Once upon a time, many adults suffering an out-of-hospital cardiac arrest received therapeutic hypothermia with a target temperature of 33°C. Then, along came the Targeted Temeperature Managment trial – in which 36°C seemed to be just as good as 33°C. Now, just to throw another confounder in the mix, we have a trial comparing 33°C to “therapeutic normothermia” – 36.8°C – and we’re doing it in children to address concerns regarding generalizability from adults.
Very detailed summaries of the numbers, methods, and enrollment can be found on other #FOAMed sites – particularly St. Emlyns and ALiEM. But, the high points:
- Many – 1,355 – were screened, but ultimately only 260 were randomized and included in their primary analysis.
- Adherence to temperature management protocols was good or adequate in ~90% of cases.
- Hypothermia was implemented for 48 hours, followed by normothermia up to 120 hours total to match the normothermia group.
- In contrast to adults, the great majority (72%) of this pediatric cohort suffered a respiratory arrest.
The outcome: no statistical difference, with 20% of the hypothermia group alive and functional at 1 year, compared with 12% of the normothermia group, a p-value of 0.14. Regarding safety, arrhythmias and culture-proven infections favored the normothermia group, 1% vs. 5%, and 39% vs. 46%, but these also did not reach statistical significance. Finally, both 28-day and 1-year mortality favored hypothermia, with an absolute difference of ~10% in each, but this was not statistically significant, either.
I will let the authors speak for me here:
“One important potential limitation of the trial is that, on the basis of the observed confidence limits for treatment differences, a potentially important clinical benefit cannot be ruled out despite the lack of a significant difference in the primary outcome measure. A larger trial might have detected or rejected a smaller intervention effect. Indeed, there was a significant difference in survival time with therapeutic hypothermia, although this was a secondary outcome measure.”
The relative likelihood of benefit for hypothermia in this trial was 1.54, with a 95% CI of 0.86 to 2.76. Now, this result crosses 1, and therefore requires interpretation in two contexts. The first is the normal distribution:
In which we visualize the frequency of potential outcomes, and the important realization the more frequent “true” outcome is most likely to occur near the center of the 95% CI range.
And, the more important context:
In which we interpret these data in the context of prior results, generalized from other settings. In this case, our prevailing opinion is one in which we suspect hypothermia – with much uncertainty regarding the details – is beneficial. As you can see, the effects of even “statistically significant” findings have only limited practical impact on the “good bet” or the “long shot”. Hence, the results of this study – which simply barely fail to reject the null hypothesis – do not hardly move the needle against the prevailing opinion.
I tend to side with the authors of this “negative” study: it is mostly likely underpowered to detect the expected benefit, and it is still reasonable to cool children following OHCA. There are many questions that remain regarding the temperature, duration, and other details – not limited only to children – but it would be erroneous to say this trial refutes the practice of hypothermia in children.
“Therapeutic Hypothermia after Out-of-Hospital Cardiac Arrest in Children”
http://www.nejm.org/doi/full/10.1056/NEJMoa1411480