This article is the latest in a series of randomized controlled trials attempting to find the safest target oxygen saturations for premature infants. It's also an extremely complex study design and report, pretty challenging to wade through, but perhaps brings us closer to the "real answer."
Source: The BOOST-II Australia and United Kingdom Collaborative Groups. Outcomes of two trials of oxygen-saturation targets in preterm infants. NEJM. 2016;374(8):749-760; doi:10.1056/NEJMoa1514212. See AAP Grand Rounds commentary by Dr. Jonathan Mintzer (subscription required).
PICO Question: Among premature infants
Question type: Intervention
Study design: Randomized controlled trial
This report is really 2 combined studies, 1 from Australia and the other from the United Kingdom, encompassing over 2000 newborns less than 28 weeks' gestation randomized according to the target oxygen saturation percentage clinicians would aim for during their treatment. The "lower target" group aimed for a range of 85-89% saturation, while the higher target was 91-95%. When all the dust/confusion (see below) had cleared, the investigators found a statistically significant higher mortality in the lower target group, 24.5% versus 16.9%. However, it's still not that clear-cut.
Because the investigators wanted the clinicians to be blinded to group assignment, they told all the clinicians to aim for 88-92% saturations, but then rigged the pulse oximeters to display false readings depending on the group assignment. Oximeters for the lower target group were modified to display a reading up to 3 percentage points higher than the actual reading (e.g. a display reading of 90% was really 87%) and up to 3 percentage points lower for the higher target group (90% display was really 93% saturation).
If you're not confused already, just wait. Unfortunately, the UK investigators discovered a problem with oximeter settings after the study had begun, resulting in slightly incorrect shifting of the saturation readings. The machines were fixed, but now the UK group was left with some patients managed using the oximeters with the glitch, while others managed with the glitch corrected and therefore more comparable to the Australian group. So, now the investigators were stuck performing post hoc subgroup analyses. That always raises eyebrows, because it brings up a whole range of statistical problems and requires altering the statistical methodology; not surprisingly, statisticians disagree as to how these analyses are best performed.
In addition to this report, 3 other similar trials of oxygen saturation targeting in premies are in various stages of progress, as part of the Neonatal Oxygen Prospective Meta-analysis Collaboration. When this analysis is complete, we should have a better view of this issue. In the interim, it seems that most authorities in this field, including Dr. Mintzer in his commentary, suggest avoiding the lower oxygen saturation targeting.
