Optimizing Cerebral Autoregulation May Decrease Neonatal Regional Hypoxic-Ischemic Brain Injury

Therapeutic hypothermia offers only partial neuroprotection for neonates with hypoxic-ischemic encephalopathy (HIE). Even with hypothermia, neurologic disabilities persist in 35-55% of survivors at 6-7 years [1, 2]. Ensuring robust cerebrovascular autoregulation might improve outcomes. In pilot studies, we showed that blood pressure deviations from the optimal mean arterial blood pressure (MAP_OPT) at which autoregulation is most robust are associated with brain injury on MRI [3, 4] and 2-year neurodevelopmental outcomes [5] after HIE. However, these studies were too small to control for confounders that affect autoregulation, such as the arterial partial pressure of carbon dioxide (PaCO₂) [6] and vasopressors [7, 8]. Here, we studied a larger cohort to evaluate autoregulation independent of PaCO₂, vasopressors, perinatal insult severity, and seizures.

Autoregulation can be measured indirectly with near-infrared spectroscopy (NIRS) [3, 4, 5, 9, 10]. NIRS measures deoxy- and oxyhemoglobin optical densities from the regional cortex, and the sum of these densities - the relative total tissue hemoglobin (rTHb) - is a surrogate measure of regional cerebral blood volume. NIRS rTHb measurements reflect the fluctuations in cerebral blood volume that occur during autoregulatory vasoactive responses to changes in arterial blood pressure. The hemoglobin volume index (HVx) is calculated as the correlation coefficient between rTHb and MAP [11].

Because NIRS interrogation focuses on the frontal cortex, whether autoregulation measurements can gauge injury risk in other anatomic regions is unclear and must be studied to determine the relevance for HIE. Our objective was to measure autoregulation with HVx and brain injury with MRI in neonates who received therapeutic hypothermia for HIE to test the hypotheses that (1) HVx identifies the blood pressure range with optimal autoregulatory vasoreactivity; (2) autoregulation affects brain injury independent of perinatal insult severity; and (3) these relationships vary by anatomic region.

This observational, prospective study was approved by the Johns Hopkins University School of Medicine Institutional Review Board. Written informed consents were obtained until May 2013, when NIRS became standard of care for HIE treatment at our hospital. After that, we were granted a waiver of consent. We sequentially screened all neonates admitted for HIE between September 2010 and July 2015 using HIE diagnostic criteria from the NICHD Neonatal Research Network's trial of therapeutic hypothermia in HIE [12] and the eligibility criteria from our prior studies [3, 4, 5]. The neonates reported here include those from our pilot studies [3, 4, 5] with an additional 36 neonates. We excluded neonates who received extracorporeal membrane support (ECMO).

Neonates received therapeutic hypothermia for 72 h per previously described protocols [3, 4, 5]. We examined HVx during hypothermia, rewarming, and the first 6 h of normothermia. Clinicians determined the neonates' blood pressure goals and could view the NIRS regional cerebral oximetry (rSO₂) but were blinded to HVx. When necessary, dopamine was initiated followed by dobutamine and epinephrine. Seizures were diagnosed by electroencephalogram and treated with phenobarbital. Hydrocortisone was initiated for adrenal suppression or persistent hypotension refractory to vasopressors when needed.

Clinical data were obtained from the medical record by investigators (R.C.-V. and M.O'C.) blinded to the autoregulation and brain MRI data. PaCO₂ levels were classified into four categories [13]: (1) all PaCO₂ levels 35-45 mm Hg; (2) some <35 mm Hg but none >45 mm Hg; (3) none <35 mm Hg but some >45 mm Hg; and (4) some <35 mm Hg and some >45 mm Hg. An investigator blinded to the autoregulation and MRI data (R.C.-V.) created a perinatal insult score to grade the birth asphyxia severity (Table 1)

Neonates received bilateral forehead NIRS monitoring with an INVOS 5100 (Medtronic, Minneapolis, MN, USA). HVx was calculated from the NIRS and arterial blood pressure signals as previously described during hypothermia, rewarming (defined as rectal temperature 34.1-36.5°C), and the first 6 h of normothermia [3, 4, 5, 11]. Briefly, HVx was determined from deoxygenated and oxygenated hemoglobin optical densities [11], which decreases the sensitivity of this index to changes in systemic oxygenation when compared to metrics based solely on oxyhemoglobin. We synchronously sampled MAP from the arterial blood pressure catheter and NIRS signals using ICM+ software (Cambridge Enterprises, Cambridge, UK). HVx is calculated by a continuous, moving correlation coefficient between MAP and the NIRS rTHb (rTHb = 1 - optical density_A*50), a surrogate measure of cerebral blood volume [11], after removal of signal artifacts and high-frequency waves from respiration and pulse [3, 4, 5, 14]. When autoregulatory vasoreactivity is functional, rTHb and MAP have negative or near-zero correlation, and HVx is negative or near zero. During periods of dysfunctional autoregulation, rTHb and MAP become positively correlated, and HVx approaches +1 with progressive impairments in autoregulation [11]. We verified that neonates did not have unilateral intracranial lesions and averaged the right and left HVx values for sorting into 5-mm Hg bins of MAP. The most negative HVx identified the MAP_OPT at which autoregulation was most robust with maximal vasoreactivity to changes in MAP during hypothermia, rewarming, and normothermia. The neonate was coded as “unidentifiable MAP_OPT” if a nadir in HVx could not be identified [3, 4, 5]. An investigator (J.K.L.) blinded to outcomes and medical histories identified the MAP_OPT values with corroboration by additional investigators (F.J.N. and M.M.G.). Blood pressure was analyzed as the (1) maximal blood pressure deviation below or above MAP_OPT; (2) duration with blood pressure below, within, or above MAP_OPT analyzed as a percentage of the autoregulation monitoring period; and (3) area under the curve (AUC; min × mm Hg/h) to combine time (min) spent with blood pressure below MAP_OPT and blood pressure deviation (mm Hg) below MAP_OPT normalized for the monitoring duration (h) in each period [4, 15]. We also calculated the percentage of the hypothermia, rewarming, and normothermia periods that neonates spent with MAP below gestational age in weeks +5 mm Hg, a common clinical guide for neonatal hemodynamic goals [16].

Neonates received brain MRIs after completing hypothermia and autoregulation monitoring according to our published protocol [3]. MRI studies were performed on a 1.5-T clinical scanner (Avanto; Siemens, Erlangen, Germany) by using a standard neonatal 8-channel head coil under natural sleep (without general anesthesia). Standard neonatal brain MRI with sagittal T1-weighted, axial T2-weighted, and axial susceptibility-weighted imaging was obtained. A single-shot, spin echo, echo planar axial diffusion tensor imaging sequence with diffusion gradients along 20 noncollinear directions was acquired. Trace of diffusion images and apparent diffusion coefficient (ADC) maps were automatically calculated by the vendor-specific software in the MRI scanner. Image analysis was performed on the PACS (picture archiving and communication system) workstation [4]. Two experienced pediatric neuroradiologists (A.T. and T.A.G.M.H.) categorized the brain injury as none, mild, moderate, or severe [3] in the paracentral gyri, white matter, posterior limb of the internal capsule (PLIC), putamen and globus pallidus, thalamus, and brain stem. These regions are associated with adverse neurologic outcomes from HIE [17, 18, 19, 20, 21]. The radiologists were blinded to the neonates' HVx, blood pressures, and clinical history.

Data were analyzed with Rv3.2 (Vienna, Austria) and presented as means with standard deviations (SD) or medians with interquartile ranges as appropriate. Associations between brain injury on MRI and the blood pressure autoregulation parameters were analyzed separately for each period (hypothermia, rewarming, or normothermia). Classifications of regional brain injury on MRI were analyzed for their associations with the predictors (MAP_OPT; blood pressure in relation to MAP_OPT [the percentage of time spent with MAP below, within, or above MAP_OPT; the maximal MAP deviation above or below MAP_OPT; and the AUC of MAP below MAP_OPT in each period]; the percentage of time spent with MAP below gestational age +5 mm Hg in each period; and the mean rSO₂ averaged between the right and left sides across each period) with ordered polytomous regression for proportional increase in injury [22]. Each analysis was adjusted for PaCO₂ category, presence of seizures, vasopressor treatment, and perinatal insult score. These covariates were selected due to their potential influence on cerebral autoregulation and neurologic injury [6, 7, 8, 23, 24, 25]. Relationships between perinatal insult score and brain injury were estimated with polytomous regression for ordered categorical outcomes. Finally, we examined the neonates' blood pressures in relation to MAP_OPT as they progressed through hypothermia, rewarming, and normothermia using a Pearson's correlation coefficient to compare the percentages of the autoregulation monitoring period with blood pressure below, within, or above MAP_OPT between hypothermia versus rewarming; hypothermia versus normothermia; and rewarming versus normothermia.

We screened 122 newborns with HIE. Forty-seven (39%) were ineligible for the study because of an unreliable arterial catheter blood pressure tracing (16), parents' refusal to consent to the study (9), transfer for ECMO (6), death before initiation of HVx monitoring (5), technical difficulties (5), inadequate resources (3), complex heart disease (1), coagulopathy (1), or parents' inability to speak English or Spanish (1). Therefore, 75 (61%) neonates with HIE were monitored with HVx. Eleven did not have diagnostic brain MRI for evaluation because of motion artifacts (7) or withdrawal of support before MRI acquisition (4).

The final sample size was 64 neonates (38 boys, 26 girls), and data were analyzed among those with an identifiable MAP_OPT (see below). All 64 neonates with brain MRI data had HVx monitoring during hypothermia (mean duration 46.5 h [SD, 19.8]), 59 during rewarming (duration 6.3 h [SD, 2.6]), and 57 during normothermia (duration 5.3 h [SD, 1.6]). HVx monitoring was stopped after hypothermia in 2 neonates because of technical problems and in 3 who were transferred to another unit; these 3 neonates underwent the therapeutic hypothermia protocol, did not receive ECMO, and had brain MRI data analyzed in the study. Two neonates had HVx monitoring stopped after rewarming because of early removal of the NIRS (1) or arterial blood pressure catheter (1). Patient descriptions are listed in Table 2. The hemoglobin levels were 15.5 g/dL (SD, 2.0) during HVx monitoring. Twelve neonates received hydrocortisone.

The neonates' blood pressure distributions are shown in Figure 1. We identified MAP_OPT values in 55/64 neonates (86%; 32 boys, 23 girls) during hypothermia, 54/59 (92%; 31 boys, 23 girls) during rewarming, and 55/57 (97%; 35 boys, 20 girls) during normothermia. The mean MAP_OPT values were 50 mm Hg (SD, 10) in each period. (Fig. 2a). The neonates who were coded as having an unidentifiable MAP_OPT did not display a clear HVx nadir in the bar graph of MAP versus HVx. The neonates' blood pressures in relation to MAP_OPT are shown in Figures 2b-f.

We also compared neonates who did not have brain MRI (and were excluded from the study) with those who did. The perinatal injury scores among neonates who did or did not receive an MRI were 6 (SD, 1.3) and 6 (SD, 1.5), respectively. The MAP_OPT values between neonates with or without MRI were 50 mm Hg (SD, 10; n = 55) and 45 mm Hg (SD, 10; n = 10) during hypothermia (p > 0.10) and 50 mm Hg (SD, 10; n = 54) and 45 mm Hg (SD, 10; n = 4) during rewarming (p > 0.10), respectively. However, MAP_OPT during normothermia was lower in neonates without MRI (mean, 40 mm Hg; SD, 10; n = 5) than in those with MRI (mean, 50 mm Hg; SD, 10; n = 55; p = 0.026). Blood pressure in relation to MAP_OPT during hypothermia and rewarming was similar between neonates with and without MRI (p > 0.05; data not shown). Neonates without MRI had greater duration (p = 0.018) and deviation (p = 0.017) in blood pressure above MAP_OPT during normothermia than neonates with MRIs.

MRIs were obtained at 8.5 days of life (SD, 2.5; range, 4-16), and brain injury was graded in all 64 neonates (Table 3). The perinatal insult score was not associated with injury in the white matter, PLIC, putamen and globus pallidus, thalamus, or brain stem (p > 0.05; n = 64). However, more severe perinatal insult was related to greater injury in the paracentral gyri (p = 0.009; n = 64).

In the analysis adjusted for PaCO₂, seizures, vasopressors, and perinatal insult severity, greater AUC below MAP_OPT during rewarming (β = 0.002; p = 0.047, n = 54) was associated with more severe injury in the white matter. More time with blood pressure within the 5-mm Hg range of MAP_OPT during rewarming was related to less white matter injury (β = -0.044, p = 0.017, n = 54) (online suppl. Table 1; for all online suppl. material, see www.karger.com/doi/10.1159/000452833). The autoregulatory parameters were not associated with PLIC injury (p > 0.05 for all comparisons; data not shown).

Both perinatal insult severity (p = 0.009; n = 64) and blood pressure affected paracentral gyrus injury. The analyses, therefore, included adjustments for perinatal insult severity. Among neonates with an identified MAP_OPT, greater duration (β = 0.041; p = 0.007; n = 55), deviation (β = 0.104; p = 0.012; n = 55), and AUC (β = 0.005; p = 0.020; n = 55) of blood pressure below MAP_OPT during hypothermia correlated with more severe injury in the paracentral gyri. In addition, greater duration (β = -0.027, p =0.020, n = 55) and deviation (β = -0.069, p = 0.038, n = 55) of blood pressure above MAP_OPT during hypothermia was related to decreased paracentral gyrus injury (online suppl. Table 2).

Spending more time with blood pressure within MAP_OPT during normothermia was associated with less injury in putamen and globus pallidus (β = -0.034, p =0.040, n = 55) (online suppl. Table 3). Finally, greater duration of blood pressure within MAP_OPT during normothermia was related to less brain stem injury (β = -0.035, p =0.027, n = 55) (online suppl. Table 4). Blood pressure relative to MAP_OPT was not associated with thalamic injury (p > 0.05 for all comparisons; data not shown).

In the adjusted analysis of all neonates with an identified MAP_OPT, rSO₂ was unrelated to brain injury in any anatomic region (p > 0.05; data not shown). Moreover, the duration of blood pressure below the gestational age +5 mm Hg in any period was not associated with injury in any brain region in the adjusted analysis of neonates with an identified MAP_OPT (p > 0.05; data not shown).

Neonates with greater duration of blood pressure below MAP_OPT during hypothermia also spent more time with blood pressure below MAP_OPT during rewarming (r = 0.49; p <0.001). The durations of blood pressure below MAP_OPT were not correlated between hypothermia and normothermia (r = 0.20; p = 0.17) or between rewarming and normothermia (r = 0.20; p = 0.16). Neonates who spent more time during hypothermia with blood pressure above MAP_OPT also spent more time with blood pressure above MAP_OPT during rewarming (r = 0.50, p <0.001). Time with blood pressure above MAP_OPT was not correlated between hypothermia and normothermia (r = 0.21; p = 0.15) or rewarming versus normothermia (r = 0.22; p = 0.12). Finally, neonates with greater duration of blood pressure within MAP_OPT during hypothermia also spent more time with blood pressure within MAP_OPT during rewarming (r = 0.40; p = 0.007) and normothermia (r = 0.32; p = 0.025). Time with blood pressure within MAP_OPT during rewarming and normothermia were not correlated (r = 0.16; p = 0.273).

We provide evidence that blood pressure deviations from the range of optimal autoregulation during and after therapeutic hypothermia are associated with brain injury on MRI in neonates with HIE, but this relationship is complex and varies by anatomic region. Brain injury was not affected by perinatal insult severity in most regions; thus, blood pressures that do not optimize autoregulation during and after therapeutic hypothermia independently affect subsequent brain injury measurements on MRI. Greater duration and deviation of blood pressure below MAP_OPT were associated with greater injury in the white matter and paracentral gyri. Blood pressure within MAP_OPT related to lesser injury in the white matter, putamen and globus pallidus, and brain stem. Hence, hemodynamic management to optimize autoregulation using HVx could serve as a therapeutic adjunct to hypothermia in HIE.

Brain MRI is a biomarker of neurodevelopmental outcome in HIE [26, 27]. We used T1/T2-weighted images, trace of diffusion images, and ADC maps given the high specificity and sensitivity of these sequences in predicting neurodevelopmental outcomes [28]. In addition, qualitatively scoring brain injury on MRI during the first 2 weeks of life and after therapeutic hypothermia has consistent and high predictive accuracy for neurodevelopmental outcome [29]. Signal abnormalities in conventional MRI, trace of diffusion images, and ADC maps in the cortex, PLIC, white matter, putamen and globus pallidus, thalamus, and brain stem predict disability or death after HIE [9, 18, 19, 20, 21]. Thus, identifying methods to reduce brain injury on MRI may improve neurodevelopmental outcomes.

The potential of targeting the autoregulatory blood pressure range as a therapeutic adjunct to hypothermia is unclear. Hypothermia after hypoxic-ischemic brain injury is known to decrease cerebral blood flow [30, 31], although the blood pressure limits of autoregulation and MAP_OPT can still be determined during hypothermia [3, 14, 31]. Severe birth asphyxia may cause brain damage with dysfunctional autoregulation and/or hemodynamic instability. Alternatively, some neonates' outcomes might improve if their blood pressure is maintained to support autoregulation. We developed a perinatal insult severity score to account for the effects of birth asphyxia on autoregulation and subsequent brain injury. Reliable and accurate risk stratification methods to differentiate neonates with favorable or unfavorable neurologic outcomes are not available in HIE, although the 10-min Apgar score may be associated with neurocognitive outcomes at 5-6 years of age [32]. The perinatal insult score, which is derived from common clinical perinatal parameters, did not relate to brain injury in any region except for paracentral gyri. New and noninvasive methods that can identify neonates who are at high risk of permanent neurologic injury soon after birth are urgently needed. These high-risk babies may benefit from adjuvant treatments, such as methods that support autoregulatory function. Alternatively, identifying neonates with poor autoregulation using HVx may indicate those at highest risk of neurologic injury on MRI. Nonetheless, the fact that the perinatal insult score was not associated with brain injury in most regions suggests blood pressures that do not support autoregulation during the first 4 days of life may independently relate to injury evolution in the white matter, putamen and globus pallidus, and brain stem. Such an independent relationship does not strictly distinguish whether blood pressure deviation from MAP_OPT causes additional injury or if the greater injury, particularly in the brain stem, produces greater blood pressure lability.

Blood pressure deviation from MAP_OPT measured by frontal NIRS, which predominantly measures the frontal cortex, related to injury in regions not captured by NIRS, including other cortical (i.e. paracentral gyri) and noncortical regions. Autoregulation involves large cerebral arteries and pial arterioles, and thus has a prominent macrocirculatory component upstream of local parenchymal tissue. Much of the autoregulatory vasoactivity is attributed to the intrinsic myogenic response to transmural pressure. Thus, large regional differences in MAP_OPT are not expected after global cerebral ischemia, and large and prolonged deviations of blood pressure from MAP_OPT measured in the frontal lobe will likely affect perfusion throughout much of the brain. While more precise methods for determining regional autoregulatory properties would be ideal, using MAP_OPT from the frontal lobe is a reasonable first approximation for targeting hemodynamic management.

We identified associations between the autoregulation parameters and brain injury during hypothermia (paracentral gyrus injury), rewarming (white matter injury), and normothermia (putamen and globus pallidus and brain stem injuries). Larger studies are needed to define the interactions between temperature and autoregulation in specific regions. Hypothermia and rewarming do not fully protect cortical gray and white matter from apoptotic cell death [33, 34, 35], and blood pressure below the optimal autoregulatory level may render these regions more vulnerable to cell death. By contrast, the neuroprotection afforded by hypothermia in the putamen [36] may delay vulnerability to suboptimal autoregulatory function until normothermia.

Blood pressure within MAP_OPT was associated with lesser injury in the white matter, putamen and globus pallidus, and brain stem. Blood pressure above MAP_OPT was related to lesser injury in paracentral gyri. Although it is tempting to speculate that targeting or exceeding MAP_OPT would confer neuroprotection, caution must be exercised when raising blood pressure to not induce cardiopulmonary injury with increases in afterload, intravascular volume, and cardiogenic strain. Nonetheless, identifying MAP_OPT related better to brain injury than did rSO₂ or blood pressure based on gestational age +5 mm Hg. Larger studies across multiple institutions are needed to further explore whether hemodynamic management that targets MAP_OPT reduces neurologic injury compared to conventional neonatal clinical guidelines, including those based on the gestational age [16].

Neonates who spent a greater duration of time with blood pressure below MAP_OPT during hypothermia also spent more of rewarming with blood pressure below MAP_OPT. Conversely, neonates with greater duration of blood pressure above MAP_OPT during hypothermia also had more time with blood pressure above MAP_OPT during rewarming. The neonates' time with blood pressure within the 5-mm Hg bin of MAP_OPT were also correlated across hypothermia, rewarming, and normothermia. This consistent relationship between blood pressure and MAP_OPT across time and temperature phases emphasizes the importance of using continuous HVx monitoring to identify neonates with suboptimal autoregulation early in hypothermia because these neonates are at high risk of continued suboptimal autoregulation during rewarming and normothermia.

Causal relationships were not addressed by this observational study. Because of the small sample size, the results may underidentify associations between injury and autoregulation in this single-center study. The exclusion of neonates who received ECMO or had withdrawal of care and did not receive HVx monitoring or brain MRIs creates selection bias. Neonates excluded from the study because they lacked MRI data were similar in perinatal insult score and autoregulation data to those included in the study with the exception of MAP_OPT during normothermia; 5 neonates without MRI had lower MAP_OPT values and greater blood pressure deviation above MAP_OPT during normothermia than the 55 neonates who received MRI. HVx monitoring could only begin after an arterial blood pressure cannula was established; therefore, early autoregulatory instability may not have been captured. We analyzed the data as a percentage of the autoregulation monitoring period and normalized the AUC for the monitoring duration to account for different monitoring durations in hypothermia. We monitored HVx more consistently during rewarming and normothermia. Steroid use and delays in initiating hypothermia may affect blood pressure and neurologic injury, and we did not adjust for these potential confounders.

NIRS-derived HVx autoregulation monitoring can identify the blood pressure range with most robust autoregulatory vasoreactivity in neonates who receive therapeutic hypothermia for HIE. The relationship between cerebrovascular autoregulation and MRI-documented neurologic injury in HIE patients is complex, may vary by anatomic region, and is largely independent of the perinatal injury insult severity. Determination of MAP_OPT that supports maximal cerebrovascular reactivity to perturbations in MAP may identify hemodynamic ranges that limit the progression of neural injury in several regions. The influence of hypothermia and rewarming on the effects of autoregulatory hemodynamics and brain injury requires further study.

We are grateful to Claire Levine, MS, ELS, for her editorial assistance.

Support was provided by NIH R01HD070996, R01HD086058, and R01HD074593 (F.J.N.); K08NS080984 and R21HD072845 (J.K.L.); R01NS060703 (R.C.K.); Johns Hopkins University Clinician Scientist Award and American Heart Association Grant-in-Aid (J.K.L.); and the Sutland-Pakula Endowment for Neonatal Research (R.C.-V.).

Drs. Lee, Northington, Gilmore, and Chavez-Valdez received research support from Medtronic for a separate study.