Introduction

The ability of the cerebral vasculature to regulate its resistance in response to both local and systemic factors is defined as cerebrovascular reactivity (CR) and is aimed at maintaining adequate cerebral blood flow (CBF) to meet cerebral metabolic demand. Neonatal CR was first studied using radioactive tracer methods and Doppler sonography. The application of non-invasive and operator-independent near-infrared spectroscopy (NIRS) technique for cerebral regional tissue oxygen saturation (rStO2) monitoring in neonatal settings has shed further light on CR in several physiological and pathological conditions. Nevertheless, the methodological heterogeneity, along with the observational nature and the small sample sizes of current reports, hinders a routine application of CR monitoring for neonatal neuroprotection. On these premises, we aim to provide an updated review of NIRS-based evidence on neonatal CR and to identify critical issues and gaps that should be addressed with targeted trials to implement CR monitoring in neonatal intensive care.

Physiological mechanisms of cerebrovascular reactivity

Pressure-flow autoregulation

The modulation of vascular tone in relation to intraluminal pressure is a leading mechanism of CBF regulation, mediated by the mechanoreceptor properties of smooth muscle cells lining cerebral arteries. In response to increased intraluminal pressure, membrane depolarisation and calcium-dependent vasoconstriction occur, while the opposite happens at low intraluminal pressure, resulting in cerebral vasodilation.1 The classic depiction of cerebral pressure-flow autoregulation is a sigmoidal curve (Fig. 1), with stable CBF over a range of cerebral perfusion pressure (CPP), of which a main determinant in neonates is arterial blood pressure (ABP). When ABP values fall below the lower limit or rise above the upper limit of autoregulatory capacity, pressure-passive circulation occurs, with potential ischaemic or haemorrhagic complications.2 Recent evidence from adults, however, has revised this classical view by showing a much shorter plateau which still has a gentle slope, indicating some degree of pressure-passive CBF.3,4 Notably, ‘cerebral autoregulation’ is often used to refer to ‘cerebrovascular reactivity’. However, although the pressure-flow autoregulation of CBF is frequently involved in the interplay between CR mechanisms, the two terms are not synonymous, and their interchangeable use may contribute to the heterogeneity seen in current literature.

Fig. 1: Relationship between pressure-flow, metabolic and biochemical regulation of cerebral blood flow (CBF).
figure 1

The main determinant of cerebral perfusion pressure (CPP, shown in the x-axis) in the neonatal brain is arterial blood pressure. The black line indicates pressure-flow regulation under physiological homoeostatic conditions (i.e., normal pH, PaCO2, PaO2, normoglycaemia, basal autonomic status). The red line illustrates changes in CBF and in its relationship with CPP under hypercapnia and other conditions determining cerebral vasodilation, which may reduce the vasodilatory reserve and shorten the autoregulatory plateau. The blue line illustrates the changes in CBF and in its relationship with CPP under hypocapnia or associated conditions, which may widen the autoregulatory plateau. The dashed black line indicates the effect of sympathetic activation on the pressure-flow regulation curve. The dotted vertical lines indicate the upper (ULA) and lower (LLA) limits of the autoregulatory plateau in different conditions.

Full size image

Biochemical factors

Partial arterial pressure of oxygen (PaO2) and carbon dioxide (PaCO2) are potent chemo-modulators of cerebral vasculature, independent of intravascular pressure. While hypoxia and hypercapnia exert a vasodilatory effect on CBF, hyperoxia and hypocapnia lead to cerebral vasoconstriction. Vascular reactivity to PaO2 and PaCO2 is mediated by H+/K+ homoeostasis, secondary to changes in perivascular pH.5,6,7 CR to PaO2, PaCO2 and ABP can functionally interact (Fig. 1). For example, if the vasodilator pathway has been activated during hypercapnia, the slope of the autoregulatory plateau increases, predisposing to larger CBF fluctuations even when ABP/CPP are within the normal range.8

Metabolic factors

Cerebral metabolic demands can influence cerebral perfusion, as supported by the evidence of CBF changes in relation to glucose availability. In preterm newborns, hypoglycaemia is associated with a significant compensatory increase in CBF;9 following the restoration of normoglycaemia, CBF gradually decreases.10 A recent systematic review11 has also reported a similar negative correlation between blood glucose levels and either cerebral rStO2 or haemoglobin concentration,12,13,14 used as a proxy for CBF, in term and preterm neonates.

Sympathetic nervous system

The cerebral vasculature has abundant adrenergic receptors and is under precise autonomic control.15 Sympathetic activation shifts the autoregulatory plateau towards higher CPP, thereby protecting the brain against hyperperfusion (see Fig. 1).16 The sympathetic system appears to play a greater role in CBF regulation in the perinatal period than later in life.17 The relative immaturity of nitric oxide (NO)-induced vasodilatory mechanisms during early development, the greater sensitivity to exogenous norepinephrine and the higher sympathetic nerve density in neonatal compared to adult pial arteries may contribute to this finding.18

Functional activation

The CBF response to neuronal activation is referred to as neurovascular coupling. This response is mediated by the neurovascular unit (Fig. 2), which represents an interactive network of cerebral vessels, vascular cells (pericytes, smooth muscle and endothelial cells), glia (astrocytes and microglia) and perivascular neurons. Upon neuronal activation, vasoactive substances such as prostaglandins, NO and adenosine are released from both neurons and astrocytes, leading to the modulation of vascular smooth muscle.19 Studies using functional NIRS have provided insight into the neurovascular coupling response in newborns. In the adult brain, the classic ‘positive’ response is characterised by an increase in oxygenated haemoglobin (O2Hb) and a decrease in deoxygenated haemoglobin (HHb). In neonates, however, variable cerebral haemodynamic patterns have been reported in response to neuronal activation, including a ‘negative’ response with local O2Hb reduction, which may indicate that the oxygen consumption triggered by neuronal activity transiently outpaces the concomitant CBF increase.20,21,22,23,24 This variability may be due to developmental changes in the capacity of cerebral vasculature to produce functional hyperaemia in the low-density capillary bed of the neonatal brain.25,26

Fig. 2: Illustration of the neurovascular unit and of the complex interplay between arteriole, interneuron, astrocyte, and neuron-secreted factors on smooth muscle cells, which regulate the vascular arteriole diameter.
figure 2

CO carbon monoxide, SMC smooth muscle cell, EC endothelial cell, Ach acetylcholine, PGs prostaglandins, EETs epoxyeicosatrienoic acids, VIP vasoactive intestinal polypeptide, CGRP calcitonin gene-related peptide, NA noradrenaline, DA dopamine, NPY neuropeptide Y, EDHF endothelium-derived hyperpolarizing factor, ET endothelin, GABA γ-aminobutyric acid, ROS reactive oxygen species. Reproduced with permission from Brew et al.38.

Full size image

NIRS-based assessment of cerebrovascular reactivity: surrogate signals and neonatal applications

The assessment of CBF fluctuations in response to physiological changes or stimuli is fundamental to evaluate the integrity of CR. This assessment evaluates the relationship between the input and the output (i.e., CBF) signals,27 and can be classified into static, which includes quantitative methods performing point measurements of CBF (e.g.,133Xe clearance), or dynamic.28,29 While in adults, transcranial Doppler sonography (TCD) can be used for dynamic CR monitoring, the small size of neonatal vessels hinders an accurate measurement of the vessel diameter, which is necessary to calculate blood flow within a specific artery.30 The small vessel size also causes frequent loss of signal, limiting the application of neonatal TCD to static CBF assessments.31,32,33

NIRS exploits the relative transparency of near-infrared light (700–950 nm) in biological tissue and the oxygen-dependent absorption of haemoglobin at different wavelengths to measure rStO2, derived from the changes in concentration of O2Hb and HHb within the vascular beds (Fig. 3). Fluctuations in cerebral rStO2 can reflect changes in CBF if other determinants of cerebral metabolism and oxygen delivery (i.e., arterial oxygen saturation, haemoglobin concentration, arterial–venous volume ratio, fractional inspired oxygen, tissue oxygen diffusivity) remain relatively constant.34 NIRS monitoring can be performed non-invasively and continuously for relatively prolonged periods; hence, cerebral rStO2 has been used as a surrogate for dynamic CBF monitoring, mainly for slow CBF changes. The changes in total cerebral haemoglobin concentration (∆tHb) or tissue haemoglobin index (THI), as a sum of ∆O2Hb and ∆HHb or directly measured using the isosbestic 805 nm wavelength, have been described as surrogate measurements of cerebral blood volume.12,35,36,37

Fig. 3: Near-infrared spectroscopy: absorption spectra and photon trajectory.
figure 3

a Absorption spectra for oxygenated haemoglobin (O2Hb), deoxygenated haemoglobin (HHb), and water. b Example of photon trajectory through biological tissue. S source, D detector, ρ source-detector distance. The shaded part highlights the area crossed by the photon. Adapted with permission from Martini et al.172.

Full size image

The interaction between CBF and CPP is mediated by cerebrovascular resistance as follows, using an analogy of Ohm’s law:

$${{{{{\rm{CBF}}}}}} = \frac{{{{{{{\rm{CPP}}}}}}}}{{{{{{{\rm{Cerebrovascular}}}}}}\,{{{{{\rm{resistance}}}}}}}}$$

Cerebrovascular resistance is determined by the vascular tone of the arterial smooth muscle cells. During brain development, the muscularis layer of the extra-striatal arterioles is initially limited to the pial vessels and superficial penetrators; consequently, in preterm infants, cerebral vasoreactivity occurs predominantly in the superficial and peripheral parenchyma of the brain.38

CPP is dependent on ABP and intracranial pressure. As intracranial pressure is assumed stable in neonates because of the open cranial sutures, the slow waves of ABP can presumably be used as a surrogate for low-frequency CPP changes. With high-frequency CPP changes and oscillations (>0.20 Hz), the autoregulatory processes become less able to stabilise CBF in the face of changing CPP.39 Therefore, these fast CPP oscillations are passed along unimpeded into CBF oscillations. In contrast, slower frequency oscillations (<0.20 Hz, but most effectively <0.05 Hz) can be counteracted by the cerebral arterioles and are dampened.40,41 Hence, continuous measurements of slow rhythmic oscillations in ABP and CBF have been used to assess the integrity of pressure-flow autoregulation36,42,43,44,45,46 and, in different neonatal cohorts, have also identified individual optimal ABP (ABPopt) ranges within which autoregulatory mechanisms are most effective, defined by the lowest values of the related CR coefficient indicating functional reactivity.47,48,49,50,51,52 Continuous ABP monitoring requires an indwelling arterial catheter53 and may not always be feasible. To date, evidence on the use of non-invasive continuous monitoring for blood pressure (e.g., beat-to-beat finger arterial devices) to assess neonatal pressure-flow autoregulation is limited,54,55 and the reliability of non-invasive blood pressure monitoring against invasive methods needs further improvements.56,57 Using blood pressure data measured non-invasively by arm cuffs at 15 min after birth, functional pressure-flow autoregulation was demonstrated in term neonates during the immediate postnatal transition, but not in preterm infants at this early phase.58

Being a direct determinant of cardiac output, heart rate (HR) can be considered a surrogate of systemic blood flow. Since continuous HR monitoring is non-invasive and universally available, HR can represent an alternative input signal to assess CR in response to systemic blood flow changes. A moving correlation coefficient between cerebral rStO2 and HR (TOHRx) has been proposed for CR monitoring in preterm infants, with rising values indicating loss of CR.47,48,59,60,61 Low or negative TOHRx values have been used to define ABPopt ranges during the first 24 h after birth.48 The recent neonatal application of non-invasive bedside devices for continuous cardiac output monitoring may facilitate the investigation of CR to systemic blood flow; current data, however, are limited. Two reports have found no association between cardiac output and cerebral rStO2 in term infants immediately after delivery62 or at different sleep positions.63

PaCO2 can also be used as an input to evaluate CBF-CO2 reactivity.8,64 End-tidal (etCO2) or transcutaneous CO2 (tCO2) monitoring allow continuous, non-invasive estimation of PaCO2. The relationship between tCO2 and THI in neonates has been evaluated by Dietz et al.,65 who documented a trend towards increased CO2 reactivity in healthy term neonates between days 1 and 4, whereas Aly et al.66 found that lower gestational age (GA), mechanical ventilation and increased PaCO2 were associated with stronger CO2 reactivity in the first week of life in preterm infants, with possible implications on the risk of brain injury. EtCO2 fluctuations have also been associated with concomitant changes in cerebral rStO2 and electroencephalographic brain activity.67 Other studies using punctual PaCO2 from blood gas analysis yielded variable results: in preterm infants, Kaiser et al. observed a progressive loss of pressure-flow autoregulation for PaCO2 ≥45 mmHg during the first week of life,8 whereas Hoffman et al. failed to demonstrate an overall association between PaCO2 and pressure-flow autoregulation during the transitional period.68 During the immediate postnatal period (i.e., 15 min after birth), Wolfsberger et al. observed that the vasodilative effect of PaCO2 on cerebral rStO2 was less pronounced in preterm compared to term neonates.69

Electroencephalographic techniques have been used to investigate CBF responses to neurophysiological changes. In preterm neonates, Roche-Labarbe et al. described that spontaneous bursts of electroencephalographic activity were coupled to a haemodynamic response characterised by an O2Hb decrease, followed by an increase and then a return to baseline.70 Tataranno et al. documented a decreased cerebral rStO2 and increased tissue oxygen extraction in extremely preterm infants with increasing intensity of spontaneous brain activity.71 Seizures are paroxysmic bursts of abnormal electrical brain activity. Available neonatal data during different types of seizures consistently report a reduction of cerebral rStO2,72,73,74 followed by a variable rStO2 increase.73,74 The observed rStO2 reduction was associated with an O2Hb decrease and an HHb increase mimicking a ‘negative functional response’ pattern,73 consistent with the elevated oxygen consumption associated with seizures which exceeds cerebral oxygen delivery.

Notably, CR assessment is only reliable if considerable variability in the input signal is present; if there is no variation, dependency cannot be determined. Some researchers have proposed adding more weight to epochs with high variability to correct for this factor.75,76,77 Furthermore, when compared to outer input signals such as those derived by electroencephalographic or electrocardiographic techniques, some NIRS instruments may have a lower sampling rate; this may affect the calculation accuracy of some CR metrics such as those for neurovascular coupling, especially for long-term monitoring periods.19,25,78

Proposed mathematical methodologies for CR assessment

Recent overviews, which also provide specific methodological details of the studies included, have confirmed the feasibility of examining CR in the neonatal population.79,80 To date, however, there is no clinical gold standard methodology for CR assessment; this is reflected in the numerous mathematical methods used for these purposes. These methodologies have mostly been applied to the investigation of pressure-flow autoregulation. Signal processing techniques for assessing neurovascular coupling have recently been reviewed by Hendrikx et al.19 and will not be addressed in this review.

Invasively measured ABP and cerebral rStO2 as surrogates for CPP and CBF, respectively, are included in different mathematical models to address pressure-flow autoregulation. The lack of standardisation on different levels in the quantification of pressure-flow autoregulation (Table 1) hinders the reproducibility of studies that have related the analyses with clinical outcomes.

Table 1 Variability on different levels in the quantification of cerebrovascular reactivity in infants.
Full size table

To study cerebral autoregulation, a synchronised capture of the physiological signals from multiple monitoring systems into a single aggregated file is required. Several available data capture platforms compatible with current NIRS monitors have been recently summarised by Vesoulis et al.81 A major step in pre-processing of the acquired data is the removal of artefacts; although this essential component is often performed manually, ideally a reliable automated approach would speed up this stage, avoid any biases, and enable real-time analysis. As two key assumptions for the use of rStO2 as a surrogate for CBF are stable SpO2 and cerebral metabolic rate (CMRO2), correction for SpO2 and CMRO2 would ideally also be necessary. The total cerebral haemoglobin concentration has been proposed by Grubb et al. as a surrogate for CBF independent of the CMRO2.82

Pressure-flow autoregulation analysis, making use of the spontaneous oscillations in cerebral oxygenation and ABP, has been explored both in the time domain and frequency domain by correlation and coherence analysis, respectively. These approaches assume that the physiological parameters are considered stationary signals.

Time-domain methods

The time-based method of Pearson correlation (r) is the most straightforward approach. The correlation coefficient between rStO2 and ABP is used as a cut-off value to assess the presence or absence of pressure-flow autoregulation (the closer to 0, the higher the autoregulatory capacity); significant positive correlation is then considered to represent pressure-passive cerebral circulation in that epoch.83,84,85,86,87,88 A question remains whether the cerebrovascular transit time (i.e., the time needed for cerebral rStO2 to fully respond to a CBF change) is considered. In this regard, the moving correlation coefficient between cerebral rStO2 and ABP, which has been validated in hypotensive piglets and has been shown to correlate with TCD measurements of pressure-flow autoregulation in adult patients,44 has been largely used in neonatal clinical observational studies.85,89,90,91,92 The percentage of epochs with impaired pressure-flow autoregulation (defined as r above a predefined threshold), the strength of correlation, but also the amount of r variability during measurement can be assessed.93 These principles have been applied to other NIRS-based parameters to study pressure-flow autoregulation. For example, the moving correlation between THI and ABP has been used to determine individualised ABPopt.49,50,52,94,95 Similarly, the Pearson correlation between THI and tCO2 has been used to investigate CR to CO2 in the neonatal population.65,66

Frequency-domain methods

Coherence analysis has been applied to explore the relationship between NIRS measurements and ABP changes in the frequency domain.96 Notably, according to the Nyquist theorem, the sampling frequency of the collected data needs to be at least twice the highest frequency of interest in the signal. Similar to time-domain methods, a predefined threshold is used to define impaired pressure-flow autoregulation (the closer to 0, the higher the autoregulatory capacity). To correct for different epoch lengths and sample frequencies, Monte-Carlo simulations are performed to identify significant coherence in the given data segments;75,97 adaptations to the basic model of coherence are also made.76,98 Coherence describes the impairment of pressure-flow autoregulation quantitatively. The degree of dependence between two variables (e.g., ABP and rStO2) is measured (no units), and by performing transfer function analysis, the gain and phase are calculated. The gain estimates the degree of this impairment by describing the magnitude of change in the output signal (e.g., rStO2) in relation to a unity change in the input signal (e.g., ABP) at a given frequency. However, causality, i.e., what signal induces changes in the other signal, cannot be established by these methods. Wong et al.45 first used coherence and gain to assess rStO2-derived pressure-flow autoregulation in sick infants. This method was later validated by Hahn et al. in piglets.99 Transfer function with logarithmic transformation of the gain coefficient to provide the amplitude of the dampening response is described by Vesoulis et al.100 The phase describes the time delay between coherent oscillations of the two signals and can be seen as an additional factor in pressure-flow autoregulation analyses. Pressure-flow autoregulation can be studied in different frequency bands, ranging from ultra-low to high-frequency bands. Low-frequency bands correspond to slow oscillations in ABP and rStO2. Slow and prolonged periods of hypotension or hypertension of higher magnitudes are considered to be more injurious than fast and transient ABP changes of small magnitudes (‘high-pass filter’ principle of pressure-flow autoregulation).101

Time-frequency methods

Since biological signals tend to be non-linear and non-stationary, alternative methods based on linear equations, such as wavelet cross-correlation, have also been described.102 This approach incorporates a time element to frequency analysis. It makes no assumption about the stationarity of input signals, providing a framework for the analysis of non-stationary effects in cerebral haemodynamics. Another time-frequency method, defined as bivariate auto-regressive coherence (BiAR-COH), has been proposed by Riera et al.98 The main difference between the BiAR-COH and standard coherence methods is that the former demands both temporal and frequency dependence, whereas coherence only evaluates frequency dependence. This key feature discriminates changes in the two signals that are closely related in time from those that are not time-related and, therefore, have no mutual dependence. The same authors introduced the partial directed coherence (PDC) method to address the condition of directionality,76 so that the system is forced to consider only those events in which changes in CPP induce changes in CBF. Therefore, this approach not only analyses pressure-flow autoregulation, but also infers causality.76

Impaired cerebrovascular reactivity and neonatal outcomes

Impaired CR is a marker of disease severity and adverse outcome, both in term85,103,104,105,106,107,108,109,110 and preterm infants.60,76,83,84,91,98,111,112,113 The patient populations most often studied are neonates with hypoxic-ischaemic encephalopathy (HIE) and preterm infants during the postnatal transition, which will be discussed hereafter; specific study details on these populations are available in other reviews.80,114,115 Insights on other neonatal conditions are also available in recent targeted reviews.2,116

There is longstanding evidence of the association between perinatal hypoxia-ischaemia and altered CR, resulting in hypoxia-ischaemia-reperfusion injury.117,118,119 Cerebral hypoxia-ischaemia induces compensatory overproduction of NO,120 leading to persistent cerebral vasodilation that may disrupt the vessels’ autoregulatory capacities (the so-called vasoparalysis).115 Upon reperfusion, the disrupted CR results in a significant increase in CBF with no change in CMRO2.121 In animal models of hypoxia-ischaemia, this cerebral hyperaemia has been associated with histopathological evidence of brain damage.106,115 Consistent evidence of post-asphyxial cerebral hyperaemia has been obtained from human neonates with HIE using NIRS.103,115,122,123 This CBF increase positively correlates with the severity of the ischaemic hit, and is accompanied by an impaired CR to acute ABP and CO2 changes.119 The burden of cerebral hyperaemia and of CR impairment was greater in infants who later developed brain injury on MRI or showed poorer neurodevelopmental outcomes.85,103,104,105,106,107,108,110,124 Using wavelet coherence between NIRS and electroencephalographic signals, Das et al.109 recently showed that HIE infants with MRI brain abnormalities had poorer neurovascular coupling during the first 24 h compared to those with normal neuroimaging. Notably, the cerebral rStO2- electroencephalographic coherence during this early period was superior to the Sarnat score in predicting abnormal brain MRI.

Given the altered CR following perinatal asphyxia, defining ABPopt ranges of pressure-flow autoregulation is particularly important. Prolonged ABP deviations below ABPopt during the hypothermic treatment have been associated with increased MRI abnormalities in both the deep grey matter51,52 and white matter.50 Infants with more prolonged deviations below ABPopt also had greater motor and cognitive impairments at 21–32 months.49 Based on this evidence, monitoring pressure-flow autoregulation to establish and target ABPopt ranges with a tailored haemodynamic management (e.g., adjusting pharmacological cardiovascular support) is potentially an adjunctive neuroprotective strategy in neonatal HIE.

In the preterm population, the cardiovascular changes occurring during the first 72 h after birth, which define the so-called transitional period, are often associated with a significant haemodynamic instability together with CBF fluctuations that may cause disruption of the germinal matrix endothelium, increased intravascular pressure and125 result in intraventricular haemorrhage (IVH).126 Accordingly, specific NIRS patterns of cerebral haemodynamics have been reported in infants developing this complication, characterised by reduced cerebral rStO2 during the first 24 h60,91,127,128 followed by a transient increase,60,127 which suggest the hypoperfusion-hyperaemia alternance.

Numerous NIRS studies have shown impaired CR in preterm infants developing IVH.60,61,76,83,84,91,98,111,112,129 Severe IVH development has been associated with a significantly higher time burden of a pressure-passive circulation on day 2, which was also the median age at IVH detection.84 An independent association between high-magnitude cerebral pressure-flow passivity in the low-frequency range and IVH development has also been found.112 With regard to ABPopt, infants with greater deviations below or above ABPopt ranges had higher IVH rates compared to those whose ABP laid close to optimal ranges during the transition.47,48 Significantly higher TOHRx values were also reported in infants who developed IVH compared to those who did not.60,61 In contrast, one study observed lower amplitudes of cross-correlation, semblance and gain between cerebral rStO2 and HR, measured by wavelet analysis, in a small number of preterm infants developing IVH/pulmonary haemorrhage compared to those who did not, possibly reflecting the extreme haemodynamic instability associated with haemorrhages.130

The link between CR indices and systemic blood flow, as well as their relationship to IVH has also been reported. Low superior vena cava (SVC) flow has been pathophysiologically linked to intracranial bleeding in preterm neonates.125,131 Low SVC flow infants also had higher BiAR-COH and PDC, indicating impaired pressure-flow autoregulation, and were more prone to developing severe IVH.76,98 A negative correlation between COx and left ventricular output at 24 h of life in a cohort of neonates that developed IVH has also been reported.91 These findings align with the report of lower left ventricular output and cerebral rStO2 within the first day of life in infants who later developed IVH,127 and may reflect CR impairment associated with fluctuations of systemic blood flow before or around the time of bleeding. The association between other factors influencing CR, such as PaCO2 and IVH,132 warrants further targeted investigations.

Dopamine and impaired pressure-flow autoregulation: causality or association?

Circulatory failure and the need for cardiovascular support using inotropic/vasopressor medications have been associated with impaired cerebral pressure-flow autoregulation. Due to its alpha-, beta-adrenergic and dopaminergic activity, dopamine has long been used as a vasopressor-inotrope to support circulation and maintain adequate perfusion of critical organs, such as the brain.133 Dopamine’s impact on cerebral pressure-flow autoregulation, secondary to its effects on vascular tone or inotropism, remains inconclusive. Several NIRS studies reported that, in hypotensive preterm infants treated with dopamine, both CBF and ABP increase together,134,135,136,137 indicating that cerebral pressure-passive circulation or a small positive slope of the autoregulatory plateau92,134 may persist over a range of ABP. Moreover, time periods with impaired pressure-flow autoregulation in hypotensive infants were reported to increase with dopamine treatment in a dose-dependent fashion.138,139,140 However, the methodology and design of these studies could not address whether dopamine directly impaired the autoregulatory capacity or was merely an indicator of illness. In contrast, when dopamine effects were investigated in newborn piglets with induced hypotension, an improvement of cerebral pressure-flow autoregulation at low ABP, proportional to dopamine dose, was observed.141 Recent data from the HIP trial, where hypotensive preterm neonates were randomly assigned to either dopamine or placebo infusion,142 reported a significantly impaired pressure-flow autoregulatory capacity in hypotensive compared to normotensive infants, but not in relation to dopamine treatment.111 This, however, was not an adequately powered study; therefore, no firm conclusions can be drawn. Although currently available evidence suggests that autoregulatory capacity may be impaired by hypotension and its underlying causes rather than by dopamine treatment, larger targeted studies are needed to validate these findings and to define the complex relationship existing between these factors. In addition, data on the impact of other inotropic and vasoactive medications on pressure-flow autoregulation in preterm infants remain very limited and scarce.

Discussion

Current evidence on neonatal CR is based on many observational studies which, over the years, have pointed towards physiological associations between altered CR and an increased risk of brain injury. To move forward towards a preventive and therapeutic approach in neonatal CR research, multiple aspects need to be considered.

First, most of the available neonatal literature is derived from single-centred studies, based on small and heterogeneous cohorts which are under-powered and potentially involve biases and confounders (e.g., different types of brain injury, lack of PaCO2 data, use of different monitoring windows etc.), hindering comparison between studies to quantify the independent impact of CR impairment.

Second, an agreement on the multiple methods proposed for CR estimation is needed. The heterogeneity in the monitoring devices, recording methods, pre-processing steps and mathematical models applied for CR assessment hampers the identification of possible gold standard methodology that shows the best sensitivity and specificity for outcome prediction. The comprehensive integration of different multimodality monitoring signals, as well as the adoption of the Findable, Accessible, Interoperable, and Reusable approach143 for an open database with high-resolution data on cerebral rStO2, ABP, HR and standardised outcomes would facilitate the comparison and evaluation of their performances in predicting outcome.

Third, if an acceptable and reliable methodology for CR assessment is agreed upon, continuous real-time monitoring of CR may allow a personalised approach to neonatal intensive care, aimed at optimising CR and reducing neonatal brain injury. Continuous CR assessment combining rStO2, ABP and tCO2 monitoring should ideally be measured in at-risk neonates. However, such a comprehensive and multi-modal monitoring may be technically challenging (e.g., skin frailty, signal noise, availability of intra-arterial catheter), especially in extremely preterm neonates. In this regard, the validation of CR measurements using less invasive parameters (e.g., HR) as input signals for outcome prediction would facilitate CR monitoring and application in neonatal settings. Further improvements in the design of NIRS probes to suit the fragile skin of extremely preterm infants may also support long-term CR monitoring.

Some of the above aspects have been tackled by adult neurointensive care research groups, especially following traumatic brain injury (TBI), sepsis and stroke.144,145,146 In this setting, a Delphi consensus stating that CR status is uniquely dependent on an individual patient at any specific time, excluding the use of universal and absolute thresholds for CPP, was formalised, and a research agenda was proposed to establish and validate CR assessment methods against outcome, together with prospective safety, feasibility and efficacy studies to investigate the application of CR-guided clinical management.147 Of note, a study in adults with TBI, investigating the feasibility of automated assessment of optimal CPP based on individualised CR monitoring, is currently recruiting.148,149,150

A consensus approach similar to the abovementioned one may help to establish a future research agenda including collaborative clinical NIRS-based trials, targeted to address the current questions on neonatal CR listed in Table 2. These feasibility trials may represent the first step towards a randomised controlled trial based on continuous real-time CR monitoring, using dedicated software and aimed to assess whether a proposed treatment strategy (i.e., actively maintaining ABP within an optimal CR range) may improve neurological and neurodevelopmental outcomes. The goals are to identify infants with CR impairment predictive of brain injury, followed by interventional trials to prevent or correct the CR impairment for neonatal neuroprotection. Indeed, the SafeBoosC trials have pioneered the multi-centred approach in assessing the benefit of clinical interventions to optimise cerebral rStO2 in preterm neonates.151,152,153,154,155 In this regard, the formation of an international, multicentre working group to coordinate clinical trials and collated data management would support achieving these research targets.

Table 2 Proposal for a series of feasibility trials aimed at addressing the currently open questions on neonatal cerebrovascular reactivity (CR).
Full size table