Experimental Evaluation of Dry Powder Inhalers During In- and Exhalation Using a Model of the Human Respiratory System (xPULM™)

Dry powder inhalers are used by a large number of patients worldwide to treat respiratory diseases. The objective of this work is to experimentally investigate changes in aerosol particle diameter and particle number concentration of pharmaceutical aerosols generated by five dry powder inhalers under realistic inhalation and exhalation conditions. The active respiratory system model (xPULM™) was used as a model of the human respiratory system and to simulate a patient undergoing inhalation therapy. A mechanical upper airway model was developed, manufactured and introduced as a part of the xPULM™ to represent the human upper respiratory tract with high fidelity. Integration of optical aerosol spectrometry technique into the setup allowed for evaluation of pharmaceutical aerosols. The results show that the upper airway model increases the resistance of the overall system and act as a filter for bigger particles (>3 µm). Furthermore, there is a significant difference (p < 0.05) in mean particle diameter between inhaled and exhaled particles with the majority of the particles depositing in the lung. The minimum deposition is reached for particle size of 0.5 µm. The mean particle number concentrations exhaled are 2.94% (BreezHaler®), 2.66% (Diskus®), 10.24% (Ellipta®) 2.13% (HandiHaler®) and 6.22% (Turbohaler®). In conclusion, the xPULM™ active respiratory system model is a viable option for studying interactions of pharmaceutical aerosols and the respiratory tract in terms of applicable deposition mechanisms. The model can support the reduction of animal experimentation in aerosol research and provide an alternative to experiments with human subjects.


Introduction
According to the report on the global impact of respiratory disease published by the Forum of International Respiratory Societies from 2017 [1] approximately 65 million people globally suffer from mild to severe chronic obstructive pulmonary disease (COPD) and 334 million people suffer from asthma. In conjunction with acute lower respiratory tract infections, these diseases are among the most prevalent severe illnesses and causes of death.
[1] Based on Eurostat statistics [2] from 2016, diseases of the respiratory system accounted for approximately 7,5% of all deaths in the former EU-27. Targeted delivery of pharmaceuticals directly into the affected part of the respiratory region via inhalation drug therapy is crucial for managing cases of obstructive respiratory diseases [3].
Inhalation therapeutic devices can be categorised into four main types, including nebulisers, pressurised-metered dose inhalers (pMDI), soft mist inhalers (SMI) and dry powder inhalers (DPI) [3]. In terms of units sold in 2014, pMDIs and DPIs constitute the majority of devices for inhalation drug delivery [4]. For this reason, this article focuses purely on the evaluation of DPIs. In contrast to pMDIs, DPIs work with larger lactose particles carrying the active substance and are environmentally preferable due to the absence of hydrofluorcarbons [5]. Nevertheless, DPIs require a minimum peak flow during inhalation, created by the patient, to detach and propel the aerosol in direction of the lung regions. The lack of required synchronity between activation and inhalation of the DPI is reducing a potential source of misuse but does not cover for the problems caused by different activation mechanisms [6]. The optimum flow profile varies for the currently available DPIs and may lead to a suboptimal delivered dose for the patients [6]. Most recently developed DPIs only deliver a low dose of medication while the users have to be able to create a minimum inspiratory flow and have the cognitive ability to properly operate the DPI [7]. This is accompanied by the need for an adequate lung volume of the user, therefore excluding children below the age of 5 years as users [7]. Considering only a single peak inspiratory flow rate (PIFR) value as the main criterion for determining the capability of the patient to use an inhaler efficiently may be an insufficient criterion as the DPIs vary in their design and resistance to airflow [8]. While several available studies evaluating DPIs focus mainly on inspiratory flow rate [9][10][11][12][13] a more suitable criteria has proven to be ensuring a sufficient pressure drop of ≥1 kPa across the device. Inhalation under these conditions leads to delivery of an adequate lung dose of the pharmaceutical [8]. Focusing on the pressure drop across the device could help prevent exclusion of patients from DPI usage due to insufficient or excessive peak inspiratory flow rate. Both have been shown to negatively impact pulmonary drug delivery [9,14]. Therefore, the pressure drop over the DPI has been taken as the main evaluation criterion for successful inhalation processes for this work.
In vitro pharmaceutical aerosol test systems often include either sample collection tubes or cascade impactors, such as the Andersen non-viable impactor, or the Next Generation Impactor to collect the particles for classification [15][16][17]. The results using such systems provide insights about the properties of the inhaled aerosol, such as the sizes of the inhaled particles and the deposition fraction, which can be used for comparison with radionuclide imaging studies [18] or to validate in-silico dosimetry models [19][20][21]. Cascade impactors consist of stages, each containing impaction plates which represent obstacles for an incoming airstream [22]. These plates create an abrupt bend in the airstream causing the particles, whose inertia exceeds a cut off size, to deposit [23]. Due to the operating principle of cascade impactors and sample collection tubes, evaluation of aerosol during consecutive inhalation and exhalation is not feasible [24]. The aerosol particles deposit on the impaction plates during inhalation and are consequently not present in the exhalatory airstream . For this reason, optical aerosol spectrometry was utilised in this work allowing for evaluation of particles within both inhalation and exhalation airstream.
Pulmonary drug delivery is based on the primary mechanisms of aerosol deposition, which are defined as inertial impaction, gravitational sedimentation, Brownian diffusion, turbulent deposition, electrostatic precipitation, and interception [25]. The effect of deposition mechanisms on aerosol particles depends on the particle characteristics such as particle size, overall size distribution, shape, composition, surface characteristics and charge [26]. Moreover, the processes resulting from molecular transfer between particles and their respective surrounding gas are nucleation, condensation, evaporation hygroscopicity and coagulation [27]. Inhalation drug therapy aims at targeted delivery of pharmaceuticals into the lung. The inhaled particles must overcome filtration mechanisms in the upper airways causing them to deposit within this region [28]. The deposition mechanism occurring mainly in the upper airways is inertial impaction affecting mostly large particles (>2-5 µm) with a strong dependency on the airflow rate. The deposition in this region of the respiratory tract results from direction changes of flow when the particles deviate from the streamline and collide with the airway walls. The probability of such deviation can be described by the Stoke's number where particle diameter, carrier gas viscosity and airway diameter are used to calculate the probability of deposition. [29] In the respiratory tract, gravitational sedimentation of particles in the size range of (>1-8 µm), refers to the settling of particles under the influence of gravity. Brownian diffusion results from random motion and the collision of the particles with the carrier gas molecules. The effect of mutual repulsion due to electric charges concerning the inhaled particles is defined as electrostatic 3 of 15 precipitation. The described mechanisms arise mostly in the upper and conducting airway region of the respiratory tract, whereas diffusion and electrostatic precipitation is also taking place in the acinus region of the pulmonary system for particles <3 µm. [25] The objective of this work is to experimentally investigate changes in aerosol particle diameter and particle number concentration of pharmaceutical aerosols under realistic inhalation and exhalation conditions, resulting in a calculated lung deposition. The active respiratory system model (xPULM) used in this work includes two core elements; a computed tomography (CT) derived mechanical upper airway model (UAM), and a primed porcine lung serving as human lung equivalent. This setup is used to represent a patient undergoing inhalation therapy. In contrast to widely spread measurement setups, this work integrates an optical aerosol spectrometer for inhalation and exhalation measurements to eliminate the drawbacks introduced by cascade impactors [24,30]. To cover a wide spectrum of devices used in clinical practice, five commonly used DPIs are investigated. Instead of focusing on PFIR, the focus was put on reaching a pressure drop of at least (P DROP ≥1 kPa) for all inhalers. This article aims to provide an alternative respiratory model suitable to reduce animal experimentation in aerosol research. Furthermore, the work aspires to mitigate the shortage of experimental data, viable to substitute demanding and constrictive experiments with human subjects. Moreover, the experimental setup including the xPULM™ model, can be seen as a basis for an alternative to animal testing, as the porcine lung, included in this trial, has been salvaged from an abattoir.

Measurement Setup and Procedure
The following two measurement trials were conducted during this study: A) Characterisation measurements and B) Respiration measurements. To assess the particles generated by the DPIs, Characterisation measurements were performed using a simple connection element to the respiratory model xPULM™ . This connector is characterised by a simplified version of the human laryngeal space, in form of a 90 degree bend and includes a sampling nozzle. This aerosol sampling point is in-line with the inhalatory airstream to ensure isoaxial aerosol sampling. Moreover, the control loop of the optical aerosol spectrometer ensures isokinetic sampling by maintaining a constant sampling flow, regardless of the inhalation flow profile. The active model of the human respiratory system, xPULM™, was used with polymer breathing bags, to simulate the inhalation effort of a patient during particle characterisation measurements. In the second step, Respiration measurements (see Figure 1) were conducted to investigate changes in aerosol particle diameter and particle number concentration during inhalation and exhalation. For this purpose, a primed porcine lung was used as a anatomically realistic lung equivalent. The porcine lung has been proven to be a suitable model of the anatomy of the human lung [31] and has been used in previous studies to research the pathogenesis of diseases such as cystic fibrosis [32]. The lung equivalent was connected to a mechanical UAM which was rapidly manufactured using 3D-printing techniques. The UAM is based on a clinically annotated CT examination of a healthy subject. In contrast to the characterisation measurements, sampling took place at the lower end of the mechanical UAM to assess the influence of its geometry on the measured values.The DPIs were mounted to the UAM using custom mouthpiece adapters for ensuring airtight connection. The measurement procedure of the Respiration measurements consists of three phases (i) inhalation, (ii) breath-hold, (iii) exhalation. Inhalation with maximum effort was simulated until the pressure drop across the DPI, measured with the FlowAnalyser PF-300 (IMT Analytics, Switzerland), reached at least the recommended pressure drop of ≥1 kPa [8]. However, if achieveable, a pressure drop of 4 kPa, was targeted [33]. The driving force of the inhalation was terminated When the peak value of the pressure drop was reached. However, inhalation continued briefly, due to inertia and compliance of the lung equivalent. Inhaler-specific inhalation profiles were recorded using mass flow sensors SFM3000 (Sensirion, Switzerland).
The inhalation manoeuvre was followed by a 5 s breath-hold period prior to slow and steady exhalation at a flow of 30 L/min for the duration of 6 s. For each tested DPI, the measurements were repeated 12 times (n=12). The in-/exhalation airstream was sampled by the optical aerosol spectrometer Promo 2000 (PALAS, Germany) connected to a white light aerosol sensor Welas 2070 (PALAS, Germany) with a constant flow rate of 5 L/min. This allowed for isokinetic measurement of particle size distribution and particle number concentration. The sensor is capable of measuring particles in the range of 0.2 µm to 10 µm.

Model of the Human Respiratory System
The active model of the human respiratory system xPULM™ has been used in this study. The xPULM™ replicates human breathing efforts exerted during the use of DPIs. Fundamental respiratory characteristics (e.g., flow, pressure, and volume) of a rapidly inhaling human are captured during the simulation with high fidelity. Properties of the human respiratory system such as airway resistance and lung compliance are represented by using lung equivalents (porcine lungs) and mechanical UAMs (based on CT examinations). The displacement of gasses during spontaneous breathing occurs due to the pressure difference between the atmosphere and the human lung. This physiological process is recreated by the xPULM™ . During the breathing simulation, pressure changes in the thoracic chamber are induced by the movement of a bellows system. For inhalation, a negative pressure is created in the chamber by expanding the bellows, leading to air following the pressure gradient resulting in inflation of the lung equivalent. During exhalation the opposite process occurs. The bellows is moved back to its original position, increasing the pressure in the chamber and deflating the lung equivalent. The movement of the bellows system can be precisely adjusted in the control software of the xPULM™ . This allows for the simulation of different breathing scenarios under various conditions as demonstrated in [34]. A detailed description of the xPULM™ functionality and components including validation measurements are presented in our previous work [35]. The mechanical UAM includes the oral cavity, oropharynx, larynx, and trachea. A CT examination of a 28-year-old, healthy, non-smoking, male was used for the UAM reconstruction. The subject has been annotated as healthy by clinical staff and did not show any sign of abnormal restrictions or geometrical limitations. Therefore, this CT examination has been considered to serve as a valid alternative to standardised airway models [36]. The selected dataset contained 280 images with a slice thickness of 0.75 mm and was exported in a Digital Imaging and Communications in Medicine (DICOM) format for further processing. The upper respiratory tract was segmented using a combination of thresholding and region growing techniques. The outcomes of the semi-automatic segmentation were inspected on a slide-by-slide basis and the segmentation parameters were adapted to obtain a precise segmentation of the upper airways. The resulting 3D model was exported as a Standard Triangle Language (STL) file and post-processed to be 3D-printable. The final 3D mechanical UAM was manufactured using rapid prototyping techniques and coated with resin. An emphasis was placed on positioning the model to minimise usage of support structure in the flow path. The dimensions for each section of the final model are summarised in Figure 2 and Table 1. Custom connectors were designed based on the geometry of each DPI to ensure an airtight connection between the inhaler and the mechanical UAM.

Representation of the Upper Respiratory Tract
All rapidly produced components were manufactured from Polylactic acid (PLA) with a wall thickness of 2 mm and a layer height of 0.2 mm. The lower respiratory tract consists of the bronchi, bronchioles, and alveoli, which form the lung. During breathing simulations, these structures have been represented by a primed porcine lung. The lung was salvaged from a slaughterhouse process and is therefore compliant with the 3R principles [37] which denote responsible use of animal or animal organs during experiments. The Nasco-guard® (Nasco, Wisconsin, USA) preservation process keeps the porcine lung inflatable, elastic and covered with the parietal pleura. These properties are necessary for physiologically and anatomically realistic simulations of human breathing.

Dry Powder Inhalers
In total, five DPIs were evaluated in this study, grouped into single-dose and multiple dose inhalers. The single-dose devices (BreezHaler® and HandiHaler®) are loaded with a capsule containing the dose which is punctured prior to use. The remaining three were multi-dose DPIs (Diskus®, Ellipta®, Turbuhaler®) which store multiple doses within the devices. Outlets of all DPIs were modified with custom rapidly manufactured adaptors to enable well-fitted, airtight, connection to the oral cavity of the mechanical UAM. The optical aerosol spectrometry measurements were conducted with 128 intervals per decade. The arithmetic centre of the intervals (x i ) is then: For further calculations, the differential particle number distribution q 0 (x i ) is defined as: where n i is the measured particle number within the interval limits x i,lower and x i,upper . Leading to the mean particle diameter M 1 calculation: Further information about the inhaled aerosol is obtained by calculating the particle number concentration: where the measured volume V m is defined as: where u is particle velocity and Iw is the cross-section of the optical sensor.

2
The measurement data is evaluated with non-parametric methods as the requirements    Table 3.

Flow profiles measured during Characterisation and Respiration measurements for
10 the evaluated DPIs are presented in Figure 3. During Characterisation measurements, the 11 resistance of the system is primarily resulting from the inner resistance of the DPIs. The   The flow results of the different DPIs, as shown in Figure 3, allow conclusions on the 28 handling of the inhaler and its characteristics during use. The HandiHaler® for example 29 shows a wider range of flow values as well as higher volatility in pressure drop values 30 (see Figure 4), than most of the other inhalers. This is mainly caused by the propelling 31 mechanism, which is based on the mechanical movement of the aerosol loaded capsule. 32 Based on the user guide, the capsule has to move (also acoustically noticeable) within the 33 inhaler, in order to disperse the powder. This oscillating movement leads to a volatile flow 34 and oscillating pressure drop measurements, therefore, characterisation of this inhaler is 35 influenced by the handling of the capsule and inhaler. 36 A comparable observation can be made for the use of the second capsule-based DPI, the 37 Breezhaler®. This product is also based on the oscillation of the capsule in order to propel 38 the aerosol properly. These oscillations are moreover influenced by the holding position 39 and angle of the device during inhalation. In contrast to the HandiHaler®, the capsule 40 within the Breezhaler® is not limited in movement mechanically, but mainly by gravitation. 41 When the Breezhaler® is moved to a horizontal position the likelihood of the capsule 42 dropping out of the holding cavity increases, impacting the aerosol production mechanism. 43 The correspondingly changed flow profiles caused by different lung equivalents can be 44 observed in Figure 3. The compliance of the introduced lung tissue (depicted by the flow 45 curves in Figure 3B) influences the peak flow as well as the flow profile. The anatomic 46 components of the used porcine lung and its geometric properties lead to a prolonged 47 inhalation time and flattened flow profile when using identical inhalation settings as with 48 the polymer breathing bags as lung equivalent. drop across the inhalers, see Figure 4B). The resistance of the measurement system increases 53 significantly (p < 0.05) with all inhalers (see Table 3) and a pressure drop of 4 kPa is reached 54 for lower inspiratory flow rates.

55
The measurements revealed limitations in reaching the pressure drop of 4 kPa con-  Respiration measurements are summarised in Table 3. These parameter values complement 65 the graphical result shown in Figure 3 and Figure 4. Additionally, they provide further 66 inside about the relationship between the inner resistance of the DPIs, inhaled volume, 67 inhalation time and peak inspiratory flow at particular pressure drop values.

68
The difference between the inner resistances of DPI measured during characterisa-69 tion and the values extracted from the literature is in an acceptable tolerance range of 70 ± 0.01 kPa 1/2 /L/min, with the exception of Diskus® (see Table 2 and Table 3). Results

71
show, that the Diskus® inhaler, which was included in the measurement setup, exhibits for an Ellipta® DPI [8,38]. This discrepancy was potentially caused by suboptimal storage 76 conditions for this inhaler. constantly changing inner geometry of the lung tissue, which influences the mean particle 102 diameter. Additionally, the high relative humidity within the lung tissue may lead to 103 hygroscopic growth and therefore also to adhesion of particles.  Figure 7.

111
There is a statistically significant difference (p < 0.05) between particle number concen-112 tration in inhaled and exhaled airstream for all tested inhalers (K-W test, H = 17.29, p = 113 0.00003). This is caused by particles depositing in the primed porcine lung.

114
The generated drug particles from the DPIs are inhaled through the mechanical UAM 115 which represents the naso-oro-pharyngo-laryngeal region (extrathoracic region). Larger 116 particles (>3 µm) deposit in this region mainly due to effects of inertial impaction [43]. The and Brownian diffusion [21,43].

121
Regional lung deposition and bronchodilator response of pharmaceutical aerosols was 122 studied extensively in previous works [44,45]. Their results confirm that small particles are 123 exhaled with exhalation fractions for particle diameters 1.5 µm, 3 µm and 6 µm being 22%, 124 8%, and 2% respectively [45]. A lung deposition study in healthy human subjects showed 125 a exhalation fraction of exhaled dose of 1.2% [46]. In this study, however, a MAGhaler DPI 126 was used to aerosolise the powder.  Table 3.

134
For a large number of patients, DPIs are the device of choice for the delivery of phar-135 maceuticals to manage asthma and COPD [9,48]. The number of commercially available 136 DPIs is growing [8] with inhalers varying in their design, operating mechanisms, and 137 resistance to inhaled airflow [8,49]. Accounting for these properties and the patient's ability 138 to use the specific device is essential for efficient drug delivery. Testing setups provide an option to evaluate aerosolised dry powders generated by DPIs and allow for further 140 insights into DPI performance under various conditions [50][51][52][53].

141
In this work, aerosol particle diameter and particle number concentration of pharma-

176
• The majority of particles entering the porcine lung deposit within, minimum deposi-177 tion is reached for the particle size of (0.5 µm). The primed porcine lung is therefore a 178 suitable lung equivalent and representation of the human lung.

179
• Sampling of the airstream during inhalation and exhalation and its subsequent evalua-180 tion using optical aerosol spectrometry techniques is a viable alternative to impactors 181 for evaluating pharmaceutical aerosols.

182
In conclusion, the xPULM™ active respiratory system model in combination with 183 the introduced mechanical UAM and the optical aerosol spectrometer is a viable option to quantify the regional deposition of pharmaceutical aerosols in lung tissue obtained 190 by 3R compliant processes. Additionally, coating of the inner surface of the mechanical 191 UAM will be considered, to ensure the least possible artifacts and interference of the 3D 192 printing materials on the particle transportation effects. Besides regional deposition, also 193