Blow some Os up that nose? The PARIS 2 RCT

The PARIS 2 trial
Cite this article as:
Morgenstern, J. Blow some Os up that nose? The PARIS 2 RCT, First10EM, March 20, 2023. Available at:

I don’t think I have seen any medical intervention adopted as quickly as high flow nasal oxygen.  However, despite its prevalence, there are still many questions about its role. Although the original PARIS trial was technically positive, the results did not look great to me. Here, we will cover the PARIS 2 trial, a multicenter RCT comparing high flow nasal oxygen to standard oxygen in children with acute hypoxic respiratory failure. 

The paper

The PARIS 2 trial: Franklin D, Babl FE, George S, et al. Effect of Early High-Flow Nasal Oxygen vs Standard Oxygen Therapy on Length of Hospital Stay in Hospitalized Children With Acute Hypoxemic Respiratory Failure: The PARIS-2 Randomized Clinical Trial. JAMA. 2023 Jan 17;329(3):224-234. doi: 10.1001/jama.2022.21805. PMID: 36648469

The Methods

PARIS 2 is an open-label multi-center RCT from 14 hospitals in Australia and New Zealand.


Pediatric patients aged 1-4 years with acute hypoxic respiratory failure of any cause who required admission to hospital. Specifically, they needed to meet 4 inclusion criteria:

  1. Increased work of breathing due to acute respiratory disease.
  2. Oxygen requirement to keep sats at or above 92% at most hospitals, but 90% in one hospital.
  3. Respiratory rate over 34.
  4. Admission to hospital.

Exclusions: Craniofacial abnormalities, upper airway obstruction, cyanotic heart disease, and those who required immediate higher-level care in the ICU, or required noninvasive or invasive mechanical ventilation.


High flow humidified nasal cannula (HFNC) using the AIRVO-2 system.

Flows were set at 2 L/kg/min in children up to 12 kg, 30 L/min in 12-15 kg, 35 L/min in 16-30kg, and 40 L/min in 31-50 kg. FiO2 was titrated to oxygen saturation.


Standard oxygen either using a nasal cannula with a maximum flow rate of 2 L/min or a Hudson facemask with a maximum flow rate of 8 L/min.


The primary outcome was hospital length of stay. 

The Results

They include a total of 1567 children, with a median age of 1.9 years, median oxygen saturation of 88% at enrollment, respiratory rate of 47, and only about 15 hours of total symptoms.

The primary outcome of hospital length of stay was longer with high flow nasal oxygen (1.77 days versus 1.5 days, HR 0.83, 95% CI 0.75-0.92).

Length of oxygen therapy (1.07 days vs 0.75 days) and ICU admission (13% vs 7%) were also higher with HFNC. Tolerance of therapy was the same in both groups. There were 3 complications (2 nosebleeds and 1 death) with HFNC and none with standard therapy.

My thoughts

This is important research, and I think the results are believable, but there are a number of problems with this trial which significantly increase the risk of bias. It is an unblinded trial, with a primary outcome that is heavily reliant on clinical assessment, and therefore easily influenced by that lack of blinding. (My guess is that all children were ready to go home at the exact same time, but kids wearing the HFNC just look sicker, and so were weaned slower, and therefore discharged later. Of course, this is likely to happen outside of the trial as well.) Selection bias is also a potential problem, with only 51% of eligible patients being included, with no explanation provided for the vast majority of patients missed.

The cap on the flow rates in the standard care group seems artificial and unnecessary. Many children are treated with more than 2 L/min by nasal prongs. Facemask oxygen is frequently provided at rates higher than 8 L/min. It is not clear to me why this cap was necessary. It seems to bias the trial against the standard care group. Why not just allow clinicians to provide oxygen like they always do without artificial caps?

Some interventions are so obvious that they push against the limits of evidence based medicine. Last week, I suggested a novel approach to lateral canthotomy that has only been tested in cadavers. That evidence is extremely poor, but success with the technique will be so obvious that I didn’t think we needed better evidence. Some technology is also obviously better. A video laryngoscope with identical geometry and design is clearly better than an identical blade lacking video. We don’t need science to prove that, but science is still important, for example, to ask questions of cost/benefit.

Anyone who has used high flow humidified nasal oxygen knows that it is a game changer in resuscitation. Like the canthotomy approach I suggested with limited evidence, HFNC has almost no risk, and the benefit is immediately apparent. So one might ask: do we really need studies demonstrating that it is a good idea to provide oxygen to patients who need oxygen?

I think this study points at the biggest problem with HFNC. The device will definitely provide patients with oxygen. The problem is not with the device, but with our tremendous capacity for overuse in medicine.

How many of these children needed any oxygen at all? With a baseline oxygen saturation of 88% and a median hospital length of stay under 2 days, it seems very likely that the vast majority of these children would have been fine without any intervention. Indeed, if this trial is ever replicated, I think the addition of a no oxygen arm is warranted and would be very informative.

However, we need to be careful not to extrapolate these results to a different group of patients. It is not surprising that HFNC provides no value in a group of patients that may not have required any oxygen at all. But this study tells us absolutely nothing about the child who presents with an oxygen saturation in the 70s, because that child is not represented here.

That this trial was focusing on the wrong population seems pretty obvious from the methodology and sample size calculation. Their primary outcome was hospital length of stay, but because they only expected these children to have a 2 day hospital stay, they decided to power their trial to find a 9.6 hour difference between the groups. They state that difference would be clinically significant, but I beg to differ. In the real world, there is no way these kids are getting discharged in the middle of the night, so being ready 9 hours earlier just means waiting until the next day anyway. A 9 hour difference is incredibly unlikely to be clinically significant, and a group with a total length of stay less than 2 days is incredibly unlikely to benefit from high flow oxygen. It seems like this trial was doomed to fail from the outset. 

It is interesting to note that almost 90% of the patients treated with high flow nasal oxygen were managed on a regular ward rather than the ICU. That might be value information for some hospitals, where silly rules limit the use of this device. On the other hand, if your patient isn’t sick enough for the ICU, do their really need this device?

Bottom line

In a group of patients with very mild hypoxia, there is no benefit of early high flow humidified nasal oxygen. This trial should caution us about the massive indication creep we have seen with these devices over recent years.

Other FOAMed

The PARIS Trial: HFNC in Infants with Bronchiolitis



Franklin D, Babl FE, George S, Oakley E, Borland ML, Neutze J, Acworth J, Craig S, Jones M, Gannon B, Shellshear D, McCay H, Wallace A, Hoeppner T, Wildman M, Mattes J, Pham TMT, Miller L, Williams A, O’Brien S, Lawrence S, Bonisch M, Gibbons K, Moloney S, Waugh J, Hobbins S, Grew S, Fahy R, Dalziel SR, Schibler A. Effect of Early High-Flow Nasal Oxygen vs Standard Oxygen Therapy on Length of Hospital Stay in Hospitalized Children With Acute Hypoxemic Respiratory Failure: The PARIS-2 Randomized Clinical Trial. JAMA. 2023 Jan 17;329(3):224-234. doi: 10.1001/jama.2022.21805. PMID: 36648469

Photo by Rafael Kellermann Streit on Unsplash

Leave a Reply

2 thoughts on “Blow some Os up that nose? The PARIS 2 RCT”

%d bloggers like this: