Hypothermia for Hypoxic-Ischemic Encephalopathy: Second Generation Trials to Address Gaps in Knowledge

Summary:

Hypothermia treatment for moderate or severe HIE represents a landmark therapy because it is the first and only treatment to date that has been demonstrated to reduce mortality and improve neurodevelopmental outcome. Randomized trials of hypothermia have laid the groundwork for implementation and dissemination of this treatment. The success of hypothermia has triggered important questions which were not studied in the initial randomized trials. This article has summarized and discussed 3 randomized controlled trials for moderate or severe HIE to address the use of longer and/or deeper cooling, initiation of hypothermia after 6 hours of age, and use of hypothermia in preterm neonates. The results of these trials expand the evidence base available to clinicians when using hypothermia to treat moderate or severe HIE.

Introduction

Prior to 2005 management of hypoxic-ischemic encephalopathy (HIE) among term newborns was limited to intensive supportive care without a specific brain-oriented therapy. Between 2005 and 2011 six randomized trials demonstrated that hypothermia for newborns with moderate or severe HIE reduced death or disability or increased survival without neurologic abnormality at 18–22 months of age^1–6. These trials were remarkably homogenous in many key characteristics such as post-natal age, a stepwise process for inclusion, randomization to cooling or usual care, duration of cooling, rate of rewarming, and a similar primary outcome. Almost all enrolled infants were ≥36 weeks gestation and either body cooling (target core temperature, Tc, 33.5°C) or head cooling combined with a more modest reduction in Tc (target Tc, 34.5°C) was used. A meta-analysis based on results from these 6 trials and several smaller studies (1344 infants analyzed) indicated a 25% relative risk reduction with hypothermia compared to usual care with normothermic temperatures⁷.

Based upon these studies, the Eunice Kennedy Shriver National Institute of Child Health and Development (NICHD) convened two workshops to identify knowledge gaps for treatment of HIE with hypothermia^8,9. Although the initial trials demonstrated a reduction in death or disability at 18–22 month follow-up, 40–50% of newborns treated with hypothermia still died or survived with serious neurological disability. The workshops established multiple research priorities, one of which was refinements of hypothermia therapy for newborns with HIE. As a response to these workshops, the NICHD Neonatal Research Network (NRN) performed 3 randomized trials to address selected gaps in our understanding of the use of hypothermia treatment (second generation trials). These trials include the Optimizing Cooling Trial (NCT01192776), the Late Hypothermia Trial (NCT00614744) and the Preterm Hypothermia Trial (NCT01793129) and each will be reviewed in this article.

Optimizing Cooling Trial

Rationale:

The initial trials of hypothermia for moderate or severe HIE used a Tc of either 33.5°C for whole body cooling or 34.5°C for head cooling over 72 hours^1–6. These cooling parameters were based on preclinical investigation and pilot human studies and reflected the best estimate of conditions to provide neuroprotection and minimize potential adverse effects. However, multiple observations suggested that neuroprotection could be enhanced by modifying the cooling regimen. Specifically, cooling to a depth >3–6°C below baseline in preclinical studies of timed hypoxia-ischemia demonstrated that cooled animals compared to those maintained normothermic had more histopathological protection¹⁰, less brain swelling, greater preservation of energy metabolism,¹¹ differential neuroprotection in white and gray matter¹², and less reduction in secondary energy failure¹³. In addition, longer cooling was protective against cell necrosis and inflammation initiated during reperfusion and evolving over time^14,15. Two clinical reports published during the conduct of the Optimizing Cooling (OC) trial are pertinent to the rationale for OC. Non-invasive measurement of brain temperature indicated that neonates with severe HIE had higher brain temperatures than those with moderate HIE during and after hypothermia at 33.5°C for 72 hrs¹⁶. In a second report, despite the expected decrease in brain temperature among neonates with HIE during the first day of cooling, a subsequent increase of brain temperature was noted among those who developed injury on imaging¹⁷.

Trial design:

Based on the above observations, the NICHD NRN undertook the Optimizing Cooling (OC) randomized controlled trial (RCT) to determine the impact of longer and/or deeper cooling on death or disability of infants with moderate or severe HIE^18,19. Inclusion and exclusion criteria of the OC RCT were similar to the prior NRN RCT² except that infants who were overcooled at screening (Tc < 32.5°C for at least 2 hrs.) were excluded. Neonates ≥36 weeks gestational age, with severe birth acidosis or need for resuscitation and moderate or severe encephalopathy diagnosed by trained and certified physician examiners, were enrolled within 6 hours of age. In a 2×2 factorial design, neonates were randomly assigned to one of 4 hypothermia groups, 33.5°C for 72 hrs., 33.5°C for 120 hrs., 32.0°C for 72 hrs. and 32.0°C for 120 hrs. The cooling and rewarming procedure was similar to the prior NRN trial², except that, to prevent over cooling for the 32.0°C group, the Tc (esophageal) was initially lowered to 33.5°C and once stable for 15 minutes, further lowered to 32.0°C.

The primary outcome was death or moderate or severe disability at 18 to 22 months of age¹¹. Death was included as it is a competing outcome for disability. Neurodevelopmental outcomes were assessed by trained examiners, certified to reliability on an annual basis, and were unaware of treatment status. Severe disability included any of the following: Bayley Scales of Infant Development III cognitive score <70, a Gross Motor Function Classification System (GMFCS) level of 3 to 5, or blindness or hearing loss despite amplification. Moderate disability was defined as a cognitive score of 70 to 84 and either GMFCS level 2, active seizures, or hearing loss requiring amplification. A sample size of 363 neonates was required for marginal comparison of usual depth (33.5°C) to deeper cooling (32.0°C) and 363 neonates needed for the marginal comparison of usual duration (72 hrs.) to longer cooling (120 hrs., total 726 neonates). The sample size was based on a 2-tailed α=.05, a power of 80%, a 5% loss to follow up to detect a 10% difference in death or disability from 37.5% to 27.5% in the 2 depths and cooling duration groups. The analyses accounted for differential exposure periods (72 hrs. and 120 hrs.), controlled for clinical center and level of HIE, and examined interactions between deeper and longer cooling. Prespecified secondary outcomes included mortality, level of disability by stage of encephalopathy, rates of vision, hearing, and multiple disabilities, cognitive and motor scores, cerebral palsy, rehospitalizations and growth measurements. Interactions between the 2 depths and durations of cooling were assessed for the primary outcome and key secondary outcomes. Pre-defined severe adverse events included cardiac arrythmia, persistent acidosis, major vessel thrombosis and bleeding. The independent data safety and monitoring committee (DSMC) serially evaluated safety after trial initiation in October 2010.

Results:

NICU Mortality and Safety:

The NICU mortality rate was 7% for the 33.5°C for 72hrs. group (n=95), 14% for the 32.0°C for 72hrs. group (n=90), 16% for 33.5°C for 120 hrs. group (n=96) and 17% for 32.0°C for 120 hrs. group (n=83). The mortality rates were not different between the cooling depths, 33.5°C (12%) and 32.0°C (12%) and the cooling durations, 72 hrs. (11%) and 120 hrs. groups (16%), respectively. There was no interaction between depth and duration of cooling and mortality adjusted for center and level of encephalopathy. Safety outcomes were similar between the 2 cooling depths and the 2 cooling durations, except for the following: major bleeding occurred in 1% of the 120 hrs. group vs 3% in the 72 hrs. group. The incidence of arrhythmia and anuria was higher and hospital stay was longer for the 120 hrs. group vs the 72 hrs. group. During the NICU course, a higher incidence of bradycardia, inhaled nitric oxide and ECMO use, and more days of oxygen therapy was observed in the 32.0°C group compared with the 33.5°C group. Futility analysis determined that the probability of detecting a statistically significant benefit for longer cooling, deeper cooling, or both for NICU survival was less than 2%. Following the eighth evaluation (in November 2013) the trial was halted because of mortality and safety concerns after 364 neonates had been enrolled, 185 to the 72 hours group, 179 to the 120 hours group, 191 to the 33.5°C group and 173 to the 32.0°C group¹⁰.

Outcome at 18–22 months:

Outcome assessments were completed in January 2016, with data available on 347 of 364 infants (95%)¹⁹. The primary outcome of death or moderate or severe disability was not different between the 2 depth or the 2 cooling duration groups (Tables 1 and 2). The primary outcome was also similar between the 2 depth and the 2 cooling duration groups for infants with moderate or severe HIE at randomization. The secondary outcomes were not different between groups except among infants in the 2 cooling duration groups; there were more deaths (19 vs.13%, p=0.04), fewer rehospitalizations after discharge (16 vs. 29%, p=0.02) and fewer infants with motor scores <70 (13 vs. 19%, p=0.04) with 120 hours vs 72 hours cooling. The latter reflects lower survival with longer cooling compared to usual duration of cooling. There was an interaction between depth and duration of cooling and the primary outcome (p=0.048).

Table 1:

Outcomes at 18–22 months of age by the depth of cooling

	Esophageal Temp 33.5°C	Esophageal Temp 32.0°C	Adjusted risk ratio (95% CI)a
Death or moderate or severe disability	59/185 (32%)b	51/162 (33%)	0.92 (0.68–1.26)
Death	26/185 (14%)	30/162 (19%)	1.17 (0.67–2.04)
Moderate or severe disability	33/159 (21%)	21/132 (16%)	0.71 (0.36–1.39)
Cerebral Palsyc	26/158 (16%)	21/132 (16%)	0.94 (0.51–1.73)

^aAll models were adjusted for center and severity of encephalopathy. Risk ratios were derived using the 33.5°C group as the reference.

^bResults are the number of neonates affected divided by the total and expressed as a percent

^cCerebral palsy of any severity

Table 2:

Outcomes at 18–22 months of age by the duration of cooling

	Duration 72 hours	Duration 120 hours	Adjusted risk ratio (95% CI)a
Death or moderate or severe disability	56/176 (32%)b	54/171 (32%)	0.92 (0.68–1.25)
Death	23/176 (13%)	33/171 (19%)	1.39 (1.02–1.90)
Moderate or severe disability	33/153 (22%)	21/138 (15%)	0.68 (0.41–1.11)
Cerebral Palsyc	29/153 (19%)	18/138 (13%)	0.65 (0.36–1.16)

^aAll models were adjusted for center and severity of encephalopathy. Risk ratios were derived using the 72 hrs. group as the reference.

^bResults are the number of neonates affected divided by the total and expressed as a percent

^cCerebral palsy of any severity

The rates of the primary outcome among the 4 hypothermia groups were 29.3% in the 72 hrs. at 33.5°C group, 34.5% in 72 hrs. at 32.0°C, 34.5% in 120 hrs. at 33.5°C and 28% in the 120 hrs. at 32.0°C group. Pre-planned Bayesian analyses indicated that the probability of increasing death with deeper cooling (32.0°C for 72 hr.), longer cooling (33.5°C for 120 hr.), or both (32.0°C for 120 hrs.) compared with usual care cooling (33.5°C for 72 hrs.) was 66%, 93% and 89%, respectively (see Late Hypothermia analytic plan regarding Bayesian analysis). The OC trial provides important detail regarding early childhood outcomes for infants cooled with usual parameters that are currently used (33.5°C for 72 hrs.), and maybe helpful for clinicians when counseling families (Table 3).

Table 3:

Outcomes at 18–22 months of neonates cooled to a depth of 33.5°C for 72 hr.

	No/Total	Percent
Death or moderate or severe disability	27/92	29
Death or moderate/severe disability
Among neonates with moderate HIE	14/71	20
Among neonates with severe HIE	13/21	62
Death	8/92	9
Among neonates with moderate HIE	4/71	6
Among neonates with severe HIE	4/21	19
Among survivors:
Severe disability	18/84	21
Moderate disability	1/84	1
Mild disabilitya	16/81	20
No disabilityb	46/81	57
Any Cerebral palsy	16/84	19
Disabling cerebral palsyc	14/84	17
Bayley III cognitive scoresd
Median, interquartile range	90 (80–100)
≥ 85	51	65
70–84	14	18
< 70	13	17
Rehospitalization after discharge	16/80	36
Height < 10%	19/75	25
Weight < 10%	7/78	9
Head Circumference < 10%	15/76	20

^aMild disability: cognitive score of 70–80 or a cognitive score ≥ 85 and any of the following: GMFCS level 1 or2, seizure disorder, or hearing loss not requiring amplification

^bNo disability: cognitive score ≥ 85 and absence of neurosensory deficits and seizures

^cDisabling cerebral palsy: moderate or severe with infants requiring support for sitting, unable to crawl or requiring adult assistance to move

^dBayley III cognitive score normative data, mean±sd. 100±15; data based on 78 infants due to loss to follow-up and inability to test selected infants

Conclusions:

The OC trial confirmed that deeper and longer cooling was not superior to the usual depth of cooling (33.5°C for whole body cooling) and the usual duration of cooling (72 hrs.). Although the trial may have been underpowered for determination of the primary outcome due to the early closure, the results are consistent with preclinical studies conducted after initiation of the OC trial that demonstrated longer and deeper cooling were not neuroprotective^20–22.

An interesting observation was that the rate of death or moderate/severe disability in the OC trial for infants cooled with usual parameters (33.5°C for 72 hrs.) was 29%. This is considerably lower than death or disability after hypothermia treatment in multiple initial cooling trials (eg., 44% in the NICHD whole body cooling², 45% in the TOBY trial³, 51% for the neo.nEURO.network⁴). An important contributing factor was the lower rate of severe encephalopathy among infants randomized in the OC trial compared to earlier trials. Other variables include earlier recognition of at-risk neonates, improved stabilization, earlier initiation of hypothermia treatment, and more intensive neuromonitoring²³.

Without the OC trial, enthusiastic clinicians may reason that deeper or longer cooling may provide more neuroprotection than presently achieved with Tc of 33.5°C maintained for 72 hrs. The results of the OC trial emphasize the importance of avoiding drifts in practice and adhering to established protocols based on randomized trials.

Late Hypothermia Trial

Rationale:

A key element of hypothermia therapy for neonatal moderate or severe HIE is initiation within 6 hours following birth. The 6-hour interval represents a putative therapeutic window during which hypothermia can be implemented and still favorably alter outcome. The strongest support for a 6-hour therapeutic window is well-done preclinical fetal sheep studies in which 72 hours of hypothermia was initiated at either 1.5, 5.5 or 8.5 hours after brain ischemia, and histologic injury was compared to sham controls^24–26. Neuronal loss was reduced only with the 2 earlier initiation times. Although neuronal loss scores were not lower when hypothermia was initiated at 8.5 hours (p=0.11), the results were based on a limited sample size (5 hypothermic animals) and cannot exclude the possibility of a longer therapeutic window²⁶. Furthermore, direct extrapolation from animals to newborns may be limited given variables of the insult in the clinical setting that cannot be easily replicated in the laboratory (e.g., duration, repetitiveness, severity of hypoxia-ischemia). There is also strong clinical rationale for study of hypothermia beyond 6 hours of age. First, encephalopathy may evolve from absent or mild in extent at < 6 hours to moderate or severe after 6 hours. Second, preclinical studies employ well timed hypoxic-ischemic events and the assumption in clinical medicine is that hypoxia-ischemia occurs near or at delivery. However, the frequency of sentinel events at delivery (e.g., abruptio placenta, ruptured uterus, prolapsed cord etc,) was approximately 60% in one trial²⁷ and leaves many neonates with an unclear timing of hypoxia-ischemia. It seems possible that hypoxia-ischemia in some neonates may have its onset hours before birth, and hypothermia treatment initiated within 6 after birth may represent a much longer interval from event to treatment. Third, hypothermia treatment may not be initiated within 6 hours of age due to births in remote locations, although controlled cooling on transport may diminish this issue.

Trial design:

The Late Hypothermia trial was designed to estimate the probability that hypothermia initiated at 6 to 24 hours after birth reduced the risk of death or disability at 18–22 months among neonates with moderate or severe HIE²⁸. This was a randomized trial conducted between April 2008 and June 2016 by 21 centers of the NRN. Inclusion criteria were infants ≥ 36 weeks gestation, either severe fetal acidemia or need for resuscitation at birth, and either moderate or severe encephalopathy or a seizure. Encephalopathy or seizure was either recognized between 6–24 hours of age or infants arrived at a referral center for cooling after 6 hours. Newborns were randomized to an esophageal temperature of 33.5°C or 37.0°C for 96 hours. A longer duration of hypothermia (96 vs 72 hrs.) was used based on preclinical data prior to 2008²⁹. The primary outcome was death or moderate or severe disability at 18–22 months similar to a prior NRN hypothermia trial².

Sample size and analytic plan:

A major challenge for this trial was the number of available patients. The sample size was predefined at 168 which was thought to be the largest number of neonates that could be enrolled in a feasible time interval (6 years) using prior experiences^{2, 30}. A Bayesian analysis was pre-specified given the limited sample size and power in conventional frequentist analyses to identify a treatment effect. A Bayesian analysis has been recommended for trials of rare conditions or uncommon features of previously studied disease processes³¹. A Bayesian analysis unlike a conventional frequentist analysis provides a direct assessment of the probability that the hypothesis is true based on the observed data and a prior distribution within a plausible range.

A hypothetical Bayesian analysis is illustrated in Fig 1. Distinct from frequentist analyses, a Bayesian analysis uses pre-existing data from trials to establish a prior distribution which represents the probability of a hypothesized treatment effect. If there is no available information to indicate benefit or harm from the intervention, a neutral prior can be used, centered at a risk ratio (RR) of 1 indicating a 50% probability of benefit (RR < 1.0 for a reduction in death or disability) and a 50% probability of harm (RR > 1.0). Alternatively, the prior can be centered at RR < 1.0 or > 1.0 if existing data suggests benefit or harm, respectively. The prior distribution is described by a 95% credible interval (95% probability that the true RR lies in this interval) reflecting the minimum and maximum likely treatment effects based on effect sizes identified for major outcomes in randomized trials³². In the Late Hypothermia trial, a neutral prior was used given the paucity of data for hypothermia initiated beyond 6 hours from birth. The prior distribution is combined with the observed data from the trial to yield a probability of a posterior distribution. The posterior distribution can be characterized by a point estimate, 95% credible intervals, and the area under the curve which is < RR of 1.0 (treatment benefit) and the area under the curve which is > RR of 1.0 (treatment harm). Analyses were adjusted (aRR, adjusted RR) for age at randomization (≤ 12h, > 12h) and level of encephalopathy (moderate, severe).

Figure 1:

A hypothetical Bayesian analysis is depicted. On the x axis is the risk ratio (RR) and the Y axis is the probability density which is the frequency distribution of observed values. For the Late Hypothermia trial, the non-cooled group is the reference and a RR < 1.0 is the desired outcome indicating a reduction in the outcome of interest in the hypothermic group. In this example, the prior distribution (dashed line) is shifted to the left of a RR of 1.0 reflecting encouraging pilot studies in this hypothetical scenario that suggest benefit from the intervention. The prior distribution is asymmetric since a RR < 1.0 can only go as low as zero but a RR > 1.0 can go to infinity. The curve would be symmetric if the x axis was on a log scale. When combined with the observed trial results, the posterior probability of treatment effect, indicated by the solid gray curve, is shifted to the right of a RR of 1.0. The area under the gray curve < or > a RR of 1.0 provides the posterior probability of benefit and harm, respectively. In this hypothetical case, most of the posterior probability distribution is a RR > 1.0 indicating a greater likelihood of harm than benefit.

Results:

The primary outcome is depicted in Figure 2. Death or disability occurred in 24.4% of the hypothermic group and 27.9% of the non-cooled group. Combining the trial results with a neutral prior shifted the distribution to a 76% posterior probability of any reduction in death or disability and a posterior aRR of 0.86 (95% credible intervals, 0.58–1.29). The probability of a treatment benefit for death alone, and survival with moderate or severe disability were 73% and 68% with a neutral prior, respectively. When considered as an absolute risk difference (hypothermia minus non-cooled), there was a 64% probability of at least a 2% absolute risk difference of less death or disability for hypothermia compared to non-cooled treatment. For an absolute risk difference of 2%, hypothermia was associated with a 3.2-fold lower risk of death or disability compared to non-cooled neonates assuming a range of risk differences around zero considered equivalent between groups.

Figure 2:

This figure depicts the results of the primary outcome for the Late Hypothermia trial. A neutral prior (dashed line) was used reflecting the absence of other trials examining hypothermia therapy for HIE after 6 hours of age. The neutral prior is asymmetric as described for Figure 1. When the observed results from the trial were combined with a neutral prior distribution, the posterior probability of treatment effect was shifted to the left of an adjusted RR of 1.0 (adjusted for stratifying variables of age at randomization and level of encephalopathy). The area under the curve < RR of 1.0, indicated by the gray portion of the posterior distribution, was a 76% probability of any treatment benefit. The area under the curve > RR of 1.0, indicated by the black portion of the posterior distribution, was a 24% probability of any treatment harm.

This figure has been adapted with permission from Figure 2A of Laptook A. et al, JAMA 2017;318(16):1550-1560. doi10.1001/jama.2017.14972.

Conclusions:

The priority for treatment of moderate or severe HIE with hypothermia is to initiate therapy as soon as possible after birth and within 6 hours of age. As discussed, there is strong biologic and clinical rationale to justify study of hypothermia after 6 hours of age. Bayesian analysis provides a probability of benefit or harm even with a limited sample size and results are easily interpretable by clinicians. A 76% probability of any reduction in death or disability, and a 64% probability of at least 2% less death or disability at 18 months with late hypothermia is suggestive but not conclusive evidence. Some clinicians may view the uncertainty as insufficient to justify use of late hypothermia. Alternatively, some clinicians may choose to use late hypothermia in the setting of moderate or severe HIE given the potential for death or a disastrous neurodevelopmental outcome, the absence of harm and lack of any other rigorously tested brain specific therapy.

Preterm Hypothermia

Rationale:

All 6 initial trials of hypothermia for HIE were limited to neonates ≥ 36 weeks gestational age (GA)^1–6 except for one study⁶. The latter trial and one pilot study included an unspecified number of neonates at 35 weeks GA.^{6,33, 34} Personal communication with those authors (SE Jacobs; DJ Eicher) revealed that there were only 7 neonates enrolled. Based on this group of studies, the Committee on the Fetus and Newborn (COFN) of the American Academy of Pediatrics recommended hypothermia for HIE for infants ≥ 35 weeks GA³⁵. Despite the lack of compelling data, many clinicians have adopted the COFN recommendation and provide hypothermia for neonates with HIE at 35 weeks. Furthermore, some centers have elected to provide hypothermia for neonates with HIE below 35 weeks fueled by multiple single center non-randomized experiences.^36–40

Irrespective of the COFN recommendation and ongoing clinical practice, there is a knowledge gap regarding the effectiveness of hypothermia treatment for moderate and severe HIE at ≤ 35 weeks gestation. Hypothermia may place preterm neonates at greater risk for initiation and/or exacerbation of prematurity-associated problems (e.g., intracranial hemorrhage, necrotizing enterocolitis, coagulopathy) compared to more mature neonates.

Trial Design:

To address this knowledge gap, the Eunice Kennedy Shriver NICHD NRN conducted a randomized, controlled trial to determine if hypothermia initiated at < 6 hrs. of age among neonates born at 33⁰-35⁶ weeks gestation with HIE would reduce death or moderate or severe disability at 18–22 months corrected age. This was a randomized trial conducted between July 2015 and December 2022 by 19 centers of the NRN.

Many features of this trial were identical or very similar to previous hypothermia trials conducted by the NRN, including 1) inclusion criteria (severe acidosis and/or perinatal sentinel event with resuscitation at birth, followed by the presence of moderate or severe encephalopathy using modified Sarnat score assessed by certified examiners or clinical seizures at <6 h), 2) device to achieve whole body hypothermia and monitor esophageal temperature (Tes), 3) randomization to hypothermia (target Tes 33.5°C) or normothermia (target Tes 37.0°C), 4) duration of intervention (72 hours), 5) re-warming rate (0.5°C/hour), 6) schedule of Tes surveillance during intervention and re-warming periods, and 7) the primary outcome, death or moderate or severe disability and its definition.^2,18,19,28

All neonates admitted to the Neonatal Intensive Care Units (NICU) of participating NRN centers at 33⁰ to 35⁶ wk GA and <6 hrs. postnatal age with a diagnosis of encephalopathy, perinatal asphyxia, neurologic depression, or birth depression were screened for eligibility. Because of concerns about prematurity associated changes in the neurologic exam and Sarnat scoring, an abnormal level of consciousness of moderate or severe degree was required. Similarly, criteria for posture and Moro were adjusted to account for maturational changes between 33–35 weeks GA.⁴¹ Some of the salient exclusion criteria included: 1) birthweight < 1500 grams, 2) temperature < 34°C for > 1 hr. at screening, 3) paralytic or sedative agents precluding a meaningful exam, and 4) presence of a major anomaly. Passive cooling during transport was discouraged. Because of the association of hyperthermia with adverse outcomes, steps were incorporated to prevent or limit hyperthermia > 37.5°C in the normothermia arm.^42–44 Thermal management before and after the intervention and re-warming period as well as other aspects of care such as laboratory and imaging surveillance, respiratory support, sedation, anticonvulsant therapy were per the standards at each center. Infants were not fed during the intervention and re-warming period.

Surveillance was conducted for pre-specified safety events as in previous NRN hypothermia trials. In addition, there was surveillance for events pertinent to prematurity (e.g., intracranial hemorrhage, necrotizing enterocolitis [NEC], days of mechanical ventilation). Cranial ultrasounds were to be obtained within 24 hours of randomization and brain MRI at 7–21 days of age. The Data Safety Monitoring Committee (DSMC) reviewed accumulating data at 6 intervals during the trial. After its third review of cumulative data on September 17, 2017, the DSMC requested additional measures to prevent or correct Tes < 32°C. The following steps were implemented: 1) change pre-cooling the blanket prior to the intervention from 5 to 15°C, 2) mmediate notification of research staff if Tes < 32.0°C, 3) Tes recorded every 10 minutes until target Tes attained, and 4) documentation of corrective actions.

Review of admission records of participating NICUs suggested that a maximum of 168 neonates could be enrolled in a 5-year period. Given the anticipated moderate number of subjects, Bayesian analysis was pre-specified. A neutral prior was planned due to absence of randomized studies of hypothermia in this population (see Late Hypothermia analytic plan regarding Bayesian analysis). The sample size was preplanned for 168. All analyses were intention to treat and were adjusted for level of encephalopathy and study center.

Results:

Of the 168 subjects enrolled, 88 were randomized to hypothermia and 80 to normothermia. Demographic and clinical characteristics of mothers and neonates were balanced between groups (Table 4). The average gestational age was 34 weeks and more than 50% of the neonates were outborn. Stabilization interventions at delivery, and the extent of encephalopathy (moderate/severe) at randomization were similar between groups.

Table 4:

Maternal and Neonatal Characteristics in the Preterm Hypothermia Trial

	Hypothermia (n=88) % or Mean ± SD	Normothermia (n=80) % or Mean ± SD
Maternal
Age, yrs.	30.9 ± 6.1	28.8 ± 6.4
Black	31	39
Pre-eclampsia/Hypertension	37	40
Diabetes	24	29
Abruption	41	46
Intrapartum Hemorrhage	20	20
Uterine Rupture	7	1
Intrapartum Temperature > 37.5 C	2	4
Fetal Deceleration	79	72
Cord Mishap	10	9
Emergent C-Section	77	84
Infant
Gestational – wk	34.0 ± 0.8	34.1 ± 0.8
Birth Weight – g	2464 ± 634	2371 ± 608
Male	52	56
Outborn	53	59
Delivery Room
Intubation	64	63
Chest Compressions	46	38
Epinephrine	30	33
Apgar Scores ≤ 5 at 10 minutes	51	43
pH, Cord or Postnatal blood gas < 1 hr.	6.90 ± 0.20	6.90 ± 0.20
Randomization
Age in hours	4.5 ± 1.2	4.5 ±1.3
Moderate Encephalopathy	69	71
Severe Encephalopathy	31	29
Seizures Alone	0	0

The primary outcome was determined beyond 22 months in 52 subjects due to Covid-19 center specific restrictions. Death or moderate/severe disability occurred in 29 of 83 (35%) children in the hypothermia group versus 20 of 69 (29.0%) in the normothermic group. The adjusted risk ratio (aRR; hypothermia/normothermia) using a neutral prior probability was 1.11 (95% credibility interval [CrI] 0.74 – 2.00), resulting in a 74% probability of harm (Table 5). When stratified by moderate versus severe encephalopathy at randomization, the primary outcome also revealed no benefit with hypothermia. Stratification by gestational age revealed a higher incidence of death or disability in hypothermic neonates and of death alone at each GA except for 33 weeks (Table 6).

Table 5:

Posterior Probabilities of the Primary Outcome and its Components, Death Alone and Survival with Disability: Neutral Prior

	Hypothermic (N=83)a		Normothermic (N=69)b		Bayesian Results
	n/N	%	n/N	%	aRR (95% credibility interval)	Probability of treatment harm
Death or moderate or severe Disability	29/83	35	20/69	29	1.11 (0.74–2.00)	74%
Death	18/83	22	9/69	13	1.38 (0.79–2.85)	87%
Survival with moderate or severe Disability	11/83	13	11/69	16	0.86 (0.46–1.63)	32%

^aThe primary outcome was available in 83 of 88 neonates randomized to the hypothermic group. Three neonates were lost to follow-up and 2 had incomplete follow-up

^bThe primary outcome was available in 69 of 80 neonates randomized to the normothermic group. Seven neonates were lost to follow-up, 2 had incomplete follow-up, and 2 were withdrawn (parental decision) immediately after randomization and prior to initiation of the intervention.

Table 6:

Treatment Comparison of Primary Outcome and Death Alone Stratified by Gestational Age Based on Best Obstetrical Estimate

Gestational Age	Death or Moderate/Severe Disability n/N (%)		Death n/N (%)
	Hypothermic (N=83)	Normothermic (N=69)	Hypothermic (N=85)	Normothermic (N=71)
33 weeks	13/31 (42)	6/17 (35)	5/31 (16)	4/19 (21)
34 weeks	9/24 (38)	10/32 (31)	8/26 (31)	2/32 (6)
35 weeks	7/28 (25)	4/20 (20)	5/28 (18)	3/20 (15)

Formal Bayesian analysis to determine aRR and probability of benefit/harm is not reported due to small numbers.

Death alone occurred in 18 (22%) and 9 (13%) neonates randomized to hypothermia and normothermia, respectively. The aRR for death alone was 1.38 (95% CrI 0.79–2.85) with probability of harm of 87% (Table 5). Three deaths occurred after discharge from the initial hospitalization, and all but two deaths occurred following decisions to re-direct care or withhold resuscitation. Death was attributed to asphyxial brain injury by site investigators in 15/18 (83%) among hypothermic and 5/9 (56%) normothermic neonates and to multi-organ failure in 2/18 (11%) and 2/9 (22%) neonates. Survival with moderate or severe disability occurred in 11 (13%) and 11 (16%) of the hypothermic and normothermic groups, respectively (aRR 0.86, 95% Crl 0.46, 1.63) yielding a 32% probability of harm.

There were no notable differences between groups in pre-specified safety events during the intervention (arrhythmia, persistent metabolic acidosis, major bleeding, skin alterations). Hyperglycemia (serum glucose >180 mg/dL) and thrombocytopenia (platelet count <100,000/mm³) were noted in >10% of neonates in both groups. Hyperglycemia occurred in 23% and 12% of the hypothermic and normothermic groups, respectively (aRR 1.64, 95% Crl 0.92, 3.27) yielding a 95% probability of harm. The number of neonates with intracranial hemorrhage during the intervention were 6 and 5 for hypothermic and normothermic groups, respectively. The frequency of intracranial hemorrhage was 8% in both groups. MRI results are pending. There were no cases of NEC and rates of pulmonary artery hypertension were 6% and 5% for hypothermic and normothermic groups, respectively.

Conclusions:

This study found no evidence of benefit from hypothermia at <6 hours among neonates 33⁰ to 35⁶ weeks gestation with moderate or severe HIE, unlike previous trials in neonates ≥ 36 weeks GA. The findings were consistent for both death or moderate/severe disability, and death alone. The 32% probability of harm for survival with moderate or severe disability among the hypothermia group needs cautious interpretation given the higher mortality in the hypothermic group. The sample size was limited but pre-specified at 168 reflecting the GA criteria over which a smaller fraction of births occurs compared to trials enrolling neonates at ≥ 36 weeks. A traditional frequentist analysis would require a larger sample size despite recruitment at 19 large academic centers that receive transports from a wide community net. Even with the limited sample size, recruitment and follow-up took 7.5 years to complete.

The reasons for lack of benefit from hypothermia for HIE among neonates born at 33⁰-35⁶ GA are unclear. There may be differences in the risk-benefit profile among less mature neonates. However, there were no differences in major bleeding, thrombocytopenia, intracranial hemorrhage, pulmonary hypertension or NEC compared to normothermic neonates. Interestingly, rates of pulmonary hypertension were lower than the 22% reported in two prior NRN hypothermia trials⁴⁵. Maternal complications of pregnancy and the in-utero environment may differ compared to hypothermia trials among more mature neonates. Specifically, maternal hypertensive disease and placental abruption were considerably higher compared to the few reports with similar data among more mature neonates.

Many centers provide hypothermia for neonates with HIE at 35 weeks gestation but the primary outcome and post-hoc analysis stratified by gestational age does not support benefit at any GA. Overall, the results support that targeted normothermia with interventions to prevent or minimize hyperthermia may be the most evidence-based approach for this group of neonates.

Key points:

• Use of deeper (32°C) and/or longer hypothermia (120h) for neonates with moderate or severe HIE does not improve outcome and may increase mortality.

• Hypothermia initiated after 6 hours of age for neonates with moderate or severe HIE may have a possible treatment benefit, but the results are not conclusive.

• Hypothermia initiated at < 6 hours of age in infants 33⁰-35⁶ weeks gestation with moderate or severe HIE does not decrease death or disability at 18–22 months corrected age.

Synopsis:

Multiple randomized controlled trials of hypothermia for moderate or severe neonatal hypoxic-ischemic encephalopathy (HIE) have uniformly demonstrated a reduction in death or disability at early childhood evaluation. These initial trials along with other smaller studies established hypothermia as a standard of care in the neonatal community for moderate or severe HIE. The results of the initial trials have identified gaps in knowledge needing further investigation. This article describes 3 randomized controlled trials of hypothermia (second generation trials) to address refinement of hypothermia therapy (longer and/or deeper cooling), late initiation of hypothermia (after 6 hours following birth), and use of hypothermia in preterm newborns.

Best Practices.

What is the current practice when using hypothermia treatment for neonatal moderate or severe hypoxic-ischemic encephalopathy?

• Hypothermia therapy should be used for treatment of HIE when there is a reasonable certainty of a hypoxic-ischemic event which occurred at or proximate to birth using inclusion criteria from randomized trials.

• Hypothermia should be initiated within 6 hours after birth but as early as feasible along with stabilization of cardiopulmonary, metabolic, and other critical processes.

• Cooling parameters for whole body hypothermia are a target core temperature (esophageal or rectal) of 33.5°C continued for 72 hours. Rewarming is typically done at 0.5°C/hour.

Recommendations:

• Adhere to hypothermia regimens that have been studied in rigorous randomized controlled trials.

• For neonates with moderate or severe HIE who meet criteria for hypothermia therapy but are beyond 6 hours of age, late hypothermia can be considered as an option.

• There is no benefit of using hypothermia for moderate or severe HIE among infants < 36 weeks gestation.