At the PAS meeting in Washington DC, that has just finished, the COT trial results were presented, as well as the short term outcomes of the BOOST trials.

These are a series of clinical trials which have compared higher saturations 91 to 95% to lower saturations 85 to 89%. As I mentioned before when discussing the SUPPORT trial, which had the same target ranges, and the same technique of using masked pulse oximeters, these target ranges are within the ranges that were being used routinely in NICUs around the world, but we did not know if one target range was better or worse than the other, overall. There has been a remarkable international collaboration to get several similar trials performed at around the same time, so that we can then get even more power by putting the results together.

The COT trial showed no significant difference between the high and low sat target ranges. The BOOST trials (BOOST UK (and Ireland) BOOST Australia and BOOST NZ) showed the increased mortality, in the low saturation group, that we have already heard about, and the increased retinopathy in the high saturation group similar to the SUPPORT trial. Both of these studies have also just been published, COT here and the BOOSTs here

Now what?

Closer examination of the COT results shows that there were more deaths in the low saturation group, 16.6% vs 15.3%. Although this is not significant (at usual levels of statistical significance) it is not very different in magnitude to the increased mortality in the other trials results. Even more concerning, the difference in mortality between the groups occurred entirely after introducing the new software in the Masimo oximeters. After changing the software, the mortality in the 2 groups was 16.8% in the low saturation group, and 14.1% in the high saturation group.

Also, in COT, retinopathy of prematurity (RoP) was not different between the groups, this is harder for me to understand, there was really no difference at all between the saturation groups. If COT had been published first, then I would have thought that maybe the increase in RoP with oxygen mostly occurs with higher sats, and is not different between the high 80s and the low 90s, but the other trials now all show more RoP between these target ranges.

So probably, and this question was asked by Richard Martin in the meeting, a meta-analysis will show that the confidence limits of the COT mortality difference have a wide overlap with the other oxygen trials, and most likely the overall analysis will still show a significant increase in mortality with lower saturations, and, I would guess, with very little heterogeneity.

One of the strengths of these oxygen trials, as I mentioned above, was that even from before they started there was a plan to perform a prospective meta-analysis, whereby the individual patient data are entered into a database, but with variables defined a priori, and the data to be collected agreed from the start. All of the oxygen trials (SUPPORT, COT, the BOOSTs) agreed to participate in this NeoProm collaboration, one of the reasons that many of us were very happy to participate in COT , was the participation in NeoProm (I actually initiated COT in two hospitals, the Royal Victoria Hospital in Montreal, where it was ably taken over by Nabeel Ali when I left, and then at Sainte Justine).

I think it is vitally important that we all continue this collaboration, to be able to answer the important questions, that were raised several years ago, with the greatest confidence.

Lisa Askie, the expert in this kind of analysis in the perinatal field, has written a review of the current situation (already a bit out of date with these new presentations, and publications). She notes the preliminary data regarding increased mortality, and describes the NeoProm collaboration in some detail. As she notes, the BOOST trials were stopped early as a combined preliminary analysis showed the increase in mortality that I am describing, COT did not contribute to this preliminary analysis, and an article published in Trials (open access) recently by the BOOST steering committee describes the history of what happened as the steering committees were struggling to decide what to do with the data showing increased mortality with lower saturations.

I think the time has come for complete openness with trial data. Iain Chalmers notes that the enrollment in the BOOSTs was stopped after the interim analysis, a process which would have been easier and done with more confidence if the COT data had been available, such a decision has major consequences whichever way it goes, reducing sample size and reducing power, exposing subjects to risks, or truncating a trial which did not truly have differential risks. These are all potentially of great importance. The only way to make a good decision is to have all the available relevant information.

NeoProm will help us to finally answer the question posed in the title of this post: each of the 2 new publications is accompanied by an editorial, (here and here) both of them state that we should now target 90 to 95% saturation, which should avoid the increased risk of death with the lower saturations, and although there will probably be some more retinopathy, the risk of serious visual problems in the long term is not increased because of the high quality ophthalmological screening and treatment we currently have.

The data from clinical trials do not belong to the PIs and the steering committees of the trials. They should be considered to be the property of the whole community: the babies and families who participated, the funding agencies and the societies that support them, and the investigators and health care workers who facilitate the trials.

Chalmers paper mentioned above includes the following sentence ‘If the academic community wishes to be seen to be putting the interests of patients and science above individual agendas, concerns and benefits, the issue of sharing interim analyses needs to be given more attention’ they give examples of 2 situations in which similar on-going trials have agreed to share data during the performance of those trials. Only by doing this will we be able to avoid situations where continuing trials, or stopping them too early, causes unnecessary harm.