Nasal High Flow Clears Anatomical Dead Space in Upper Airway Models

: 33 Recent studies showed that Nasal High Flow (NHF) with or without supplemental oxygen can 34 assist ventilation of patients with chronic respiratory and sleep disorders. The hypothesis of 35 this study was to test whether NHF can clear dead-space in two different models of the 36 upper nasal airways. The first was a simple tube model consisting of a nozzle to simulate the 37 nasal valve area, connected to a cylindrical tube to simulate the nasal cavity. The second was 38 a more complex anatomically representative upper airway model, constructed from 39 segmented CT-scan images of a healthy volunteer. After filling the models with tracer-gases, 40 NHF was delivered at rates of 15, 30 and 45 L/min. The tracer gas clearance was determined 41 using dynamic infrared CO 2 spectroscopy and 81m Kr-gas radioactive gamma camera imaging. 42 There was a similar tracer-gas clearance characteristic in the tube model and the upper 43 airway model: clearance half-times were below 1.0 s and decreased with increasing NHF 44 rates. For both models, the anterior compartments demonstrated faster clearance levels 45 (half-times < 0.5 s) and the posterior sections showed slower clearance (half-times < 1.0 s). 46 Both imaging methods showed similar flow-dependent tracer-gas clearance in the models. 47 For the anatomically-based model, there was complete tracer-gas removal from the nasal 48 cavities within 1.0 s. The level of clearance in the nasal cavities increased by 1.8 mL/s for 49 every 1.0 L/min increase in the rate of NHF. The study has demonstrated the fast-occurring 50 clearance of nasal cavities by NHF therapy, which is capable of reducing of dead-space re- 51 breathing. 52


ABSTRACT: 33
Recent studies showed that Nasal High Flow (NHF) with or without supplemental oxygen can 34 assist ventilation of patients with chronic respiratory and sleep disorders. The hypothesis of 35 this study was to test whether NHF can clear dead-space in two different models of the 36 upper nasal airways. The first was a simple tube model consisting of a nozzle to simulate the 37 nasal valve area, connected to a cylindrical tube to simulate the nasal cavity. The second was 38 a more complex anatomically representative upper airway model, constructed from 39 segmented CT-scan images of a healthy volunteer. After filling the models with tracer-gases, 40 NHF was delivered at rates of 15, 30 and 45 L/min. The tracer gas clearance was determined 41 using dynamic infrared CO 2 spectroscopy and 81m Kr-gas radioactive gamma camera imaging. 42 There was a similar tracer-gas clearance characteristic in the tube model and the upper 43 airway model: clearance half-times were below 1.0 s and decreased with increasing NHF 44 rates. For both models, the anterior compartments demonstrated faster clearance levels 45 (half-times < 0.5 s) and the posterior sections showed slower clearance (half-times < 1.0 s). 46 Both imaging methods showed similar flow-dependent tracer-gas clearance in the models.

INTRODUCTION 59 60
Respiratory failure is a common complication in a range of pulmonary conditions (12). 61 Recent studies report that an open nasal cannula system for delivering of Nasal High Flow 62 (NHF) can assist ventilation in patients with chronic respiratory failure (2, 3, 5, 10) and sleep 63 disorders (14,18). The concept of delivering high flow through the open nasal cannula is not 64 entirely new (24) but advancements in technology that efficiently warm and humidify 65 respiratory gases have been the key factor for clinical application of NHF. This form of 66 respiratory support is commonly used with a wide range of flow from 2 L/min in preterm 67 newborns to 60 L/min in adults with or without supplemental oxygen (7,25). NHF can also 68 be combined with a delivery of aerosolized drugs into the airways (1, 4 ). 69 70 A number of clinically relevant benefits have been associated with NHF therapy: reduction in 71 respiratory rate, an increase or decrease in minute ventilation, improved alveolar 72 ventilation, a reduction of wasted ventilation and the work of breathing (5, 10). However, 73 the mechanisms of how NHF produces these benefits are poorly understood. A mechanistic 74 study of NHF proposed two different ventilatory responses, one when awake and another 75 during sleep (16). The reduction of dead-space ventilation was proposed to be the principal 76 driver for the response during sleep. The mechanisms of dead-space clearance are difficult 77 to study due to the anatomical complexity and inability to visualize the gas flow in the upper 78 airways. However, many researchers have proposed dead-space clearance during NHF as the 79 major physiological mechanism that improves respiratory support (17,20,22). 80 Measurement of carbon dioxide (CO 2 ) concentration in the trachea confirmed this 81 hypothesis (21), and other studies have also reported a reduction of arterial and tissue CO 2 82 in response to NHF therapy (3,6). 83

84
In this study, the clearance of gas in dead-space with NHF was investigated using two upper 85 airway models. The first was a simple Tube Model (TM), which consisted of a nozzle and a 86 cylindrical tube. The nozzle represented the nasal valve area, which is the narrowest 87 constriction of the upper airways, and the tube characterized the volume of the upper 88 airways. The second was an Upper Airway Model (UAM), which more accurately represented 89 the complexity of the upper airways. For the simple TM, the gas clearance rates were 90 used to compare the imaging systems with different gas compositions, while the second 123 imaging system enabled visualization in the more anatomically accurate geometry. CO 2 124 MWIR absorption spectroscopy and 81m Kr-gas radioactive gamma camera imaging were used 125 as imaging systems in this study. The first imaging system used Carbogen, which better 126 represented the expired gases, and the second used air labelled with 81m Kr-gas which could 127 also be imaged through the more complex upper airway model. 128

MWIR absorption spectroscopy 130
The first imaging system used MWIR absorption spectroscopy to visualize CO 2 tracer-gas 131 clearance. It comprised a custom-made blackbody ( Figure 1A), controlled to 230 °C, to be 132 used as the heat radiation source, and a MWIR camera system (SC7600, FLIR, France), 133 together with a narrow-band pass filter 4260/20nm (Spectrogon, Sweden). The CO 2 -filled TM 134 was placed between the IR heat source and the MWIR camera system ( Figure 1A), which 135 detected heat radiation absorption by CO 2 in the tube. A Carbogen gas mixture of similar 136 composition to expired air (6% CO 2 , 21% O 2 and 73% N 2 , BOC Gases, Auckland, New Zealand) 137 was used as a tracer-gas in order to closely represent gas behavior during NHF therapy. 138 139 81m Kr tracer-gas imaging 140 The second imaging system used a planar gamma camera (Orbiter,Siemens,Erlangen,141 Germany) to visualize radioactive 81m Kr tracer-gas clearance. For these experiments, the 142 81m Kr-gas (TycoHC Covidien, Neustadt, Germany) was generated at a concentration of 1 to 143 2% in the entrained air. The models filled with the 81m Kr tracer-gas were placed in front of 144 the planar gamma camera. The gamma camera sampled images at a 25 Hz frame rate. The 145 clearance curves were then fitted with exponential functions and the clearance half-times 146 calculated. This data was corrected to allow for the level of natural 81m Kr-gas decay (T 1/2 = 13 147 s) and presented as corrected clearance half-times. The simplified geometry of the TM had two distinct compartments. The first was a single 151 nozzle representing the combined nasal valve area (inner diameter of 12 mm) of both noses, 152 machined from a single sodium chloride (NaCl) crystal. Directly coupled to the nozzle was the 153 second tube compartment, which represented the volume of the upper airways. This was fabricated from a grown sapphire crystal tube of dimension ID = 26 mm, OD = 31 mm, length 155 = 130 mm ( Figure 1A). This model was used for both CO 2 absorption spectroscopy imaging 156 and 81m Kr tracer-gas X-ray imaging. Both the NaCl and sapphire compartments exhibited 157 excellent transmission characteristics in the MWIR spectrum, allowing high-efficiency CO 2 158 imaging within the 4.26 μm absorption band, as well as the 81m Kr tracer-gas imaging. For 159 both tracer-gases, the TM was filled with the tracer-gas from the side opposite to the nozzle. 160 Then the administration of NHF was carried out via a cannula, which was placed coaxially in 161 the nozzle of the TM ( Figure 1A). The induced dilution and wash-out of the tracer-gases, or 162 dead-space clearance, was recorded by the appropriate imaging system, allowing direct 163 comparison of clearance rates within the TM. 164

165
As illustrated in Figure 1B, two regions of interest (ROIs) were defined in the simple TM, 166 those being the anterior section of the nozzle (TM1, volume 28 cm 3 ) and posterior part of 167 the nozzle (TM2, volume 25 cm 3 ). CO 2 clearance profiles for these ROIs were analyzed and 168 characterized by fitting exponentially decaying functions. The CO 2 clearance half-times for 169 these ROIs were then calculated. For the experiments using the 81m Kr-gas imaging system 170 with the simplified TM, images were captured for both the 81m Kr-gas filling and subsequent 171 15 s of NHF clearance. Similar ROIs to those used in the CO 2 imaging were applied to analyze 172 the 81m Kr-gas clearance characteristics from the TM ( Figure 1B An anatomically accurate 3D upper airway model was developed to better represent the 182 expected gas clearance when using NHF therapy in practice ( Figure 3A). The UAM was based 183 on segmented images from a Computer Tomography (CT) scan of a healthy volunteer, which 184 was then constructed using a High Definition (HD) 3D printer (Projet HD3000, 3D Systems 185 Inc., USA). Nasal valve area in both noses was 56 mm 2 . There was no anatomical structure beyond the nasal cavity and the unit led into an 18 mm ID tubing, which exited at the 187 bottom of the model. The material in the 3D printer is highly absorbent in the MWIR 188 spectrum. Therefore, only the 81m Kr-gas gamma camera imaging was used. The UAM was 189 then integrated into a plastic head model to enable the attachment of the cannula interface. 190 The position of the UAM in front of the gamma camera and the attachment of the NHF 191 cannula interface are shown in Figure 3A  condition were compared using a two-sided t-test with a significance level of p < 0.05. In 211 addition, differences were assessed using a paired t-test with the same significance level. A 212 Pearson correlation analysis was performed to assess correlations among study variables. 213

217
CO 2 -and 81m Kr-gas clearance in the TM 218 The clearance half-times for the two ROIs versus the NHF rates for the CO 2 clearance 219 experiments are shown in Table 1 and plotted in Figure 4A. The comparable clearance half-220 times of 81m Kr-gas are shown in Table 2  The 81m Kr-gas clearance half-times observed in the UAM at the NHF rates tested are 245 summarized in Table 3 and illustrated in Figure 4C. The characteristic dependencies of gas 246 clearance from the UAM are comparable to those during use of the TM. Clearance half-time 247 decreases by increasing the NHF rate (r = -0.84, p < 0.001). For both the TM and UAM 248 experiments, the anterior ROI (TM1 and UAM1, respectively) demonstrated the fastest 249 clearance rates (p < 0.01), and the clearance half-times decreased by increasing the NHF rate 250 for both ROIs. In addition, clearance half-times of 81m Kr-gas from the UAM's ROIs correlates 251 with the comparable TM's ROIs (r = 0.95, p < 0.001). 252

253
The total clearance rate from both UAM nasal cavity ROIs expressed in mL/s ( Figure 5)  being hypothesized as the clearance of dead-space resulting in a reduction of CO 2 -re-263 breathing (5, 13). However, dead-space clearance is very difficult to study in vivo due to the 264 challenges of quantifying this rapid process in an anatomically complex environment. 265 Animal models are of limited use because of their significant differences in the anatomy of 266 their upper airways. In this study, the dead-space clearance rates for two very geometrically 267 different models were studied. The effects of breathing on dead-space clearance were 268 excluded in order to guarantee reproducible experimental conditions. Therefore, the studies 269 were performed to simulate quasi-static breath-holding conditions. The tracer-gases had 270 been introduced into the models before NHF was commenced, to record sequences of 271 images. NHF therapy (16). The airway model was filled with either CO 2 or 81m Kr tracer-gas, and then 280 the wash-out and clearance characteristics caused by the delivery of high-flow air via the 281 cannula were quantified using either dynamic MWIR or radioactive gamma camera imaging, 282 respectively. Spectroscopic imaging of the Carbogen gas mixture within the TM allowed 283 visualization of the gas flow at 1000 frames per second. This was then post-processed to 284 allow quantification of the gas clearance rates, as CO 2 is a key component of the expired gas 285 and imaging this gas is physiologically relevant. However, only a small number of materials 286 that are very difficult to manufacture are transparent in the MWIR spectrum. This limited 287 the complexity of the upper airway model to a simple tube, with a valve region. Therefore, 288 the technique of gamma camera imaging of 81m Kr-gas was used, as this allowed much more 289 complex upper airway models to be imaged from a wider range of materials. One weakness 290 of the gamma camera imaging was that it had lower spatial and temporal resolution than 291 the MWIR spectroscopy. Gamma camera imaging can also be used in vivo to visualize the 292 airways in healthy volunteers because Krypton is an inert noble gas and the isotope 81m Kr 293 produces a very low radiation dose. 294

295
The results from both the CO 2 and 81m Kr-gas imaging have demonstrated very fast clearance 296 of the tracer-gas following the application of high flow through the cannula. The clearance 297 half-times in the simple TM nasal cavity were less than 0.6 s for both tracer-gases. There was 298 a similar flow-rate dependency of the compartmental clearance half-time for both tracer-299 gases, and higher flow rates significantly reduced the clearance half-times. The gases leaked 300 around the cannula and the clearance profiles suggest two specific characteristics: 1) for all 301 flows studied, clearance is faster in the anterior ROI and slower in the posterior ROI; 2) 302 clearance half-time decreases as the NHF rate increases for all ROIs. The strong correlation 303 between the data from both gases in the TM experiments shows that the two imaging 304 systems are comparable. The significantly slower acquisition rate of the 81m Kr-gas gamma 305 camera imaging system can still be used to study kinetics of these very rapid processes in 306 more complex anatomically realistic models where CO 2 visualization is inhibited. These 307 comparable results are not unexpected, as both tracer-gas mixtures had comparable 308 physical properties. The Carbogen gas is similar to ambient air with an elevated CO 2 309 concentration of 6%, while the 81m Kr-gas concentration was only 1 to 2% with the remainder 310 being ambient air. 311

312
In this study the upper airway models were filled with the tracer-gas. When flow of the 313 tracer-gas was stopped NHF was introduced. This quasi-static setup allowed for comparable 314 and reproducible experiments to be performed. This is in contrast to the condition in vivo, 315 where clearance of dead-space already starts before the end of expiratory flow. This was 316 demonstrated when expiring through the TM while a cannula flow of 30 L/min was applied 317 ( Figure 2B). Initially the model was always "filled" with exhaled CO 2 (Figure 2A), but the 318 same CO 2 was significantly cleared at the end of expiration when a cannula flow such as 30 319 L/min was present ( Figure 2B). Fresh gas from a cannula was observed within the model 320 even as the expiratory flow rate decreased from that of the peak flow, due to the difference 321 of dynamic pressure between the nozzle and the cannula flows. This highlights that the clearance of dead space can be significantly affected by the breathing pattern and that the 323 static condition in the experiments did not include the effects of breathing, thus 324 underestimating the speed and effectiveness of the clearance. 325

326
In the 3D-printed UAM, only the radioactive 81m Kr-gas tracer-gas clearance protocol was 327 followed, as the materials used to fabricate the model were not MWIR transparent. 328 Similarly, to the TM imaging analysis, the clearance rates were assessed in two adjoining 329 ROIs, those being the anterior (UAM1) and posterior (UAM2) regions of the nasal cavities. 330 The relationship between the clearance level and the NHF rate for the UAM was comparable 331 to those obtained using the TM, with the clearance half-time of the 81m Kr-gas decreasing by 332 increasing NHF rates. In both ROIs the change in clearance half-times was greater when the 333 NHF was increased from 15 to 30 L/min than from 30 to 45 L/min. However, the clearance 334 rates for both ROIs were calculated, and were shown to have a linear response to an 335 increase of the NHF rate ( Figure 5). An increased NHF rate of 1 L/min corresponded to ≈1.8 336 mL/s increase in the cleared nasal cavities volume in the UAM. In the deeper compartments 337 beyond the soft palate (oropharynx, trachea), clearance half-times were greater than 1 s 338 (see also supplementary video SupplVideo-UAM-Kr.avi). These deeper regions of the 339 conducting airways were not included in the data analyses, as clearance in these regions is 340 more likely to be subjected to changes in flow restriction due to variability in the shape of 341 the soft palate, the vocal cords or the mouth opening. Consequently, clearance of the 342 deeper regions of conducting airways has to be studied in vivo. 343 344

Physiological and clinical implications 345
It has been shown that NHF influences the gas exchange in the lungs with an increase in 346 oxygen blood saturation (6, 17) and a reduction of arterial CO 2 (17). A study on healthy 347 volunteers revealed that the effects of NHF on ventilation was also dependent on whether 348 the subject was awake or asleep: a reduction of tidal volume when awake and changes to 349 minute ventilation when asleep (16). The authors of this paper speculated that the reduction 350 of ventilation during sleep may be due to either a wash-out of anatomical dead-space or a 351 reduction in CO 2 production. 352 It is known that during normal breathing at rest approximately one third of tidal volume is 354 re-breathed from the anatomical dead-space (11). At the end of expiration, the dead-space 355 is filled with gas depleted in oxygen (15 to 16 % compared to 21 % in ambient air) and rich in 356 CO 2 (5 to 6% compared to 0.04% in ambient air). Therefore, an NHF-induced reduction of re-357 breathed CO 2 volume should either decrease the tidal volume or respiratory rate to maintain 358 the same alveolar ventilation. All of these conditions could eventually improve gas exchange. 359 Improving the gas exchange through the reduction of dead-space may either affect arterial 360 blood gases or reduce minute ventilation, with a potential reduction in the work of 361 breathing. Our data is limited to the dead-space in the nasal cavities, but supports both the 362 conclusions of Mundel et al. (16), that the reduction of dead-space is the primary 363 mechanism of decreased tidal volume with minute ventilation during sleep, and of Bräunlich 364 et al. (3) on the reduction of the respiratory rate and minute ventilation. Reduction of 365 respiratory rate, either through a decrease in re-breathing or through pressure effects as 366 previously described in detail (16), may lead to a further reduction of dead-space, as shown 367 by the strong time dependence of clearance in this study. We speculate that a reduction in 368 the respiratory rate may improve clearance by NHF therapy. Even in the absence of an end-369 expiratory pause, the slower respiration rate may lead to a more efficient clearance of dead-370 space and a reduction of re-breathing. 371

372
The ratio of dead-space to tidal volume (V D /V T ) increases during shallow breathing, which in 373 turn requires an increase of breathing frequency to maintain an adequate level of alveolar 374 ventilation. Physiological dead-space can be significantly increased in conditions like 375 emphysema in Chronic Obstructive Pulmonary Disease (COPD) or pulmonary embolism by 376 the elevated alveolar dead-space volume (11) that would lead to a high V D /V T ratio, or in 377 Acute Respiratory Distress Syndrome (ARDS) (9), which is associated with higher mortality. In 378 these cases, even a small reduction of dead-space would lead to a relatively high increase in 379 alveolar volume. In this study the ROIs were limited to the nasal cavities, which includes 380 between the nasal valve area and the soft palate, and have a combined volume of 55 cm 3 . 381 Typical nasal cavity volumes of 40 to 50 cm 3 were reported in healthy adults (23). Even this 382 anatomical volume comprises at least one third of the anatomical dead-space (8) in adults 383 and is significantly higher in children (19). . Clearing this upper airway dead-space could 384 therefore be quite significant for patients who have elevated V D /V T ratios. Similarly, the reduction of dead-space volume has been proposed as a mechanism that improves 386 ventilation during purse-lip breathing, through forcing the flow to be unidirectional and 387 bypassing the nasal cavities when exhaling (8). 388

389
The effectiveness of clearing the nasal cavities in the UAM ( Figure 5) has a linear positive 390 dependency with NHF treatment. Therefore, the nasal cavity clearance level rises with 391 increasing NHF rates, with every 1 L/min NHF increase leading to a 1.8 mL/s increased 392 clearance in the nasal cavities. Independent of the variations between the geometries of the 393 two upper airway models, the different tracer-gas properties, and imaging techniques used, 394 all the results demonstrated very similar clearance levels during changes in the NHF rates. 395 This further contributes to the notion that clearance of the dead-space, especially within the 396 nasal cavities, is strongly affected by the NHF rates. The size of the cannula may also have an 397 influence on dead-space clearance due to the higher velocity of gas for a given flow. Smaller 398 cannula may lead to more efficient dead-space clearance due also to more space, and hence 399 leak, around the cannula. However, Mundel et al. (16) reported that the use of larger 400 cannula leads to higher expiratory pressures, which may potentially reduce the respiratory 401 rate and increase the tidal volume during wakefulness; this may, in turn, improve the 402 efficiency of clearance and alveolar ventilation. 403

404
This study has investigated dead-space clearance under a quasi-static breathing condition, 405 which occurs in the period between expiration and inspiration. At this stage of the breathing 406 cycle the flow rates are reduced from low to no flow, shortly before reversing direction. 407 During a normal breathing cycle this may take between 0.5 and 1.0 s, which would allow 408 sufficient time for a tracer-gas to be washed-out, based on the experimentally determined 409 clearance half-times in the airway models used in this study ( Figure 5 and Tables 1, 2 and 3). 410 The quasi-static experimental condition resulted in an underestimation of the level of total 411 clearance by NHF therapy, but allowed the clearance rates of these tracer-gases to be 412 studied without the added complexity of respiration. The second key strength is that the upper airway clearance has been investigated using two 438 distinct imaging modalities that use different tracer-gases. Visualization and analysis of CO 2 439 using MWIR transmission spectroscopy provided high temporal and spatial resolution for 440 studying the clearance rates in a model with a simple geometry. Gamma imaging of 81m Kr-441 gas also produced comparable clearance rates in the same simple geometry model, as well 442 as in a more complex but realistic upper airway model. Both methods and models produced 443 comparable results that demonstrated the same clearance dynamics with increasing the NHF 444 rates. The gamma ray imaging of 81m Kr-gas is of particular importance as this technique can 445 be implemented for in vivo experiments. 446

447
There are several limitations in this study. The main drawback is that all experiments were 448 performed with in vitro models, and only static clearance rates in the absence of breathing were quantified. The clearance responses to a range of tidal volumes and breathing patterns 450 were not investigated during this study. The addition of breathing will only accelerate the 451 clearance of the tracer-gas from dead-space; therefore, the results presented in this study 452 underestimated the clearance levels. It was decided to limit the scope of this study to allow 453 the accurate quantification of the NHF clearance rates in a simplified but repeatable In summary, this study has shown effective clearance of the tracer-gas, demonstrating 461 similar dynamic characteristics despite the very different geometries of the upper airway 462 models. The clearance is linearly related to the NHF rate with an anterior portion of the nasal 463 cavities clearing faster than the posterior portion. We conclude that clearance of the nasal 464 component of the anatomical dead-space with NHF therapy is a rapid process, which may 465  Table 1: Half-times (T 1/2 ) of CO 2 -gas clearance (mean +/-standard deviation, medium, 569 minimum and maximum) in the proximal (TM1) and medium (TM2) ROIs of the Tube Model 570 (TM) during flow from a cannula for Nasal High Flow (NHF) rates of 15, 30 and 45 L/min (**: 571 p < 0.01 compared to 15 L/min NHF, ++: p < 0.01 for TM2 compared to TM1).  Table 2: Half-times (T 1/2 ) of 81m Kr-gas clearance (mean +/-standard deviation, medium, 575 minimum and maximum) in the proximal (TM1) and medium (TM2) ROIs of the Tube Model 576 during flow from a nasal cannula for Nasal High Flow (NHF) rates of 15, 30 and 45 L/min (**: 577 p < 0.01 compared to 15 L/min NHF, ++: p < 0.01 for TM2 compared to TM1).