Skip to content

PARIS in the autumn



As the days get shorter for those of us in the Southern hemisphere it inevitably means that bronchiolitis season is on its way. So far, studying bronchiolitis has left us with a big list of things that don’t work – steroids, salbutamol, adrenaline, and antibiotics. Chest x-rays don’t help, and NPAs don’t give us any information we can actually use – but what about high flow?

Through the PARIS trial, Donna Franklin and our friends at PREDICT have considered nearly 21,000 infants presenting to 17 different hospitals to try and shed some light on this for us.

Franklin et al, A Randomized Trial of High-Flow Oxygen Therapy in Infants with Bronchiolitis. NEJM. 2018. 378(12):1121-1131

Who did they look at?

Infants <12 months old with a clinical diagnosis of bronchiolitis and a need for supplemental oxygen to keep their SpO2 in the range 92-98% or 94-98% (target sats range varied by hospital as hospitals had different guidelines at the time of the trial). Of the 20,795 children screened, 1638 were randomised. The bulk of exclusions were for disease too mild to warrant treatment or for having an alternative diagnosis. Other exclusion criteria all have pretty logical reasons (e.g. cyanotic heart disease, craniofacial abnormalities, or the need for home oxygen). After removing those for whom consent could not be obtained, 1472 made it through to analysis.

What did they do?

The intervention group were placed on heated humidified high-flow oxygen at 2L/kg/min via an Optiflow system.* Notably, this flow rate is double the 1L/kg/min used by Elizabeth Kepreotes and her team in what, prior to PARIS, had been the largest trial of bronchiolitis in high-flow conducted in Australia. The PARIS trial adjusted the FiO2 to achieve sats in the target range, followed by weaning down to 0.21 to use the lowest FiO2 possible to achieve those target sats. Kids whose FiO2 had been at 0.21 for 4 hours were taken off high-flow. I’m labouring the details of this process because I think PARIS is setting a standard as I will explain below.

What were they comparing it with?

Infants in the standard therapy group got plain old wall oxygen through standard nasal cannulae up to a maximum flow rate of 2L/min (note this is NOT weight dependent) and again weaned so as little oxygen as possible was being used to keep the sats in the target range.

What did they find?

The primary outcome was treatment failure that resulted in the escalation of care during the hospital admission. Treatment failure was defined by the child meeting any 3 out of 4 clinical criteria:

  • Heart rate remained unchanged or increased since admission;
  • Respiratory rate remained unchanged or increased since admission;
  • Needing oxygen at either >2L/min (standard group) or FiO2 >0.4 (high-flow group) to maintain minimum target sats; or
  • Scoring high enough on the hospital’s early warning tool (CEWT/between the flags/Victor/etc) to warrant escalation of care.

Potentially problematic, from the study point of view, but unavoidable in real life, clinicians were also able to escalate care based on any other concern – and many did so. For those in the standard therapy arm, escalation of care meant high flow. Escalation of care occurred in 12% (87/739) of the high-flow group and 23% (167/733) of the standard therapy group. Via some maths that sounds plausible to me, this gave a number needed to treat of 9 to avoid one escalation of care.

Hospitals with an onsite ICU escalated care in more of the patients on high flow (14% vs 7%) and fewer of the patients on standard flow (20% vs 28%) than hospitals without an on-site ICU. This makes sense, given the practical barriers to escalating beyond high-flow in facilities without a PICU onsite. 34% of infants who had their care escalated had it done on the basis of clinician concern, not the pre-defined physiological parameters, and this was more common in the standard therapy group. This, perhaps, reflects the widespread belief that high flow is helpful in this situation, making it harder for people to hold back. Looking only at the infants who triggered the physiological criteria, escalation was still more common in the standard group than the high-flow group (16% vs 7%), supporting the conclusion that high-flow does indeed reduce the objective need for escalation of care.

Historical factors such as prematurity and previous admission with bronchiolitis made no difference to the findings, nor did positive identification of RSV in the patient’s nasopharynx.

Secondary outcomes showed no differences in the duration of hospital stay, duration of ICU stay or duration of oxygen therapy. 12 infants were intubated, giving an intubation rate of just under 1% of the bronchiolitis infants who made the inclusion criteria. Importantly, adverse event rates were both low and equivalent in both groups, with one pneumothorax in each group, neither of which needed draining.

The bottom line

This is the biggest and most robust trial yet done to assess the value of high-flow in bronchiolitis. The primary outcome shows that there is a role for high-flow in the non-ICU management of this disease. Importantly PARIS has shown in a large cohort of children that high-flow, when used within the parameters of the trial protocol, does not lead to an increase in adverse events which in-turn suggests the increased patient:nurse ratios for kids on high-flow that are often mandated by hospital policies may not be necessary (depending on the severity of disease of course). Some caution must be used around the potential for erroneous use of the high-flow circuits themselves and the interpretation of early warning scores in the context of high-flow use.

PARIS was supported with significant nursing education resources potentially reducing errors to a level that were below what could be expected with the standard resourcing of mixed EDs and other environments where high-flow use in children may be infrequent. As with many grey areas in medicine protocols as to how we use high-flow vary by institution with little more than opinion to guide them.

Though neither the intention nor the conclusion of this paper in showing the progress of such a large number of children on high-flow, this trial also provides a basis for more robust decision making around how we use high-flow itself.

Ben discussed this paper with Ken Milne on the Skeptic’s Guide to Emergency Medicine blog and podcast episode 228, which you can access here.

PARIS in the summer

The original Paediatric Acute Respiratory Intervention Study (PARIS) was published 5 years ago and saw authors attempt to compare standard oxygen with high-flow nasal cannula oxygen (HFNC) to treat bronchiolitis. The authors’ primary outcome was an escalation of treatment between the groups, but posthoc analysis and discussion around the study show that their methodology didn’t quite allow this comparison.

Authors of the second trial published by the PARIS group asked a different question, focusing on the effects of early HFNC, compared to standard oxygen, on the length of stay for patients with acute hypoxemic respiratory failure. To address this question, authors enrolled and randomised more than 1500 children across 14 hospitals in Australia and New Zealand.

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;329(3):224–234. doi:10.1001/jama.2022.21805

Who did they look at?

Children between 1 and 4 years old were recruited IF they had increased work of breathing, a respiratory rate of 35 breaths per minute or more, an oxygen requirement (as per the recruiting hospital’s oxygen target saturation range), and required hospital admission. Those with upper airway obstruction, craniofacial abnormalities, cyanotic heart disease and who required immediate HDU or ICU were excluded. The remaining patients were randomised to HFNC or standard oxygen unless their caregiver refused consent. 753 children were randomised and consented to the HFNC group, and 764 children were randomised and consented to the standard oxygen group.

What did they do?

HFNC oxygen was given at set flow rates, depending on the weight of the child, with the FiO2 titrated to achieve target oxygen saturations. Flows were not changed during the period of HFNC use, but FiO2 was weaned to the lowest possible whilst still achieving target oxygen saturations. When the FiO2 was able to be weaned to 0.21, the study protocol called for HFNC to be stopped. This is similar to the protocol used in the initial PARIS Trial, the difference being that the previous protocol advised ceasing HFNC only after 4 hours at a FiO2 of 0.21. Patients who did not tolerate HFNC (as determined using a visual analogue scale recorded by caregivers and nursing staff) were changed to standard oxygen therapy.

What were they comparing it with?

Children randomised to the standard group received oxygen at 2L/min via nasal cannulae or up to 8L/min via Hudson masks, depending on their oxygen requirements. Oxygen was weaned to the lowest possible flow rate to maintain target oxygen saturations. Treatment was escalated for children who could not meet target oxygen saturations within the standard group at the decision of the treating team.

What was their question and how did they tackle the data?

The team’s primary outcome was the length of stay, from time of randomisation to time of hospital discharge or death, and they had 9 secondary outcomes. These included the length of oxygen therapy after randomisation and the proportion of children who had a change in oxygen therapy during their hospital admission.

Length of stay was assessed using Cox Proportional Hazards Modelling, which allows authors to take into account any effect on the results from patients being treated at different sites, having obstructive or non-obstructive airways disease at presentation, having wheeze or no wheeze, and being older or younger. Secondary outcomes were assessed using multiple tests, which can lead to false positives (if we perform multiple comparisons with a significance level of 0.05, then the outcome of one in twenty statistical tests will be incorrect). Authors have therefore stated that the secondary outcomes are exploratory only.

What did they find?

For the primary outcome, length of hospital stay after randomisation, authors found that this was significantly higher in children on HFNC who stayed for a median of 1.77 days compared to a median of 1.5 days in the standard oxygen group. This difference was only seen when comparing HFNC vs standard oxygen and did not persist when looking at obstructive vs non-obstructive disease, wheezy vs non-wheezy children, or if children were stratified by age.

Although the secondary outcomes are only exploratory, the authors found that the median length of oxygen therapy after randomisation was greater in the HFNC group (1.07 vs 0.75 days). There was also a higher number of HFNC patients admitted to ICU compared to those admitted in the standard oxygen group (94 vs 53 patients).

The trial’s analysis was performed regardless of whether the patient’s treatment was changed during the study period (for example, those who were escalated from standard oxygen, or reduced from HFNC, were analysed in their original randomised groups). This is called an intention to treat analysis and is used commonly as it best mirrors what would happen clinically in real life and so leads to less risk of bias.

The bottom line

Data from this RCT of standard oxygen vs HFNC oxygen therapy for acute hypoxaemic respiratory failure showed that children on HFNC stayed in hospital for longer and were more often escalated to ICU than children on standard oxygen therapy.

It is unclear why this is the case, but authors suggest it may be due to a subconscious thought that children receiving HFNC are more unwell than those on standard oxygen, leading to a more cautious weaning regime. This would make sense as one of the limitations of this trial was that clinical teams could not be blinded to treatment allocation.

Regardless of the reason for the findings, they suggest that there is no length of stay benefit related to early initiation of HFNC in acute hypoxaemic respiratory failure. Perhaps more education is required on which children benefit most from the therapy.

*Fisher and Paykel who manufacture the Optiflow system have provided financial support to both the DFTB17 and DFTB18 conferences


Kepreotes, E., Whitehead, B., Attia, J., Oldmeadow, C., Collison, A., Searles, A., Goddard, B., Hilton, J., Lee, M. and Mattes, J., 2017. High-flow warm humidified oxygen versus standard low-flow nasal cannula oxygen for moderate bronchiolitis (HFWHO RCT): an open, phase 4, randomised controlled trial. The Lancet389(10072), pp.930-939.

Franklin, D., Dalziel, S., Schlapbach, L.J., Babl, F.E., Oakley, E., Craig, S.S., Furyk, J.S., Neutze, J., Sinn, K., Whitty, J.A. and Gibbons, K., 2015. Early high flow nasal cannula therapy in bronchiolitis, a prospective randomised control trial (protocol): A paediatric acute respiratory intervention study (PARIS). BMC Pediatrics15(1), pp.1-8.


  • Ben Lawton is a paediatric emergency physician interested in education, retrieval medicine and simulation. Lives in Brisbane where he enjoys falling off his mountain bike and being outsmarted by his pre-teen children. @paedsem | + Ben Lawton | Ben's DFTB posts

  • Cameron escaped the cold and wet of Scotland to work as an ED registrar in sunny Melbourne, pursuing interests in paediatric critical care and evidence-based medicine. Outside of work he can be found cycling, climbing or with friends exploring brunch options


No data was found

Leave a Reply

Your email address will not be published. Required fields are marked *

4 thoughts on “PARIS in the autumn”

  1. Escalation in one group is not equivalent to escalation in the other group.
    If some in the low-flow group were then stabilised by escalation to high flow , then why start anyone on high flow early? start on low flow early, and only escalate to HF in those in whom it is required (there must have been 570-odd patients randomised to the high-flow group who would have done perfectly well on low flow, thus reducing both equipment and nursing costs)

    1. I’m with you, Brendan – so, 167 of 733 (23%) of the low-flow group required escalation to high-flow. Of these, 102 responded meaning that only 65 of 733 (9%) who started on low-flow needed escalation beyond high-flow compared to 87 of 739 (12%) who started on high-flow. In the absence of any other obvious benefits, this suggests that low-flow is still a good place to start, with escalation to high-flow if necessary…he says from his hospital sans PICU, having admitted a 4 month old with moderate bronchiolitis last night, and started him on high-flow.

  2. This study DOES NOT show any real benefit for high flow in this patient group. Same length of stay, duration of O2 and need for ICU. In addition, the “3 out of 4” factors for escalation of care were often ignored by clinicians and care escalated for “other” reasons. Clearly it is easier to escalate care from low flow to high flow than in is from low flow to nasal CPAP or intubation.

  3. We now have two large studies which fail to show any benefit of high flow oxygen on length of stay, duration of oxygen therapy, or need for ICU.