Whole Body Hypothermia for Neonatal Encephalopathy In Preterm Infants 33–35 Weeks Gestation: A Randomized Clinical Trial

Abstract

Importance

Hypothermia begun <6 hours after birth reduces death or disability in infants with encephalopathy due to hypoxia-ischemia at ≥36 weeks gestation. Trials of hypothermia for infants <36 weeks gestation are lacking.

Objective

To assess the probability that hypothermia at <6 hours after birth decreases death or disability in infants 33 to 35 weeks gestation with moderate or severe hypoxic-ischemic encephalopathy.

Design

A randomized clinical trial was conducted between July 2015 and December 2022 for infants 33 to 35 weeks gestation with moderate or severe hypoxic-ischemic encephalopathy at <6 hours after birth. Bayesian analyses and intention to treat were pre-specified.

Setting

Nineteen US Neonatal Research Network centers.

Intervention

168 infants received unblinded targeted esophageal temperature management. Eighty-eight hypothermic infants were maintained at 33.5°C (acceptable 33–34°C) for 72 hours and then re-warmed. Eighty normothermic infants were to be maintained at 37°C (acceptable 36.5–37.3°C).

Main Outcome and Measures

Composite of death or disability (moderate or severe) at 18 to 22 months corrected age adjusted for level of encephalopathy and center.

Results

Hypothermic and normothermic infants were preterm (mean [SD], 34.0 [0.8] and 34.1 [0.8] weeks gestation, respectively), while 46/88 (52%) and 45/80 (56%) were male. Randomization occurred at mean [SD] 4.5 [1.2] and 4.5 [1.3] hours for hypothermic and normothermic groups, respectively. The primary outcome occurred in 29/83 (35%) hypothermic and 20/69 (29%) normothermic infants (adjusted relative risk [hypothermic/normothermic] 1.11, 95% credibility interval 0.74–2.00) while death occurred in 18/88 (20%) hypothermic and 9/78 (12%) normothermic infants (adjusted relative risk 1.38, 95% credibility interval 0.79–2.85). Bayesian analysis with neutral prior indicated 74% probability of increased death or disability and 87% probability of increased death with hypothermia.

Conclusions and Relevance

Among infants 33 to 35 weeks gestation with hypoxic-ischemic encephalopathy, hypothermia at <6 hours age did not reduce death or disability at 18–22 months corrected age.

Introduction

Perinatal hypoxia-ischemia (HI) is a major cause of brain injury and death at all gestational ages (GA). The only effective treatment supported by multiple randomized controlled trials (RCTs) is therapeutic hypothermia for infants ≥36 weeks (wk) GA.^1–5 Despite minimal evidence for efficacy and safety at <36 wk, use of hypothermia in such infants has increased.^6,7 Multiple reports describe such experience without randomized controls.^8–11 Infants with GA <36 wk may be at increased risk for problems that may be triggered or respond adversely to therapeutic hypothermia (e.g., intracranial hemorrhage [ICH], necrotizing enterocolitis [NEC], coagulopathy, shock) as well as death.

We conducted a randomized trial to assess effectiveness and safety of therapeutic hypothermia in infants 33 to 35 wk GA. We hypothesized that therapeutic hypothermia (esophageal temperature [Tes] 33.5°C) for 72 hours (h) will decrease death or moderate/severe disability at 18–22 months corrected age in infants with moderate or severe encephalopathy due to HI at <6 h of age compared to infants treated with targeted normothermia (Tes 37.0°C)

Methods

Many features of this trial were similar to previous therapeutic hypothermia trials conducted by the Eunice Kennedy Shriver National Institute of Child Health and Human Development Neonatal Research Network (NRN).^2,12,13 All infants admitted to Neonatal Intensive Care Units (NICUs) of participating centers at 33–0/7 to 35–6/7 wk GA by best obstetrical estimate and <6 h age with a diagnosis of encephalopathy, perinatal asphyxia, neurologic depression or similar condition were screened for eligibility. Inclusion criteria required both: 1) blood pH (cord or neonatal at <1 hour) ≤7.0 or base deficit ≥16 mEq/L or, if no blood gas or lesser acidosis (pH 7.01–7.15 or base deficit 10.0–15.9 mEq/L), an acute perinatal event with either 10-minute Apgar score ≤5 or ventilation initiated at birth and continued ≥10 minutes; 2) moderate or severe encephalopathy using modified Sarnat score assessed by certified examiners, or clinical seizures at <6 h.^2,12,13 Clinical seizures did not require EEG confirmation. Because of concern about changes in Sarnat scoring attributable to prematurity, abnormal level of consciousness (moderate or severe) was required to be present. Similarly, criteria for posture and Moro were modified to account for maturational changes between 33 and 35 wk GA.¹⁴ Exclusion criteria included: 1) core temperature <34.0°C for >1h before screening, 2) receipt of paralytic or sedative agents obscuring Sarnat exam, 3) encephalopathy unlikely due to HI, 4) major anomaly, 5) moribund and not receiving intensive care, 6) birthweight <1500 gm, 7) clinician declined enrollment. Thermal management before randomization was per practice at each center. Passive cooling (i.e., withholding external heat) during transport was discouraged. Randomization was performed by telephone with the Research Triangle Institute (RTI) data center using computer-generated randomized permutated block algorithm with block sizes 2 and 4 in 1:1 ratio and stratified for encephalopathy level (moderate versus severe) and center.

After randomization, all infants underwent placement of a temperature monitoring probe in the distal esophagus. Placement time was considered time 0. By 108 hours, esophageal probes were removed and further thermal management implemented per local practice.

Those randomized to hypothermia underwent whole body cooling with Cincinnati Subzero Blanketrol Hyper-Hypothermia^R II or III device (CSZ). This device was used with an FDA Investigational Device Exemption (IDE) because of the study population GA. Target Tes was 33.5°C (33.0–34.0°C) for 72 h. Rapid cooling with this device is accompanied by transient Tes fall below the target (overshoot) followed by warming to return to target, then maintained. Overshoot occurs most often with cooling initiation. Tes in this group was recorded every 15 minutes for the first 4 h, every hour for the next 8 h and every 2 h for the remaining intervention period. Rewarming proceeded at 0.5°C/h.

Those randomized to normothermia had target Tes 37.0°C. Temperature monitoring was similar except for hourly temperatures during the first 4 h. Skin temperature associated with Tes 37.0° C for individual infants was identified and servo-controlled by radiant warmer or incubator to maintain Tes 36.5–37.3°C. Because of the association of hyperthermia with adverse outcomes, steps were incorporated to minimize hyperthermia in this group.^15–17 If Tes >37.3°C occurred, standard thermoregulatory management was verified. If Tes was >37.5°C., a single tepid sponge bath was implemented and if unsuccessful, active cooling with the CSZ and cooling blanket implemented until target Tes attained.

Cranial ultrasound was required within 24 h of randomization and brain MRI required at 7–21 days (d) postnatal age in survivors. MRI results will be reported separately. Laboratory and imaging studies were obtained per standard care. Infants were not fed during the intervention. Other aspects of management such as sedation/analgesia, respiratory support and anticonvulsant therapy were managed by local standards.

The primary outcome was death or disability (severe or moderate) at 18–22 months corrected age. Certified examiners trained to reliability and blinded to group assignment performed assessments at follow-up, including post-discharge history, growth, neurologic exam, Bayley Scales of Infant Development third edition (Bayley III), Gross Motor Function Classification System level (GMFCS), and vision and hearing status. Severe disability was deemed present by any of: Bayley III cognitive score <70, GMFCS 3–5, blindness, or hearing loss with inability to hear commands despite amplification. Moderate disability was defined by cognitive score 70–84 and any of GMFCS 2, treated seizure disorder, or hearing loss requiring amplification or implant to understand commands. Infants with cognitive score ≥85 and no deficits were considered normal.

Pre-specified secondary outcomes included death alone, severe or moderate disability only, death or profound disability (defined as severe disability with assignment of lowest possible cognitive score because infant untestable due to impairment), survival with normal outcome, each component of severe and moderate disability, and cause of death.

Study documents and consent forms were reviewed and approved by Institutional Review Boards at all participating institutions. Written informed consent by a parent or legal guardian was required.

Safety

Adverse events included arrhythmia requiring treatment, persistent metabolic acidosis, thrombosis, bleeding, altered skin integrity and death as in prior NRN hypothermia trials.^2,12,13 Additionally, intracranial hemorrhage (ICH); seizures after randomization; necrotizing enterocolitis (NEC) of Bell stage ≥II¹⁸; spontaneous intestinal perforation (SIP); major bleeding (prompting blood product administration); thrombocytopenia (<100,000/mm³); hypoglycemia (<30 mg/dL); hyperglycemia (>180 mg/dL); bronchopulmonary dysplasia (BPD); pulmonary hypertension (PPHN); late onset culture-proven bloodstream infection; esophageal probe injuries; abnormalities of electrolytes, calcium, phosphorus or magnesium; receipt of extracorporeal membrane oxygenation (ECMO) and other morbidities were prospectively tracked. The NRN Data Safety Monitoring Committee (DSMC) reviewed cumulative outcomes and adverse events at 6 specified intervals during the trial: after 20, 40, 60, 80, 100 and 130 infants reached NICU discharge, were alive in the NICU at 60 days or had died by 60 days.

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 blanket before intervention from 5 to 15°C, 2) immediate 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.

Sample Size and Statistical Methods

Intention-to-treat was pre-specified. CONSORT-C reporting guidelines were followed. Using retrospective review of NICU admission records at participating centers in 2012, the number of eligible infants for a 5-year period was estimated to approach 168. Simulations showed that the trial design had >75% chance of observing final posterior probability of ≤0.80 for RR<1, when true RR is near 0.70. The trial was designed using Bayesian principles to assess primary and other outcomes. The Bayesian approach assesses probability of treatment benefit/harm based on observed data and allows for formal inclusion of any prior data.^19,20 Binary outcomes were modeled using logistic regression, and predicted outcome probabilities were post-processed using the method of Gelman to estimate posterior distribution of relative risk (RR) and risk difference (RD).²¹ Models were adjusted for level of encephalopathy and center as random effects. Posterior distributions of adjusted risk ratio (aRR) and adjusted risk difference (aRD) were used to estimate 95% credibility intervals (CrI) and posterior probabilities of benefit/harm for outcomes. aRR was determined using hypothermia group results as numerator and normothermia as denominator; aRD was determined by subtracting normothermia results from those of the hypothermia group. Neutral prior for aRR (centered on 1.0, with 50% prior probability of better outcome and 50% of worse outcome) was pre-selected given no pre-existing trials for this intervention in the target GA range. Pre-specified assessments of enthusiastic (centered on aRR 0.75) and skeptical (centered on aRR 1.1) prior probabilities were also conducted. These priors and credibility intervals were based on largest effect sizes identified for major outcomes in randomized trials and exclude implausible effect sizes.²² For death or severe disability and other binary outcomes, the three priors were centered at −0.2877, 0, and 0.0953 respectively on the log odds ratio (OR) scale, with 95% credible intervals on that scale of (−1.39, 1.39) for neutral, (−1.67, 1.10) for enthusiastic, and (−1.29, 1.48) for skeptical. To produce neutral prior for aRR within a linear regression model, a normal prior with mean 0 and standard deviation 0.7072 was placed on the treatment parameter in the logistic regression. Probability of treatment benefit and harm was assessed by determining areas under the posterior probability distribution curve with aRR <1.0 and >1.0. All analyses were conducted in SAS 9.4 or R version 4.2.2 and all models were assessed for convergence using visual observation of trace plots within SAS and Gelman-Rubin statistics based on 3 Markov Chain Monte Carlo chains. Full Statistical Analysis Plan is in Supplemental files available online.

Results

From July 2015 to September 2020, 168 subjects were randomized (Figure 1). No subject qualified by clinical seizures alone. Among enrollees, 88 were randomized to hypothermia. One infant randomized to hypothermia did not receive that intervention due to late discovery of temperature <34°C for >1h before screening but was analyzed as receiving hypothermia per intention-to-treat. Of the 80 randomized to normothermia, 2 were withdrawn by parents immediately after randomization but before intervention and were not analyzed beyond features present at randomization.

Figure 1: CONSORT Flow Diagram.

^aOther reasons for non-eligibility included 14 with major anomaly, 12 unable to randomize by 6 hrs postnatal age, 7 with equipment/staff unavailable, 6 moribund and not receiving intensive care, 3 with temperature <34°C for >1 hr, 2 requiring ECMO, and 1 with no acute perinatal event

Demographic and clinical characteristics of mother and infant before randomization were similar between the groups (Table 1). Mean Tes of the two groups during the intervention demonstrates expected differences (Supplementary Figure 1). In the normothermia group, 16 infants met threshold for treatment of Tes >37.5°C.

Table 1.

Maternal and Neonatal Characteristics at Randomization^a

Maternal	Infant Hypothermic (N=88)	Infant Normothermic (N=80)
Age (years)	30.9 ± 6.1	28.8 ± 6.4
Married	43/88 (49%)	47/80 (59%)
Raceb
Black	27/88 (31%)	31/80 (39%)
White	52/88 (59%)	42/80 (52%)
Other	9/88 (10%)	7/80 (9%)
Hispanicb	15/88 (17%)	10/80 (12%)
Gravida (number of pregnancies)	3 (2, 5) n=87	2 (2, 4) n=80
Parity (number of births)	2 (1, 3) n=87	2 (1, 3) n=80
Education
High school or less	20/83 (24%)	22/73 (30%)
Any college or more	63/83 (76%)	51/73 (70%)
Pregnancy Complications
Preeclampsia/Hypertension	32/87 (37%)	31/78 (40%)
Antepartum hemorrhage	25/87 (29%)	21/79 (27%)
Thyroid dysfunction	4/87 (5%)	8/77 (10%)
Diabetes	21/88 (24%)	22/77 (29%)
Intrapartum Complications
Fetal decelerations	69/87 (79%)	57/79 (72%)
Cord mishap	9/86 (10%)	7/78 (9%)
Uterine rupture	6/86 (7%)	1/78 (1%)
Maternal fever (37.6 °C)	2/84 (2%)	3/75 (4%)
Placental problem (Any)	38/86 (44%)	36/78 (46%)
Abruption	35/86 (41%)	36/78 (46%)
Previa	1/86 (1%)	0/78 (0%)
Accreta	1/86 (1%)	0/78 (0%)
Maternal trauma	1/86 (1%)	5/79 (6%)
Maternal hemorrhage	17/85 (20%)	16/78 (20%)
Shoulder dystocia	3/86 (4%)	3/78 (4%)
Rupture of membranes (yes)	36/81 (44%)	20/75 (27%)
Duration, hr prior to delivery	4.2 (0.3, 27.6) n=32	1.5 (0.1, 13.3) n=19
≤ 18 hr	22/32 (69%)	16/19 (84%)
> 18 hr	10/32 (31%)	3/19 (16%)
Histologic Chorioamnionitis	11/59 (19%)	7/45 (16%)
Emergent cesarean delivery	68/88 (77%)	67/80 (84%)
Infant
Gestational age (weeks)	34.0 ± 0.8	34.1 ± 0.8
Birthweight (grams)	2464 ± 634	2371 ± 608
Length (cm)	46.0 ± 3.2 n=86	45.1 ± 2.8 n=76
Head circumference (cm)	32.0 ± 1.8 n=86	31.7 ± 1.8 n=77
Male	46/88 (52%)	45/80 (56%)
Outborn	47/88 (53%)	47/80 (59%)
Delivery Room Resuscitation
Intubation	56/88 (64%)	50/79 (63%)
Chest compressions	40/88 (46%)	30/79 (38%)
Epinephrine	26/88 (30%)	26/79 (33%)
Time to spontaneous breaths, minutes	2.0 (1.0, 3.0) n=80	2.0 (1.0, 3.0) n=74
Apgar Score <5
At 5 min	54/86 (63%)	48/79 (61%)
At 10 min	36/70 (51%)	29/67 (43%)
Cord blood or if unavailable, neonatal blood gas at <1 hour postnatal age
pH	6.9 ± 0.2 (n=69)	6.9 ±0.2 (n=68)
Base deficit	17.7 ± 7.0 (n=60)	17.0 ± 7.7 (n=59)
Age at randomization, hours	4.5 ± 1.2	4.5 ± 1.3
Level of Encephalopathy
Moderate	61/88 (69%)	57/80 (71%)
Severe	27/88 (31%)	23/80 (29%)
Clinical seizures at randomization	14/88 (16%)	11/80 (14%)

^aPlus-minus values are means ± SD. Interquartile ranges are separated by commas within parentheses below median values.

^bRace and ethnic group were reported by parent or guardian

The last patient completed follow-up in December 2022. Follow-up intended for 18–22 months was delayed in 52 infants due to Covid-related restrictions. In those cases, data from the earliest delayed follow-up visit were employed; mean ± SD corrected age for these infants was 26 ± 4 months (range 23–40). One infant whose age exceeded Bayley III exam limit was considered ‘lost to follow-up’.

The primary outcome of death or moderate/severe disability occurred in 29/83 (35%) infants randomized to hypothermia versus 20/69 (29.0%) normothermic infants (Table 2). aRR using neutral prior probability was 1.11 (95% CrI 0.74 – 2.00), yielding probability of benefit 26% and of harm 74% (Figure 2; see Supplementary Table 1 for analyses using skeptical and enthusiastic priors). aRR for death alone was 1.38 (95% CrI 0.79–2.85) with probability of benefit 13% and of harm 87%. Assessment for treatment heterogeneity with hypothermia revealed benefit probability 38% for males and 25% for females, while this probability was 35% for whites and 11% for blacks. Stratification by GA in exploratory analysis revealed higher incidence of primary outcome in hypothermic infants at each GA and of death at each GA except 33 wk (Supplementary Table 2).

Table 2.

Comparison of Primary and Secondary Outcomes in Infants Using Neutral Prior^a

	Hypothermia Group (N=88) n/N (%)	Normothermic Groupb (N=78) n/N (%)	Bayesian Effect	Medianc	95% CrId	Posterior Prob of Benefite
Primary Outcome
Death or moderate or severe disability	29 / 83 (35)	20 / 69 (29)	aRDf	0.04	(−0.08,0.18)	26%
			aRRf	1.11	(0.74,2.00)	26%
Secondary Outcomes g
Any Death	18 / 83 (22)	9 / 69 (13)	aRD	0.05	(−0.05,0.26)	13%
			aRR	1.38	(0.79,2.85)	13%
Survival with moderate or severe disability	11/83 (13)	11/69 (16)	aRD	−0.02	(−0.15,0.09)	68%
			aRR	0.86	(0.46,1.63)	68%
Death or severe disability	27 / 83 (32)	20 / 69 (29)	aRD	0.02	(−0.11,0.15)	38%
			aRR	1.05	(0.67,1.82)	38%
Death or moderate or severe disability with initial moderate NE	9 / 56 (16.)	6 / 48 (12)	aRD	0.03	(−0.09,0.14)	32%
			aRR	1.18	(0.59,2.41)	32%
Death or moderate or severe disability with initial severe NE	20 / 27 (74)	14 / 21 (67)	aRD	0.06	(−0.14,0.26)	28%
			aRR	1.09	(0.81,1.53)	28%
Cause of death: Asphyxial brain injury	15 / 18 (83)	5 / 9 (56)	aRD	0.10	(−0.13,0.36)	18%
			aRR	1.17	(0.80,2.19)	18%
Cause of death: Multi-organ failure	2 / 18 (11)	2 / 9 (22)	aRD	−0.04	(−0.27,0.15)	69%
			aRR	0.8	(0.32,2.04)	69%
Clinical seizures after randomization	13/86 (15)	11/75 (15)	aRD	0.0	(−0.12, 0.13)	47%
			aRR	1.02	(0.55, 1.92)	47%

^aModels adjusted for level of encephalopathy and included center as a random effect.

^bExcluded two withdrawn infants.

^cPosterior median of probability distribution

^dPosterior 95% credible interval

^ePosterior Probability of Benefit due to whole body hypothermia. For Relative Risk this is Pr(RR < 1 | Trial Data), and for Risk Difference this is Pr(RD < 0 | Trial Data). For relative risk, the comparison is whole body hypothermia over normothermia, so relative risk values less than 1 indicate benefit for whole body hypothermia. For risk difference, the comparison is whole body hypothermia minus normothermia, so risk difference values less than 0 indicate benefit for whole body hypothermia.

^faRD = adjusted risk difference; aRR = adjusted risk ratio

^gAlthough ‘profound disability’ was pre-specified as a secondary outcome of interest, only 1 infant (normothermic) met criteria and results are therefore not included in the Table.

NE, neonatal encephalopathy

Figure 2.

Posterior probability of relative risk for primary outcome (relative risk = probability with hypothermia/ probability with normothermia)

In exploratory analysis, it was noted that 32 infants randomized to hypothermia and 1 to normothermia attained Tes<32°C. during the intervention (with duration <1 hour in 26). Exclusion of all such infants (Supplementary Table 3) revealed no clinically important difference between groups for the primary outcome or death alone.

Death was attributed by site investigators to asphyxial brain injury in 15/18 (83%) hypothermic and 5/9 (56%) normothermic infants. Multi-organ failure was considered cause of death in 2/18 (11%) hypothermic and 2/9 (22%) normothermic infants. Cause of death in 2 remaining normothermic infants were pulmonary hypoplasia with chronic pulmonary hypertension and severe bronchopulmonary dysplasia, while cause in the remaining hypothermic infant was cardiomyopathy. All but 2 deaths followed decisions to re-direct care or forego resuscitation. Three deaths (2 hypothermia, 1 normothermia) occurred following NICU discharge, including one discharged on hospice care.

The frequency of pre-specified non-death safety events during intervention was generally comparable in the two groups (Table 3). The posterior probability that hypothermia was beneficial in reducing these events ranged from 25 to 81%, except for hyperglycemia (5%), hyponatremia (1%), days on mechanical ventilation (1%) and hypoglycemia (87%). Details of intracranial hemorrhage on ultrasound before, during and after the intervention during NICU course are in Supplementary Table 4. Two normothermic infants had intestinal perforation after intervention, one due to spontaneous intestinal perforation and one due to previously undiagnosed ileal atresia.

Table 3.

Neonatal Safety Events During Intervention Period (Relative Risk: Hypothermic/Normothermic Groups)

Adverse Event During Intervention	Hypothermic (n=88) n / N (%)	Normothermic (n=78) n / N (%)	Posterior Prob of Benefit	Posterior aRR (95% CrI)
Arrhythmia needing treatment	1 / 88 (1.1)	1 / 78 (1.3)	54%	0.95 (0.33,2.73)
Persistent metabolic acidosisa	4 / 88 (4.5)	5 / 78 (6.4)	65%	0.84 (0.35,1.99)
Thrombosis	0/88 (0)	0/78 (0)	Cannot Be Estimated	Cannot Be Estimated
Intracranial bleedingb	6/88 (8)	5/78 (8)	53%	1.03 (0.46, 2.33)
Major Bleeding	1 / 88 (1.1)	4 / 78 (5.1)	81%	0.65 (0.24,1.69)
Thrombocytopenia	11 / 88 (12.5)	12 / 78 (15.4)	67%	0.87 (0.44,1.67)
Treatment with vasopressors	28/88 (32)	23/78 (30)	43%	1.03 (0.65,1.71)
Treatment with steroids	9/75 (!2)	7/70 (10)	38%	1.12 (0.54,2.38)
Oliguria/Anuriac	6/85 (7)	3/74 (4)	30%	1.24 (0.53,2.98)
Liver dysfunctiond	40/88 (46)	38/78 (49)	68%	0.94 (0.67,1.26)
Serum sodium <120 mEq/L	8/88 (9)	0/78 (0)	4%	2.25 (0.94,5.90)
Serum sodium >150 mEq/L	5/88 (6)	1/78 (1)	20%	1.51 (0.59,3.92)
Serum potassium <3.0 mEq/L	21/88 (24)	22/78 (28)	72%	0.87 (0.55,1.38)
Serum potassium >6.0 mEq?L	16/88 (18)	19/78 (24)	80%	0.80 (0.47,1.34)
Serum calcium <7.0 mg/L	29/82 (35)	26/71 (37)	57%	0.97 (0.66,1.43)
PPHN	5 / 88 (5.7)	4 / 78 (5.1)	48%	1.03 (0.43,2.48)
Hyperglycemia	20 / 88 (22.7)	9 / 78 (11.5)	5%	1.64 (0.92,3.27)
Hypoglycemia	1 / 88 (1.1)	5 / 78 (6.4)	87%	0.58 (0.22,1.48)
NEC	0/88 (0)	0/78 (0)	Cannot be Estimated	Cannot be Estimated
Esophageal probe issue	0/88 (0)	0/78 (0)	Cannot be Estimated	Cannot be Estimated
ECMO	0/88 (0)	0/78 (0)	Cannot be Estimated	Cannot be Estimated
Altered skin integrity
Erythema	4/88 (4)	1/78 (1)	25%	1.38 (0.53,3.73)
Sclerema	0/88 (0)	0/78 (0)	Cannot be estimated	Cannot be estimated
Cyanosis	1/88 (1)	0/78 (0)	Cannot be estimated	Cannot be estimated
Subcutaneous fat necrosis	0/88 (0)	0/78 (0)	Cannot be estimated	Cannot be estimated
For entire hospitalization f
Late-onset sepsise	1/88 (1)	1/78 (1)	55%	0.94 (0.32,2.71)
Days on oxygen	4 (0,29)	2 (0,87)	55%	0.98 (0.70,1.37)
Days on mechanical ventilation	3.5 (0,64)	2 (0,57)	1%	1.43 (1.08,1.90)
Length of hospital stay in survivors	22 (N=67) (7,73)	22.5 (N=60) (6,115)	32%	1.04 (0.87,1.24)

^aBlood pH <2 SD and base deficit >2 SD from time-specific values developing >3 hours after beginning of intervention period and persisting >3 hours.

^bIncludes intraventricular, parenchymal and cerebellar hemorrhage found on cranial ultrasound during or after intervention period, since cannot exclude possibility that hemorrhage detected after intervention actually occurred during the intervention.

^cOliguria <0.5 mL/kg/hour

^dLiver dysfunction any of aspartate aminotransferase >200 international units/liter, alanine aminotransferase>100 international units/liter or direct bilirubin >1.5 milligrams/deciliter

^ePositive Blood culture at >3 days of age

^fFor continuous outcomes below ‘late-onset sepsis’, columns under treatment groups list median (minimum, maximum) rather than frequency (%)

aRR, adjusted relative risk; CrI, credibility interval; PPHN, pulmonary hypertension; NEC, necrotizing enterocolitis; ECMO, extracorporeal membrane oxygenation

Discussion

Our findings indicate that therapeutic hypothermia initiated by 6 h postnatal age did not reduce the primary outcome of death or moderate/severe disability or death alone at 18+ months in infants born at 33 −35 wks GA with moderate or severe neonatal encephalopathy. A 74% probability of treatment harm for the primary outcome and 87% for death alone with hypothermia was observed. Other pre-specified safety events were generally comparable between groups. Although survival with moderate or severe disability appeared to be slightly better in the hypothermic group, the absolute difference was small (13 vs 16%) and this potential small benefit was accompanied by a much higher incidence of death.

Most studies demonstrating effectiveness of hypothermia for neonatal encephalopathy limited their population to ≥36 wk GA.^1–5 Two RCTs with follow-up at 12–24 months included enrollment at 35 wk but the publications did not specify how many such infants were included nor their outcomes.^23–25 The American Academy of Pediatrics Committee of Fetus and Newborn considered one of those trials in their 2014 recommendation for hypothermia in infants ≥35 weeks.²⁶ Personal communication in 2014 with the authors of those two RCTS revealed the total number of 35 wk GA infants in those studies to be 7, with 2 randomized to control (1 death, 1 normal survivor) and 5 to cooling (2 deaths, 2 survivors with moderate disability, 1 normal survivor). Unlike those two trials, our trial required abnormal level of consciousness and may select sicker patients. Our findings for 48 infants born at 35 weeks GA provide no support for hypothermia at that GA.

Reasons for absent benefit with hypothermia among preterm infants in our trial are unclear and may be multifactorial. First, developmental differences may alter risk-benefit balance with hypothermia compared to infants ≥36 wks GA. Second, antepartum environment and perinatal events may differ between preterm and term infants. Rates of placental abruption were more than 3-fold higher in the present study than in the nEuro (11.6% with ‘placental problems’ including abruption) and Late Hypothermia trials (11.3%).^5,12 Maternal hypertensive disorders were 2-fold higher than in the Late Hypothermia trial (17.8%).¹² Finally, protocol-directed correction of hyperthermia in the normothermic group may have altered outcomes. Non-cooled comparison groups from prior RCTs that experienced higher core temperatures were associated with increased risk of death or moderate/severe disability.^15–17 The effect of measures to avoid high Tes in our normothermic infants is unclear.

The observation that many infants randomized to hypothermia attained Tes<32.0°C. during the intervention is not surprising since preterm infants in general are known to have limited thermoregulatory ability. It is unclear if increased frequency of primary outcome and death alone seen in those infants reflects a direct effect of overshoot or if overshoot is simply a marker for infants who had sustained more severe hypoxic-ischemic insult and were at higher risk for adverse outcomes. Newer generations of whole body cooling devices are reported by manufacturers to have reduced likelihood of overshoot but it is unknown if that will yield any meaningful difference in outcomes. It is noteworthy in our study that if all infants with Tes<32.0°C. are excluded, there was still no notable benefit with hypothermia.

Prospective estimates of the incidence of neonatal encephalopathy among infants 33–35 wks GA are lacking. A single-center retrospective review of such infants who fulfilled inclusion criteria for a prior hypothermia trial estimated an incidence of 5 per 1000 live births.²⁷ Studies of neonatal encephalopathy at 33–35 wks GA will thus have limited sample size given the small number of births at these GAs compared to trials enrolling infants ≥36 wks even with multiple, large referral centers. Frequentist analysis with typical assumptions of types I and II error and outcome difference require a larger number of subjects than this multi-site study could accrue over 5 years. Bayesian analysis is particularly recommended for uncommon conditions to provide an estimate of the probability of benefit or harm due to an intervention based on trial results combined with pre-existing data.^19,20 Estimating the probability of benefit or harm may yield meaningful information to clinicians, patients and family.

This study offers significant information that may strongly influence clinical practice. It offers strengths of being a pragmatic randomized rrial conducted prospectively in 19 centers; well-defined inclusion and exclusion criteria; certified examiners to determine degree of encephalopathy at randomization; standardized interventions including steps to avoid confounding hyperthermia in the normothermic arm; systematic follow-up by certified examiners blinded to treatment assignment; and use of intention-to-treat design. Limitations include moderate sample size; inability to assign primary outcome to 11% of normothermic and 6% of hypothermic infants; loss to follow-up of more patients in the normothermic than hypothermic group; that intervention was unblinded to clinicians, although follow-up investigators were unaware of treatment assignment; and the fact that 47 infants were excluded from the study by clinicians opting for hypothermia without randomization (mean GA ± SD 34.8 ± 0.6 wk) .

Our trial provided evidence for no benefit from hypothermia in infants 33 to 35 wk GA with moderate or severe hypoxic-ischemic encephalopathy. Indeed, our findings suggest that it may increase death or impairment. In the absence of strong supportive evidence from future trials, use of hypothermia is not indicated in infants born with hypoxic-ischemic encephalopathy at ≤35 wk GA.

Key Points.

Question

Does hypothermia initiated at <6 hours after birth reduce the probability of death or disability at 18–22 months corrected age in infants 33 to 35 weeks gestation with neonatal encephalopathy due to hypoxia-ischemia?

Findings

In this Bayesian randomized clinical trial of 168 newborns of 33 to 35 weeks gestation with hypoxic-ischemic encephalopathy, treatment with hypothermia resulted in a 74% probability of increased death or disability and 87% probability of increased death at 18–22 months corrected age.

Meaning

This trial provided no evidence that hypothermia begun at <6 hours after birth in infants 33 to 35 weeks gestation with hypoxic-ischemic encephalopathy decreases death or disability at 18–22 months corrected age.