Alasdair Munro. High flowing controversy: A return ticket to PARIS, Don't Forget the Bubbles, 2019. Available at:
Today we are all familiar with high flow nasal cannula (HFNC) as respiratory support. We see it in our emergency departments, on our wards and in the PICU. People rave about it, and people rant about it. People talk about it so much that there must be some high-quality evidence about it, right?
There certainly wasn’t until last year, when the much-anticipated Paediatric Acute Respiratory Intervention Study (PARIS) was published in the prestigious New England Journal of Medicine (if you haven’t read about the trial, most of this will make little to no sense to you. Have a quick read of the DFTB run down here, or read the article itself in full here).
This RCT of HFNC in bronchiolitis has stirred up strong opinions in some, and a few pretty convincing and controversial myths have arisen from it over the past year.
This take on HFNC and the PARIS trial may be a little controversial, so buckle your seatbelts (and please comment if you disagree).
Myth 1 – PARIS was a badly conducted study
If you hate HFNC and are looking for a take-down of the PARIS trial, I’m afraid this isn’t it. PARIS was a fantastically well-conducted trail, with an impressive write up. It was performed to the highest standards of clinical research and it has been clearly analysed. Most importantly, it finally provided the first attempt at high-quality evidence about an intervention which has already become a standard of care, despite having no evidence to support its use. The authors should be congratulated.
If PARIS was so well conducted, where have these myths come from? Almost all stem from a misunderstanding of myth number 2…
Myth 2 – Paris was a trial of HFNC vs standard oxygen therapy
This is a confusing myth, as the publication of the study states this is the case in the methods section and throughout the article. If this is what the authors say it is, how can this be untrue?
It can be argued that PARIS cannot be a trial of HFNC vs standard oxygen therapy, for two reasons:
- In both groups, 100% of escalated children get HFNC at some point (you cannot compare HFNC vs something else if everyone gets HFNC).
- The primary outcome for the two arms, despite being called the same thing, are actually different. For both arms the outcomes are called “escalation of care”, but for the HFNC group this means escalation to PICU, whereas for the standard O2 group this means escalation to…you’ve got it – HFNC.
You don’t have to be a research wizard to realise that having two different outcomes for your groups is a no-no. It means you cannot compare outcomes between the two groups, as you are not comparing like with like. The HFNC group has apples, and the standard O2 therapy group has oranges.
So, what is PARIS a trial of?
PARIS is a trial of a strategy of immediate HFNC for hypoxaemia in bronchiolitis, vs a strategy of rescue HFNC for hypoxaemia when standard oxygen therapy fails.
If we are not comparing the original primary outcomes, what should we compare?
In PARIS we should compare treatment escalation to PICU, as this is an equal outcome for both arms.
If we assess the study this way, we can see quite clearly that when comparing these two strategies, there is no difference in outcome of escalation to PICU (9% of standard therapy, 12% of HFNC, p=0.08). In fact, when we discount the original primary outcome definition, there is no difference in any other outcome between the two groups (except that the HFNC group were sicker when they got escalated).
The PARIS trial was an amazing piece of clinical research, but the write up has asked a strange question of the data and described the groups in a confusing way.
Myth 3 – PARIS demonstrated immediate HFNC is superior
Based on a misunderstanding of the above, plus a frequently stated statistic from the trial that:
The NNT of HFNC vs standard oxygen therapy to prevent one episode of escalation of care is 9
Some people have taken home the message that a strategy of immediate HFNC is superior to standard oxygen, as it prevents treatment failure. However, as we have already discussed, treatment escalation for the standard oxygen therapy group consists of being started on HFNC. All this means then, is using HFNC is superior for preventing you from later putting the patient on HFNC. Obviously, this makes no sense at all.
If we are discounting this mind-bending interpretation, and (as above) we see that an immediate strategy of HFNC changed no outcomes compared to a strategy of rescue HFNC, how do we decide which is better?
This is simple: when faced with two interventions which provide equivalent outcomes, you always pick the one which is cheapest and least invasive. Here, a strategy of standard oxygen therapy with rescue HFNC is the winner.
Myth 4 – PARIS tells us HFNC is effective in bronchiolitis
Sadly, PARIS cannot tell us anything about HFNC in and of itself, because we did not compare HFNC to no HFNC. Everyone got HFNC in the end, so there is no true control group. Although we can try and guess what would have happened in the standard oxygen therapy arm if there was no HFNC available, we can’t say this for sure as it didn’t happen.
Many people certainly believe that HFNC works and prevents PICU admission (which it may well do), but unfortunately PARIS cannot answer that question for us, and as yet no trial has done so.
What PARIS CAN tell us about HFNC is that it is safe. There were very few adverse events, of which it is unlikely any were related to HFNC.
If PARIS cannot tell us if HFNC is effective, how do we find out?
We want to design a study to tell us if HFNC is useful in bronchiolitis. First, we need to decide what we actually want HFNC to do. We are not expecting it to save lives, as children rarely die of bronchiolitis, and if they look like they might die then we have an effective intervention, which is to send them to PICU and intubate them. We also don’t expect HFNC to get children home any quicker, as it doesn’t treat the underlying disease process.
So, what do we want from HFNC? I would argue it is to keep children out of PICU, and on the ward. This is cheaper, nicer and less invasive. A win all around.
So how could we tell if HFNC keeps children out of PICU? We need a trial which compares rescue HFNC (which PARIS has demonstrated is preferable to immediate HFNC, as there’s no difference between the two), to standard care – basically normal treatment of bronchiolitis without using HFNC at all (hard to remember a time when this was the norm…). You then see if there is any difference between how many children go to PICU, or how long they stay there. Let’s put our dream study in a PICO:
P – Infants with bronchiolitis with hypoxaemia
I – Standard care + HFNC as treatment escalation
C – Standard care (no HFNC)
O – Admission to PICU/ PICU LOS
If the above looks strange to you, it’s worth remembering that not too long ago there was no HFNC, and there is certainly no way that the same number of children we currently put on HFNC all used to go to PICU instead.
Why hasn’t this trial happened?
If that’s the dream trial, where is it?
Unfortunately, because medical devices are not regulated in the same way as drugs, there is no requirement to prove their efficacy or added value prior to them being approved for any indication. HFNC made its way into every paediatric unit faster than a bronchiolitic baby can sneeze, and quickly became the standard of care without ever having any evidence it was effective. The horse has now left the stable, and there is no way of convincing people to run a trial in which half of the children do not get HFNC.
In the absence of any evidence, the world has now been split into a group of “HFNC believers” and “HFNC non-believers”, neither of which have any evidence that they are right (but many getting confused about the PARIS trial and sometimes misusing it to confirm their opinions).
Myth 5 – But I’ve SEEN HFNC work
Some people may read all this, and think,
“Yes, but who cares. I use HFNC and I have seen it help children and stop them go to PICU. We don’t need a trial, because we can SEE it works”.
Sadly, we have learnt from previous experience that even when people believe they have seen interventions work, sometimes when the evidence is finally produced it turns out we were wrong all along. This is particularly a problem for medical devices, due to the way they are regulated. Just look at the recent NEJM study for IVC filters to prevent pulmonary emboli, or the mind blowing ORBITA trial last year for coronary stenting for chronic angina. This is why randomised controlled trials are so vital. They have demonstrated millions of pounds have been wasted on ineffective procedures and devices in these scenarios, even when health professionals truly believed that they had seen them work. Time has proven again and again that we will often just see what we want to see.
- PARIS was a fantastically well conducted trial
- PARIS was NOT a trial of HFNC vs standard care. It was a trial of immediate HFNC vs rescue HFNC, and no differences between the two were demonstrated.
- When two interventions provide the same outcome, you always pick the cheapest and least invasive option, which in this case is rescue
- PARIS cannot answer questions about HFNC efficacy, but did demonstrate its safety
- To demonstrate efficacy, we require a trial of rescue HFNC vs standard care (without HFNC), with an outcome of PICU admission/LOS
- Current problems with device regulation mean they can be instituted without evidence of efficacy, and following widespread implementation it becomes very difficult to test their effects. We owe our patients better than that.
A recent meta-analysis of HFNC in bronchiolitis, including (and heavily weighted by) the PARIS trial, was recently published in the Journal of Paediatrics and Child Health. You can read our analysis here.