atherosclerosis, defined by CAC >0. When considering OSA as a categorical measure, individuals with severe OSA, defined by an AHI>30, had an approximately 10% increased adjusted prevalence of CAC, compared to individuals without OSA. In contrast, high CAC burden (CAC>400) was significantly associated with high sleep fragmentation (arousal index) and low sleep quality (time in stage N3, slow wave sleep). In particular, each standard deviation increase in the arousal index was associated with a 14% higher prevalence of CAC>400, while each standard deviation decrease in stage N3 (slow wave sleep) time was associated with a 30% higher prevalence of CAC>400. Both of these sleep exposures are associated with increased levels of sympathetic nervous system activation and hypertension²³. The current data provide the first evidence we are aware of linking these sleep quality metrics to CAC burden, and suggest that disturbances in sleep architecture associated with higher nightly levels of sympathetic nervous system activation associate with a high burden of subclinical atherosclerosis.

OSA and CAC Prevalence

Our results extend and support finding from prior studies evaluating the association between OSA and CAC prevalence. These MESA findings are unique as OSA was objectively measured using state-of-the-art in-home polysomnography equipment, and the study sample was diverse, community-based, and quite large relative to other studies which have objectively measured sleep. OSA is known to be highly underdiagnosed in the community²⁴, and study samples identified in clinical settings (most often sleep clinics) are likely not representative of the general population. Of existing studies on OSA and CAC prevalence, our design compares most closely to that of the population-based Heinz Nixdorf Recall study, which recently evaluated this association in 1604 German participants⁹. This study used limited channel sleep monitoring, which precluded quantitative assessments of sleep architecture, sleep fragmentation and hypoxemia, and which may underestimate sleep apnea severity and obstructive/central subtypes. The study also did not collect objective measurements of sleep duration, as was done in MESA with the use of actigraphy. They found objectively-measured OSA to be associated with CAC prevalence among men aged ≤ 65 years and women of any age. Although in MESA there was no interaction by sex, associations between OSA and CAC tended to be stronger among younger (≤67 years) MESA participants. The idea that OSA may exert a stronger influence on atherosclerotic risk among younger individuals is also supported by findings from the Sleep Heart Health Study which revealed that OSA was an independent predictor of coronary heart disease in men ≤ 70 years of age, but not in older men or women of any age². Our MESA findings furthermore show that the relationship with OSA persisted after adjusting for sleep duration.

The two other population-based studies of OSA and prevalent CAC were much smaller than the MESA and Heinz Nixdorf Recall study samples, and in both instances the associations were not independent of BMI^{5, 6}. In MESA the association between OSA and CAC was modestly attenuated after adjustment for BMI. Obesity is a strong risk factor for the development of both OSA²⁵ and CAC²⁶, and as such is an important potential confounder of the association. However, in both MESA and the Heinz Nixdorf Recall study associations between the AHI and CAC prevalence persisted even after accounting for BMI and other established CVD risk factors, which have been hypothesized to mediate the association. These results support converging evidence which suggests that severe OSA is a risk factor for subclinical atherosclerosis. MESA participants with moderate or mild OSA did not have a higher CAC prevalence.

In our sample, no statistically significant association was present between OSA and CAC >400. However, as fewer people had CAC >400, power was more limited to detect an association. The fact that the magnitudes of the prevalence ratios were similar to those for CAC >0, and in some instances stronger than those for CAC >0, suggests that OSA may also be associated with CAC >400.

OSA causes repetitive acute hypoxemic episodes and sleep disruption, which are believed to initiate a range of pathophysiological mechanisms, including sympathetic nervous system activation, that may act to promote CVD. As has been reviewed elsewhere, OSA has been implicated in the pathogenesis of systemic inflammation, oxidative stress, a prothromotic state, hypertension, and diabetes¹. Hypoxemia occurring in association with OSA has been associated with both metabolic disturbances²⁷ and atherosclerosis²⁸. Furthermore, converging evidence also indicates that sleep interventions may improve vascular risk factors¹. Multiple guidelines recognize OSA as a secondary cause of hypertension and advise assessment and treatment of OSA in patients with refractory hypertension²⁹. Some evidence also suggests that OSA treatment may improve glucose metabolism^{30, 31} and inflammatory marker³² profile. Therefore, though speculative, it is possible that if OSA does influence CAC development through these existing cardiovascular risk factors, that intervening on OSA may slow CAC progression.

It is of interest that despite the strong potential pathophysiological links between OSA and hypoxemia-related stresses and atherosclerosis, the association between OSA and CAC was relatively modest, and only present with severe OSA. This is consistent with epidemiological literature that has shown relatively weaker associations of OSA with coronary heart disease compared to OSA and stroke^{2, 33}, and suggests the possibility that the influence of OSA on CVD risk may vary across CVD outcomes.

Other Sleep Phenotypes and CAC Prevalence

In addition to evaluating the association between OSA and CAC prevalence, we also assessed whether other objective measurements of sleep architecture were associated with CAC. Notably, for CAC >400, a level of CAC that indicates severe CAC burden, sleep disruption as measured by an elevated arousal index and lower stage N3 were both associated with a higher prevalence. Frequent arousals are accompanied by chemoreflex-mediated increases in sympathetic activity to the peripheral blood vessels and consequent vasoconstriction, endothelial dysfunction, and higher blood pressure¹. Stage N3 sleep is the stage when parasympathetic tone is highest and overnight blood pressure the lowest. Reductions in N3 sleep have been associated with an increased incidence of hypertension²³ as well with decreased insulin sensitivity³⁴, and both elevated blood pressure and diabetes have been shown to be associated with CAC development³⁵. These data indicate that polysomnography measures that quantify the degree of disruption in sleep may provide unique information regarding sleep-related CVD risk.

The CARDIA study previously reported an association between short sleep duration and CAC¹⁶. Short sleep duration may reflect curtailed time in bed, or reduced sleep time occurring as a consequence of frequent arousals or increased wake time after sleep onset. The differences in findings from MESA and CARDIA in regards to sleep duration and CAC prevalence may reflect differences in the ages of the two populations. The MESA findings suggest that in older populations, measures of sleep fragmentation are more strongly associated with CAC than are measures of sleep duration.

Strengths and Limitations

Strengths of this study are the relatively large population-based multi-ethnic sample, the use of both polysomnography and multiple-day actigraphy to objectively measure sleep characteristics, and the standardized assessment of CAC. Importantly, sleep studies conducted with in-home polysomnography, such as those conducted here, have been shown to be highly consistent with hospital polysomnography (log-transformed, r = 0.96 for apnea-hypopnea index) and highly reproducible³². Furthermore, the analytic sample was free of established CVD, allowing inferences to be made for subclinical disease. Although relatively large for a study which used polysomnography and 7-day actigraphy to objectively measure sleep characteristics, power was limited to detect effect sizes of small magnitude and for subgroup analyses. This was particularly evident when CAC >400 was used as the outcome. Furthermore, the design was cross-sectional. As such, it is not possible to determine temporality of the association between sleep characteristics and CAC prevalence, and selection bias may have occurred. To date, no studies have been published exploring the association between OSA and CAC incidence or progression.

Conclusions

Using objective measurements, severe OSA and other indices of nocturnal hypoxemia were associated with greater CAC prevalence, independent of BMI and traditional cardiovascular risk factors, in this cross-sectional analysis of data from the community-based Multi-Ethnic Study of Atherosclerosis. There also was evidence that an elevated arousal index and a low percentage of time spent in stage N3 sleep were associated with a high CAC burden (CAC >400). These specific measures of sleep disruption have been linked to elevations in sympathetic nervous system activation. Overall, associations were somewhat stronger in younger individuals relative to older individuals, but there was no evidence of differences by sex or race/ethnicity. Our findings support existing evidence suggesting that OSA is associated with risk of incident coronary artery disease. However, prospective data are needed to evaluate the temporal relation between OSA and CAC development.

Key Messages.

What is the key question?

Are indices of obstructive sleep apnea (OSA) and sleep quality associated with coronary artery calcification (CAC) prevalence independent of obesity?

What is the bottom line?

In this large population-based sample there was evidence that indices of sleep apnea, arousal and sleep quality were all associated with prevalent CAC and/or high CAC burden.

Why read on?

The current findings enhance the existing literature suggesting an association between OSA and CAC, and provide novel evidence that more frequent arousals and less slow wave sleep are associated with a high burden of subclinical atherosclerosis.

Abbreviations List

AHI

apnea-hypopnea index

BMI

body mass index

CVD

cardiovascular disease

CAC

coronary artery calcification

eGFR

estimated glomerular filtration rate

MESA

Multi-Ethnic Study of Atherosclerosis

OSA

obstructive sleep apnea

SD

standard deviation

WASO

time wake after sleep onset