The role of endothelial frequency in the cerebral blood flow control during neonatal asphyxia: a retrospective longitudinal study

Abstract

Background

Cerebral blood flow dynamics can be explored through analysis of endothelial frequencies. Our hypothesis posits a disparity in endothelial activity among neonates with perinatal asphyxia, stratified by the presence or absence of neuronal lesions.

Methods

We conducted a retrospective longitudinal study involving newborns treated with hypothermia for moderate to severe asphyxia. Participants were grouped based on the presence or absence of neuronal damage to investigate temporal endothelial involvement in cerebral blood flow regulation. Regional cerebral oxygen saturation (rScO2) was measured using near-infrared spectroscopy (NIRS), and temporal series were analyzed in the frequency domain, utilizing the original frequency of the INVOS™ device.

Results

The study included 88 patients, with 53% (47/88) being male and 33% (29/88) demonstrating brain lesions on magnetic resonance imaging. Among them, 86% (76/88) had a gestational age exceeding 37 weeks according to the Ballard scale, and 81% (71/88) had a birth weight exceeding 2500 g. Cohen’s d effect size was calculated to assess differences in endothelial frequency between groups, indicating a small effect size based on cerebral MRI findings (Cohen’s d values for Day 2 = 0.2351 and Day 3 = 0.2325).

Conclusion

NIRS represents a valuable tool for monitoring cerebral autoregulation in neonates affected by perinatal asphyxia, underscoring the utility of assessing endothelial frequency or energy on rScO2 measured by NIRS using the original INVOS™ device frequency.

Introduction

Perinatal asphyxia remains a significant clinical concern, contributing substantially to neurological complications in neonates and ranking as the second leading cause of neonatal mortality and morbidity worldwide [1, 2]. This condition, characterized by reduced oxygen and glucose supply to the infant during or immediately after birth, initiates a cascade of complex pathophysiological processes that can lead to both immediate and long-term neurological consequences, accounting for a considerable portion of neonatal mortality [2, 3]. These processes encompass distinct stages, including primary and secondary energy failure, latency, reperfusion injury, and a reparative phase, each playing a critical role in the spectrum of perinatal asphyxia [4].

Current therapies, notably therapeutic hypothermia, have demonstrated efficacy in mitigating long-term complications. However, enduring neurological and renal impairments, along with other multiorgan dysfunctions, persist [5,6,7]. Throughout the progression of the condition, alterations in cerebral blood flow manifest in patients with adverse neurological outcomes. These changes in blood flow dynamics are attributed to neuronal damage occurring in the initial and subsequent phases of perinatal asphyxia [8, 9]. Variations in endothelial frequencies observed during hypothermia among patients with and without neurological injury can be attributed to neuronal damage occurring in the early and advanced stages of perinatal asphyxia [8,9,10]. This damage leads to an upsurge in extracellular glutamate, which influences cerebral endothelial cells via the N-methyl-D-aspartate (NMDA) receptor and consequent vasodilation [10].

This study comprehensively assessed the proposed hypothesis by specifically investigating endothelial frequencies and utilizing Regional Cerebral Oxygen Saturation (rScO2) values measured via Near-Infrared Spectroscopy (NIRS) as prognostic indicators for perinatal asphyxia [10,11,12]. The findings are anticipated to underscore the pivotal role of NIRS values in predicting the severity of perinatal asphyxia and the likelihood of neuronal damage, thereby reshaping clinical strategies for managing this significant medical condition. The primary objective was to examine the correlation between endothelial frequency within the specified spectral band and rScO2 values obtained via NIRS with cerebral magnetic resonance imaging (MRI) injury in neonates undergoing hypothermia treatment for moderate and severe perinatal asphyxia.

Methods

Study design and setting

A retrospective longitudinal study was conducted on neonates treated with hypothermia for moderate to severe perinatal asphyxia at the neonatal intensive care unit of the Fundación Cardio Infantil-Instituto de Cardiología in Bogotá, Colombia, from November 2021 to August 2022. The study population comprised term neonates (determined by a Ballard score of ≥ 37 weeks of gestation) within ≤ 12 h postnatal and a birth weight of ≥ 1800 g. The study was reviewed and approved by the ethics committee of the Fundación Cardio Infantil-Instituto de Cardiología (code: CEIC-0602-2022).

This study focused on analyzing cerebral saturation values obtained via NIRS in newborns who experienced perinatal asphyxia and were treated with therapeutic hypothermia, with continuous monitoring using cerebral NIRS. The cohort was divided into two groups: newborns without neuronal lesions and newborns who developed neuronal lesions confirmed by MRI after the treatment period.

Eligibility criteria

Inclusion criteria required the presence of moderate to severe perinatal asphyxia, confirmed by umbilical cord arterial blood gases with pH ≤ 7.0 or base deficit ≥ -16, or postnatal blood gases within the first hour of life with pH 7.01–7.15 or base deficit of -10 to -15.9. Additionally, inclusion required a history of acute perinatal events and an Apgar score ≤ 5 at 5 min or the need for at least 10 min of positive pressure ventilation. Only neonates with moderate to severe hypoxic-ischemic encephalopathy, defined by Sarnat stages 2 and 3, were eligible. Exclusion criteria encompassed neonates with intrauterine growth restriction leading to a birth weight below 1800 g, congenital anomalies, or chromosomal disorders.

Clinical variables

We described the following variables: sex, gestational age in weeks determined by Ballard score, route of delivery (vaginal and caesarean), type of resuscitation at birth (basic and advanced), encephalopathy severity according to Sarnat staging, APGAR score, severity of asphyxia, pH, bicarbonate (HCO3), base excess, lactate, hemoglobin, and use of inotropic agents. Patient data were obtained from hospital medical records.

Moderate neonatal encephalopathy is classified as stage II according to the Sarnat staging, whereas severe neonatal encephalopathy is classified as stage III according to Sarnat [13, 14]. Severe perinatal asphyxia is characterized by the presence of at least three of the following criteria: Apgar score ≤ 5 at 5 min, pH < 7.0 in the first hour of life in arterial, venous, or capillary cord blood samples, base excess deficit ≤ -16 mmol/L in the first hour of life, moderate to severe encephalopathy (Sarnat stages II–III), and lactate ≥ 12 mmol/L during the first hour of life [15]. On the other hand, moderate perinatal asphyxia requires the presence of at least two of the following criteria: Apgar score ≤ 7 at 5 min, pH < 7.15 in the first hour of life in arterial, venous, or capillary cord blood samples, and mild to moderate encephalopathy (Sarnat stages I–II) [13,14,15]. Basic neonatal resuscitation included initial respiratory support through thermoregulation, positioning, secretion aspiration, and positive pressure ventilation and/or endotracheal intubation [16]. Advanced resuscitation involved chest compressions, umbilical catheterization, and medication administration [16].

Therapeutic hypothermia was initiated within six hours after birth and maintained for 72 h, followed by a 6-hour rewarming period. During this time, cerebral oxygenation was continuously monitored using neonatal cerebral oximetry sensor INVOS™ (Irvine, California, United States of America) placed on the frontal cranial region. Data were recorded every 30 s and averaged over the first 24 (day 1), 48 (day 2), and 72 (day 3) hours of hypothermia, as well as during the 6-hour rewarming period.

The hypothermia protocol involved total body cooling with the ThermoWrap^® hypothermia blanket and temperature monitoring using an esophageal thermal sensor placed in the lower third of the esophagus. The core temperature was reduced to between 33 and 34 °C within 30 to 40 min during the induction phase. Subsequently, the target temperature was maintained at 33.5 °C ± 0.5 for 72 h during the maintenance phase. Reheating was conducted over 6 h at a rate of 0.5 °C per hour until a final temperature of 36.5 °C was reached. Continuous monitoring of rScO2 was performed with the INVOS™ sensor (Irvine, California, United States of America) throughout all phases of hypothermia.

The machine generated continuous readings, recording data every 30 s. These records were exported to an Excel database using the INVOS software Shortcut to Invos Analytics Tool. Missing data points in the brain NIRS time series were identified and interpolated using MATLAB R2023a.

Following the completion of therapeutic hypothermia and subsequent warming, MRI scans were performed to classify neonates into two groups based on the presence or absence of cerebral lesions. All patients underwent brain imaging at one week of age (between 5 and 7 days). Brain injury severity was evaluated using conventional T1- and T2-weighted spin echo sequences, diffusion-weighted imaging, and apparent diffusion coefficient maps.

The brain injury criteria followed ASCON16 recommendations. Patterns of acute asphyxia included: (a) abnormal signal in the basal ganglia and peri-rolandic cortex, (b) alteration or disappearance of normal signal intensity in the posterior limb of the internal capsule, (c) prolonged partial asphyxia pattern with signal involvement in vascularization areas bordering the middle cerebral artery and posterior cerebral artery, affecting white matter, (d) patterns of perinatal ischemic or hemorrhagic stroke and/or venous sinus thrombosis. Presence of at least one criterion classified MRI findings as indicative of cerebral alteration.

Additionally, neonates who did not survive long enough to undergo MRI evaluation were excluded from the study. The independent variable analyzed was spectral power localized within the range of 0.0095 Hz to 0.021 Hz (endothelial frequencies) in the NIRS values [11, 12]. The dependent variables were the presence or absence of cerebral lesions identified through MRI evaluation. All subsequent analyses were performed using MATLAB R2023a.

Data processing and analysis

The data underwent the following processing and analysis steps: 1) Data Loading: Separate data files were prepared for each subgroup within the two cohort groups, delineating 24-hour periods of hypothermia treatment and the final 6 hours of warming, and imported into the MATLAB environment. 2) Detrending: Cerebral saturation values from NIRS were detrended within each subgroup to eliminate linear trends, ensuring that frequency analysis could focus on underlying physiological phenomena. 3) Power Spectral Density Estimation: Welch’s method for power spectral density (PSD) estimation was applied using the ‘pwelch’ function in MATLAB. Parameters included a window length of 500 samples, 60% overlap, FFT length of 500 points, and a frequency resolution of 0.033 Hz. The PSD was computed over the frequency range of interest (0.0095–0.021 Hz), targeting endothelial frequencies.

Frequency range selection

The frequency range of interest (0.0095–0.021 Hz) was selected based on physiological considerations associated with endothelial frequencies and the sampling rate of the NIRS device used. These values correspond to wave effects in blood flow attributed to nitrous oxide-mediated endothelial activity.

Energy Calculation: The energy within the specified frequency range (0.0095–0.021 Hz) was computed using the trapezoidal numerical integration method (trapz). This calculation represents the energy associated with the specific wave effects observed in blood flow.

Group Comparison: Calculated energy values were stored for each subgroup, enabling subsequent statistical comparison between groups with and without neuronal lesions.

Statistical analysis

Qualitative variables were summarized using absolute and relative frequencies, while quantitative variables were described using measures of central tendency and dispersion. The normal distribution assumption was assessed using the Shapiro-Wilk test. To compare energy values within the specified endothelial frequency range between the two main groups each day, a Mann-Whitney U test was employed [14]. To evaluate the clinical significance of differences in endothelial spectral power, effect sizes were calculated using Cohen’s d test [17]. Effect size interpretations were as follows: values below 0.20 indicate no significant effect, 0.21 to 0.49 suggest a small effect, 0.50 to 0.70 indicate a moderate effect, and values above 0.80 indicate a large effect.

Results

A total of 88 patients were enrolled in the study, with 67% (59/88) exhibiting normal MRI findings (Fig. 1). Among them, 53% (47/88) were male, 86% (76/88) had a gestational age greater than 37 weeks according to the Ballard scale, and 81% (71/88) had a birth weight exceeding 2500 g (Table 1). Patients diagnosed with MRI-detected anomalies showed a mean pH 0.1 lower compared to those without brain lesions (6.9 vs. 7.0; p-value = 0.02). The mean rScO2 values were 80% on Day 1 (24 h), 82% on Day 2 (48 h), 81% on Day 3 (72 h), and 81% during the rewarming period (6 h). Regarding the change in rScO2 between two groups, there was no change in the value during treatment (p > 0.05).

Fig. 1

Study flow chart. Notes: ICU: Intensive Critical Unit; MRI: magnetic resonance imaging; NIRS: Near-Infrared Spectroscopy

Table 1 Clinical characteristics of the newborns

Analysis of NIRS data over multiple days revealed a significant difference in the mean area under the power spectral density curve between groups with and without neuronal lesions (Fig. 2). Although the average energy curves displayed noticeable differences, the box plots illustrating the same data indicated trends that did not achieve statistical significance (Fig. 3a-d).

Fig. 2

Mean energy of Near-Infrared Spectroscopy. Notes: Mean energy of Near-Infrared Spectroscopy (NIRS) value, per group, over the days of hypothermia treatment and subsequent warming phase. Both cohorts begin around the same value at instance 0. The small effect size observable with Cohen’s d is represented here as well on days 2 and 3

Fig. 3

Comparison of endothelial energy according to neurological injury in cerebral magnetic resonance imaging, represented by box plots Notes: A: Day 1 comparison; B: Day 2 comparison; C: Day 3 comparison; D: Day 4 comparison

Cohen’s d effect size was computed to assess the magnitude of difference between the means of both groups. The results of Cohen’s d indicated a small effect size in endothelial frequency between groups based on cerebral MRI findings (Table 2), with Cohen’s d values for Day 2 = 0.2351 and for Day 3 = 0.2325. This subtle distinction is more evident in the Gardner-Altman plots (Fig. 4a-d) than in the box and whisker plots.

Table 2 P-Values and Cohen’s d value per time period

Fig. 4

Comparison of endothelial energy according to neurological injury in cerebral magnetic resonance imaging, represented by Gardner-Altman plots. Notes: A: Day 1 comparison; B: Day 2 comparison; C: Day 3 comparison; D: Day 4 comparison

Discussion

We conducted a study to examine the relationship between endothelial frequency within a specific spectral band, rScO2 values measured via NIRS, and cerebral MRI abnormalities in neonates receiving hypothermia treatment for moderate to severe perinatal asphyxia. The analysis of effect size using Cohen’s d revealed a small standardized difference between the means of the two groups. This suggests that NIRS could be a valuable tool for monitoring cerebral autoregulation in neonates affected by perinatal asphyxia. The spectral power parameter in the working frequency band of the endothelium appears to be a promising biomarker that can be used in real-time based on neonatal cerebral NIRS measurements. This can assess over time the frequency of the endothelial component of cerebral blood flow control.

While no statistically significant differences were observed between the two groups across the four time periods, a detailed examination using Cohen’s d revealed a small effect size difference between them, particularly evident on days 2 and 3, coinciding with the reperfusion phase. This observation may be attributed to the varying severity of asphyxia and encephalopathy, which likely influenced the measured brain values using NIRS in newborns undergoing therapeutic hypothermia. Specifically, newborns with perinatal asphyxia exhibited distinct brain value differences compared to those who developed brain injuries or remained unaffected, particularly on the second and third days of monitoring. Our analysis focused on NIRS brain data within the frequency range of 0.0095 Hz to 0.021 Hz, known to correlate with nitric oxide-induced endothelial frequency.

Initially, both groups showed similar energetic signatures, suggesting comparable baselines. However, as treatment progressed, the group with neuronal injuries exhibited notable peaks in high-energy values, a trend less pronounced in the group without injuries. This increase in energy among the group with neuronal injuries may indicate heightened neuronal activity or increased metabolic demand, potentially in response to the injury. These peaks offer critical insights into the ongoing neural consequences of the injury over time.

The theory of hypoperfusion-hyperperfusion is crucial to our findings, as constant monitoring via NIRS allows for detecting changes in cerebral oxygenation [18,19,20]. During perinatal asphyxia, hypoperfusion leads to a decrease in oxygen and nutrient supply to the brain, potentially resulting in irreversible neuronal damage if not quickly reversed [20, 21]. On the other hand, the hyperperfusion phase serves as a compensatory mechanism to restore blood flow to the brain. However, this can lead to increased intracranial pressure and disruption of the blood-brain barrier, causing cerebral edema. Therefore, our findings highlight important results for quickly identifying hypoperfusion and hyperperfusion phases, which could be used to optimize therapeutic interventions, maintain adequate cerebral oxygenation, and prevent additional complications [22, 23].

We hypothesize that the modest effect size observed between the two patient groups may be attributed to variations in endothelial frequency. These frequencies, influenced by nitric oxide (NO), a pivotal molecule in vasodilation, could plausibly be elevated due to increased NO release following perinatal asphyxia [11, 12]. The pathophysiology of brain hemorrhages resulting from perinatal asphyxia has been previously detailed. Evidence includes the presence of NMDA receptors in brain endothelial cells responsive to heightened glutamate release, which triggers increased NO synthesis during phases of energy depletion in perinatal asphyxia [8,9,10].

Current tools such as MRI and electroencephalogram often fail to predict early brain damage, typically detecting changes only after injury onset [24, 25]. NIRS offers a reliable, real-time, and non-invasive method for continuous monitoring of rSO2 and blood flow alterations, serving as an early biomarker for organic injury [26,27,28]. It provides an accurate assessment of early changes in tissue oxygenation and blood flow, particularly in cases of perinatal asphyxia [29, 30]. While some studies have demonstrated NIRS’s efficacy in mitigating cerebral hypoxia and detecting neurological injuries [31], others have not established a definitive association between NIRS data and neurological damage beforehand [32].

Limitantions

Our study is constrained by a small sample size, which limited our ability to detect subtle differences between groups. Methodological limitations, including the absence of a control group and variability in the timing of NIRS measurements, may have introduced confounding variables. However, the observed peaks in endothelial frequency align consistently with current medical literature findings. The inclusion of the 6-hour warming period in our analysis after day 3 may have constrained our ability to precisely assess the effects of hypothermia or warming on our results. The absence of changes on MRI during the first week of life in these patients can be attributed to the rapid temporal evolution of brain damage, imaging limitations, and the impact of hypothermia intervention. Thus, continuous patient follow-up is crucial.

Furthermore, our study did not evaluate vascular endothelial injury in children relative to NIRS data, nor did it compare these findings with physiological data (such as heart rate, blood pressure) or echography commonly used in clinical practice. Future research should prioritize larger sample sizes to enhance statistical power and deepen understanding of endothelial frequency measured by NIRS in perinatal asphyxia. Standardizing these variables across studies will be essential for achieving more reliable and generalizable results. Additionally, future investigations should explore the association between endothelial frequency measured by NIRS in perinatal asphyxia and clinical outcomes such as length of hospital stay, neurological status at discharge, and maternal risk factors.

Conclusion

NIRS could be a valuable tool for monitoring cerebral autoregulation in neonates affected by perinatal asphyxia, emphasizing the added value of calculating endothelial frequency or energy on rScO2 measured by NIRS using the original frequency of the INVOS™ device, 0.13 Hz. Analysis of NIRS data over several days revealed a significant difference in the mean area under the power spectral density curve between the groups with and without neuronal lesions assessed by MRI. Furthermore, a more pronounced difference in endothelial frequency was observed on days 2 and 3, coinciding with the reperfusion phase. This suggests that reperfusion injury may be more severe in the group with abnormalities in MRI.