Predictors of slow-wave sleep in a clinic-based sample

Summary

Slow-wave sleep has been associated with several physiological phenomena, including glucose metabolism, sympathetic nervous system activity, hormonal secretion and blood pressure regulation. The aim of these analyses was to determine which sociodemographic and medical factors were associated with slow-wave sleep duration in a large clinical sample. We conducted cross-sectional analysis of clinical data from 1019 consecutive adults over a 10-month period who had their first in-laboratory polysomnogram for suspicion of obstructive sleep apnea. Patients either underwent in-laboratory full-night polysomnogram followed by full-night continuous positive airway pressure titration or split-night polysomnogram. Patients also completed questionnaires to assess race, education, marital status and medical co-morbidities. A multiple linear regression model that predicted the natural log of slow-wave sleep in minutes indicated that African Americans had approximately 48% less slow-wave sleep than non-African Americans. Increasing age and male gender were also associated with less slow-wave sleep. Overweight and obese individuals had significantly less slow-wave sleep than those not overweight, even after adjustment for obstructive sleep apnea severity. Finally, those with severe obstructive sleep apnea had significantly less slow-wave sleep than those with less severe obstructive sleep apnea even after adjustment for obesity. Results remained unchanged when patients who had a split-night polysomnogram were excluded. We observed less slow-wave sleep in African Americans, a group at increased risk of diabetes and hypertension compared with Caucasians, and in those who are overweight and obese and those with severe obstructive sleep apnea. Future research needs to explore potential reasons for reduced slow-wave sleep in these individuals.

Introduction

Sufficient, good quality sleep is important for a variety of health domains, but it appears that not all sleep stages contribute equally. In particular, slow-wave sleep (SWS), the deep stages of non-rapid eye movement (non-REM) sleep, has been associated with glucose metabolism, decline in sympathetic nervous system activity compared with quiet wakefulness and lighter stages of non-REM sleep and the secretion of certain hormones (Hornyak et al., 1991; Somers et al., 1993). For example, release of growth hormone is temporally linked to SWS (Van Cauter et al., 2000), and SWS is also associated with a decrease in cortisol secretion (Bierwolf et al., 1997). In addition, two experimental studies that suppressed SWS, intentionally or unintentionally, observed significantly reduced insulin sensitivity after SWS suppression (Stamatakis and Punjabi, 2010; Tasali et al., 2008). Notably, total sleep time remained similar in both studies. Both studies also demonstrated that SWS suppression was associated with increased sympathetic nervous system activity based on spectral analysis of heart rate variability. Chronically elevated sympathetic nervous system activity is associated with increased blood pressure, which could predispose individuals to the development of hypertension and cardiovascular disease (Malpas, 2010). Poor sleep quality with increased time spent in lighter stages of non-REM sleep has been associated with blunted nocturnal blood pressure dipping (Matthews et al., 2008), a phenomenon that is an independent predictor of cardiovascular mortality (Fagard et al., 2009; Hansen et al., 2011). Indeed, SWS is an important predictor of nocturnal blood pressure dipping in healthy subjects (Loredo et al., 2004). When compared with Caucasians, African Americans have a blunted nocturnal blood pressure dipping independent of age, gender or sleep-disordered breathing (Ancoli-Israel et al., 2002; Routledge and McFetridge-Durdle, 2007; Sherwood et al., 2002). Therefore, non-dipping in African Americans can be partly mediated by reduced sleep quality and SWS. Lastly, during SWS the upper airway dilator muscle activity is increased (Basner et al., 1991). This improvement in upper airway tone can lead to a more stable breathing pattern and a reduction in obstructive respiratory events compared with stage 2 non-REM sleep and REM sleep. Together, these data suggest that SWS may play an important role in the risk of cardiometabolic diseases, such as hypertension or diabetes. Therefore, it is important to identify factors that may limit time spent in SWS.

Although SWS is remarkably stable within individuals, it can vary greatly between individuals (Tan et al., 2001). Age and gender differences in SWS have been well documented. Aging is associated with a major decline in SWS beginning at least by midlife (Redline et al., 2004; Van Cauter et al., 2000), and some have shown a linear decline in amount of SWS between ages 20 and 59 years (Carrier et al., 1997). Sex differences have also been observed, such that women exhibit higher levels of SWS than men, at least up until menopause (Dijk et al., 1989; Redline et al., 2004). African American race has also been associated with reduced SWS in several previous studies, but covariates included in these analyses have varied. For example, two small studies, which included a total of 96 subjects, and the Sleep Heart Health Study found a lower percentage of SWS in African Americans as compared with Caucasians (Profant et al., 2002; Redline et al., 2004). All of these analyses adjusted for respiratory disturbance index, but there was no adjustment for socioeconomic status. A third study that examined the sleep of 27 Caucasians and 24 African Americans reported lower levels of SWS in African Americans, and they included a measure of social class, but social class did not actually differ between the two race groups (Stepnowsky et al., 2003). The Pittsburgh SleepSCORE project examined the impact of both race and socioeconomic status on SWS, and socioeconomic status had no effect on the racial differences in SWS (Mezick et al., 2008). This study, however, did not adjust for body mass index (BMI). The SWAN Sleep Study, which included only older women (average age was 50 years), also found lower SWS percentage in African Americans (Hall et al., 2009). This association persisted after adjustment for both educational attainment and financial strain; however, they did not include a measure of sleep-disordered breathing (Hall et al., 2009). Finally, a recent study of 128 adults found that the lower amount of SWS in African Americans remained after adjustment for both childhood socioeconomic status and current adult socioeconomic status (Tomfohr et al., 2010). Thus, although all these studies have found less SWS among African Americans, the covariates for which these different studies adjusted were not consistent.

Given the well-known racial disparities in cardiometabolic diseases, such as obesity, diabetes and hypertension (Hall et al., 2003), the potential racial association with SWS is important to examine. The aim of the present analyses was to determine whether racial differences persist after adjusting for several sociodemographic and health-related factors in a large clinical sample of men and women who were studied in the Sleep Laboratory at the University of Chicago.

Materials and Methods

Sample

Our sample is drawn from clinical data from 1019 consecutive adults over a 10-month period who were referred to the University of Chicago Sleep Laboratory for their first in-laboratory polysomnogram (PSG). We excluded patients who were missing information about their race (n = 61) and who were missing information about their insurance status (n = 4), which left a final analytic sample of 954 patients. The primary reason for PSG referral was suspicion of sleep apnea (94%), insomnia (3%), periodic limb movement disorder (2%) and all others (1%). Analysis of these clinical data was approved by the University of Chicago Institutional Review Board.

PSG

Patients either underwent in-laboratory full-night PSG followed by full-night continuous positive airway pressure titration or split-night PSG. Split-night PSGs were performed on select patients whose first 2–3 h of sleep had an apnea–hypopnea index (AHI) ≥ 30 as long as there were at least 3–4 h of titration time remaining. Split-night PSGs were performed in 263 (27.6%) patients. The PSG included standard recordings of electroencephalograms, bilateral electrooculograms, chin and leg electromyograms, oxygen saturation by pulse oximetry (3-s average signaling time), airflow with oronasal thermistor and nasal cannula pressure transducer, snoring, chest and abdominal respiratory effort with the use of either thoracoabdominal piezoelectric belts or respiratory inductance plethysmography.

Polysomnograms were scored according to the 2007 American Academy of Sleep Medicine Manual for the Scoring of Sleep and Related Events (Iber et al., 2007). Obstructive apneas were identified by a decrease in the oronasal airflow by at least 90% from baseline for at least 10 s in the presence of respiratory effort. Hypopneas were scored if the magnitude of ventilation signal decreased by at least 50% of the baseline amplitude of the nasal pressure transducer for at least 10 s, and were associated with either at least a 3% drop in oxygen saturation as measured by finger pulse oximetry or an electroencephalogram microarousal. The AHI was calculated as the total number of apneas and hypopneas per hour of sleep. Severity of obstructive sleep apnea (OSA) was defined as follows: none, AHI < 5; mild, AHI 5–14; moderate, AHI 15–29; severe, AHI ≥ 30.

Covariates

On the night of the PSG, patients completed a questionnaire that included questions on race, education, marital status and medical co-morbidities. The provided categories of race in the questionnaire included: African American, Caucasian, Hispanic/Latino, Asian/Pacific Islander, or other. For the purposes of analysis race was categorized into African American and non-African American. Education level was categorized as high school or less, more than high school, and unknown/unreported. Marital status was categorized into three categories: married; unmarried, which included never married; divorced and widowed; and unknown/unreported. Co-morbidities and medication use were self-reported as either present or absent. In these analyses we examined self-reported hypertension, diagnosed anxiety and use of antidepressants (all categorized as yes or no), which was reported on the questionnaire. The Epworth Sleepiness Scale was also administered to patients; higher scores indicate greater daytime sleepiness (Johns, 1991, 1992). Finally, the questionnaire asked the patient to estimate how much sleep they usually obtained each night. The patient’s age, sex and medical insurance type were obtained from the patient’s electronic medical record. Medical insurance was categorized into Medicaid and non-Medicaid. The patient’s BMI was calculated using weight and height measurements that were taken the day of the PSG. BMI was categorized into the following categories: not overweight, BMI < 25 kg m⁻²; overweight only, BMI 25–< 30 kg m⁻²; obese, BMI ≥ 30 kg m⁻².

Statistical analysis

Demographic, medical and PSG data were analysed for the full sample and separately by the two race groups (African American and non-African American). Differences between the racial groups were tested using either the Student’s t-test (continuous variables) or the chi-square test (categorical variables). Finally, a multiple linear regression model was estimated to predict minutes of SWS, but because this variable was right skewed we used the log-transformed value of SWS in minutes as the dependent variable. Covariates in this model included all the demographic and medical information, as well as total recording time from the PSG data and severity of OSA. All analyses were performed using Stata (v10.1, College Station, TX, USA).

Results

Table 1 describes the demographics of the sample as a whole and by racial group. For the full sample, mean (SD) age was 49 (15) years and ranged from 18 to 88 years. Over half our sample was women and 61% was African American. Most of the demographic and health-related variables differed significantly by race. The African Americans in our sample included a greater proportion of women, were more likely to be obese, were less likely to be married, had less education, were more likely to have Medicaid insurance and were more likely to report having hypertension. The non-African Americans were more likely to report anxiety and antidepressant use. Despite the greater prevalence of obesity in the African American group, there was no difference in OSA severity by race. Table 2 presents the sleep architecture data from the PSG and two self-reported measures of sleep. In terms of absolute minutes of each stage, the only significant difference was that African Americans had fewer minutes of SWS. However, the percentages of both stage 2, which was higher in African Americans, and SWS, which was lower in African Americans, were significantly different.

Table 1.   Description of sociodemographic and health characteristics of full-sample and by race

Variable	Full sample	African Americans	Non-African Americans	P-value
OSA, obstructive sleep apnea.
n	954	580	374
Age (years); mean ± SD	49.1 ± 15.0	49.6 ± 15.1	48.4 ± 14.7	0.21
Women (%)	57.0	63.8	46.5	< 0.001
Weight category (%)
Not overweight	10.2	5.7	17.1	< 0.001
Overweight	19.5	12.8	30.0
Obese	70.3	81.6	52.9
OSA severity (%)
None	9.0	7.6	11.2	0.27
Mild	21.4	21.6	21.1
Moderate	24.2	24.1	24.3
Severe	45.4	46.7	43.3
Marital status (%)
Married	38.7	27.4	56.2	< 0.001
Unmarried	50.1	60.2	34.5
Unknown	11.2	12.4	9.4
Education (%)
≤ High school	27.9	33.5	19.3	< 0.001
≥ Some college	70.4	65.2	78.6
Unknown	1.7	1.4	2.1
Insurance (%)
Non-Medicaid	77.2	67.1	92.8	< 0.001
Medicaid	22.9	32.9	7.2
Self-reported hypertension (%)	54.3	62.3	42.0	< 0.001
Self-reported anxiety (%)	18.2	15.1	23.0	0.002
Self-reported antidepressant use (%)	19.9	15.4	26.7	< 0.001

Table 2.   Description (mean ± SD) of PSG and self-reported sleep characteristics of full-sample and by race

Variable	Full sample	African Americans	Non-African Americans	P-value
REM, rapid eye movement sleep; SWS, slow-wave sleep; TRT, total recording time; TST, total sleep time; WASO, wake after sleep onset.
n	954	580	374
TRT (min)	412.9 ± 48.7	411.1 ± 44.7	415.7 ± 54.3	0.16
TST (min)	314.9 ± 76.5	314.2 ± 72.7	316.0 ± 82.2	0.72
Stage 2 (min)	173.0 ± 69.4	172.1 ± 69.2	174.4 ± 69.9	0.62
Stage 2 (%)	64.2 ± 13.4	65.2 ± 13.2	62.6 ± 13.6	0.004
SWS (min)	22.4 ± 28.6	18.2 ± 24.7	28.9 ± 32.6	< 0.001
SWS (%)	8.0 ± 9.9	6.9 ± 9.5	9.8 ± 10.2	< 0.001
REM (min)	38.4 ± 31.5	37.9 ± 30.7	39.1 ± 32.8	0.56
REM (%)	12.7 ± 8.8	12.9 ± 8.8	12.4 ± 8.8	0.44
WASO (min)	73.3 ± 53.6	72.8 ± 52.2	74.1 ± 55.6	0.71
Sleep efficiency	75.2 ± 16.5	75.4 ± 16.1	74.8 ± 17.0	0.58
Self-reported habitual sleep duration (h)	7.0 ± 1.8	6.8 ± 2.0	7.2 ± 1.5	0.0006
Epworth Sleepiness Score	9.0 ± 5.2	9.5 ± 5.4	8.3 ± 4.8	0.0008

Table 3 presents the results from the multiple linear regression model that predicted the natural log of SWS in minutes. Even after adjustment for the sociodemographic and health-related variables, African Americans had approximately 48% less SWS than non-African Americans. Increasing age and male gender were also associated with less SWS. Individuals who were overweight and who were obese had significantly less SWS than those who were not overweight, even after adjustment for OSA severity. Finally, those with severe OSA had significantly less SWS than those with less severe or no OSA, even after adjustment for obesity. None of the other variables was associated with SWS. If we added sleep efficiency or self-reported habitual sleep duration to the model, the associations remained the same. Results also remained virtually unchanged when patients who had a split-night PSG were excluded from the analyses. When we examined only those subjects with an AHI < 15 (n = 290), the association between SWS and African American race remained significant (beta = −0.87, P = 0.002). The significance of the association with overweight (beta = −0.63, P = 0.11) and obesity (beta = −0.55, P = 0.11) were attenuated in patients with an AHI < 15.

Table 3.   Multiple linear regression model predicting the natural log of SWS (min)

Variable	Regression coefficient	95% CI	P-value
Bold text indicates P < 0.05. OSA, obstructive sleep apnea; TRT, total recording time.
TRT (min)	0.003	−0.0002, 0.006	0.07
Age (years)	−0.04	−0.05, −0.03	< 0.001
Male gender	−0.69	−1.01, −0.37	< 0.001
African-American race	−0.48	−0.81, −0.14	0.006
Weight category
< 0.001	referent
< 0.001	−0.56	−1.11, −0.002	0.049
< 0.001	−0.73	−1.24, −0.23	0.005
Medicaid insurance	−0.30	−0.69, 0.09	0.13
Marital status
Married	referent
Unmarried	0.11	−0.22, 0.44	0.52
Unknown	0.17	−0.32, 0.67	0.49
Education
≤ High school	−0.13	−0.47, 0.21	0.45
≥ Some college	referent
Unknown	−0.39	−1.52, 0.74	0.50
OSA severity
None	referent
Mild	0.20	−0.37, 0.77	0.49
Moderate	0.11	−0.46, 0.68	0.71
Severe	−1.34	−1.92, −0.77	< 0.001
Self-reported antidepressant use	0.04	−0.34, 0.42	0.82
Self-reported hypertension	−0.40	−0.72, −0.07	0.017
Self-reported anxiety	0.07	−0.32, 0.45	0.74
Constant	2.85	1.34, 4.36	< 0.001

Discussion

Our results demonstrated that African Americans had significantly less SWS even after adjustment for socioeconomic indicators and co-morbidities including obesity and OSA. In the unadjusted comparison, the difference in SWS minutes between African Americans and non-African Americans was only 11; however, 11 min represented approximately half a standard deviation for the African Americans and also represented 60% of their mean. Nonetheless, whether 11 fewer minutes of SWS can impair health needs to be examined by future studies. The higher percentage of stage 2 suggests that this is the sleep stage replacing SWS. African Americans also reported obtaining less sleep habitually (6.8 versus 7.2 h); however, average recording time in the clinic was only 2–3 min longer than average self-reported sleep duration. We do not know how much actual sleep these patients obtained in the home, and some have found self-reported sleep to be approximately 1 h longer than actigraphically estimated sleep on average (Lauderdale et al., 2008). Thus, the sleep characteristics observed during the clinical PSG may be similar to what is experienced at home. Finally, although African Americans report sleeping less than non-African Americans, a SWS rebound is not likely to occur if provided a longer sleep opportunity, as several studies of partial sleep restriction observed no change in SWS (Akerstedt et al., 2009; Spiegel et al., 1999; Van Dongen et al., 2003). Furthermore, overweight and obesity were associated with less SWS, even after accounting for OSA. Conversely, severe OSA was associated with reduced SWS even after adjustment for obesity. We also observed reduced SWS in men and with increasing age, which has been well documented previously.

The racial association observed in our study is similar to what has been reported previously (Hall et al., 2009; Mezick et al., 2008; Profant et al., 2002; Redline et al., 2004; Stepnowsky et al., 2003; Tomfohr et al., 2010), despite the adjustment for several potential confounders, including obesity, OSA and socioeconomic indicators. These results suggest that the racial differences are not entirely explained by confounding due to OSA, obesity, demographics (age and gender) or socioeconomic status. Therefore, alternative explanations for this association need to be identified, particularly if the reduced SWS puts African Americans at greater risk of health conditions like insulin resistance and cardiovascular disease.

A novel finding in our study was that overweight and obesity was associated with reduced SWS independently of sleep-disordered breathing. When we restricted to those with an AHI < 15, the statistical significance of these associations was reduced, which could indicate that the association between weight status and SWS is attenuated in those without OSA or it could be due to reduced statistical power in the smaller sample size. Shorter average sleep duration has been associated with obesity (Cappuccio et al., 2008). Two large community-based studies in elderly men and middle-aged women reported an inverse relationship between SWS and BMI or measures of central adiposity (Rao et al., 2009; Theorell-Haglow et al., 2010). The underlying reason for less SWS among overweight and obese people, however, is not fully understood. Overweight and obesity per se may affect sleep homeostasis independently of OSA or, conversely, reduced SWS could increase the risk of weight gain (Beccuti and Pannain, 2011). Finally, we found that those with severe OSA had reduced SWS, even after adjusting for all the covariates. Interestingly, mild or moderate OSA was not associated with reduced SWS.

This study has some limitations that need to be acknowledged. First, this is a clinic-based sample and therefore these results may not generalize to the entire population. However, it is valuable for clinical practice to understand the factors associated with disease risk, and in our patient sample we found that African Americans, the overweight and obese, and those with severe OSA are at increased risk of less SWS, which may have important health consequences. Another limitation is that we did not have sufficient numbers of individuals in other racial/ethnic categories and could not examine whether other racial/ethnic groups, such as Hispanic/Latino or Asian groups, have less SWS compared with non-Hispanic Caucasians. A third limitation is that because this was a retrospective analysis, we were limited to the information that was collected in the clinic. Future studies should consider which additional factors may be important predictors of SWS and select appropriate methodologies for measuring them. For example, better tools to assess psychosocial measures, including depressive symptoms, or wrist actigraphy to assess habitual sleep may be included. Finally, another limitation is that the PSGs were only performed during one night, and the night-to-night variability in our patient population is therefore unknown. However, the Sleep Heart Health Study investigators demonstrated that SWS was not significantly different between home PSG and in-laboratory PSG, and in fact SWS had the least amount of variability compared with other sleep measures (Iber et al., 2004; Quan et al., 2002). The strengths of our study include a large percentage of women (57%) and a large percentage of African Americans (61%) in a large sample of 954 individuals.

In conclusion, SWS is linked to important physiological processes, including hormonal secretion and glucose metabolism, as well as reduced sympathetic nervous system activity. Reductions in SWS may be associated with increased risk of cardiometabolic disorders, such as diabetes and hypertension. Our analyses confirm previous findings that demonstrate lower amounts of SWS in African Americans, a group at increased risk of diabetes and hypertension compared with Caucasians. Additionally, we observed less SWS in those who are overweight and obese and those with severe OSA. Future research needs to explore the potential reasons for reduced SWS in these individuals, particularly if SWS partially mediates the racial disparity in diabetes risk.

Acknowledgement

This work was partly supported by a grant (1 P30 HL101859-01) from the National Heart, Lung and Blood Institute (NHLBI) of the National Institutes of Health (NIH).

Conflict of Interest

Authors have no conflict of interest to report related to the content of this manuscript.