Hypothermia for Neonatal Hypoxic-Ischemic Encephalopathy: Different Cooling Regimens and Infants Not Included in Prior Trials

Abstract

Perinatal hypoxic-ischemic encephalopathy is a neurological condition which is etiologically linked to death and neurodevelopmental impairment later in childhood. Until clinical trials of therapeutic hypothermia were conducted, treatment was limited to aggressive supportive care. Six large randomized clinical trials have established that therapeutic hypothermia reduced the composite outcome of death or life-long disability when assessed at 18 months of age. However, the results also indicate that there are many infants who do not benefit from cooling treatment. This chapter provides the rationale for continued investigation of therapeutic hypothermia. Randomized trials that address potentially better hypothermia regimens are reviewed. Knowledge gaps in the application of therapeutic hypothermia to other groups of newborns not included in prior trials (mild encephalopathy, preterm infants) are also reviewed.

Keywords

deeper cooling, late cooling, longer cooling, low-and middle-income countries, mild encephalopathy, preterm infants, therapeutic hypothermia

•

There is strong rationale for further investigation of therapeutic hypothermia.
•

A longer duration of cooling or a lower temperature for cooling when treating hypoxic-ischemic encephalopathy can be associated with harm.
•

There is biologic and clinical rationale for initiating therapeutic hypothermia after 6 hours of age and this deserves clinical testing.
•

Cooling on transport is effective when performed with a device that can control core temperature.
•

There is little data to indicate benefit from hypothermia for hypoxic-ischemic encephalopathy among with infants with a mild encephalopathy, preterm infants or infants in low or middle-income countries.

Therapeutic hypothermia is an effective therapy for neonatal encephalopathy when the likelihood of a hypoxic-ischemic origin is high. Multiple randomized trials demonstrated that relatively small reductions in core temperature either alone, or in combination with reduced head temperature, reduced death or disability at 18 months. The Cochrane meta-analysis indicates that therapeutic hypothermia reduced the composite outcome of death or major neurodevelopmental disability in survivors from 61% in noncooled infants compared with 46% in infants treated with hypothermia, yielding a risk ratio of 0.75 (95% confidence interval [CI] 0.68–0.83). Components of the primary outcome were also reduced by hypothermia; death was decreased from 34% to 25% (risk ratio 0.75, 95% CI 0.64–0.88), and disability was reduced from 24.9% to 19.2% (risk ratio 0.77, 95% CI 0.63–0.94). Disability was typically severe and could be any of cognitive, motor, or sensory deficits. Neuroprotective effects of therapeutic hypothermia persisted even at 6 to 7 years. The importance of this therapy extends beyond the benefits provided to infants and their families; it signifies that hypoxic-ischemic brain injury is modifiable and has accelerated testing other potential neuroprotective interventions either with or without therapeutic hypothermia. Given the beneficial effects of hypothermia, the Committee on the Fetus and Newborn of the American Academy of Pediatrics has provided an overview of the available data and expectations for centers that provide this therapy.

Hypothermia regimens have multiple components ( Box 4.1 ) that contribute to three phases of the therapy: induction, maintenance, and rewarming. Induction represents the time from initiation of cooling to reaching target temperature; maintenance represents the duration of keeping the infant at the target temperature; and rewarming represents the reestablishment of a normothermic temperature. The initial trials of hypothermia used remarkably similar cooling regimens. Specifically, the age of initiation was always less than 6 hours after birth, the extent of temperature reduction was 33.5°C for whole-body cooling and 34.5°C for head combined with body cooling, the duration of cooling was 72 hours, and the rate of rewarming was 0.5°C/hr. The similarity in hypothermia regimens facilitated meta-analyses of multiple trials to provide more accurate estimates of patient outcomes. The only major component of cooling regimens that differed among the initial trials was the mode of cooling: whole-body versus head with body cooling. Meta-analysis indicated that outcomes are similar irrespective of the mode of cooling.

Box 4.1

•

Age of initiation of cooling
•

Target temperature
•

Duration of cooling
•

Rate of rewarming
•

Mode of cooling

Components of a Therapeutic Hypothermia Regimen

Rationale for Further Investigations of Therapeutic Hypothermia

There are several justifications to perform further study of hypothermia even when the results of multiple trials demonstrated reductions in death or disability. First, the Cochrane meta-analysis indicates that 46% of infants treated with hypothermia either die or are diagnosed with moderate or severe impairment. There is obviously a need for improvement in outcome. Additional trials of hypothermia could be viewed as diverting valuable research resources from investigation of other potential neuroprotective agents, such as erythropoietin, xenon, and melatonin. This raises the second justification that refinement of the cooling regimen holds promise that outcomes could be further improved since hypothermia was studied as a package with little variability in the components of the regimen. Furthermore, many of the hypothermia components used in the first series of trials represented “a best estimate” based on animal investigation and pilot human studies. Third, the application of therapeutic hypothermia was limited to a specific group of newborns—that is, those with a diagnosis of hypoxia-ischemia presenting at less than 6 hours of age and with a gestational age of at least 36 weeks. These trials did not address other cohorts of newborns (preterm, mild encephalopathy, or cooling in low-resource countries).

This chapter addresses two major domains concerning therapeutic hypothermia. First, will there be clinical trials designed to improve the hypothermia regimen? Second, will there be clinical trials and observational studies that address extending hypothermia to infants not enrolled in the first series of hypothermia trials? These domains are consistent with the priorities for investigation of neuroprotective therapies for newborn infants published by the National Institute of Child Health Consensus Workshop on therapeutic hypothermia.

What Is the Optimal Temperature and Duration for Therapeutic Hypothermia?

The depth of temperature reduction used in the first series of hypothermia trials was extrapolated from preclinical investigation and pilot studies in newborns. Recognition that small changes in brain temperature modified the extent of hypoxic-ischemic brain injury in adult animals prompted perinatal animal investigations to examine the effects of depth and duration of “modest hypothermia” at varying times after brain hypoxia-ischemia. Modest hypothermia encompassed a range of temperature reductions from as little as 2°C to as much as 5°C using newborn swine, rat pups, and fetal sheep. This range reflected concerns regarding a possible trade-off between the potential benefit of lower temperature and adverse effects of hypothermia, which increase with greater reductions in core temperature. An interesting caveat in using animal investigation to plan clinical trials is that the basal temperature of newborn animals differs; for example, the resting temperature of newborn swine is 38.5°C to 39.0°C, whereas the nesting temperature of newborn rat pups is 36.0°C to 36.5°C. Comparison of the extent of neuroprotection associated with hypothermia using animals needs an awareness of the relative temperature reduction and the absolute temperature achieved. The optimal temperature to use for hypothermic intervention in clinical trials was not known but was guided by existing animal data and the assessment of incremental reductions in core temperature in pilot human studies of head cooling combined with body cooling and whole-body cooling. Based on this body of work, clinical trials of head cooling combined with body cooling and whole-body cooling alone were conducted at a rectal temperature of 34.5°C and an esophageal temperature of 33.5°C, respectively.

The optimal duration of hypothermia was also uncertain when the first series of clinical trials of hypothermia were undertaken in newborn infants. Available data were derived primarily from adult animals and indicated that increasing the duration of hypothermia reduced brain injury compared with shorter cooling intervals. Although not studied as extensively in newborn animals, 21-day-old rat pups subjected to hypoxia-ischemia had reduced brain injury when cooling was extended to 72 hours compared with 6 hours. Pilot studies in preparation to study hypothermia after brain ischemia in fetal sheep by Gunn et al. indicated rebound epileptiform activity when cooling was stopped after 48 hours but not observed if cooling was continued for 72 hours. Based on these series of studies, clinical trials of head cooling with mild body cooling and whole-body cooling used a 72-hour cooling intervention.

The absence of strong evidence for the depth and duration of cooling and the approximate 46% of infants with an outcome of death or disability despite hypothermia therapy provided unequivocal rationale for further investigation. The rationale became even stronger with a growing appreciation that the temporal profile of the pathways to injury is not confined to the early hours following hypoxia-ischemia but extends to days following the insult ( Fig. 4.1 ).

In response to these knowledge gaps, the National Institute of Child Health and Human Development (NICHD) Neonatal Research Network (NRN) conducted a randomized clinical trial of cooling to a lower temperature and for a longer duration (Optimizing Cooling Trial, NCT 01192776). The Optimizing Cooling Trial was a randomized 2 × 2 factorial design performed at 18 centers to determine if longer cooling (120 hours), deeper cooling (32.0°C), or both initiated before 6 hours of age are superior to cooling at 33.5°C for 72 hours among infants of at least 36 weeks’ gestation with moderate or severe hypoxic-ischemic encephalopathy (HIE) ( Fig. 4.2 ). The primary outcome was death or disability at 18 to 22 months adjusted for center and level of encephalopathy. The trial was closed to patient enrollment after 364 of a planned 726 infants were enrolled based on recommendations of an independent Data Safety Monitoring Committee. In-hospital mortality rates for cooling of 72 compared with 120 hours’ duration were 11% and 16%, respectively (adjusted risk ratio [aRR] 1.37, 95% CI 0.92–2.04). In-hospital mortality rates for cooling at 33.5°C compared with 32.0°C were 12% and 16%, respectively (aRR 1.24, 95% CI 0.69–2.25). Although not statistically different, the risk ratio and boundary of the 95% CI suggest that longer cooling maybe associated with an increase in mortality. Cooling for 120 hours was associated with more arrhythmias, anuria, and a longer length of hospital stay compared with 72 hours of cooling, whereas cooling to 32.0°C was associated with a higher use of inhaled nitric oxide, extracorporeal membrane oxygenation, longer use of supplemental oxygen, and a higher incidence of bradycardia compared with cooling to 33.5°C. A preliminary report of the 18-month follow-up indicated that death or disability occurred in 32% of infants cooled for 72 hours and 32% for infants cooled for 120 hours (aRR 0.92, 95% CI 0.68–1.25), 32% of infants cooled to 33.5°C, and 31% for infants cooled to 32.0°C (aRR 0.92, 95% CI 0.68–1.26).

The results of this trial indicate that among infants of at least 36 weeks’ gestation with HIE, longer cooling was not superior to 72 hours of cooling and deeper cooling was not superior to cooling to 33.5°C. Despite the paucity of data to support the depth and duration of hypothermia in the first series of cooling trials, the Optimizing Cooling Trial supports the continued practice of whole-body hypothermia at 33.5°C for 72 hours and drift from this practice could be associated with increased mortality and morbidity.

How Late Can Hypothermia be Initiated?

The time of initiation of hypothermia represents the component of a hypothermia regimen studied in the most systematic fashion in preclinical investigations. Gunn et al. performed a series of fetal sheep studies where 30 minutes of brain ischemia was followed by 72 hours of hypothermia (cooling cap positioned on the fetal head in utero) initiated at 1.5, 5.5 and 8.5 hours following ischemia. A neuronal loss score in different brain regions was assessed at 48 hours after completion of hypothermia and demonstrated that the extent of neuroprotection was time sensitive. Earlier cooling was neuroprotective (initiation at 1.5 more so than 5.5 hours) but later cooling (initiation at 8.5 hours) was not. These experiments provided the rationale for initiation of cooling within 6 hours of birth in the first series of human cooling trials. These trials, however, could not determine if initiation of cooling earlier in the 6-hour window is more effective since the vast majority of enrolled infants had hypothermia initiated between 4 and 5 hours. A single-center retrospective cohort analysis of hypothermia reported that initiation of cooling at 3 hours or earlier is associated with higher psychomotor developmental scores using the Bayley II Scales of Infant Development compared with cooling initiated after 3 hours.

Although these data suggest that studying initiation of hypothermia after 6 hours of age would not be of benefit, there is strong biologic and clinical rationale for further investigation. As noted for the preclinical studies of the depth of temperature reduction, data from animal models may not be readily extrapolated to newborns. A therapeutic window of 6 hours in fetal sheep seems reasonably established, but there is no information of the duration of the therapeutic window in newborns with HIE because all cooling trials initiated hypothermia by 6 hours. Whether in utero preconditioning events prolong the therapeutic window is unknown. Enrollment in clinical trials of hypothermia at less than 6 hours of age assumes that hypoxic-ischemic events occur proximate to delivery. However, precise timing of hypoxic-ischemia in utero among newborns with HIE may be inaccurate in the absence of a sentinel event, and prior trials likely enrolled infants beyond 6 hours from hypoxia-ischemia. Other important considerations are births in rural communities remote from centers that provide hypothermia, evolution of encephalopathy after 6 hours of age, and late recognition of encephalopathy. All these variables may limit application of hypothermia within a putative narrow therapeutic window. However, initiation of hypothermia after 6 hours has been reported despite the absence of evidence. A comprehensive registry established in the United Kingdom reported on the implementation of hypothermia among 1331 infants born between December 2006 and July 2011; initiation of hypothermia between 6 and 12 hours of age occurred in approximately 9% to 10% of infants and beyond 12 hours of age in 2.2% of infants ( Fig. 4.3 ).

Given the knowledge gap concerning initiation of hypothermia beyond 6 hours, the NRN has conducted a randomized trial to obtain an unbiased estimate of the probability of benefit or harm from “late” initiation of hypothermia (NCT 00614744). Infants at least 36 weeks of gestational age with moderate or severe HIE assessed at or after 6 hours up to 24 hours of age were randomly assigned to an esophageal temperature of 33.5°C maintained for 96 hours, compared with an esophageal temperature maintained at 37.0°C to determine the risk of death or disability at 18 months. Enrollment has been completed and publication of the results is pending.

A major challenge for this trial was determination of the sample size. In most clinical trials, a frequentist analytic approach is used to determine the probability of the observed data or more extreme data if the null hypothesis is true. Results of the NRN’s first Hypothermia trial can be used to illustrate the dilemma for studying late initiation of hypothermia. In this trial, death or disability occurred in 62% of noncooled infants and 44% of cooled infants (aRR 0.72, 95% CI 0.54–0.95, P = .01). Based on preclinical data, it would be reasonable to postulate a smaller effect size than the results of hypothermia initiated at less than 6 hours. Using a frequentist approach, 392 infants would be needed in each group if death or disability were to occur in 60% of infants randomly assigned to 37°C and a 10% absolute reduction in death or disability were postulated. If death or disability were to occur less frequently (e.g., 30% of infants randomly assigned to 37°C) given a decade of experience in stabilizing infants with HIE and a 6% absolute reduction in death or disability is postulated (20% relative reduction), 859 infants would be needed in each group. Dissemination of hypothermia at less than 6 hours of age across the neonatology community made it prohibitive to plan such a large trial.

An alternative approach is to use a Bayesian analysis, which provides different information for hypothesis testing than a frequentist analysis and was prespecified in the NRN trial of late initiation of hypothermia. A Bayesian analysis provides the probability that the hypothesis is true based on the observed data. It is a formal statistical method to assess the range of treatment effects compatible with the best available evidence and is recommended for trials with limited sample size. In contrast to a traditional frequentist analysis, a Bayesian analysis uses preexisting data (pilot studies, clinical trials, observational reports, and animal work) to establish a prior distribution representing the probability of a hypothesized treatment effect ( Fig. 4.4A ). The position of the prior distribution along an axis of risk ratio values depends on the available data at the start of the trial; if there are few data available concerning the intervention, the prior distribution can be centered at 1.0, indicating an equal number of infants may benefit or be harmed (neutral prior). If existing data support benefit, or alternatively, harm of the intervention before the trial, the prior distribution can be shifted to a risk ratio that best represents the available data (enthusiastic or skeptical prior distribution). The prior distribution is then combined with the observed data from the trial to yield a posterior probability of treatment effect ( Fig. 4.4B ). The latter can be characterized by the area under the curve less than aRR of 1.0 and represents the posterior probability of a treatment benefit (e.g., reduction in death or disability). The level of probability that clinicians feel justified to use a specific therapy in practice will reflect the severity of the outcome targeted by the intervention and possible hazard associated with the treatment. A Bayesian analysis can thereby provide clinicians with the best estimate of treatment effect when definitive results using a frequentist approach are not feasible.