Early Glycemic State and Outcomes of Neonates With Hypoxic-Ischemic Encephalopathy

We aimed to characterize the glycemic profile in the first 12 hours after birth and its association with death and/or neurodevelopmental impairment (NDI) in infants enrolled in the High-dose Erythropoietin for Asphyxia and Encephalopathy (HEAL) Trial.

Data were collected as part of the phase III, multicenter, double-blinded, randomized, placebo controlled HEAL Trial.⁴  The trial included 500 infants born ≥36 weeks’ gestational age and treated with TH for moderate or severe HIE who were randomized to either 5 doses of erythropoietin (1000 IU/kg/dose) or placebo. Primary outcome was defined as the composite outcome of death or NDI of any severity assessed at 22 to 36 months. The window was extended from 22 to 26 months to 22 to 36 months because of the SARS-CoV-2 pandemic. Any NDI was defined as Gross Motor Function Classification System (GMFCS) level ≥1, or cerebral palsy diagnosed on a standardized neurologic exam, or Bayley Scales of Infant Toddler Development, third edition, cognitive score <90. The combined outcome of moderate or severe NDI was defined as a GMFCS level of 1 and cerebral palsy, a GMFCS level of ≥2, quadriplegic cerebral palsy, or a Bayley Scales of Infant Toddler Development, third edition, cognitive score of <85, corresponding to 1 standard deviation below the mean. Details of the trial protocol including the CONSORT diagram have been published previously.^4,15  Blood glucose measurements were performed as part of clinical practice.

A total of 2568 glucose values were documented for 491 study subjects who had at least 1 glucose value measured in the first 12 hours after birth. Data were visually inspected and cleaned to remove spurious measurements (eg, those that occurred because of a contaminated sample from a line with concurrent dextrose administration). Unusually high-appearing blood glucose levels (>400 mg/dL) were considered contaminated and eliminated (n = 27; 1.1%) if a corresponding normal glucose level was documented within 2 hours without documented insulin administration. The remaining 2541 values were included in this analysis. Any remaining values above 400 mg/dL were winsorized to 400 mg/dL to minimize the effect of influential outliers. A histogram of the final included values, split by severity of HIE at randomization, is shown in Fig 1A. Average glucose over time stratified by severity of HIE is depicted in Fig 1B, using the locally estimated scatterplot smoothing (LOESS) method.

In previous studies of similar populations, any hypoglycemic or hyperglycemic episodes during the first 6 to 12 hours after birth were used to characterize neonates as hypoglycemic, hyperglycemic, or both. Often the hypoglycemic episode preceded the hyperglycemic episode, raising the question of whether the hyperglycemic episode was treatment induced rather than purely associated with the initial injury.^12,14  Previous studies have also used a range of values to classify infants as hypo- (<36-50 mg/dL) and hyperglycemic (>144-150 mg/dL).^{6,7,9–14,16}  We performed an exploratory analysis of the association between highest and lowest glucose levels for each infant in the first 12 hours after birth and the primary outcome of death or NDI, using LOESS curves (Fig 2A and 2B). We defined hyperglycemia as a blood glucose level of >200 mg/dL, hypoglycemia as <50 mg/dL, and euglycemia as blood glucose levels between 50 and 200 mg/dL, based on visual inspection of the LOESS curve inflection points at which >50% of neonates experienced an abnormal outcome. A similar method was used to examine average glucose profiles over time by outcome. Infants who experienced both hypo- and hyperglycemia (n = 13) were classified based on the abnormal glycemic state that occurred first.

To examine the association between glycemic state and outcome, we calculated the adjusted odds ratios (aOR) of death or NDI in infants who were hypo- or hyperglycemic in the first 12 hours after birth, compared with those who were euglycemic, using logistic regression. Similar analyses were performed to examine the associations between glycemic state and death and glycemic state and NDI among survivors. Sensitivity analyses were performed where infants who experienced both hypo- and hyperglycemia in the first 12 hours were removed (Supplemental Fig 5). Because blood glucose levels are influenced by severity of illness and medication exposures, models were adjusted for sentinel events during labor (shoulder dystocia, placental abruption, prolapsed cord, or uterine rupture), treatment group (erythropoietin versus placebo), small for gestational age, large for gestational age, severity of encephalopathy, 10-minute Apgar score, need for intubation, use of inotropic medications and/or hydrocortisone on day of life 1, and documented severe adverse events within 24 hours of birth, specifically disseminated intravascular coagulation (clinical bleeding warranting transfusion of blood products), intracranial hemorrhage (intraparenchymal or intraventricular blood seen on brain imaging), and severe pulmonary hypertension (requiring inhaled nitric oxide or extracorporeal membrane oxygenation). Analyses were conducted in RStudio using the R statistical package (version 4.1.2, Foundation for Statistical Computing, Vienna, Austria).¹⁷ 

Within the first 12 hours after birth, 491 of the 500 (98.2%) trial participants had documented blood glucose values and were included in this analysis. A total of 2541 glucose values were documented in the period of interest, with a median of 5 (interquartile range, 3-6.5) values per baby. Of those, 593 (23.3%) were serum laboratory measurements, 1935 (76.2%) were point-of-care measurements and 13 (0.5%) did not have a documented measurement type. Hypoglycemia occurred in 98 (20.0%), hyperglycemia in 111 (22.6%), and euglycemia in the remaining 282 (57.4%) neonates.

Of the 491 included infants, 379 (77.2%) were diagnosed with moderate and 112 (22.8%) with severe HIE. Hyperglycemia was significantly more common among neonates with severe compared with those with moderate HIE (42.3% vs 16.9%; P < .001), whereas the rate of hypoglycemia was similar among both groups (severe, 21.4% vs moderate, 19.5%; P = .69).

Infants with abnormal glucose levels had a greater degree of perinatal acidosis, a higher rate of severe HIE and kidney or liver dysfunction, and a higher rate of receiving medications that may alter glucose levels such as hydrocortisone (Table 1). Compared with the percentage of euglycemic infants who experienced a documented sentinel event at birth (28%), hyperglycemic infants were more likely (35%), and hypoglycemic infants less likely (22%), to have a documented sentinel event.

Glucose profiles within the first 12 hours after birth differed between neonates with moderate and severe HIE (Fig 1A). Neonates with moderate HIE were more likely to be euglycemic compared with those with severe HIE (63.6% vs 36.6%; P < .001). The average blood glucose levels remained relatively constant over the first 12 hours after birth in neonates with moderate HIE, whereas mean glucose levels in neonates with severe HIE were more dynamic, with peak levels occurring by 6 hours followed by a plateau period before decreasing again around 10 to 12 hours after birth (Fig 1B). Additionally, infants with severe HIE had a much wider range of initial blood glucose measurements and, on average, a lower initial blood glucose than those with moderate HIE (Fig 1B).

Outcome data at 22 to 36 months was available for 472/491 (96.1%). Both minimum and maximum measured blood glucose levels during the first 12 hours after birth were associated with death or NDI, with U-shaped associations between highest/lowest glucose and death/NDI (Fig 2A and 2B). Qualitatively, infants with blood glucose levels <25 mg/dL had the highest rates of abnormal long-term outcome. An increased proportion of death or NDI was also seen in infants with a minimum glucose value >200 mg/dL (Fig 2A). Maximum blood glucose levels in the first 12 hours after birth were also associated with death or moderate-severe NDI (Fig 2B). Infants with a maximum blood glucose level >200 mg/dL had increasing adverse primary outcome proportional to the maximum blood glucose level.

When separated by outcome, average glucose profiles in the first 12 hours of life showed notable differences (Fig 3). Neonates who died displayed group-averaged initial blood glucose levels of approximately 65 mg/dL, which increased to nearly 200 mg/dL on average by 6 hours after birth and remained higher than the other outcome groups for the majority of the 12-hour period. In contrast, neonates who survived without NDI had average initial glucose levels of approximately 100 mg/dL, which increased to approximately 125 to 130 mg/dL on average at 4 to 6 hours before slowly declining again. Those who survived with any NDI displayed group-averaged glucose levels between the other 2 outcome categories; initial average blood glucose was around 85 mg/dL, increasing over the 12-hour period to approximately 135 mg/dL.

Euglycemic neonates were least likely to experience an adverse outcome of death or NDI and were used as the reference group. Primary outcome was more likely to occur in neonates with hypoglycemia (aOR, 2.62; 95% confidence interval [CI], 1.5–4.67) and hyperglycemia (aOR, 1.77; 95% CI, 1.03–3.03) (Fig 4). When separating primary outcome by death and NDI, neonates with hypoglycemia had an increased risk of both death (aOR, 2.85; 95% CI, 1.09–7.43) and NDI in survivors (aOR, 2.50; 95% CI, 1.09–7.43). In contrast, neonates with hyperglycemia had a significantly increased risk of death (aOR, 2.52; 95% CI, 1.10–5.77), but not NDI (Fig 4B and C). Removing infants who experienced both hypo- and hyperglycemia, did not alter the results (sensitivity analyses; Supplemental Fig 5).

In this large cohort, we show that glucose profiles in the first 12 hours after birth differ in neonates with moderate versus severe encephalopathy and that baseline values are associated with severity of illness. After adjusting for severity of illness and degree of encephalopathy, early glycemic state was still associated with primary outcome at 22 to 36 months. Specifically, hypoglycemia was associated with both mortality and higher rates of NDI in survivors, whereas hyperglycemia during the 12 hours following birth was associated with an increased risk of death, but not NDI.

Hypoglycemic and hyperglycemic states have both been reported to be associated with adverse outcome in newborns affected by moderate to severe HIE.^6,12,13  Although many studies have assessed hypoglycemia and hyperglycemia as contributing or causative factors for adverse outcome, few studies have investigated glycemic trends and profiles during the first 12 hours after birth. Emerging literature on small patient samples supports the idea that timing and duration of injury might determine which neonates become hypoglycemic versus hyperglycemic.^9,12  Our data suggest that healthier babies display a greater ability to compensate for the increasing metabolic demands of HIE and can achieve glucose homeostasis despite a perinatal insult. In comparison, sicker babies with HIE maintain a higher baseline blood glucose concentration, have greater glucose lability in the first 12 hours after birth, and these findings are associated with unfavorable outcomes. These results are similar to those described in a post hoc analysis of the CoolCap trial and in a small prospective study by Montaldo et al, both of which showed that newborns who have glucose lability in the first 12 hours have a trend toward unfavorable outcome.^6,12 

Early hypoglycemia within the first 6 to 12 hours of life has been shown in several studies to be associated with severity of HIE and gravity of metabolic acidosis, suggesting that early hypoglycemia may reflect depleted fetal glucose stores and/or inadequate gluconeogenesis secondary to multiorgan dysfunction.^9,14,16,18  This theory is supported by previous observations that markers of liver injury are more prominent in hypoglycemic neonates, which was also the case in our cohort.¹⁶  A decreased ability to mobilize glycogen stores or undergo gluconeogenesis may result in decreased ability to compensate during transition from fetal to neonatal life, exacerbating any accompanying hypoxic-ischemic injury. Furthermore, hypoglycemia is associated with an injury pattern on magnetic resonance imaging that is consistent with prolonged and partial ischemic insults, a higher rate of neurodevelopmental impairment in survivors, as well as concern for less effectiveness of TH in this population, which may further hint at a subacute and prolonged injury underlying the acute presentation.^7,9,13,14,16  Hypoglycemia in neonates with moderate or severe HIE often presents in conjunction with other markers of end-organ injury, indicating a multiorgan insult that therefore might suggest that hypoglycemia can be seen as a marker of severity of injury. For example, several studies found that hypoglycemia was no longer significantly associated with adverse outcome in HIE after adjusting for severity of illness.^13,19  However, previous studies had significantly smaller sample sizes and were conducted partially before the era of TH. In the current study, we adjusted for markers of severity of illness including degree of encephalopathy and organ dysfunction, and the results indicate that hypoglycemia appears to be independently associated with an increased adverse long-term outcome compared with euglycemic newborns, but to a lesser degree than those who experience hyperglycemia.

The relationship between hyperglycemia and outcomes in neonates with HIE has not been widely studied yet, possibly because the commonly used definition of hyperglycemia may have resulted in inclusion of a significant proportion of glucose levels in the 150- to 200-mg/dL range, for which pharmacological interventions such as insulin are rarely indicated. We defined hyperglycemia as >200 mg/dL according to clinical significance determined by visual inspection of continuous LOESS curves of average highest glucose and proportion of infants experiencing death/NDI. Using this clinically meaningful cutoff, we show that hyperglycemia is not only associated with a significantly higher rate of death but also that more severely affected newborns have higher median blood glucose levels than their less-compromised peers.

Hyperglycemia in the early stages of asphyxia may reflect depressed cerebral glucose uptake and acute ongoing utilization of adenosine triphosphate and H⁺ release as suggested by lower measured pH values.^9,20  The hypothesis that hyperglycemia may reflect a very acute and recent insult is supported by a study in piglets where glucose levels increased threefold immediately following asphyxia.¹⁹  These findings are similar to those seen after cardiac arrest or traumatic brain injury in animals, adults, and pediatric patients in which hyperglycemia has been associated with a more profound brain injury.^21–25  Early hyperglycemia has also been more commonly reported in cases of HIE with an identified acute sentinel event, which appeared to be the case in our cohort, in which 35% of hyperglycemic infants had a documented sentinel event compared with 28% of euglycemic and 22% of hypoglycemic infants.⁹  Additionally, the post hoc analysis of the CoolCap trial demonstrated that TH was more effective in hyperglycemic infants, supporting the theory that hyperglycemia is associated with more recent acute asphyxial events.⁹ 

This study has several limitations. The lack of a standard approach to glucose sampling could have caused sampling bias driven by extreme values resulting in more frequent sampling. The HEAL dataset did not include daily markers of liver and kidney function, and although we adjusted for a number of factors reflecting severity of illness, residual confounding may still be present. Additionally, the observational nature of this study precludes any conclusions regarding whether correction of either hypo- or hyperglycemia would improve outcomes. A significant strength of this study is that the HEAL Trial dataset is the largest cohort with standardized follow-up available to elucidate the potential causation and impact of deranged glycemic states in neonates with moderate or severe HIE undergoing TH. In contrast to many of the published studies, we had the opportunity to established cutoffs of hypo- and hyperglycemia that appeared to be more directly associated with outcomes and to adjust for severity of illness to eliminate its potential confounding effect on the glycemic state. As a result, we were able to show that both hyper- and hypoglycemia are independently associated with adverse long-term outcomes.

Glycemic profiles during the first 12 hours after birth differ between neonates with moderate and severe HIE. Both hypo- and hyperglycemia were associated with an increased risk of death after adjusting for severity of illness. Among survivors, hypoglycemia was also independently associated with an increased risk of NDI at 22 to 36 months of age.

aOR

adjusted odds ratio

BG

blood glucose

CI

confidence interval

GMFCS

Gross Motor Function Classification System

HEAL

High-dose Erythropoietin for Asphyxia and Encephalopathy

HIE

hypoxic-ischemic encephalopathy

LOESS

locally estimated scatterplot smoothing

NDI

neurodevelopmental impairment

TH

therapeutic hypothermia