Aerosol drug delivery to the lungs during nasal high flow therapy: an in vitro study

Long-term oxygen therapy can improve survival in patients with chronic obstructive pulmonary disease (COPD) and chronic respiratory failure [1, 2]. Nasal high-flow (NHF) therapy is a form of respiratory support used in the hospital or emergency unit [3], mainly for management of acute hypoxaemic respiratory failure [4]. NHF therapy delivers oxygen (often warm and humidified) to patients at flow rates higher than that used in traditional oxygen therapy. Warm and humidified air may eliminate the side-effects associated with conventional oxygen therapy including upper airway dryness and irritation plus mucociliary clearance interference [3, 5]. A substantial number of COPD patients suffer from exacerbations, which are defined as an acute worsening of respiratory symptoms [6]. Acute exacerbations can be treated and sometimes prevented with inhaled antibiotics, bronchodilators or corticosteroids [7‐9].

Hypoxemic patients using an NHF system may benefit from combined aerosol therapy as the etiology of hypoxemia might justify the administration of aerosolized medication [10]. In vitro studies have investigated whether pressurized metered-dose inhaler (pMDI), nebulizers or dry powder inhalers (DPI) can be combined with NHF systems for simultaneous administration of oxygen and pharmaceutical aerosols [11‐16]. Réminiac et al. [11, 12] found that the position of the nebulizer or pMDI in the NHF circuit is profoundly important. Placing a nebulizer before the humidification chamber resulted in 26–32% emitted dose from the nasal prongs [11], whereas placing a pMDI immediately upstream of the nasal cannula resulted in 12% emitted dose. Ari et al. [13] and Bhashyam et al. [14] performed experiments with similar nebulizer setups and achieved 2–11%, and 19–27% emitted doses, respectively. Perry and his team [15] placed a nebulizer further away from the humidification chamber (closer to the nasal prongs) and found the emission efficiency being only 2.5% in the study. Dugernier et al. [10] reported that lung deposition in vivo was 4 and 1% with a vibrating-mesh nebulizer and a jet nebulizer, respectively.

Dry powders were thought to be incompatible with an NHF system because of humidified air [17]. Water may adsorb to the surface of dry powders when the humidity is high, thereby compromising the flowability and dispersibility of the powders due to agglomeration and increased adhesiveness [18]. The use of dry powders in such systems has been neglected for that reason [16, 19, 20]. Nevertheless, we have previously shown that heated and humidified air could disperse mannitol powders as effectively as dry air [16]. However, the predicted lung dose was only 8% in that in vitro setting, limiting its clinically utility [16].

In the present study, we aimed to develop an efficient nose-to-lung delivery system using a DPI device coupled to a NHF system that can overcome the current clinical and technical limitations, with improved delivery (FPF > 15%) of powder aerosols to the lungs.

Our study demonstrated promising results compared to those in the literature [11, 13, 15, 16]. The highest FPF previously [16] was 7.99 ± 0.75% versus 32.15 ± 0.81% in the present study. A nebulizer was used in a similar setup by Reminiac and colleagues [11], where the highest respirable mass found was 10%. A pMDI was used in another study by Reminiac et al. [12], where the highest emitted dose was 12% during normal breathing. In other studies with nebulizers, the highest reported emitted doses were 27 and 11% [13, 14]. Only one in vivo study was found in the literature [10]. However, aerosol delivery to the lungs through a NHF system was only 1–4% of the nomimal dose leading the authors to conclude that the concept should be optimized further before we can expect a significant effect with nebulized antibiotics [10]. Our study may have clinical relevance, as our setup is capable of delivering high quantities of an aerosol powder to the lungs. It allows clinical evaluation of powder aerosol delivery to the lungs during NHF therapy in humans.

The high FPFs could be attributed to the relatively low DFRs. Generally, powder is dispersed more efficiently at higher dispersion flow rates [16, 33, 34]. However, higher DFRs also lead to more impaction loss [11, 16]. Therefore, if a powder were dispersible, it will be desirable to use a lower flow rate. High resistance devices can achieve efficient dispersions at a lower flow rate, e.g., the Handihaler™. A legitimate concern with a higher resistance inhaler is whether an adequate flow rate can be generated [33]. However, our setup does not rely on a patients’ ability to inhale. Instead, the Handihaler™ is activated by an external air source. Second, the good dispersibility of our Man+Leu formulation was an essential reason. The addition of 20% leucine improved the flowability and dispersibility of the powder compared with that of pure mannitol (Table 3). The D₅₀ and span were smaller for Man+Leu than mannitol. The PSD values of Man+Leu were essentially the same irrespectively of the DFR. In contrast, mannitol required a higher flow rate to achieve a more satisfactory PSD, which is also what we observed previously [16].

The nasal cannula and connection tube (Fig. 6c) accounted for substantial deposition loss ranging from 23.46 ± 0.59% up to 41.54 ± 11.42%. The geometry inside this region presumably caused the large deposition (Fig. 8). It can be appreciated that the inside of the Handihaler chamber connection region has flow constrictions where powder can deposit. Likewise, aerosols could be trapped in possible recirculating regions around the entry of the connection tube. Deposition could also occur in the tube. At DFR of 20, 25, and 30 L/min, Reynolds number (Re) values were approximately 2872, 3591 and 4309, respectively. Re was calculated with Reynolds equation \( \left(\mathit{\operatorname{Re}}=\frac{\rho Vd}{\eta}\right) \) [35]. Thus, at 20 and 25 L/min, Re was in the transitional region for a circular pipe, while 30 L/min, the flow would be turbulent, although the velocity profile is likely not fully developed due to the relatively short length of the connection tube. Regardless, it is likely that turbulent dispersion likely plays some role in wall deposition in the connection tube. When the aerosols exit the connection tube and enter the nasal cannula, the flow direction changes 90 degrees. A sudden change in direction may also cause impaction in the back of the cannula, especially for particles with greater inertia. The Stokes number (Stk) is valuable to predict whether aerosols are likely to deposit in the cannula bend. According to theory, particles with Stoke number much less than one (Stk < < 1) are expected to follow gas streamlines. When Stk > > 1, particles will continue its original direction when the gas turns, rather than following the flow streamlines [35]. Using the D₅₀-value of Man+Leu and a flow rate of 30 L/min, Stk = .05 which is probably small enough that particles are not much affected by changes in airflow direction. However, using the D₉₀-value we found Stk = .7, indicating particles of that size probably would be affected when the airflow changes direction, which may explain our recovery of a significant amount of powder in the cannula.

We found the overall replica deposition was unaffected by the dispersion flow rate (Table 4). In contrast, it has been reported that replica deposition is expected to increase with increasing DFR [11, 16]. One possible explanation is the relatively small PSD and good dispersability of the present Man+Leu formulation. Just like the situation in the nasal cannula, small particles will follow the flow irrespective of the airflow. This indicates that the particle size predominantly determines the replica deposition. This is supported by a computational fluid dynamic deposition study, where regional deposition of nasal sprays in the airways of the nose was explored across different physical parameters [36]. The percentage of particles reaching the lungs was found to be relatively insensitive to the injection velocity whereas particle size showed a bigger influence on the deposition in the nose [36]. However, we did observe increased deposition in the nasal turbinates and nasopharynx when the IFR was increased (Fig. 7). We observed the same previously [16]. This may be due to some combination of enhanced impaction and turbulent deposition at the higher flow rates. Reynolds number Re in the replica can be calculated using the modified equation provided by Golshahi et al. [23]. At IFR 20, 25, 30 and 40 L/min, the replica specific Re were 2087, 2609, 3131 and 4174, respectively. Abrupt local diameter changes in the nasal cavity can trigger the onset of turbulence. (If air flows through a diverging duct, then the transition from laminar to turbulent can happen at a Re considerably lower than 2000 [37]). Thus, the onset of local turbulence or increases in separated flow region size could have caused more deposition in the replica at the higher flow rates. In addition, Stokes number Stk increases with flow rate, which may result in increased impaction.

IFR had a positive effect on the FPF in the experiments (Fig. 7b). At higher IFR more powder was drawn into the NGI. More powder in the NGI lead to a higher FPF. We had similar observations previously [16].

Our work has limitations that need to be addressed and improved for future studies. First, the current setup was not integrated into a clinically-approved NHF system. The AIRVO was not compatible with our setup. The AIRVO system could not reach a specific airflow quickly enough to be used for the short ‘burst’ in our experiments. Second, the air source was dry and not humidified oxygen. Conceptually, it would have been more accurate to use humid air instead of dry air. However, Okuda et al. [16] found the dispersibility of spray-dried mannitol was not affected by the air source due to the low exposure time. For Man+Leu powder, the effect of humidified air would probably be negligible due to the presence of leucine on the surface of the particles. Li et al. [21] found l-leucine protects powders from moisture-induced deterioration. Even if we used humidified air, the powder would only have been exposed for a short time. Additionally, the viscosity of air and oxygen do not differ much (both kinematic and dynamic viscosity). Thus, using air and not oxygen is not likely to have altered our results much since neither Re or Stk was much affected. Third, our design can be significantly improved. The dead space volume can be reduced by reducing the volume of the ‘Handihaler Chamber’. The silicon adapter inside the Handihaler chamber was not a perfect fit for the Handihaler™. If the Handihaler is not firmly inserted, air might bypass the device and ruin the dispersion. The connection tube should be replaced with a type with a smooth, rather than corrugated, inner surface. The Optiflow™ used here comes with a spiral corrugated connection tube. It has a relatively rough surface that could cause additional deposition. Fourth, in vitro studies with replicas of nasal airways have limitations. The extrathoracic geometries vary significantly between individuals [38]. Our replica was based on the MRI data of a single human being [23]. Finally, a vacuum pump was used during the experiment to simulate inspiration of a person. The flow was constant and is not realistic. A more realistic inspiratory airflow would use e.g. a sine function vs time. By replacing the constant flow condition with realistic breathing profiles, more representative results can be obtained. Although this study has limitations, our results demonstrate that the system can effectively deliver aerosols to the lungs.

Particle deposition in the nose and extrathoracic area is affected by the size of the airway [23, 38, 39]. Therefore, it would be valuable to validate our findings in healthy human subjects. Investigating subject-specific deposition in humans would be relevant as well. The topography of the nasal airway can be accurately determined with an acoustic rhinometer [40]. Furthermore, in vivo studies can be used to validate in vitro models. Results obtained from people with a Caucasian background may not apply to people with an Asian background as the nasal geometry is different [41]. Thus, investigating potential differences in deposition between human beings with different race would also be an entirely new topic to consider.

In conclusion, we have successfully developed an in vitro physical model capable of delivering large quantities of aerosols to the lungs with a nasal cannula. The highest fine particle fraction obtained was 32%, and the lowest fraction was 21%. Our work demonstrates that dry powder inhalers may be practical for NHF systems. Our results may lay the foundation for clinical evaluation of powder aerosol delivery to the lungs during NHF therapy in humans.