Results of in vitro experiments evaluating tracheal pressures and EtCO2 during delivery of CPAP or NHF to child airway replicas are reported above. Several differences between CPAP and NHF warrant further discussion, as do the potential sources of variability in pressure and gas washout between airway replicas.
For the delivery of CPAP, PEEP was observed to be approximately constant across the 10 airway replicas at either 5 cmH2O or 10 cmH2O (Fig. 5), indicating that the CPAP machine was working as intended, and delivered targeted positive airway pressures. In contrast, PEP, MIP, and AIP were observed to vary between replicas, indicating that these three pressure parameters were influenced by additional factors including breathing flow rates and the airway geometries of each subject (Figs. 5 and 6). This was expected, as airway pressure was evaluated at the exit of each replica (representative of a tracheal pressure), such that pressure drop through the replica influenced the airway pressure in all cases where flow was nonzero. In contrast, PEEP was measured at a point on the breathing cycle of zero flow, such that the instantaneous pressure drop through the replica is also zero.
Unlike the CPAP machine, the NHF system does not adjust delivered flow rate to maintain a constant pressure. As such, all pressure parameters, including PEEP, were observed to be variable across the 10 airway replicas for the delivery of NHF, with negative pressures observed during inhalation for 3 of 10 replicas (Figs. 5 and 6). With a set flow rate of 20 L/min, the average PEEP across the 10 airway replicas was approximately 5 cmH2O, which is similar to a CPAP setting of 5 cmH2O. Accordingly, though NHF can generate positive airway pressures, the pressures are variable and subject-dependent. McGinley et al. [20] reported on the delivery of NHF as an alterative to CPAP for children aged 10 ± 1 years (mean ± SEM; n = 12) at a set flow rate of 20 L/min. In their study, they found similar reductions in apnea–hypopnea index, comparable to CPAP prior to the study, when using NHF in a majority of the children studied [20]. Prior to NHF, the average CPAP setting used for therapy was 9 ± 1 cmH2O (mean ± SEM; n = 10) [20].
An increase in EtCO2 from baseline was observed during CPAP therapy across all 10 upper airway replicas. The presence of a mask increased EtCO2, due to added dead space of the mask. This increase was smallest for CPAP at 10 cmH2O (Fig. 7), owing to the greater average flow rate delivered from the CPAP machine at the higher CPAP setting. In contrast, a reduction in EtCO2 from baseline was observed during NHF therapy across all 10 upper airway replicas. This is consistent with a known mechanism of NHF: washout of the nasopharyngeal dead space, leading to reduced rebreathing of expired air [24, 32]. It is notable that, due to differences between the NHF cannula interface and CPAP mask interface, effective washout was observed for NHF at a flow rate of 20 L/min, whereas no, or limited, washout was observed for CPAP with an average delivered flow rate of 18.8 L/min (for CPAP at 5 cmH2O), or 26.1 L/min (10 cmH2O). During exhalation, any flow delivered by the CPAP machine is diverted through the exhalation port, such that little mixing occurs with gases in the mask or upper airway.
No significant difference was observed in tracheal pressures nor change in EtCO2 between the three different NHF cannulas for the subset of five tested replicas. An average PEEP of 5.4 ± 1.6 cmH2O, 4.3 ± 1.5 cmH2O, and 3.5 ± 0.5 cmH2O were generated through the Optiflow 3S, +, and Junior 2 nasal cannula, respectively (Fig. 8). Though not statistically significant, differences in average PEEP between cannula models may be associated with different cannula prong sizes, as has been noted to influence PEEP in previous studies [33, 34]. All three nasal cannulas also had similar reductions in EtCO2 from baseline: − 0.5 ± 0.3% for the Optiflow 3S, − 0.4 ± 0.2% for the Optiflow +, and − 0.4 ± 0.2% for the Optiflow Junior 2 (Fig. 10). However, only five replicas were tested because two of the three nasal cannula models, the Optiflow 3S and the Optiflow +, did not fit the five remaining replicas. This indicates that the selection of nasal cannula for NHF is important for fit and preventing blockage of the nares during delivery of therapy. Relationships between reduction in EtCO2 from baseline with tidal volume and replica volume were also investigated; however, no correlation was observed. It may be that variability in gas washout during NHF was influenced by the shape of the replica airways, especially the nasal vestibule in immediate proximity of cannula prongs; however, this was not investigated in detail in the present study.
The increased variability between replicas in tracheal pressures generated during NHF as compared to CPAP is noticeable in Figs. 5 and 6. Variability in PEEP between replicas was accounted for in part by modeling the pressure drop through the annular space between the prongs and nostril walls as a minor loss. Such a model is frequently adopted in fluid mechanics to calculate the pressure drop associated with flow through a constriction or past an obstruction. On average, calculated minor loss coefficients did not vary appreciably between the three NHF cannulas studied. Furthermore, minor loss coefficients remained approximately constant across the range of Reynolds numbers studied (Re = 950–1350), as is typically observed for flow through a constriction [30]. Similarly, Katz et al. [35] previously adopted a minor loss model for the pressure drop through extrathoracic and bronchial airways, and observed that minor loss coefficients approached constant values as Reynolds numbers exceeded ~ 1000. In the present work, this relationship suggests that PEEP generated in the replicas by NHF was related primarily to the occlusion of the nares by the cannula prongs. For a fixed flow rate of gas supplied to the cannula, the greater the extent of occlusion, the larger the PEEP that will be generated [36].
Some variability in calculated minor loss coefficients persisted between replicas, and can be attributed primarily to the dissimilar shape of the annular space for different replicas, which is not fully accounted for in the use of a single length scale, namely the hydraulic diameter calculated in Eq. 2. Variation in the percentage of the nostrils’ inlet area occluded by cannula prongs may also have contributed to variability between replicas in the minor loss coefficients. The greater variability in minor loss coefficient between replicas for the Optiflow Junior 2 cannula, as compared with the other two NHF cannulas studied, likely resulted from the larger number of replicas investigated with this cannula. For the subset of five replicas tested with all three NHF cannulas, the percent of occlusion ranged from 34 to 47% for the Optiflow 3S, 33–45% for the Optiflow +, and 20–27% for the Optiflow Junior 2. When tested over the larger set of 10 replicas, the percent of occlusion ranged from 20 to 41% for the Optiflow Junior 2.
Previously, Moore et al. [33, 34] identified predictive correlations for PEEP generated during application of NHF based on a characteristic air speed through the non-occluded nares area, as in Eq. 1 of the present study, but also influenced by an additional characteristic air speed exiting the cannula prongs. In the present work, consideration of this additional characteristic air speed did not further improve our ability to account for variability in PEEP between nasal cannulas. This may in part be due to the limited range of air speeds exiting cannula prongs in the present study, which was conducted with a single flow rate supplied to nasal cannula. Furthermore, the Moore et al. studies included high flow nasal cannula from a different manufacturer, which are intentionally designed with smaller inner diameters to influence washout of the upper airway [37].
A limitation of this study is the use of rigid airway replicas. They did not deform during breathing or under positive airway pressures, and thus the dynamic effects of breathing are not fully captured. Additionally, airway replicas used in the present study were fabricated based on scans of children that were obtained for indications other than airway pathology, whereas children with OSA may have reduced upper airway dimensions compared to controls [38]. We tried to minimize these limitations by testing multiple airway replicas to cover a range of differing airway geometries. Variation in, e.g., airway volume or cross-sectional areas between different airway replicas is expected to be much greater than variation that occurs dynamically over an individual’s breathing cycle. Furthermore, the range of airway dimensions measured in children with OSA overlaps that measured in controls [38], such that we expect the conclusions of the present work to extend to airway geometries representative of children with OSA. A second limitation is the testing of only one flow rate setting for NHF, 20 L/min, for our airway replicas with a subject age range of 4–8 years old. Previous studies have shown both airway pressures and washout to be flow rate dependent [33, 39]. However, clinical studies by McGinley et al. and Amaddeo et al. both used 20 L/min when investigating the use of NHF therapy as a treatment for OSA in children, aged 10 ± 1 years and 8.9 ± 6.2 years respectively [20, 21]. In both studies, NHF therapy at 20 L/min had a positive effect in treating OSA [20, 21]. Therefore, we focused on NHF at 20 L/min as a clinically-relevant flow rate for children with OSA.
