Association between Obstructive Sleep Apnea and Left Ventricular Structure by Age and Gender: the Multi-Ethnic Study of Atherosclerosis

Abstract

Study Objectives:

The presence and severity of obstructive sleep apnea (OSA) are associated with impaired left ventricular (LV) structure and function. Our goal was to quantify the associations between LV systolic function and mass with severity of OSA in an ethnically diverse cohort, assessing variations by age and sex.

Methods:

We conducted a cross-sectional analysis of data from 1,412 racially/ethnically diverse participants across 6 US communities from the Multi-Ethnic Study of Atherosclerosis who underwent both overnight polysomnography and cardiac magnetic resonance imaging from 2010–2012. We evaluated the association between the obstructive apnea-hypopnea index (AHI) by clinical category (< 5, 5–15, 15–30, 30–50, > 50) and secondary measures of sleep apnea with the outcomes left ventricular (LV) mass adjusted for height, LV mass/volume ratio, and LV ejection fraction.

Results:

After adjusting for potential confounders and mediators, LV mass was significantly increased with increasing AHI category for subjects age 65 y or younger (β = 1.84 ± 0.47 g/m, P = 0.0001). The association between the AHI and LV mass appeared stronger in whites and Chinese compared to blacks and Hispanics, although interaction terms were not statistically significant. Additionally, while both LV mass and LV mass/volume ratio were significantly associated with hypoxia, ejection fraction was not associated with any OSA severity index. Comparable associations were observed in men and women.

Conclusions:

Independent of confounders, higher levels of AHI are significantly associated with increased LV mass in both men and women younger than 65 y from a community-based cohort.

Citation:

Javaheri S, Sharma RK, Wang R, Weng J, Rosen BD, Bluemke DA, Lima JA, Redline S. Association between obstructive sleep apnea and left ventricular structure by age and gender: the Multi-Ethnic Study of Atherosclerosis. SLEEP 2016;39(3):523–529.

Significance.

This study demonstrated significant associations between severity of obstructive sleep apnea (OSA) and left ventricular (LV) structure using cardiac magnetic resonance imaging. While prior research has demonstrated an association between obstructive sleep apnea and increased LV mass in men, this study extends the literature by demonstrating comparable associations between OSA severity and LV mass as well as LV mass/volume ratio in both men and women from an ethnically diverse, community-based cohort. Further, this association is significantly modified by age such that the associations are present in those 65 and younger but are attenuated in those > 65 years of age.

INTRODUCTION

Obstructive sleep apnea (OSA) is highly prevalent in patients with cardiovascular disease (CVD), particularly heart failure (HF). Approximately 35% to 60% of HF patients have OSA^1–5 and the presence of comorbid OSA is associated with adverse outcomes including increased hospitalizations, morbidity, and mortality.^6–8

The mechanisms underlying this relationship likely relate to several adverse pathophysiologic effects of OSA: recurrent upper airway obstruction with intrathoracic pressure swings, intermittent hypoxemia, and sympathetic hyperarousal, leading to increased left ventricular (LV) afterload and cardiac dysfunction.⁹ Hypertension¹⁰ and diabetes,¹¹ along with chronic sequelae of untreated OSA, can also be contributing factors.

Though growing literature supports a strong association between OSA and cardiac hypertrophy in patients with CVD, less is known regarding cardiac structural features associated with OSA in the general population. Additionally, prior research has largely utilized echocardiography to characterize these alterations with limited representation of women and minorities.^6,8,12–14 Given that structural changes such as cardiac hypertrophy and concentric remodeling predict future CVD and are associated with increased mortality,^15–21 we sought to use rigorously obtained measures from cardiac magnetic resonance imaging (cMRI) in an ethnically diverse community cohort representative of women and men to quantify the relationship of LV anatomy with objective measures of OSA.

We conducted an analysis of cross-sectional data from the Multi-Ethnic Study of Atherosclerosis (MESA) to address (1) the association of OSA with LV structure in a middle-aged to older community sample; (2) the influence of hypertension and diabetes on these relationships; and (3) possible variation by age, sex, and race. We hypothesized that (1) OSA severity is associated with impaired LV structural and functional parameters; (2) this association will be stronger in men, minorities, and younger individuals than in women, white individuals, and older individuals, in accordance with previous studies showing that men²² and minorities²³ have longer lifetime cumulative burden of OSA and that younger individuals have stronger associations with OSA and impaired LV structure.²⁴

METHODS

Study Design and Population

The sample was derived from MESA, a prospective multicenter study designed to detect markers of subclinical CVD in an ethnically diverse, community-based cohort. Details of this study design have previously been published.²⁵ Briefly, individuals aged 45–84 y and initially free of CVD were recruited from six different sites (Baltimore, MD; Chicago, IL, Forsyth County, NC; Los Angeles County, CA; Northern Manhattan, NY; and St. Paul, MN) between July 2000 and August 2002. Participants underwent five exams at 2-y intervals. The Exam 5 visit (2010– 2013) included standardized anthropometric measurements (height, weight, waist circumference, and blood pressure (BP) and cMRI. Also at Exam 5, participants other than those reporting regular use of oral devices, nocturnal oxygen, or nightly positive airway pressure (PAP) devices were invited to participate in the Multi-Ethnic Study of Atherosclerosis (MESA) Sleep Ancillary Study, which consisted of polysomnography (PSG), 7-d wrist actigraphy, and sleep questionnaire data collected during an in-home examination. Details of the sleep ancillary study have also been previously published.²⁶ Of 4,077 participants in the MESA Exam 5 who were approached, 147 (6.5%; 95 due to PAP, 52 due to oxygen use, oral device use, or other medical reason) were ineligible and 141 participants lived too far away to participate. Of the remaining 3,789 participants, 2,261 participated in the sleep examination (59.7%). In total, 2,057 participants had successful PSG data, of which 1,412 also had cardiac MRI in conjunction with Exam 5 (the other 645 with successful PSG data either elected not to undergo MRI or were ineligible due to medical reasons).

Measures

Polysomnography

Participants underwent a single night, unattended 15-channel PSG (Somté PSG, Computmedics Ltd., Abbotsford, Victoria, Australia), using methods adapted from the Sleep Heart Health Study (SHHS).²⁴ Recorded channels included electroencephalograms (EEGs) (Fz, Cz, Oz, C4, M1), bilateral electrooculograms, electrocardiogram, chin electromyogram, thoracic and abdominal respiratory inductance plethysmography, airflow (via oral/nasal thermistor and nasal pressure transducer), oxyhemoglobin saturation (finger pulse oximetry), leg movements, and body position. Unit setup, including connection of sensors and electrodes and checks of signal calibrations and impedance, was performed by trained, certified staff. Studies were scored at a centralized sleep reading center (Brigham and Women's Hospital, Boston, MA) by registered polysomnologists. Events were classified as obstructive based on the presence of respiratory effort as measured by thoracic and abdominal plethysmography. An apnea was defined as a complete reduction in the thermocouple signal for more than 10 sec. A hypopnea was defined as a ≥ 30% reduction in airflow or breathing amplitude for 10 sec coupled with a 4% oxygen desaturation. Oxygen desaturation events were automatically scored from the oxygen saturation channel and manually edited for artifact. Interscorer and intrascorer reliability for the reported summary measures was excellent, with intraclass correlation coefficients exceeding 0.90 for key parameters.

Our primary exposure, the obstructive AHI, was defined as the sum of obstructive apneas plus hypopneas divided by total sleep time (hours) and was categorized using clinical cutoffs; normal (AHI < 5), mild (5 to < 15), moderate (15 to < 30), moderately severe (30 to < 50) and severe (≥ 50).²⁷ Our secondary exposures included the arousal index and average oxygen saturation (SaO₂). An arousal was defined as an abrupt shift in EEG frequency lasting at least 3 sec, and occurring after 10 sec of continuous sleep. Arousal index was calculated as number of arousals per hour of sleep over total sleep time. Average oxygen saturation represents the average saturation over the entire sleep period.

Cardiac MRI

cMRI was performed with 1.5-T magnets using steady state free precession cine images with temporal resolution less than or equal to 40 msec. Assessment of LV mass, volumes and function was performed using CIM software (version 6.2, University of Auckland, New Zealand)²⁸ at a central reading center (Johns Hopkins University, Baltimore, MD). The difference between the epicardial and endocardial areas for all slices was multiplied by the slice thickness and section gap which was then further multiplied by the specific gravity of myocardium (1.04 g/mL) to derive the ventricular mass. Our primary outcome was LV mass. LV mass was adjusted for body height in meters. Our secondary outcomes were LV mass to volume ratio (LVMVR) and left ventricular ejection fraction (LVEF). LVMVR was derived by dividing the LV mass in grams by end-diastolic volume in milliliters.

Actigraphy

Measures of sleep efficiency and sleep duration were obtained using the Actiwatch Spectrum wrist actigraph (Philips Respironics, Murrysville, PA) worn for 7 consecutive days and output was scored at the centralized sleep reading center using a self-actuated event marker, sleep diary, and light sensor. Using Actiware-Sleep version 5.59 analysis software (Mini Mitter Co., Inc., Bend, OR) and a validated sleep/wake algorithm,²⁹ sleep duration, and efficiency were calculated with intrascorer reliability of 0.91 and 0.97, respectively. Sleep efficiency was defined as sleep time divided by time in bed and sleep duration was defined as total time spent asleep.

Blood Pressure

Resting blood pressure (BP) was measured in the right arm in triplicate after 5 min in a seated position with a Dinamap automated device (Model PRO 100, GE Healthcare). The average of the second and third measurement was used for analyses.

Covariates

Diabetes was defined as glucose level ≥ 126 and/or use of diabetic medication. Alcohol use was defined as currently drinking alcohol (yes/no), sleep duration as total time spent asleep, and sleep efficiency as total sleep time divided by time in bed. CVD was defined as prevalent CVD at Exam 5 (as all participants were free of clinically overt CVD at recruitment) and includes having a myocardial infarction, angina, resuscitated cardiac arrest, or stroke. Systolic blood pressure, diastolic blood pressure, and use of antihypertension medication were also included as covariates. Hypertension was defined as systolic blood pressure ≥ 140, diastolic blood pressure ≥ 90, and/or use of antihypertension medication.

Statistical Analysis

Between-group differences were assessed with the Fisher exact test for categorical variables, the two-sample t test for normally distributed variables, and the Wilcoxon rank-sum test for non–normally distributed continuous measures. Continuous cMRI outcomes were modeled using multivariate linear regression models. Models were adjusted for age, sex, waist circumference, race, alcohol use, systolic BP, diastolic BP, antihypertension medication use, and diabetes. Additional covariates included actigraphy-derived sleep duration and sleep efficiency. LVEF was additionally dichotomized using a threshold of < 45% and modeled using logistic regression.

Additional regression models used nocturnal SaO₂ and arousal index as exposures. Analyses were also repeated after excluding subjects with LVEF < 45% and subjects with hypertension to address potential residual confounding. Given a priori hypotheses that results may vary across age, sex, and race, interactions were tested between all sleep exposures and each of these covariates. A significance level of 0.05 was used for main effects and 0.10 for interaction effects. SAS 9.3 (SAS Institute, Cary, NC) was used to conduct analyses.

Informed consent was obtained from all participants and Institutional Review Board approval was obtained from all study sites.

RESULTS

Table 1 describes sample characteristics overall and by AHI category. Mean age was 68 ± 8.8 y, 46.4% were men and, as designed, the sample was ethnically diverse consisting of 38% white, 13.1% Chinese, 26.4% black, and 22.5% Hispanic individuals. Approximately half of subjects had hypertension, slightly more than one-tenth had diabetes, and 22 had an ejection fraction (EF) < 45%. The median AHI was 13.8 (interquartile range 6.5; 26) and > 40% had moderate or severe OSA (AHI ≥ 15). Individuals with more severe OSA were more likely to be male, overweight, diabetic, hypertensive, and have elevated left ventricular mass indexed for height (LVHi), LVMVR, or lower LVEF.

Table 1.

Sample characteristics by apnea-hypopnea index category.

Table 2 shows sleep characteristics stratified by LVHi quartiles. In unadjusted analyses higher AHI, arousal index, and lower SaO₂ levels, sleep duration, and efficiency were significantly associated with increased LVHi. Results were similar after adjustment for age, sex, race, alcohol use, waist circumference, systolic and diastolic blood pressure, antihypertensive medication use, and diabetes for our primary exposure, obstructive AHI, as well as for hypoxemia, though not for arousal index which was no longer significantly associated with LVHi after covariate adjustment (data not shown).

Table 2.

Sleep characteristics by left ventricular mass/height quartiles.

Based on a priori hypotheses that associations may be stronger in men and minorities and prior evidence that associations may be stronger in younger participants,²⁴ interactions were tested between sleep exposures and age, sex, and race for each outcome. Significant interactions were observed between age and both AHI category (P interaction = 0.006, see Figure 1) and SaO₂ (P interaction = 0.0046) for LVHi. Interactions between sleep indices and sex were nonsignificant for all outcomes. Interactions between exposures and race did not meet prespecified significance criteria (P interaction < 0.10), but there was a tendency toward a steeper rise in LVHi by AHI category (P interaction = 0.13) in whites and Chinese compared to blacks and Hispanics (Figure 2). There were no significant interactions for LVMVR or LVEF.

Figure 1.

Left ventricular (LV) mass/height versus obstructive sleep apnea severity by age groups. This figure demonstrates a significantly steeper rise in LV mass/height with increasing apnea-hypopnea index (AHI) category in those age 65 y or younger, depicted by the blue line, compared to those older than 65 y as depicted by the red line (P interaction = 0.006). Dotted lines represent 95% confidence intervals.

Figure 2.

Left ventricular (LV) mass/height versus obstructive sleep apnea severity by race. This figure demonstrate a tendency toward a steeper rise in LV mass/height with increasing apnea-hypopnea index (AHI) category (P interaction = 0.13) in whites (blue line) and Chinese (red line) compared to blacks (green line) and Hispanics (brown line). Dotted lines represent 95% confidence intervals.

Table 3 presents results of linear regression models between AHI and average SaO₂ for LVHi, including age interactions. After adjustment for age, sex, race, alcohol use, waist circumference, systolic blood pressure, diastolic blood pressure, use of antihypertensive medication, and diabetes, AHI severity was significantly associated with increasing LVHi in those age 65 y or younger but not those older than 65 y. The model predicted that each elevation in AHI category was associated with a 1.82 ± 0.47 g/m increase in LVHi (P = 0.0001) in those age 65 y or younger. These results are also depicted in Figure 1, which demonstrates a steeper rise in LVHi by AHI category in those age 65 y or younger (P interaction = 0.006). Table 3 also shows a significant association between average SaO₂ and LVHi in those age 65 y or younger. Lower average SaO₂ was associated with increased LVHi in individuals younger than 65 y, such that for every 1% decrease in SaO₂, LVHi increased by 1.1 ± 0.34 g/m (P = 0.001).

Table 3.

Adjusted associations between sleep exposures and LV mass/ height.

After adjusting for confounders, LVMVR was not significantly associated with increasing AHI severity. However, SaO₂ was associated with LVMVR such that for every 1% decrease in SaO₂, LVMVR increased by 0.008 g/mL (β = −0.008 ± 0.004, P = 0.03). Additionally, after adjusting for confounders, there were no significant associations between any exposures and LVEF (not shown). When dichotomizing at LVEF < 45% (n = 22), adjusted logistic regression analyses also did not demonstrate any significant associations. There were no significant associations between arousal index and any cMRI outcomes (not shown).

Additional Analyses

The primary analyses were repeated after excluding 813 participants with hypertension, defined as anyone with systolic blood pressure ≥ 140, diastolic blood pressure ≥ 90, any self-report of hypertension diagnosis, and/or any use of antihypertension medication (n = 599). Among those age 65 y or younger in this subset, the relationship between AHI and LVHi remained significant (β = 1.8 ± 0.6; P = 0.006). Because diabetes and hyper-tension may be intermediary pathways linking OSA and LVHi, we compared models that did and did not adjust for diabetes, systolic BP, or anti-hypertension medications and found that the beta coefficient for AHI and LVHi increased by 13% (from 1.82 to 2.1) in those age 65 y or younger (P < 0.0001) when those covariates were not included, suggesting that diabetes and hypertension explain a portion of the association. In order to determine how much of this effect was from diabetes or hypertension, we then compared models adjusting for one versus the other. When we excluded only diabetes as a covariate the effect estimate increased from 1.82 to 1.84 (P = 0.0001) in those age 65 y or younger, and when we excluded systolic and diastolic blood pressure and antihypertension medication as covariates the effect estimate increased from 1.82 to 2.01 (P < 0.0001) suggesting that the majority of this effect is explained by hypertension. In those older than 65 y, associations remained nonsignificant. Addition of sleep duration or sleep efficiency as covariates did not change any results (not shown). Analyses that excluded subjects with EF < 45% and that adjusted for BMI rather than waist circumference revealed no appreciable changes.

DISCUSSION

In a large, multiethnic, community-based cohort studied with both PSG and cMRI, we demonstrated that increasing AHI category and hypoxia are significantly associated with increased LV mass in individuals age 65 y or younger. Results are consistent with prior echocardiography literature demonstrating significant associations between LV hypertrophy and OSA^{8,14,15,24,30–32} even in the absence of hypertension³² and diabetes,¹³ and extend the literature by identifying comparable associations in men and women and quantifying these changes with cMRI. We also demonstrate that average oxygen saturation is significantly associated with LVMVR. LVMVR, suggestive of concentric remodeling, is of particular interest given evidence that concentric ventricular remodeling is predictive of increased cardiovascular events and mortality^18,33 even in individuals with normal LV mass.³³ In addition, increased LVMVR may represent an early sign of heart failure (HF) with preserved EF, which is increasingly recognized to contribute to morbidity and mortality and is also associated with OSA.^1,34

Potential mechanisms underlying an association between OSA severity and LVHi and LVMVR include (1) increased LV afterload due to large negative intrathoracic pressure swings resulting from forceful inspiration against an occluded upper airway,³⁵ and (2) elevations in nocturnal BP³⁶ secondary to nightly surges in sympathetic activity from airway occlusion, sleep disruption and hypoxemia. Notably, results did not appreciably change after accounting for BP and diabetes. Although daytime BP may partly mediate the observed associations (explained approximately 9% of the association), these analyses suggest that OSA negatively influences cardiac structure by additional factors such as nocturnal BP, oxidative stress, or inflammation, which were not measured.

Our study is the first of this magnitude to demonstrate comparable associations between OSA severity and adverse LV structural changes in both men and women in a multiethnic population. Prior work has demonstrated associations in men but not women^13,14,31 or found stronger associations in men.²⁴ Our findings support the need to consider OSA as a contributor to cardiac dysfunction in women as well, particularly given that HF is the leading cause of death in women age 65 y and older.³⁷

Stronger associations were observed between AHI severity, nocturnal SaO₂ and LVHi in individuals age 65 y or younger compared to individuals older than 65 y. Possible explanations include age variation in OSA characteristics influencing susceptibility to CVD such as (1) magnitude of inspiratory effort generated with each apneic event; (2) patterns of apneas across the night; and (3) autonomic response to apneas which may be less pronounced in individuals older than 65 y. The weaker associations in older individuals may also reflect survival biases. Cross-sectional analyses from the SHHS also demonstrated that AHI > 30 and time with SaO₂ < 90% were more significantly associated with echocardiographically assessed LV mass and concentric hypertrophy in younger subjects.²⁴ However, results were weaker in women and those analyses did not differentiate obstructive from central apnea.

A stronger association between AHI and LVHi was suggested in whites and Chinese as compared to Hispanics and blacks, though statistical interactions were nonsignificant and should be viewed cautiously. Although limited by modest samples within race/ethnicity strata, the suggestive stronger associations in whites and Chinese compared to African Americans and Hispanics may indicate that risk factors for OSA and increased LV mass differ by racial group or that black and Hispanic groups may have a different distribution of risk factors for increased LV mass and support a need for further investigation of racial differences.

We observed no association between OSA severity and LVEF, though < 2% of the sample had LVEF < 45%, reflecting recruitment of a cohort free of overt CVD at baseline. Further cohort follow-up will clarify whether the observed cardiac structural changes influence incident HF.

Strengths of this study include use of a community-based sample of ethnically diverse women and men recruited 10 y earlier when free of CVD, making results generalizable to the US population at large and minimizing referral biases or reverse causality. Use of standardized, objective PSG measures and standardized cMRI reading reduced measurement error, reporting biases and increased precision. Use of obstructive AHI increased specificity for OSA.

Study limitations include the cross-sectional design, such that reported associations do not prove causality. Because it is unknown in how many subjects OSA developed during the course of the study, it is not clear that the structural changes developed in response to OSA. Measurements of daytime BP do not fully capture the hypertension burden associated with OSA,³⁸ and thus assessment of hypertension as a mediating factor is limited. Though our sample is the most diverse for a study of this nature and magnitude, we were not powered to detect modest interactions with race/ethnicity.

CONCLUSIONS

In an ethnically diverse community-based cohort, after adjusting for possible confounders, AHI level is significantly associated with increased LV mass and concentric ventricular remodeling, both of which predict CVD events and mortality.^18,33,39 Associations are not fully explained by daytime BP levels or antihypertensive medication use, and associations with LV mass are stronger in individuals age 65 years or younger. In contrast to prior studies, findings are comparable in women and men, suggesting that OSA contributes to sub-clinical cardiac disease in both groups. These results support the utility of assessing cardiac structure for both sexes in future OSA intervention studies aimed at CVD risk reduction.

DISCLOSURE STATEMENT

This was not an industry supported study. Funding was provided by NIH 5T32HL007901, 1R01HL083075, R01HL098433, R01 HL098433-02S1, 1U34HL105277-01, 1R01HL110068-01A1, and 1R01HL113338-01. Dr. Redline has received research support from Jazz Pharmaceuticals. The other authors have indicated no financial conflicts of interest.

ABBREVIATIONS

AHI

apnea-hypopnea index

SaO₂

average oxygen saturation

BP

blood pressure

cMRI

cardiac magnetic resonance imaging

CVD

cardiovascular disease

HF

heart failure

LV

left ventricular

LVEF

left ventricular ejection fraction

LVHi

left ventricular mass indexed for height

LVMVR

left ventricular mass volume ratio

MESA

Multi Ethnic Study of Atherosclerosis

OSA

obstructive sleep apnea

PSG

polysomnogram

SHHS

Sleep Heart Health Study