Influence of Hypoxia and Hypercapnia on Sleep State-Dependent Heart Rate Variability Behavior in Newborn Lambs

Abstract

Study Objectives:

Although hypercapnia and/or hypoxia are frequently present during chronic lung disease of infancy and have also been implicated in sudden infant death syndrome (SIDS), their effect on cardiac autonomic regulation remains unclear. The authors' goal is to test that hypercapnia and hypoxia alter sleep-wake cycle-dependent heart rate variability (HRV) in the neonatal period.

Design:

Experimental study measuring HRV during sleep states in lambs randomly exposed to hypercapnia, hypoxia, or air.

Setting:

University center for perinatal research in ovines (Sherbrooke, Canada). INSERM-university research unit for signal processing (Rennes, France).

Participants:

Six nonsedated, full-term lambs.

Interventions:

Each lamb underwent polysomnographic recordings while in a chamber flowed with either air or 21% O₂ + 5% CO₂ (hypercapnia) or 10% O₂ + 0% CO₂ (hypoxia) on day 3, 4, and 5 of postnatal age.

Measurements and Results:

Hypercapnia increased the time spent in wakefulness and hypoxia the time spent in quiet sleep (QS). The state of alertness was the major determinant of HRV characterized with linear or nonlinear methods. Compared with QS, active sleep (AS) was associated with an overall increase in HRV magnitude and short-term self-similarity and a decrease in entropy of cardiac cycle length in air. This AS-related HRV pattern persisted in hypercapnia and was even more pronounced in hypoxia.

Conclusion:

Enhancement of AS-related sympathovagal coactivation in hypoxia, together with increased heart rate regularity, may be evidence that AS + hypoxia represent a particularly vulnerable state in early life. This should be kept in mind when deciding the optimal arterial oxygenation target in newborns and when investigating the potential involvement of hypoxia in SIDS pathogenesis.

Citation:

Beuchée A; Hernández AI; Duvareille C; Daniel D; Samson N; Pladys P; Praud JP. Influence of hypoxia and hypercapnia on sleep state-dependent heart rate variability behavior in newborn lambs. SLEEP 2012;35(11):1541-1549.

INTRODUCTION

Sustained hypoxic and/or hypercapnic episodes are a common occurrence in infants, especially in cases of chronic lung disease or in conditions suspected to be involved in sudden death syndrome (SIDS). In addition, the current standard of care in neonatal intensive care units tends to favor prolonged mild Hc and Hx in an attempt to increase the survival of neonates, as well as to decrease the frequency of chronic lung disease and retinopathy of prematurity.^1,2 However, although the autonomic nervous system (ANS) is influenced by both sleep-wake cycle and acute Hc or Hx,³ the effect of these gas conditions on cardiac autonomic adaptation to the sleep-wake cycle is far from being fully understood, especially in the newborn.

Although there are several published reports on the effects of sleep states on cardiac autonomic regulation, certain unresolved issues still persist, especially with regard to rapid eye movement (REM) sleep. In the normal adult human, the transition from wakefulness to nonrapid eye movement (NREM) sleep is associated with an increase in cardiac parasympathetic activity and/or a decrease in sympathetic activity.^3,4 On the other hand, an increase in sympathetic activity has generally been reported in REM sleep.⁵ Similarly, in the perinatal period, current knowledge generally ascribes a higher cardiac sympathetic nerve activity to active sleep (AS) than to quiet sleep (QS),^6,7 together with decreased cardiac parasympathetic activity.⁸ However, certain studies have conversely suggested the presence of a concomitant vagal activation in REM sleep compared with NREM sleep in both adult and perinatal life, the net effect of this cardiac sympathovagal coactivation on heart rate and heart rate variability (HRV) being dependent on the balance between the two efferent branches of the ANS.^5,9,10 Interestingly, several recent reports on ANS function contend that the traditional view of reciprocal actions of cardiac vagal and sympathetic outflows is hardly ever substantiated, with cardiac sympathovagal coactivation being present in most physiologic conditions.¹¹ However, in most of these studies, autonomic regulation has been explored with linear methods, neglecting the nonlinear effects imposed by the ANS on the cardiovascular system.

Both hypoxia (Hx) and hypercapnia (Hc) are known to induce complex, regionally and centrally mediated changes in heart rate, cardiac contractility, and peripheral vascular resistance. Numerous studies on the effects of Hx on HRV have been performed. Overall, although most studies in adult humans concluded that Hx decreases cardiac vagal activity and increases cardiac sympathetic activity,¹² others failed to find any changes in sympathetic¹³ or parasympathetic activity,¹⁴ and some even reported a global decrease in HRV.¹⁵ Results obtained on Hx in the perinatal period have not provided a clearer picture. During gestation, although results ranged from a global increase¹⁶ to a global decrease¹⁷ in HRV in human fetuses during Hx, studies in the fetal lamb reported either an increase in cardiac parasympathetic neural activity only¹⁸ or a sympathovagal coactivation.¹⁹ In the early postnatal period, available results on Hx are again in apparent contradiction. Indeed, although hypercapnic Hx in human newborns did not induce any significant changes in HRV,²⁰ it led to a global increase in HRV in anesthetized piglets.²¹ In contrast, isocapnic Hx increased cardiac sympathetic activity but decreased cardiac vagal activity in anesthetized piglets.²² Finally, hypocapnic Hx increased cardiac sympathetic activity with no change in cardiac parasympathetic activity in the emu.²³ Compared with Hx, very few studies on the effect of isolated Hc on HRV have been reported. Two studies in the human fetus showed either no change²⁴ or a minimal decrease in HRV²⁵ whereas in anesthetized paralyzed newborn piglets, Hc induced sympathovagal coactivation in the 30-day-old group only.²⁶

Responses to physiologic changes such as sleep-wake cycling²⁷ or to external challenges such as Hx or Hc²⁸ result in alterations in the magnitude and/or dynamic behavior of HRV, which is associated with maintenance of homeostasis. Numerous linear and nonlinear analyses methods of HRV time-series have been developed to better understand normal HRV as well as to characterize abnormal responses, which can threaten this homeostasis.^29–31 Linear analysis allows the characterization of HRV by the estimation of time- and frequency-domain indices, which in turn are used in an attempt to infer cardiac sympathetic and/or vagal autonomic regulation of sinus rhythm. Nonlinear analysis methods are complementary to linear analysis and provide detailed information regarding the complexity of HRV dynamics, related to multiple influences from and beyond control by the ANS.³² Although the concomitant use of various linear and nonlinear analyses is advocated as the best option for a complete characterization of HRV in a given situation, this is far from being the rule, as witnessed in recent studies on HRV. This observation is especially true for the newborn.

In the current study, we hypothesized that cardiac sympathovagal coactivation may be present during AS in healthy, nonsedated, newborn lambs and that this autonomic adaptation to the sleep-wake cycle could be altered by sustained Hx or Hc. Consequently, we first aimed at providing a full feature of the changes in HRV dynamics during sleep, using nonlinear analyses of HRV in addition to traditional linear methods. Second, we aimed at testing the effects of sustained Hx and Hc on HRV in these newborn lambs, in an attempt to infer sympathetic and vagal control of the sinus node in Hx and Hc in relation to sleep states.

MATERIAL AND METHODS

Animals

Six mixed-bred lambs, born at term by spontaneous vaginal delivery, were involved in the current study. The same lambs took part in another study to assess the effect of Hx and Hc on nonnutritive swallowing throughout sleep stages.³³ The protocol was approved by the Committee for Animal Care and Experimentation of the Université de Sherbrooke.

Surgical Preparation

Aseptic surgery was performed on the second day of life under general anesthesia (isoflurane 1-2% + NO₂ 30%, balance O₂) after an intramuscular injection of atropine sulfate (0.1mg/kg), ketamine (10 mg/kg), midazolam (100 μg/kg), and antibiotics (5 mg/kg gentamicin and 7,500 IU/kg duplocillin, which were administered daily thereafter until the end of the experiment). One dose of ketoprofen (3 mg/kg intramuscularly) was systematically given immediately after induction of anesthesia for analgesia and repeated if needed the next day. Chronic instrumentation was performed as previously described³³ and included two needle-electrodes into the parietal cortex for electrocorticogram (ECoG) and two subcutaneous needle electrodes into the forelegs for electrocardiogram (ECG). One needle-electrode was also inserted subcutaneously on the scalp to serve as a ground. Furthermore, an arterial catheter was introduced in the brachial artery for measuring blood gases. Correct electrode positioning was systematically verified at autopsy.

Recording Equipment

The instrumentation was completed immediately prior to the recording session. Two needle electrodes were inserted subcutaneously near the right eye for electrooculogram (EOG) recording. Nasal airflow was recorded using a type J thermocouple (response time 0.002 sec), and respiratory thoracoabdominal movements were monitored with respiratory inductance plethysmography. Inspiratory and expiratory fraction of oxygen and carbon dioxide were continuously monitored using a nasal cannula and a Capnomac II (Datex-Ohmeda, Mississauga, ON, Canada). Our custom-built radiotelemetry system³⁴ was used to continuously transmit signals of nasal flow, ECG, EOG, and ECoG. All signals were sampled at 1,000 Hz and recorded on a PC, using the MP100A data acquisition system and Acknowledge 3.7.3 software (Biopac Systems Inc., Goleta, CA, USA).

Design of the Study

Lambs were studied without sedation 48 hr after surgery (age: 4 ± 1 day, weight: 3.9 ± 0.5 kg), between 6:00 and 11:30. The lambs were housed with their mother between all experiments. All recordings began at least 40 min after their last feeding. Each lamb underwent three recordings on three subsequent days while in a Plexiglas chamber (volume 1.2 m³). The chamber was continuously flowed (12 L/min) with medical air or 21% O₂ + 5% CO₂ (Hc) or 10% O₂ + 0% CO₂ (Hx). Consistency of the gas mixture in the chamber was monitored throughout the recordings using both an oximeter and a capnometer (Sable Systems O₂ FC-1B and CO₂ CA-1B, Las Vegas, NV, USA). Each day corresponded to a different gas mixture, administered during 4 hr and in random order. The chamber temperature (24°C) and humidity (70%) were maintained constant throughout the recordings. Lambs were monitored throughout the three recordings, and an observer was always present in the laboratory to note all events.

Data Analysis

Sleep States

Standard electrophysiologic and behavioral criteria were used to define wakefulness (W), QS, and AS from ECoG and EOG tracings, together with careful, continuous observation of the lambs.³⁵ Briefly, in W, lambs were standing or lying prone, with the head up; in QS, they were lying prone, usually with the head placed backward on their shoulder; in AS, they were lying prone with the head and neck extended forward on the floor, while experiencing numerous twitches. Percentages of time spent in each state of alertness as well as mean duration and frequency of W, QS, and AS epochs were calculated. Respiration and heart rate were calculated for all epochs of W, QS, and AS recorded in all lambs, when at least 60 sec were spent in that epoch.

Signal Processing

Custom-built signal processing tools designed with the Matlab software (The Mathworks, Inc., Meudon, France) at the Laboratoire de traitement du signal et de l'image (INSERM U1099, Université de Rennes 1) were used.

Editing of Electrocardiographic Recordings:

A rigorous, multistep editing process was applied to all ECG recordings. In a first step, epochs with major reactions such as arousal with gross body movements, bleating, or coughing were excluded from the analysis. Second, on all selected periods, QRS complexes were automatically detected from the ECG signal by applying a modified version of the Pan and Tompkins detector,³⁶ using specific filter coefficients adapted to the lamb's QRS frequency content. The interval between successive peaks of the QRS complex on the electrocardiogram (RR) series created from this automatic detection was manually verified and corrected by an experienced operator. Third, to avoid artifacts on RR series related to abnormal QRS morphologies or transient (less than 2 beats) signal saturation, a linear interpolation was applied between adjacent normal RR intervals, and the obtained RR series were upsampled to 10 Hz. Finally, to analyze the best 2-min segment in terms of stationarity and lack of artifacts, an automatic selection was performed on the available RR series³⁷ for each state and for each condition. This also allowed rejecting segments with minor albeit significant changes such as those occurring with quiet arousal, opening of eyes, and swallowing. The selection was performed by (1) dividing the entire available RR series in subsegments of 10 sec; (2) calculating the mean value on each subsegment; and (3) calculating the standard deviation (SD) of these mean values for all sets of 12 consecutive subsegments. The set of 12 consecutive subsegments (2 min) presenting the lowest SD was selected for further analysis.

Analysis Methods of Heart Rate Variability:

Results from both linear (time-domain and frequency domain) and nonlinear (Poincaré plot, entropy, and fractal organization) analyses were compared to investigate the effect of sleep state on HRV and assess the influence of Hc and Hx on that particular sleep state effect.

Linear Analyses of HRV:

Time domain analysis of HRV provides measures of the overall magnitude of RR interval fluctuations around its mean value. In the current study, it consisted of extraction of the mean cardiac cycle length and SD, which is an estimate of the magnitude of global HRV, and the square root of the mean squared differences of successive RR intervals (rMSSD), which reflects parasympathetic control of sinus rhythm. Results were expressed in absolute value (ms).

Frequency domain analysis provides measures of the magnitude of RR fluctuations in certain predetermined frequencies. It was performed by an autoregressive estimation of the power spectrum, and integration on the low-frequency (LF, 0.02-0.25 Hz) and high-frequency (HF, 0.25–2 Hz) spectral bands, yielding values expressed in ms². These frequency bands have been adapted to the newborn lamb.³⁸ Although variations in HF power mainly reflect variations in cardiac vagal control, LF power reflects both vagal and sympathetic control of the sinus node, with LF/HF ratio being an index of cardiac sympathovagal balance (e.g., increased LF/HF reflects more sympathetic and/or less vagal cardiac control). Very low frequency variations (0-0.02 Hz) were not considered in those short RR series, to avoid the possibility of artifacts due to slow long-term oscillations.

Nonlinear Analyses of HRV:

As stated earlier, nonlinear analysis methods are complementary to linear analyses and provide information regarding the complexity of HRV dynamics.³² Poincaré plot analysis was performed to further assess short-term and long-term HRV on the selected RR series. Poincaré plots are scatter graphs where each point represents a given RR interval, for example, interval RR_i+1, plotted against the preceding RR interval RR_i(Figure 1, second line). They can be analyzed quantitatively by measuring the SD of the data along two orthogonal axes, one of which corresponds to the identity line of the Poincaré plot. Parameter SD1 measures the SD of the points projected onto the line orthogonal to the identity line, and SD2 is obtained as the SD of the points projected onto the identity line. SD1 reflects short-term HRV whereas SD2 is interpreted as a measure of both short-term and long-term variability. SD1, SD2, and the Poincaré plot area (S) correlate significantly with the time frequency domain of HRV. In particular, SD1 strongly correlates with rMSSD, whereas SD2 and S strongly correlate with SD.³⁹

Figure 1.

Examples of 2-minute RR series and their nonlinear analysis in one lamb. The RR series were from the best 2-min stationary periods obtained in each combination of gas condition and wakefulness in the same lamb. As shown in these examples, short- and long-term heart rate variability observed on the tachograms (first line) and the Poincaré plots (second line) were greater in AS and W than in QS, irrespective of gas condition. Similarly, the scale invariance coefficients, estimated through log-log spectral analysis with frequencies encompassing 0.02 to 2 Hz (third line), and detrended fluctuation analysis with n encompassing 4 to 45 beats (fourth line), were greater in AS and W than in QS. AS, active sleep; HC, hypercapnia; HX, hypoxia; PSD, power spectral density (total heart rate variability power); QS, quiet sleep; SpEn, sample entropy; W, wakefulness.

Sample entropy (SampEn) was calculated to assess the irregularity of the selected RR series. SampEn is a modification of the approximate entropy,⁴⁰ shown to be appropriate for studying biologic signals in most instances.^41,42 However, SampEn is less dependent on data length and presents more consistency in many cases.⁴³ Given groups of n points in a series, the sample entropy is related to the probability that two sequences, which are similar for N points, remain similar at the next point. Higher values of entropy reflect lesser regularity and predictable sequences, in relation to the multiple influences normally affecting RR from and beyond the ANS.

Finally, detrended fluctuation analysis (DFA) was used to measure the fractal characteristics of RR fluctuations normally present in healthy subjects. A fractal system is characterized by scale invariance (self-similarity), i.e., the same features repeat themselves on different measurement scales. A fractal organization is flexible, and breakdown of this scale invariance may lead to a more rigid and less adaptable system with either random or too highly correlated behavior of heart rate dynamics.³² To find the most appropriate time scales to apply for the calculation of the DFA coefficients in newborn lambs, the mean DFA curve and its SD were analyzed for all available stationary segments. We identified a common inflection point at 45 heartbeats. This finding is consistent with values reported in the literature for preterm infants^44,45 but much higher than that found in adults.^46,47 Because we chose to perform analysis on stationary segments only, the maximal time scale was 2 min and the mean length of the RR series was 397 ± 95 beats. This limited length did not permit a reliable calculation of long-term scale invariance, i.e., for n > 45 heartbeats, due to the statistical error related to finite sample size.^46,47 Therefore, we measured the scale invariance exponent α_4–45 for n ranging from 4 to 45 heartbeats, corresponding to the classic short-term scale invariance α1 exponent. This time scale corresponds to intermediate and low frequencies in the power spectrum; thus, α_4–45 quantifies the fractal temporal organization of heartbeat fluctuations generated by the sympathetic, but also vagal, components of ANS.

Statistical Analysis

Quantitative variables were expressed as mean ± SD. Normality was first tested using Q-Q plots and the Shapiro-Wilk test. Statistical analyses were performed on raw data for all variables using a general linear model two-way analysis of variance for repeated measures with gas mixture and state of alertness as the independent variables. PROC MIXED of SAS software (version 9.2 TS level 2M3, Cary, NC) was used. Differences were deemed significant if P < 0.05.

RESULTS

Total duration of polysomnographic recordings in the six lambs was 24.4 hr in air (mean ± SD per lamb: 4.1 ± 0.2 hr), 24.6 hr in Hc (4.1 ± 0.5 hr), and 25.1 hr in Hx (4.2 ± 0.6 hr). Arterial blood gases were maintained constant throughout the recordings under each gas condition, as shown by results obtained in the first and last hr of the recordings: in air, PaO₂ = 86 ± 21 and 82 ± 15 mm Hg respectively, PaCO₂ = 38 ± 0 and 38 ± 2 mm Hg, pH = 7.47 ± 0.01 and 7.47 ± 0.002; in Hc, PaO₂ = 105 ± 1 and 99 ± 1 mm Hg, PaCO₂ = 44 ± 0 and 43 ± 3 mm Hg, pH = 7.40 ± 0.01 and 7.44 ± 0.04; in Hx, PaO₂ = 26 ± 4 and 26 ± 4 mm Hg, PaCO₂ = 25 ± 0 and 28 ± 1 mm Hg, pH = 7.57 ± 0.01 and 7.55 ± 0.00.

Sleep Architecture and Respiratory Parameters

Active sleep was obtained in all lambs under all three gas conditions. Neither Hc nor Hx had a significant effect on time spent in AS, expressed as a percentage of the total recording time (%AS). Although Hc decreased %QS and increased %W, Hx increased %QS and had no significant effect on %W. Regardless of the sleep state, respiratory rate was increased in both Hc and Hx, with no difference being observed between Hc and Hx (Table 1).

Table 1.

Time spent in each state of alertness and respiratory rate during baseline, air breathing, and in hypercapnic and hypoxic conditions

Heart Rate Variability

The major determinant of HRV was the state of alertness (Table 2). Overall, mean RR interval duration and amplitude of variability measured on stationary periods were increased in AS compared with QS, irrespective of the gas condition. Moreover, a significant interaction was observed between gas condition and state of alertness for global (SD) and low frequency indices (LF, SD2, and area of Poincaré plot) of HRV.

Table 2.

Heart rate variability values (time domain, frequency domain, and nonlinear analysis)

Influence of Sleep State on HRV During Air Breathing

When compared with W, QS was marked by a decrease in the magnitude of global (SD), short-term (rMSSD and HF), and long-term variability (LF) of HRV, together with a decreased LF:HF ratio, reflecting an overall decrease in the control of sinus rhythm by the ANS, and especially its sympathetic component. This was accompanied by an increase in SD2, Poincaré plot area, and SampEn, reflecting an increase in complexity, which cannot be measured with linear methods.

On the contrary, when compared with QS, AS was marked by an increase in total (SD, area of Poincaré plot) and long-term (LF, SD2) HRV indices, suggesting a sympathetic activation from QS to AS. In addition, an increase in short-term (rMSSD, HF, SD1) HRV indices was observed, suggesting cardiac sympathovagal coactivation from QS to AS. This is coherent with the higher values of α_4–45 in AS in comparison with QS reflecting higher fractal organization of sinus rhythm fluctuations in relation with sympathovagal regulation. Entropy was also significantly lower in AS, indicating a greater regularity and predictability in the autonomic control of heart rate.

Even if the differences between AS and W were less consistent, albeit significant, global (SD and area of Poincaré plot) and LF index (LF and SD2 of Poincaré plot) values were significantly higher in AS when compared with W state, suggesting a sympathetic activation from W to AS. Similarly, all of these parameters aside from Poincaré plot indices were significantly higher during W than QS in air.

Influence of Hypoxic and Hypercapnic Breathing on Heart Rate Variability During W and Sleep

The effects of Hx and Hc on HRV in relation to sleep states are reported in Table 2. The main differences observed between AS and QS in air, especially cardiac sympathovagal coactivation associated with a loss of complexity in AS, persisted in Hx and Hc. In Hx, this AS-related coactivation was even more pronounced than in air, as shown by a significantly greater increase in SD, LF, SD2, and area of the Poincaré plot in comparison with QS and W. The consistent and global AS-related increase in variability amplitude in all gas conditions is illustrated in Figure 1 (top two lines).

DISCUSSION

The current findings add new knowledge on the changes in HRV profile occurring during sleep in newborn lambs, which can be summarized as follows. First, compared with QS, AS is associated with a cardiac sympathovagal coactivation and a loss of heart rate complexity. This new finding for the neonatal period strengthens certain previous reports of sleep in adults and the general view that sympathovagal coactivation is frequent in normal life. Second, we confirm a small number of prior reports in adults that the passage from QS to AS is associated with increased fractal organization (self-similarity) of sinus rhythm fluctuations. Third, we show for the first time that AS-related sympathovagal coactivation is still present during Hc and even further enhanced in Hx, two frequent clinical conditions in infants. Overall, our findings bring new insights on the adaptation of the ANS to the various sleep stages in air as well as in Hx and Hc in the early neonatal period.

Active Sleep: A State of Cardiac Sympathovagal Coactivation

In the controlled conditions of the current HRV analyses, our results are highly suggestive of cardiac sympathovagal coactivation in REM sleep and extend identical findings from a few previous studies in adults to the newborn.^5,9,10 Cardiac sympathovagal coactivation is a common behavior of ANS response to stressful challenges as shown by the many conditions with which it has been previously associated, including the arterial chemoreflex, the diving reflex, the nociceptor reflex, the oculocardiac reflex,¹¹ and the laryngeal chemoreflexes.³⁸ In such situations, the increase in cardiac sympathetic activity is not associated with the reciprocal withdrawal of cardiac vagal outflow but rather with its persistence or even its simultaneous increase. Allegedly, the benefit of cardiac sympathovagal coactivation is to increase cardiac output by increasing stroke volume as well as increase cardiac sympathetic activation by its inotropic effect and cardiac vagal activation by limiting the chronotropic effect of cardiac sympathetic activation, thus increasing the time available for ventricular filling.¹¹ The increased cardiac output brought about by cardiac sympathovagal coactivation may be beneficial in REM sleep, because a high level of cortical activity is associated with an increase in cerebral blood flow, likely influenced by heightened cerebral metabolic oxygen consumption.⁴⁸ This adaptation may be even more crucial in the newborn, in whom a high cerebral metabolic rate in REM sleep is observed more consistently than in adults⁴⁹ and in whom limited autoregulation of the cerebral circulation might lead to cerebral vascular compromise in case of systemic arterial hypotension.⁴⁹ Recent results, however, further underlined the high complexity of autonomic control in REM sleep, where a specific increase in the sympathetic contingent to cerebral vessels has also been observed, likely as a means to finely modulate cerebral blood flow and maintain such flow in a both metabolically appropriate and nondangerous range.⁵⁰

Alterations in heart rate nonlinear dynamics have also been reported during the sleep-wake cycle in the human fetus,^51,52 infant,⁴² and adult.²⁷ Our current finding of more regular cardiac cycle length dynamics (as measured by entropy) and greater short-term self-similarity (as measured by α_4–45) in AS may be explained by the cardiac sympathovagal coactivation, albeit with a different behavior compared with that observed in adults.⁵³ Of note, although a sustained regularity was present during AS, irrespective of the gas condition, the greater short-term scale invariance was only observed during air breathing. Overall, AS appears to be associated with a simplification of cardiac control mechanisms that could lead to an impaired ability of the cardiovascular system to react to cardiovascular adverse events.⁵⁴

Effects of Hc and Hx on Sleep Organization and Heart Rate Variability

The interactions between sustained and moderate Hc or Hx, sleep architecture, and heart rate variability have been sparsely described.

Sleep Architecture

In the current study, sleep organization was altered by both Hx and Hc, which is in accordance with previous studies in the perinatal period. Hx was associated with an increase in the time spent in QS, with no changes in W or AS duration, as previously observed in hypoxic kittens⁵⁵ and in agreement with the theory of an enhanced hypoxic hypometabolism and inactivity as a protective response and energy conservation in the neonatal period.⁵⁶ Hc, on the contrary, was associated with a decrease in the time spent in QS and an increase in W. This is consistent with the known stimulating effect of Hc on several chemosensitive neuronal populations (e.g., the serotonergic medullary raphe neurons) known for enhancing W.⁵⁷

Heart Rate Variability

As reported in the introduction, several studies have previously assessed the effect of Hx on HRV magnitude using traditional linear analyses in adults as well as during the perinatal period with apparent conflicting results. Our results on an enhanced cardiac sympathovagal coactivation in AS are in agreement with studies in fetal lambs reporting cardiac sympathovagal coactivation during isolated Hx¹⁹ or during asphyxia,⁵⁸ as well as with the study on hypercapnic Hx in anesthetized piglets.²¹ Of note, none of the previously mentioned perinatal studies considered the potential effects of sleep state. To our knowledge, the only study to assess the effect of both sleep state and Hx on HRV was performed in 7- to 29-mo-old infants with severe bronchopulmonary dysplasia; mild Hx was reported to induce cardiac sympathetic activation in NREM sleep stage 2 but decreased inhibition of the cardiac vagal component in REM sleep.⁵⁹

Results from the current study performed in healthy, full-term newborn lambs indicate that the AS-related cardiac sympathovagal coactivation observed in air is enhanced during sustained Hx. To our knowledge, this additive effect of AS and Hx on cardiac sympathovagal coactivation has not been reported previously. Again, it is likely aimed at increasing cardiac output, as reported with Hx in full-term, newborn infants and lambs.^60,61 Of note, the current results obtained in healthy, full-term lambs may not be applicable in preterm lambs at birth. Indeed, we have previously shown that the increase in heart rate observed in Hx from the very first days of life in full-term lambs is delayed for at least the first postnatal wk in preterm lambs.⁶² Finally, the presence of Hc may have enhanced the cardiac sympathetic activation secondary to Hx. Indeed, one study in normal adults reported that hypocapnic Hx led to a greater sympathetic efferent activity to muscle vessels than isocapnic Hx, due to inhibition of sympathetic activation by both the afferent activity from lung stretch receptors and the arterial baroreflex in the latter.⁶³ Whether this holds true in newborn lambs and for cardiac sympathetic activation is unknown.

Results from the few previous studies on nonlinear dynamics of HRV during Hx have provided contradictory results, which is likely at least partly related to the various experimental conditions. In human adults sleeping at high altitude after a progressive ascent, short-term fractal behavior was either unchanged⁶⁴ or decreased (short-term correlation).⁶⁵ Acute exposure to Hx during experiments with simulated high altitude did not alter fractal properties,^15,66 whereas complexity of HRV either decreased⁶⁶ or increased.¹⁵ The only neonatal study performed on anesthetized piglets acutely exposed to isocapnic Hx revealed an increased complexity in the first min of Hx only.²⁶ In the current study, the AS-related loss of complexity in air was also observed in Hx.

The current observation of minimal changes in HRV during Hc compared with air breathing is in agreement with certain previous studies performed during the perinatal period. In the near-term human fetus, maternal Hc was found to be associated with either no change²⁴ or a minimal decrease in HRV.²⁵ Similar results were observed in newborn piglets²⁶ and in newborn infants after induction of hypoxic Hc.²⁰ However, other results suggest some effects of Hc on autonomic cardiovascular control. A decrease in regional sympathetic activity to cerebral vessels has been observed in 21-day-old lambs after acute (60-sec) and moderate (50 mm Hg) Hc, leading to an increase in cerebral blood flow, which was especially manifest during REM sleep.⁶ In addition, Hc has been associated with HRV changes, suggesting cardiac sympathovagal coactivation in piglets beyond the neonatal period.²⁶ Overall, it appears that mild Hc in the neonatal period is associated with minimal effects on cardiac ANS control and that significant changes may be observed beyond the neonatal period,^6,26 or within the context of global neonatal asphyxia whereas severe Hc is associated with Hx and acidosis.²¹ To our knowledge, there is no data on the effect of Hc on complexity or fractal organization of HRV.

Potential Limitations of our Study

Previous authors have underlined the difficulty in inferring the physiologic counterparts of HRV analysis with regard to cardiac sympathetic and parasympathetic activity, given the multiple influences participating in alterations of basal sinus node rhythm.^67,68 Moreover, the underlying physiologic mechanisms of the newer indices such as entropy and fractal scaling properties are not straightforward.^{41,46,47,53,69} This may be even more critical in the neonatal period, for which there are fewer studies aimed at linking HRV parameters and determinants of sinus rhythm.^39,70–72 In the current study, however, the use of multiple linear and nonlinear HRV indices allowed description of consistent changes in HRV behaviors, which enabled reasonable speculation as to the evolution of the autonomic control of heart rate with sleep states during Hx and Hc in the neonatal period.^39,73 In addition, because linear and nonlinear HRV indices have distinct theoretical background and significance, they contribute to complementary information regarding signal characteristics allowing precise and fine description of the changes occurring in autonomic control of heart rate.^30,73 Moreover, the current study was carried out on very carefully selected stationary RR series.

The animals were studied 48 hr after instrumental surgery, and pain could have theoretically affected the results. However, within the last 20 yr, we have refined our surgical and anesthetic procedures and postoperative care to reduce pain to a minimum. Consequently, we did not find any increase in sympathetic activity beyond 48 hr post-surgery in a previous, longitudinal study performed after identical surgical procedures (unpublished results). In addition, the radiotelemetry system that we routinely use enables the recording of nonsedated and freely moving lambs and greatly minimizes stress.³⁴

Clinical Implications

Attempting to infer potential clinical consequences of the sympathovagal coactivation observed during AS in healthy newborn lambs, as well as its further enhancement during Hx, must be made with caution.⁷⁴

The preterm newborn with chronic lung disease in the newborn intensive care unit can be exposed to moderate Hx and/or permissive Hc during weeks, even months. Clearly, our short 4-hr-long exposure in full-term lambs does not mimic such a prolonged clinical condition. However, the possibility that Hx enhances AS-related sympathovagal co-activation of the sinus node should be taken into account when deciding to target an oxygenation level in these infants. Conversely, our results suggest that HRV is not a determinant factor when considering permissive Hc.

Although it must be highlighted that lambs in the current study were only a few days of age, the observation that Hx further enhances AS-related sympathovagal coactivation may be relevant to SIDS pathogenesis. Indeed, AS is associated with high oxygen extraction and low cerebral venous oxygenation that worsen up to the age of 2 to 3 mo in prone sleeping, when risk of SIDS is at its greatest.⁷⁵ In this context, the enhancement of sympathovagal coactivation associated with low cerebral oxygen delivery may lead to a vulnerable state for the sleeping infant in AS. In this regard, our observation of an increased regularity of cardiac cycle length in AS, which is usually interpreted as a loss of physiological adaptability of the ANS to its environment,⁴² is noteworthy.

DISCLOSURE STATEMENT

This was not an industry supported study. The authors have indicated no financial conflicts of interest.

ABBREVIATIONS

ANS

autonomic nervous system

AS

active sleep

DFA

detrended fluctuation analysis

ECG

electrocardiogram

ECoG

electrocorticogram

EOG

electrooculogram

Hc

hypercapnia

HF

high-frequency component of heart rate variability

HRV

heart rate variability

Hx

hypoxia

LF

low-frequency component of heart rate variability

QS

quiet sleep

REM

rapid eye movement

rMSSD

square root of the mean squared differences of successive RR intervals

RR

interval between successive peaks of the QRS complex on the electrocardiogram

SampEn

sample entropy

SD

standard deviation

W

wakefulness