Core body temperature changes before sleep are associated with nocturnal heart rate variability

Abstract

Core body temperature (CBT) reductions occur before and during the sleep period, with the extent of presleep reductions corresponding to sleep onset and quality. Presleep reductions in CBT coincide with increased cardiac parasympathetic activity measured via heart rate variability (HRV), and while this appears to persist into the sleep period, individual differences in presleep CBT decline and nocturnal HRV remain unexplored. The purpose of the current study was to assess the relationship between individual differences in presleep CBT reductions and nocturnal heart rate (HR) and HRV in a population of 15 objectively poor sleeping adults [10 males, 5 females; age, 33 ± 4 yr; body mass index (BMI) 27 ± 1 kg/m²] with the hypothesis that blunted CBT rate of decline would be associated with elevated HR and reduced nocturnal HRV. Following an adaptation night, all participants underwent an overnight, in-laboratory sleep study with simultaneous recording of polysomnographic sleep including electrocardiography (ECG) and CBT recording. Correlations between CBT rate of change before sleep and nocturnal HRV were assessed. Blunted rate of CBT decline was significantly associated with increased heart rate (HR) in stage 2 (N2; R = 0.754, P = 0.001), stage 3 (N3; R = 0.748, P = 0.001), and rapid-eye movement (REM; R = 0.735, P = 0.002). Similarly, blunted rate of CBT decline before sleep was associated with reduced HRV across sleep stages. These findings indicate a relationship between individual differences in presleep thermoregulatory processes and nocturnal cardiac autonomic function in poor sleeping adults.

NEW & NOTEWORTHY Core body temperature (CBT) reductions before sleep onset coincide with increases in heart rate variability (HRV) that persist throughout the sleep period. However, the relationship between individual differences in the efficiency of presleep core temperature regulation and nocturnal heart rate variability remains equivocal. The present study reports an association between the magnitude of presleep core body temperature changes and nocturnal parasympathetic activity, highlighting overlap between thermoregulatory processes before sleep and nocturnal cardiac autonomic function.

INTRODUCTION

Thermoregulation is heavily intertwined with sleep processes (1). Core body temperature (CBT) follows a circadian rhythm, with sleep onset typically occurring during the downslope of CBT (2–4), and awakening following a few hours after CBT reaches its nadir and begins to rise. CBT begins to decline before sleep onset, and the rate of this decline is predictive of sleep initiation (3, 5). This presleep reduction of CBT is a function of both heat dissipation and heat production, with increased heat dissipation due to skin vasodilation primarily driving the CBT decline (6) and further predicting sleep timing (2, 7). Specifically, around the sleep period, changes in distal skin temperature are opposite to that of CBT, representing increased blood flow to distal extremities and skin, allowing for heat transfer from internal organs to the distal regions where heat is lost to the environment. Experimentally induced increases of skin temperature or heat dissipation result in more pronounced nocturnal reductions in CBT, reduced sleep onset latency, and increased sleep depth in healthy younger and older adults, as well as individuals with insomnia (8–12). Conversely, disruptions in sleep thermoregulatory processes may precede and further exacerbate chronic sleep disturbance (13).

Diurnal fluctuations in CBT are observed in tandem with changes in cardiac activity. Specifically, as CBT declines, parasympathetic activity at the level of the heart increases. Cardiac autonomic activity is influenced by both sleep onset, sleep architecture (14, 15), and the circadian system (16–18). There is likely a bidirectional relationship between cardiac autonomic activity and sleep parameters (19). Proper cycling toward nocturnal parasympathetic predominance is relevant to health outcomes, as elevated nocturnal heart rate (HR) and blunted parasympathetic activity during sleep are associated with an elevated risk of mortality and cardiovascular complications (20–23).

HR decline before sleep (24) occurs concomitantly with the steepest CBT reduction (3, 5). These cardiac changes before sleep onset are likely mediated through a shift toward parasympathetic predominance occurring in tandem with reductions in CBT (25). However, evidence indicates that once asleep, heart rate variability (HRV) within a given sleep stage does not change as a function of time asleep (14), suggesting that presleep processes establish a shift toward parasympathetic dominance that persists into the sleep period. Disruption to these presleep processes may have lasting effects upon nocturnal cardiovascular control (26–28).

The simultaneous changes in CBT, HR (24), and parasympathetic activity (25) before sleep onset suggests an association between the efficiency of presleep thermoregulation and nocturnal cardiac autonomic function. Manipulations that facilitate a greater reduction in CBT, such as high-heat capacity mattresses, reduce nocturnal HR in older men (12) and women (8), in whom age-associated impairments in thermoregulation are observed (29). However, the relationship between presleep thermoregulation and nocturnal cardiac autonomic control in poor sleepers has not been adequately investigated. Studies to date have largely been restricted to young healthy populations (25, 30), whereas recent evidence from our laboratory has highlighted nocturnal cardiac dysfunction in habitually poor sleeping adults (31). It is well documented that chronic poor sleep is associated with a higher risk of cardiovascular disease (32–34). Therefore, investigation into the relationship between presleep thermoregulation and nocturnal cardiac function in poor sleepers is warranted.

The purpose of the current study was to assess the relationship between the rate of presleep CBT reductions, used as a metric of presleep thermoregulatory efficiency, and nocturnal cardiac activity in a population of subjectively and objectively poor sleepers. We hypothesized that a blunted rate of CBT decline before sleep would be associated with elevated nocturnal HR. Furthermore, we hypothesized that blunted CBT dipping would be associated with a reduced nocturnal HRV.

METHODS

Participants

Subject characteristics.

Participants were recruited from Michigan Technological University via community flyers, social media, and word-of-mouth. All participants were between the ages of 18 and 60 yr and had a body mass index (BMI) less than 35 kg/m². All participants were nonsmokers and nondiabetic and did not have any diagnosed cardiovascular, autonomic, or sleep disorders. Participants were not shift workers. Participants were not allowed to take any psychoactive, hypnotic, analgesic, or autonomic nervous system medications within the month preceding the study date. Eligible female participants were required to have regular menstrual cycles or were required to be 5 yr postmenopausal and not on hormone replacement therapy.

Subjective and objective poor sleep assessment.

After completion of a preliminary phone screen and eligibility determination, each participant was provided with an informed overview of the study, followed by voluntary written informed consent from each participant. Participants were required to be both subjective and objective poor sleepers. To assess subjective sleep quality, participants were asked to complete the Pittsburgh Sleep Quality Index (PSQI) and were required to have a score greater than 5, which distinguishes poor from good sleep (35). Participants were additionally asked to take the insomnia severity index (ISI) and Epworth sleepiness scale (ESS) to assess insomnia symptoms and daytime sleepiness levels. Participants were then sent home with an at-home sleep apnea kit (ApneaLink; ResMed, San Diego, CA) and actigraphy wristwatch (Actiwatch Spectrum Pro, Philips Respironics, Bend, OR). Briefly, the ApneaLink is an at-home system that uses nasal cannula and pulse oximetry to estimate the apnea-hypopnea index (AHI) to screen for obstructive sleep apnea (OSA). Any participant with an apnea-hypopnea index (AHI) > 15 episodes/h was disqualified from further participation. The wrist actigraphy watch was worn for 7–14 days. To qualify as objectively poor habitual sleepers, participants were required to have a sleep onset latency (SOL) > 20 min and/or a sleep efficiency (SE) < 85%. Participants then underwent a familiarization night within the laboratory and were excluded if overnight polysomnography (PSG) revealed any other sleep disorders (i.e., OSA, restless leg syndrome, etc.).

A total of 62 individuals were screened for participation. However, after excluding individuals based on exclusionary criteria or insufficient data collection, 15 participants (10 males, 5 females; age, 33 ± 4; BMI 27 ± 1) were included in the final analysis. Michigan Technological University Institutional Review Board approved all testing procedures, and each participant provided written informed consent before testing.

Experimental Design

Following detailed consent and at-home apnea-link and actigraphy testing, participants underwent a familiarization PSG night in the sleep laboratory that was maintained between 20°C and 21°C (68–70°F) during all sleep sessions. During the familiarization visit, all participants were acclimated to the laboratory environment, and any other sleep disorders were screened. Following the familiarization night, participants once again arrived at the sleep laboratory for their official testing session on a separate occasion. Participants were asked to arrive at the sleep laboratory at ∼5:00 PM, and were asked to refrain from exercise and caffeine for 12 h, and alcohol for 24 h. In addition, participants were required to fast for at least 3 h before arrival. Upon arrival, participants were provided with a standard dinner following the recommendation of a registered dietitian based on caloric need and a fixed ratio of proteins, carbohydrates, and fats. Next, each participant was asked to ingest a temperature monitoring pill (CoreTemp, HQ Inc., Palmetto, FL), which recorded a temperature measurement every 10 s as it passed through the gastrointestinal tract. Participants went about typical evening routines until 8:30 PM when overnight PSG was set up by trained sleep laboratory technicians and students following American Academy of Sleep Medicine (AASM) guidelines. Participants continued their nightly routine and were asked to put away all electronics 1 h before lights out. All participants brought sleep clothing of their choosing. Participants entered the bed 15 min before lights out and laid supine during which time adequacy of PSG monitoring was determined. Lights out occurred between 10:00 and 11:00 PM based on the preference of each participant. Participants were provided with a total of 8-h sleep opportunity during which simultaneous PSG and HR were monitored. Upon awakening, participants were allowed to leave the laboratory.

Measurements

Polysomnography.

Overnight PSG (Natus Medical; Middleton, WI) was used to quantify both rapid-eye movement (REM) and non-REM (stage I, II, and III) sleep. Sleep electroencephalography (EEG) was conducted based on guidelines from the AASM, including electrode placements with pairs of central, frontal, and occipital leads. Reference leads were placed contralaterally on the opposite mastoid processes of the head. Electrooculography (EOG) monitored eye movement with two electrodes placed near the eyes, whereas electromyography (EMG) recorded muscle activity with three electrodes placed on the chin. Respiratory effort was monitored using thoracic and abdominal piezoelectric effort belts. A thermistor and a nasal cannula were used to monitor respiratory flow. Pulse oximetry measured blood oxygen saturation and detected any potential desaturations indicating an apneic event. Overnight PSG recordings, subsequent sleep staging, apneic events, limb movements, and arousals were reviewed and scored first by a graduate student or registered polysomnography sleep technician and were confirmed by a board-certified sleep physician.

Nocturnal heart rate.

Continuous HR was recorded with two-lead electrocardiogram placed bilaterally inferior of the clavicle. HR was recorded at a frequency of 250 Hz throughout the entirety of the 8-h allotted sleep opportunity.

Core body temperature.

Continual monitoring of CBT was recorded using a temperature sensing, ingestible telemetry pill (CorTemp, HQ Inc., Palmetto, FL). Each participant ingested the pill with dinner and was then equipped with a wireless recorder that was worn around the waist, allowing the research team to capture measures of CBT throughout the study period. All participants wore the recorder on their waist throughout the entirety of the sleep period.

Data Analysis

EEG spectral analysis.

EEG spectral analysis was performed in MATLAB (MathWorks, Natick, MA) using the EEGLAB toolbox and graphic interface (36). The data were preprocessed using a band-pass filter (0.3–35 Hz), referenced to the linked-mastoids, and sampled at 250 Hz. Artifact-free, 30-s epochs of stage 2 (N2), stage 3 (N3), and REM sleep were isolated throughout the night. These sections were then used to calculate the absolute power density in delta (Δ; 0.5–4 Hz), alpha (α; 8–12 Hz), and beta (β; 16–32 Hz) frequencies, across C3 and C4 electrodes, using a fast Fourier transform (FFT). All analyses presented were performed using the average absolute power across the central (C3 and C4) leads. However, in three participants, a lead was lost, in which case the lead that remained was used for analysis, rather than the average of two leads.

Heart rate variability.

HRV measures were assessed throughout the sleep period using previously defined parameters (14) that have since been used in subsequent studies (27, 37, 38). This methodology was used as longer analysis time periods (5–10 min) were rare within the current sample population of objectively poor sleepers due to sleep disruption, and a minimum of 2-min of ECG recording are sufficient to assess frequency domain measure of HRV (39). Briefly, 2-min periods of stable sleep that were not preceded or followed by sleep disruptions (i.e., arousal, leg movements, apnea, etc.) were isolated and used in subsequent analysis. Frequency and time domain measures of HRV were acquired from overnight ECG recordings and served as a measure of nocturnal cardiac autonomic, specifically parasympathetic, function. The ECG recording from the 8-h sleep opportunity was imported into analysis software (Lab Chart 8; ADInstruments, Sydney, Australia) that automatically marked each R-wave. The research team manually confirmed each R-wave. Spectral analysis was performed on all 2-min periods of nocturnal ECG of stable PSG sleep to determine frequency domain HRV. Data were passed through a 60-Hz Notch filter before analysis. The integrated area within the high-frequency (HF; 0.15–0.4 Hz) and low-frequency (LF; 0.04–0.15 Hz) ranges are reported. Time domain HRV was expressed using the percentage of R-R intervals that varied by 50 ms or more (pNN50) and the root mean squared of successive difference of R-R intervals (RMSSD). Higher HF-HRV, pNN50, and RMSSD are indicative of higher HRV and greater parasympathetic activation.

Following analysis of HRV during each 2-min segment of sleep, these 2-min segments were grouped by sleep stage (N2, N3, and REM) within each participant. N1 sleep was not used as it is a transitionary stage, and the participants did not have sufficient quantity of N1 sleep to meaningfully assess. Within each participant, average HRV measures during each sleep stage were used in subsequent analysis.

Core temperature rate of change.

From the overnight CBT tracing, one measurement was taken every 5 min beginning 1 h before lights out until final awakening. Data points were removed if they were more than 2 standard deviations away from the average of the preceding three data points. In these cases, CBT was estimated by determining the slope of the closest accurate data points before and after the inaccurate point and placing the estimated point on that slope line. The rate of change in CBT from 60 min before lights out to sleep onset was utilized as a metric of presleep thermoregulation and was used in the remainder of analyses. A representative image of the decline in CBT surrounding sleep onset from a participant is presented in Fig. 1.

Figure 1.

A representative core body temperature (CBT) tracing from one participant beginning 60 min before sleep onset until 120 min after sleep onset. CBT begins to decline before sleep onset and continues to decline until it reaches its nadir during the night. The dotted line denotes lights out, as does the adjacent gray shaded region.

Statistical analysis.

All data were analyzed using SPSS statistical software (SPSS 25.0; SPSS, Chicago, IL). Descriptive statistics of participant anthropometrics, subjective, and objective sleep parameters are reported. All variables were assessed for normality using the Shapiro–Wilk assessment. The only measures that did not meet normality were the frequency-domain measures of HRV (i.e., HF, LF, and LF/HF) and the EEG spectral analyses. These variables were transformed using a log₁₀ transformation and used in subsequent analysis. To assess the effects of sleep stage on HRV measures, a repeated-measures analysis of variance (ANOVA) was conducted on the average HRV across individuals in each sleep stage. In cases of sphericity violations, Greenhouse–Geisser corrections were used. If the effect of sleep stage was significant, pairwise comparisons using a Bonferroni adjustment were used to assess post hoc differences between sleep stages. Finally, a Pearson correlation was used to assess the relationship between presleep rate of CBT change and nocturnal HRV measures during each sleep stage. An a priori power analysis was not conducted. To determine the minimum effect size that could be reliably detected (40, 41), a post hoc sensitivity analysis was conducted to assess the sensitivity of our analysis in detecting associations between presleep CBT rate of change and the primary outcome variable, nocturnal HR. Sensitivity analysis indicated that a Pearson’s correlation coefficient with 15 participants would be sensitive to effects of R = 0.65 with at least 80% power (α = 0.05, two-tailed). All data are presented as means ± SD. A significance level of α = 0.05 was used for all statistical tests.

RESULTS

Demographics and Sleep Parameters

Table 1 highlights participant characteristics and questionnaire responses. All participants reported a PSQI greater than 5 (8 ± 3 a.u.). The subject population reported an insomnia severity index that meets the cutoff for subthreshold insomnia symptoms (ISI: 10 ± 2 a.u.). Table 2 depicts actigraphy and PSG sleep parameters. Participants received less than the recommended 7–9 h of total sleep time (TST) when assessed habitually using actigraphy wristwatch monitoring (411 ± 41 min) and one night of PSG monitoring (416 ± 33 min). Furthermore, actigraphy-defined habitual SE was poor across the participants (81 ± 5%), and SOL was also prolonged (28 ± 33 min).

Table 1.

Participant characteristics

Sex (M/F)	10/5
Age, yr	33 ± 15 [20–59]
BMI, kg/m2	27 ± 5 [19–33]
PSQI, a.u.	8 ± 3 [6–13]
ISI, a.u.	10 ± 2 [4–21]
ESS, a.u.	7 ± 5 [2–12]

Values are means ± SD [ranges]. BMI, body mass index; ESS, Epworth Sleepiness Scale; ISI, insomnia severity index; PSQI, Pittsburgh Sleep Quality Index.

Table 2.

Sleep parameters

Actigraphy
Bedtime, hh:mm:ss	23:33:55 ± 01:36:58
TST, min	411 ± 41
SE, %	81 ± 5
SOL, min	28 ± 13
Awakenings, No.	37 ± 7
WASO, min	42 ± 13
Polysomnography
AHI, episodes/h	0.8 ± 1
Bedtime, hh:mm:ss	22:37:54 ± 00:27:42
TST, min	416 ± 33
SE, %	87 ± 7
SOL, min	9 ± 11
Arousals, No.	59 ± 33
Awakenings, No.	29 ± 8
WASO, min	50 ± 26

Values are means ± SD. AHI, apnea-hypopnea index; SE, sleep efficiency; SOL, sleep onset latency; TST, total sleep time; WASO, wake after sleep onset.

Impact of Sleep Stage on Nocturnal HRV

Table 3 depicts changes in HRV across the night by sleep stage. A significant increase in HR was observed in REM sleep when compared with stages N2 and N3. Likewise, REM sleep was associated with an increased LF/HF_log10 when compared with NREM stages. LF_log10 was lowest in N3 when compared with both N2 and REM.

Table 3.

Effect of sleep stage on nocturnal HRV

Measure	N2	N3	REM	P Value	Direction
HR, beats/min	56 ± 8	57 ± 9	59 ± 9	0.002	REM > N2, N3*
RMSSD, ms	66 ± 35	58 ± 25	65 ± 43	0.217
pNN50, %	37 ± 23	34 ± 24	31 ± 23	0.079
LFlog10, a.u.	3.1 ± 0.3	2.9 ± 0.4	3.2 ± 0.4	<0.001	N3 < N2, REM
HFlog10, a.u.	3.2 ± 0.5	3.1 ± 0.5	3.1 ± 0.7	0.484
LF/HFlog10, a.u.	−0.1 ± 0.3	0.2 ± 0.3	0.3 ± 0.4	<0.001	N2, N3 < REM

Values are means ± SD. HF_log10, log transformed high frequency-HRV; HR, heart rate; HRV, heart rate variability; LF_log10, log transformed low frequency-HRV; LF/HF_log10, log transformed low frequency/high frequency HRV ratio; N2, stage 2 sleep; N3, stage 3 sleep; pNN50, percentage of R-R intervals that varied by 50 ms or more; REM, rapid eye movement sleep; RMSSD, root mean squared of successive differences of R-R intervals. *P = 0.059.

Relationship between Presleep Core Temperature and Sleep Architecture

A significant inverse correlation was observed between the rate of presleep CBT and total time spent in N3 sleep (R = −0.602, P = 0.018). However, presleep CBT was not associated with time spent in N1 (R = 0.462, P = 0.083), N2 (R = 0.370, P = 0.175), or REM sleep (R = −0.176, P = 0.531). Similarly, CBT dipping was not significantly associated with TST (R = −0.421, P = 0.118) or SE (R = −0.450, P = 0.093).

There was a significant positive association between CBT rate of change and nocturnal α_log10 power (R = 0.576, P = 0.025), although there was no significant association with β_log10 (R = 0.422, P = 0.117) or Δ_log10 (R = −0.125, P = 0.656) power.

Relationship between Presleep Core Temperature and Nocturnal HRV

Figure 2 displays the correlation between presleep CBT rate of change and HR, RMSSD, and pNN50 for sleep stages N2, N3, and REM. During stage N2, the rate of presleep CBT change was correlated with HR (R = 0.754, P = 0.001, Fig. 2A), RMSSD (R = −0.557, P = 0.031, Fig. 2D), and pNN50 (R = −0.598, P = 0.019, Fig. 2G). During N3, the CBT rate of change was associated with HR (R = 0.748, P = 0.001, Fig. 2B), RMSSD (R = −0.591, P = 0.020, Fig. 2E), and pNN50 (R = −0.542, P = 0.037, Fig. 2H). Although HR was related to CBT rate of change during REM sleep (R = 0.735, P = 0.002, Fig. 2C), the associations with RMSSD (R = −0.479, P = 0.071, Fig. 2F) and pNN50 (R = −0.502, P = 0.056, Fig. 2I) did not reach significance.

Figure 2.

Relationship between presleep core body temperature (CBT) changes and nocturnal time-domain heart rate variability (HRV) measures. Bivariate correlations were used to assess the relationship between CBT decline before sleep and heart rate (HR) during stage 2 (N2; A), stage 3 (N3; B), and rapid-eye movement (REM; C) sleep. Similarly, bivariate correlations were used to assess the relationship between CBT decline and the root mean squared of successive differences of R-R intervals (RMSSD) during N2 (D), N3 (E), and REM (F) sleep. Finally, correlations between CBT changes and percentage of R-R intervals that varied by 50 ms or more (pNN50) during N2 (G), N3 (H), and REM (I) sleep were assessed. n = 15.

Figure 3 depicts log₁₀ transformed HF HRV, a marker of parasympathetic activity at the level of the heart, during N2, N3, and REM sleep. HF HRV_log10 was inversely correlated to the rate of presleep CBT change during stages N2 (R = −0.592, P = 0.020, Fig. 3A) and REM sleep (R = −0.602, P = 0.018, Fig. 3C), but this relationship did not reach significance during N3 (R = −0.496, P = 0.060, Fig. 3B). LF HRV_log10 was significantly negatively associated with CBT rate of change during N3 (R = −0.577, P = 0.024) and REM (R = −0.566, P = 0.028), but not during N2 (R = −0.468, P = 0.078). LF/HF_log10 was positively associated with the rate of CBT change during REM sleep (R = 0.540, P = 0.038), but not during either N2 (R = 0.442, P = 0.099) or N3 (R = 0.093, P = 0.743).

Figure 3.

Relationship between presleep core body temperature (CBT) changes and nocturnal high-frequency (HF) heart rate variability (HRV). Bivariate correlations were used to assess the relationship between presleep CBT changes and HF_log10 HRV during stage 2 (N2; A), stage 3 (N3; B), and rapid eye movement (REM; C) sleep. n = 15.

DISCUSSION

The present study assessed the relationship between the rate of CBT change before sleep onset and nocturnal vagal tone (estimated via HRV) in a sample of objectively and subjectively poor sleepers. Three primary findings were observed. First, a blunted CBT decline before sleep was associated with increased HR throughout the sleep period. Second, a blunted rate of CBT decline before sleep onset was associated with decreased nocturnal HRV. A decrease in HR and all time domain measures (RMSSD and pNN50) was observed in both N2 and N3, whereas only increased HR was observed in REM sleep. In addition, the blunted rate of CBT decline was negatively associated with the HF HRV in both N2 and REM sleep. Third, a blunted rate of CBT decline before sleep onset was associated with decreased total N3 time, and increased α-EEG frequency, an indicator of poor sleep and/or nocturnal arousal (42–44). Overall, our findings support the hypothesis that the CBT rate of change is associated with various measures of nocturnal cardiac activity and suggest a potential link between presleep thermal regulation and nocturnal cardiac autonomic dysregulation.

CBT in humans follows a diurnal rhythm, with sleep occurring following the maximal rate of reduction in CBT in the evening (3, 5, 6). The reduction in CBT before sleep is a product of the interplay between heat production and heat loss. Heat dissipation through skin vasodilation primarily mediates this reduction in evening CBT, whereas a greater increase in heat production facilitates the opposite effect in the morning (6). Efficiency of heat dissipation through distal vasodilation is impacted by postural changes, relaxation, and other factors in normal settings (45, 46). The current study utilized CBT rate of change as a measure of presleep thermoregulation, with the notion that a greater presleep CBT rate of decline would indicate efficient thermoregulatory processes before sleep. Our findings demonstrate that blunted CBT dipping before sleep was significantly associated with reduced N3 and increased wake EEG-rhythm (α) intrusion during the sleep period, indicating that those with blunted presleep CBT dipping exhibited impaired sleep quality. This finding aligns with previous reports of a significant association between CBT and slow wave activity (47), particularly within the first sleep cycle. Further support comes from experimental manipulations resulting in greater heat dissipation, which facilitate greater reductions in CBT, and ultimately increased duration and depth of N3 sleep (8, 11, 12).

In addition to its relationship to poorer sleep quality and depth, CBT dipping before sleep was associated with nocturnal cardiac parasympathetic control. Previous findings have indicated that presleep thermoregulatory processes occur in tandem with changes in HRV (25), which persist into the sleep period and are unaffected by the time spent asleep (14). However, findings connecting presleep thermoregulation and nocturnal HRV are sparse, with exploration of individual differences equivocal. Only two studies have explored the sleep onset period and its intersection with thermoregulation and HRV and reported contrasting results (25, 48). Although Okamoto-Mizuno et al. (25) observed tandem reductions in CBT and increases in HRV before sleep onset in young healthy individuals, Anders et al. (48) did not observe associations between CBT and HRV in individuals who experience difficulty with sleep initiation and discomfort from cold extremities. The dichotomous findings of Okamoto-Mizuno et al. (25) and Anders et al. (48) may underscore the importance of individual thermoregulatory differences and nocturnal HRV. Although findings relating thermoregulation to nocturnal HRV are sparse, utilization of a high-heat capacity mattress has been shown to increase heat dissipation and CBT reductions along with a simultaneous modest reduction in nocturnal HR (8, 12). However, within these studies, HRV was not measured. The present findings support those of Okamoto-Mizuno et al. (25), and indicate a relationship between individual differences in presleep CBT dipping and nocturnal HRV, although the directionality of this relationship remains unknown.

Vigorous exercise before bed has been shown to simultaneously increase CBT and HR before sleep, with the latter remaining elevated during the sleep period (49). The authors, however, reported unchanged HRV during the subsequent sleep period, indicating parasympathetic recovery following exercise (49). Conversely, eliciting acute psychological stress before sleep results in reduced HRV that persists into the sleep period (26–28). Reductions in presleep arousal result in lasting increases in HRV during the sleep period (50), indicating an impact of presleep regulatory processes on nocturnal parasympathetic control. It is possible that presleep hyperarousal in the current data set of objectively poor sleepers contributes to the current findings. As noted in the results, reduced CBT rate of change was associated with elevated HR and reduced HRV throughout the entirety of the night. The participants in the current study may habitually experience hyperarousal, which then impairs thermoregulatory processes around the sleep period. Whether presleep arousal interacts with thermoregulatory processes to inhibit heat dissipation before sleep in the current data set remains unknown.

Although heat dissipation primarily drives CBT dipping before sleep (6), excessive heat production may also play a role in the present study. Changes in heat production are highly correlated to HR (6). In individuals with insomnia maintained under a 26-h constant routine schedule, distal skin temperature exceeded levels in healthy individuals (51). Conversely, CBT remained higher in those with insomnia, suggesting a role for excessive heat production rather than inadequate heat dissipation (51). Because the participants in the current study were objectively poor sleepers, it is possible that excessive heat production rather than impaired heat dissipation before sleep may be driving the blunted CBT decline. Measures of metabolic rate and heat production were not utilized in the current study, and future studies are necessary to disentangle the associations between heat production versus dissipation and nocturnal HRV in poor sleeping adults.

Along with heat dissipation and production, which are heavily influenced by the autonomic nervous system, circadian influences may play a significant role in the observed relationships. Utilization of constant routine and forced desynchrony protocols have elucidated circadian variations in HRV measures (16–18). Experimental evidence supports a role for circadian alignment in cardiac autonomic function (23). Sleep restriction in healthy adults reduces nocturnal HRV when paired with circadian misalignment (23). This is relevant to the current data set, in which the average TST of the participants fell below 7 h when measured via actigraphy and PSG, indicating chronic sleep curtailment. The relationship observed between CBT decline and nocturnal HRV may be influenced by a certain level of circadian misalignment present in the current data set of objectively short sleepers. Some of the participants in the current study did not exhibit reductions in CBT before sleep, suggesting they may have been sampled at a circadian rise or plateau in CBT coinciding with their “wake maintenance zone” (13, 52). However, when the habitual (actigraphy) and experimental (PSG) bedtimes were compared with measures of HRV and CBT rate of decline, no significant relationships were present (all P > 0.20). Moreover, declines in CBT are temporally and causally linked to plasma melatonin concentrations (53). Nocturnal melatonin concentrations account for ∼40% of the amplitude observed in circadian CBT rhythms through the promotion of skin blood flow (54). Recently, an in vitro study reported that melatonin administration improved HRV when directly applied to cardiomyocytes (55). This offers mechanistic underpinnings between the in-phase relationship between HF-HRV, a measure of cardiac parasympathetic activity, and melatonin metabolites (urinary 6-sulfatoxy-melatonin; 56), and further explains previous studies indicating a beneficial impact of melatonin supplementation on cardiac function in adults (57). However, although reductions in HR are observed following melatonin administration (57–59), the effect of melatonin on cardiac parasympathetic activity is not as well established in humans with many studies observing no effect (58–60). Despite null effects on cardiac autonomic activity, the same studies observed significant reductions in systolic blood pressure (58, 60). This may be mediated through direct effects of melatonin on cardiomyocyte activity (55) and peripheral vascular tone (61) rather than directly on central autonomic outflow. Nonetheless, the present study was not designed to assess the influence of circadian mechanisms and cannot rule out the influence of melatonin onset or concentrations.

The current study holds implications for populations who suffer from impaired thermoregulation surrounding the sleep period, namely those with insomnia (13). The present study is the first, to our knowledge, to assess the relationship between presleep thermoregulation and HRV in objectively and subjectively poor sleepers, with attentiveness to differences in presleep CBT decline. Clinical populations with altered thermoregulation often exhibit disrupted circadian rhythms and autonomic regulation of cardiac function. For instance, some studies have indicated impaired peripheral vasodilation and heat dissipation in individuals with insomnia (62), whereas others have shown unchanged, or elevated distal skin temperature before sleep onset (51). Indeed, studies have reported an overall augmented CBT in individuals with insomnia compared with healthy adults (51, 63), indicating a potential increase in heat production and supporting the idea of physiological hyperarousal in these individuals (64). This finding is bolstered by evidence of an augmented metabolic rate in individuals with insomnia, which might subsequently elevate CBT (65). Additional observations of delayed circadian phase in sleep onset insomnia (66) and advanced circadian phase in sleep maintenance insomnia (67) suggest a key role for circadian rhythm on dysfunctional thermoregulation around the sleep period in individuals with insomnia, although alterations in circadian rhythmicity have not been reported by others (63, 68). Although both circadian and autonomic influences are likely to impact thermoregulation in individuals with insomnia, decrements in HRV in this population are far less consistent. Studies investigating autonomic regulation of cardiac function have indicated a reduction in HRV in individuals with insomnia, particularly those who exhibit objectively poor sleep quality (69–71). Unfortunately, these findings are largely inconsistent as pointed out in a review by Dodds et al. (72), who indicated that consensus regarding cardiac autonomic deficiencies in individuals with insomnia cannot be reached with the current findings in the field due to numerous limitations. A recent study by Rösler et al. (73) utilized continuous 4-day, at-home ambulatory assessment of sleep and ECG to characterize cardiac autonomic activity in a large sample of individuals with mild to severe insomnia. The authors observed a significantly augmented HR before nocturnal postural changes, as well as a prolonged HR response following the postural change (73). Our laboratory has noted similar HR responses to nocturnal arousals in objectively poor sleeping adults (31). Rösler et al. (73) similarly reported a smaller nighttime reduction in HR, and diminished sleep HRV in individuals with insomnia compared with controls, a finding that was pronounced in individuals with augmented insomnia severity. The use of multiday testing within the home setting, in tandem with both cardiac and sleep testing, suggest the findings of Rösler et al. (73) provide evidence of impaired cardiac autonomic function in individuals with chronic insomnia that is dependent on the severity of the symptoms. Future studies may benefit from integrating individual differences in insomnia severity and sleep difficulty into assessment of both thermoregulation and cardiac autonomic function, as this may better highlight the relationship between the two processes in a disease state.

Aging is associated with worsening sleep quality, marked by reductions in slow wave sleep, total sleep time, sleep efficiency, and an increased frequency of sleep disturbances (74, 75). Aging is similarly marked by alterations in numerous thermoregulatory processes. Raymann et al. (10, 76) have demonstrated that subtle changes in skin temperature impact sleep propensity similarly in elderly and younger adults despite older adults having a reduced subjective ability to detect changes in temperature. These findings indicate that sensitivity of the sleep system to changes in body temperature appears to be maintained with aging, although processes that facilitate heat dissipation, as well as the subjective determination of ideal environmental conditions, are diminished. The current study consisted of both younger and older adults between the ages of 20 and 59 yr, meaning a potential influence of age on both CBT patterns and nocturnal autonomic activity may be present. We did not have an adequate sample size to assess the influence of aging on the observed results within the current study.

We acknowledge the following limitations in our study. First, given the associative nature of the findings, we cannot definitively determine what mechanisms are underlying the present relationship between presleep CBT rate of change and nocturnal HRV. Furthermore, the study design did not allow for separate assessment of autonomic and circadian effects that may be underlying individual differences in presleep thermoregulation. Whether impaired heat loss, heat production, circadian influences, or some combination of these factors are underlying this association remains unknown. Furthermore, whether chronic cardiovascular dysfunction and elevated cardiac activity are causing the blunted CBT decline before sleep is unknown. In populations with chronic cardiovascular dysfunction such as heart failure, skin vasodilatory responses to heat are blunted (77), although the impact of cardiovascular disease on nocturnal thermoregulation has not been adequately characterized, and remains an avenue of future research. The current study provides evidence that supports future assessment of individual differences in circadian output, autonomic control of thermoregulation, and other variables that may underly the presented data. Previous findings have indicated a clear impact of menstrual phase on both autonomic function (38, 70) and CBT (78). As menstrual phase was not controlled for, we are unable to assess any menstrual phase-dependent effects on the presented relationships. Whether the current findings are similarly observed in healthy sleeping adults is unknown due to the specific recruitment of poor sleepers for this study. Future studies assessing whether sleep thermoregulation and cardiac parasympathetic control are altered in healthy sleepers may be warranted. In the current sample, quiet wake cardiac activity was not assessed, but will be important in future work. Finally, we were unable to assess the impact of sex and age on this association given the limited sample size and distribution of males, females, and ages.

Conclusions

The present study demonstrates that individual differences in CBT decline before sleep are associated with nocturnal HRV and HR. Individuals with blunted CBT dipping exhibit reduced nocturnal parasympathetic activity throughout the night. These findings suggest an intertwined relationship between sleep, thermoregulation, and nocturnal autonomic dysregulation in objectively poor sleepers. However, altered presleep thermoregulation could be caused by numerous mechanisms, including an increase in heat production, reduced distal heat dissipation, or potential circadian influences. Furthermore, whether the reduced HRV precedes or follows impairments in presleep thermoregulation cannot be inferred from the present study. The mechanism behind the association between thermoregulation and nocturnal cardiac autonomic activity in poor sleepers, including a likely circadian influence, warrants further investigation.

DATA AVAILABILITY

Data will be made available upon reasonable request.

GRANTS

This study was supported by National Institutes of Health Grants AA0024892, U54GM115371, and P20GM103474.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

J.A.B. and J.R.C. conceived and designed research; J.A.B. and J.R.C. performed experiments; J.A.B., E.L.C., E.B., J.E.G., and J.R.C. analyzed data; J.A.B., E.L.C., E.B., J.E.G., and J.R.C. interpreted results of experiments; J.A.B., E.L.C., and J.R.C. prepared figures; J.A.B., E.L.C., E.B., J.E.G., and J.R.C. drafted manuscript; J.A.B., E.L.C., E.B., J.E.G., and J.R.C. edited and revised manuscript; J.A.B., E.L.C., E.B., J.E.G., and J.R.C. approved final version of manuscript.