Brain Injury Outcomes after Adjuvant Erythropoietin Neuroprotection for Moderate or Severe Neonatal Hypoxic-Ischemic Encephalopathy: A Report from the HEAL Trial

Abstract

Introduction

Erythropoietin (Epo) is a putative neuroprotective therapy that did not improve overall outcomes in a phase 3 randomized controlled trial for neonates with moderate or severe hypoxic-ischemic encephalopathy (HIE). However, HIE is a heterogeneous disorder, and it remains to be determined whether Epo had beneficial effects on a subset of perinatal brain injuries.

Methods

This study was a secondary analysis of neuroimaging data from the High-dose Erythropoietin for Asphyxia and Encephalopathy (HEAL) Trial, which was conducted from 2016 to 2021 at 17 sites involving 23 US academic medical centers. Participants were neonates >36 weeks’ gestation undergoing therapeutic hypothermia for moderate or severe HIE who received 5 doses of study drug (Epoetin alpha 1,000 U/kg/dose) or placebo in the first week of life. Treatment assignment was stratified by trial site and severity of encephalopathy. The primary outcome was the locus, pattern, and acuity of brain injury as determined by three independent readers using a validated HIE Magnetic Resonance Imaging (MRI) scoring system.

Results

Of the 500 infants enrolled in HEAL, 470 (94%) had high quality MRI data obtained at a median of 4.9 days of age (IQR: 4.5–5.8). The incidence of injury to the deep gray nuclei, cortex, white matter, brainstem and cerebellum was similar between Epo and placebo groups. Likewise, the distribution of injury patterns was similar between groups. Among infants imaged at less than 8 days (n = 414), 94 (23%) evidenced only acute, 93 (22%) only subacute and 89 (21%) both acute and subacute injuries, with similar distribution across treatment groups.

Conclusion

Adjuvant erythropoietin did not reduce the incidence of regional brain injury. Subacute brain injury was more common than previously reported, which has key implications for the development of adjuvant neuroprotective therapies for this population.

Introduction

Neonatal hypoxic-ischemic encephalopathy (HIE) remains a leading cause of death and long-term neurodevelopmental disability worldwide [1]. While therapeutic hypothermia has increased disability-free survival among infants with moderate and severe HIE, 40–50% continue to suffer adverse outcomes [2], motivating an urgent need for adjuvant neuroprotective therapies. Erythropoietin (Epo), a cytokine with demonstrated neuroprotective and neuroregenerative properties as a monotherapy [3], has shown promise in both preclinical [4–7] and phase 1–2 clinical trials [8–11]. However, the recently completed phase 3, High-dose Erythropoietin for Asphyxia and Encephalopathy (HEAL) trial found that the addition of Epo to hypothermia did not reduce the rate of death or neurodevelopmental impairment over hypothermia alone, nor did it reduce the overall likelihood of brain injury [12].

Still, neonatal HIE is a heterogenous disorder resulting from a wide spectrum of conditions including chronic intermittent hypoxia, sensitization by infection/inflammation, or acute hypoxia ischemia with a sentinel event. The distribution of brain injury identified by MRI varies with the underlying cause and timing of the inciting insult [13–15]. In preclinical models of HIE, adjuvant Epo has shown the most robust effects on injuries to the cerebral cortex [4], white matter (WM) [4, 5], and cerebellum [6]. Similarly, in a phase 2 randomized trial, infants treated with Epo were found to have a reduction in subcortical and cerebellar injury as compared to those who received placebo [9]. Furthermore, unlike hypothermia, Epo has been shown to have beneficial effects on acute [4–6] and subacute [7] brain injuries. In this post hoc secondary analysis of the HEAL trial, we now explore brain injury outcomes in a more detailed manner and determine whether adjuvant Epo is associated with a reduction in brain injury according to the locus, pattern, or acuity of injury.

Materials and Methods

Design

This study was a post hoc exploratory analysis of data from the HEAL trial (NCT02811263). HEAL was a multicenter, phase 3 clinical trial in which 500 neonates with moderate or severe HIE undergoing therapeutic hypothermia were randomly assigned (double-blind) to receive either Epo (1,000 U/kg) or normal saline placebo at 1, 2, 3, 4, and 7 days of age. The design, rationale, baseline characteristics, and primary results of the trial have been previously published [12, 16]. Trial enrollment took place from January 25, 2017 through October 9, 2019, and data were analyzed between February and November 2022. The study protocol was reviewed and approved by the Institutional Review Boards at all participating sites and was registered with the Food and Drug Administration (Investigational New Drug application 102,138). Parents provided written consent for their child’s participation in the research study. This study followed the Consolidated Standards of Reporting Trials (CONSORT) reporting guidelines.

Study Population

Neonates ≥36 weeks of gestational age undergoing therapeutic hypothermia for moderate or severe HIE were eligible to participate in HEAL. Inclusion criteria included the following: (1) either an Apgar score of <5 at 10 min, or need for resuscitation at 10 min, or pH <7.00 or base deficit >15 within the first hour of age and (2) moderate or severe encephalopathy between 1- and 6-h following birth as determined by the modified Sarnat exam. Exclusion criteria included known genetic or congenital abnormalities, birth weight <1,800 g, head circumference <30 cm, hematocrit >65%, and redirection of care being considered due to moribund condition.

MRI Acquisition

The HEAL trial protocol included a brain MRI scan to be performed between 96 and 144 h of age, or as soon thereafter as possible if the infant was clinically unstable. A standardized protocol harmonized across 9 different 3T MR scanners and 23 enrolling hospitals was used as previously described [17]. The MRI protocol included 3D T1-weighted (T1w) images at isotropic 1 × 1 mm resolution, 2D T2-weighted (T2w) images at 1 × 1 × 2 mm resolution, and diffusion-weighted images at 2 × 2 mm resolution from which trace-weighted and apparent diffusion coefficient (ADC) maps were derived. Additionally, short-echo, single-voxel MR spectroscopy (MRS) data were acquired from a 3.4 to 4.9 mL voxel in the left thalamus and left parietal WM. Following acquisition, MRI data were transferred to the HEAL Neuroimaging Core, where they underwent quality assurance procedures and were analyzed blinded to the infant’s treatment assignment, hospital course, and neurodevelopmental outcome.

MRI Measures

The primary MRI outcome measures were semi-quantitative scores derived from a validated HIE MRI scoring system [18] as previously described [17]. Briefly, this scoring system indicates the extent of signal abnormality (i.e., none, ≤25%, 25–50%, ≥50%) observed on the T1w, T2w, and trace/ADC images in 8 regions of interest: caudate, lentiform nucleus (putamen + globus pallidus), thalamus, posterior limb of the internal capsule, cortex, WM, brainstem, and cerebellum. Each MRI scan was scored independently by two of three experienced raters (JLW, RCM, AMM), with any discrepancies resolved by consensus. A deep gray injury score was derived by summing the caudate, lentiform, and thalamic scores while a global injury score was also derived by summing all regional subscores.

Pattern(s) of injury were scored categorically based on the global topography of the injuries observed. We defined 6 patterns of injury in accordance with the published literature [13–15]: central (basal ganglia/thalamus [BGT] ± perirolandic cortex); peripheral (parasagittal cortex and/or WM, i.e., “border-zone or watershed” pattern); global (BGT and total or near total involvement of cortex and/or WM); punctate WM (discrete foci of injury 1 mm–10 mm in size localized to the periventricular WM or centrum semiovale); arterial ischemic stroke (infarct localized to the vascular territory of the middle, anterior, or posterior cerebral arteries); other focal lesion (i.e., venous infarct and/or intraparenchymal hemorrhage, contusion, unilateral focal lesion in BGT). Any injuries not classified into one or more patterns above were subsumed into a seventh category: other pattern (e.g., cerebellar injury, hypoglycemia pattern, bilirubin encephalopathy-like pattern, multifocal pattern). Injury patterns were not mutually exclusive. Infants without observable injury were classified as no injury, including infants for whom the only observed finding was hypomyelination of the posterior limb of the internal capsule on T1w and T2w images, in what was judged to be consistent with incomplete brain maturation.

Acuity of brain injury was scored categorically as acute, subacute, acute + subacute, or chronic based on the co-occurrence of signal abnormalities across the three MRI sequences as previously described [17, 19]. Acute lesions were defined as foci of restricted diffusion with or without corresponding signal abnormalities on T1w and T2w MRI. Subacute lesions were defined by signal abnormalities on T1w and/or T2w MRI without corresponding diffusion restriction (i.e., reduced ADC). Chronic lesions were defined by the presence of volume loss or parenchymal cysts. Each scan received a global acuity score indicating whether there was (1) no injury, (2) acute injury only, (3) subacute injury only, (4) acute + subacute injury, or (5) chronic (+/− acute or subacute injury) (Fig. 1). Last, the locus (intraparechymal, intraventricular, subdural, subarachnoid) and extent (none, trace, mild/mod, severe/mass effect) of any intracerebral blood products were recorded based on T1w and T2w scans.

Fig. 1.

Examples of acute, subacute, and acute + subacute brain injuries as observed in the first week of life in the HEAL trial. Injury acuity was classified based on the signal abnormalities observed on ADC and conventional T1w and T2w MRI, independent of injury pattern. a, b Acute injury was defined by areas of restricted diffusion in either the BGT +/− perirolandic cortex (a, blue arrows, central pattern), cortex/WM in the intervascular borderzone along the parasagittal convexity (b, orange arrows, peripheral pattern), or other loci (not pictured). Subacute injury was defined as signal abnormalities on T1w or T2w MRI without concomitant restricted diffusion and was observed in cases with central (c, e) and peripheral (d, f) patterns of injury. Cases with both acute and subacute injury present as shown in (e) (acute peripheral + subacute central) and (f) (acute central + subacute peripheral) were classified as acute + subacute. Ages at MRI are as follows: a (4.6d), b (4.9d), c (5.1d), d (4.1d), e (3.7d), f (5.4 d).

MRS Processing and Quality Assurance

Raw MRS data were processed centrally using a customized LCModel (V6.3-1L, Stephen Provencher Inc., Oakville, ON, Canada) pipeline as described previously [17]. As above, upon transfer to the HEAL Neuroimaging Core, MRS data were processed and reviewed for protocol compliance and quality. Protocol deviations were logged, reported back to the site, and those data were excluded from subsequent analyses. Voxel position for each ROI was reviewed, and major deviations were flagged, reported back to site, and excluded from subsequent analyses. Last, the spectra were visually inspected for artifacts and any spectra with major artifacts (e.g., skull lipids, motion), poor linewidth (defined as FWHM >0.08), or poor signal to noise (defined as SNR <6) were excluded from subsequent analyses.

MRS Measures

For this study, we focused on the quantitative ratios of lactate to N-acetylaspartate (NAA), lactate + lipids to NAA, and NAA to total creatine. Lactate is a byproduct of anaerobic metabolism and increases in the setting of acute hypoxic-ischemic brain injury [20]. At short TE MRS, lactate levels peak early after acute hypoxia [21] and are later surpassed by overlapping lipid signals, which arise from free lipids accumulating in damaged cells [22]. In this study, we measured both using lactate as a marker of acute brain injury and lactate + lipids as a marker of both acute and subacute injury. By contrast, NAA is synthesized in the mitochondria of neurons and axons as a byproduct of energy metabolism [23]. We analyzed NAA signal in reference to total creatine (free creatine plus phosphocreatine) as a marker of neuronal/axonal integrity.

Outcomes

The primary outcome of interest was the locus, pattern, and acuity of brain, as defined above from the MRI scoring system. Secondary outcomes included the regional MRS measures and incidence of any intracerebral hemorrhages.

Statistical Analysis

Analyses were based on a modified intention-to-treat approach, in which all randomized infants who had received at least one study drug dose and who had MRI and MRS data that met quality assurance thresholds were included. Descriptive statistics (frequency, mean, standard deviation) were used to describe baseline infant, maternal, pregnancy, and delivery characteristics among infants who were included in the MRI and MRS analyses as compared to those who were not. Treatment randomization was stratified by site, and most sites used a single MR scanner, although a limited number (n = 7) used more than one scanner during HEAL due to the presence of multiple hospitals (n = 3) or hardware upgrades (n = 4). We directly tested for inequality in proportion for MR vendor or field strength across treatment groups and found there were no differences (vendor: p = 0.44; field strength: p = 0.49).

The primary analysis compared the incidence of brain injury among those randomized to receive Epo versus placebo using logistic regression, reporting relative risks [RRs] and corresponding 95% confidence intervals (CIs). Brain injury was determined according to brain region, pattern, and acuity, and analyses were adjusted for HIE severity and study site. Due to the expected timing of pseudonormalization (i.e., 10–14 days for acute injuries in HIE infants undergoing hypothermia) [24], we conservatively excluded all infants for whom MRI was completed after the 7th day (>168 h of age) from analyses related to injury acuity to ensure we had adequate sensitivity for distinguishing between acute and subacute injuries. In addition, because HIE severity was used to stratify treatment randomization in the primary trial, subgroup-specific treatment effects were estimated for the primary outcome measures according to severity of encephalopathy using proportion odds ratios, logistic regression, or bootstrapped differences in medians, as appropriate. Secondary analyses used the bootstrapped difference in medians (95% CIs) to evaluate whether the degree of injury measured by (1) MRS biomarkers and (2) brain hemorrhages differed across treatment groups. Box plots were used to evaluate distributions of spectroscopy ratios in groups of interest. Analyses were performed between February 2023 and September 2023 using R Statistical software version 4.0.3 (R Foundation for Statistical Computing).

Results

Description of the Cohort

MRI data were obtained from 474 (95%) infants enrolled in HEAL at a median of 4.9 days of age (IQR: 4.5–5.8). Of those, 4 were excluded from MRI analyses and 115 were excluded from MRS, leaving a final sample of 470 with MRI or MRS and 359 with MRI and MRS (online suppl. eFig. 1; for all online suppl. material, see https://doi.org/10.1159/000534618). Baseline clinical and demographic characteristics of the infants with and without neuroimaging data were similar, except that infants without neuroimaging were more likely to have severe HIE and more likely to have died in the first week of life (online suppl. eTable 1).

Two hundred and forty infants included in the primary analyses received Epo and two hundred thirty received placebo (online suppl. eFig. 1). The median (IQR) age at MRI in the Epo (4.9 [4.5, 5.8]) and placebo (5.0 [4.5, 5.8]) and the number of study drug doses received prior to MRI were similar in the Epo (4 [4, 4]) and placebo (4 [4, 4]) groups.

Primary Outcomes by Treatment Group

Locus of Injury

Brain injury was observed on MRI in 324/470 (69%) infants, with a median (IQR) global brain injury score of 8 (2, 22). As shown in Figure 2, injury to the deep gray nuclei was evident in 119 out of 240 (50%) of Epo-treated infants and 110 out of 230 (48%) of placebo-treated infants (RR, 1.03; 95% CI: 0.85–1.24). Injury to the cortex was evident in 92 (38%) of Epo-treated infants and 89 (39%) of placebo-treated infants (RR: 1.04; 95% CI: 0.89–1.21). Injury to the WM was evident in 147 (61%) of Epo-treated infants and 134 (58%) of placebo-treated infants (RR: 1.04; 95% CI: 0.89–1.21). Other regions also showed similar effects across groups (Fig. 2).

Fig. 2.

Forest plot of primary brain injury outcomes, stratified by treatment group.

The degree of injury in each locus was also similar between treatment groups. As shown in online supplementary eFigure 2, the median (IQR) difference in injury to the deep gray nuclei between Epo and placebo groups was 0 (−2, 2). Similar findings were observed in the caudate, putamen, and thalamus, individually, as well as in the cortex, WM, brainstem, and cerebellum (online suppl. eFig. 2).

Pattern of Injury

The central pattern of brain injury was most common, observed in 94 (39%) of Epo-treated infants and 78 (34%) of placebo-treated infants (RR: 1.12; 95% CI: 0.9–1.38). The peripheral/watershed pattern was observed in 55 (23%) of Epo-treated infants and 62 (27%) of placebo-treated infants (RR: 0.96; 95% CI: 0.73–1.29). Other patterns, including punctate WM lesions, other focal lesions and arterial ischemic strokes were observed at similar rates in both groups (Fig. 2).

Acuity of Injury

As noted above, to account for pseudonormalization of the ADC signal, acuity analyses were restricted to infants who underwent MRI in the first week of life (n = 414/470). Of those, 23% (94/414) evidenced acute brain injury only, 22% (93/414) evidenced subacute only, and 21% (89/414) acute and subacute injuries. Only 2% (8/414) evidenced chronic injures (+/− acute or subacute). The distribution of injury acuity was similar in Epo and placebo groups (Fig. 2).

Locus, Pattern, and Acuity Stratified Severity of Encephalopathy

In the primary trial, treatment randomization was stratified by severity of encephalopathy. We found that HIE severity was independently associated with the locus, pattern and acuity of brain injury (Table 1). Specifically, infants who presented with severe encephalopathy in the first 6 h of life had more injury to the deep gray nuclei, in particular, thalamus and lentiform nucleus, as well as cortex and WM. They also had an increased likelihood of the central gray pattern of injury and of acute or acute + subacute injury while infants with moderate encephalopathy had an increased likelihood of no injury. The rate of intracerebral hemorrhages was similar among infants with moderate and severe encephalopathy.

Table 1.

Brain injury outcomes by treatment group, stratified by severity of clinical encephalopathy

	All moderate (n = 370)	Moderate HIE			Severe HIE				HIE group comparison (95% CI)
		Epo-treated (n = 189)	placebo (n = 181)	treatment group comparison (95% CI)	all severe (n = 100)	Epo-treated (n = 51)	placebo (n = 49)	treatment group comparison (95% CI)
Total MRI Injury  median (IQR)	6 (1.25, 16)	6 (1, 16)	6 (2, 16)	0 (−3, 2)a	33 (8, 72.25)	34 (12, 75.5)	22 (6, 71)	12 (−28, 27)a	27 (14, 39)
Regional subscores, median (IQR)
Deep gray  nuclei	0 (0, 4)	0 (0, 6)	0 (0, 4)	0 (0, 0)a	13 (2, 42)	13 (4, 37)	10 (1, 42)	−3 (−18, 16)a	13 (8, 22)a
Caudate	0 (0, 0)	0 (0, 0)	0 (0, 0)	0 (0, 0)a	0 (0, 12)	0 (0, 9)	0 (0, 12)	0 (−3, 1)a	0 (0, 0)a
Putamen/GP	0 (0, 3)	0 (0, 4)	0 (0, 2)	0 (0, 0)a	6 (0, 14)	6 (2, 16)	4 (0, 14)	−2 (−8, 8)a	6 (4, 9)a
Thalamus	0 (0, 2)	0 (0, 2)	0 (0, 2)	0 (0, 0)a	7 (0, 16)	8 (1.5, 16)	6 (0, 16)	−2 (−6, 8)a	7 (4, 11)a
PLIC	2 (0, 4)	2 (0, 4)	0 (0, 4)	−2 (−2, 2)a	4(0.75, 8.25)	4 (2, 8)	4 (0, 9)	0 (−2, 4)a	2 (2, 6)a
WM	2 (0, 6)	2 (0, 6)	2 (0, 6)	0 (−1, 2)a	6 (0, 14)	6 (1.5, 15.5)	6 (0, 13)	0 (−3, 6)a	4 (2, 6)a
Cortex	0 (0, 2)	0 (0, 2)	0 (0, 2)	0 (0, 0)a	3 (0, 10)	2 (0, 10)	4 (0, 10)	2 (−5, 3)a	3 (1, 6)a
Brainstem	0 (0, 0)	0 (0, 0)	0 (0, 0)	0 (0, 0)a	0 (0, 6)	0 (0, 7)	0 (0, 6)	0 (−4, 3)a	0 (0, 2)a
Cerebellum	0 (0, 0)	0 (0, 0)	0 (0, 0)	0 (0, 0)a	0 (0, 2)	0 (0, 2)	0 (0, 2)	0 (0, 1)a	0 (0, 0)a
MRI pattern(s) of injuryc
Central gray	106 (29)	59 (31)	47 (26)	1.13 (0.85, 1.51)b	66 (66)	35 (69)	31 (63)	1.13 (0.91, 1.4)b	2.30 (1.86, 2.85)b
Peripheral/ watershed	87 (24)	42 (22)	45 (25)	0.95 (0.68, 1.31)b	29 (29)	13 (25)	16 (33)	1.11 (0.71, 1.74)b	1.23 (0.86, 1.76)b
Other
Punctate  WM lesions	83 (22)	45 (24)	38 (21)	1.11 (0.79, 1.56)b	15 (15)	11 (22)	4 (8)	2.26 (0.97, 5.3)b	0.67 (0.40, 1.10)b
Other focal  lesion	42 (11)	17 (9)	25 (14)	0.65 (0.36, 1.16)b	13 (13)	6 (12)	7 (14)	1.21 (0.52, 2.81)b	1.15 (0.64, 2.05)b
Arterial  Ischemic Stroke	14 (4)	8 (4)	6 (3)	1.28 (0.45, 3.61)b	7 (7)	3 (6)	4 (8)	1.17 (0.32, 4.31)b	1.85 (0.77, 4.46)b
Atypical  injury	93 (25)	44 (23)	49 (27)	0.94 (0.69, 1.28)b	27 (27)	5 (10)	3 (6)	1.39 (0.86, 2.24)b	1.07 (0.74, 1.55)b
Timing of brain  injury		(n = 173)	(n = 157)			(n = 42)	(n = 42)
No injury	117 (35)	64 (37)	53 (34)		13 (15)	5 (12)	8 (19)		0.44 (0.26, 0.73)b
Acute (only)	61 (18)	25 (14)	36 (23)	0.69 (0.46, 1.05)b	33 (39)	19 (45)	14 (33)	1.24 (0.85, 1.81)b	2.13 (1.50, 3.01)b
Subacute (only)	83 (25)	43 (25)	40 (25)	0.93 (0.67, 1.30)b	10 (12)	6 (14)	4 (10)	1.64 (0.60, 4.43)b	0.47 (0.26, 0.87)b
Acute +  subacute	62 (19)	35 (20)	27 (17)	1.05 (0.70, 1.58)b	27 (32)	11 (26)	16 (38)	1.03 (0.66, 1.62)b	1.71 (1.17, 2.51)b
Chronic (+/−  acute/sub.)	7 (2)	6 (3)	1 (1)	4.63 (0.58, 36.80)b	1 (1)	1 (2)	0 (0)	NA	0.56 (0.07, 4.50)b

n (%) shown unless otherwise specified.

PLIC, posterior limb of the internal capsule.

^aBootstrapped difference in medians (95% CI), unadjusted.

^bRR (95% CI) estimated with a generalized linear model, unadjusted.

^cNot mutually exclusive.

As shown in Table 1, within the subgroups of infants with moderate and severe encephalopathy, regional brain injury scores were similar among Epo-treated and placebo infants. Likewise, the distribution of injury patterns and acuity were similar among Epo-treated and placebo groups (Table 1).

Secondary Outcomes by Treatment Group

Regional MRS Biomarkers

After quality assurance procedures, 359 infants were included in the final MRS analyses (online suppl. eFig. 1). As shown in online supplementary eTable 1, the infants with MRS were similar to the infants with MRI alone. The median (IQR) ratio of lactate to NAA obtained from the BGT region similar in the Epo-treated group 0.07 (0.00, 0.14) as compared to placebo 0.08 (0.00, 0.17) (Fig. 3). Likewise, the median ratio of lactate to NAA in the parietal WM region was similar in the Epo-treated group 0.09 (0.00, 0.29) as compared to placebo 0.07 (0.00, 0.32). Similar findings were observed for the ratios of lactate + lipids to NAA and NAA to creatine (Fig. 3).

Fig. 3.

Regional MRS biomarkers. The median and IQR brain injury scores for the Epo and placebo groups are shown for each of the 3 MRS biomarkers obtained from the left basal ganglia/thalamus (BGT) region of interest (ROI) (top) and left parietal WM ROI (bottom). Bootstrapped difference in medians with (95% CIs) was used to identify significant differences, which would be indicated by 95% CIs that do not contain 0. Corresponding median differences (95% CI) from the BGT ROI are as follows: Lac/NAA −0.01 (−0.03, 0.03); Lac+lipids/NAA −0.12 (−0.35, 0.12); NAA/creatine −0.01 (−0.04, 0.02). Corresponding median differences (95% CI) from the parietal WM ROI are as follows: Lac/NAA 0.02 (−0.06, 0.09); Lac+lipids/NAA 0.21 (−0.19, 0.48); NAA/creatine 0.01 (−0.03, 0.03).

Brain Hemorrhages

Intracerebral hemorrhages were observed in 197 out of 470 infants. Subdural hemorrhages were the most common, observed in 152 infants (32%) followed by intraventricular and intraparenchymal hemorrhages. Across all loci, severe hemorrhages were rare (1–2%), with the majority being trace followed by mild/moderate without mass effect. The distribution of intracranial hemorrhages was similar across Epo-treated and placebo groups (Table 2) and among infants with moderate or severe HIE (online suppl. eTable 2).

Table 2.

Brain hemorrhages by treatment group

	All (n = 470)	Epo (n = 240)	Placebo (n = 230)	Group comparison (95% CI)
Intraparenchymal				1.30 (0.77, 2.19)*
None	404 (86)	203 (85)	201 (87)
Trace	41 (9)	20 (8)	21 (9)
Mild/mod	18 (4)	12 (5)	6 (3)
Severe	7 (1)	5 (2)	2 (1)
Intraventricular				1.54 (0.98, 2.44)*
None	376 (80)	184 (77)	192 (83)
Trace	70 (15)	41 (17)	29 (13)
Mild/mod	13 (3)	7 (3)	6 (3)
Severe	11 (2)	8 (3)	3 (1)
Subdural				1.09 (0.75, 1.59)*
None	318 (68)	162 (68)	156 (68)
Trace	89 (19)	40 (17)	49 (21)
Mild/mod	55 (12)	32 (13)	23 (10)
Severe with mass effect	8 (2)	6 (3)	2 (1)
Subarachnoid				0.85 (0.34, 2.13)*
None	451 (96)	231 (96)	220 (96)
Questionable/trace	12 (3)	6 (3)	6 (3)
Focal	7 (1)	3 (1)	4 (2)
Diffuse	0 (0)	0 (0)	0 (0)

*Odds ratio and 95% CI estimated with proportional odds regression model, adjusted for HIE severity and site.

Discussion

Treatment with Epo as an adjuvant to hypothermia did not reduce the likelihood of injury to deep gray nuclei, cortex, WM, brainstem, or cerebellum, nor did it alter the distribution of injuries by pattern or acuity in neonates with moderate to severe HIE. While these findings contrast with earlier reports from preclinical studies [4–7] and phase 1/2 clinical trials [9, 11, 25, 26], they parallel the primary outcome from the parent phase 3 HEAL trial [12] and underscore that adding Epo to hypothermia does not provide neuroprotection for infants with moderate or severe HIE, regardless of the location of the underlying brain injury. At the same time, the present findings broaden our understanding of HIE brain injury in the era of therapeutic hypothermia, with key implications for the development of adjuvant neuroprotective therapies.

First, among the 69% of infants in the present study with documented brain injury on MRI, nearly equal proportions had acute, subacute, and acute + subacute brain injuries. Early studies, conducted in the pre-therapeutic hypothermia era, suggested that the vast majority of brain injuries in neonates who met criteria for moderate to severe HIE were acute, with chronic injuries being rare [13, 27]. While our finding is that 3% of the sample evidenced chronic injuries is in agreement with the earlier literature, half of the sample had evidence of subacute injuries contrasts with these earlier reports. One possible explanation for this discrepancy is that the increased signal to noise afforded by modern 3T scanners has increased our ability to detect injury, particularly subacute injuries, which are often more subtle on MRI. Another possibility is that therapeutic hypothermia has altered the distribution of lesions observed by MRI. Clinical and experimental studies suggest that therapeutic hypothermia is only effective when initiated shortly after the hypoxic-ischemic insult [28, 29]. Thus, it is possible that a larger proportion of acute lesions are successfully treated by hypothermia, thereby shifting the distribution of residual lesions apparent after hypothermia to include a higher proportion of non-responsive subacute lesions. Regardless of the underlying reasons, the high rate of subacute has key implications for the timing of injury and the development of adjuvant therapies for this population.

Prior studies have shown that ADC decreases acutely after hypoxic-ischemic brain injury reaching a nadir at approximately 2–3 days and then rises thereafter, pseudonormalizing between 10 and 14 days after injury [24]. In the present study, 45% of the infants had injuries that were associated with either normal or elevated ADC values in the first week of life and no volume loss, i.e., subacute injuries (Fig. 1). Given the age at MRI (≤ 1 week) and expected rate of pseudonormalization, we speculate that the onset of injury was likely incurred days before birth. To date, advances in intrapartum care and fetal monitoring have had little effect on reducing the rate of HIE [30, 31]. Our findings suggest that reducing the incidence of HIE will require the development of new approaches to fetal monitoring to identify at-risk infants well before they present in labor and delivery. Furthermore, they suggest that successful translation of adjuvant neuroprotective therapies for moderate to severe HIE will require efficacy on subacute brain injury.

The findings from this study underscore importance of stratifying treatment randomization by HIE severity while also demonstrating that moderate to severe encephalopathy is not simply a continuum of worsening brain injury. Rather, our findings indicate that the locus, pattern, and acuity of brain injury are different among neonates with moderate as compared to severe encephalopathy. Specifically, we found that infants with severe HIE were more likely to exhibit injury to the deep gray nuclei, cortex, and WM, more likely to exhibit the central pattern and more likely to exhibit acute or acute + subacute brain injuries as compared to infants with moderate HIE. The loci and pattern suggest that infants with severe HIE are at highest risk for cerebral palsy and global developmental delay while the acuity findings suggest that they also may be the infants who benefit most from therapies that mitigate the early phases of HIE brain injury. Moreover, collectively, the neuroimaging findings suggest that infants with severe HIE are most likely to have suffered a prolonged, near total asphyxia event, or multiple events. Further research is needed to determine whether the development of individualized neuroprotective therapies for moderate and severe HIE improves outcomes.

Strengths of our study include the prospective study design, large sample size, harmonized neuroimaging protocol, consistent early imaging (in the first week of life), and use of a validated MRI scoring system [18]. Additionally, 76% of the cohort had quality MRS data obtained from the left basal ganglia – thalamus region and parietal WM, with findings reinforcing the primary MRI outcome. Limitations include the lack of imaging in the first day of life prior to neuroprotective therapy. As such, our measures reflect residual brain injuries and do not provide an indication of which injuries were successfully treated with neuroprotection. Additionally, most infants were imaged at 4–5 days of age, prior to the final dose of study drug. Thus, we were not able to detect therapeutic effects on later phases of HIE brain injury, including tertiary reparative effects. While outcome data at 2–3 years showed no differences between groups [12], there remains a possibility that Epo-induced late reparative effects might manifest at school age or later as cognitive demands increase. Finally, we employed a semiquantitative scoring system together with quantitative MRS measures; however, it is possible that quantitative voxelwise imaging analyses would have been more sensitive to potential group differences. Further research is underway.

Conclusion

There was no apparent benefit of adjuvant erythropoietin on the location, type, or acuity of brain injury in infants with HIE. Subacute brain injury was more common than previously reported and suggests that in many cases the onset of injury may start well before the time of delivery. Successful translation of adjuvant neuroprotective therapies will likely require approaches that mitigate subacute brain injuries.

Appendix

The members of the HEAL Consortium who contributed to this paper include (in alphabetical order of participating site): Kaashif Ahmed, Ping-Sun Keven Chen, James Dix (Children’s Hospital of San Antonio and Methodist Children’s Hospital), Ellen Bendel-Stenzel, Andrea Lampland, Richard Patterson, Yanerys M. Ramos (Children’s Hospitals and Clinics of Minnesota), Taeun Chang, Stanley Fricke, Matthew Whitehead (Children’s National Medical Center), Tai-Wei Wu, Stefan Bluml, Jessica L Wisnowski (Children’s Hospital Los Angeles), Toby Yanowitz, Ashok Panigrahy (Children’s Hospital of Pittsburgh of UMPC and the Magee Women’s Hospital), John Flibotte, Jeff Berman, Arastoo Vossough (Children’s Hospital of Philadelphia), Brenda Poindexter, Jean Tkach, Beth Kline-Fath (Cincinnati Children’s Hospital Medical Center), David Riley, Hayden W. Head (Cook Children’s Hospital), Nathalie Maitre, Mark Smith, Jerome Rusin (Nationwide Children’s Hospital), Mariana Beserga, Michael Oveson, John Rampton, (Primary Children’s Hospital and the University of Utah Hospital), Ulrike Mietzsch, Gregory M. Sokol, Chang Y. Ho (Riley Children’s Hospital), Dennis Mayock, Ulrike Mietzsch, Seth Friedman, Dennis Shaw (Seattle Children’s Hospital and the University of Washington Medical Center), Krisa Van Meurs, Shreyas Vasanawala, Kristen Yeom (Stanford University), Fernando Gonzalez, Yvonne Wu, Trevor Flynn, Duan Xu (UCSF Benioff Children’s Hospital), Joern-Hendrik Weitkamp, Sumit Pruthi (Vanderbilt University Medical Center), Fernando Gonzalez, Trevor Flynn, Duan Xu (UCSF Benioff Children’s Hospital), Lina Chalak, Nancy Rollins (University of Texas Southwestern), Amit M. Mathur, Rakesh Rao, Zachary Vesoulis (Washington University).

Acknowledgments

We thank Ashok Panigrahy for his role as advisor to the HEAL Neuroimaging Core and John Feltner, Kelleen Nelson, and Stephanie Hauge for their contributions as central study coordinators. We thank Ronnie Guillet, Jody Ciolino, Michael Cotten, Robin Ohls, Renee Shellhaas, and Janet Soul for their contributions as members of the Data Safety Monitoring Committee. Finally, we are indebted to the parents of the infants enrolled in this trial for being willing to take part in this research effort, which would not have been possible otherwise.

Statement of Ethics

This study protocol was reviewed and approved by the UCSF Human Research Protection Program Institutional Review Board, approval number 16-19260; Cincinnati Children’s Hospital Institutional Review Board, approval number 2017-2040; Children’s Minnesota Institutional Review Board, approval number 1608-092; The Children’s Hospital of Philadelphia Research Institute Institutional Review Board, approval number 17-013928; Children’s Hospital Los Angeles Institutional Review Board, approval number CHLA-16-00535; Children’s National Medical Center Institutional Review Board, approval number PR000007971; University of Texas Southwestern Medical Center Institutional Review Board, approval number STU-072016-075; Cook Children’s Health Care System Institutional Review Board, approval number 2017-010; Indiana University Institutional Review Board, approval number 1606206292; Nationwide Children’s Institutional Review Board, approval number IRB16-00569; University of Pittsburgh Institutional Review Board, approval number PRO16070554; Stanford University Institutional Review Board, approval number 38029; CHRISTUS Health Institutional Review Board, approval number 2016-074; Methodist Healthcare Institutional Review Board, approval number 926129; Intermountain Healthcare Research Institutional Review Board, approval number 1050339; The University of Utah Institutional Review Board, approval number IRB_00093313; Seattle Children’s Institutional Review Board, approval number STUDY00000181; Vanderbilt University Medical Center Human Research Protections Program, approval number 162107; Washington University in St. Louis Institutional Review Board, approval number 201608022. Written informed consent was obtained from parents prior to their child’s participation in the research study.

Conflict of Interest Statement

The authors have no conflicts of interest to declare.

Funding Sources

The High-dose for Erythropoietin for Asphyxia and Encephalopathy (HEAL) trial was funded by the National Institute of Neurological Disorders and Stroke of the National Institutes of Health under Award Numbers U01NS092764 and U01NS092553. JLW is supported by a Research Career Development Award funded by the Eunice Kennedy Shriver National Institute of Child Health and Human Development Institute (K23HD099309). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Author Contributions

Jessica L. Wisnowski made substantial contributions to conception and design, acquisition of data, and analysis and interpretation of data; wrote the first draft of the article; and provided final approval of the version to be published. Sarah E. Monsell, Bryan A. Comstock, and Patrick J. Heagerty made substantial contributions to analysis and interpretation of data; revised the paper critically for important intellectual content; and provided final approval of the version to be published. Stefan Bluml and Robert C. McKinstry made substantial contributions to conception and design, acquisition of data; revised the paper critically for important intellectual content; and provided final approval of the version to be published. Amy M. Goodman and Yi Li made substantial contributions to conception and design and interpretation of data; revised the paper critically for important intellectual content; and provided final approval of the version to be published. Sandra E. Juul, Yvonne W. Wu, and Amit Mathur made substantial contributions to conception and design, acquisition of data, and analysis and interpretation of data; revised the paper critically for important intellectual content; and provided final approval of the version to be published.

Funding Statement

The High-dose for Erythropoietin for Asphyxia and Encephalopathy (HEAL) trial was funded by the National Institute of Neurological Disorders and Stroke of the National Institutes of Health under Award Numbers U01NS092764 and U01NS092553. JLW is supported by a Research Career Development Award funded by the Eunice Kennedy Shriver National Institute of Child Health and Human Development Institute (K23HD099309). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Data Availability Statement

A complete de-identified patient data set will be available at https://www.ninds.nih.gov/CurrentResearch/Research-Funded-NINDS/ClinicalResearch/Archived-Clinical-Research-Datasets on July 1, 2023. Further inquiries can be directed to the corresponding author.