Association of Incident Cardiovascular Disease With Progression of Sleep-Disordered BreathingClinical Perspective
Background—Prospective data suggest that sleep-disordered breathing enhances risk for incident and recurrent cardiovascular disease (CVD). However, a reverse causal pathway whereby incident CVD causes or worsens sleep-disordered breathing has not been studied.
Methods and Results—A total of 2721 Sleep Heart Health Study participants (mean age 62, standard deviation=10 years; 57% women; 23% minority) without CVD at baseline underwent 2 polysomnograms 5 years apart. Incident CVD events, including myocardial infarction, congestive heart failure, and stroke, were ascertained and adjudicated. The relation of incident CVD to change in apnea-hypopnea index between the 2 polysomnograms was tested with general linear models, with adjustment for age, sex, race, study center, history of diabetes mellitus, change in body mass index, change in neck circumference, percent sleep time spent in supine sleep, and time between the 2 polysomnograms. Incident CVD occurred in 95 participants between the first and second polysomnograms. Compared with participants without incident CVD, those with incident CVD experienced larger increases in apnea-hypopnea index between polysomnograms. The difference in adjusted mean apnea-hypopnea index change between subjects with and without incident CVD was 2.75 events per hour (95% confidence interval, 0.26 to 5.24; P=0.032). This association persisted after subjects with central sleep apnea were excluded. Compared with participants without incident CVD, participants with incident CVD had greater increases in both mean obstructive and central apnea indices, by 1.75 events per hour (95% confidence interval, 0.10 to 1.75; P=0.04) and by 1.07 events per hour (95% confidence interval, 0.40 to 1.74; P=0.001), respectively.
Conclusions—In a diverse, community-based sample of middle-aged and older adults, incident CVD was associated with worsening sleep-disordered breathing over 5 years.
Weight, age, gender, neck circumference, and craniofacial morphology are risk factors for sleep-disordered breathing (SDB).1 Although weight gain was shown to be the strongest determinant of change in apnea-hypopnea index (AHI) in the Sleep Heart Health Study (SHHS), SDB progressed over 5 years even in subjects whose weight remained stable or decreased.2 Thus, additional factors may explain the unfavorable progression of SDB.
Editorial see p 1265
Clinical Perspective on p 1286
Cross-sectional studies have linked SDB to cardiovascular disease (CVD) including coronary artery disease, congestive heart failure (CHF), and stroke.3,–,10 Although stroke and heart failure are known risk factors for central SDB,11,12 whether CVD contributes to the development or progression of obstructive SDB is unclear. The cross-sectional association between CVD and SDB is most frequently approached with the assumption that obstructive SDB enhances the risk of CVD. Although prospective data support this idea,13,–,16 these data do not preclude a reverse causal pathway in which obstructive SDB may, in some cases, be the consequence of CVD rather than the cause. There is growing evidence that ventilatory control instability may contribute to the pathogenesis of obstructive as well as central SDB.17,–,19 Cardiac dysfunction is associated with ventilatory instability20,21 through mechanisms that may include enhanced chemoreflex sensitivity to carbon dioxide21 and stimulation of pulmonary irritant receptors by pulmonary congestion,22 suggesting a mechanism whereby cardiac disease might cause or worsen obstructive SDB.
In this report we seek to determine whether incident myocardial infarction (MI) or CHF causes or worsens obstructive SDB.
The SHHS is a cohort of individuals aged ≥40 years recruited from participants in several ongoing cohort studies of cardiovascular and obstructive pulmonary diseases in the United States. Details of the design and quality control procedures were reported previously.23 In this report, we examine SHHS participants who had 2 polysomnograms performed ≈5 years apart.
Of the 5681 SHHS participants with a baseline polysomnogram that met quality standards, 783 (13.7%) were excluded because they had prevalent CVD. Of the remaining participants, 2850 (58%) had a follow-up polysomnogram ≈5 years after baseline. Of these, 7 (0.2%) had surgery for obstructive sleep apnea, and 122 (4%) were missing covariate data. The analysis sample therefore consisted of 2721 individuals.
The study was approved by the institutional review board at all participating sites.
SHHS participants underwent in-home polysomnography with the use of the Compumedics P-series portable monitor (Abbotsford, Victoria, Australia). The polysomnogram recordings included the following channels: electroencephalogram, electrooculogram, ECG, chin electromyogram, pulse oximetry, chest and abdominal excursion, airflow (by thermocouple), and body position. Polysomnogram recordings were analyzed and scored centrally with the use of scoring guidelines and quality-assurance procedures that have been published.24,25 As reported previously, the intraclass correlation coefficient of AHI in the SHHS was >0.9 for interscorer reliability25 and 0.8 for night-to-night variability.26 Apneas were identified if there was a decrease in airflow amplitude to <25% of baseline lasting for at least 10 seconds. Hypopneas were defined as a decrease in airflow or chest wall movement amplitude to <70% of baseline lasting for at least 10 seconds. Hypopneas required an associated 3% oxygen desaturation. Central apneas were scored if there was no evidence of effort from both the abdominal and thoracic channels. Obstructive apneas were scored if effort persisted in either channel. The AHI was computed as the number of apneas plus hypopneas per hour of sleep associated with at least a 3% drop in oxyhemoglobin saturation. The central apnea index (CAI) was calculated as the number of central apneas per hour of sleep regardless of associated oxyhemoglobin desaturation. The obstructive apnea index (OAI) was defined as the number of obstructive apneas per hour of sleep regardless of associated oxyhemoglobin desaturation. Alternative definitions of the AHI, CAI, and OAI were used in sensitivity analyses requiring that apneas and hypopneas be associated with at least a 4% unit drop in oxyhemoglobin desaturation. The follow-up polysomnogram was performed a mean of 5 years and 89 days (SD=98 days) after the baseline polysomnogram.
During the baseline home visit, a study technician collected information on medical history using a standardized questionnaire. On the night of the baseline and follow-up polysomnograms, weight was measured with the use of a portable scale with the participant in light clothes. Neck circumference was measured just below the laryngeal prominence. Height was obtained at the baseline home visit only if not already measured within 3 months in the parent cohort. Information on gender, race, and history of diabetes mellitus was obtained from parent cohorts. Body mass index (BMI) was calculated as weight in kilograms divided by the square of height in meters. The change in AHI, BMI, and neck circumference between baseline and follow-up was calculated as follow-up level minus baseline level.
The association of incident CVD to AHI change was evaluated by multiple linear regression (Proc GLM). The dependent variable for the main analysis was change in AHI. Although the distribution of AHI was skewed, change in AHI was normally distributed, which allowed treating it as a continuous variable without logarithmic transformation. Change in CAI and change in OAI were the dependent variables in secondary analyses. The primary exposure variable for the main analysis was incident CVD between the baseline and follow-up polysomnograms (yes/no). Incident MI and CHF between the baseline and follow-up polysomnograms were analyzed separately in supplementary analyses.
The results are presented unadjusted and in a model adjusted for demographic variables (age, sex, race, and study site), change in BMI, change in neck circumference, history of diabetes mellitus, length of follow-up, and difference in percent sleep time spent in the supine position. Models were further adjusted for change in medication use including estrogen, antihypertensive medications, benzodiazepine, and antidepressant therapy and for the CVD risk factors blood pressure, cholesterol level, and smoking.
The main analysis was repeated with apneas and hypopneas associated with at least a 4% drop in oxyhemoglobin saturation for calculation of the AHI, CAI, and OAI.
The main analysis was stratified by baseline BMI (<25, 25 to 30, and >30 kg/m2), by weight change categories (increase in weight >5 kg, stable weight [change 0 to 4.9 kg], decrease in weight >5 kg), and by presence of obstructive sleep apnea at baseline (baseline AHI <5 or ≥5). Additional analyses examined interactions between incident CVD and weight change, between incident CVD and sex, and between incident CVD and both baseline AHI and BMI. Analyses were performed with the exclusion of subjects with central sleep apnea, defined as baseline CAI >5. Multinomial logistic regression was used to determine the effect of incident CVD on categorical AHI change with stable AHI used as the reference group. In this analysis, AHI change was defined as 5 ordinal categories: increase in AHI >15 events per hour, increase in AHI by 5 to 15 events per hour, stable AHI (change of −4.9 to 4.9), decrease in AHI by 5 to 15 events per hour, and decrease in AHI >15 events per hour. Logistic regression was also used to compare the odds of significant SDB progression (defined as change in AHI from baseline of ≥5 events per hour) between subjects with incident CVD and subjects without.
The association of AHI change and incident CVD that occurred after the second polysomnogram was evaluated in supplemental analyses with multiple linear regression and logistic regression. Statistical analyses were performed with SAS, version 9.1 (SAS Institute Inc., Cary, NC).
Characteristics of the study sample are shown in Table 1. Ninety-five subjects had incident CVD events over an average follow-up of 5.2 (SD=0.27) years. Fifty-seven subjects had MI, and 57 subjects had CHF. Participants who had incident CVD between the 2 polysomnograms were older on average and were more likely to be male, to be diabetic, and to lose weight over the length of follow-up (Table 1).
The mean change in AHI was larger in subjects who had an incident CVD event compared with those who did not (Table 2). This difference persisted and remained statistically significant after adjustment for demographic variables, change in BMI, history of diabetes mellitus, change in neck circumference, length of follow-up, and percent sleep time in the supine sleep position (Table 2). Further adjustment for change in medication use and cardiovascular risk factors led to further exclusion of 118 subjects, with no meaningful impact on the observed association (difference in mean AHI change of 2.80; 95% confidence interval [CI], 0.26 to 5.34; P=0.03) between subjects with and without incident cardiovascular events.
Findings were similar in a sensitivity analysis in which apneas and hypopneas associated with ≥4% (rather than ≥3%) oxyhemoglobin desaturation were used to calculate the AHI (results not shown). Consistent with the linear models, subjects with incident CVD had higher odds of a categorical increase in AHI over the duration of follow-up compared with subjects without incident CVD, with an adjusted odds ratio of 1.65 (95% CI, 1.11 to 2.43; P=0.01) for AHI change being in 1 higher ordinal level. Furthermore, subjects with incident CVD were more likely to experience significant SDB progression (defined as change in AHI of ≥5 events per hour) compared with subjects without incident CVD (47.4% versus 31.7%; P=0.001). The odds ratio of significant SDB progression in subjects with incident CVD compared with subjects without was 1.94 (95% CI, 1.28 to 2.92; P=0.001) in the unadjusted model and 1.65 (95% CI, 1.07 to 2.92; P=0.02) in the adjusted model.
No evidence of interaction between CVD and either gender or weight change was observed.
To assess whether the association of incident CVD and SDB progression is different between subjects with and without SDB at baseline, the main analysis was stratified by baseline AHI <5 or ≥5. The difference in adjusted mean AHI change between subjects with incident CVD and subjects without incident CVD was 3.25 events per hour (95% CI, −0.04 to 6.54; P=0.053; n=1805) in the group with baseline AHI ≥5 and was 1.3 events per hour (95% CI, −1.56 to 4.18; P=0.3; n=916) in the group with baseline AHI <5 (Figure, panel A). However, the interaction term between incident CVD and baseline AHI was nonsignificant.
The main analysis was also stratified by baseline BMI. The difference in adjusted mean AHI change between subjects with incident CVD and subjects without incident CVD was 7.62 (95% CI, 2.52 to 12.72; P<0.003; n=673), 0.84 (95% CI, −2.33 to 4.01; P=0.6; n=1113), and 3.4 (95% CI, −2.23 to 9.18; P=0.23; n=935) in the groups with baseline BMI <25, 25 to 29.9, and ≥30, respectively (Figure, panel B). However, the interaction term between incident CVD and baseline BMI was nonsignificant.
When MI and CHF were considered individually, the difference in AHI was more pronounced and statistically significant among subjects with incident MI (Table 3). To assess whether the observed association was driven by CVD inducing central sleep apnea, the mean changes in CAI and OAI were analyzed separately, and the main analysis was repeated with the exclusion of subjects with predominantly central sleep apnea. Mean change in both CAI and OAI was larger in subjects who had incident CVD (Table 4). The magnitude of difference in mean OAI change was larger than that in mean CAI change. The findings were similar in a sensitivity analysis in which a definition of apneas requiring a 4% desaturation to calculate the CAI and OAI was used (results not shown). The findings from the main analysis were not meaningfully altered after exclusion of 18 subjects with baseline CAI ≥5.
To further evaluate whether worsening SDB was a cause of incident CVD or whether the association reflects common risk factors, change in AHI among subjects who had incident CVD after the second polysomnogram was compared with that among subjects without incident CVD. No difference was observed between the 2 groups (Table 5). The odds ratio of incident CVD after the follow-up polysomnogram as a function of change in AHI was 1.00 (95% CI, 0.98 to 1.14).
In an ethnically diverse, community-based sample of middle-aged and older adults, incident CVD was associated with worsening SDB as measured by change in AHI after adjustment for age, sex, race, site, change in BMI, change in neck circumference, history of diabetes mellitus, length of follow-up, and difference in percent sleep time spent in the supine sleep position and with further adjustment for baseline blood pressure, cholesterol, smoking, and change in medication use. This finding persisted when the OAI was used as a measure of SDB and when subjects with central sleep apnea at baseline and subjects whose central sleep apnea worsened more than obstructive sleep apnea were excluded from the analysis. These findings suggest that not only is SDB a risk factor for CVD, but CVD can potentially exacerbate SDB. The magnitude of the observed association is clinically relevant because a significantly larger proportion of subjects with incident CVD experienced a sizable progression in SDB compared with subjects without. Although a test of interaction does not support effect modification by presence of sleep apnea at baseline, the observation that incident CVD is associated with increased AHI in subjects with baseline AHI ≥5 but not in subjects with AHI <5 suggests that incident CVD could worsen SDB in subjects already affected by SDB but may not cause SDB de novo. The observation that incident CVD is associated with increased AHI in subjects with baseline BMI <25 but not in subjects with BMI ≥25 suggests that obesity may overwhelm the effect of CVD on SDB progression in obese and overweight individuals.
Although a number of studies have explored the association of SDB at baseline with incident CVD, to our knowledge this is the first study to examine whether incident CVD might affect the occurrence or severity of SDB. Although the longitudinal prospective design of SHHS makes it likely that incident CVD is the cause of the worsening SDB that we observed, the possibility that progression of SDB is simply a marker of increased likelihood of incident CVD due to common risk factors cannot be discounted. However, the lack of association between change in AHI and incident CVD after the follow-up polysomnogram provides strong evidence against this interpretation.
It is unclear from our analysis why incident CVD is associated with worsening of SDB. One factor that has been implicated in recurrent collapse of the upper airway in obstructive SDB is ventilatory control instability.17,–,19 Loop gain, a measure of ventilatory instability, has been found to be increased in patients with obstructive SDB19 and correlates with SDB severity in patients with intermediate anatomic airway collapsibility.17 Because cardiac dysfunction can cause ventilatory control instability,20,21 a possible explanation is that development of CVD in some persons with SDB results in ventilatory control instability and consequently worsening of both obstructive and central SDB. Stimulation of pulmonary irritant receptors by pulmonary congestion22 and increased chemoreceptor sensitivity to carbon dioxide21 are key factors that predispose to ventilatory control instability in patients with cardiac dysfunction by inducing chronic hyperventilation and hypocapnea, and they may be mediating the association between incident CVD and progression of SDB observed in this study. In light of this proposed mechanism, we expected incident CHF to have a strong association with change in AHI. In contrast, the association of change in AHI with CHF was only half as large as its association with incident MI. Indeed, the association of CVD with change in AHI was statistically significant only in subjects with incident MI. Another possible mechanism relates to rostral interstitial fluid shifts during sleep, which have been associated with worsening SDB across the night in patients with heart failure.34 Because incident cardiac dysfunction may be associated with interstitial fluid buildup, it may lead to worsening SDB as the fluid moves rostrally during sleep.
Several issues need to be considered when the results of our analyses are interpreted. First, only 57% of the SHHS subjects underwent follow-up polysomnograms. This could have biased the results if the association of interest was different in subjects who did not participate in the follow-up polysomnogram. Subjects who did not have a follow-up polysomnogram were more likely to have incident CVD over the duration of the study (15.5% versus 9.4%) and were on average slightly older (64 versus 62 years). However, they did not differ with respect to baseline SDB severity, sex distribution, and diabetes mellitus and with respect to both baseline and change in BMI and neck circumference. Second, central and obstructive apnea indices could not be derived in all subjects because of technical issues that are not influenced by CVD risk factors. Third, night-to-night variability in AHI measurement accounts for some of the change in AHI between the 2 polysomnograms. However, this variability is likely nondifferential and would thus be expected to bias toward the null. Finally, despite adjustment for several covariates, unknown confounders could account for the observed association.
Balancing these limitations are several strengths, including the large ethnically diverse community-based sample, standardized polysomnogram measures obtained according to strict protocols and rigorous quality control measures, and the prospective ascertainment of covariates and rigorous adjudication of CVD events following explicit guidelines.
In conclusion, we found that incident CVD in a community-based sample is associated with the worsening of preexisting obstructive and central SDB after adjustment for known risk factors for obstructive SDB. The lack of association between the rate of worsening SDB and incident CVD that occurred after the follow-up polysomnogram suggests that this association may be causal. These findings further our understanding of the association between SDB and CVD and suggest that this association may be bidirectional.
Sources of Funding
This work was supported by National Heart, Lung, and Blood Institute cooperative agreements U01HL53940 (University of Washington), U01HL53941 (Boston University), U01HL53938 (University of Arizona), U01HL53916 (University of California, Davis), U01HL53934 (University of Minnesota), U01HL53931 (New York University), U01HL53937 and U01HL64360 (Johns Hopkins University), U01HL63463 (Case Western Reserve University), and U01HL63429 (Missouri Breaks Research).
The SHHS acknowledges the Atherosclerosis Risk in Communities Study, the Cardiovascular Health Study, the Framingham Heart Study, the Cornell Worksite and Hypertension Studies, the Strong Heart Study, the Tucson Epidemiology Study of Airways Obstructive Diseases, and the Tucson Health and Environment Study for allowing their cohort members to be part of the SHHS and for permitting data acquired by them to be used in the study. The SHHS is particularly grateful to the members of these cohorts who agreed to participate in SHHS as well. The SHHS further recognizes all of the investigators and staff who have contributed to its success. A list of SHHS investigators, staff, and their participating institutions is available on the SHHS Web site, http://www.jhucct.com/shhs. This study included data on participants covered by the Indian Health Service (U.S. Department of Health and Human Services). The opinions expressed in this article are those of the authors and do not necessarily reflect the views of the Indian Health Service.
Continuing medical education (CME) credit is available for this article. Go to http://cme.ahajournals.org to take the quiz.
Guest Editor for this article was Paul W. Armstrong, MD.
- Received June 22, 2010.
- Accepted January 11, 2011.
- © 2011 American Heart Association, Inc.
- Sanner B,
- Konermann M,
- Weiss T,
- Kreuzer I,
- Laschewski F,
- Burman-Urbaneck M,
- Sturm A
- Sun SY,
- Wang W,
- Zucker IH,
- Schultz HD
- Javaheri S
- Lorenzi-Filho G,
- Azevedo ER,
- Parker JD,
- Bradley TD
- Toole JF,
- Lefkowitz DS,
- Chambless LE,
- Wijnberg L,
- Paton C,
- Heiss G
- Fried LP,
- Borhani NO,
- Enright P,
- Furberg CD,
- Gardin JM,
- Kronmal RA,
- Kuller LH,
- Manolio TA,
- Mittelmark MB,
- Newman A
- Price TR,
- Psaty B,
- O'Leary D,
- Burke G,
- Gardin J
- Kannel WB,
- Polf PA,
- Garrison RJ
- Cupples LA,
- D'Agostino RB
- Lee E,
- Cowan L,
- Welty T,
- Sievers M,
- Howard WJ,
- Oopik A,
- Wang W,
- Yeh J,
- Devereux RB,
- Rhoades ER,
- Fabsitz RR,
- Go O,
- Howard BV
- Yumino D,
- Redolfi S,
- Ruttanaumpawan P,
- Su M,
- Smith S,
- Newton G,
- Mak S,
- Bradley TD
Sleep-disordered breathing has been associated with incident cardiovascular disease. The possibility that incident cardiovascular disease may cause or worsen sleep-disordered breathing (eg, by altering ventilatory control) has not been evaluated. We assessed this possibility in the community-based Sleep Heart Health Study. In this study, incident cardiovascular disease was associated with a more rapid progression of both obstructive and central sleep-disordered breathing over a 5-year follow-up period. This association is most pronounced in subjects with body mass index <30 and apnea-hypopnea index >5 events per hour at baseline. This finding suggests that the association between sleep-disordered breathing and cardiovascular disease may be bidirectional.