High‐flow nasal therapy – modelling the mechanism

When one of the authors of this Editorial was a trainee trying to develop a suitable curriculum vitae for a teaching hospital appointment, he was told to ‘publish or perish’, so he dutifully produced a paper on the new anaesthetic innovation of the time, the Brain airway, soon to be renamed the laryngeal mask. For senior trainees in 2018 in a similar position, high-flow nasal therapy (HFNT) seems to be the topic of choice, with an exponentially rising number of papers about the technique and its ever-expanding applications in anaesthesia and critical care. To suggest that HFNT will revolutionise anaesthesia as the laryngeal airway did may be far-fetched. However, the development of gas delivery devices which deliver ideally conditioned gases at flow rates to match or exceed a patient's peak inspiratory flow rate is significant. High-flow nasal therapy delivers inspired oxygen concentrations of 21–100% at flows of up to 60 l.min−1, and still maintains ideally warmed and humidified gases, which benefits patients by the avoidance of mucosal drying and inspissation of secretions in the airway. It is much more acceptable to patients than conventional high-flow oxygen masks and compliance with the therapy is good 1. The terminology used to describe the intervention is variable – most authors begin with the term ‘high-flow nasal’ or ‘nasal high-flow’ but this is followed by various combinations of ‘oxygen’, ‘cannulae’ or ‘therapy’. Anaesthetist authors tend to favour ‘high-flow nasal oxygen’, whereas basic science and medical papers more often use ‘cannulae’ or ‘therapy’ in the description. In this paper, we have deliberately used HFNT to better reflect the fact that, as we describe below, the intervention is much more than simply delivering supplemental oxygen to the nose. High-flow nasal therapy can be delivered to patients who are either breathing or apnoeic. The former group includes patients in acute respiratory failure 2, during elective and emergency awake airway management 3, 4 and during bronchoscopy under sedation 5. There is also interest in the use of HFNT in the postoperative setting, for example, following abdominal and cardiac surgery, to reduce the incidence of postoperative pulmonary complications, with mixed results so far 6, 7. The use of HFNT has been shown to significantly reduce the number of exacerbation days in patients with chronic airways disease 8 and it may also have a role in the management of patients with respiratory disease who are not suitable for intubation, for example, in the palliative setting. The clinical benefits of HFNT are a result of reliable and non-invasive delivery of high oxygen concentrations, the generation of a positive pressure in the pharynx, and improved carbon dioxide clearance. In spontaneously-breathing patients with respiratory failure, the last of these offers a major clinical benefit, reducing functional dead space by flushing carbon dioxide from anatomical dead space gas in the nasal cavity between the end of expiration and the beginning of the next inspiration 9. Positive pharyngeal pressure is generated predominantly during expiration as a result of continuous inward flow from the nasal cannulae, with the extent of the pressure generated dependent on both the delivered flow rate and the patient's expiratory flow rate. Mean pharyngeal pressure has been shown to increase by 1 cmH2O for each 10 l.min−1 increase in HFNT flow rate within the range 30–50 l.min−1 10. This pressure, however, varies throughout the respiratory cycle, being higher during expiration, including in the trachea 9, 11. The pressure generated is unfortunately also inconsistent, being affected by factors such as whether the mouth is open or closed as well as the presence of nasal pathology and variations in nasal anatomy 12. Although HFNT does not produce a constant pressure as continuous positive airway pressure does, the increase in end-expiratory pressure is still sufficient to benefit patients with respiratory failure. For example, HFNT has been shown to improve respiratory system mechanics including a reduction in work of breathing, an increase in dynamic lung compliance and decreased inspiratory resistance 13. During apnoea, HFNT normally involves delivering 100% oxygen. The aim is to maintain arterial oxygenation in patients who are not receiving any tidal ventilation to facilitate airway manipulation, for example, during attempted intubation or for airway surgery 4, 14. The ability of this technique to maintain oxygenation for prolonged periods of time is widely reported and the physiological basis described 15. In this respect, HFNT offers no particular advantage over apnoeic oxygenation at lower flow rates which has been practiced for many years. It is the behaviour of carbon dioxide during apnoeic oxygenation that seems to be different in HFNT, with several groups reporting a slower increase in carbon dioxide levels during apnoea with higher flows 14. The concept that carbon dioxide removal can still occur during apnoea is not new. Mass movement of oxygen into the lung during apnoea was demonstrated in human volunteers using a body plethysmograph in 1959 16. The basic physiology was fully understood at the time – buffering of carbon dioxide in body tissues means that oxygen uptake exceeds carbon dioxide output so a negative pressure in the alveolus develops, drawing gas into the lung. This explanation, however, clearly suggests there is no removal of carbon dioxide from the lung during apnoea. The development of high-frequency oscillatory ventilation led to a resurgence of interest in carbon dioxide removal in the absence of tidal breathing. Sinusoidal pressure waveforms in the airway with tidal volumes a tiny fraction the size of anatomical dead space, usually too small to measure, still resulted in substantial carbon dioxide clearance. Several mechanisms are involved, some of which are likely to occur during apnoea, including gas mixing from turbulence, cardiogenic oscillations and molecular diffusion 17. Partial clearance of carbon dioxide is achieved with apnoeic oxygenation via a catheter placed in the trachea, with animal studies showing increased clearance with greater flow rates and more distal position of the catheter in the trachea 18. These results suggest that HFNT, applied at the nostrils, should be less effective, contrary to the recent observations showing improved clearance of carbon dioxide relative to other apnoeic oxygenation techniques 14. The mechanisms of carbon dioxide clearance during apnoea are therefore now the object of much research. In-vivo studies are difficult, and apart from measuring arterial, end-expiratory or transcutaneous carbon dioxide levels there is little that can be done in patients to explain the observed changes. The answers must therefore be sought in physiological modelling, and three approaches are now being pursued. First are physical models of airways, which have been used to study airway gas flows for many years. Models developed in the 1980s consisted of clear plastic tubes bolted together to study flow from tracheal tubes within the models 19. Simple tube models have also been used more recently and may continue to provide useful information, for example, on the rate at which carbon dioxide may be flushed from the nasal cavity 20. High-flow nasal therapy has been studied in silicone models derived from computerised tomography (CT) images of airways and using glycerine/water mixtures and pumps to simulate tidal ventilation. These models have shown anatomical variations between left and right nasal cavities sufficient to affect the flow characteristics on each side 12, and that the HFNT cannulae are narrow relative to the airway causing a high velocity jet of gas to be generated to facilitate flushing of the nasopharynx 21. More recent work has used 3D printed upper airway models, again derived from CT images of individual airways, but using clearance of a radioactive tracer gas to assess the effects of HFNT on airway resistance 22 or on carbon dioxide clearance from the nasal cavity 19. A final physical model involves the use of artificial ventilation of ex vivo sheep lungs studied with sampling catheters placed into small distal airways 23. This work differs from other physical models in that airways beyond the upper trachea are now included, and also, for the first time, flow-dependent removal of carbon dioxide from lower airways was demonstrated. This model is of course also anatomically far removed from an intact human lung. The second type of model used to investigate carbon dioxide clearance in HFNT is computational fluid dynamic (CFD) modelling. This involves converting 3D images from human airways into mathematical models, from which computer software can then calculate flow patterns based on basic principles of gas flow. Gas composition, flow patterns into the model and the boundary conditions where the flowing gas meets the model wall must first be defined. Detailed calculations of flow from, and patterns within, the model are then obtained. The first study to use a CFD model to investigate HFNT within the nasopharynx included only the nasal cavity and explored the effect of anatomical features and flow rate on the positive pressure generated 24. Another CFD study extended the airway model to the proximal trachea, with results validated against a physical model of the same individual airway. This study looked specifically at carbon dioxide clearance from the nasal cavity during spontaneous breathing and confirmed the long-suspected flow-dependent flushing of carbon dioxide from the dead space in the nose 25. Almost all the modelling papers described so far have addressed the mechanism of HFNT in patients who are breathing to better understand the factors affecting the reduction in functional dead space. The physiology of apnoea is harder to study and was first tackled with the third model type – computational physiological modelling. This involves the use of known physiological principles to simulate complex interactions between the respiratory and cardiovascular systems under varied circumstances 26. This type of model is limited by its dependence on the assumed variables selected for the simulation, that is, setting up the virtual subject, but has the advantage of being able to include the many factors affecting the behaviour of an intervention like apnoeic oxygenation in vivo. This form of simulation has been validated against in-vivo studies and has even facilitated the study of apnoeic oxygenation in more challenging situations, for example, in pregnant subjects 27. Changes in arterial carbon dioxide concentration have been investigated using this model in apnoea with reasonable correlation with in vivo studies 25, although the technique has not yet been applied to HFNT. In this edition of Anaesthesia, Hermez et al. 28 report a study using physical airway models to investigate the way in which HFNT may facilitate carbon dioxide clearance. The model used is an improvement on previous versions in that it also includes the first five generations of airways, that is, distally as segmental or small bronchi of 3–4 mm diameter. This leaves of course a further 18 generations of airway before reaching the source of the carbon dioxide, the alveoli. The model also differed from its predecessors by having the mouth open for gas flow. Under these circumstances, most of the gas entering at the nares passes through the oropharynx and out of the mouth, and airway carbon dioxide must only reach the oropharynx to be effectively flushed out. This study demonstrated that the turbulence induced in the pharynx by HFNT extends beyond the glottis, and at higher flow rates extends even to the distal trachea. The turbulence, and so mixing of gases and removal of carbon dioxide, is significantly facilitated by the cardiogenic oscillations also simulated in this model. The interaction between the turbulent flow in the large airways and the cardiogenic oscillations originating in the distal lung tissue is proposed as a major contributor to the additional carbon dioxide removal seen with HFNT. All these models have limitations in common. They employ a single airway which may or may not be based on the anatomy of a human subject. There is huge variation in airway morphology and so flow patterns will vary both between and within individuals, as evidenced by, for example, sleep disordered breathing. High-flow nasal therapy is recommended to work best when the mouth is closed, and most model studies so far assume this is the case, but patients do not always have their mouth closed, so the applicability of the models to real life remains limited. All models must choose flow rates to simulate breathing, based on studies of normal inspiratory and expiratory patterns but again, flow rates in patients are highly variable depending on size, pathology and even psychology, for example, anxiety. Finally, physical models are generally rigid structures, whereas the in vivo airway is not. However, in the absence of another way to advance our understanding of HFNT, efforts should be made to further refine these techniques. Although Hermez et al. are to be applauded for modelling airways to the 5th generation, they are still a step away from a complete lung connecting the source and exit points of pulmonary carbon dioxide. We believe this is more likely to be achieved with some form of computational rather than physical model: CFD is likely to be the best approach for understanding flow patterns and gas mixing but developing a mathematical model lung with the estimated 260,000 respiratory bronchioles from where the carbon dioxide originates will be a challenge. Physiology simulators represent low-fidelity modelling of flow characteristics and anatomical structures but offer a more integrated model that is likely to better replicate the in vivo situation. However challenging, these goals are worth pursuing; HFNT remains a new innovation with room for development. For example, the current system delivers a constant flow to the nasal cavity; given the new findings of increased turbulence improving carbon dioxide clearance, different flow patterns to increase turbulence such as pulsatile flows or higher velocity flows, for example, from smaller nasal cannulae at the same flow rate, may further improve the technology. No conflicts of interest.