For the patient whose inspiratory flow rate exceeds even the generous threshold of Venturi masks, high flow nasal oxygen is an excellent option. Though the first paper to describe these devices (Dewan & Bell, 1994) gave us this terminology, subsequent authors have occasionally referred to these devices as "high flow nasal cannulae" or "high flow nasal oxygen", because presumably the word "prongs" is somehow uncivilized or intrinsically comical. All CICM trainees will be familiar with the device - it is a single-limb circuit which connects a gas blender to a heater/humidifier, and then funnels a mixture of oxygen and air into the patient, essentially using their respiratory system as a PEEP valve.

Question 2 from the first  paper of 2013 asked for indications, contraindications and complications of high flow nasal prong therapy. To cover all bases, this chapter was written to answer Question 2 as if it were a "critically evaluate" style SAQ. Then, in the first paper of 2017 Question 3 asked the candidates to critically evaluate high flow nasal prongs. LITFL notes on this topic cover the subject in enough detail to answer it. There is also a  great article by J-D Ricard (2012)  which dissects this oxygen delivery system. The more recent review by Papazian et al (2016) offers a decent overview of the evidence to support all the indications for HFNP. To say that these resources have been condensed into the summary below would be unfair to the definition of condensation. However, the time-poor exam candidate will be spared the job of filtering through self-indulgent drivel by the brief summary offered in the grey box below:

Physiological rationale for using HFNP Improved oxygenation by reservoir effect and reduced dilution of inspired O2 Improved tolerance by heating and humidification Improved tolerance of secretions Improved CO2 clearance and respiratory effort efficiency by pharyngeal dead space washout Limitations and contraindications Unprotected airway;  need for intubation Nasal, facial, base of skull injuries Need for a predictable level of PEEP. Complications  of HFNP Barotrauma and pneumothorax Nasal mucosal damage and pressure areas, irritation, epistaxis Aspiration, including of contaminated circuit rain-out, epistaxis, saliva, teeth, food Delay of intubation (i.e. time-wasting behaviour, prevents definitive management)

Scenarios where HFNP is a favoured indication Where NIV is poorly tolerated or inappropriate (eg. oesophageal surgery) Where intubation is not appropriate Apnoeic oxygenation pre-intubation Evidence for the use of HFNP: Parke et al (2011): n= 60; HFNP vs. high-flow face mask.  HFNP group did much better (10% rate of NIV vs. 30% for the standard mask) FLORALI trial (2015): n=310, HFNP vs. NIV,  no difference in intubation rate, but an improvement in 90 day mortality associated with HFNP. PREOXYFLOW (2015): n=124. HFNP vs. high-flow face mask for pre-oxygenation during intubation. There was no difference (well, a difference of 1% SpO2). THRIVE (2014): observational case series, n=25 (difficult airways). Apnoeic oxygenation times were around 14 minutes (half of these patients were obese and a third had stridor). S68 Hi-Flo study (2014): n= 72 babies under 18mth with bronchiolitis; no difference BiPOP (2015): n= 830 post-CABG patients; HFNP vs. NIV. HFNP was non-inferior but otherwise,  no difference in ICU mortality. Meta-analysis: benefit in mortality among immunocompromised patients, but not among immunocompetent ones.

Rationale for the use of high flow nasal prongs (HFNP)

Pharyngeal dead space washout

The upper airways are "rinsed" with humidified oxygen; this is called the "pharyngeal dead space washout". According to the original studies by Fowler (1948) that space is about 150ml, which makes about 25% of the  tidal volume. As such, under normal circumstances that volume ends up being filled with expired air, which might be highly CO₂-rich if the patient is significantly hypercapneic. The  next breath drags this stagnant swamp gas back into the lungs. The high-flow jet reaches deep into this anatomical dead space and flushes out the expired air with nice humidified oxygen-rich (CO₂-poor) gas, which is not something that can be accomplished by other non-invasive devices or even high--flow face masks. Probably, the tracheobronchial gas remains untouched - but at least some nasopharyngeal gas ends up being replaced in this way.

This can have two main effects. One is to produce an improvement in the elimination of CO₂. If the expired air is re-inhaled the mixture will contain expired CO₂ and this will decrease the gradient for the removal of CO₂ from the alveolar blood. HFNP should theoretically improve CO₂  clearance by increasing that gradient.

It's actually not clear whether this really happens. According to Dysart et al (2009), this effect is an extrapolation of the enhanced CO₂ clearance which is seen with tracheal fresh gas insufflation in ARDS, for example. Putting a tracheal catheter into the airway of a severe ARDS patient is one of the (relatively exotic) ways to mitigate the inevitable "permissive hypercapnea" associated with low tidal volume ventilation, so that it does not have to be quite so permissive. Results from Dewan & Bell (1994) suggest that the dead space washout effect is about the same for the nasal prongs and the tracheal catheters, so the extrapolation is probably valid. Also, Fricke et al (2016) convinced a 62 year old COPD patient to have an endotracheal catheter through his tracheostomy for the purposes of measuring the concentration of gas in their anatomic dead space while on high flow nasal prongs, and found that yes - it does wash out something like 50% of rebreathed CO₂. The authors implied that if sustained, this CO₂ removal effect would be comparable with what is achieved by NIV; they were able to drop the arterial PCO₂ by 7.4% over 15 minutes. 

The othermore boring  effect of dead space washout is an improvement in oxygenation. According to Chatila et al (2004), this is mainly because the dead space volume is replaced by oxygen, essentially turning it into a reservoir. This is based on an abstract presented by Tiep and Barnett, who built an airway model and recorded videos of ultrasonic flow studies.  Chatila et al took this information, mixed it with their own findings (improved arterial oxygenation) and made several inferences on this basis, among which one was that this "reservoir effect" contributes substantially to the oxygenation improvement. The debate as to exactly how substantially it contributes trespasses into the territory of academic pointlessness. 

The ultimate upshot of all these factors is that pharyngeal dead space washout improves the efficiency of the respiratory effort. Per unit effort (however you measure it), more CO₂ is expired, and more oxygen is inspired. The pharyngeal washout effect is probably the most important way the HFNP device improves respiratory function over the short and medium-term ventilation timeframes.

Improved oxygenation by PEEP effect

People rave about the PEEP effect of high flow nasal prongs, and protocols are built on the basis of it. The effect is probably a fairly minor contributor to the overall benefit from high-flow nasal prongs. It seems to only be about 3cm H2O with 60L/min flow, when the mouth is open. Tobin and Grove (2007) demonstrated this by convincing some of the staff of St Vincent's Hospital in Melbourne to have their airways topicalised and then passing 10Fr suction catheters into them to transduce the pressure. With their mouths closed, the volunteers had an average PEEP of about 7.5 cm H₂O while the HFNP was set to 60L  (the maximum was 9.7, in the female subjects). 

Why was the PEEP higher in females than in males? Nostril size. By fitting too loosely in large masculine nares, the nasal prongs had sufficient leak around them to depressurise the airway. In contrast, nasal prongs fit more snugly into dainty ladylike nares, the leak is less, and therefore the pressure is higher. This is a purely speculative statement from the Tobin and Grove study, as the authors failed to report important details such as nare diameter (though they did observe that PEEP increased proportionally to decreasing staff member height, suggesting that nare diameter and height are somehow related). This probably has only comedic value to the intensivist, with the exception of those who routinely practice on neonates (as neonatal high flow nasal prongs can easily fit too snugly, produce too much pressure and generate a pneumothorax).

If this PEEPish effect works, then it has all the benefits of "proper" PEEP - recruitment of atelectatic lungs, decreased work of breathing, and so forth. On top of that, it is supposed to overcome the "nasopharyngeal resistance" of obese OSA patients.  In fact the benefits seem to be most pronounced in the obese patients- and the degree of improvement in gas exchange tends to be related to the degree of increase in end-expiratory lung volume, which suggests that there is a real alveolar recruitment effect happening here (Corley et al, 2011).

Improved oxygenation by oxygen dilution reduction

The patient in respiratory failure typically struggles for breath, and has a high inspiratory flow rate, in tens of litres per minute. To use some meaningful comparison, the peak inspiratory flow rate of moderately athletic humans under the load of light exercise was approximately 30L/min according to Anderson et al (2006). If such a human is receiving oxygen by conventional means, that oxygen is being delivered at a sluggish flow rate, say 2-6 litres per minute. Thus, the panicking respiratory failure patient will inhale a gas mixure which will have an inordinately large proportion of room air, and very little of their supplemental oxygen.  High flow nasal prongs ensure that no matter how high the patients' inspiratory flow, the inhaled gas mixture will contain a large amount of oxygen. Most high flow nasal gas delivery systems max out at 60L/min flow, which probably represents something close to the realistic maximum of a hypoxic patient with respiratory failure.  One can make the assumption that they are probably hypoxic because of some sort of problem with their respiratory system, and therefore their diseased respiratory system is insufficiently powerful to generate flows higher than that.

Benefits of humidification: prevention of heat loss and improved secretion clearance

Humidified oxygen is theoretically better than raw untreated wall oxygen. Wall oxygen comes form a tank where the super-low temperature excludes the contribution of any added water: there may be residual water ice inside the tank, but it remains frozen solid at the temperature at which the liquid oxygen turns to gas. The wall oxygen is therefore is very cold and completely dry. The effect of breathing cold dry gas is the loss of both moisture and heat. Heating and humidification therefore theoretically prevents heat loss and moisture loss. Prevention of moisture loss is particularly important to prevent the inspissation of secretions (a topic discussed in some detail in the chapter on heat and moisture exchangers). Mucociliary function is better preserved when the mucus is moist, and with heating and humidification, it is possible to blast high flow oxygen into a patient for 30 days without serious consequences (Boyer et al, 2011). Having said this, we have no direct evidence that HFNP increase the rate or volume of secretion clearance; all this is extrapolated from the (decades-old) findings that in the absence of good quality humidification, ventilator circuits reduce secretion clearance. Nobody has ever measured the airway mucus of HFNP patients and remarked on how much less viscous it was. 

Increased comfort

Increased in relation to what, one might ask. To intubation? Asphyxia? Apparently, when comfort comes into the equation, high flow nasal oxygen is compared to CPAP. The main reason for this is the fact that the mouth is left alone, unlike most forms of CPAP. Additionally, the humidification of oxygen tends to decrease the nasty side effects of oxygen therapy, such as raw stripped mucosa.  Because there is no need for a tight mask, there is no claustrophobia. The patient is able to eat, drink and communicate without the NIV mask in the way. With improved tolerance, there is less need for chemical behaviour control in the delirious or demented population. Tellingly, trials of HFNP like Sztrymf et al (2011) practically  always report that none of the patients asked for the HFNP to be discontinued because of intolerance. 

HFNP may be appropriate in circumstances where NIV is not

Specific situations favouring the use of HFNP may include oropharyngeal surgery and oesophageal surgery such as oesophagectomy. Question 22 from the first paper of 2014 is a fine example of such a situation. Nasopharyngeal surgery however might be off-limits. You wouldn't want 60L/min of gas pneumodissecting its way into your sella turcica after transspenoidal pituitary surgery.

Apnoeic oxygenation peri-intubation

HFNP may be used for apnoeic oxygenation as an alternative to the standard mask. However, this does not seem to be an improvement over the normal methods - in the PREOXY-FLOW trial the HFNP group did not experience any fewer desaturaton events as compared to the standard bag-valve mask  (Vourc'h et al, 2015). It is  not clear what the effect of this on the airway manipulator would be as they stand to face the patient - 60L/min of gas can throw a whole lot of aerosolised pathogens at you (eg. if the patient has active tuberculosis), whereas at least the bag-valve mask poses something of a physical barrier. 

Limitations of HFNP and contraindications to their use

• The patient must be able to protect their airway
• The nose must be intact (i.e. its not obstructed, fractured, or
• The base of skull should be intact (in base of skull fracture you might induce a pneumocephalus and god knows what else)
• There should be no epistaxis (or the blood will end up being aspirated)
• If there has been recent nasal surgery, HFNP may do damage the operative site
• If there has been recent oesophageal surgery, use of HFNP must be weighed against the risk of anastomotic breakdown (though it could still be used, and is probably safer than NIV)
• There is some PEEP, but it is not measured, and is completely unpredictable
• If the patient requires intubation and intubation is delayed because of people wasting time experimenting with HFNP, the outcome may be worse.

This topic of using HFNP following upper airway and GI surgery remains contentious. In short, most reasonable people who understand the function of HFNP will agree that there is minimal risk from their use, as the PEEP-like pressure exerted by these devices is trivial. Other equally reasonable people will instead point to the fact that these devices have never been demonstrated to improve any important parameters (oxygenation etc) in upper GI surgical patients, and that the risk of compromising an oesophageal anastomosis - though theoretical - is not zero. 

Where does the truth lie? Likely, there is some middle ground, but the situation is not very well studied. The evidence regarding the use of HFNP in these questionable scenarios is mixed, and can be used to promote either viewpoint. Observe:

On one hand, if you wanted to protect your precious anastomosis at any cost, you could argue against the use of HFNP using the following information:

• There does not appear to be any benefit, in terms of outcomes.  Futier et al (2016) included a wide range of abdominal surgical patients in a randomised trial and found no improvement in oxygenation (or any other secondary outcomes like pneumonia or reintubation) when comparing HFNP to conventional oxygen delivery methods. 
• HFNP is known to produce PEEP, and so any contraindications to the use of PEEP should logically apply to HFNP (Natham et al, 2019)

On the other hand, if you really wanted to use HFNP, you could point to the following:

• If PEEP is your main concern, then NIV should be even more unsafe in these scenarios, and yet multiple authors have demonstrated the safety of NIV with pressures as high as 16cm H₂O (Joris et al, 1997; Kingden-Milles et al, 2005; Pessoa et al, 2010). Applying PEEP is clearly not such a big problem.
• The UpToDate authors or other review writers (eg. Wilcox et al , 2018, or Nishimura et al, 2016) do not usually list upper GI surgery as one of the contraindications for the use of HFNP.
• Futier et al (2016) included numerous upper GI patients in their study (35 post Whipples,  4 post gastrectomy and 7 post oesophagectomy). They blasted these people with 50-60L/min without any dramatic increase in the need for re-operation within seven days. Clearly, none of those anastomoses ended up leaking.

Potential adverse effects associated with the use of HFNP

Trawling though the evidence, one can find relatively few articles where the complications of HFNP are discussed in any sort of great detail. They are generally from the paediatric literature (eg, Baudin et al,  2016)

Pressure and flow-related complications

• Overdistension of the alveoli, and barotrauma
  • in fact, in neonates this may lead to pneumothorax
• Nasal mucosal damage due to high flow
• Pressure areas due to the device

Aspiration of food

• If the patient is trying to eat and the pressure of this devise has not bean dialled down, bad things may happen. In the best case scenario the high flow gas will blow the food out of their mouth, and make an embarrassing mess. In the worst case scenario, the jet of gas will blow the food into their airway. The same goes for vomit that has been regurgitated by a semiconscious patient, or secretions from their nose.

Discomfort of the device

• The nasal prongs need to be fastened at all times; the cannulae may be irritating to the nostrils; and the devise produces a constant loud hiss, which can be so irritating as to drive you mad.

Other complications:

• Failure to achieve the desired effect because of mouth-breathing
• Overabundance of secretions (Velasco et al, 2014) - though some might view this as a desired effect
• Epistaxis
• Time-wasting (delaying the inevitable intubation)
• Aspiration of circuit condensation water (there's no evidence that this causes pneumonia, but people complained about it in a survey of paediatric ICUs conducted by Manley et al, 2012). The perception of infection risk from warm circuit water seems to be greater than the actual risk, but some literature out there does exist to support at least the theoretical possibility of this; for example Pierce & Sanford (1973) present some evidence that Pseudomonas can be spread through the humidification systems of infant incubators, and quote contemporary studies where the horrified authors cultured all sorts of microorganisms from the water baths of humidifiers. The authors wisely recommend that a good standard for humidification equipment would be make sure that the deliver air is no more contaminated than the room air of the hospital (a relatively low bar).

Clinical applications of high flow nasal prongs

• As a stand-alone therapy for hypoxic respiratory failure, in which case it could be used in any sort of respiratory failure (LITFL list a handful of different types, such as asthma, APO, pneumonia, etc etc)
• Instead of NIV:
  • When positive pressure is contraindicated, eg. oesophagectomy
  • When the patient is intolerant of NIV (eg. delirium)
  • When clearance of secretions must be maintained, but the patient is too hypoxic for conventional oxygen delivery devices (eg. in pneumonia)
• Instead of intubation:
  • In patients whom it is inappropriate to intubate (i.e. as a palliative measure)
  • In patients for whom intubation is associated with a worse outcome, eg. febrile neutropenic patients and those recovering from bone marrow transplant.
• Prior to intubation:
  • HFNP may significantly improve preoxygenation and reduce episodes of hypoxia (Miguel-Montanes et al, 2015).
  • This seems to be something that benefits difficult airway patients (THRIVE)  rather than all-comers (PREOXY-FLOW)

Evidence in support of the use of HFNP

Parke et al (2011):  one of the first studies comparing HFNP and standard high-flow face mask

• 60 patients randomised to either normal mask or HFNP
• Main outcome measure was having to resort to NIV
• HFNP group did much better (10% rate of NIV vs. 30% for the standard mask)

FLORALI trial (2015): multicenter open-label trial, 310 patients

• Only hypoxic patients were selected (P/F ratio <300) - hypercapnea was excluded
• Primary outcome was intubation rate, secondary outcome was mortality
• HNFP was compared to standard oxygen and NIV
• There was no difference in intubation rate, but somehow there was a improvement in 90 day mortality associated with the use of high-flow oxygen.
• The NIV group had 9ml/kg tidal volumes, which may have influenced their mortality by exacerbating their lung injury
• Of the intubated and NIV patients, more died of shock rather than respiratory failure.

PREOXYFLOW (2015): multicenter open-label trial,124 patients

• Randomised to either high flow oxygen mask (removed at the end of intubation) or high flow nasal prongs (kept on during the whole process).
• The main point was to see whether an apnoeic patient would benefit from high flow oxygen blowing into their airway; theoretically they should desaturate more slowly during the intubation attempt because pure oxygen from the HFNP-irrigated upper airways will be entrained into the lung by mass transfer.
• No such effect was seen (a statistically insignificant difference of 1% SpO2 was found).

THRIVE (2014): observational case series of 25 patients with difficult airways

• HFNP was commenced prior to the induction of anaesthesia
• While maintaining jaw thrust, HFNP delivered oxygen to the apnoeic patient (apnoea time was counted from the administration of muscle relaxant)
• Median apnoea time was 14 minutes (ranging from 9 to 19 minutes) and the patients did not desaturate beyond 90%.
• Given that half of these patients were obese and a third had stridor, this is an outstanding result. Amazed authors concluded that this technique "has the potential to transform the practice of anaesthesia".

S68 Hi-Flo study (2014): Randomised controlled trial of 72 babies under 18 months of age

• Bronchiolitis was the specific pathology under investigation
• Comparison of 2L vs 8L O₂
• There was a minor improvement of clinical parameters (a modified Tal score), but no real difference otherwise

BiPOP (2015): Multicenter, randomized trial in 830 post-op cardiothoracic patients

• Inclusion criteria were respiratory failure after surgery, or those who were "deemed at risk" of this following extubation
• The patients were then randomised to either HFNP or NIV
• HFNP was non-inferior: the rate of reintubation was the same in both groups
• There was also no difference in ICU mortality.
• NIV produced more skin pressure areas, but was otherwise equivalent.

Meta-analysis (Nedel et al, 2016) -  critically ill patients with respiratory failure, or at risk of it

• 9 studies met inclusion criteria
• No mortality difference was found when the data were pureed in the meta-analysis blender, because the studies were severely heterogeneous and each insisted on measuring the improvement in oxygenation in some different way
• Many of the trials excluded hypercapneic COPD patients, unfairly (as HFNP could be of substantial benefit to these patients)

Meta-analysis (Monro-Somerville, et al; 2017)

• 14 studies met inclusion criteria (n= 2,507)
• No mortality benefit
• However, the mortality was quite low (6% with HFNP, vs. 8.1% with usual care)
• Presumably selecting sicker patients will allow benefits to better manifest, might say the HFNP enthusiast

Meta-analysis (Huang et al, 2018) - immunocompromised patients with respiratory fialure

• n = 667
• Significant reduction in short-term mortality and intubation rate
• Unchanged ICU length of stay

Meta-analysis (Conte et al, 2018) - pre-term neonates

• n = 1227 
• When nasal CPAP was used, failure rate (i.e. progression to intubation) was lower. 
• i.e. HFNP failed more frequently, and in those situations the NCPAP occasionally prevent the need for intubation