Heart rate variability is depressed in the early transitional period for newborns with complex congenital heart disease

Abstract

Purpose:

To compare early changes in autonomic nervous system (ANS) tone between newborns with complex congenital heart disease (CHD) and newborns without CHD.

Methods:

We performed a case-control study of heart rate variability (HRV) in newborns with complex CHD (transposition of the great arteries [TGA] or hypoplastic left heart syndrome [HLHS]) and low-risk control newborns without CHD. Cases with CHD were admitted following birth to a pediatric cardiac intensive care unit and had archived continuous ECG data. Control infants were prospectively enrolled at birth. ECG data in cases and controls were analyzed for HRV in the time- and frequency-domains at 24 hours of age. We analyzed HRV metrics of alpha short (α_s), alpha long (α_L), root mean square short and long (RMS_s and RMS_L), low-frequency (LF) power, normalized LF (nLF), high-frequency (HF) power, and normalized HF (nHF). We used ANOVA to compare HRV metrics between groups and to control for medication exposures.

Results:

HRV data from 57 infants with CHD (TGA, n = 33 and HLHS, n = 24), and from 29 controls were analyzed. HRV metrics of α_s, RMS_L, LF, and nLF in infants with CHD were significantly lower from those of controls. Due to effect of normalization, nHF was higher in CHD infants (P<.0001) although absolute HF was lower (P=.0461). After adjusting for medications, α_s and nLF, remained lower and nHF higher in newborns with CHD (P<.0005).

Conclusions:

Infants with complex CHD have depressed autonomic balance in the early postnatal period which may complicate fetal-neonatal transition.

Introduction

The autonomic nervous system (ANS) plays an essential role in the coordinated control of the cardiovascular and respiratory systems to establish homeostasis during fetal to neonatal transition. Newborns with complex congenital heart disease (CHD) are at particular risk for hemodynamic instability in the period immediately after birth due to the anatomic cardiovascular anomaly. An additional risk for poor transition in newborns with complex CHD is impaired function of the ANS. Since fetal conditions associated with reduced oxygen and nutrient supply (e.g. complex CHD) may impair maturation of the ANS, the ANS may be unprepared for optimal transition in newborns with complex CHD [1]. Poor cardiorespiratory transition in infants with complex CHD can be seen clinically as a higher risk for hypotension after birth and greater respiratory support needs. Furthermore, the autonomic balance in the newborn has not yet established that of the mature ANS since vagal tone continues to increase through infancy [2].

Hypoplastic left heart syndrome (HLHS) and dextro-transposition of the great arteries (TGA) are two forms of complex CHD that may result in significant hemodynamic instability in the period after birth because of the abnormal circulatory anatomy; during this critical period of adaptation, ANS balance is likely of major importance. Both of these types of CHD are diagnosed in the fetal period or in the immediate postnatal period and require surgery within a few days-to-weeks after birth. In HLHS, the left side of the heart is small and may provide inadequate systemic and cerebral perfusion; central mixing of oxygenated and deoxygenated blood results in hypoxemic cerebral perfusion. Similarly, in TGA the aorta and pulmonary arteries are transposed with the more deoxygenated stream ascending through the aorta into the cerebral circulation. In this manner these cardiac anomalies may predispose to both cerebral hypoxemia and hypoperfusion during fetal brain development, with potential dysmaturation of the central ANS centers a consequence. In the period of transition after birth there is additional risk for ischemic brain injury [3].

ANS function can be measured non-invasively in the infant from the bedside electrocardiogram (ECG) by a determination of heart rate variability (HRV). HRV is the fluctuation in the length of time between successive heart beats (R-R intervals) and provides a measure of sympathetic and parasympathetic function, and therefore ANS balance [4, 5]. In the fetus and premature infant, maturation of the ANS toward term gestational age is normally associated with an increase in parasympathetic tone [6]. However, fetuses with HLHS show reduced HRV compared to healthy control fetuses at a similar gestational age, indicating a delayed maturation of the ANS even prior to birth [7]. However, the ANS balance in infants with complex CHD during fetal-to-neonatal transition immediately following birth has not been well established. The objective of this study was to compare changes in autonomic tone, as measured by HRV time and frequency domain metrics, between infants with complex CHD (TGA or HLHS) versus low-risk term control infants. We hypothesize that infants with complex CHD have lower metrics of HRV (depressed autonomic balance) in the early postnatal preoperative period compared to controls.

Methods

Participants

We retrospectively studied infants diagnosed with complex CHD types TGA or HLHS by echocardiography who underwent surgical repair between August 2012 to April 2016 at Children’s National Health System (Children’s National henceforth) in Washington, DC, and who had available continuous ECG data archived from the cardiac intensive care unit (CICU). Infants with structural brain abnormalities, chromosomal abnormalities, or birth gestational age <36 weeks were excluded. The study received a waiver of parental consent and Institutional Review Board (IRB) approval at Children’s National Health System.

As part of a prospective study at the Inova Fairfax Hospital, Fairfax, VA, term newborn control infants were prospectively enrolled following informed parental consent and IRB approval at Inova Fairfax Hospital, Fairfax, VA and Children’s National Health System, Washington, DC between May 2017 and July 2017. Eligible control infants were from uncomplicated pregnancies and deliveries, without significant maternal illness, uncomplicated labor and delivery, ≥37 weeks gestational age at birth, with a normal birth weight (10^th - 90^th percentile for gestational age), without postnatal infection, and without CHD. Infants did not require intensive care after birth.

Data Collection

The medical history of the infants with CHD (TGA or HLHS) was extracted from the clinical database and included demographic and birth clinical characteristics, need for mechanical ventilator support, and medication exposure. Medications were grouped into four broad classes: sedatives (fentanyl, morphine, midazolam), medications for hypotension (dopamine, epinephrine), neuromuscular blocking agent (vecuronium, rocuronium), and prostaglandin E1 (PGE1). For infants with CHD, continuous ECG recordings from admission to 72 hours of age were retrieved from an institutional Research Data Export archive (IntelliVue Information Center, Philips Healthcare, Andover, MA) at a sample rate of 125 Hz. For the controls, we used ECG data collected as part of a combined ECG and high-density electroencephalography study using a Geodesic EEG System 300 (Electrical Geodesics, Inc. Eugene, OR) at a sample rate of 1000 Hz. Clinical and demographic data were collected from study-site medical records. The ECG recording datasets from the controls were approximately 60 minutes long and collected at ≤48 hours of age. For the CHD group, we selected 60 minutes of ECG signals between 23 – 24 hours of age or the first available one-hour period.

EKG Processing and HRV Analysis

ECG was low-pass filtered with a filter cutoff of 70 Hz using a Butterworth filter with zero phase distortion. The QRS complex in every cardiac cycle was identified using a recently proposed approach [8] and the beat-to-beat (RRi) interval was calculated. The erroneous (missed/extra) beats were repaired using an established approach [9]. The RRi data were partitioned into 10-minute epochs with no overlap. The time domain metrics were characterized using the detrended fluctuation analysis (DFA) approach [10] and the frequency domain metrics were characterized using power spectral analysis [11]. For power spectral analysis, the RRis were interpolated using a cubic spline interpolation technique.

To ensure that the difference in ECG sampling rate between CHD infants and controls would not impact results, we systematically downsampled the ECG to 500 Hz, 250 Hz, 125 Hz and 64 Hz and calculated RRi and HRV for each sampling frequency. We compared the HRV metrics obtained for each sampling frequency against those obtained for ECG sampled at 1 KHz. We found no difference between HRV metrics calculated using 1 KHz sampled ECG and HRV metrics calculated fromECGs sampled at lower sampling frequencies until 125 Hz [12]. Thus, the ECG sampled at 125 Hz yielded consistent results as obtained from the ECG sampled at a higher rate.

Detrended Fluctuation Analysis – Time Domain Characterization

HRV can be measured by time-domain analysis which evaluates short- and long-term variability in the HR. Long-term variability is influenced by both the sympathetic and parasympathetic nervous systems, meanwhile short-term variability is influenced more by the parasympathetic nervous system based on rapid ANS needs [13]. Detrended fluctuation analysis (DFA) is a modified root mean square (RMS) analysis approach and is based on the principles of statistical physics designed to counter non-stationarity in RRi [10, 14]. Since accelerations and decelerations in infant heart rate can be caused by infant movement, the heart rate is non-stationary and DFA enables a way to account for this type of data. This approach is now widely used [15]. Briefly, this method involves (i) calculating profile function as the cumulative sum of the RRi; (ii) partitioning the profile function into non-overlapping windows containing s-number of beats; (iii) fitting the profile inside each window using a polynomial function; (iv) calculating local fluctuation function in each window as the root mean square of deviation of profile from the best fit; and (v) calculating global fluctuation function as the median of the local fluctuation functions over all windows. Steps ii – v are repeated for each 10-minute epoch. In step (iii) we used a fourth-order polynomial function (DFA4), which required the smallest window number to be six beats with the maximum window number being one fourth the total number of beats in a 10-minute epoch. We derived the following four metrics using the fluctuation functions: α_S and α_L as the slopes of the fluctuation function and window size ‘S’ using 15 – 30 beats and window size ‘L’ 35 – one fourth of the total number of samples, respectively (Figure). Similarly, the RMS fluctuation was obtained as maximum of fluctuation function in ‘S’ 15 – 50 beats and ‘L’ 100 – 150 beats, respectively. The α metrics quantify auto-correlation in the RRi whereas the RMS metrics quantify variability in the RRi. The α metrics are dimensionless quantities. The RMS metrics are in units of seconds (sec).The subscripts ‘S’ and ‘L’ denote short and long term quantities (for details, see reference [10]).

Fig: Sample Detrended Fluctuation Analysis (DFA) Plot of Infant with CHD and a Control Infant.

Shown are DFA-fourth order polynomial (DFA4) fluctuation functions obtained for 10 minutes of RRi of an infant in control group as well as of an infant in the CHD group. The fluctuation functions are plotted as a function of the scale s in a log-log representation. DFA4 indicates a fourth order polynomial was used in the detrending process (refer to text for details). The short-term exponent (\alpha_S) is calculated from the scale region shown in the black rectangle. The long-term exponent (\alpha_L) is calculated from the scale region shown in cyanide rectangle.

Spectral Analysis

HRV can also be measured by the frequency domain. High-frequency (HF) variability reflects parasympathetic function and is influenced by the respiratory rate, while low-frequency (LF) variability is due to a combination of sympathetic and parasympathetic inputs and baroreflex induced changes in HR [16]. We used Welch periodogram approach to calculate the power spectrum of the RRi in every 10-minute epoch. This method involves partitioning the RRi into 30-second windows (so that the frequency resolution of the resulting spectrum is 0.033 Hz) and calculation of periodogram as the square of the magnitude of the Fourier transform of the RRi in that window. We obtained an estimate of the power spectrum as the average of the periodograms from all windows. Using the power spectrum, we calculated the following four metrics: LF and HF powers as the logarithm of the median spectral power in 0.05 – 0.25 Hz and 0.3 – 1 Hz, respectively; we also normalized LF (nLF) and normalized HF (nHF) as the ratio of the sum of the spectral powers in 0.05 – 0.25 Hz and 0.3 – 1 Hz to the total power, respectively. Total power was calculated as the sum of the spectral powers in 0.05 – 2 Hz. The normalized spectral powers (nLF and nHF) are dimensionless quantities. The absolute spectral powers are in unit of decibel (dB). We did not include the LF/HF ratio since this may not effectively characterize changes in ANS function in all circumstances [17].

Statistical Analysis

The following metrics characterize the sympathetic tone: α_S, RMS_s (sec), RMS_L (sec) whereas HF (dB) and nHF characterize the parasympathetic tone [18]. LF (dB) and nLF metrics reflect both sympathetic and parasympathetic mediated activity. α_L characterizes ultraslow changes in the heart rate, below the frequency of sympathetic tone. All analyses were carried out off-line using MATLAB (Mathworks, Inc, MA). The figure was made with MATLAB. For statistical analysis, the HRV from all 10-minute epochs in one-hour windows were averaged.

Infant clinical characteristics were analyzed using standard measures of central tendency and were compared between TGA and HLHS infants, and between CHD infants and controls using a t-test and for categorical variables, a Chi-square test. Categorical data were summarized using counts and percentages. Medication use during the period of data collection and differences in the HRV metrics between the three groups (controls, TGA infants, and HLHS infants) were compared. ANOVA were used to examine the differences among the groups, adjusting for the covariates that were significantly different between infants with CHD and controls (medication exposures). The effect of mechanical ventilation on HRV parameters was also assessed. The statistical analysis was performed using SAS 9.3 (SAS Institute Inc, Cary, NC). To deflate the Type-I error due to multiple comparisons, we considered P<.0005 a statistically significant comparison.

Results

Clinical

Fifty-eight infants with CHD (TGA = 33; HLHS = 25) were included in the study and compared to 29 control newborns. Demographic and clinical characteristic differences between infants with CHD and controls are presented in Table 1. HRV data was analyzed at an earlier mean (SD) hour of age in infants with CHD (23.3[8.7] hours of age in TGA and 24.7[6.8] hours of age in HLHS) compared to controls (31.0[13.7] hours of age) (P<.0170).

Table 1.

Infant Demographic and Clinical Characteristics

Characteristic	TGA	HLHS	Control	P-value TGA vs. HLHS	P-value CHD vs. control
	(n = 33)	(n = 25)	(n = 29)
Gestational age, wks (mean ± SD)	38.94 ± 1.13	38.44 ± 0.90	39.24 ± 0.77	.07	.013
Birth weight, kg (mean ± SD)	3.42 ± 0.42	3.22 ± 0.52	3.39 ± 0.30	.12	.53
Head circumference, cm (mean ± SD)	33.87 ± 1.76	34.10 ± 1.33	34.83 ± 1.51	.58	.019
Male gender (n, %)	24 (72.7)	17 (68)	10 (34.5)	.69	.001
Apgar score 1 min (median, range)	8 (4-9)	8 (5-9)	8 (7-9)	.0965	.0004
Apgar score 5 min (median, range)	8 (7-9)	9 (8-9)	9 (8-9)	.0056	<.0001
Age at study, hrs (mean ± SD)	23.3 ± 8.7	24.7 ± 6.8	31.0 ± 13.7	.4975	.0170
Mechanical ventilation (n,%)	19 (58)	5 (20)	NA	.004*	NA

P-value determined by student t-test unless noted. Chi-square test used for male gender.

Abbreviations: HLHS = hypoplastic left heart syndrome; SD = standard deviation; TGA = transposition of the great arteries

HRV metrics of α_S, RMS_L, LF, and nLF in infants with CHD were significantly lower from those of controls, except nHF, which was higher in the infants with CHD (P<.0001) (Table 2). Absolute HF was however lower in infants with CHD compared to controls, but this did not reach statistical significance (P=.0461) Mechanical ventilation status did not have an effect on HRV metrics (P>.05). Due to their medical condition, infants with CHD had multiple medication exposures (Table 3), none of which were present in controls. After controlling for the effect of medications, α_s and nLF remained significantly lower in infants with CHD (TGA and HLHS) compared to controls and nHF remained significantly higher (P<.0005). In addition, α_S, RMS_L, nLF, and nHF were different between controls and infants with TGA (P<.0005), while α_S, nLF, and nHF were different between controls and infants with HLHS (P<.0001) (Table 2). In summary, the HRV metrics that measure low-frequency (sympathetic) changes were lower in the infants with CHD compared to controls: α_S, RMS_L, LF and nLF.

Table 2.

Heart Rate Variability Metrics in Infants with Congenital Heart Disease and Controls

HRV metric	TGA	HLHS	Controls	Unadjusted	Control vs CHD	Control vs TGA	Control vs HLHS	TGA vs HLHS
(mean ± SD)	(n = 33)	(n = 25)	(n = 29)	P value	adjusted P value #	adjusted P value#	adjusted P value#	adjusted P value#
αS	.59 ± .31	.64 ± .3	1.4 ± .24	<.0001*	<0.0001*	<0.0001*	<0.0001*	0.74
αL	1.16 ± .23	1.23 ± .23	1.22 ± .19	.3775	0.83	0.73	0.94	0.77
RMSS (sec)	.01 ± .01	.01 ± .03	.03 ± .03	.0014	0.26	0.14	0.51	0.43
RMSL (sec)	.04 ± .03	.05 ± .05	.13 ± .06	<.0001*	0.0014	0.0004*	0.01	0.5
nLF	.34 ± .15	.4 ± .13	.7 ± .1	<.0001*	<0.0001*	<0.0001*	<0.0001*	0.78
nHF	.5 ± .11	.47 ± .11	.26 ± .1	<.0001*	<0.0001*	<0.0001*	<0.0001*	0.15
LF (dB)	−3.77 ± .76	−3.53 ± .81	−2.71 ± .49	<.0001*	0.42	0.13	0.91	0.11
HF (dB)	−4.04 ± .6	−3.95 ± .67	−3.66 ±.56	.0461	0.52	0.92	0.24	0.08

Notes: Unadjusted P value from a One way ANOVA, #ANOVA indicates an ANOVA adjusted for medication exposure covariates listed in Table 3,

^*indicates a significant difference between groups.

Abbreviations: α_S = alpha short; α_L = alpha long; dB = decibels; HF = high frequency; HLHS = hypoplastic left heart syndrome; LF = low frequency; nLF = normalized low frequency; nHF = normalized high frequency; RMS_S = root mean square short; RMS_L = root mean square long; TGA = transposition of the great arteries

Table 3.

Medication Exposure Among Infants with Congenital Heart Disease

Medication exposure	TGA (n=33)	HLHS (n=25)	P value
Neuromuscular blocker agents (n, %)	24 (73)	7 (28)	0.0007
Medications for hypotension (n, %)	14 (42)	4 (16)	0.0312
Sedatives (n, %)	26 (79)	11 (44)	0.0063
PGE (n, %)	17 (51)	25 (100)	<.0001

Chi-square test

Abbreviations: HLHS = hypoplastic left heart syndrome; TGA = transposition of the great arteries; PGE = prostaglandin E1

Discussion

This study evaluated ANS function by HRV time- and frequency-domain analysis in the early transitional preoperative period in newborns with complex CHD. Compared to controls, we showed that newborns with complex CHD have depression of the sympathetic nervous system and to a lesser extent the parasympathetic nervous system during the early postnatal period. Although nHF was higher in infants with CHD compared to controls, this is not interpreted as greater changes in parasympathetic tone since the absolute value of HF was less than in control infants. Due to normalization, nHF increased because of relative suppression of sympathetic power (nLF) to that of parasympathetic power since normalized frequencies are modeled to add up to a value of one. Our study methods employed widely-used HRV metrics and newer DFA methods [18–23]. Importantly, the measured autonomic depression remained significant after controlling for medication exposures in the infants with CHD. Knowing that infants with CHD demonstrate both impairment in sympathetic and parasympathetic balance and have a complex anatomic abnormality that may complicate physiological adaptation during the transitional period may in future be used to improve care of newborns with complex CHD. For example, impairment in autonomic balance may precede clinically significant hypotension. Since the sympathetic nervous system is important for maintenance of cerebral blood flow and cerebral autoregulation, lower autonomic balance could disrupt cerebral blood flow below the threshold of autoregulation and affect cerebral perfusion [24]. With the additional circulatory effects of the anatomic cardiac anomalies in these infants, such impairment in autonomic balance might further increase the risk of impaired cerebral perfusion and oxygenation during fetal-to-postnatal transition.

Studies evaluating the normal early maturation of the sympathetic and parasympathetic divisions of the ANS have shown that the sympathetic division begins development earlier than the parasympathetic, which in turn undergoes more accelerated maturation during the third trimester [25, 26]. Adverse intrauterine conditions are known to increase the risk for disturbed fetal development, especially brain development [27–31]. Fetuses with complex CHD, including TGA and HLHS, undergo intrauterine brain development under conditions of tenuous cerebral oxygen delivery. Given the normal rapid development of the central ANS during the third trimester, it is reasonable to speculate that autonomic maturation may also be disturbed in newborns with complex CHD, which is supported by the postnatal HRV data in our study.

To our knowledge, this is the first study of changes in autonomic tone using HRV in infants with CHD during the critical period of fetal-neonatal transition. Several studies have focused on the effect of CHD on HRV characteristics in the fetal period [7, 32]. In a study by Siddiqui et al [7], using fetal ECG recordings, the authors found a decrease in HRV in fetuses with cyanotic CHD, compared to controls. This difference was only significant between 34 and 38 weeks gestational age in fetuses with HLHS, while TGA fetuses showed HRV not significantly different from controls at any gestational age [7]. Specific analysis of the sympathetic versus parasympathetic balance was not described [7]. In a study comparing intrapartum fetal heart rate between fetuses with CHD and controls, severe variable decelerations and prolonged decelerations occurred more often in fetuses with CHD compared to controls [33]. These features were attributed to a difference in autonomic balance with more pronounced parasympathetic response among infants with potentially less hypoxemic reserve than that of the fetuses without CHD. In infants, a hypoxic state may affect ANS function through afferent effects on chemoreceptors that modulate autonomic balance [24].

Unlike Siddiqui’s fetal HRV study [7], we found a significant early postnatal depression of autonomic balance, especially in sympathetic power, in infants with both CHD types, TGA and HLHS. In this regard, our findings are similar to those of Kaltman et al [20], who found no difference in HRV (in either the LF or HF power) during the preoperative period between infants with single- versus two-ventricle anatomy [20]. Other studies have described reduced HRV in infants with CHD during the perioperative period [20, 21]. Smith et al., found that during the preoperative period, HLHS infants had reduced HRV compared to age-matched controls [21]; this study excluded infants on mechanical ventilation. Unlike the study by Smith et al [21], our study focused instead on the very early period after birth. We included infants with TGA and HLHS on mechanical ventilation, and despite a higher number of infants with TGA being on mechanical ventilation; ventilation status did not have an effect on the HRV parameters.

Given the medical acuity in the newborn with complex CHD, we had to consider the effect of other clinical factors, such as medications, on ANS function. Infants undergoing critical care with CHD are exposed to multiple agents with potential autonomic effects, such as the sedatives, medications for hypotension, neuromuscular blocker agents, and prostaglandin E1 which were used in our study population. After controlling for these medication types, we found a persistence of significant autonomic depression in the CHD infants that was shown to be independent of medication exposure.

There are a few studies that examined the relationship between the variability in the RRi and the mean heart rate [35–39]. To mitigate the influence of heart rate on the variability analysis of RRi it has been suggested to normalize RRi either by mean heart rate or by the exponential of the mean heart rate. Since the exact nature of the relation between the variability in RRi and mean heart rate is not clear, we have not explored this option in our study. Furthermore, the normalization should not affect the normalized spectral powers. Since the results of the normalized low-frequency and absolute low-frequency show a similar trend, we believe that our results are unbiased characterizations of autonomic tones.

Our study has a number of strengths but also limitations, including those inherent to a retrospective study design. Since our normative dataset for HRV consisted of recordings performed in the first 24-48 hours after birth, it enabled us to compare infants with complex CHD to control infants at a similar postnatal age. We were, however unable to compare the control infants to CHD infants at a later time period since controls only had one ECG session. Infants with CHD were studied at an earlier hour of age than control infants, due to earlier availability of the CHD infants’ data upon admission to our tertiary care children’s hospital compared to control newborns that were prospectively recruited following birth. While this difference is statistically significant, the biological difference is likely insignificant. Since the ANS maturation after birth is unlikely to be on the scale of hours, we doubt that this small time difference accounts for the differences in ANS maturation between our cases and controls [4, 40]. Furthermore, a larger number of infants with TGA were on mechanical ventilation due to need for balloon atrial septostomy compared to infants with HLHS. We controlled for medication exposure and evaluated for the impact of mechanical ventilation in our infants with CHD, to understand if this clinical difference impacted HRV findings. Children’s National is a referral center for neonatal cardiac surgery with no delivery services. Thus, all except one of the CHD cases were out-born, and many were started on PGE1 treatment prior to arrival at our institution. Prostaglandins are known to exert effects on the ANS and thus might have played some role in the results of our study. The control infants were low-risk term newborns and therefore did not have exposure to postnatal medication or mechanical ventilation, which were present in the infants with CHD.

Conclusions

Newborns with complex CHD have depressed HRV metrics of primarily sympathetic and parasympathetic tone balance in the early postnatal transitional period, prior to cardiac surgery, compared to control newborns. Reduced autonomic tone in the early postnatal preoperative period may be due to the impact of prenatal hypoxemia and altered hemodynamics on the developing brain and ANS. Over the last decade, it has become increasingly clear that the high prevalence of long-term neurodevelopmental morbidity in survivors of certain CHD types is likely due to an accumulation of neurologic insults at multiple points in time, including the fetal, pre-, intra-, and postoperative periods. Further studies are required to evaluate the impact of impaired ANS function in newborns with CHD on their clinical course in the cardiac intensive care unit and on long-term neurologic outcome.