Heart rate variability during sleep onset in patients with insomnia with or without comorbid sleep apnea

Abstract

Objectives:

Pre-sleep stress or hyperarousal is a known key etiological component in insomnia disorder. Despite this, physiological alterations during the sleep onset are not well-understood. In particular, insomnia and obstructive sleep apnea (OSA) are highly prevalent co-morbid conditions, where autonomic regulation may be altered. We aimed to characterize heart rate variability (HRV) during sleep onset as a potential measure of pre-sleep hyperarousal.

Methods:

We described the profile of pre-sleep HRV measures and explore autonomic differences in participants with self-reported insomnia disorder (with no OSA, n = 69; with mild OSA, n = 70; with moderate or severe OSA, n = 66), compared to normal sleep controls (n = 123). Heart rate data during the sleep onset process were extracted for HRV analyses.

Results:

During the sleep onset process, compared to normal sleep controls, participants with insomnia had altered HRV, indicated by higher heart rate (p = 0.004), lower SDNN (p = 0.003), reduced pNN20 (p < 0.001) and pNN50 (p = 0.010) and lower powers (p < 0.001). Participants with insomnia and moderate/severe OSA may have further deteriorated HRV outcomes compared to no/mild OSA patients with insomnia but differences were not significant. Insomnia itself was associated with significantly higher heart rate, lower pNN20, and lower high frequency power even after adjustment for age, gender, BMI and OSA severity.

Conclusions:

Participants with insomnia had lower vagal activity during the sleep onset period, which may be compounded by OSA, reflected in higher heart rates and lower HRV. These altered heart rate dynamics may serve as a physiological biomarker for insomnia during bedtime wakefulness, or as a potential tool to evaluate the efficacy of behavioral interventions which target bedtime stress.

1. Introduction

Difficulty falling asleep is common in patients with insomnia, with stressful thoughts [1–3] and a sense of urgency about falling asleep adversely affecting sleep onset [4]. Hyperarousal is a key component in modern etiological models of insomnia disorder [5]. Pre-sleep hyper--arousal refers to a state of physiological and psychological activation that occurs before bedtime, making it difficult for individuals to transition into sleep. Beyond that, as a core feature of insomnia, hyper-arousal can manifest itself at bedtime or globally throughout the day or night, and has not been distinguished between state and trait hyper-arousal [6–8]. There has been support for the concept that hyperarousal processes from the molecular to the higher systems level play a key role in the pathophysiology of primary insomnia. However, physiological alterations during the sleep onset period are not well-understood. In addition, in clinical evaluation, insomnia is most commonly assessed through a combination of subjective and self-reported measures, and physiological biomarkers have been underexplored.

The most commonly used measurement for autonomic regulation is heart rate variability (HRV), which has been broadly used in multiple contexts and among different populations [9–13]. HRV reflects the dynamic interplay between these two branches of the autonomic nervous system (ANS), sympathetic nervous system (SNS) and parasympathetic nervous system (PNS). The sleep onset process involves a complex transition from wakefulness to sleep involving the SNS and PNS, and chronic insomnia may lead to sustained autonomic imbalance, contributing to the perpetuation of sleep disturbances and associated health problems. Therefore, tracking HRV during sleep onset may help us understand how arousal levels change as an individual attempts to fall asleep, enable the tailoring of treatments to address wake-sleep transition issues, or provide a biomarker for the effectiveness of treatments. Currently, although results are inconsistent, several studies have compared differences of autonomic control by means of HRV during different sleep stages [14–18], suggesting decreased HRV related parasympathetic activity in insomnia patients during sleep [19,20]. However, results of HRV of patients with insomnia vs controls during the transition from wake to sleep (sleep onset) are inconsistent [21–24]. Recent research has begun to explore physiological signals during sleep-onset. For example, Tsai et al. [25] reported associations between prolonged sleep onset latency and heart rate dynamics among young patients with insomnia compared to “good sleepers”, but the small sample size (19 young patients with insomnia and 14 “good sleepers”) limits the interpretation and generalizability of these findings.

HRV has also been proposed to be a marker of OSA severity during both wakefulness [26] and sleep [27]. A recent study reported that a 5-min ECG recording during wakefulness provides comprehensive insights regarding variation in HRV patterns with OSA severity [26]. There were no differences between participants with no OSA (AHI<5) and those with mild OSA (AHI 5–15), however, individuals with severe OSA had markedly reduced HRV compared to those without OSA. Insomnia and OSA are highly prevalent, with the co-occurrence of both insomnia and OSA symptoms ranging from 55 % to 84 % of patients presenting to sleep clinics [28]. Both conditions cause common clinical impairments that largely overlap [29–31], and lead to significant morbidity. Patients with comorbid insomnia and sleep apnea present challenges to clinical management and need a standardized systematic approach to identifying and treating these sleep disorders when they co-occur [32]. Physiological alterations and heart rate dynamics during sleep onset in this comorbid group has not been studied.

Further research is needed to better understand the relationship between heart rate dynamics during sleep onset and pre-sleep wakefulness in these commonly encountered sleep disorders. Within this context, we aimed to describe the profiles of heart rate dynamics during the sleep onset process, specifically comparing participants with or without insomnia, and comparing those with insomnia plus or minus comorbid OSA, to explore whether HRV measures can be used as a physiological biomarker in the assessment of autonomic alteration/activation during pre-sleep wakefulness/sleep onset.

2. Material and methods

2.1. Dataset and included participants

Data used in this analysis were obtained from the Sleep Heart Health Study (SHHS), a multi-center cohort study implemented by the National Institutes of Health, National Heart Lung & Blood Institute to determine the cardiovascular and other consequences of sleep-disordered breathing. Participants’ baseline visits were between 1995 and 1998 utilizing existing community-based cohorts as described previously [33]. Eligible participants were at least 40 years old and not receiving active treatment for OSA (e.g., continuous positive airway pressure, oral appliance, and oxygen therapy) [34]. At baseline visits, trained research investigators conducted interviews. Sociodemographic characteristics, sleep habits, cardiometabolic factors, overall health, medication use, anthropometric and blood pressure measurements, questionnaires (e.g., Epworth Sleepiness Scale, ESS) were included. The dataset included 5805 participants who successfully completed baseline PSG at-home using a portable monitor (Compumedics P-series, Abbotsford, Victoria, Australia). Sensors were placed and equipment was calibrated during the evening home visit by a certified technician. Participants were not asked to go to bed at a specific time. The sleep testing unit started recording 30 min prior to the participant’s usual bedtime. Recordings included electroencephalogram (EEG), electrooculogram (EOG), electrocardiogram (ECG), chin electromyogram EMG), pulse oximetry, chest and abdominal excursion by inductance plethysmography, airflow by thermal sensor, and body position channels. All PSG recordings were scored using Rechtschaffen and Kales criteria at a centralized reading center by trained research technicians blinded to all clinical data [35].

From the original dataset [33,36], we excluded a participant’s data if (1) overall PSG signal quality was below “good” (all data were labelled with different levels of quality after manual check), (2) more than 30 min of the sleep period had either lost or unscorable EEG data, (3) the recording either started or ended in sleep, (4) the sleep latency was not reliable based on annotations from time stamps and annotations (5) participants had substance use (e.g. liquor, coffee, “unusual medication”, etc.), prior to sleep, or (6) participants with recent use of medication that may affect sleep or physiological signals (e.g., benzodiazepines, beta blockers, tricyclic anti-depressants, etc. within two weeks of their visit).

At baseline, participants completed a Sleep Habits Questionnaire that elicited standardized information on sleep symptoms and patterns. Four insomnia-related questionnaire items queried whether participants: “Have trouble falling asleep,” “Wake up during the night and have difficulty getting back to sleep,” “Wake up too early in the morning and unable to get back to sleep,” and “Take sleeping pills or other medication to help (you) sleep” [37].

Participants with insomnia were identified based upon self-report of at least one of the above symptoms with affirmative responses occurring almost always (16–30 nights per month). Participants without insomnia were identified if they never (zero nights per month) or rarely (<1–2 nights per month) reported any of the above symptoms. Given the high co-occurrence of insomnia and OSA, we further classified the participants with self-reported insomnia into subgroups according to apnea–hypopnea index (AHI, ≥3 percent oxygen desaturation recorded per hour of sleep): insomnia with no OSA, insomnia with mild OSA, and insomnia with moderate or severe OSA. According to the American Academy of Sleep Medicine (AASM) standards, OSA is categorized into mild (5–15 events/hour), moderate (15–30 events/hour), and severe (>30 events/hour). Moderate-to-severe obstructive sleep apnea, defined as an AHI score of 15 or more apnea or hypopnea events per hour, is an independent risk factor for insulin resistance, dyslipidemia, vascular disease, and death [38–44]. Therefore, our analysis included four groups: (1) normal sleep controls (NSC, no insomnia with AHI<5 events/hour), (2) insomnia with no OSA (AHI<5 events/hour, (3) insomnia with mild OSA (AHI ≥5 events/hour and <15 events/hour), and (4) insomnia with moderate or severe OSA (AHI ≥15 events/hour).

As this was a secondary analysis with only publicly available deidentified data, it did not involve a research protocol requiring approval by an institutional review board or ethics committee.

2.2. Heart rate variability (HRV)

ECG data (sampling rate 125Hz) were extracted from PSG raw data files. Then heart rate data during the sleep onset process were extracted for HRV analysis using MATLAB (MathWorks Inc, Natick, MA). The sleep onset process was defined as the interval between ‘lights-off’ and sleep onset, with varied lengths of sleep onset latency among participants. Sleep onset was defined by three consecutive epochs of stage 1 or one epoch of any other sleep stage. In the time domain, mean intervals between normal-to-normal heart beats (mean NN), standard deviation of NN (SDNN), the root mean square of successive differences between normal heartbeats (RMSSD), percentage of adjacent NN intervals that differ from each other by > 20 ms (pNN20) and >50 ms (pNN50) were evaluated [45–48]. In the frequency domain, NN intervals were interpolated and resampled to 4 Hz for HRV frequency domain analysis [45–48]. The Welch protocol (with a Hamming window applied to each 5 min segment, and moving window approach with 50 % overlap) was used for spectral analysis. Frequency domain indices included total power (TP), very low frequency (VLF), low frequency (LF) and high frequency (HF) [45–48]. HRV power spectrum measurements were also log-transformed to normalize their distribution for analysis (LnTP, LnLF and LnHF).

HRV indicates neurocardiac function and is generated by heart-brain interactions and dynamic non-linear autonomic nervous system processes. In short-term resting recordings, the primary source of the variation is parasympathetically-mediated RSA, and HF is closely correlated with PNS activity, while lower HF or higher LF power is correlated with stress, panic, anxiety, or worry [49]. The modulation of vagal tone helps maintain the dynamic autonomic regulation important for cardiovascular health, and may facilitate sleep onset.

2.3. Statistical analysis

Statistical analyses were performed using IBM SPSS Statistics version 25 (IBM Corp., Armonk, NY, USA). Demographic characteristics included age, gender, body mass index (BMI), blood pressure, ethnicity, and race. We included standard PSG outcomes, including AHI, sleep onset latency (SOL), total sleep time (TST), wake after sleep onset (WASO) and sleep efficiency (SE). Categorical variables were tested using chi-squared test or Fisher’s Exact Test, wherever appliable. Continuous variables with normal distributions were reported as mean ± standard deviation in tables, and presented as mean with standard error in figures. Independent t-tests were used to compare participants with insomnia with normal sleep controls. Then, normal sleep controls and participants in the insomnia subgroups by different levels of OSA were compared using ANOVA. If significant differences were found among groups, post-hoc tests using LSD were performed. To understand the associations of HRV and insomnia, we first conducted linear regression models with adjustment for basic demographic information (age, gender, BMI), then we further added AHI in the models to control for OSA severity. Given the differences in the ratio of patients with hypertension and its potential impact on HRV, we performed the previous models with adjustment for hypertension. P-values less than 0.05 indicate significant differences.

3. Results

3.1. Demographic characteristics

We included 123 normal sleep controls and 205 participants with insomnia. There were differences on some demographic characteristics such as age, BMI, and ratio of individuals with hypertension, so these factors were later adjusted for in regression analyses. In terms of PSG-based sleep outcomes, significant differences were expected due to the study design and eligibility criteria.

Among those with insomnia, there were 69 participants with self-reported insomnia with no OSA, 70 participants with insomnia and mild OSA, and 66 participants with insomnia and moderate or severe OSA. By design, significant differences were expected in certain demographic characteristics including BMI, AHI, and number of participants with hypertension (Table 1), with higher morbidity in participants with insomnia and moderate/severe OSA. For gender, there were more female participants in all groups, which was a general characteristic of the overall SHHS dataset.

Table 1.

Demographic characteristics.

	Normal sleep controls	Patients with insomnia	p a	Insomnia subgroups
				Insomnia with no OSA	Insomnia with mild OSA	Insomnia with moderate/severe OSA	p b
	(n = 123)	(n = 205)		(n = 69)	(n = 70)	(n = 66)
Demographics
Age (years)	58.8 ± 11.2	64.5 ± 11.1	<0.001	63.0 ± 11.7	63.6 ± 11.5	67.2 ± 9.8	0.065
BMI (kg/m2)	25.5 ± 4.1	28.3 ± 5.0	<0.001	26.1 ± 3.5	28.8 ± 5.4	30.0 ± 5.2	<0.001
Gender
Male	46 (37.4 %)	71 (34.6 %)	0.613	12 (17.4 %)	27 (38.6 %)	32 (48.5 %)	<0.001
Female	77 (62.6 %)	134 (65.4 %)		57 (82.6 %)	43 (61.4 %)	34 (51.5 %)
Ethnicity
Hispanic or Latino	10 (8.1 %)	17 (8.3 %)	0.959	11 (15.9 %)	2 (2.9 %)	4 (6.1 %)	0.015
Not Hispanic or Latino	113 (91.9%)	188 (91.7%)		58 (84.1 %)	68 (97.1 %)	62 (93.9 %)
Race
White	94 (76.4 %)	175 (85.4 %)	0.100	54 (78.3 %)	62 (88.6 %)	59 (89.4 %)	0.027
Black	13 (10.6 %)	11 (5.4 %)		3 (4.3 %)	6 (8.6 %)	2 (3.0 %)
Other	16 (13.0 %)	19 (9.3 %)		12 (17.4 %)	2 (2.9 %)	5 (7.6 %)
Hypertension	44 (35.8 %)	104 (50.7 %)	0.008	32 (46.4 %)	35 (50.0 %)	37 (56.1 %)	0.525
Diabetes	9 (7.6 %)	15 (7.6 %)	0.997	2 (3.0 %)	7 (10.3 %)	6 (9.7 %)	0.211
PSG-based sleep measures
AHI (events/hr)	2.6 ± 1.4	13.2 ± 14.6	<0.001	2.5 ± 1.4	9.1 ± 2.7	28.6 ± 16.8	c
SOL (minutes)	19.0 ± 13.9	25.4 ± 24.8	0.003	23.3 ± 20.1	27.5 ± 28.3	25.4 ± 25.4	0.042
TST (minutes)	369.6 ± 46.1	350.8 ± 61.2	0.002	349.8 ± 67.9	353.4 ± 56.5	349.1 ± 59.4	0.910
WASO (minutes)	46.2 ± 34.4	80.9 ± 10.4	<0.001	54.3 ± 38.2	59 ± 38.1	61.3 ± 43.6	0.585
SE (%)	85.3 ± 8.1	58.2 ± 39.9	0.006	81.7 ± 10.8	80.6 ± 10.5	80.2 ± 10.0	0.690
Quality of Life
PCS score	51.7 ± 6.9	44.0 ± 11.3	<0.001	45 ± 10.4	44.5 ± 11.3	42.3 ± 12.3	<0.001
MCS score	56.0 ± 6.1	49.9 ± 10.1	<0.001	49.5 ± 9.4	50.9 ± 10	49.2 ± 10.9	<0.001

Abbreviations: AHI, apnea-hypopnea index; BMI, body mass index; MCS, mental component summary; OSA, obstructive sleep apnea; PCS, physical component summary; SE, sleep efficiency; SOL, sleep onset latency; TST, total sleep time.

^aindicates the p values from t-test for comparing normal sleep controls vs patients with insomnia.

^bindicates the p values from ANOVA test for comparing the three subgroups of insomnia.

^cIndicates that the variables were expected to be significantly different among groups by study design.

3.2. Heart rate variability during the sleep onset process

Compared to normal sleep controls (n = 123), participants with insomnia (n = 205) had significantly higher heart rates (73.2 ± 12.8 beats/min vs 69.7 ± 9.4 beats/min, p = 0.004), lower SDNN (64.2 ± 32.5 ms vs 77.4 ± 42.7 ms, p = 0.003), lower pNN20 (0.27 ± 0.09 % vs 0.31 ± 0.08 %, p < 0.001), lower pNN50 (0.12 ± 0.1 % vs 0.15 ± 0.09 %, p = 0.010), lower LnTP (8.14 ± 0.89 vs 8.51 ± 0.86, p < 0.001), lower LnLF (6.56 ± 1.03 vs 6.98 ± 0.9, p < 0.001), and lower LnHF (6.41 ± 1.09 vs 6.83 ± 0.95, p < 0.001). Among the three insomnia subgroups, there were no significant differences in the HRV outcomes (Table 2).

Table 2.

Heart rate variability during sleep onset process.

	Normal sleep controls	Patients with insomnia	p a	Insomnia subgroups
				Insomnia with no OSA	Insomnia with mild OSA	Insomnia with moderate/severe OSA	p b
	(n = 123)	(n = 205)		(n = 69)	(n = 70)	(n = 66)
Time Domain
HR	69.67 ± 9.43	73.22 ± 12.75	0.004	74.44 ± 14.99	70.85 ± 10.2	74.45 ± 12.45	0.163
SDNN	77.44 ± 42.72	64.15 ± 32.51	0.003	69.49 ± 32.72	61.62 ± 30.0	61.31 ± 34.53	0.253
RMSSD	44.27 ± 19.22	39.86 ± 21.14	0.060	43.45 ± 21.96	37.5 ± 19.90	38.62 ± 21.37	0.219
pNN20	0.31 ± 0.08	0.27 ± 0.09	<0.001	0.28 ± 0.09	0.26 ± 0.10	0.27 ± 0.09	0.268
pNN50	0.15 ± 0.09	0.12 ± 0.10	0.010	0.14 ± 0.10	0.11 ± 0.10	0.12 ± 0.09	0.218
Frequency domain
LnTP	8.51 ± 0.86	8.14 ± 0.89	<0.001	8.27 ± 0.97	8.11 ± 0.80	8.03 ± 0.88	0.293
LnLF	6.98 ± 0.90	6.56 ± 1.03	<0.001	6.67 ± 1.06	6.55 ± 10.00	6.45 ± 1.03	0.485
LnHF	6.83 ± 0.95	6.41 ± 1.09	<0.001	6.59 ± 1.10	6.36 ± 1.10	6.28 ± 1.12	0.227

Abbreviations: HR, heart rate; SDNN, standard deviation of normal beat intervals; RMSSD, root mean square of successive differences between normal heartbeats; pNN20, the proportion of consecutive RR intervals that differ by more than 20 ms; pNN50, the proportion of consecutive RR intervals that differ by more than 50 ms; LnTP, the natural logarithm of total power; LnLF, the natural logarithm of low frequency power; LnHF, the natural logarithm of high frequency power.

^aindicates the p values from t-test for comparing normal sleep controls vs patients with insomnia.

^bindicates the p values from ANOVA test for comparing the three subgroups of insomnia.

When comparing the four groups, significant differences were found in most of the HRV outcomes, including HR (p = 0.010), SDNN (p = 0.007), pNN20 (p = 0.001), pNN50 (p = 0.021), LnTP (p = 0.001), LnLF (p = 0.001), and LnHF (p = 0.001). Post-hoc comparisons revealed that normal sleep controls had the lowest HR, highest SDNN, highest pNN20 and pNN50, and the highest frequency domain measures, while participants with insomnia and moderate/severe OSA had the highest HR, lowest SDNN, lowest pNN20 and pNN50, and the lowest frequency domain measures (Supplementary Fig. 1).

3.3. Associations between HRV measures and sleep onset latency

When controlling for age, gender, BMI and sleep apnea severity (characterized by AHI), partial correlation showed that certain HR dynamic measures significantly correlate with SOL. Correlations were mild to moderate, for example, average HR (r = −0.117, p = 0.028) and SDNN (r = −0.219, p < 0.001). Linear regression models with adjustment for age, gender, BMI and AHI also indicated significant impacts of HR dynamics on SOL, including HR (β = −0.22, p = 0.036) and SDNN (β = −0.13, p < 0.001).

3.4. Effects of insomnia on heart rate variability measures

Our regression models showed that insomnia had significant impacts on HR (β = 3.249, 95%CI: 0.47–6.03; p = 0.022), pNN20 (β = −0.033, 95%CI: −0.05 to −0.01; p = 0.003), pNN50 (β = −0.024, 95%CI: −0.05 –0.001; p = 0.037), LnTP (β = −0.211, 95%CI: −0.42 –0.001; p = 0.048) and LnHF (β = −0.334, 95%CI: −0.58 to −0.08; p = 0.009), controlling for age, gender and BMI. With additional adjustment for AHI or AHI and hypertension, insomnia’s significant impacts remained on HR, pNN20, and LnHF, and the values of β were relatively unchanged (Table 3).

Table 3.

Effects of self-reported insomnia on heart rate variability measures during sleep onset period.

Outcomes	Insomnia			Insomnia			Insomnia
	Adjusted for age, gender, BMI			Adjusted for age, gender, BMI and AHI			Adjusted for age, gender, BMI, AHI and HTN
	B	95 % CI	p	B	95 % CI	p	B	95 % CI	p
HR	3.249	(0.47, 6.03)	0.022*	3.141	(0.2, 6.08)	0.036*	3.222	(0.27, 6.17)	0.032*
SDNN	−8.052	(−16.83, 0.72)	0.072	−7.581	(−16.88, 1.72)	0.110	−7.594	(−16.94, 1.75)	0.111
RMSSD	−3.676	(−8.54, 1.19)	0.138	−3.419	(−8.58, 1.74)	0.193	−3.701	(−8.88, 1.48)	0.161
pNN20	−0.033	(−0.05, −0.01)	0.003*	−0.031	(−0.05, −0.01)	0.009*	−0.032	(−0.06, −0.01)	0.007*
pNN50	−0.024	(−0.05, 0.001)	0.037*	−0.022	(−0.05, 0.002)	0.074	−0.023	(−0.05, 0.001)	0.060
LnTP	−0.211	(−0.42, 0.001)	0.048*	−0.190	(−0.41, 0.03)	0.093	−0.201	(−0.43, 0.02)	0.077
LnLF	−0.192	(−0.42, 0.04)	0.100	−0.189	(−0.43, 0.05)	0.127	−0.207	(−0.45, 0.04)	0.096
LnHF	−0.334	(−0.58, −0.08)	0.009*	−0.297	(−0.56, −0.03)	0.028*	−0.315	(−0.58, −0.05)	0.020*

Abbreviations: AHI, apnea-hypopnea index; BMI, body mass index; 95 % CI, 95 % confidence interval; HR, heart rate; SDNN, standard deviation of normal beat intervals; RMSSD, root mean square of successive differences between normal heartbeats; pNN20, the proportion of consecutive RR intervals that differ by more than 20 ms; pNN50, the proportion of consecutive RR intervals that differ by more than 50 ms; LnTP, the natural logarithm of total power; LnLF, the natural logarithm oflow frequency power; LnHF, the natural logarithm of high frequency power; HTN, hypertension.

4. Discussion

Using an existing publicly-available sleep dataset, we conducted a secondary analysis to describe the profile of pre-sleep HRV measures and explore autonomic differences in participants with self-reported insomnia with and without OSA, and normal sleep controls. We found significant differences in both time and frequency domain measures between groups. During sleep onset, participants with insomnia, regardless of OSA severity, had decreased HRV, indicated by higher heart rate, lower SDNN, reduced pNN20 and pNN50, and lower high frequency powers. Participants with moderate/severe OSA had further deteriorated HRV outcomes. Self-reported insomnia showed a significant effect on heart rate, pNN20, LnHF even after adjustment for age, gender, BMI and OSA severity. Our results suggest that participants with insomnia had lower vagal activity during the sleep onset period, reflected in higher heart rates and lower heart rate variability, which may serve as a physiological biomarker for insomnia during bedtime wakefulness.

The transition from wakefulness to sleep involves a complex interplay of physiological changes in the body, which are regulated by the circadian rhythm, homeostatic processes, and neural and hormonal factors. Heart rate dynamics play a significant role in this transition, reflecting an intricate relationship between the autonomic nervous system (ANS), circadian rhythm, and sleep physiology [50]. For example, as an individual transitions from wakefulness to sleep, parasympathetic nervous system activity typically increases, which leads to a reduction in heart rate and promotes rest and relaxation, facilitating the onset of sleep. The suprachiasmatic nucleus in the brain, known as the master circadian pacemaker, also helps to regulate heart rate by sending signals to the ANS to promote the decreased heart rate in preparation for sleep. Meanwhile, existing evidence supports the bidirectional interactions between insomnia (or other sleep disorders) and sympathetic nervous system activation and other autonomic impairments [51]. In individuals with insomnia or other sleep disorders, altered heart rate variability and arousal responses at night may lead to awakenings and difficulty maintaining sleep [52,53]. A lower wake-to-sleep HR reduction and alterations in HRV variables might mediate the increased rates of cardiovascular morbidity and mortality observed in insomnia patients [54]. Simply examining HR and SDNN, the negative correlation between SDNN and SOL was as expected (more vagal was associated with shorter SOL). However, our results also showed that average HR was inversely correlated with sleep onset latency, which we could not interpret with the current design and analysis. Future studies with a larger sample size or more advanced analysis may be needed to better understand such associations.

Clinically, addressing hyperarousal and underlying causes of stress, anxiety, or other psychological factors contributing to hyperarousal is essential. Despite the notable emphasis on developing objective measures for insomnia and hyperarousal assessment, there have been challenges and limitations in this area due to several factors (e.g., complexity of insomnia, subjective-objective sleep discrepancy, variability in sleep patterns, comorbidities, resource and accessibility constraints, etc.). For objective insomnia measures, there have been continuous efforts toward the development of biomarkers, and opportunities with advances in neuroimaging, and utilizing digital health and wearables, as well as machine learning and artificial intelligence. Most notably, current wearable technology has made heart rate recordings easily accessible, and analysis of HRV-based measures widely available. In fact, incorporating cardiac features or heart rate dynamics in the ambulatory monitoring of sleep may provide a more sensitive biomarker of insomnia than actigraphy alone [55].

Techniques for managing hyperarousal or treating insomnia, such as relaxation training, cognitive-behavioral therapy for insomnia (CBT-I), and stress-reduction strategies, aim to reduce sympathetic activity and promote relaxation, making it easier to initiate and maintain sleep [56, 57]. As an easily accessible objective approach, monitoring heart rate changes can help assess an individual’s readiness for sleep, the presence of sleep disorders, or the effects of interventions designed to improve sleep quality. It is also a valuable tool for studying the autonomic nervous system’s role in sleep physiology and the interactions between the circadian system and sleep regulation.

There are several strengths and limitations in this study. First, our analysis is a secondary analysis using an existing dataset, with characteristics of a cross-sectional, observational nature. We have made efforts to include covariates in the analysis and interpretation of the results, but were not able to include psychometric instruments or control for different sleep environments. However, given the known effects of the lab environment in sleep studies, using data collected from home settings is one strength of this study, as participants are measured in the natural environments with presumed usual bedtime routine before sleep. Also, given the limitations of the parent dataset, we were not able to address the assessment of covariates such as depression, anxiety, and other mood disorders. These factors can significantly influence heart rate variability and should be taken into account in future studies. Second, despite the numerous publications on HRV and sleep, existing studies have focused on sleep itself (e.g., different sleep stages, overnight sleep, various sleep disorders, etc.) or sleep-wake comparisons. Heart rate dynamics during immediate pre-sleep wakefulness in the sleep onset process remain underexplored. Our results thus extend the knowledge on the physiology of the wake-sleep transition, particularly the sleep onset process. Third, hyperarousal can be present at bedtime or globally throughout the day or night, but in this study, we primarily focused on bedtime. Future studies may consider exploring time-of-the-day effects on heart rate dynamics. Finally, we included participants with and without insomnia following criteria used in previous publications, but excluded participants reporting an intermediate level of sleep complaints. We did not run subgroup analyses for subtypes of insomnia. Future analyses in this dataset or others may include all participants with valid data to better understand HRV metrics in a broader population and multiple subtypes of insomnia.

5. Conclusions

Participants with self-reported insomnia with and without OSA had altered HRV measures compared to normal sleep controls, and those with co-morbid disease had the lowest HRV. As measures of autonomic activity, the HRV matrix may serve as a physiological biomarker to indicate hyper-arousal during pre-sleep wakefulness in patients with insomnia. Tracking changes of HRV during sleep onset may have potential value in the evaluation of insomnia treatments, especially those targeting sleep onset difficulties.

Abbreviations

BMI

body mass index

ECG

electrocardiogram

HF

high frequency

HR

heart rate

HRV

heart rate variability

LF

low frequency

LnHF

natural logarithm of high frequency power

LnLF

natural logarithm of low frequency power

NN

normal-to-normal heart beat intervals

OSA

obstructive sleep apnea

pNN20

the proportion of consecutive RR intervals that differ by more than 20 ms

pNN50

the proportion of consecutive RR intervals that differ by more than 50 ms

RMSSD

root mean square of successive differences between normal heartbeats

SDNN

the standard deviation of NN intervals

TP

total power