A report of secondary outcomes from a trial, prior to the report of the primary outcomes should, in general I think, be resisted. Having said that, we reported the secondary outcomes of the CAP trial before we had the primary outcomes, largely because it was suggested by the data safety monitoring committee, and the idea was that the secondary outcomes that we wanted to report were quite positive, and we had identified no safety concerns, so the ongoing widespread usage of caffeine was probably quite appropriate while awaiting the developmental outcomes.
The new NICHD hypothermia trial has just published secondary outcomes (Shankaran S.: Effect of depth and duration of cooling on deaths in the nicu among neonates with hypoxic ischemic encephalopathy: A randomized clinical trial. JAMA 2014, 312(24):2629-2639.) which I think was also a good idea in this case. Most importantly because there is some doubt about safety of longer or deeper hypothermia resulting from their findings.
This was a factorial trial examining both the effects of longer hypothermia (120 compared to 72 hours) and deeper hypothermia (32 compared to 33.5 degrees).
Multiple (pre-planned) interim analyses were performed, and after exactly half of the expected sample was enrolled (364 of 726) the study was stopped for safety concerns and for ‘futility’. For both of the comparisons there were a few more deaths with the longer or deeper groups compared to the standard approach. Neither of the comparisons in terms of death were significant at conventional levels of significance, which means that there is a 14% chance that the difference in deaths found between longer and standard cooling was due to random variation, and a 26% chance that the difference in deaths between the temperature groups was due to random effects.
What that means is that there really isn’t any proof that lower or longer was harmful in terms of death, but the Safety committee determined that the probability of finding an advantage was small (they calculated it as 2%), which was really the point of doing the study. They can certainly be criticized for stopping the study at this point, but they would have been trashed if they had continued and it turned out that there was a true increase in mortality with the new approaches to treatment.
In addition there were other negative effects found, particularly of deeper cooling to 32 degrees, with an increase in pulmonary hypertension, leading to more NO use, and more ECMO use, as well as more bradycardia, which is not surprising, but might affect systemic oxygen delivery if cardiac function is compromised.
The implication of the study is that 72 hours appears to be long enough, and that 33.5 degrees is low enough, which is a bit strange… does that mean that we got it exactly right, by chance, with the very first studies? I think that would be remarkable; the initial temperature and duration chosen were based on the best pre-clinical data that we had, but they were somewhat arbitrary choices. Maybe a slightly less severe hypothermia might be better, and then it could be for a longer time, which might have advantages, but I don’t know now if we will ever do another study like this one. Another large high quality, multicenter trial of different approaches to hypothermia will probably only ever be done if we can change our models of doing clinical research. If we could cheaply randomize infants who are undergoing cooling already, collect data that is easy available and simple to collect centrally, and then get long term outcome data, without the very expensive infrastructure of a trial like this one, (similar to the registry trial in Scandinavia that I already blogged about in the past), then we could answer many other questions that will be difficult, or impossible to answer otherwise.