The main, surprising, finding of SUPPORT, now confirmed by the other oxygen trials, is that aiming for O2 saturations that were a little lower led to higher mortality.

The big question is why? Having a saturation of 85% to 90% should surely not create such a tissue injury, immune reaction, circulatory dysfunction (or whatever else is implicated), that very preterm babies are more likely to die as a result, should it?

I think an observation of Juliann Di Fiore (free access) from Richard Martin’s group might be very relevant. This analysis of high-resolution oximetry data from SUPPORT quantified episodes of intermittent hypoxia in 115 infants from the 2 groups (from 2 hospitals in the trial). IH was defined as a fall in saturation to less than 80%, for at least 10 seconds but less than 3 minutes. The low saturation group had many more such events, especially during the first few days, and then again after 2 months. The average lowest saturation during these episodes was usually below 70%, so very far below the ranges that we thought we were testing, large numbers of profound desaturations like that could well be dangerous, causing hypoxia/re-oxygenation many times a day.

Another, new, publication is also suggestive, this time looking at cerebral saturations  Schmid MB, Hopfner RJ, Lenhof S, Hummler HD, Fuchs H: Cerebral desaturations in preterm infants: A crossover trial on influence of oxygen saturation target range. Archives of Disease in Childhood – Fetal and Neonatal Edition 2013. The authors did a cross-over study, changing the saturation limits from 80–92% to 85–96% for 4 h each. The periods of time with the lower limits had more episodes of intermittent desaturation, and more episodes of intermittent cerebral desaturation also using NIRS, confirming Di Fiore’s results, and showing that there is an impact at the level of cerebral oxygenation.

Interestingly the graphs in this new study show, very clearly, the missing saturation values between 87 and 91% that are now known to be due to mis-calibration of the Pulse Oximeter. The authors note that this mis-calibration leads to “an artefactual elevation of SpO₂ readings between 87% and 90%”. This reduces the difference between groups, depending on the target ranges chosen. Hence the separation in mortality which occurred in the BOOST2 trials and COT only after the algorithm was changed. Which begs the question, why was the mortality difference there with the old algorithm in SUPPORT? I can’t answer that, the intervention in SUPPORT started much earlier in life than the other studies, maybe that magnifies the effects of lower saturations, certainly the lower sats give you more severe intermittent hypoxia in early life. The babies in Di Fiore’s study were all from SUPPORT, so using the old algorithms, and they did see the increase in hypoxic spells in the low sat target group.