High-Efficiency Dry Powder Aerosol Delivery to Children: Review and Application of New Technologies

Abstract

While dry powder aerosol formulations offer a number of advantages, their use in children is often limited due to poor lung delivery efficiency and difficulties with consistent dry powder inhaler (DPI) usage. Both of these challenges can be attributed to the typical use of adult devices in pediatric subjects and a lack of pediatric-specific DPI development. In contrast, a number of technologies have recently been developed or progressed that can substantially improve the efficiency and reproducibility of DPI use in children including: (i) nose-to-lung administration with small particles, (ii) active positive-pressure devices, (iii) structures to reduce turbulence and jet momentum, and (iv) highly dispersible excipient enhanced growth particle formulations. In this study, these technologies and their recent development are first reviewed in depth. A case study is then considered in which these technologies are simultaneously applied in order to enable the nose-to-lung administration of dry powder aerosol to children with cystic fibrosis (CF). Using a combination of computational fluid dynamics (CFD) analysis and realistic in vitro experiments, device performance, aerosol size increases and lung delivery efficiency are considered for pediatric-CF subjects in the age ranges of 2–3, 5–6 and 9–10 years old. Results indicate that a new 3D rod array structure significantly improves performance of a nasal cannula reducing interface loss by a factor of 1.5-fold and produces a device emitted mass median aerodynamic diameter (MMAD) of 1.67 μm. For all ages considered, approximately 70% of the loaded dose reaches the lower lung beyond the lobar bronchi. Moreover, significant and rapid size increase of the aerosol is observed beyond the larynx and illustrates the potential for targeting lower airway deposition. In conclusion, concurrent CFD and realistic in vitro analysis indicates that a combination of multiple new technologies can be implemented to overcome obstacles that currently limit the use of DPIs in children as young as two years of age.

Graphical Abstract

1. Introduction

Compared with other pharmaceutical aerosol approaches, dry powder inhalers (DPIs) offer a number of advantages in respiratory drug delivery including formulation stability, rapid dose administration, the potential for high inhaled doses, minimal cleaning and automatic coordination between inhalation and dose delivery (De Boer et al., 2017; Islam & Cleary, 2012; Smith et al., 2010; Weers et al., 2010; Weers et al., 2019). Despite these advantages, the use of DPIs with children is often challenging as they may not be able to perform adequate inhalation maneuvers (Devadason, 2006; Goralski & Davis, 2014; Lexmond, Hagedoorn, Frijlink, Rottier, & de Boer, 2017), and also due to high mouth-throat (MT) depositional loss, low lung delivery efficiency and often high intersubject variability (Below, Bickmann, & Breitkreutz, 2013; Devadason et al., 1997; Lindert, Below, & Breitkreutz, 2014; Ruzycki, Golshahi, Vehring, & Finlay, 2014). For commercial products, upper airway depositional losses of 80% or higher of the emitted dose are common for children in the age range of 5 up to approximately 14 years (Below et al., 2013; Devadason et al., 1997; Lindert et al., 2014; Ruzycki et al., 2014). Across a number of current dry powder inhalers (DPIs), pediatric lung delivery efficiencies are in a range of 5–30% of nominal dose, based on in vivo (Devadason et al., 1997) and realistic in vitro experiments (Below et al., 2013; Lindert et al., 2014; Ruzycki et al., 2014). Potentially more significant than the low dose delivery efficiency is the high intersubject variability associated with pediatric DPI use. For example, using the correlations for pediatric patients from Golshahi et al. (2011), applied to respirable aerosols at typical flow rates for a child (2.5 μm and 15 LPM), results in nasal cavity deposition fractions ranging from 3.4 to 24.3%. Furthermore, Lindert et al. (2014) reported a 2-fold difference in lung dose from a DPI based on inhalation waveform conditions even without varying the airway anatomy. Devadason et al. (1997) reported coefficients of variation near 100% for pediatric DPI use by 3 to 5-year-old children. As with adults, reducing extrathoracic depositional loss with an inhaler is expected to reduce variability in the lung delivery of the medication (Borgstrom, Olsson, & Thorsson, 2006).

Beyond the physics of efficient aerosol formation and extrathoracic depositional loss, children also frequently have difficulty in using passive DPI devices that operate on negative inhalation pressure and are almost always designed for adults (Everard, 1996, 2004; Goralski & Davis, 2014). It is not until approximately 6 years of age that children can reliably perform a correct DPI inhalation maneuver with a passive device (Devadason, 2006; Goralski & Davis, 2014; Lexmond et al., 2017). A recent study by Lexmond et al. (2014) in children approximately 5–12 years old with an oral DPI simulator reported that 90% of inhalations were associated with oral obstruction, arising from the tongue and cheeks, during negative pressure inhalation against the variable resistance device. Lexmond et al. (2014) also observed that for 5-year-old subjects, only half of the trained children could successfully perform a passive DPI inhalation maneuver. Furthermore, between 10 and 44% of the trained pediatric subjects exhaled into the DPI simulator, which is expected to adversely affect aerosolization of the powder due to moisture exposure.

While the challenges of delivering dry powder aerosols to children are daunting, it is our opinion that both infants (0–2 years) and children (2–12 years) can significantly benefit from a properly designed DPI that is capable of reliable high-efficiency aerosol delivery to the lungs. Key problems related to DPI aerosol use in children can be characterized as: (i) exclusive use of oral administration with passive (adult) devices that operate on negative inhalation pressure, (ii) turbulence and jet momentum that result in high device and upper airway depositional aerosol loss, and (iii) implementation of relatively large and static aerosol sizes. Considering oral administration, it is generally held that oral aerosol inhalation produces significantly higher lung delivery efficiency compared with nose-to-lung delivery (Chua et al., 1994; Everard, Hardy, & Milner, 1993). While this is generally true, it is often not considered that for sufficiently small particles at low flow rates, oral or nasal deposition loss becomes similar and can generally be held below 5 or 10%. Furthermore, nose-to-lung delivery offers some advantages such as treating the entire respiratory airways and enabling use in infants and young children when coupled with positive-pressure devices. In contrast, considering pediatric inhalation through passive DPIs, the increased flexibility of pediatric upper airways and the oral obstructions observed by Lexmond et al. (2014) are expected to produce high depositional loss in vivo, potentially greater than predicted with realistic (but rigid) in vitro models. As the negative pressure generated by passive devices was observed to narrow the airways, it is reasonable to expect that a positive-pressure device would expand the flexible upper airways and thereby improve lung penetration of the aerosol. Use of a positive-pressure aerosol source may improve device reproducibility in the pediatric population.

It is well known that turbulent jets from DPIs significantly increase both device and extrathoracic depositional loss (DeHaan & Finlay, 2001, 2004; Longest, Tian, Delvadia, & Hindle, 2012; Longest, Tian, Walenga, & Hindle, 2012; Matida, DeHaan, Finlay, & Lange, 2003; Tian, Longest, Su, Walenga, & Hindle, 2011). As recently described by Weers et al. (2019), this explains why simply increasing device resistance (which decreases device flow rate) does not reduce MT deposition. Similarly, inhaling a smaller particle size without a properly designed inhaler will not necessarily improve lung delivery efficiency (Longest, Hindle, Das Choudhuri, & Xi, 2008).

Finally, new reportedly small particle DPIs do not produce a small “apparent aerosol diameter” based on extrathoracic depositional loss (Weers et al., 2019). For example, the recently available “extrafine particle” NEXThaler DPI has an impactor measured mass median aerodynamic diameter (MMAD) of 1.5 μm, but an apparent aerosol size of 4.7 μm based on human subject deposition data (Virchow et al., 2018). The reasons for the discrepancy between the measured MMAD and the apparent aerosol diameter are most likely turbulence and jet or spray momentum emanating from the device, as well as current pharmaceutical testing methods that exclude the induction port deposition fraction, which is often 50% or more of the aerosolized dose, in size determination. Moreover, a static aerosol size may not perform as well as an aerosol that is capable of entering the airways with a relatively small size and then increasing in size within the airways in a controlled manner (Hindle & Longest, 2010; Longest, Tian, Li, Son, & Hindle, 2012; Tian, Longest, Li, & Hindle, 2013).

In order to overcome the primary limitations associated with dry powder aerosol delivery to children, as well as many limitations that are also present in adults, a number of technologies have recently been developed or re-discovered and substantially extended. These technologies include:

1. Nose-to-lung aerosol administration with sufficiently small particles (Bass, Boc, Hindle, Dodson, & Longest, 2019; Golshahi et al., 2011)

2. Use of active positive-pressure DPIs (Farkas, Bonasera, Bass, Hindle, & Longest, 2020; Farkas, Hindle, & Longest, 2018b; Farkas, Hindle, & Longest, 2018c)

3. Patient interfaces that reduce turbulence and jet momentum effects without substantially increasing particle depositional loss (Bass & Longest, 2020; Farkas et al., 2020; Longest, Son, Holbrook, & Hindle, 2013)

4. Highly dispersible spray-dried powder formulations (Chan, 2006; Vehring, 2008; Vehring, Foss, & Lechuga-Ballesteros, 2007; Weers & Miller, 2015) that change size within the airways (Hindle & Longest, 2012; Son, Longest, & Hindle, 2012)

The following sections review recent progress in each of these areas. The use of concurrent CFD and in vitro analysis is then summarized, which has enabled rapid development of these new technologies. Following this review, concurrent analysis is implemented to explore the simultaneous application of these new technologies in order to improve the trans-nasal delivery of dry powder aerosols to children with cystic fibrosis across an age range of 2–10 years old.

1.1. Nose-to-lung aerosol administration in children using sufficiently small particles

A number of previous studies have characterized the nasal deposition of ambient inhaled particles in pediatric nasal models (Golshahi & Finlay, 2012; Golshahi, Noga, & Finlay, 2012; Golshahi et al., 2011; Golshahi, Vehring, Noga, & Finlay, 2013; Xi, Si, Zhou, Kim, & Berlinski, 2014; Xi, Berlinski, Zhou, Greenberg, & Ou, 2012; Xi, Si, Kim, & Berlinski, 2011). The term ambient particles refers to an aerosol that is inhaled from the environment without jet or spray momentum, as typically occurs with a pharmaceutical aerosol generation device, and without a patient interface, such as a mouthpiece (MP) or nasal cannula (NC). From these studies, it is well known that particles with a sufficiently small impaction parameter (d_a²Q – where d_a is the particle aerodynamic diameter and Q is the inhalation flow rate) produce low depositional loss in nasal airways, even for infants and children. For example, the d_a²Q impaction parameter for a 1.7 μm aerosol delivered at 15 LPM is approximately 40 μm² L/min, which, based on the study of Golshahi et al. (2011) across 14 pediatric nasal models (4 to 14 years old), corresponds to nasal deposition in the range of approximately 0 to 12%. In contrast, employing a typical DPI aerosol size of 6 μm at a flow rate of 15 LPM results in a predicted nasal deposition range of 25 to 75% based on the data of Golshahi et al. (2011).

The high depositional losses of ambient particles are consistent with most nebulized drug delivery studies employing pharmaceutical aerosols with a droplet size that is approximately 5 μm or greater and a mask or nasal cannula interface. For a mesh nebulized aerosol, El Taoum et al. (2015) reported lung doses of 0–3% of nebulized drug in an anatomically realistic nasal cavity model of a 5-year-old child that included cyclic respiration and a variety of patient interfaces. Other bench-top testing studies that have considered nebulized aerosol delivery with a mask interface have also reported approximately 10% or less lung delivery efficiency in pediatric subjects (Lin, Harwood, Fink, Goodfellow, & Ari, 2015; Smaldone, Sangwan, & Shah, 2007). Ari et al. (2011) considered pediatric trans-nasal aerosol delivery with a nasal cannula interface without a nasal model and also reported lung delivery efficiency of approximately 10% and below (relative to total dose).

While it is relatively well known that smaller particle size can significantly improve the delivery of pharmaceutical aerosols to infants and children, this approach has not been widely applied. One exception is the recent in vitro study of Bass et al. (2019) in infants, which demonstrated high efficiency lung delivery of a pharmaceutical aerosol using a mesh nebulizer, custom mixer-heater (Spence, Longest, Wei, Dhapare, & Hindle, 2019), and excipient enhanced growth (EEG) (Hindle & Longest, 2012) aerosol formulation. The mixer-heater device was used to heat the gas stream to a temperature that remained safe for direct inhalation and reduced the aerosol mass median aerodynamic diameter (MMAD) to approximately 1.5 μm, indicating dry particle formation. A streamlined nasal cannula interface was used to further reduce device system and nasal airway depositional loss (Longest, Golshahi, & Hindle, 2013). In vitro experiments and corresponding computational fluid dynamics (CFD) simulations demonstrated >90% delivery efficiency of the nebulized dose to a tracheal filter (Bass, Boc, et al., 2019).

Reasons that small particle aerosols may not commonly be used for pharmaceutical aerosol delivery to children include: (i) low dose delivery rates, (ii) difficulty in generating the small aerosol size, and (iii) high potential to exhale dose. Dry powder inhalers can frequently be used to rapidly generate and deliver high aerosol doses (Farkas, Hindle, & Longest, 2018a; Farkas, Hindle, & Longest, 2015; Young et al., 2013), but typically have a relatively large apparent aerosol diameter with high extrathoracic losses (Newman & Busse, 2002; Weers et al., 2019). The previous study of Farkas et al. (2020) demonstrated the generation and delivery of a small particle dry powder aerosol through a nasal interface and pediatric NT model at relatively high efficiency. As described further below, the positive-pressure air-jet DPI with a spray-dried powder formulation effectively generated a small aerosol size (approximately 1.7 μm) (Farkas et al., 2020). The device actuation speed was fast (<5 s) resulting in a high dose delivery rate (Farkas et al., 2020). Furthermore, an EEG particle formulation was used to reduce the potential for exhalation of the spray-dried aerosol and to enable targeted drug delivery (Tian, Hindle, & Longest, 2014; Tian et al., 2013).

1.2. Implementation of positive-pressure active devices

The vast majority of DPIs on the market are passive devices, which form an aerosol under negative pressure in response to a user’s inhalation through the device. In contrast, active devices use an energy source external to the user to form the aerosol. As reviewed by Longest et al. (2020), positive-pressure active devices implement an external gas source to aerosolize the powder, which can be supplied by an air-syringe, manual ventilation bag, or compressed air electromechanical system. Depending on the volume of gas used, these DPIs can be classified as high (≥200 ml) or low (<200 ml) actuation air-volume (AAV) devices. Active devices are often perceived as having the disadvantage of increased complexity and cost due to the requirement for an external gas source. However, a significant advantage of positive-pressure devices may be their ability to deliver dry powder aerosol during invasive and non-invasive mechanical ventilation (Farkas et al., 2018a, 2018c; Feng, Tang, Leung, Dhanani, & Chan, 2017; Okuda, Tang, Yu, Finlay, & Chan, 2017; Pornputtapitak, El-Gendy, Mermis, O’Brein-Ladner, & Berkland, 2014; Walenga, Longest, Kaviratna, & Hindle, 2017), and their ability to administer both the aerosol and a full inhalation breath, which can be beneficial in administering dry powder aerosol to infants (Howe, Hindle, Bonasera, Rani, & Longest, 2020; Laube, Sharpless, Shermer, Sullivan, & Powell, 2012) and young children (Farkas et al., 2020).

Considering dry powder aerosol delivery to pediatric subjects, positive-pressure DPIs that deliver the aerosol and a full inhalation breath can overcome a number of previously observed limitations. First, use of a consistent positive-pressure gas source to form and deliver the aerosol should significantly reduce inter and intra-subject variability in drug delivery, especially if extrathoracic depositional loss can also be reduced. Secondly, positive-pressure operation provides the option of oral or nasal lung delivery of the aerosol. Potential advantages of trans-nasal delivery include administering pharmaceutical aerosol to infants and children that are too young to use a mouthpiece (approximately 2–3 years old) and the ability to treat the nasal and lung airways simultaneously. Thirdly, positive-pressure gas delivery will expand rather than collapse the extrathoracic airways, which should improve lung delivery of the aerosol. Providing a known volume of gas delivery can be used to assist with deep lung inhalation and expansion of constricted or obstructed tracheobronchial airways, thereby enabling improved targeting of the deep lung regions and delivery to diseased airways. Finally, positive pressure aerosol delivery requires forming a sealed connection with the lungs via the extrathoracic region. This sealed system prevents the user from exhaling through the powder containment region, which can degrade powder performance, and can be used to encourage a brief breath-hold to improve lung retention of the aerosol.

Our group has recently developed a positive-pressure air-jet DPI concept for efficient aerosol generation and delivery to adults, children, and infants. As described in previous studies (Boc, Farkas, Longest, & Hindle, 2018; Farkas et al., 2018a; Farkas et al., 2018b; Farkas et al., 2018c), the air-jet DPI implements a small diameter inlet airflow passage, aerosolization chamber, and small diameter outlet aerosol flow passage. Positive-pressure gas passes through the inlet airflow passage and forms a high-speed turbulent jet within the aerosolization chamber (Longest & Farkas, 2019). Secondary flow velocities formed by the high-speed jet are used to initially fluidize the powder. As the fluidized powder enters the high speed jet region, additional powder deaggregation occurs (Longest & Farkas, 2019). The small diameter outlet orifice serves to both help form the secondary velocities and allow passage of sufficiently deaggregated particles out of the aerosolization chamber. Using this approach, AAVs of 10 ml and lower have been shown to effectively aerosolize 10 mg powder masses in devices that were designed to be integrated with a ventilation system, which required a small AAV so as to not increase the ventilation volume (Boc et al., 2018; Farkas et al., 2018a; Farkas et al., 2018b; Farkas et al., 2018c). For pediatric drug delivery, Farkas et al. (2019) developed a positive-pressure pediatric air-jet DPI that was operated with a ventilation bag or compressed gas supply with 750 ml of air, in order to aerosolize the powder and provide a full inhaled breath for a 5-year-old child. The AAV selected for 5-year-old children was based on adult inhalers typically being tested at 50–75% of total lung capacity (TLC). For a 5-year-old child, typical TLC is 1.55 L (ICRP, 1994), such that the 750 ml AAV is at the lower end of the 50–75% TLC range used for adults. Using a highly dispersible spray-dried formulation (Son, Longest, Tian, & Hindle, 2013), the best case pediatric air-jet DPI produced an aerosol MMAD <1.75 μm and a fine particle fraction (<5μm) ≥90% based on emitted dose. Actuation with the ventilation bag enabled lung delivery efficiency through the nasal and oral interfaces to a tracheal filter of 60% or greater, based on loaded dose. In both oral and nose-to-lung administrations, extrathoracic depositional losses were <10% (Farkas et al., 2019). Effective use of the positive-pressure device requires training for the caregiver to actuate the device correctly and for the patient to allow the device to inflate their lungs during use.

Computational fluid dynamics (CFD) studies of aerosolization within the air-jet DPI have revealed some interesting characteristics. At both high and low AAVs, increasing turbulence increases emitted dose (which is advantageous), but also increases MMAD (which is typically detrimental for efficient lung delivery) (Bass, Boc, et al., 2019; Longest & Farkas, 2019; Longest, Farkas, Bass, & Hindle, 2019). The direct relationship between internal device turbulence and MMAD is a unique characteristic of the air-jet system as most other aerosol generation units are assumed to have the opposite behavior. This behavior was attributed to a two stage aerosolization process of initial fluidization of the powder followed by turbulent deaggregation of fluidized agglomerates (Longest & Farkas, 2019). Excess turbulence was viewed to fluidize the powder too rapidly leaving less time for secondary turbulent deaggregation. Provided that sufficient emitted dose can be maintained, the air-jet DPI therefore performs better with lower flows and less turbulence (Longest & Farkas, 2019; Longest, Farkas, et al., 2019), which are ideal characteristics for efficient aerosol administration to infants and children (Bass, Farkas, & Longest, 2019). Furthermore, devices tend to produce a direct linear relationship between emitted dose and MMAD, i.e., higher emitted dose is directly proportional to higher MMAD (Bass, Farkas, et al., 2019). CFD and in vitro aerosol characterization can be used to identify and select device designs with beneficial emitted dose and MMAD relationships (Bass, Farkas, et al., 2019; Howe et al., 2020).

1.3. Patient interfaces that reduce turbulence and jet momentum effects

DPIs typically employ high turbulence and small diameter flow passages leading to the mouth-throat region in order to deaggregate dry powder formulations and form an inhalable aerosol (Coates, Chan, Fletcher, & Chiou, 2007; Coates, Chan, Fletcher, & Raper, 2005, 2006; Coates, Fletcher, Chan, & Raper, 2004; DeHaan & Finlay, 2001; Fenton, Keating, & Plosker, 2003; Shur et al., 2012). While some inhalable dose fraction can be formed with this method, depositional losses in the device and mouth-throat (MT) region are typically high (DeHaan & Finlay, 2001; Ilie, Matida, & Finlay, 2008; Longest, Tian, Delvadia, et al., 2012; Longest, Tian, Walenga, et al., 2012; Matida et al., 2003; Tian, Longest, Su, Walenga, et al., 2011), due to increased impaction deposition and turbulence dispersion. Within the DPI, different structures and airflow passage designs are implemented to generate the turbulence and particle aggregate break-up mechanisms that are needed to deaggregate the powder (Coates et al., 2007; Coates et al., 2006; Coates et al., 2004; Shur et al., 2012; Wong et al., 2010, 2011a, 2011b). Longest et al. (2013) implemented a combination of in vitro analysis and CFD simulations to evaluate eight different aerosolization units including a standard constricted tube, impaction surface, 2D mesh, inward radial jets, and newly proposed 3D rod arrays. It was determined that for a set amount of input energy (applied as a negative pressure drop across the device) a 3D rod array with unidirectional rods was most effective at aerosolizing the powder.

While the air-jet DPI improved MMAD with lower internal turbulence, a potential disadvantage is the small diameter jet of high velocity aerosol exiting the device, which can lead to unnecessary impaction loss in the patient interface and extrathoracic airways. Farkas et al. (2020) observed relatively low MT deposition with an adult air-jet DPI (<10% of emitted dose), but unexpected powder deposition on the back of the throat as a result of the high turbulence jet. The CFD study of Bass et al. (2020) explored pediatric patient interfaces that could reduce the effect of the high-intensity turbulent jet that exits the air-jet DPI. Internal structures within the interface that were considered included non-smooth surfaces, rapid and stepped expansions, impaction surfaces and various 3D rod array designs. CFD results revealed that a combination of a 3D rod array with a rapidly expanding interface in the region of the rod array best dissipated the turbulent jet while minimizing depositional loss in the mouthpiece (Bass & Longest, 2020). For oral aerosol administration, the optimal flow passage compared with previous design candidates reduced device, mouthpiece, and mouth-throat deposition efficiencies by factors of 8-, 3-, and 2-fold, respectively (Bass & Longest, 2020). For nose-to-lung aerosol administration, the optimal flow pathway compared with previous designs reduced device, nasal cannula, and nose-throat deposition by 16-, 6-, and 1.3-fold, respectively (Bass & Longest, 2020).

Following the CFD study of Bass et al. (2020), Farkas et al. (2020) considered pediatric oral aerosol delivery with a realistic in vitro MT airway model using an air-jet DPI and MP interface, which included a 3D rod array to improve secondary break-up of the aerosol and dissipate the turbulent jet before entering the MT region. A new vertical aerosolization chamber was also considered that was expected to be less sensitive to larger powder mass loadings. Devices were loaded with 10 mg doses of a spray dried formulation and actuated with positive pressure using a flow rate of 10–20 L/min and an air volume of 750 ml consistent with a 5-year-old child. Inclusion of the 3D rod array in the MP was shown to further reduce the aerosol size to an MMAD of <1.7 μm without significantly increasing aerosol loss in the device. Best case device and MP combinations produced <2% MT depositional loss and >70% lung delivery efficiency (based on loaded dose) in a realistic in vitro pediatric MT geometry. Aerosol delivery with a 3D rod array and trans-nasal delivery, which is the focus of the current study, has not been previously considered experimentally.

1.4. Highly dispersible spray-dried formulations that change size within the airways

Considering the small airway diameters and relatively low inhalation volumes involved with pediatric aerosol delivery, highly dispersible powder formulations are important. Spray drying techniques are one of the most flexible and practical methods to generate large quantities of highly dispersible powder formulations (Chan, 2006; Vehring, 2008; Vehring et al., 2007; Weers & Miller, 2015). These spray-dried formulations typically require the application of particle formulation engineering techniques to make them highly dispersible, including the formation of large porous particles (Edwards, Ben-Jebria, & Langer, 1998; Edwards et al., 1997), PulmoSpheres ™ (Geller, Weers, & Heuerding, 2011; Weers & Miller, 2015), or inclusion of dispersion enhancers (Feng et al., 2011; Son et al., 2013). Using these approaches, previous studies have reported the formation of small-particle dry powder aerosols as needed for efficient pediatric drug delivery (Edwards et al., 1998; Edwards et al., 1997; Son et al., 2013; Weers et al., 2015). Nevertheless, the use of a static particle size for efficient lung delivery of an aerosol has limitations. For example, particles that are small enough to avoid extrathoracic deposition may also lack sufficient inertia for targeting deposition in the tracheobronchial airways, or may be exhaled. Increasing particle size results in higher extrathoracic depositional loss and a net reduction in lung delivery. As demonstrated by Walenga et al. (2016), conventional MDI and DPI devices deposit approximately 1% of the inhaled dose within the entire small tracheobronchial region of adult lungs.

To address the limitations associated with the delivery of static aerosol sizes, the concept of controlled condensational growth has recently been developed (Hindle & Longest, 2010, 2012; Longest & Hindle, 2010, 2011). In this approach, an aerosol is delivered to the respiratory tract with a sufficiently small size to minimize device and upper airway deposition. Droplet size increase through condensational growth allows for retention of the aerosol, which without growth would likely be exhaled. Techniques to produce the required size increase include enhanced condensational growth (ECG) and excipient enhanced growth (EEG). In the ECG approach, the aerosol is delivered with air saturated with water vapor a few degrees above body temperature, which creates supersaturated relative humidity conditions in the lungs to foster condensational growth of the droplets (Hindle & Longest, 2010; Tian, Longest, Su, & Hindle, 2011). With EEG, formulated particles contain a combination of a drug and a hygroscopic excipient, and the natural relative humidity in the lungs provides the water vapor source for aerosol size increase (Hindle & Longest, 2012; Longest & Hindle, 2011). Combination particles that contain both the therapeutic agent and a hygroscopic excipient, in order to generate aerosol size increase in the airways, are referred to as EEG formulations (Hindle & Longest, 2012; Son et al., 2013).

The new approach of controlled condensational growth has been successful at improving lung delivery for orally-administered aerosols (Hindle & Longest, 2010; Son et al., 2013; Tian et al., 2013; Tian, Longest, Su, & Hindle, 2011) and for nose-to-lung delivery with nebulizer generated aerosols (Golshahi, Tian, et al., 2013; Longest, Tian, & Hindle, 2011) based on CFD simulations and realistic in vitro experiments. Considering nose-to-lung delivery with a nebulizer, Golshahi et al. (2013) previously demonstrated that both EEG and ECG approaches reduced cannula and nasal depositional losses by an order of magnitude and delivered approximately 80% of the loaded dose to the lungs with steady flow. Using an aerosol mixer-heater system (Longest, Walenga, Son, & Hindle, 2013) combined with a mesh nebulizer, Golshahi et al. (2014) demonstrated that synchronizing the aerosol delivery with patient breathing was important to achieve high efficiency aerosol delivery (>70%) past the nose and to the lung in an adult high flow nasal cannula (HFNC) system. Tian et al. (2014) reported CFD simulations of nose-to-lung administered EEG aerosols through the conducting tracheobronchial (TB) airways and found minimal nasal depositional loss, substantial droplet growth, and a significant dose enhancement to the lower TB region of 40-fold compared with marketed inhalers.

1.5. Concurrent CFD and realistic in vitro analysis

As reviewed or highlighted in multiple publications, both CFD and realistic in vitro experiments can be used to better understand and significantly improve multiple aspects of respiratory drug delivery (Byron et al., 2010; Carrigy, Ruzycki, Golshahi, & Finlay, 2014; Delvadia, Hindle, Longest, & Byron, 2013; Delvadia, Longest, & Byron, 2012; Delvadia, Longest, Hindle, & Byron, 2013; Delvadia, Wei, Longest, Venitz, & Byron, 1016; Longest, Bass, et al., 2019; Longest & Holbrook, 2012; Ruzycki, Javaheri, & Finlay, 2013; Wei et al., 2018; Wong, Fletcher, Traini, Chan, & Young, 2012). Considering CFD, useful insights and design optimizations can be made in the areas of particle engineering (Longest, Farkas, Hassan, & Hindle, 2020), aerosol dispersion (Bass, Farkas, et al., 2019; Longest & Farkas, 2019; Longest, Farkas, et al., 2019; Wong et al., 2011b; Wong et al., 2012), device and interface design (Coates et al., 2007; Coates, Chan, et al., 2005; Coates et al., 2006; Coates et al., 2004; Coates, Fletcher, Chan, & Raper, 2005; Hindle & Longest, 2013; Longest & Hindle, 2009b; Shur et al., 2012) and airway transport and deposition (Lambert, O’Shaughnessy, Tawhai, Hoffman, & Lin, 2011; Longest, Tian, Delvadia, et al., 2012; Longest, Tian, Walenga, et al., 2012; Tian et al., 2014; Tian et al., 2013; Tian, Longest, Su, Walenga, et al., 2011). Similarly, realistic in vitro experiments can be conducted to capture extrathoracic depositional loss, aerosol size distribution entering the lungs, transport through the upper tracheobronchial airways, and transport within lower airway segments including the alveolar region (Berg & Robinson, 2011; Berg, Weisman, Oldham, & Robinson, 2010; Byron et al., 2010; Delvadia, Hindle, et al., 2013; Delvadia et al., 2012; Delvadia, Longest, et al., 2013; Olsson, Borgstrom, Lundback, & Svensson, 2013; Ruzycki et al., 2014). In the area of pediatric aerosol delivery, Carrigy et al. (2014) reviewed both CFD and realistic in vitro experiments.

As highlighted by Longest et al. (2019), our group frequently employs the simultaneous or concurrent application of both CFD analysis and realistic in vitro experiments, which is illustrated in Figure 1. In this approach, both techniques are applied simultaneously in order to take advantage of each method’s strengths and to minimize each method’s weaknesses. Briefly, realistic in vitro experiments are used to assess initial and key prototype performance of both aerosol formation and lung delivery efficiency as a benchmark. CFD models are developed of key system aspects, such as mouthpiece aerosol transport and extrathoracic depositional loss. The CFD models are first validated with the initial in vitro data. Once validated, the CFD models are then used to generate valuable insights into system transport characteristics and explore design alternatives. Best performing model designs or strategies are then produced and tested experimentally to verify system performance improvements. In some instances, quantitative relations are formed between CFD-predicted parameters and experimental critical quality attributes, such as device emitted dose and MMAD (Bass, Farkas, et al., 2019; Bass & Longest, 2020; Hindle & Longest, 2013; Longest & Hindle, 2009b; Longest, Son, et al., 2013; Longest & Farkas, 2019; Longest, Farkas, et al., 2019). For example, in the quantitative concurrent analysis of Longest et al. (2019), aerosol dispersion parameters were used to capture both emitted dose and aerosol size for a series of air-jet DPIs. These or similar dispersion parameters where then used to optimize DPI performance for low AAV (Longest, Farkas, et al., 2019) and high AAV (Bass, Farkas, et al., 2019) applications based on CFD analysis. Optimized devices were then prototyped and tested experimentally in each respective study. This concurrent analysis approach has been applied to the development of dry powder formulations for high dispersion (Longest, Farkas, Hassan, et al., 2020), powder aerosolization within DPIs (Bass, Farkas, et al., 2019; Longest & Farkas, 2019; Longest, Farkas, et al., 2019), aerosol transmission through inhalers (Hindle & Longest, 2013; Longest & Hindle, 2009b), and aerosol delivery strategies (Longest & Hindle, 2012b; Longest, Tian, Li, et al., 2012; Tian et al., 2014; Tian et al., 2013; Tian, Longest, Su, & Hindle, 2011).

Figure 1:

Schematic of concurrent CFD and realistic in vitro analysis that has been implemented to increase the lung delivery efficiency of pharmaceutical aerosols based on improvements in the areas of powder dispersion, device performance, and delivery strategy development.

1.6. Case study: Nose-to-lung dry powder aerosol administration to children with cystic fibrosis

In order to illustrate the combined use of new technologies for high efficiency aerosol administration described above and concurrent CFD and realistic in vitro analysis, a case study is considered based on nose-to-lung dry powder aerosol administration to children with cystic fibrosis (CF). Specifically, the objective of this case study is to use multiple new techniques simultaneously in order to overcome the primary limitations associated with poor dry powder aerosol administration to children and enable high efficiency trans-nasal DPI use in this population, based on concurrent CFD and realistic in vitro analysis. Techniques used to improve lung delivery efficiency of the dry powder aerosol include nose-to-lung administration in subjects as young as 2-years-old, use of a positive-pressure active DPI, implementation of patient interfaces that improve aerosol deaggregation and dissipate the flow field, and controlled condensational growth of the aerosol within the airways. Dry powder aerosol administration is assessed through nose-throat (NT) and upper tracheobronchial (TB) models of children in the age ranges of 2–3, 5–6, and 9–10 years-old. The realistic in vitro upper TB airway models are based on airway scans of pediatric subjects that have CF with moderate lung damage. As previously implemented (Longest et al., 2015; Longest, Tian, Li, et al., 2012), the upper TB airway models are enclosed in an aerosol growth chamber that is sized to represent aerosol residence time in the lungs and connect to a Next Generation Impactor (NGI) for aerosol sizing in the experiments.

The delivery system evaluated includes a newly developed positive-pressure air-jet DPI, nose-to-lung patient interface with a 3D rod array used to dissipate the turbulent jet leaving the aerosol generation unit, and an excipient enhanced growth (EEG) powder formulation. Potential pediatric-CF drug delivery applications include the administration of inhaled antibiotics, anti-inflammatories, mucus clearance agents, medications that address airway constriction, and potential corrective gene therapies. Consistent with the concurrent analysis approach, in vitro experiments are implemented to benchmark system performance in terms of device emitted dose, system delivery efficiency, extrathoracic and upper TB depositional loss and aerosol size increase through the growth chamber for the 5–6 year-old airway model. Once validated, the CFD model is used to illustrate regional aerosol deposition, aerosol size increase, and lung delivery efficiency across multiple pediatric age ranges. In contrast with the expected 5–10% lung delivery efficiency for pediatric subjects with commercial systems (Below et al., 2013; Devadason et al., 1997; Devadason, 2006; Lindert et al., 2014), the concurrent CFD and realistic in vitro analysis shows that the implementation of multiple new technologies increases the lung delivery efficiency to >70% of the loaded dose, with little effect of pediatric age and the potential for targeted deep lung delivery through aerosol size increase.

2. Case Study: Materials and Methods

To demonstrate the advantages of utilizing a rod array to reduce inlet jet intensity into the patient interface and NT region, the current study compares aerosolization performance in the best-case nasal cannula from Farkas et al. (2020), both with and without a rod array, by using concurrent in vitro testing and CFD analysis. The numerical and experimental methods used in the evaluation of these nasal cannulas is consistent with our previous studies (Bass, Farkas, et al., 2019; Bass & Longest, 2020; Farkas et al., 2020; Farkas et al., 2019) and are omitted in the current case study in the interest of brevity. 3D CAD model renderings of the DPI and nasal cannula combination are shown in Figure 2 and schematics of the device with labels of key components are shown in Figure 3. Note that the chosen delivery system employs the four key technologies described in the Introduction, and this study aims to demontrate how these techonologies maximize available lung dose in pediatric patients with concurrent experimental and numerical evaluation.

Figure 2:

CAD model rendering of combined dry powder inhaler and nasal cannula interface with rod array showing (a) translucent view with inner components and (b) expanded section view that illustrates the inner flow pathway and 3D rod array.

Figure 3:

Overview of flow pathways for the best-case nose-to-lung delivery system from Farkas et al. (2020) showing (a) the air-jet dry powder inhaler with inlet and outlet flow passages and powder aerosolization chamber and (b) the 3D rod array nasal cannula.

To evaluate upper airway losses and aerosol growth with the chosen delivery system, three growth chamber models were developed to evaluate aerosol transport and growth in a 2–3-, 5–6-, and 9–10-year-old patient. An in vitro growth chamber model for a 5–6-year-old patient (see Figure 4) was prototyped and tested with the best-case DPI and rod-array nasal cannula from Farkas et al. (2020) (see Figure 3) to provide experimental data for CFD model validation. Thereafter, numerical models for all three age groups were developed to evaluate particle deposition and growth in the representative geometries. The airway geometries used in the in vitro and numerical models consists of an upper airway (NT to B3) extracted from CT scans and a growth chamber that was designed to provide an aerosol residence time of approximately two seconds throughout the entire model (see Figure 4). Characteristic airway dimensions for the three upper airway models are provided in Table 1. Characteristic length scales in the NT region are consistent with nasal deposition studies by Storey-Bishoff et al. (2008) and Golshahi et al. (2011), and the length of the trachea was measured directly from CT scans. Measurements beyond the trachea are not reported as the diseased state of the CF airways may not provide a clear comparison between age groups. As expected, characteristic dimensions for older patients are generally larger than the younger patients, which will likely influence aerosol deposition in the upper airways.

Figure 4:

Schematic of the experimental model showing the device (air-jet DPI), patient interface (Nasal Cannula), in vitro model (Upper Airway), and two-second residence time chamber (Growth Chamber). Air flow in and out of the model is labelled (1) 10 LPM actuation air into the device, (2) 35 LPM make-up air through two one-way valves, and (3) 45 LPM to the NGI.

Table 1:

Characteristic dimensions for the 2–3-, 5–6-, and 9–10-year old upper airway models.

Dimension	2–3-year-old	5–6-year-old	9–10-year-old
V [mm3]	18,411	24,186	41,323
As [mm2]	13,742	16,202	23,060
V/As [mm]	1.34	1.49	1.79
LCP [mm]	124.4	126.4	128.5
$\sqrt{V/L_{CP}}\left[\text{mm}\right]$	12.2	13.8	17.9
Dh,G [mm]	5.7	6.7	6.8
LT [mm]	64.6	75.3	97.7

V: NT-B3 volume

A_s: NT-B3 surface area

L_cp: Central path length from nostrils to glottis

D_h,G: Hydraulic diameter of the glottis

L_t: Length of trachea from glottis to carinal ridge

2.1. Upper Airway and Growth Chamber Geometries

The NT region of the upper airway models were developed by segmentation of CT scans with the Mimics software suite (Materialise, Leuven, Belgium) from patients with healthy airways, as no nasal scans were available for patients diagnosed with CF. This approach was deemed acceptable as the lung disease attributed to CF and Pseudomonas aeruginosa infections predominantly affects the lower airways (Tiddens, Donaldson, Rosenfeld, & Pare, 2010). Scans for the NT region of the 2–3-, 5–6-, and 9–10-year-old upper airway models were selected from our database of medically necessary CT scans, which were reviewed for quality and completeness under an active Institutional Review Board protocol. When selecting scans for the study, preference was given to scans that had a slice resolution of 1.0 mm or less and of patients that had heights consistent with average values for the chosen age ranges. For the 2–3-, 5–6-, and 9–10-year-old models, the patient heights were 90 cm, 114 cm, and 144 cm, respectively, which are within the 20^th and 80^th percentiles based on the WHO growth charts (WHO, 2006). The segmentation and CAD model development generally followed the same method as outlined by Bass et al. (2019), with the exception that the latest automated skin surfacing capabilities in SpaceClaim v19.3 were utilized, which reduced the manual effort required to convert STL surfaces to CAD data.

The tracheobronchial (TB) regions of the upper airway model (trachea to B3) were also developed by segmentation of CT scans with Mimics, but the chosen scans were from patients with moderate CF lung damage as scored by the PRAGMA-CF system (Rosenow et al., 2015). The CF CT scans for all three age groups were provided and evaluated for PRAGMA-CF score by Erasmus University Medical Center. For the 2–3-, 5–6-, and 9–10-year-old models, the patient heights were 100 cm, 122 cm, and 131 cm respectively, which are also within the 20^th and 80^th percentiles based on the WHO growth charts (WHO, 2006) and consistent with the NT scans. After segmentation and CAD model development of the diseased TB regions, the geometry was coupled to the NT region of the healthy patient to form the upper airway (NT-B3). The two regions were coupled along the trachea at a point where there was less than a 1.0 mm deviation between the cross-section profiles of the NT and TB regions and preserved the expected tracheal length. A 10 mm coupling region along the length of the trachea (5 mm either side of the coupling point) was then created to define the transition between the NT and TB regions. Due to the selection of a coupling point where there is minimal difference between two regions and a relatively short coupling section, the effect of artificially joining the healthy NT and diseased TB regions was assumed to be negligible. This assumption will be tested by identifying any abnormal deposition patterns in the coupling region from the CFD results.

The cylindrical growth chamber was designed with insight from preliminary CFD work to provide minimal aerosol loss (<5% deposition) and a residence time of approximately two seconds, which was sufficient to reach maximum aerosol growth. The resulting chamber size for all age groups was a diameter of 101.6 mm (4”) and length of 155.0 mm. For the in vitro model, the chamber was constructed from a section of clear cast acrylic pipe, sourced from ePlastics (San Diego, CA), and a 3D-printed lid, base, and outlet duct.

2.2. Experimental Materials and Powder Formulation

Albuterol sulfate (AS) USP was purchased from Spectrum Chemicals (Gardena, CA) and Pearlitol® PF-Mannitol was donated from Roquette Pharma (Lestrem, France). Poloxamer 188 (Leutrol F68) was donated from BASF Corporation (Florham Park, NJ). L-leucine and all other reagents were purchased from Sigma Chemical Co. (St. Louis, MO).

Multiple batches of a spray-dried AS EEG powder formulation were produced based on the optimized method described by Son et al. (2013) using a Büchi Nano spray dryer B-90 HP (Büchi Laboratory-Techniques, Flawil, Switzerland). The AS EEG powder formulation contained a 30:48:20:2% w/w ratio of AS, mannitol, l-leucine, and Poloxamer 188. The AS EEG powder was used as a model test spray dried formulation in place of antibiotic EEG formulations (e.g tobramycin). It is expected that antibiotic EEG powder formulations with the same hygroscopic properties as the AS EEG formulation will perform comparably in regard to targeted lung delivery.

2.3. Realistic In Vitro Experimental Methods

The delivery system chosen for experimental testing was the best-case DPI and rod-array nasal cannula combination from Farkas et al. (2020). Details of the pediatric air-jet DPI (see Figure 3a) and patient interfaces (see Figure 3b) have been discussed extensively in our previous publications related to the development of this delivery system (Bass, Farkas, et al., 2019; Bass & Longest, 2020; Farkas et al., 2020; Farkas et al., 2019). Briefly, the chosen air-jet DPI operates at an approximate flow rate of 10 LPM (from a positive pressure gas source) and actuation time of 4.5 s (for an inhalation volume of 750 mL for a 5–6-year-old) with inlet and outlet airflow passage (i.e. capillary) diameters of 1.40 mm and 2.39 mm, respectively. The design parameters and operating conditions for the air-jet DPI have previously been shown to give the best aerosolization performance and high-efficiency nose-to-lung aerosol delivery (Farkas et al., 2020). The chosen nasal cannula utilizes a 3D rod array to attenuate the high-velocity, highly turbulent air jet that leaves the DPI outlet orifice, which has previously been shown to reduce interface and NT losses (Bass & Longest, 2020).

The in vitro tests of the 5–6-year-old model for upper airway losses and aerosol growth were conducted under room (22.0°C and 33.5% RH) and humid airway (37.0°C and 99.0% RH) conditions, with the humid airway conditions controlled in an environment cabinet. In addition, for the humid airway conditions, the model walls were pre-wetted by running heated, humid air (40.0°C and 99.0% RH) through the model before actuating the device in the humid airway experiments. Cooling of the 40°C saturated air in contact with the interior model walls resulted in water vapor condensing onto the walls forming a thin liquid layer. All experimental model components were developed by utilizing the CAD modelling capabilities in SolidWorks 2019 (Dassault Systèmes, Paris, France) and most parts were then 3D printed. The exceptions to prototyped parts were the acrylic growth chamber cylinder, as mentioned previously, and the face of the NT region, which was made from cast silicone to create a soft nose. The soft nose was required to accommodate the existing cannula design (which was developed to fit a different NT model) and to provide an air-tight seal around the cannula prongs. The silicone material from which the soft nose is made is also more consistent with the flexible soft tissue of the human nose than rigid 3D printer plastic. Cast molds were developed in SolidWorks 2019 such that the void represents the anterior region of the NT model. The cast was then filled with Dragon Skin™ 20 silicone (Smooth-On, Lower Macungie, PA), left to cure following the product instructions, removed from the cast, and then fixed into the NT model with epoxy. As an additional precaution against air leaks in the model, all connections between model components included O-rings to ensure air-tight sealing.

The device actuation and experimental testing in the present study generally follows the methods presented by our previous in vitro development and performance evaluation of the pediatric air-jet DPI and patient interfaces (Bass, Farkas, et al., 2019; Farkas et al., 2020). Briefly, the DPI aerosolization chamber is loaded with 10 mg of AS EEG powder and actuated with a 6 kPa positive-pressure air source, using a compressed air line and solenoid valve device, which efficiently aerosolizes the powder. Previous work has shown negligible differences in aerosolization performance when using compressed air or ventilation bag gas sources to actuate the device (Farkas et al., 2019). Characterization of the aerosol that leaves the growth chamber was performed using a Next-Generation Impactor (NGI) and AS drug masses were assayed with high-performance liquid chromatography (HPLC). Preliminary experimental tests of the growth chamber model showed that device actuation time was not sufficient to clear all the aerosol from the growth chamber to the NGI, which lead to low recovered doses as powder remained dispersed in the cylinder and was not assayed. Therefore, a second solenoid valve was utilized that switched between pulling air from the chamber to ambient air, which was controlled by a separate timer from the device actuation and ran for 8 s longer than device actuation. In summary, the testing process was as follows: (i) switch the DPI solenoid valve to deliver compressed air to the DPI at 10 LPM (see Label 1 in Figure 4) and at the same time switch the NGI solenoid valve to pull air through the growth chamber at 45 LPM (see Label 2 in Figure 4) allowing make-up air at 35 LPM to be drawn through the top of the growth chamber via one-way valves (see Label 3 in Figure 4); (ii) after 4.5 s, the DPI solenoid valve closes and device actuation stops while the NGI solenoid valve stays open for a total of 12 s, drawing all the aerosol from the chamber and delivering it to the NGI. Using this approach, all recovered doses from experimental runs were greater than 90% (average of 96.5%). The device emitted dose (ED) was defined as the difference between the loaded AS dose and the mass of AS retained in the DPI after actuation, divided by the loaded dose, and expressed as a percentage. The delivery system ED was defined with a similar method, with the mass of AS retained in the DPI and nasal cannula divided by loaded dose. The aerosol MMAD was identified with linear interpolation of a cumulative percentage drug mass vs. cut-off diameter plot from the NGI. The cut-off diameters of each NGI stage were calculated using the formula specified in USP 35 (Chapter 601, Apparatus 5) for the operating flow rate of 45 LPM. t-test were used with JMP-Pro® 12 (SAS Institute, Cary, NC) for statistical analysis. The p-value < 0.05 was considered as significant.

2.4. CFD Models

In the interest of brevity, complete details of the CFD models used to evaluate losses in the nasal cannula, both with and without a rod array, are omitted from the current case study, and the reader is referred to our previous publication on this subject (Bass & Longest, 2020). In summary, the nasal cannula CFD models presented in this study follow the Roache method for mesh independence (Roache, 1994) and adhere to our previously established and validated meshing and solution guidelines for respiratory drug delivery (Bass, Boc, et al., 2019; Bass & Longest, 2018). These guidelines include corrections to the flow and turbulence field in the near-wall region to account for turbulence anisotropy and particle-wall hydrodynamic damping (Longest, Hindle, Das Choudhuri, & Byron, 2007). The only difference from the current CFD methods compared with our previous work (Bass & Longest, 2020) is a change to the modeling of particle-wall interaction with the Teflon coated stainless steel 3D rod array in the patient interface. Previous studies assumed that all particles bounce off the rods, as deposition on the smooth Teflon coated rods is expected to be low. However, in the current study it is assumed that particles bounce off the front half of the cylindrical rods and stick to the rear half of the rods, if they should come into contact with these rear-facing surfaces. This assumption is justified by visual inspection and validation of the CFD-predicted deposition compared with experimental data in the Results section. The remaining CFD methods presented in the current study focus on development of the growth chamber models that were used to evaluate upper airway loss and targeted drug delivery capabilities with the chosen pediatric delivery system.

2.4.1. Computational Domain and Spatial Discretization

The computational domain for the 5–6-year-old upper airway and growth chamber replicates the experimental model as closely as possible, with numerical extensions added as needed to ensure fully-developed flow enters the model inlets. As the CFD models for the delivery system were evaluated separately, the inlet to the computational domain for the growth chamber model begins 5 mm downstream of the nostrils, which is consistent with insertion of the cannula prongs into the soft nose of the in vitro model. The inlet boundary is the elliptical cross-section of the cannula prongs, with the surface between the nostrils and prongs defined as walls to represent the air-tight seal between the cannula and soft nose. All other inlet, outlet, and fluid-wall boundaries are consistent between the computational domain and in vitro growth chamber model. The geometry was generated by combining the flow path in the upper airway with the internal volume of the growth chamber, and the solid parts of the TB region were subtracted from the air-side volume to include the thickness of the model components in the chamber region. For CFD evaluation of all three age groups, a similar approach was used to generate the 2–3- and 9–10-year-old models, using the same size growth chambers, with all geometries created in SpaceClaim v19.3. Figure 5a shows the full geometry, including the upper airway and growth chamber, for the 5–6-year-old CFD model, with Figure 5b–d showing only the upper airway geometry for the 2–3-, 5–6-, and 9–10-year-old CFD models, respectively. The cannula prong inlets on all three models were dimensioned to have similar areas, and hence provide similar inlet velocities given a flow rate of 5 LPM, but the major and minor radii of the elliptical prongs differed slightly to accommodate the different nostril shapes. For the 2–3- and 9–10-year old models, the major and minor radii were 2.5 and 4.5 mm respectively, which gave an average inlet velocity of approximately 2.36 m/s. For the 5–6-year-old model, the major and minor radii were 2.25 and 5.25 mm respectively, which gave an average inlet velocity of 2.25 m/s. As there is only a 0.1 m/s absolute difference in inlet velocity between the models (5.0% relative difference) the effect of changing cannula prongs on NT deposition is expected to be negligible and is a necessary modification to accommodate differing nostril dimensions. Also note that the angular rotation of prongs was different among all three models to fit the specific characteristics of each patient. The effect of these alterations to the cannula design on patient interface deposition is not known at this time.

Figure 5:

Summary of computational domains showing (a) the full CFD model of the nose-throat, tracheobronchial region, and growth chamber for the 5–6-year-old child, (b) the nose-throat to B3 (NT-B3) region for the 2–3-year-old, (c) 5–6-year-old, and (d) 9–10-year-old child.

Spatial discretization of the computational domain was performed by utilizing the meshing capabilities available in FLUENT v19.3 (ANSYS Inc., Canonsburg, PA). An unstructured meshing approach, as opposed to structured hexahedral cells, was used to accurately resolve the complexity of the nasal passage surfaces and the transition from the upper airway to growth chamber regions. Instead of traditional tetrahedral-type cells, polyhedral cells were used to discretize the computational domain, which our group has previously shown to be more computationally efficient and equally as accurate for modeling particle transport through the upper airways (Bass, Boc, et al., 2019). Our previous model development work has also identified that the near-wall (NW) mesh resolution is critical to obtaining a successful experimental validation of the CFD model (Bass & Longest, 2018). The growth chamber CFD models in the present study follow our current best practices for polyhedral meshes of the respiratory airways (Bass, Boc, et al., 2019; Bass & Longest, 2018) with a wall y+ value of approximately one, five prismatic NW cell layers, and a layer-to-layer growth ratio of 1.2. As a final step, cell nodes were smoothed in FLUENT v19.3 until the orthogonal quality metric was greater than 0.15, which ensures a high-quality spatial discretization of the computational domain.

Mesh independence of the 5–6-year-old growth chamber CFD model was established using the Roache method for mesh refinement studies (Roache, 1994) by comparing the volume-average velocity magnitude (v_mag) and turbulent kinetic energy (k) between three meshes with increasingly higher degrees of spatial resolution. The three meshes had total cell counts of 1.17, 1.98, and 4.06 million cells, which resulted in normalized grid spacing ratios of 1.51, 1.27, and 1.00 respectively. Comparing the 1.98 and 4.06 million cell meshes, the grid convergence index was less than 1% for both volume-average v_mag and k.Using Richardson Extrapolation (Richardson & Gaunt, 1927) to estimate the exact solution of the volume-average field quantities, and comparing these values to the 1.98 million cell case, the relative error in v_mag and k was 0.07% and 0.45%, respectively. Therefore, the 1.98 million cell case was chosen for evaluation of the 5–6-year-old model, and its degree of spatial discretization was subsequently applied to the 2–3- and 9–10-year old models. Furthermore, the 1.98 million cell count is consistent with the previous growth chamber study by Longest et al. (2012) which used a total of 1.49 million cells.

2.4.2. Numerical Models and Solver Settings

The Reynolds number at each nostril inlet, given a flow rate of 5 LPM and hydraulic diameter of 6.1 mm (from the elliptical inlet boundary), was approximately 1,000, which suggests laminar flow conditions. However, changes in cross-section throughout the entire upper airways are known to change local flow conditions and induce turbulence in the respiratory airways (Bass, Boc, et al., 2019). As such, the low-Reynolds number (LRN) k-ω turbulence model was implemented to model the transitional-to-turbulent flow regime. The LRN k-ω model includes an eddy viscosity damping coefficient that allows it to provide an accurate representation of the flow field in regions of both low and high levels of turbulence. As this study evaluated CFD predictions of water absorption by the EEG aerosol in the growth chamber, both water vapor and air are modeled as components of the continuous phase in the computational domain to account for varying degrees of relative humidity, which required the addition of multi-species and energy transport equations to the standard mass, turbulence, and continuity equations. As the gas phase is composed of both air and water vapor, and temperature is not a constant variable, the incompressible ideal gas law was used to model the fluid density.

Preliminary work on model development showed a jet formed as the flow passed through the constriction of the glottis in the larynx (laryngeal jet), and this jet exhibits transient behavior by oscillating back and forth through the trachea. The laryngeal jet phenomenon is consistent with previous observations by Xi et al. (2008) and has a significant effect on aerosol transport and deposition in the TB region. To capture this phenomenon accurately and achieve a converged solution, a transient formulation of the flow and turbulence transport equations was implemented in the CFD models. A time step of 0.005 s was sufficient to reach convergence within at most 100 iterations per time step and accurately modelled the oscillatory behavior of the laryngeal jet. A negligible difference in flow field and particle deposition results was also observed when decreasing the time step to 0.002 and 0.001 s, so using a time step of 0.005 s was deemed sufficient for all CFD models. The flow field was initialized with quiescent conditions and a water vapor mass fraction that was consistent with either room or humid airway conditions. Monitoring plots of volume-average v_mag and k against flow time showed that the transient start-up plateaued after approximately 0.5 s, so a total simulation time of 1.0 s was used for all CFD models. As an alternative and more computationally efficient method of tracking particles simultaneously with the flow field solution, particle trajectories were calculated at a single representative time step, as opposed to tracking particles throughout the full simulation duration. To ensure particle deposition and residence time was not vastly different between time steps, trajectories were calculated in the 5–6-year-old model every 0.25 s, with the maximum absolute deviation from the mean deposition fraction (DF) being 0.4% and the maximum absolute deviation from the mean residence time being 0.3 s.

FLUENT v19.3 (ANSYS Inc., Canonsburg, PA) was used to obtain solutions for all flow and turbulence equations, and all model settings for particle transport in the respiratory airways followed our previously defined best practices (Bass, Boc, et al., 2019; Bass & Longest, 2018). The spatial discretization of the flow and turbulence transport equations were second-order accurate, the gradient discretization used the Green-Gauss Node-based method, and the SIMPLEC pressure-velocity coupling scheme was used. Further details on the mass, momentum, and turbulence transport equations are available in our previous publications (Longest et al., 2007; Longest, Vinchurkar, & Martonen, 2006). All inlet boundaries used mass flow inlet conditions, with the nostril inlets (see Label 1 in Figure 4) each assigned a mass flow that gave a volumetric flow rate of 5 LPM (based on the 10 LPM device flow rate). The two one-way valve inlets (see Label 2 in Figure 4) were each assigned a mass flow rate that gave a volumetric flow rate of 17.5 LPM, which totaled a flow rate of 45 LPM through the model outlet to match the NGI operating conditions (see Label 3 in Figure 4). Considering the water vapor entering the domain under humid airway conditions, the water vapor mass fraction at the one-way valve inlets was consistent with the 99% RH at 37°C in the environmental chamber. As the DPI was actuated with compressed air from a wall outlet, the water vapor mass fraction at the nostril inlets was consistent with the 10% RH that is expected from dry wall air. All wall boundaries in the CFD model use the no slip shear condition, the effects of surface roughness on the flow field and particle trajectories were neglected, and the deposit-on-touch particle boundary condition was used to model deposition. As the walls of the in vitro model were pre-wetted to be representative of upper airway conditions, the water vapor mass fraction on all model walls was 99% RH at a wall temperature of 37°C.

2.4.3. Particle Transport and Growth

The discrete phase model (DPM) available in FLUENT v19.3 was used to calculate the particle trajectories and deposition through the domain, with the Runge-Kutta scheme selected to integrate the particle equations of motion. All DPM settings followed our previously defined best practices and modeling recommendations (Bass, Boc, et al., 2019; Bass & Longest, 2018; Longest et al., 2007; Walenga & Longest, 2016), which have been validated against experimental deposition data in the upper airways. NW corrections, implemented via FLUENT user-defined functions (UDFs), were applied to correct the over-prediction of micro-particle deposition that is often observed with the LRN k-ω model. These NW corrections have been discussed in detail in our previous studies (Bass, Boc, et al., 2019; Bass & Longest, 2018; Longest et al., 2007) and provide an accurate match between CFD and experimental deposition data. The effect of two-way coupling between the flow field and particles, which is where the particle trajectories and growth influence the continuous phase and vice versa, was evaluated as a possible improvement to matching numerical and experimental growth results. This is in contrast to one-way coupling where only the flow field influences the particle trajectories and growth. For two-way coupling, changes to water vapor mass fractions in the continuous phase, based on mass transfer in and out of the discrete phase, is implemented via a UDF and was previously described by Longest and Hindle (2011). When evaluating two-way particle tracking, reducing processing times is an important consideration due to the additional computational requirements. The initial particle diameter entering the nostrils is based on the 1.53 μm MMAD for the best-case device and nasal cannula combination from Farkas et al. (2020). For the nasal cannula models, a polydisperse size distribution was used to demonstrate how increasing particle size affects patient interface losses. For the growth chamber models, preliminary work showed that a monodisperse aerosol, as opposed to using a polydisperse size distribution, was capable of providing the expected aerosol growth and minimized computational expense. To ensure particle convergence, a total of 10,000 particles were introduced to the domain, with 5,000 particles at each nostril inlet.

The current study implemented a modification to the NW correction UDFs regarding the threshold, below which, the damping of the wall-normal velocity is applied to model particle-wall hydrodynamic interactions. Previous versions of the NW correction UDFs used a critical wall-normal distance between the particle and wall (called the NW limit), and the wall-normal velocity was damped in the region below this value, with CFD models using a typical NW limit value of 1–2 μm. A drawback to this approach in the current study is that there is a vast difference between the average velocity in the upper airways and growth chamber, so a single NW limit value was not able to achieve adequate model validation for both regions. Therefore, a local Stokes number was implemented to define the criterion for damping the wall-normal velocity as a non-dimensionalized approach. The local Stokes number (Stk_l) was defined as:

$${\text{Stk}}_l=\frac{v_p{\rho }_p{d}_p^2{C}_c}{18{\mu }_f{h}_p}$$ (1)

where: v_p is the particle velocity magnitude, ρ_p is the particle density, d_p is the particle diameter, C_c is the Cunningham slip correction factor, μ_f is the fluid viscosity, and ℎ_p is the wall-normal distance between the particle and wall. With this implementation of the NW correction UDFs, a critical Stk_l of 15 was used to determine where particle-wall hydrodynamic interactions are modeled (as opposed to the NW limit), which gave the best validation with experimental deposition data throughout the model geometry.

As a dry particle absorbs water from the humid airways it becomes a droplet, with this condensation as well as evaporation of water implemented in FLUENT v19.3 via a user defined function (UDF). Longest and Xi (2008) originally reported the heat and mass transfer model for multicomponent hygroscopic aerosol in the respiratory airways that was used in the current study. These aerosol growth UDFs have been successfully validated against experimental data in a coiled tube geometry (Longest & Hindle, 2012a), adult growth chamber models (Longest et al., 2015; Longest, Tian, Li, et al., 2012), and for noninvasive ventilation high flow therapy (Golshahi, Tian, et al., 2013). Further details on the specifics of the particle growth UDFs with regard to growth chamber models were described by Longest et al. (2012), and the material and hygroscopic properties for the drug and the excipients were presented by Longest and Hindle (2011).

3. Case Study: Results

3.1. Influence of Rod Array on Nasal Cannula Losses

Table 2 compares the experimentally determined aerosolization performance of delivery systems that employ a nasal cannula both with and without a rod array. These results show no statistical significance between the two cannula designs in terms of DPI retention or cannula emitted dose (p-value of 0.21 and 0.08, respectively). However, the cannula retention and particle size (as MMAD) is significantly lower for the device that does utilize a rod array for jet attenuation (p-values of 0.01 and <0.001, respectively). This demonstrates that reducing the intensity of the inlet jet that enters the patient interface reduces losses in the cannula, and hence maximizes available lung dose to the patient. Furthermore, the rods provide secondary powder break-up mechanisms that reduce the aerosol size, which in turn improves delivery through the nose.

Table 2:

Experimentally determined aerosolization performance of the dry powder inhaler and nasal cannula delivery system both with and without a rod array utilized in the patient interface.

	Nasal Cannula without Rod Array	Nasal Cannula with Rod Array
DPI Retention [%]	17.4 (1.2)	18.2 (0.9)
Cannula Retention [%]	8.9 (0.3)	6.0 (1.0)*
Cannula Emitted [%]	73.7 (0.9)	75.9 (1.8)
MMAD [μm]	1.94 (0.03)	1.67 (0.02)*
Recovered [%]	97.1 (3.0)	97.6 (1.3)
FPF <5μm [%]	85.4 (0.3)	95.5 (0.7)*
FPF <1μm [%]	15.7 (0.6)	18.7 (0.5)

MMAD: Mass-median aerodynamic diameter

FPF: Fine particle fraction

^*p < 0.05; paired t-test; significant improvement in aerosolization performance with implementation of 3D rod array

Figure 6 illustrates the CFD-predicted flow field and particle deposition patterns in the nasal cannulas, as well as validation of the CFD deposition results against the experimental data. First, the CFD models, including the assumption of particles bouncing off the front half of the rods and sticking to the rear, shows good validation against the experimental testing, with CFD-predictions of losses in the patient interface falling within the experimental SD in both cases. The iso-surfaces of velocity magnitude demonstrate the jet attenuation capabilities of the rod array, as in Figure 6a the high-velocity jet extends into the cannula up to approximately the cannula bifurcation, but in Figure 6b the jet is completely dissipated by approximately 25% of the cannula length. Furthermore, the deposition pattern without rods in Figure 6a shows that particles in the 1–5 μm range, which accounts for the bulk of the aerosol size distribution, readily deposit on the cannula bifurcation and prongs. Conversely, Figure 6b with rods shows less deposition of particles in the 1–5 μm range in these regions, with only the smaller (<1 μm) particles (which account for much less aerosol mass) being lost in the patient interface walls due to more secondary flow induced by the rod array.

Figure 6:

Comparison of flow field and particle deposition patterns in the nasal cannula with and without a rod array (half-plane view). Velocity isosurfaces are for 5, 10, and 15 m/s. The rod array is observed to attenuate the inlet jet, which reduces aerosol losses in the patient interface and minimizes particle momentum in the ET region.

One of the technologies identified as being key to maximizing delivery to pediatric CF patient is the use of appropriately sized aerosols. Figure 7 shows CFD-predicted deposition profiles (deposition fraction vs. aerodynamic particle diameter) in the nasal cannula both without and with rods. The plot labels points at an MMAD of 3.5 μm and 5.0 μm (consistent with sizes typical of adult commercial DPIs) which lead to an approximate 2- to 5-fold increase in patient interface losses over the pediatric air-jet DPI presented in this study. Furthermore, the small particle size (1.67 μm (0.02 μm) MMAD) that was achieved with utilization of the rod array in this nasal cannula is expected to maximize nasal transmission downstream of the patient interface. Applying the 1.67 μm MMAD with the 10 LPM device flow rate to the correlations for pediatric NT losses from Golshahi et al. (2011) suggests average (SD) NT losses of 1.6% (1.1%), and ranging from 0.5% to 4.9% across the 13 subjects in their study. Using a 3.5 μm MMAD and 10 LPM flow rate gives losses of 8.0% (4.7%) (ranging from 2.8% to 21.0%), and a 5.0 μm MMAD gives losses of 16.0% (7.9%) (ranging from 6.3% to 36.3%). In summary, appropriately sized particles and the aerosolization performance of the pediatric air-jet DPI with rod-array nasal cannula can produce high efficiency lung delivery of the aerosol.

Figure 7:

Deposition profiles for the nasal cannula designs (a) without rods and (b) with a rod array for jet dissipation.

3.2. Experimental Upper Airway Deposition and Growth

Table 3 summarizes the aerosol deposition and growth results from experimental testing of the 5–6-year-old model under room (22.0°C and 33.5% RH) and humid airway (37.0°C and 99.0% RH) conditions. The mean (standard deviation (SD)) delivery system ED (exiting the cannula and based on the device loaded dose) was 80.1% (3.6%) and 79.4% (2.8%) for room and airway conditions, respectively. This is consistent with the previous experimental testing of the same device and nasal cannula combination by Farkas et al. (2020) (reported as 79.5% (2.6%)) without the nasal model. Nose-to-lung aerosol delivery was very good and provided high-efficiency delivery to the lower airways (B4+), with mean (SD) NT-B3 deposition losses of 4.2% (0.9%) and 5.3% (2.3) for room and airway conditions, respectively. This is also consistent with the correlations from Golshahi et al. (2011) that were mentioned in the previous section. Comparing upper airway losses between the room and airway conditions results shows EEG aerosol growth in the humid conditions does not affect impaction deposition in the NT and upper TB regions, and is hence capable of targeting the lower airways, as there is no statistical significance between the room and airway NT-B3 DFs (p-value = 0.25). In vitro losses in the growth chamber were within an acceptable limit (<5%) under both the room and airway conditions with mean (SD) DFs of 3.7% (2.1%) and 0.9% (0.6%), respectively. There was a significant increase in growth chamber losses under room conditions compared to the humid experimental test (p-value = 0.03), which may be attributed to increased static charge on the cast acrylic cylinder when humidity is relatively low. Comparing aerosol growth between the room and airway conditions, the MMAD significantly increased from 1.70 μm to 3.19 μm (with an absolute difference of 1.49 μm or a growth ratio of 1.9) due to humidity in the respiratory airways, which is expected to provide targeted drug delivery to the lower airways.

Table 3:

Summary of aerosol deposition and growth from experimental testing under room and humid airway conditions. Deposition fractions are defined based on device loaded dose. Experimental values are given as means with standard deviations shown in parenthesis [n=3].

	Room Conditions (22.0°C and 33.5% RH)	Airway Conditions (37.0°C and 99.0% RH)
Aerosol Deposition
DPI DF [%]	13.3 (2.1)	13.8 (3.1)
Cannula DF [%]	6.6 (1.5)	6.9 (0.5)
System ED [%]	80.1 (3.6)	79.4 (2.8)
NT-B3 DF [%]	4.2 (0.9)	5.3 (2.3)
Chamber DF [%]	3.7 (2.1)	0.9 (0.6)*
Model PF [%]	70.6 (5.0)	70.1 (1.4)
Aerosol Growth
MMAD [μm]	1.70 (0.05)	3.19 (0.30)*

RH: Relative humidity

DF: Deposition fraction

ED: Emitted dose

NT-B3: Nose-throat to Bifurcation 3

PF: Penetration fraction

MMAD: Mass-median aerodynamic diameter

^*p < 0.05; paired t-test; significant difference in aerosol characteristics under humid conditions

3.3. Validation of the Growth Chamber CFD Models

The validation of CFD-predicted aerosol deposition and growth with experimental data is illustrated in Figure 8. Figure 8a shows deposition results in the computational domain under room conditions, with DFs based on the ED from the delivery system (as opposed to loaded dose), due to the fact that the CFD model does not include the DPI or patient interface. Numerical deposition predictions in the NT-B3 model and growth chamber regions match the in vitro data well and are within the experimental SD, which indicates successful validation of the particle tracking models. CFD predictions of depositional losses in the growth chamber are somewhat lower than the experimental results, but as mentioned previously, it is suspected that static charge on the acrylic cylinder walls in the experiments increased losses in this region. The CFD models do not include the effect of static charge on wall boundaries, and hence this behavior was not accounted for in the numerical deposition results. As stated in the Methods, a 10 mm coupling region between the healthy NT and CF-diseased TB regions was required in the model geometry. Deposition patterns in the trachea, where the coupling region is located, do not show any signs of abnormal deposition, which suggests the steps taken to connect the NT and TB regions do not adversely affect interpretation of the results. Due to the relatively small amount of hygroscopic particle growth that occurs under room conditions, aerosol MMADs are not compared between experiments and CFD results in Figure 8a.

Figure 8:

Validation of the numerical models showing comparison of deposition fractions (DF) in the airway and chamber regions between the CFD and in vitro experimental (Exp.) results (a) under room (i.e. no particle growth) conditions (22.0 °C and 33.5% RH) and (b) under upper airway conditions (37°C and 99% RH). Panel (b) also includes a comparison of MMAD at the model outlet between the in vitro and CFD predictions for validation of aerosol growth. Note that DFs here are given based on system emitted dose, instead of loaded dose, as the CFD models do not include the delivery system.

Figure 8b compares aerosol deposition and growth between experimental and numerical data under humid airway conditions, with one-way coupling between the continuous and discrete phases. As with the results from room conditions, CFD-prediction of DFs in the NT-B3 model and growth chamber regions are well validated against in vitro data and fall within the experimental SDs. A small decrease in airway model deposition fraction was observed for humid airway conditions, which is surprising considering that aerosol size increase should increase deposition by impaction. However, this small decrease can be attributed to the random nature of both the turbulent dispersion model and anisotropic NW turbulence corrections, and may also stem from small variations in the flow fields for each case. Nevertheless, both CFD models are well validated and fall within the experimental SDs for aerosol deposition, so the CFD predictions are considered an accurate representation of particle transport. There was also no statistical significance between NT-B3 losses when comparing experimental results under room and airway conditions, so the DF increase that was observed during in vitro testing cannot be attributed only to the presence of humidity. Looking at losses in the growth chamber, static charge on the cylinder walls is expected to be lower under humid conditions, and this supports the decrease in depositional losses in this region that is seen in Figure 8b. The CFD-predicted deposition on the growth chamber walls also shows a closer match to experimental results in Figure 8b, which further supports the claim that growth chamber losses under room conditions were due to static charge on the cylinder walls. The CFD-predicted MMAD at the growth chamber outlet compares well to the in vitro data, which indicates successful validation of the one-way coupled particle transport models and evaporation/condensation UDFs.

Comparisons between the one-way and two-way coupled particle tracking methods for aerosol deposition and growth are provided in Table 4. Results show there is little difference between the two methods in the current study, with a maximum absolute difference between two-way and one-way particle tracking for NT-B3 DF, Chamber DF, and outlet MMAD of 0.1%, 0.7%, and 0.06 μm, respectively. For two-way coupling, results are presented for both 10 and 20 coupled cycles between the continuous and discrete phase, which shows that doubling the number of coupled cycles does not influence aerosol deposition and growth. Two-way coupled particle tracking consumes water vapor (and reduces the RH) from the continuous phase as the particles grow under humid conditions, but the relatively large volume of the growth chamber region means a large amount of humid air is available, which is the most likely reason why there is little apparent difference between one-way and two-way coupled particle tracking in the current study. When considering particle paths through narrow airway bifurcations, the effect of using two-coupled particle tracking may be more pronounced. As the CFD model is well validated against experimental results using one-way particle tracking (see Figure 8), this method is used throughout the remainder of this study.

Table 4:

Comparison of aerosol deposition and growth between the one-way and two-way coupled particle tracking models. Deposition fractions are based on delivery system emitted dose.

	One-way Coupling	Two-way Coupling 10 Cycles	Two-way Coupling 20 Cycles
Aerosol Deposition
NT-B3 DF [%]	4.8	4.7	4.8
Chamber DF [%]	1.7	1.2	1.0
Aerosol Growth
MMAD [μm]	3.38	3.39	3.32

NT-B3: Nose-throat to Bifurcation 3

DF: Deposition fraction

MMAD: Mass-median aerodynamic diameter

3.4. Flow Field Characteristics in the Upper Airways

Figure 9 shows plots of velocity magnitude and relative humidity to illustrate the flow field characteristics in the computational domain. Figure 9a presents a contour plot of velocity magnitude on a plane that cuts through the left nostril, nasal passage, trachea, and growth chamber. The figure demonstrates the wide range of velocity magnitudes in the CFD model; for example, the velocity magnitudes at Point A and B were 8.55 m/s and 0.16 m/s respectively. This justifies the implementation of a local Stokes number to determine the upper limit of the NW region, below which, particle-wall hydrodynamic interactions are modeled. The inset in Figure 9a shows a close-up view of flow through the larynx and clearly illustrates the laryngeal jet phenomena. Similarly, Figure 9c shows the three-dimensional nature of the laryngeal jet with an iso-surface of 7.0 m/s velocity magnitude through the trachea.

Figure 9:

Summary of flow field characteristics in the 5–6-year-old model showing (a) contours of velocity magnitude (Vel. Mag.) with a close-up view of the laryngeal jet, (b) contours of relative humidity (RH) with a close-up view of the nostril inlet (with dry wall inlet air) entering the in vitro model, (c) iso-surface of 7 m/s velocity magnitude depicting the 3D nature of the laryngeal jet, and (d) iso-surface of 80% RH in the nasal cavity.

Figure 9b shows contours of relative humidity on the same plane as Figure 9a, with the inset providing a close-up view of the nostrils and anterior nose. This illustrates the introduction of the dry actuation air (at 10% RH) to the computational domain, which is delivered from the cannula prong outlets in the experimental model. By the time the flow reaches the nasopharynx region, observations from the CFD model show the dry air has mixed with the humid air and RH values are greater than approximately 90%. Conditions in the growth chamber are close to 99% RH, which facilitates aerosol growth and provides an accurate representation of the respiratory airways. Figure 9d shows an iso-surface of 80% RH in the nasal passage and further illustrates rapid mixing of dry and humid air in the anterior region of the nose.

3.5. Residence Time and Maximum Particle Growth

Figure 10 plots the particle diameter (as MMAD) vs. residence time for the 5–6-year-old CFD model. The points represent the diameter and time for each individual particle that was tracked through the domain, at the time that each particle either deposited or exited the mixing chamber, up to a maximum time of 5.0 s. The red dotted line shows a logistic best-fit curve that represents the average particle diameter vs. residence time profile, which is given by:

$$d_p=\frac{2.67}{0.75+\text{exp}\left(-4.5t\right)}$$ (2)

where d_p is the particle diameter and t is the particle residence time. This shows that the MMAD begins to plateau around 1.2 s and levels off completely by a residence time of 2.0 s. As t → ∞ in Equation (2), the particle diameter (d_p) approaches an MMAD of 3.56 μm. From this, the t₉₀ is defined as the residence time used in Equation (2) that gives an MMAD that is 90% of 3.56 μm (i.e., 3.2 μm), which for this case is 0.55 s. Therefore, a residence time of approximately 2.0 s is more than sufficient to maximize aerosol growth and target drug delivery to the lower airways.

Figure 10:

Plot of particle diameter vs. residence time, with each individual point representing a single particle at the time it deposited or exited the computational domain. The red dashed line shows a logistic best-fit curve and the dotted line marks the average chamber residence time (~2 seconds), which shows that chamber residence time is sufficient to maximize particle growth.

3.6. Comparison of Deposition and Growth between Age Groups

CFD-predicted particle trajectories (colored by MMAD) and regional DFs are presented in Figure 11. All three models show that the particle MMADs were less than approximately 2.5 μm when the aerosol leaves the upper airways (NT-B3) and enters the growth chamber, which shows the hygroscopic properties of the powder formulation are capable of targeting drug delivery to the lower airways (B4+). This claim was further supported with CFD predictions of relatively low aerosol deposition in the NT-B3 region for all age groups. The MMADs at the outlet of the growth chamber for the 2–3- (Figure 11a) and 9–10-year-old (Figure 11c) models were 3.40 μm and 3.47 μm respectively, which is consistent with the 3.38 μm MMAD from the 5–6-year-old model (Figure 8b and Figure 11b). As stated above, a residence time of approximately 0.55 s was required to reach 90% of the average final particle size for the 5–6-year-old model. Similar expectations can be applied to the 2–3- and 9–10-year-old age groups, based on the consistency in CFD predictions of aerosol growth between all three models. The relatively large increase in aerosol size (absolute increase of 1.49 μm or a growth ratio of 1.9) suggests the EEG powder formulation is capable of targeting deposition in the lower airways. The larger particle diameters in the lower airways are expected to reduce exhalation losses with an increase in particle momentum and impaction, and improve lung retention with an increase in sedimentation deposition (if a breath hold maneuver is employed).

Figure 11:

Examples of CFD-predicted particle trajectories and growth through the domain for the (a) 2–3-year-old, (b) 5–6-year-old, and (c) 9–10-year-old models, with annotations of nose-throat to Bifurcation 3 deposition fraction (DF_NT-B3), growth chamber deposition fraction (DF_GC), and outlet mass-median aerodynamic diameter (MMAD_Out).

As mentioned previously, the DPI delivery system and powder formulation provide high-efficiency nose-to-lung transmission in the 5–6-year-old model with a NT-B3 DF of 4.8%, as shown by both Figure 8b and Figure 11b. For the 2–3-year-old model in Figure 11a, the NT-B3 losses are higher than the 5–6-year-old with a CFD-predicted DF of 10.9%. This increase in upper airway losses can be attributed to the smaller airway dimensions in the younger patient (see Table 1), which leads to an increase in impaction deposition. For the 9–10-year-old model in Figure 11c, the NT-B3 losses are also higher than the 5–6-year-old with a CFD-predicted DF of 7.0%. In this case, the increased losses are most likely due to the increase in length of the upper airways, which leads to longer residence times and more hygroscopic growth in the NT-B3 region. Looking at the particle trajectories in Figure 11c, the MMADs in the 9–10-year-old model are approximately 2.5 μm around the glottis, whereas the MMADs in a similar region for the 2–3- and 5–6-year-old models is less than approximately 1.75 μm. The increased particle size in the upper airways for the 9–10-year-old caused more impaction losses and accounts for the small increase in NT-B3 DFs. Despite the 2–3- and 9–10-year-old models having slightly higher deposition in the NT-B3 region than the 5–6-year-old model, upper airway losses were relatively low for all three cases and drug delivery to the lower airways (B4+) was high. Upper airway losses may be improved by designing air-jet DPIs specifically for the 2–3 and 9–10-year-old age groups with different air flow rates. For example, increasing inlet velocity in the 9–10 year age range may be expected to reduce hygroscopic growth and thereby decrease nasal depositional loss.

4. Discussion

The selected case study has satisfied the initial objective of using multiple new techniques simultaneously in order to overcome the primary limitations associated with dry powder aerosol administration to children and enables high efficiency trans-nasal DPI use in this population, based on concurrent CFD and realistic in vitro analysis. Techniques used to improve lung delivery efficiency of the dry powder aerosol included nose-to-lung administration in subjects as young as 2-years-old, use of a positive-pressure active DPI, implementation of patient interfaces that improved aerosol deaggregation and dissipation of the flow field, and controlled condensational growth of the aerosol within the airways. Resulting upper airway losses (NT-B3) for the 2–3-, 5–6-, and 9–10-year-old age groups were 10.9%, 4.8%, and 7.0%, respectively, which is a vast improvement to lung doses in pediatric patients compared to commercial devices. For example, Below et al. (2013) reported that the Novolizer and Easyhaler provide a lung dose (with respect to nominal dose) of 5% and 22%, respectively, on the tracheal filter of a 4–5-year-old in vitro model. For the Cyclohaler, HandiHaler, and Spinhaler, Linder et al. (2014) reported lung delivery efficiencies of 9% to 11%. The validated CFD models showed the aerosol MMAD is expected to grow to at least 3.2 μm in the lower airways after a residence time of approximately 0.6 seconds. Both nose-to-lung delivery and lung retention may be improved further by designing devices and powder formulations that are specific to each age group. Results in Figure 11 showed differences in NT-B3 losses between the three models, which are attributed to differences in airway dimensions between patients at different ages (see Table 1) and perhaps intersubject variability within each age group.

As a secondary outcome, this case study demonstrated the benefits of concurrent CFD analysis and experimental testing (as outlined by Figure 1) for the optimization and advancement of novel delivery systems and evaluation of aerosol delivery to in vitro models. The experimental methods provide initial testing of the delivery system components, and provide aerosol characterization and validation data to facilitate the CFD models. Evaluation of the validated CFD models allows for greater insight into the delivery system behavior and permits the testing of numerous design iterations.

As a key point in the case study, multiple technologies were implemented in the delivery system that maximized the available lung dose in pediatric patients. Sufficiently small particles (combined with good aerosolization performance from the device) can traverse the natural filtration of the nasal cavity, thereby reducing losses. Also, employing nose-to-lung administration allows the device to be used with younger patients that may not be able to use a mouthpiece correctly. The positive pressure actuation of the DPI, with a ventilation bag, improves patient compliance as the responsibility is on the caregiver to actuate the device, instead of the child actuating the device passively. It is also expected that the positive pressure will expand the patient’s airways, leading to further reductions in upper airway losses. Comparisons made between the experimentally-tested and CFD-predicted performance of the nasal cannula, both with and without rods, demonstrated that utilizing a rod array in the patient interface can both minimize losses in the patient interface and reduce the aerosol size that enters the NT region. This small particle size reduces impaction deposition losses in the nasal cavity, as demonstrated by the small (approximately 5%) upper airway loss from the 5–6-year-old growth chamber model. Finally, the highly-dispersible spray-dried EEG powder, which grew to an MMAD of approximately 3.2 μm after 0.55 sec residence time, is expected to target delivery in the lower airways, where bacterial infection is more difficult to eradicate.

A primary limitation of the current study for evaluating antibiotic delivery to CF-diseased airways is the use of AS EEG powder as a surrogate test aerosol. Despite this limitation to the current study, inferences on the ability for the delivery system to provide highly-efficient nose-to-lung delivery are still valid, providing future antibiotic (e.g. tobramycin) EEG powder formulations have similar hygroscopic capabilities as the AS EEG aerosol. The EEG aerosol may also pose problems in terms of moisture ingress during storage, due to the hygroscopic excipients. This is an issue for dry powders in general, and further steps are planned for this system to prolong shelf life, such as foil-wrapped capsules for long-term dose storage. Other limitations include the use of a growth chamber to capture particle growth in the lung pathways, modifications to the cannula prongs specific to each patient, and a lack of evaluation regarding interpatient variability. The growth chamber was designed to provide a residence time of two seconds, which is consistent with transport through the airways, but particle trajectories through lung bifurcations are vastly different and may have an influence on aerosol growth. As such, evaluation of EEG aerosol transport and growth through complete airways of CF-diseased lungs is required, which may be modelled using the stochastic individual pathway methodology (Longest et al., 2015; Longest & Hindle, 2009a; Longest et al., 2007; Longest, Tian, Delvadia, et al., 2012; Longest & Vinchurkar, 2007; Tian, Longest, Su, Walenga, et al., 2011; Walenga & Longest, 2016). Modifying the cannula prong dimensions and angular rotation for the 2–3- and 9–10-year-old age groups may influence upstream losses in the patient interface. If device development of age-specific devices is conducted, the effect of modifying the cannula design should be tested at the same time. Finally, the current study evaluated a single upper airway model for each of the three age groups, but interpatient variability is known to be significant for nasal deposition of aerosols (Garcia, Tewksbury, Wong, & Kimbell, 2009; Golshahi et al., 2012; Golshahi et al., 2011; Storey-Bishoff et al., 2008) and should be taken into consideration.

In conclusion, three growth chamber models were developed that represent the diseased upper airways of pediatric patients diagnosed with CF in the age ranges of 2–3, 5–6, and 9–10 years old. The CFD models successfully validated aerosol deposition and growth through the computational domain against experimental data via the implementation of NW corrections and aerosol condensation and growth models. Results show that the chosen air-jet DPI and nasal cannula delivery system, which combines multiple new technologies for DPI aerosol administration, was capable of providing highly-efficient nose-to-lung aerosol delivery to the lower airways (B4+) of children, with NT-B3 losses of 10.9%, 4.8%, and 7.0% for the 2–3-, 5–6-, and 9–10-year-old models, respectively, based on concurrent CFD predictions and highly realistic in vitro experiments. Lung retention of the delivered aerosol is also expected to be high, as the aerosol grows to an MMAD of approximately 3.2 μm in all three cases after a residence time of approximately 0.6 seconds.

• Key aerosol and device technologies are presented that maximize delivery to pediatric airways:

  • Nose-to-lung administration with sufficiently small particles
  • Active positive-pressure DPIs
  • Patient interfaces that minimize turbulence and jet momentum effects
  • Highly dispersible spray-dried powder formulations that change size within the airways

• Using these technologies gave <11% total loss from the nasal cavity to the third bifurcation branch

• The aerosol grows from an MMAD of 1.5 to 3.2 μm, after 2 s residence time

• 90% of aerosol size increase occurs in first ~0.6 seconds

Biographies

Karl Bass completed his BEng at the University of West England and his PhD at Virginia Commonwealth University. He has 10 years’ experience applying computational fluid dynamics models to engineering problems in the fields of internal combustion engines and respiratory drug delivery.

Dale Farkas completed his PhD in Mechanical and Nuclear Engineering at Virginia Commonwealth University. He is currently working to design devices that enable high efficiency delivery of dry powder aerosols to the lungs for the treatment of respiratory diseases.

Amr Hassan received his PhD in Material Science and Nanotechnology under a joint program between Ain Shams University, Cairo, Egypt and Clarkson University, NY, USA. His research is focused on the development of novel next-generation engineered inhalation formulations for effective respiratory drug delivery.

Serena Bonasera completed her MPharm at the University of Messina, Italy. She is currently a PhD student at VCU and her research is focused on preparation, optimization, and characterization of engineered spray-dried powder formulations for aerosolization using novel dry powder inhalers.

Michael Hindle has a BPharm and PhD from the University of Bradford, UK. His current research projects at VCU include combining key in vitro characterization studies with computational fluid dynamics (CFD) simulations to improve inhalation drug delivery for the most challenging patient groups such as neonates and children.

Worth Longest completed his PhD at NC State University in Mechanical Engineering with a specialization in multiphase transport and performed a postdoc at the US EPA in respiratory dosimetry. At VCU he has focused his research in the areas of developing effective new methods for generating and delivering medical aerosols, and developing numerical models and airway geometries to assess and improve respiratory drug delivery.