Reliability of heart rate variability during stable and disrupted polysomnographic sleep

Abstract

Heart rate variability (HRV) is commonly used within sleep and cardiovascular research, yet HRV reliability across various sleep stages remains equivocal. The present study examined the reliability of frequency- and time-domain HRV within stage-2 (N2), slow-wave (SWS), and rapid-eye-movement (REM) sleep during both stable and disrupted sleep. We hypothesized that high-frequency (HF) HRV would be reliable in all three sleep stages, low-frequency (LF) HRV would be reliable during N2 and SWS, and that disrupted sleep via spontaneous cortical arousals would decrease HRV reliability. Twenty-seven participants (11 men, 16 women, 26 ± 1 yr) were equipped with laboratory polysomnography for 1 night. Both frequency- and time-domain HRV were analyzed in two 5- to 10-min blocks during multiple stable and disrupted sleep cycles across N2, SWS, and REM sleep. HF HRV was highly correlated across stable N2 (r = 0.839, P < 0.001), SWS (r = 0.765, P < 0.001), and REM (r = 0.881, P < 0.001). LF HRV was moderate-to-highly correlated during stable cycles of N2 sleep (r = 0.694, P < 0.001), SWS, (r = 0.765, P < 0.001), and REM (r = 0.699, P < 0.001) sleep. When stable sleep was compared with disrupted sleep, both time- and frequency-domain HRV were reliable (α > 0.90, P < 0.05) in N2, SWS, and REM, except for LF HRV during SWS (α = 0.62, P = 0.089). In conclusion, time- and frequency-domain HRV demonstrated reliability across stable N2, SWS, and REM sleep, and remained reliable during disrupted sleep. These findings support the use of HRV during sleep as a tool for assessing cardiovascular health and risk stratification.

NEW & NOTEWORTHY Heart rate variability (HRV) is a commonly employed indirect estimate of cardiac autonomic activity during sleep with limited reliability studies. Nocturnal frequency-domain HRV was reliable across differing stable sleep cycles of stage-2 (N2), slow-wave (SWS), and rapid-eye-movement (REM) sleep. Moreover, frequency- and time-domain HRV were reliable during stable and disturbed sleep, except SWS low-frequency HRV. Our finding supports nocturnal HRV as a potential tool for cardiovascular risk stratification.

INTRODUCTION

Sleep disorders and insufficiencies are associated with both neural and cardiovascular dysfunction (1–3). As a result, heart rate variability (HRV) has been increasingly implemented within sleep research as a means of quantifying nocturnal cardiac autonomic activity (4, 5), with the notion that sleep may be an opportune time to quantify HRV due to the natural control of respiratory sinus arrhythmia (both frequency and depth of the breath) (6). Previous studies have shown that nocturnal HRV is significantly impacted in response to experimental (7, 8) and chronic, pathological sleep disruptions such as obstructive sleep apnea (9), insomnia (10), and periodic limb movement (11). However, the sleep disturbances in these populations may influence the reliability of nocturnal HRV. Moreover, a recent systematic review by Dodds et al. (12) documented inconsistencies regarding the associations between insomnia and HRV, with the vast majority of studies being unable to replicate originally reported reductions of HRV in participants with chronic insomnia (13). Despite these inconsistencies and concerns, HRV continues to be widely used within sleep research.

Although frequency-domain HRV analysis in the high-frequency (HF; 0.15–0.40 Hz) range has been widely accepted as a reliable estimate of parasympathetic cardiac control, low-frequency (LF; 0.04–0.15 Hz) HRV is still being interpreted by some researchers as an index of cardiac sympathetic control, or sympathovagal balance, despite numerous limitations (14). Unlike HF HRV, the LF component of HRV has demonstrated poor intraindividual reliability during wake recordings (15) and often divergent responses to other measures of sympathetic activity such as microneurography (16, 17). This compromised reliability may be further exacerbated by changes in autonomic balance that naturally occur in differing stages with nonrapid-eye-movement (NREM) sleep, including slow-wave sleep (SWS) and stage-2 (N2) sleep, which demonstrate augmented parasympathetic control (i.e., increased HF and time-domain HRV indices) (18, 19), and rapid-eye-movement (REM) sleep that is associated with an increase in peripheral sympathetic activity (20, 21).

To our knowledge, only three previous studies have attempted to assess the reliability of frequency- and time-domain HRV during polysomnographic sleep, and results are inconsistent (18, 22, 23). Varying experimental sleep paradigms (i.e., HRV compared across naps, hours of the night, and multiple nights) are likely contributors to these inconsistencies (18, 22, 23). Herzig et al. (18) remains the only study that examined reliability across a single night of sleep during multiple sleep stages and determined acceptable reliability of HF HRV during SWS, but a lack of reproducibility of both HF and LF HRV in all other stages. Moreover, the findings of Herzig et al. (18) focused exclusively on uninterrupted, stable sleep periods. Despite the high prevalence of repeated nocturnal awakenings and sleep disturbances within even healthy populations, the literature is presently void of HRV reproducibility that includes disturbed sleep periods via cortical arousals. This is clinically relevant because there are a number of studies examining the associations between sleep HRV and cardiometabolic health that have used only stable periods of sleep (24) or have not accounted for the impact of arousal when collapsing HRV across the night (25, 26). Cortical arousals can cause momentary and/or sustained fluctuations in the heart rate (27, 28), suggesting the need to better understand the potential impact of sleep disruption on nocturnal HRV reliability.

Rigor and reliability of physiological studies is a stated priority of the National Institutes of Health (29) and several journals, including the American Journal of Physiology-Heart and Circulatory Physiology (30). Accordingly, the purpose of this study was to determine the reliability of frequency- and time-domain HRV across stage-specific cycles of N2, SWS, and REM sleep within healthy individuals during both stable and disturbed sleep. Reliability within a night will provide a starting point to further determine reliability between nights, an even stronger metric of reliability for HRV. We hypothesized that HF HRV, along with time-domain measures, would be reliable across N2 sleep, SWS, and REM sleep cycles. Because of natural respiratory sinus arrhythmia control in N2 and SWS, we hypothesized that LF HRV would be reliable in N2 sleep and SWS, but not REM sleep. Finally, we hypothesized that the inclusion of sleep with cortical arousals would decrease the overall reliability of frequency- and time-domain HRV in all sleep stages.

METHODS

Participants

The participants were recruited from Michigan Technological University, Montana State University, and the communities surrounding both institutions. Participant inclusion criteria included age 21–45 yr and body mass index (BMI) of 18.5–35 kg/m². Exclusion criteria included patients with diabetes, active smokers, people with current prescription of blood pressure or autonomic medications, or people with the diagnosis and/or treatment of obstructive sleep apnea (OSA). Women could not be prescribed oral, hormonal, or intrauterine birth control in the last 6 mo, could not be pregnant or breastfeeding, and needed to self-report a normal menstrual cycle of 25–32 days. All participants were screened for OSA using an at-home sleep apnea kit. In addition, overnight polysomnography (PSG) studies did not show any indication of other sleep disorders (i.e., OSA, restless leg syndrome, etc.).

Data were obtained both retrospectively (7, 31) and prospectively. Twenty-nine participants completed an in-laboratory polysomnography study night, with stable PSG and ECG recorded for 27 participants (11 men, 16 women, 26 ± 1 yr, 27 ± 1 kg/m²). Michigan Technological University and Montana State University Institutional Review Board approved all the testing procedures, and procedures were in accordance with the Declaration of Helsinki. All participants provided written, informed consent.

Experimental Design

Participants arrived at the laboratory for PSG at 9:00 PM and were set up with the routine PSG setup detailed elsewhere (7, 31), including continuous two-lead electrocardiography (ECG). Once the PSG setup was complete, the participants entered bed at 10:45 PM. Following confirmation of PSG and ECG signal integrity, lights off occurred at 11:00 PM. The participants were allotted an 8-h sleep opportunity with lights on at 7:00 AM. A registered polysomnography sleep technician observed the overnight PSG and ECG over the entire 8-h recording period.

Measurements

Polysomnography.

The overnight PSG (Natus Medical; Middleton, WI) included a 10–20 sleep electrode placement to record and score sleep electroencephalography (EEG) with pairs of central, frontal, and occipital leads. The leads were referenced to the contralateral electrodes on the opposite mastoid process of the head. Electrooculography (EOG) and electromyography (EMG) were recorded using American Association of Sleep Medicine (AASM) standards, with two electrodes lateral to the eyes and three on the chin. A continuous two-lead electrocardiogram measured heart rate (HR) and rhythm. Respiratory inductance plethysmography (RIP) on the thorax and abdomen measured respiratory depth and frequency, nasal airflow was measured with a thermistor and a nasal pressure sensor, and a pulse-oximeter for blood oxygen saturation.

Data Analysis

Sleep stage identification.

AASM guidelines were used to score the overnight PSG recordings in 30-s epochs and confirmed by a board-certified sleep physician (C. A. Smoot). Two 5- to 10-min periods of stable sleep stage (i.e., cycle 1 vs. cycle 2), free of cortical arousals, were selected for each participant in N2 (C1: 5 ± 0 min, C2: 5 ± 0 min) sleep, SWS (C1: 9.5 ± 2 min, C2: 9.8 ± 1 min), and REM (C1: 7.6 ± 3 min, C2: 7.0 ± 2 min) sleep. To address the hypothesis regarding disrupted sleep, two additional 5- to 10-min periods of sleep that contained at least one cortical arousal were found for each participant within N2 (C1: 5 ± 0 min, C2: 5 ± 0 min) sleep, SWS (C1: 5 ± 0 min, C2: 5 ± 0 min), and REM (C1: 5 ± 0 min, C2: 5 ± 0 min) sleep.

Heart rate variability.

Heart rate variability was derived from the overnight ECG and used as an index of cardiac autonomic activity in both frequency- and time-domains. Briefly, the ECG during the 8-h sleep opportunity was imported into a custom software (WinCPRS, Absolute Aliens, Turku, Finland) that marked each individual R-wave and was confirmed by the research team to determine the R-R interval (RRI). All ectopic beats were excluded from the analysis. A fast Fourier power spectral analysis was performed to quantify frequency-domain HRV from the 5–10 min of stable or disturbed sleep. Power spectra were obtained and expressed within the high-frequency (HF; 0.15–0.4 Hz) and the low-frequency (LF;0.04–0.15 Hz) range (32). Time-domain HRV included the root mean square of successive differences of R-R intervals (RMSSD) and the percentage of R-R intervals that varied by 50 ms or more (pNN50) (33).

Statistical Analysis

All data were analyzed using statistical software (SPSS 26.0; IBM SPSS, Armonk, NY). Z-score normality assumption tests were conducted on each variable to determine variable skewness, with a z-score greater than three standard deviations considered nonnormally distributed. Raw LF and HF HRV were nonnormally distributed and log₁₀ transformed. Pearson correlations were used to determine the reliability of frequency- and time-domain HRV during stable N2 sleep, SWS, and REM sleep. Bland–Altman plots were used to assess proportional bias between stable and disrupted sleep cycles, with stable sleep serving as the standard (34). A one-sample t test was run to determine whether mean differences between stable and disrupted HRV metrics significantly differed from zero (i.e., no difference between stable and disrupted sleep). If significant, proportional bias was evident. If nonsignificant, a Bland–Altman plot was constructed to show the mean HRV value between stable and disrupted sleep against the difference in stable and disrupted sleep HRV measure between each cycle of N2 sleep, SWS, and REM sleep. The plot included upper and lower limit bounds of the 95% confidence interval. Intraclass correlation (ICC) analyses were performed for each of the primary variables of LF, HF, pNN50, and RMSSD to determine the HRV reproducibility in stable and disturbed sleep. The ICC analyses were performed between each variable and for each sleep stage (SWS, N2, and REM) at four points: two arousal-free sections and two cortical arousal sections. The Cronbach’s alpha (α) value of the ICC was reported as the reproducibility statistic. Results are expressed as means ± SD or medians [interquartile ranges] with appropriate Cronbach’s α-coefficient. The null hypothesis was rejected (i.e., unrelated samples) when P < 0.05.

RESULTS

Table 1 depicts subject demographics. Participants were young and healthy, and both total sleep time (TST) and sleep efficiency determined via laboratory PSG were near the threshold for adequate sleep. Participants did not exhibit obstructive or central sleep apnea (apnea-hypopnea index of <5 episodes/min). The distribution of TST spent in N2, SWS, and REM were within normal distributions for healthy adults (35).

Table 1.

Participant demographics and sleep profile

Variable
Age, yr	26 ± 6
Sex, male/female	11/16
BMI, kg/m2	26.9 ± 4
TST, min	408 ± 68
SE, %	83 ± 10
N2, %	51 ± 9
SWS, %	22 ± 6
REM, %	19 ± 5
AHI, events/min	1 ± 2

Values are means ± SE. AHI, apnea hypopnea index; BMI, body mass index; N2, percentage spent in stage-2 sleep; REM, percentage spent in rapid-eye-movement sleep; SE, sleep efficiency; SWS, percentage spent in slow-wave sleep; TST, total sleep time.

Figure 1 depicts the correlations between HF HRV during stable cycles of N2 sleep, SWS, and REM sleep. The log₁₀ transformed HF HRV was highly correlated for all three sleep stages. Figure 2 depicts the reliability of log₁₀ transformed LF HRV across stable cycles of N2 sleep, SWS, and REM sleep, which was moderate-to-highly correlated for N2 sleep, SWS, and REM sleep.

Figure 1.

High-frequency (HF) component of heart rate variability (HRV) across two stable 5- to 10-min portions of stage-2 (N2) sleep (A, n = 26), slow-wave sleep (SWS; B, n = 25), and rapid-eye-movement (REM) sleep (C, n = 21) using a bivariate correlation. HF HRV was highly correlated between sleep cycles in each sleep stage (P < 0.001).

Figure 2.

Low-frequency (LF) component of heart rate variability (HRV) across two stable 5- to 10-min portions of stage-2 (N2) sleep (A, n = 26), slow-wave sleep (SWS; B, n = 25), and rapid-eye-movement (REM) sleep (C, n = 20) using a bivariate correlation. LF HRV was moderate-to-highly correlated between sleep cycles in each sleep stage (P < 0.001).

For the Bland–Altman plot analysis, HF HRV revealed no proportional bias between stable and disrupted sleep cycles in N2 sleep, SWS, and REM sleep. However, proportional bias was evident in cycles 1 and 2 comparisons in N2 sleep and cycle 1 comparison in REM sleep for LF HRV via a significant one-sample t test suggesting higher LF HRV in disrupted sleep periods. SWS cycles 1 and 2 comparisons, in addition to REM sleep cycle 2, demonstrated no proportional bias.

Table 2 depicts frequency- and time-domain HRV across cycles of both stable and disturbed sleep. HF HRV demonstrated excellent reliability in N2, SWS, and REM sleep. LF HRV exhibited excellent reliability in N2 and REM sleep, but not SWS sleep. Time-domain HRV depicted as pNN50 and RMSSD were both reliable in N2 sleep, SWS, and REM sleep. The number of cortical arousals in disrupted sleep periods was similar between cycles in N2 sleep (58 vs. 53 arousals), SWS (24 vs. 13 arousals), and REM sleep (41 vs. 37 arousals). Overall, parasympathetic HRV indices all exhibit excellent reliability in N2 sleep, SWS, and REM sleep. LF HRV demonstrated questionable reliability in SWS, but otherwise strong reliability in N2 and REM sleep.

Table 2.

Cronbach’s α for frequency and time-domain HRV

	Stable		Fragmented
	Cycle 1	Cycle 2	Cycle 1	Cycle 2	Cronbach’s α
N2
LF	3.1 [2.7–3.3]	3.2 [2.9–3.6]	3.4 [3.0–3.6]	3.5 [3.1–3.7]	0.90
HF	3.0 [2.5–3.3]	3.1 [2.7–3.4]	3.2 [2.5–3.4]	3.0 [2.6–3.5]	0.96
pNN50	44 [12–54]	56 [21–63]	50 [20–57]	47 [21–62]	0.93
RMSSD	46 [34–87]	72 [43–113]	41 [40–108]	58 [44–126]	0.94
SWS
LF	2.9 [2.3–3.2]	2.6 [2.5–3.2]	2.9 [2.7–3.3]	2.9 [2.5–3.1]	0.62*
HF	2.9 [2.6–3.4]	2.5 [2.4–3.3]	2.8 [2.7–3.2]	2.8 [2.4–3.0]	0.96
pNN50	44 [14–65]	22 [9–65]	34 [19–56]	43 [16–44]	0.96
RMSSD	71 [37–100]	42 [36–106]	70 [45–123]	55 [26–74]	0.96
REM
LF	3.3 [2.7–3.4]	3.3 [2.9–3.6]	3.3 [2.8–3.4]	3.4 [3.1–3.7]	0.91
HF	3.0 [2.0–3.4]	3.2 [2.5–3.5]	3.0 [2.5–3.2]	3.3 [2.6–3.5]	0.92
pNN50	47 [4–50]	42 [12–61]	28 [11–40]	33 [17–61]	0.95
RMSSD	51 [29–106]	43 [38–112]	39 [39–101]	59 [42–121]	0.96

Cronbach’s α analysis of reproducibility within stage-2 sleep (N2; n = 23), slow-wave sleep (SWS; n = 5), and rapid-eye-movement sleep (REM; n = 15). Results are medians [25–75th percentile]. LF, low frequency (log₁₀ transformed); HF, high frequency (log₁₀ transformed); HRV, heart rate variability; pNN50, proportion of R-R intervals that exceed 50 ms; RMSSD, root mean square of successive differences. All Cronbach’s α are P < 0.001 unless denoted with *.

DISCUSSION

The present study investigated the reliability of HRV during N2 sleep, SWS, and REM during both stable and disrupted sleep. Previous studies examining the reliability of frequency- and time-domain HRV during polysomnographic sleep have been inconsistent (18, 22, 23). Consistent with prior work, HF HRV indices, along with RMSSD, were reliable during stable SWS (18).In contrast to a study by Herzig et al. (18), we report high reliability of HF HRV across all recorded sleep stages, as well as moderate-to-high reliability of LF HRV across all recorded sleep stages. Importantly, a novel aspect of the present study examined sections of each sleep stage with cortical arousals, and these results also demonstrated strong intraindividual reliability when compared with the stable sections of sleep. Overall, both time- and frequency-domain HRV were reliable across both stable and disturbed sleep. Our findings support the use of HRV as a physiological measure during various sleep stages and provide evidence that mild cortical arousals do not substantially impact the HRV intraindividual reliability.

The LF component of HRV is often met with a degree of controversy within interventional study designs. First, wake recordings of LF HRV can exhibit poor within-session and within-participant reliability, as well as divergent responses to other direct measures of peripheral sympathetic activity (12, 14, 15, 36, 37), making differences in LF HRV across interventional studies challenging to interpret. Second, the sleep literature often directly implicates LF HRV as a noninvasive estimate of cardiac sympathetic activity (38), rather than an indirect estimate of cardiac sympathetic and parasympathetic activity (39). To date, only two studies have assessed the reliability of LF HRV during sleep (18, 22). Herzig et al. (18) in their study demonstrated poor between-night and within-night reliability of LF HRV in N2 sleep, SWS, and REM sleep in circadian-aligned sleep. Cellini et al. (22) in their study examined the reliability for short periods of sleep during short daytime sleep opportunities (i.e., naps) and determined LF HRV was not reproducible across three consecutive nap opportunities over a 1-mo period (ICC <0.6) in all sleep stages, but did not examine within nap reliability. Results from the present study suggest strong within-subject sleep cycle reliability of LF HRV in N2 sleep, SWS, and REM sleep, which is in contrast to a study by Herzig et al. (18) during sleep and other wake HRV recordings (22). However, the proportional bias was seen in the Bland–Altman plots for N2 sleep, and cycle 1 of REM sleep for LF HRV, supporting our initial hypothesis that LF HRV may not be comparable between stable and disrupted sleep. This finding also supports existing literature that LF HRV demonstrates poor intraindividual reliability (15–17). Interestingly, no proportional bias was found in LF HRV during SWS supporting a study by Brandenberger et al. (6) who suggested SWS may be the optimal period to assess LF HRV, but only for risk stratification purposes rather than indices of cardiac sympathetic activity (40–42).

Frequency- and time-domain estimates of cardiac parasympathetic activity (i.e., HF, RMSSD, and pNN50) are known to provide reliable estimates during wake (43, 44) and sleep (18, 23). Two studies have examined the day-to-day reliability of HF HRV during sleep (22, 23). Cellini et al. (22) in their study reported poor-to-moderate HF HRV reliability across NREM (i.e., N2 sleep and SWS) and REM sleep during the three afternoon nap opportunities separated by 2 wk. In contrast, Israel et al. (23) in their study demonstrated excellent reliability across three back-to-back circadian aligned sleep periods. This divergence may be partially explained by differing autonomic profiles between an afternoon nap period and a full 8-h sleep opportunity (45). The study by Herzig et al. (18) remains the only one to assess reliability within one sleep period, and reported that the HF component and RMSSD were reliable only in SWS, with deteriorations in reliability in N2 and REM sleep. Our data support the reliability of frequency- and time-domain cardiac parasympathetic estimates across repeated cycles of N2 sleep, SWS, and REM sleep. However, in recent methodological advancements, even HF HRV may not be a sufficient indicator of cardiac parasympathetic activity given new evidence suggesting a complete divergence and lack of correlation between direct vagal recording and several parasympathetic HRV estimates including the HF component and RMSSD (46). This highlights the need to be critical even to long-standing techniques and their physiological interpretation.

The current study also investigated the reliability of HRV indices during modestly fragmented sleep, in contrast with previous investigations that have only examined stable sleep stages (18, 22). Sleep is a highly heterogenous process that is frequently interrupted by nocturnal disruptions, even in healthy populations. Nocturnal cortical arousals are further associated with a consistent cardiovascular response including a transient increase in heart rate, blood pressure, and peripheral sympathetic activity (27, 28). Although the heart rate response appears to recover quickly after cessation of the arousal (47), sustained reductions in cardiac parasympathetic activity have been reported (48). In addition, these cardiovascular responses can be impacted by numerous factors including changes in arousal intensity (49, 50), limb movements (49, 50), short habitual sleep duration (27), among others. The variability in cardiovascular responsiveness to unique arousals may further impact HRV indices sampled during disrupted sleep periods.

Support for a modifying role of arousals in HRV measurement is further observed in a recent study of adults with primarily mild-to-moderate obstructive sleep apnea, where the authors noted that the arousal index alone was significantly related to nocturnal HRV metrics (51). The combination of these findings infers that the presence of nocturnal arousals might heavily influence the within-night reliability of nocturnal HRV indices, particularly when compared with stable sleep periods. Contrary to this hypothesis, our findings indicate that even during stages of modestly fragmented sleep, HRV metrics remained reproducible (Table 2). Within our dataset, we did not discriminate based on the type of arousal (respiratory related, movement associated, etc.). This increased the ecological validity of our findings, supporting the reliability of studies performing HRV analyses over the course of the entire sleep period in healthy adults (26).

Despite these findings of HRV reliability across both stable and disrupted sleep, it is important to note that our data set includes primarily modest sleep disruption. Each 5- to 10-min period of fragmented sleep used to assess reliability contained approximately one to two arousals which, if maintained, equates to ∼12–24 arousals/h. The present study was conducted in healthy adults without pathological sleep conditions; thus, we cannot say with certainty whether the observed HRV reliability during sleep extends into populations with sleep disorders such as OSA, restless leg syndrome, insomnia, etc. Interestingly, experimental manipulations of airflow and inhaled gases in healthy participants do not appear to impact cardiovascular reactivity to evoked arousals (52–56). Furthermore, although arousal intensity is related to subsequent cardiovascular responsiveness (50), respiratory events preceding cortical arousals in OSA do not appear to have an underlying role in arousal severity (57). These findings support our indiscriminate use of arousal type within the current study and the potential extension of our findings to some pathological conditions such as mild OSA. Nonetheless, apneic events occurring alone or in concert with cortical arousals are associated with elevated peripheral sympathetic activity, as well as cardiovascular and ventilatory instability (58–60), all of which can significantly impact HRV measurements and reliability. Although HRV has proven to be reproducible in primary insomnia (23), further research is necessary to extend these findings, as well as those in the current study to other sleep-disordered populations who exhibit high levels of sleep fragmentation and disruption.

Although the current study offers strong support for nocturnal HRV reliability during various sleep stages, we acknowledge the following limitations. First, analysis conducted during periods of SWS had a smaller sample size than the rest of the sleep stages when examining the arousals. This discrepancy can be explained by the overall lower frequency of arousals during SWS (61). Even with this slightly decreased sample size, HF HRV was still considered reliable during modestly fragmented SWS. Second, the number of arousals during each sleep stage was variable, as the basis for selection was more than one arousal in the analyzed timeframe. Future work assessing heavily fragmented sleep periods, as well as specific types of arousals, may offer further insight into the reliability of nocturnal HRV. Third, and as noted in the methods, the HF and LF HRV data were not normally distributed, requiring a log transformation of the data. However, previous HRV studies have also transformed the HRV frequency-domain data via natural logarithmic transformation (18, 22). Fourth, we did not control for menstrual phase in the 16 female participants. The female participants phase distribution was as follows: luteal (63%), ovulatory (25%), and follicular (12%) phases. The current study was also not powered for sex comparisons, but future work in this area within a consistent menstrual phase for female participants might be warranted. Finally, our study was performed in a young, healthy population without known sleep disorders. Therefore, our results are generalizable to a healthy population, but may have limited generalizability to sleep-disordered populations. Although one study has shown reproducibility between nights in individuals with chronic insomnia (23), further studies in populations with other sleep disorders would be necessary to confirm HRV reliability in specific populations.

In summary, HRV is a commonly used measure of nocturnal cardiac autonomic activity, despite limited studies examining its reproducibility across various sleep stages. Previous studies have demonstrated ambiguous results in terms of nocturnal HRV reliability (18, 22, 23) and have not examined the reliability of HRV during sleep with cortical arousals. Our findings advance this scientific gap, with results indicating that both frequency- and time-domain HRV demonstrate strong reproducibility within key sleep stages in healthy young adults, including during periods of modestly disrupted sleep. The present study supports the use of nocturnal HRV in healthy adults as a tool for cardiovascular health and risk stratification. Future work may be warranted in populations with frequent nocturnal arousals and cardiovascular/respiratory instability, such as obstructive sleep apnea, to assess whether these HRV reproducibility findings extend to sleep-disordered populations.

GRANTS

This project was funded by National Institutes of Health Grants R01AA024892, P20GM103474 (Montana INBRE), and U54GM115371 (American Indian-Alaska Native Clinical and Translational Research Center).

DISCLOSURES

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

Jason Carter is an associate editor of American Journal of Physiology-Heart and Circulatory Physiology and was not involved and did not have access to information regarding the peer-review process or final disposition of this article. An alternate editor oversaw the peer-review and decision-making process for this article.

AUTHOR CONTRIBUTIONS

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