Neighborhood stress and autonomic nervous system activity during sleep

Abstract

Study Objectives

Stressful neighborhood environments are known to adversely affect health and contribute to health disparities but underlying mechanisms are not well understood. Healthy sleep can provide a respite from sustained sympathetic nervous system (SNS) activity. Our objective was to evaluate relationships between neighborhood stress and nocturnal and daytime SNS and parasympathetic nervous system (PNS) activity.

Methods

Eighty-five urban-residing African Americans (56.5% female; mean age of 23.0) participated. Evaluation included surveys of neighborhood stress and sleep-related vigilance, and continuous electrocardiogram (ECG) and actigraphic recording in participants’ homes from which heart rate variability (HRV) analysis for low frequency/high frequency (LF/HF) ratio and normalized high frequency (nHF), as indicators of SNS and PNS activity, respectively, and total sleep time (TST), and wake after sleep onset were derived.

Results

All significant relationships with HRV measures were from the sleep period. Neighborhood disorder correlated negatively with nHF (r = −.24, p = .035). There were also significant correlations of HRV indices with sleep duration and sleep fears. Among females, LF/HF correlated with exposure to violence, r = .39, p = .008, and nHF with census tract rates for violent crime (r = −.35, p = .035). In a stepwise regression, TST accounted for the variance contributed by violent crime to nHF in the female participants.

Conclusions

Further investigation of relationships between neighborhood environments and SNS/PNS balance during sleep and their consequences, and strategies for mitigating such effects would have implications for health disparities.

Statement of Significance

Neighborhood environments have been shown to contribute substantially to variance in the health of populations though the mechanisms by which this occurs are not well understood. Optimal sleep and parasympathetic activity off-setting sympathetic nervous system activity induced by stress are important to maintaining health. We found associations between indicators of stressful neighborhood environments with increased sympathetic and decreased parasympathetic activity during sleep, and the reverse associations for sleep duration. Further research is indicated to establish nocturnal autonomic arousal as a link between threatening neighborhood environments and adverse health outcomes, and whether extending sleep, and/or reducing arousal prior to sleep onset can improve autonomic balance during sleep.

Introduction

There is strong evidence supporting the role of environmental factors in the health disparities that disproportionately burden African Americans [1]. Using data from the National Health Interview Survey collected between 1989 and 1994, Do and colleagues [2] found 15 to 75 per cent of the disparities between Black and White respondents to be accounted for by residential context. They cite factors contributing to the effects of place on health as including environmental exposures (e.g. toxins), built environment (e.g. availability of safe recreational facilities), and social conditions (e.g. exposure to neighborhood violence and drugs). In a study by Steptoe and Feldman [3], self-ratings of poor health and impairment were 2 to 3 times higher among those in the highest quartile for neighborhood problems compared with the lowest quartile. These findings were independent of individual risk factors and neighborhood socioeconomic status which led the authors to infer an important contribution of chronic stress to the effect of environment on health. Further elucidation of mechanisms by which threatening environments affect health is needed to ameliorate health disparities.

Excess activity of the sympathetic nervous system (SNS) has long been recognized to be a major effector of the relationship between stress and cardiovascular disease [4]. Increased SNS activity occurs with responses to threat and stress, and elevates blood pressure and heart rate. Sustained SNS activity strains the cardiovascular system by directly increasing peripheral resistance as well as stimulating immune and other humoral contributors to arterial stiffness [5], leading to hypertension and over time, adverse cardiovascular events. SNS activity is counterbalanced in the autonomic nervous system (ANS) by its other major branch, the parasympathetic nervous system (PNS). Sleep appears to provide an opportunity for PNS to offset SNS activity. Trinder et al. [6] demonstrated changes in ANS activity across the night of sleep in healthy participants using multiple measures that indicated shifts toward PNS dominance, particularly during nonrapid eye movement sleep. The direct effect of sleep on this shift is supported by demonstration that PNS activity increases with both nocturnal and daytime sleep [7]. Conditions where PNS dominance during sleep is compromised are associated with adverse cardiovascular consequences. For example, increased nocturnal SNS activity has been observed with the absence of nocturnal blood pressure reduction which is an established risk factor for hypertension and other cardiovascular events [8]. Apneic events trigger bursts of SNS activity during sleep, and sleep apnea is strongly associated with hypertension [9]. Blunted nocturnal reduction of low frequency heart rate variability (HRV), an indicator of SNS activity, was observed among participants with histories of myocardial infarction [10]. One might therefore infer that in the healthy situation, sleep provides an important respite from sustained SNS arousal.

There is evidence that the shift toward PNS dominance during sleep can also be compromised in insomnia [11] and with post-traumatic stress disorder (PTSD) [12], which have both been associated with cardiovascular risk [13, 14]. Short habitual sleep durations have also been linked to health risk [15, 16].The contribution of altered ANS activity to the effects of compromised sleep on health, however, remains largely unexplored.

Impact on sleep may be an important pathway by which stressful/threatening neighborhood characteristics influence health. A population study indicated that sleep durations of less than 6 hr were more common among blacks compared with whites and other racial/ethnic groups in the United States, and this association was partially explained by blacks more often living in high density urban environments [17]. Perception of threat could be an environmental factor that has important negative influence on sleep and ANS activity during sleep. Brosschot et al. [18] have argued that perseverative cognitions, such as worry and intrusive recollections, are more significant in sustaining ANS arousal than discreet stressful events. Presleep time in bed is a time of enhanced susceptibility to perseverative cognitions [19] and anticipation of a challenging morning task has been shown to increase ANS arousal during sleep [20]. In addition, sleep is a state that necessitates reduced vigilance which can enhance feelings of vulnerability if the external environment is perceived to be threatening.

In the present study, we derived HRV measures from ambulatory cardiac monitoring to evaluate relationships between neighborhood stress and ANS activity among young adult African American men and women living in urban environments. We hypothesized that measures of neighborhood stress would be positively associated with SNS and negatively associated with PNS indicators during the sleep period and these relationships would be influenced by nocturnal vigilance/sleep fears and sleep duration. We further explored whether relationships to nocturnal ANS indicators would be similar to those from daytime wake hours. In addition, in view of evidence that ANS activity and reactivity can differ between males and females [21], we also conducted subanalyses within men and women.

Methods

Participants

Participants were recruited between 2008 and 2012 from the Washington, DC metropolitan area primarily via posting flyers and participant referrals. Evaluations were conducted in the Clinical Research Unit (CRU) of the medical center for Howard University (HU), a Historically Black institution in urban DC. The protocol was approved by the HU Institutional Review Board and all participants received detailed description of the study and reviewed and signed the informed consent document. Initial advertisement included the University campus; however, in order to balance study recruitment, flyers were only posted in community settings that represented the study target demographics (e.g. near the University and Southeast D.C.) for the latter two-thirds of the recruitment period.

During initial phone and in-person screening, potential participants were ineligible if they reported height and weight with a calculated body mass index ≥ 40, excessive use of caffeine (>5 cups of coffee or its equivalent of other caffeinated drinks per day), heavy smoking (>20 cigarettes per day) and hazardous drinking (>14 drinks per week in men, >7 drinks per week in women), chronic medical conditions or severe psychiatric illnesses (i.e. schizophrenia, bipolar, chronic depression) that required consistent use of medications, and having regular night shift work or unusual habitual sleep or rise times (i.e. after 3 am, after 11 am, respectively).

Participants for the laboratory and monitoring phase were selected from those who filled out questionnaires based on availability and interest, and to balance the sample by gender, and to increase representation from the community rather than the college campus setting. To fulfill initial study aims, selection was also determined by the goal of recruiting similar sized groups of participants with PTSD, remitted PTSD, trauma exposure without PTSD, and a group without trauma exposure. Additional exclusion criteria for the final study group were a current psychiatric disorder other than PTSD, phobic disorders, or depression that was secondary in temporal onset and severity to PTSD, and current alcohol or drug abuse/dependence elicited through the structured clinical interview, positive urine toxicology for illicit drugs, and sleep apnea defined as an apnea/hypopnea index (AHI) of ≥5 on a screening sleep recording. In prior publications reporting sleep architecture [22] and nocturnal blood pressure dipping [23], a higher threshold of ≥10 was used. The lower AHI threshold of 5 was used for this analysis due to the effect of apnea on ANS activity [9]. The flow of study participants is illustrated in Figure 1.

Figure 1.

Flow chart of study population.

Self-report measures

These included a demographic questionnaire that obtained information on age, gender, race, highest educational level, and home address. Neighborhood stress was assessed by the City Stress Index (CSI) which is an 18-item self-report measure with items that query indicators of stressful urban environments such as abandoned houses, the sound of gunshots, and dilapidated buildings. It has a four-point response format ranging from never (0) to often or none to most (3). The CSI has demonstrated modest correlations with census indices of social disadvantage and indicators of emotional distress in urban adolescents [24]. The scale generates a composite score of neighborhood stress with subscales measuring Neighborhood Disorder and Exposure to Violence (toward family members or friends). The Fear of Sleep Inventory (FOSI) assesses vigilant behaviors and fear of loss of vigilance in relation to sleep, and dread of nightmares. It features a five-point scale format that is anchored by 0 “not at all” and 4 “nearly every night.” The FOSI has demonstrated good reliability and significant correlations with sleep quality, PTSD, and insomnia including in our urban young adult African-American population [25]. The Insomnia Severity Index (ISI) assesses the amount of difficulty falling asleep, staying asleep, and waking early, satisfaction with sleep, impact of insomnia on daily functioning, degree to which others notice, and the amount of distress caused by lack of sleep for a 2 week period using a five-point Likert scale rating. A cut-off score of 10 has been established as optimal for detecting clinically significant sleep difficulties in a community sample [26].

Census tract indicators

To compliment self-reported data on neighborhood environments, census tracts were identified using the US Census Bureau’s online geographic locator (www.factfinder.census.gov) from the home address filled out on the demographic form. For the participants residing in the District of Columbia, rates for violent crime were obtained from the website of Neighborhood Info DC—a project of The Urban Institute and Washington DC Local Initiatives Support Corporation (http://www.neighborhoodinfodc.org). Information for the Maryland (n = 31) and Virginia (n = 2) neighborhoods was obtained from the US Census website (http://factfinder2.census.gov).

Interviews

The Clinician Administered PTSD Scale for DSM-IV (CAPS) is a structured clinical interview designed to determine lifetime and current PTSD diagnostic status according to the Diagnostic and Statistical Manual of Mental Disorders 4th edition [27]. The CAPS also provides a continuous score of symptom severity based on their frequency and intensity. The Structured Clinical Interview for the DSM-IV (SCID) was used to determine lifetime and current mood disorders, psychotic disorders, other anxiety disorders, substance abuse/dependence, and eating disorders [28]. All interviews were conducted by trained staff (psychology graduate students and clinical psychology postdoctoral fellows) and were reviewed by the first author (T.A.M.) who is a board certified psychiatrist.

Monitoring procedures

Two overnight polysomnography recordings were obtained 1–2 weeks prior to the ambulatory cardiovascular monitoring. Other than apnea screening from the first night, polysomnography data are reported elsewhere [22]. Two 24 hr ambulatory cardiovascular recordings were conducted 1 week apart. Monitors included actigraphs (MicroMini-Motionlogger, AMI, Ardsley, NY) to measure sleep duration and continuity, and ambulatory blood pressure monitors that utilized an oscillometric method (Spacelabs 90207) and were programmed to inflate and record blood pressure every hour. Monitors were attached in the morning at the CRU during the morning hours. Participants left the CRU and assumed normal daytime activities and returned the following morning. A night of actigraphy recording alone was obtained after the first full ambulatory monitoring so that estimates of possible sleep disruption by blood pressure cuff inflation could be obtained. Participants recorded their time of going to bed and turning the lights out and waking up on a short sleep diary form.

Continuous monitoring of the electrocardiogram (ECG) was obtained using the Nasiff Cardio System (Brewerton, NY; www.Nasiff.com). After skin preparation, five electrodes were placed over the sternum and lower ribs to enable three-channel recording to a pocket-size recorder. Participants were instructed to avoid getting the recording device and electrodes wet and contact with potential interference, and pushing or pulling on electrodes. After setup, participants left the CRU and returned 24 hr later to terminate the recording. Signal quality was checked before participants’ departures. The device was set to digitize recordings at a sampling frequency of 256 Hz. After returning and detaching equipment, data stored in a flash card were downloaded to CardioCard software (Nasiff) for Windows on a desktop or laptop computer. To allow for acclimation to the study equipment, recordings from the second full day of ambulatory monitoring (1 week after the first) were analyzed for HRV. The entire recording was reviewed to confirm or correct automatic marking of R-waves. Five minute epochs with artifact that obscured successive R-waves were excluded from analyses. Records with ≥30 per cent of epochs that were unusable were excluded from the analysis. The CardioCard software computed the power spectral density for low frequency (LF) (0.04–0.15 Hz) and high frequency (HF) (0.15–0.40 Hz) for each nonoverlapping 5 min epoch using fast Fourier transformation.

Analyses

Variables selected for analyses included age, BMI and gender, the two subscale scores from the CSI, the total score for the FOSI, annual rates for violent crime for participants’ home census tract, actigraphy-measured total sleep time (TST) and wake after sleep onset (WASO), and LF/HF and normalized high frequency (nHF) during wake and during the sleep period denoted by time from lights out to the final awakening, i.e. time in bed, on the sleep diary. Relationships of the HRV measures to gender were explored via t-tests and to age, gender, and the concurrent neighborhood measures, sleep measures, and sleep fears with Pearson correlations. Correlations were also calculated within gender in exploratory analyses. Stepwise regression models were calculated in order to evaluate the independent contributions of the indicators of neighborhood stress, nocturnal vigilance, or actigraphy-derived sleep measures that demonstrated significant bivariate correlations, with the HRV signals.

Results

This study featured a subset of the 136 participants who completed the home monitoring with selection and recruitment as described in Methods and in a previous publication [23]. The more stringent criterion for exclusion for sleep apnea (AHI ≥ 5) eliminated 32 of the previously included participants. In addition, 2 recordings were not utilized due to minimal sleep occurring (<3 hr), 2 due to frequent ectopic heart beats, and 15 due to loss of leads, and/or artifact affecting greater than 30 per cent of either wake or sleep epochs (Figure 1).

All participants identified as black or African American. The analyzed sample of 85 was 56.5 per cent female, mean age was 23.0 (SD = 4.76; range 18–35). Seventy (82.4%) of these 85 participants reported exposure to a trauma meeting DSM-IV PTSD criterion A, 12 (14.1%) met current criteria for PTSD, and an additional 27 (31.8%) met criteria during their lifetime and had recovered. None of the participants had current depression. Thirty-four (40%) of the sample met or exceeded the ISI cutoff of 10 for detecting insomnia. Thirty-seven (43.5%) of the sample were healthy weight, 36 (42.4%) were overweight, and 11 (12.9%) were obese. Four (4.7%) endorsed smoking tobacco (less than 20 cigarettes per day). Thirty-three (38.8%) participants lived on the Howard University campus and the remaining 52 (61.2%) resided in the community.

There were no significant or trend-level associations between the four HRV-derived measures and age or BMI. Differences by gender were found for LF/HF during wake (males—1.027 ± .0167, females—1.035 ± .1529; t (83) = −.206; p = .04) and LF/HF during the sleep period (males—1.028 ± .021; females—1.040 ± .019; t = (83) 2.87; p = .005). None of the four HRV measures differed by whether the participant had current or lifetime or never had PTSD. On the nights, HRV measures were obtained: mean TST was 339.2 ± 115.7 min and mean WASO was 67.4 ± 68.0 min. These measures were not significantly different from, and were similar to, those obtained with actigraphy absent blood pressure monitoring during the preceding week (TST = 354.2 ± 100.3; WASO = 77.2 ± 78.4).

As indicated in Table 1, LF/HF and nHF during the sleep period correlated with TST (r = −.25, r = .23), nHF with neighborhood disorder (r = −.24), and LF/HF with FOSI (r = .25) at p < .05. While correlated with FOSI (r = .47, p < .001) and TST (r = −.22, p < .05), correlations of ISI with the nocturnal HRV measures were weak and nonsignificant (p’s > .65). Among the women in the sample (n = 48), LF/HF and nHF from the sleep period correlated with TST (r = −.39, r = .36) and nHF during sleep with census track rates for violent crime (r = −.35) (p’s < .05) and the Exposure to Violence subscale of the CSI (p < .01). There were no significant correlations within the male subgroup nor during wake periods for the entire sample. (All p values were >0.1.)

Table 1.

Relationships between heart rate variability and measures for neighborhood stress, sleep, and sleep fears

N = 85	Neighborhood disorder†	Exposure to violence†	Rate of violent crime	TST‡	WASO§	FOSI||
LF/HF¶ in bed	.15	.13	−.04	−.25*	.04	.23*
nHF# in bed	−.24*	−.07	−.15	.23*	.04	−.12
Females only (n = 48)
LF/HF in bed	.28	.39**	.22	−.39*	.22	.14
nHF in bed	−.28	−.23	−.35*	.36*	−.16	−.13
Males only (n = 37)
LF/HF in bed	.03	−.22	−.19	−.07	−.29	.18
nHF in bed	−.05	.21	.03	−.01	.29	−.02

*p ≤ .05; **p < .01.

^†From the City Stress Index.

^‡Total sleep time.

^§Wake after sleep onset.

^||Fear of Sleep Index.

^¶Low frequency/high frequency.

^#Normalized high frequency.

Since none of the sleep period HRV signals were significantly correlated with age or BMI, no other covariates were included in the regression analyses which are presented in Table 2. In summary, for the whole group, neighborhood disorder, but not TST was significant in predicting nHF (p < .01), and TST but not fear of sleep, was significant in predicting LF/HF (p < .05). Within the female subsample, Exposure to Violence, but not TST was significant in predicting LF/HF (p < .02); however, TST, but not violent crime rate, was significant in predicting nHF (p < .01). (All of the dependent measures were of HRV signals during the sleep period.)

Table 2.

Stepwise regressions

Whole sample
DV—nHF
Model—R2 = .094
IVs—Neighborhood disorder*—β = −.31, p = .009 Total sleep time—β = .18, p = .13
DV—LF/HF
Model—R2 = .065
IVs: Total sleep time*—β = −.25, p = .03 Fear of sleep—β = .15, p = .19
Females only
DV—LF/HF
Model—R2 = .15
IVs—Exposure to violence*—β = .38, p = .015
Total sleep time—β = −.28, p = .08
DV—nHF
Model—R2 = .15
IVs—Total sleep time*—β = .51, p = .003 Violent crime rate—β = −.12, p = .48

*Retained IV, probability of F ≤ .05.

Discussion

This study provides unique information from 24 hr assessments of autonomic activity that emphasizes signals from sleep periods and their relationships to measures of neighborhood stress, sleep, and nocturnal vigilance. The relationships that had statistical significance of p ≤ .05 were all with HRV indices from the sleep period. Potential relationships with HRV during wake could have been obscured by the effects of varying activity levels on ANS activity compared with the relatively quiescent sleep period. Our hypotheses that LF/HF, the study index of SNS activity, during the sleep period, would be positively related, and that nHF, the study index of PNS activity, during the sleep period, would be negatively related to measures of neighborhood stress have preliminary support from several of the study findings. The findings further suggest that nocturnal ANS balance may be more sensitive to reactions to environmental factors in women, particularly exposure to violence to others. While correlated with LF/HF, we did not find evidence that sleep fears/nocturnal vigilance contributed to relationships between neighborhood stress measures and ANS activity during the sleep period. The FOSI queries tendencies over a 1 month period. It is possible that a more momentary assessment of nocturnal vigilance would show more direct influence on relationships between environmental concerns and nocturnal ANS activity. Nocturnal ANS measures were also not related to self-reported insomnia. We also found evidence that the duration of sleep (TST) was negatively related to LF/HF and positively related to nHF which is consistent with prior studies relating sleep to PNS dominance [8, 9]. In a regression, TST accounted for the relationship between census tract–determined violent crime rates on nHF during the sleep period in women. These findings collectively raise questions regarding the contribution of autonomic dysregulation on the adverse health effects of habitual short sleep duration [17, 18].

Due in part to concern regarding the potential sleep disruptive effects of concurrent, intermittent inflation of a blood pressure cuff, relationships between HRV indices and actigraphy-measured WASO were examined and were found to be small and nonsignificant. In addition, WASO and TST from a night that featured BP and electrocardiogram (ECG) measurement were very similar to measures from the night where monitoring only included wrist actigraphy.

While suggesting a potentially important pathway from environmental stress to adverse health consequences, there are limitations to the present study. First are concerns regarding interpretations of HRV signals including that LF/HF and nHF are mathematically inter-related and uncertainty regarding LF/HF uniquely reflecting SNS activity [29]. There is extensive validation, however, of spectrally derived HRV measures as indices of autonomic balance, imperfections notwithstanding [30], and they have also provided an essential tool for the limited number of investigations of autonomic activity during sleep. To the best of our knowledge, our utilization of HRV measures from continuous ECG recording over 24 hr with delineation of sleep and wake is unique. Editing 24 hr of ECG recordings is quite time consuming and we are therefore relegated to the limitation of using only one day and night of data. The nonrandom selection of participants for monitoring introduced selection bias that could have confounded the examined relationships. We previously reported a difference in nocturnal ANS balance between a subset of the current study participants who met criteria for PTSD and a subgroup who had been exposed to “high impact” trauma but never developed significant PTSD symptoms [12]. However, ANS indices were not associated with a PTSD diagnosis in this study population, and the strongest finding of the present study, the relationship between violence exposure and LF/HF in women, was not attenuated by adding PTSD severity to a regression model. The previous study analyses further revealed a very strong correlation between sleep duration and the nocturnal HRV indices in the resilient group, suggesting that the high level of resilience in the comparison group contributed to the differences in HRV between the PTSD and resilient groups in that study. Conclusions are also limited but the modest strength of the study findings and given the number of exploratory correlations we cannot exclude chance findings. The modest strength of study correlations could relate both to imprecision of the measures and multiple influences on autonomic activity.

Limitations notwithstanding, the cross-sectional relationships found in this study are generally consistent and suggest that reactions to environmental stress can have adverse impact on autonomic arousal during the sleep period and that longer sleep durations could sometimes mitigate these effects. As noted in the Introduction, conditions where there is evidence for the rise in PNS activity in sleep being compromised [13, 14] are known to be associated with cardiovascular and other health morbidity [15, 16]. Additional research on the influence of sex and gender on these relationships, relationships between overnight autonomic balance and cardiovascular risk, and whether reducing presleep cognitive and emotional arousal and/or extending sleep are effective in influencing nocturnal autonomic activity could contribute to understanding how to diminish the negative impact of stressful neighborhood environments on health.

Funding

This research was supported by National Heart, Lung, and Blood Institute grant R01HL087995 to Thomas A. Mellman and National Center for Advancing Translational Sciences grant UL1RR031975 for the Georgetown-Howard Universities Center for Clinical and Translational Science.

Notes

Conflict of interest statement. T.A.M. has received consulting fees from Tonix Pharmaceuticals and speaking honoraria and grant support from Merck.