The Melbourne group must get fed up of publishing important research, here is another in the PNEJM.
This is one of the first non-inferiority studies in neonatology, they designed a trial with 303 very preterm babies (less than 32 weeks) who were randomized at the time of extubation to either HFNC with the Optiflow device at either 5 or 6 liters per minute, or CPAP at 7 cm H2O. The study was designed to find out if HFNC were worse at preventing re-intubation than CPAP or not. They expected a 25% failure rate in the CPAP group, which was what they actually found, 25.8%, and they found a 34% failure rate in the HFNC group, which was not significantly worse.
Many of the HFNC failures were rescued by starting CPAP, and there was much less nasal trauma with HFNC than CPAP.
Of note, most of the treatment failures were in the first day after extubation, and failures were very common in both groups of infants less than 26 weeks (61% CPAP, 81% HFNC). Which goes to show that we need better ways of predicting which extremely preterm infants are ready for extubation, and better ways for keeping them breathing, most of the failures being for apnea.
This study can be added to another trial from Melbourne (this time from the Mercy Hospital, although the ubiquitous Peter Davis manages to be an author on both) and also the multicenter RCT of Yoder and others. Clare Collins’ study also included infants under 32 weeks, with no stated minimum and randomized babies to the Vapotherm at 8 liters per minute or CPAP at 7 or 8 cm H2O; Brad Yoder’s study only included infants over 27 weeks and over 1 kg, but went right up to term, they used whatever device the NICU had for HFNC at 3, 4, or 5 liters per minute depending on size, and used CPAP at 5 to 6 cmH2O.
Extubation failure was equally common in those trials also.
It appears that we finally have sufficient evidence of efficacy of HFNC for post-extubation care of the preterm neonate, including the very preterm, and even including a few term babies. The non-significant differences in Collins’ study were in the opposite direction to Brett Manley’s data, and the numbers in Yoder’s trial were almost identical. So an eye-ball meta-analysis (Barrington method) shows no difference in extubation failure.
HFNC lead to less nasal trauma, and it doesn’t look like there is any other clinically important difference in outcomes.
Neonatal Research 
I think this is a question that often comes up on rounds: do we know the actual PEEP delivered with HFNC? Thx
The quick answer is ‘it depends’. I know not very helpful… going back 20 years Jay Greenspan published a paper showing that high flow cannulae could create PEEP, which they called ‘inadvertent’ PEEP (Locke RG, Wolfson MR, Shaffer TH, Rubinstein SD, Greenspan JS: Inadvertent administration of positive end-distending pressure during nasal cannula flow. Pediatrics 1993, 91(1):135-138.) in that study the pressures were between 0 and 17 cmH2O if I remember correctly (and I usually do;-)).
The pressures vary according to the flow, the ‘fit’ of the cannulae, whether the mouth is open or closed, the body weight of the baby… etc. In fact that is the reason that I have been very negative about using HFNC, I think there are safety issues, and efficacy issues, that still have not been adequately addressed, but then, there are similarly issues with nasal CPAP; we measure the pressure in the circuit, but no-one measures the pharyngeal pressure routinely, and they are not the same. At least with CPAP we have an upper limit of pressure. With some HFNC systems we have a pressure limit now, but not all. Personally I think the Fisher-Paykell system has the best pressure limit, but there is no objective data to support that opinion.
To answer your question further, when the flow is less than 2 liters per minute, positive pressures in general are small or undetectable, above 2 liters per minute they increase, but you can never be sure what pressure you are giving.