Time to Adopt Cooling for Neonatal Hypoxic-Ischemic Encephalopathy: Response to a Previous Commentary

On the basis of opinions of problematic evidence, Kirpalani et al1 stated that the “standard of care” criteria for adopting hypothermia for neonatal hypoxic-ischemic encephalopathy (HIE) had not been met. They argued for a “conservative approach.” Clinicians who are “impressed” by their evidence are not mandated to offer cooling and can ethically continue to randomly assign patients “into ongoing, or new trials.” Yet, the authors considered the evidence to be “certainly” sufficiently strong for cautious use of this treatment by clinicians who are “impressed with the results.” These apparently conflicting statements are reinforced by previous confusing statements by co-author Barks2: “Many of us, including this author and the lead author [Dr Shankaran] of the Network trial,[3] who caution against uncontrolled adoption of cooling, have lost equipoise, and we are now cooling babies who meet the eligibility criteria of our original trials,” and the semantically elusive, “In a small number of experienced centers, it is a novel therapy that has become ‘standard care’ (without randomization), but it is not yet the ‘standard of care’ (emphasis in the original).Evolution in this field has outdated this conservative approach. One cooling trial (Infant Cooling Evaluation [ICE])4 was stopped prematurely, the research protocol of another has been altered,5 a third trial (Trial of Whole Body Hypothermia for Perinatal Asphyxia [TOBY])6 was stopped at the end of the funding period before achieving its revised sample size, and 3 independent systematic reviews4,7,8 have confirmed the efficacy and safety of cooling.Today's question is whether it is acceptable for individual hospitals or physicians to withhold from parents information about and access to cooling. We question the lingering concerns about the methodology of the trials raised by Kirpalani et al.1 The importance of the answer is illustrated by predictive calculations: Of the 15 to 18 infants born daily in the United States with moderate-to-severe asphyxia, 10 to 12 of them die or develop moderate-to-severe disability; providing hypothermia to all would likely prevent 3 infants from either death or moderate-to-severe disability without causing any clinically significant adverse effects.8 This is an enormous effect size.We address the issues raised by Kirpalani et al1 in the order in which they were presented originally.We agree that withdrawal of life support (or nonwithdrawal for patients with similar prognoses) is pivotal in determining the outcome of infants with HIE by tipping the balance between death and significant disability. In our experience, based on a well-defined, transparently documented process, the immediate cause of death was withdrawal of life support in 91% of infants with HIE who died after 8 hours of age.9 Kirpalani et al's line of reasoning is that a reduced rate of withdrawal is expected to tip the balance, decreasing the death rate and increasing the severe disability rate among survivors. They acknowledged that the latter did not occur. Although withdrawal of intensive therapy occurred more frequently in the intervention group, there was no evidence that the cost of the reduced mortality rate was an increased disability rate; on the contrary, all outcomes improved.7,8,10 Thus, the “primum non nocere” principle was not violated. Kirpalani et al pointed out that the failure to demonstrate an increased morbidity rate does not exclude the possibility of underlying bias but made no mention of the potential for cooling to reduce the severity of the HIE, hence less reason for withdrawal of support.Kirpalani et al1 also mentioned the use of the composite outcome end point as a source of potential bias but did not develop this point. In fact, the component outcomes of disability and death are different expressions of the same outcome, viz, severe brain injury (or a strong suspicion thereof). The terms “complementary outcomes” or “reciprocal outcomes” may be more appropriate than “composite outcome” in the HIE context. The components death and disability are both clinically important and carry similar weight (with the latter permitting withdrawal of support). Both are likely to be influenced by cooling, and together they characterize the net effect of the therapy. Reporting of the composite outcome in hypothermia trials conforms to the recommendations for the use and interpretation of composite outcomes.11 We agree with Kirpalani et al that criteria for withdrawal cannot be mandated in a clinical trial, but we disagree with the notion of blinded adjudication in a research context. It is unnecessary to audit the withdrawal process in this context; the great majority of physicians who are involved with the parents in decision-making are personally risk-aversive and do not participate in or implement withdrawal-of-support decisions unless the best interests of child and family are served.Kirpalani et al indicated that a proportion of control patients had periods of hyperthermia and suggested that “some active cooling mechanism may be required to avoid excessive [body] temperatures.”1 In contrast, the control-group patients may have benefited from the Hawthorne effect. Kirpalani et al suggested avoiding hyperthermia but did not suggest how to do it. Some patients develop hyperthermia in response to brain injury,12 and the only way to control it would be to cool patients actively. Unless the results of ongoing trials change the results of additional meta-analyses,8 the new research advocated by Kirpalani et al will long delay decisions about care.We agree with the methodologic concerns of Kirpalani et al1 regarding the different subgroup criteria and the inadequate sample sizes. However, their critique of subgroup analyses has no bearing on the results of the primary outcome of the studies.We challenge the calculated sample size of Kirpalani et al1 on 3 grounds: First, a choice of 80% rather than 90% power would have sufficed for a therapy not shown to have any clinically significant adverse effects.8 Second, the control event rate was 64%, not 61.3%.4,8 Third, the authors did no justify an arbitrary 50% primary outcome rate in controls, a rate that requires inclusion of patients with mild HIE. Recalculation of the required sample size on the basis of the above grounds is 482, and with more than the expected effect size (actual relative risk: 0.76 [ie, 24% effect size]), the sample size would be even lower.8 Thus, the published trials certainly meet the suggested “optimal information size.”The authors reported historical instances of harmful interventions adopted prematurely but failed to mention the other side of the story. Failure to conduct meta-analyses and adopt new standards in an appropriate and timely fashion was described by Chalmers as a “scandalous failure of science to cumulate evidence scientifically.”13 Examples include the recommendation to adopt the supine sleeping position in 1970, which could have averted 50 000 cases of sudden infant death syndrome in Europe alone,14 and the use of streptokinase for myocardial infarction in 1973 (P < .01), which could have averted random assignment of the next 17500 patients (in 25 trials) to control groups.15 We are obliged to navigate our patients between the extreme positions of precocious recognition of novel therapies and unjustifiable delay in instituting beneficial therapies.Clinicians and health care providers who make patient-care decisions and ethics and safety committees that make human-experimentation decisions may have difficulty explaining their disregard of the evidence (the significant results of trials and meta-analyses4,7,8) on the basis of the problems perceived by Kirpalani et al.1 Individual clinicians and institutions are obliged to review the evidence and make their own decisions, as did members of the network trial.2 It is time to stop postponing the decision to accept hypothermia (at experienced centers that use established protocols) as an effective treatment. It is our duty to explain the benefits and unknowns of cooling and to offer this treatment to every eligible patient with moderate or severe neonatal HIE.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.