Heart rate variability is associated with self-regulation and stress neurobehavior in preterm infants

Abstract

Background:

The NICU (neonatal intensive care unit) Network Neurobehavioral Scale (NNNS) quantitatively assesses several distinct neurobehavioral domains. Neurobehavioral states interface with the autonomic nervous system. We sought to determine if autonomic tone, measured by heart rate variability (HRV), is associated with performance on NNNS neurobehavioral subscales.

Methods:

Using continuous electrocardiogram recordings, we derived the following HRV measures: short (α1), and long-term (α2) fractal scaling exponents and short (RMS1) and long-term (RMS2) root mean square fluctuations. The NNNS was completed around term equivalent age (TEA). Linear mixed effects models tested associations between six NNNS subscales and HRV metrics. Secondary analyses explored the role of postmenstrual age (PMA) on associations between NNNS and HRV. These associations were further tested in the earliest and latest postnatal NICU weeks for each infant.

Results:

We enrolled 39 premature infants (birth gestational age 23–36 weeks). Several HRV metrics were associated with Self-Regulation and Stress. After controlling for common NICU exposures (i.e. ventilation, steroids, caffeine and antibiotics), self-regulation was positively associated with alpha1 and RMS1 (standardized β=.262–.284). Stress was associated negatively with alpha1 and RMS1 (standardized β=−.283 – −.286) and positively with alpha2 (standardized β=.245). Associations between Self-Regulation and Stress on RMS1 differed by PMA and NICU week, respectively. α1 and α2 at TEA predicted Stress (AUC=.730–.775).

Conclusion:

Of the six subscales examined, only Stress and Self-Regulation scores were associated with autonomic metrics in preterm infants. For short-term autonomic fluctuations, PMA and NICU week influenced this relationship. Furthermore, the fractal scaling exponents predict abnormal Stress scores.

1. Introduction

Early life experiences may shape autonomic nervous system (ANS) development[1], [2]. The challenges and stressors to which infants are exposed, alter how their autonomic and allostatic systems are deployed which enables the infant to better adapt to their internal and external environments. In preterm infants, these experiences occur while the ANS is still underdeveloped. The ‘unnatural’ premature extrauterine environment exposes these infants to excessive sensory stimulation from light, sound, and pain, disrupted sleep/wake cycles, and limited parental contact, which may impact autonomic development. As a result, differences in both autonomic tone[3], [4] and neurobehavior[5] are observed in ex-preterm infants.

Heart rate variability (HRV), which describes differences between successive heart beats, is a widely accepted, noninvasive indicator of autonomic tone. The brainstem communicates with the heart through both branches of the ANS. Parasympathetic input to the sinoatrial node decreases heart rate and increases HRV, while sympathetic activation accelerates heart rate subsequently decreasing HRV. According to the neurovisceral integration model, the mature brainstem also communicates with higher limbic and cortical centers responsible for neuro-affective and - behavioral function[6]. During late gestation and early life maturation, these centers as well as their communication are especially active and therefore sensitive to disruptive environmental influences[7]. Disturbed programming of these systems may result in long-term dysregulation of responses to environmental stimuli, including the stress response[8], [9], [10].

There are a number of neonatal assessments aimed at predicting infants at risk of poor neurodevelopmental outcomes [9], [11] as well as newer approaches such as the WECS[12] that assess the strength of the mother-infant emotional connection, however the NNNS has certain advantages to other scales. The NNNS incorporates principles of many other assessments such as the NBAS, NAS, NAPI and APIB[9], [11]. Furthermore, it has been validated and extensively used in preterm infants and there is comparative data from a normative population.

The NICU Network Neurobehavioral Scale (NNNS)[13] estimates the risk of neurodevelopmental delays in premature infants[14], which opens potential windows for amelioration of such risks[15]. Likewise, certain HRV metrics in the newborn have been used to predict neurodevelopmental outcomes[15], [16]. The NNNS consists of neurologic components and behavioral components, the latter adapted from the Neonatal Behavioral Assessment Scale[17], [18]. Though the NNNS recognizes the bidirectional relationship between newborn behavioral cues and physiological systems[18], few studies have investigated the relationship between neonatal neurobehavior and physiological measures of stress such as HRV[19], [20] or cortisol levels[21]. The NNNS measures neurobehavior originating from physiological processes, through qualitative observation of externally apparent behavioral responses. These are often binary (i.e., present or absent) and don’t necessarily indicate the degree of autonomic dysregulation that these infants are experiencing. An understanding of how HRV relates to neonatal neurobehavior (as measured by the NNNS) may supplement our knowledge of how aberrant scores in certain NNNS subscales relate to internal autonomic regulation.

The primary objective of this study is to determine whether HRV measures of autonomic regulation, derived using detrended fluctuation analysis (DFA)– a method for quantifying long-range correlations in non-stationary time series are associated with NNNS subscale scores in infants undergoing third trimester development in an extrauterine environment. Our secondary aim is to determine when these associations develop. Subsequent analyses will determine if HRV can be used to identify abnormal performance on the NNNS.

2. Methods

2.1. Sample

This is a retrospective observational study using data from two pre-existing cohorts of a larger study. Inclusion criteria were preterm infants born <37 weeks gestation, recruited after admission to our level IV NICU, with continuous ECG recordings through their NICU stay and term-equivalent NNNS scores. Exclusion criteria were other parenchymal brain lesions, chromosomal, genetic, syndromic or metabolic abnormalities.

All procedures were performed in compliance with relevant laws and institutional guidelines and have been approved by our institutional review board (Pro00002391 approved 4/3/2012). The privacy rights of human subjects were observed and written informed consent was obtained from each parent. There was no monetary compensation for enrolling.

2.2. Procedure

Participants were recruited after admission to the NICU. The physiological signals including ECG sampled at 250 Hz were retrieved from a Phillips Data Warehouse Connect system (PIIC iX Rev B, Phillips Healthcare, Andover, USA) and archived in a local biorepository. For every infant, the ECG signal was retrieved from the repository from the time of admission to the time of discharge. Using ECG, the QRS complexes were identified, and the heart rate was calculated[22]. The heart rate was partitioned into 10-minute non-overlapping epochs. For data in every epoch, HRV metrics were calculated. The NNNS[13] assessment was administered by a trained clinician between 37–48 weeks PMA. Assessments were entered in and retrieved from a REDCap database[23].

2.3. Neurobehavioral Assessment

The NNNS is a neurobehavioral assessment developed for high-risk and premature infants[18]. It assesses neurologic integrity and behavioral functioning. Unlike its predecessor, the Neonatal Behavioral Assessment Scale (NBAS), the NNNS has established norms[24], [25] and is designed to be independent of the examiner’s skill level[17]. Furthermore, this scale tends to have good reliability (α = .87 - .90)[17]. We applied six behavioral NNNS subscales, including arousal, attention, excitability, lethargy, self-regulation and stress/abstinence, since these were most likely to interface with the autonomic system.

2.4. Detrended Fluctuation Analysis

The methodology for attaining HRV using DFA has been described in detail elsewhere[26]. The α exponents characterize autocorrelations in the beat-to-beat intervals and the RMS metrics characterize variability in the beat-to-beat intervals[26]. The α1 and RMS1 characterize short-term changes (parasympathetic tone) in the heart rate and RMS2 characterize long-term changes (sympathetic tone) in the heart rate.

2.5. Clinical Characteristics

Small for gestational age was defined as having a birth weight at or below the 10^th percentile established in Oken et al., 2003 for the infant’s respective sex[27]. Ventilation describes infants receiving nasal cannula, CPAP, NIPPV, conventional ventilation and high flow ventilators. Infections were defined as positive microbial tests and 5 or more days on antibiotics. Histological chorioamnionitis was based upon placental pathology. Necrotizing enterocolitis was defined as any necrotizing enterocolitis based on modified Bell’s staging criteria[28]. Bronchopulmonary dysplasia was defined as respiratory support administered at 36 weeks PMA[29]. Patent ductus arteriosus was defined as the presence of moderate to severe patent ductus arteriosus on an echocardiograph. Brain injury was assessed using the Kidokoro scale[30]. This scale objectively assess the full extent of brain abnormalities in conventional magnetic resonance images. Global scores below 3 are considered normal, between 4–7 are mild, between 8–11 are moderate and above 12 are severe.

2.6. Statistical Analysis

Statistical analyses were performed using Stata 18 SE[31]. Descriptive statistics such as means and standard deviations are reported for continuous measures. The number of participants and percentages of the whole sample are reported categorical measures.

Due to repeated measures of HRV within participants, a linear mixed effects (LME) model was used to test associations between NNNS subscales and HRV metrics. Each NNNS subscale and PMA, which is the sum of gestational age (GA) and PNA, were included in the model as fixed effects. Both an unadjusted and an adjusted model with dynamic clinical covariates included were considered. The random intercept model was clustered by the participant identifier. Since 6 NNNS subscales were being tested, the Benjamini-Hochberg procedure[32] was applied to control for the false positive rate in multiple comparisons.

Given the known associations of ventilation, antibiotics, steroids and/or caffeine with HRV, adjusted models included these measures as dynamic, binary covariates for each week to determine if relationships between HRV and NNNS subscales were driven by exposure to treatments.

As part of our secondary analysis, remaining significant associations were tested for significant interactions between the NNNS subscale and PMA. The Johnson-Neyman procedure[33] was used to identify the PMA at which NNNS and HRV are no longer correlated. For this and subsequent analysis, the significance level was set at p < .05.

Generalized linear models (GLM) were used to determine if the relationship between NNNS and HRV differed between the first and last week of NICU hospitalization. The average HRV for each participant’s first and last available week in the NICU was calculated as the independent variable. The coefficients of each GLM based on each week were compared.

To determine if HRV could predict abnormality, a binary logistic regression model was used to estimate the odds and the associated 95% confidence interval of having abnormal NNNS subscale scores. Normality was defined as scoring within one standard deviation of the published means[24] from a sample of healthy, full-term newborns. Abnormality was defined as scoring less than 1 SD from the mean in stress or higher than 1 SD from the mean in self-regulation compared to the normal participants. HRV for each infant was averaged over the week closest to term-equivalent age for which HRV data was available. Aside from PMA, no other covariates were included in the model.

3. Results

We included 39 infants born in the United States from April 2017 to September 2022 who met our eligibility criteria.

Demographics:

Descriptive statistics for the sample are summarized in Table 1. All infants were between 23 and 36 weeks GA at birth. Moderate-late (32–37 weeks) preterm infants made up 12.8%, very (28–32 weeks GA) preterm infants made up 33.3% and extremely preterm infants (less than 28 weeks GA) made up 53.8% of the sample. Twenty-two (56.4%) of the infants were male. Mothers were between 17 and 40 years old at the time of delivery. Nineteen (55.9%) of the participants were evaluated with the NNNS after discharge. Of the 6 infants discharged prior to TEA, 3 were released to a hospital or facility.

Table 1.

Descriptive Statistics (n = 39).

Birth Characteristics
Female	17 (43.6)
GA at Birth (weeks)	28.20 ± 0.69
PMA during NNNS (weeks)	40.72 ± 0.61
Small for GA	2 (5.13)
Apgar 1-minutec	5.31 ± 2.36
Apgar 5-minutec	7.38 ± 1.52
NICU Treatments
Ventilator	38 (97.4)
Caffeine	33 (84.6)
Antibiotics	36 (92.3)
Steroids	22 (56.4)
Clinical Characteristics
Duration of Hospitalization (days)b	76.88 ± 34.37
Kidokoro Scorea	6.61 ± 6.50
No diagnosisd	2 (8.33)
Infectiond	22 (54.2)
Respiratory Distressd	18 (75.0)
Bronchopulmonary Dysplasiad	12 (50.0)
Retinopathy of Prematurityd	14 (58.3)
Patent Ductus Arteriosusd	11 (45.8)
Sepsisd	6 (25.0)
Chorioamnionitisd	3 (12.5)
Necrotizing Enterocolitisd	2 (8.3)
Grade 1	1 (4.2)
Grade 3	1 (4.2)
Hydrocephalusd	4 (16.7)
Ventriculomegalyd	6 (25.0)
Parental Education (n = 70 parents)
Graduate School
Both Parents	11 (15.7)
Single Parent	1 (1.4)
College or University
Both Parents	11 (15.7)
Single Parent	0 (0.0)
High School
Both Parents	38 (54.3)
Single Parent	2 (2.9)
Less than High School
Both Parents	7 (10.0)
Single Parent	0 (0.0)
Parental Employment (n = 66 parents)
Employed
Both Parents	50 (75.8)
Single Parent	3 (4.5)
Unemployed
Both Parents	11 (16.7)
Single Parent	0 (0.0)
Homemaker
Both Parents	2 (3.0)
Single Parent	0 (0.0)
NNNS Subscale
Arousal	3.72 ± 0.08*
Attention	4.67 ± 0.25*
Excitability	2.10 ± 0.35*
Lethargy	4.76 ± 0.40*
Self-Regulationa	5.44 ± 0.15*
Stress / Abstinence	0.10 ± 0.01*

Number of cases (percentages %) or means ± standard deviations.

^*denotes means that are significantly different from published norms.

Abbreviations: Gestational age (GA), postmenstrual age (PMA), and neonatal intensive care unit (NICU) network neurobehavioral scale (NNNS).

^acalculated from 36 infants

^bcalculated from 34 infants

^ccalculated from 32 infants

^dcalculated from 24 infants

The portion of infants that were treated with antiobiotics, ventilations, steriods or caffeine at any point during their NICU hospitalization are displayed in Table 1. On average, infants spent 5 days on high-flow ventilators, 47 days on caffeine, 16 days on antibiotics and 12 days on steroids. Kidokoro scores ranged from normal (38.5%), mild (23.1%), moderate (12.8%), to severe (25.6%).

Compared to normative means, this sample had significantly lower levels of arousal (∆ = .62, t(378) = 5.683, p < .001), attention (∆ = 1.3, t(299) = 7.342, p < .001), excitability (∆ = 2.7, t(381) = 7.128, p < .001) and stress (∆ = .04, t(378) = 4.979, p < .001). They had significantly higher levels of lethargy (∆ = −.75, t(381) = −1.699, p < .001) and self-regulation (∆ = −.36, t(343) = −2.908, p = .002).

Associations between NNNS and HRV:

Table 2 displays the unadjusted and adjusted LME models. The unadjusted model includes each observation of HRV throughout their NICU hospitalization and the corresponding PMA as predictors of NNNS. The adjusted model includes common NICU exposures (use of antibiotics, caffeine, steroids and ventilation) as dynamic, binary covariates to the unadjusted model.

Table 2.

Unadjusted & Adjusted LME Model

NNNS	HRV	Without Covariates			With Covariates
		Coefficient	SE	p	Coefficient	SE	p
Arousal	Alpha1	.075	.053	.158	.092	.058	.115
	Alpha2	-.023	.030	.441	-.019	.029	.511
	RMS1	.203	.218	.352	.208	.217	.338
	RMS2	.660	.567	.245	.738	.660	.263
Attention	Alpha1	.024	.019	.209	.019	.022	.392
	Alpha2	-.019	.011	.081	-.013	.010	.228
	RMS1	.140	.077	.069	.094	.080	.240
	RMS2	.284	.205	.165	.276	.245	.262
Excitability	Alpha1	-.031	.018	.083	-.032	.020	.107
	Alpha2	.015	.010	.140	.014	.010	.142
	RMS1	-.147	.072	.041^	-.124	.073	.089
	RMS2	-.272	.194	.161	-.264	.228	.246
Lethargy	Alpha1	-.007	.013	.582	-.008	.015	.607
	Alpha2	.008	.007	.298	.003	.007	.652
	RMS1	-.027	.055	.618	-.022	.055	.697
	RMS2	.004	.143	.975	-.038	.169	.820
Self-Regulation	Alpha1	.108	.035	.002*	.115	.038	.003*
	Alpha2	-.049	.021	.018^	-.044	.020	.024^
	RMS1	.475	.140	.001*	.424	.145	.004*
	RMS2	1.008	.361	.005*	1.011	.432	.019^
Stress / Abstinence	Alpha1	-1.348	.426	.002*	-1.504	.482	.002*
	Alpha2	.847	.227	<.001*	.715	.231	.002*
	RMS1	-6.294	1.649	<.001*	-6.098	1.717	<.001*
	RMS2	-11.359	4.778	.017^	-14.429	5.650	.011^

^{^}indicates p < .05,

^*indicates p-value is significant after Benjamini-Hochberg correction.

Covariates include post-menstrual age, ventilation, antibiotics, steroids, and caffeine.

Abbreviations: heart rate variability (HRV), linear mixed effects (LME), and root mean square (RMS).

Secondary analysis for the significant associations between NNNS and HRV:

For associations that were significant in both the adjusted and unadjusted models, we explored potential interactions with PMA in a secondary analysis. The association between self-regulation and RMS1 differed by PMA (β = .024, SE = .009, p = .009), however no other interactions were significant. In Figure 1, the Johnson-Neyman[33] procedures is applied to the significant interaction to detect the point at which the association between PMA and self-regulation becomes significant.

Figure 1. Self-regulation’s association with RMS1 depends on PMA:

Dashed vertical line represents the point at which self-regulation scores correlate with RMS1 (PMA = 25.4). No covariates are included. Low self-regulation values are offset by 1 day to prevent overlapping. Abbreviations: postmenstrual age (PMA) and root mean squared (RMS).

Next, we analyzed the first and last week of NICU hospitalization in statistically significant adjusted models (see Table 3.). RMS1 was negatively associated with stress in the first week (β = −3.329, SE = 1.464, p = .029). No other associations were significant.

Table 3.

Associations between NNNS subscales and HRV by NICU Week.

NICU Week	NNNS	HRV	Coefficient	SE	p
First	Self-Regulation	Alpha1	.027	.036	.459
		RMS1	.114	.103	.276
	Stress / Abstinence	Alpha1	-.705	.489	.158
		Alpha2	.283	.391	.474
		RMS1	-3.329	1.464	.029*
Last	Self-Regulation	Alpha1	.061	.039	.124
		RMS1	.343	.194	.086
	Stress / Abstinence	Alpha1	-.376	.506	.463
		Alpha2	.527	.297	.085
		RMS1	-3.798	2.516	.140

^*indicates a significant interaction.

Postnatal age and gestational age at birth are covariates.

Abbreviations: heart rate variability (HRV), neonatal intensive care unit (NICU) network neurobehavioral scale (NNNS), and root mean square (RMS).

Prediction of abnormal NNNS using HRV measures:

Table 4 displays the results of the GLM logistic regression to determine if HRV measured near TEA could predict abnormally low stress and high self-regulation scores in preterm infants. HRV averages for the week nearest to TEA were used as the predictor. The mean PMA when HRV was collected was 37.744 ± 2.584 weeks. An α1 cutoff value of .597 resulted in 54.55% sensitivity, 88.24% specificity and 69.23% accuracy. For each .001 increase in α1 above the mean, the odds of abnormally low stress increased by 4.72%. An α2 cutoff value of .548 resulted in 72.73% sensitivity, 82.35% specificity and 76.92% accuracy. For each .001 increase in α2 above the mean, the odds of abnormally low stress decreased by 9.25%. Furthermore, for each additional week of PMA above the mean, the odds of abnormally low stress decreased by 29.33%.

Table 4.

Logistic Regression predicting Abnormal Stress from HRV.

NNNS	HRV	OR	SE	p	AUC
Self-regulation	α1	5.518	11.481	.412	.708
	α2	0.007	0.027	.176	.679
	RMS1	2.141	0.963	.090	.721
	RMS2	1.272	0.269	.255	.679
Stress / Abstinence	α1	100.752	234.657	.048*	.730
	α2	<0.001	<.001	.017*	.775
	RMS1	2.280	.961	.051^	.746
	RMS2	1.219	.231	.296	.685

^{^}indicates p < .10,

^*indicates p < .05;

Abbreviations: area under the curve (AUC), heart rate variability (HRV), neonatal intensive care unit (NICU) network neurobehavioral scale (NNNS), odds ratio (OR), and root mean square (RMS).

4. Discussion

In this study, we sought to explore associations between neurobehavioral measures and autonomic tone in preterm infants. As expected, we found that autonomic tone during NICU hospitalization was positively associated with self-regulation and negatively associated with stress. The relationship between autonomic tone and both self-regulation and stress remained significant even after adjusting for covariates known to influence HRV. Short-term autonomic tone between 25 and 40 weeks PMA was associated with self-regulation scores at TEA. Furthermore, short-term autonomic tone during the first week of NICU hospitalization, but not the last week, was associated with stress scores at TEA. We also found that higher HRV predicted lower than normal stress behaviors, but it was not predictive of self-regulation. Although we included six neurobehavioral NNNS subscales in our analysis, only stress and self-regulation scores were significantly related to autonomic tone, while the arousal, attention, excitability or lethargy subscale scores were not.

Available literature that examines how performance on the NNNS relates to autonomic function is limited. Of the published research, HR[20], respiratory sinus arrythmia (RSA)[20], high-frequency (HF) HRV[19], and cortisol[21] have been used as measures. Conradt and colleagues (2013) found no relationship between baseline HRV and either stress or self-regulation, but they did see an association with lethargy[20], though this was in a population of term infants with prenatal substance use, and therefore they faced different adverse exposures than our preterm subjects[20]. Using cortisol as a biomarker of stress, Zhang and coauthors (2022) found a correlation with self-regulation in preterm infants, but no association with stress scores on the NNNS[21]. Aldrete-Cortez and peers (2025), similarly found no association between vagal tone and CNS or skin stress signs, but they did find an association with lethargy and excitability in a smaller cohort of late-preterm infants[19]. The infants in our study had a significantly lower birth gestational age than either of these studies, which might explain the difference in our results.

Neither arousal, attention, excitability nor lethargy were associated with autonomic tone. Though this neurobehavior may be associated with changes in heart rate, their relationship to autonomic function might not be as direct as the relationship the ANS has with self-regulation and stress. Further studies are needed to determine why these subscales showed no relationship with autonomic tone.

This study provides a nuanced understanding of the ANS in relation to neurobehavior by using several HRV measures that have not been previously explored in this context. Unlike linear techniques, DFA HRV is robust to noise and reflects the complex fractal nature of the cardiovascular system[34]. As such, these measures might be detecting patterns that RSA and HF cannot. Ongoing studies are needed to provide further insight into this relationship.

After birth, nonverbal communication with caregivers through behavior is essential for infants to convey either distress or contentment. In preparation, the connection between infant behavior and internal distress might begin to develop prior to birth. Based on our findings, the third trimester of development may be a critical period for the establishment of neurobehavioral self-regulation. The 3rd trimester is also a period of rapid increase in parasympathetic development due in part to myelination of the vagus nerve[7]. Between the 25–30th gestational week, the vagus nerve starts to influence heart rate[35]. It is possible that the association between self-regulation and short-term parasympathetic activity is dependent on the maturation of vagal tone. Ongoing studies are needed to determine how early the connections between infant behaviors and internal states are established.

There are several benefits to understanding the relationship between the NNNS and HRV. Firstly, infants cannot be assessed on the NNNS until they reach TEA, however HRV can be measured far earlier. Based on the results of this study, after birth clinicians can predict poor self-regulation and elevated stress scores in infants with low HRV. Additionally, it is now apparent that lower scores on the stress subscale are characterized by higher parasympathetic tone proportional to sympathetic tone and increased autonomic complexity which are both indicative of a more mature autonomic nervous system. Better understanding of the relationship between HRV and NNNS scores will advance our knowledge of the neural underpinnings of infant behavior.

There are several limitations to our findings. First, our study population is not large, which limits our ability to account for additional confounding variables. Furthermore, our study population had higher self-regulation and lower stress scores than other preterm[9], [14], [36] or even full-term newborns[24], [25] in other reports. Given that preterm infants are exposed to an overstimulating extrauterine environment and painful procedures, this raises questions as to whether the lower stress and higher self-regulation scores seen here and in other reports[21], [37], [38] are due to resilience or allostatic load. Lastly, only 14 infants were born before 25 weeks GA, which might explain why significant associations were not observed between HRV and self-regulation before 25 weeks and 3 days. Additional studies with a larger sample size across a broader GA are needed.

This study did not touch on the impact of parents on the HRV of preterm infants. Though research supports the notion that parental factors, such as maternal socioeconomic status, skin-to-skin care[39], emotional connection[12], can have a favorable impact on the infant’s parasympathetic nervous system as well as their neurobehavioral outcomes[38], our analysis focused solely on the infant’s neurobehavior instead of the dyadic relationship. Due to the fact that this was a retrospective analysis of data from an observational study, data on the duration and frequency of skin-to-skin care was limited. Future studies will need to determine if the relationship between HRV and neurobehavior is mediated by maternal attachment, socioeconomic status, or mental health.

The present study examined the relationship between autonomic tone and a set of NNNS subscales in preterm infants receiving care in a level IV NICU. Our findings emphasize the need for longer term follow-up to assess whether these associations are enduring. Furthermore, additional studies are required to determine if our findings are supported by those in other high-risk populations of infants, including full-term newborns. We do not know if the relationship we observed between the NNNS subscales and HRV is unique to infants spending their third trimester ex utero or is also evident in infants born later from unsupportive intrauterine environments, such as fetal growth restriction.

The findings of this study have implications for the assessment and treatment of preterm infants at risk of neurodevelopmental delays. As more hospitals implement near real-time HRV analysis at the bedside [41], clinicians will be able to assess neonatal well-being through the use of HRV. Since the development of the NNNS, promising therapies for the restoration of infant autonomic tone have been developed[42], namely, Family Nurture Intervention (FNI)[43]. This approach focuses on the autonomic emotional connection of the mother-infant dyadic relationship instead of infant pathology as a focus of intervention[44]. As such, similar therapies can be appropriate for improving autonomic tone in preemies identified as having poor autonomic regulation. In conclusion, the results of our study support the notion that autonomic tone in the weeks leading up to TEA is associated with self-regulation and stress behavior in preterm infants. Researchers using the NNNS scores to evaluate an infant’s neurobehavioral status should consider additional physiological measures, such as autonomic dysregulation.

Highlights:

• Self-regulation and stress are associated with heart rate variability metrics.

• These associations remain after controlling for common preterm exposures.

• Prior to 25 weeks postmenstrual age, variability is not related to self-regulation.

• Associations of stress or self-regulation and variability may be time dependent.

• Alpha1 and Alpha2 predict abnormally low stress at term-equivalent age.

Glossary

ANS

Autonomic Nervous System

DFA

Detrended Fluctuation Analysis

GA

Gestational Age

HF

High Frequency

HR

Heart Rate

HRV

Heart Rate Variability

NBAS

Neonatal Behavioral Assessment Scale

NICU

Neonatal Intensive Care Unit

NNNS

NICU Network Neurobehavioral Scale

PMA

Postmenstrual Age

PNA

Postnatal Age

RMS

Root Mean Square

RSA

Respiratory Sinus Arrythmia

TEA

Term-equivalent Age