In Vitro Tests for Aerosol Deposition. I: Scaling a Physical Model of the Upper Airways to Predict Drug Deposition Variation in Normal Humans

Abstract

Background:

In vitro–in vivo correlations (IVIVCs) are needed to relate in vitro test results for deposition to mean data from clinical trials, as well as the extremes in a population. Because drug deposition variations are related to differences in airway dimensions and inhalation profiles, this article describes the development and validation of models and methods to predict in vivo results.

Methods:

Three physical models of the upper airways were designed as small, medium, and large versions to represent 95% of the normal adult human population. The physical dimensions were validated by reference to anatomy literature. The models were constructed by rapid prototyping, housed in an artificial thorax, and used for in vitro testing of drug deposition from 200 μg Budelin Novolizers using a breath simulator to mimic the inhalation profiles used in the clinic. In vitro results were compared to those reported in vivo.

Results:

The “average” model was scaled to produce “small” and “large” versions by multiplying linear dimensions by 0.748 or 1.165, respectively, based on reports of the mean and standard deviation of airway volume across a normal adult population. In vitro deposition variation under fixed test conditions was small. Testing in the model triplet however, using air flow rate versus time profiles based on the mean and the extremes reported in the clinic, produced results for total lung deposition (TLD) in vitro consistent with the complete range of drug deposition results reported in vivo. The effects of variables such as flow rate in vitro were also predictive of in vivo deposition.

Conclusions:

A new in vitro test method is described to predict the median and range of aerosol drug deposition seen in vivo. The method produced an IVIVC that was consistent with 1:1 predictions of total lung deposition from a marketed powder inhaler in trained normal adults.

Introduction

Lung deposition in different human subjects is often highly variable, even for a single inhaler.⁽¹⁾ Because of this, collecting proof of equivalent deposition between inhalers or inhaler prototypes is challenging because it demands clinical trials with the power to discriminate between devices; large variability demands large trials. Unfortunately, expensive trials constrain product development by precluding inhaler device changes once Phase 2 clinical trials begin.⁽²⁾ As a result, in vitro–in vivo correlation (IVIVC) discussions are frequent in regulatory circles, as these correlations may provide a way to predict and improve device performance without repeating large trials in different human cohorts.⁽³⁾

To be useful to the industry and its regulators, IVIVCs need to relate in vitro test results not only to mean data from human clinical trials, but also to the small and large extremes in a population that are best represented by lower and upper 95% confidence limits. Because the magnitude of the deposition variance seen in vivo from a given inhaler has not been well correlated to morphologic measurements of the upper airways [mouth–throat (MT), trachea (TB), and upper bronchi], this article describes the development and initial validation of new in vitro methods that seek to provide this information. A triplet of airway models is described that can be partnered with realistic inhalation profiles to provide in vitro estimates of the mean and 95% limits of in vivo deposition seen in normal human volunteers of both genders. Scaled models are described with reference to the existing literature on normal human airway dimensions. These have been constructed and used to collect drug deposition results for a marketed powder inhaler. To create and validate an IVIVC, in vitro results are compared to the clinical deposition results for the same inhaler. The approach that was used was similar to one described by Olsson et al.⁽⁴⁾

Materials and Methods

Physical models of the upper airways

An existing physical model of the MT and upper airways (Fig. 1)⁽²⁾ was scaled to accord with the literature describing the regional morphometry of the respiratory tract (RT) to produce the hollow tube models shown in Figure 2. As an initial hypothesis, we assigned the dimensions of the model shown in Figure 1 to those of an average RT or “medium-sized” model representing the upper oral airways of an averagely sized normal human of either gender. The total internal volume of this “average RT model” was 100.6 cm³, comprised of at least two distinct sections: MT=61.6 cm³ and TB=39.0 cm³. The hypothesis, that this was a “medium-sized” RT model was backed by considerable preliminary data^(5–8) and efforts to develop an in vitro test that successfully employed the model to predict drug deposition from inhalers when coupled to a breath simulator.^(3,9) Notably, the model (Fig. 1) incorporates MT (including the larynx), and trachea and upper bronchi, TB; the latter section includes the trachea (generation 0) extending through the upper bronchi (generation 3). The MT geometry was based on Xi and Longest,⁽¹⁰⁾ while TB was designed using the classic morphometric data of Yeh and Schum⁽¹¹⁾ and described in the studies of Tian et al.^(12,13) Unlike the symmetrical branching described by Weibel,⁽¹⁴⁾ the TB model contains asymmetric branching angles, tube dimensions, and more realistic angles of inclination to gravity.⁽¹¹⁾ To prevent the model reaching a physical size that would make in vitro testing impractical, the number of TB generations was limited. Nevertheless, the TB airways extend to the approximate point of entry to each lung lobe (right upper, middle, and lower, left upper and lower, or RU, RM, RL, LU, LL, respectively). To generate similar pressures at the model outlets in the experiments, three bifurcations were used in each branching pathway. As a result, the model contained eight outlets with two outlets extending into the LU, LL, and RU lobes and one outlet extending into the RM and RL lobes (Fig. 1).

FIG. 1.

Internal appearance of the average or “medium sized” mouth–throat (MT) according to Xi and Longest⁽¹⁰⁾ with upper airways (tracheobronchial (TB) segment based on Yeh and Schum.⁽¹¹⁾ Scaled hollow models were constructed with an integral mouthpiece to fit Budelin inhalers as described in Figure 2. These were subdivided for drug analysis at various locations using airtight, friction-fit junctions.

FIG. 2.

The small, medium, and large MT-TB models used in the present study. Complete dimensional description is available at http://www.rddonline.com/resources/tools/models.php.⁽³¹⁾ Models were constructed of Accura 60 (3D System, Valencia, CA) using rapid prototyping (Viper SLA, 3D Systems). MT models were made to snap fit (at arrow) on top of companion TB models to form MT-TB_S, MT-TB_M, and MT-TB_L with internal volumes of MT=26.6, 61.6, 96.1, and TB=16.3, 39.0, 61.6 cm³, respectively. The integral Budelin mouthpiece adapter added an additional volume of 5.4 cm³ to all models.

Scaling the airway model to represent small and large humans

Large and small geometry models were scaled to represent the upper and lower 95% confidence limits for the MT and TB regions based on literature reports of their dimensions in different adult populations. The MT geometry of Figure 1 was scaled volumetrically by adding and subtracting 37.8 cm³ to the original “medium-sized” volume (65 cm³) of Xi and Longest.⁽¹⁰⁾ In practice, this was accomplished by multiplying each linear dimension of the model by length scale factors of 1.165 [e.g., (102.8 cm³/65 cm³)^0.333] and 0.748, respectively. The value of 37.8 cm³ was assigned based on 2 times the standard deviation of the average MT volume reported by Burnell et al.⁽⁵⁾ a working assumption that this volume was normally distributed across a mixed-gender adult population and the statistical generalization that the mean value±2SD should embrace >95% of the population. The circular inlet diameter shown in Figure 1 was then adapted to fit the Novolizer mouthpiece. This resulted in the medium MT volume of 61.6 cm³ and adaptor volume of 5.3 cm³ (see Table 1) reported in this study.

Table 1.

Actual Dimensions of the MT Region of Our Medium, Small, and Large Models (Without Mouthpiece Adapter; Fig. 2) in Comparison with the Mean (2^*SD) Data of Burnell et al.⁽⁵⁾

Parameters ^a	MT-TB_M	MT-TB_S	MT-TB_L	Burnell et al.
MT volume (cm³)	61.6	26.6	96.1	60.7 (37.8)
Buccal volume (cm³)	33.9	15.0	52.3	35.1 (28.0)
Angle bcd	340	340	340	350 (10.5)
Angle cda	260	260	260	251(18.6)
A_min (mm²)	408	228	554	293 (328)
A_max (mm²)	614	344	833	1032 (908)
B_min (mm²)	251	140	340	272 (277)
B_max (mm²)	531	297	720	368 (294)
C_min (mm²)	84	47	114	127 (109)
C_max (mm²)	209	117	284	431 (276)
D_min (mm²)	104	58	141	133 (96)
D_max (mm²)	188	105	255	263 (95)
Length ad (mm)	23	17.2	26.8	27.1 (19.9)

Outliers (values in which the small or large models had dimensions outside of Burnell's 95% CI) are bold and underscored. The volumes of the TB regions are not included in this table.

Terms describing dimensions (e.g., B_min, D_max, etc.) are defined in Table 2 and Figure 3 of Burnell et al.⁽⁵⁾

Similarly, the “medium-sized” TB geometry shown in Figure 1 was derived from Yeh and Schum⁽¹¹⁾ after first following their advice to scale the model to a lung volume of interest. Their limiting dimensions describe a geometry corresponding to total lung capacity (TLC) of “standard-sized man” (5.6 liters).⁽¹⁵⁾ We created our “medium-sized” TB model to correspond to a lung volume of 3.5 L (e.g., a lung volume between functional residual capacity and TLC seen typically during inhaler use by males and females.^(2,9,16) Thus, dimensions reported in Yeh and Schum were multiplied by a length scale of 0.855 [(3.5/5.6)^0.333] to create a medium-sized TB geometry; this was paired with the “medium-sized MT” to produce MT-TB_M (Fig. 1). Small and large TB geometries, for pairing with small and large MT models were designed by scaling the medium-sized TB model using the same factors that were used for MT, to represent the 95% confidence limits in the normal adult population. A literature validation of these models was performed by comparing their physical dimensions to data in the literature for normal human adults.

Model construction

Small, medium, and large three-dimensional (3D) airway geometries were constructed in SolidWorks^® computer-assisted design (CAD) software (SolidWorks, Concord, MA) by pairing MT and their companion TB geometries. These designs were constructed as hollow plastic MT-TB models (Accura 60, 3D System, Valencia, CA) using a rapid prototyping process (Viper SLA, 3D Systems). MT models were made to snap fit on top of their companion TB models to form MT-TB_S, MT-TB_M, and MT-TB_L, where subscripts represent small, medium, and large, respectively (Fig. 2); in practice, similar airtight snap-fit junctions can be created elsewhere, when deposition in different regions is of interest.

In vitro deposition testing

Each airway model was installed in an identical custom-built cylindrical Plexiglas housing (internal diameter and height were 13.9 and 12.6 cm, respectively; volume was 1.9 L) with minimal dead space (Fig. 3). To evaluate drug deposition in this setup in a relatively small number of experiments and to compare the resulting in vitro deposition with estimated lung deposition in the clinic, a marketed powder inhaler was selected that possessed reproducible dosing paired with good-quality published deposition data; the latter in a group of trained normal humans inhaling at different, but well-defined flow rates. Novolizer was used as the multidose powder inhaler with reproducible dosing.^(17,18) Budelin Novolizers (budesonide 200 μg) were purchased from the supplier (Meda Pharmaceuticals, Bishops Stortford, UK). The deposition of radiolabeled budesonide from Budelin Novolizers was assessed previously in 13 healthy volunteers by Newman et al.⁽¹⁹⁾ Those authors described the deposition of single 200-μg doses of budesonide from the inhaler at peak inspiratory flow rates (PIFR) of 99±13, 65±3, and 54±7 L·min⁻¹ paired with mean inhalation volumes (V) of 3.13±1.01, 2.96±0.83, and 2.77±0.63 L, respectively (note that the Asta Medica device referred to by Newman et al. is marketed by Meda Pharmaceuticals as Budelin in the EU). Before formally testing Budelin in our models, we confirmed that delivered doses of budesonide from test inhalers fell within USP limits when tested by withdrawal of 4 L air at 83 L/min (corresponding to a 4-kPa pressure drop across the inhaler).⁽²⁰⁾ Single-metered dose deposition testing in each of the MT-TB models was then performed after priming each Novolizer and inserting it into a mouth opening designed and manufactured to fit the inhaler mouthpiece (Fig. 3). Airtight seals between the inhaler mouthpiece, MT, and TB models were maintained in all cases. During simulated inhalations, air was drawn through the inhaler and airway model via a low resistance filter (Pulmoguard II^®, SDI Diagnostics, MA) capable of retaining all aerosolized drug that passed through the model. The filter was connected to a computer programmed breath simulator (ASL 5000, IngMar Medical, Pittsburgh, PA) equipped with digital recording software (LabVIEW^™) to vary and record the rate and volume of air drawn through the setup. In all experiments, the internal surfaces of the MT-TB models were coated with a silicone spray (Dow Corning^® 316 Silicone Release Spray, Dow Corning Corp., Midland, MI) followed by solvent evaporation before each experiment. Powder aerosols were collected as unit doses following each simulated inhalation. Drug retained in the inhaler mouthpiece, MT, TB, and Plexiglas housing plus filter (the latter designated as “peripheral deposition”; P) was recovered and analyzed by high-performance liquid chromatography (HPLC) after each dose, meaning that each experiment or experimental replicate began with equipment that was clean and drug free.

FIG. 3.

Diagram of physical test apparatus used to predict deposition of drugs from dry powder inhalers (DPIs). The externalized MT region is connected to TB through generation 3. Differently scaled versions are shown in Figure 2. Drug retained in the artificial thorax and filter is designated P (pulmonary deposition) following actuation of the breath simulator to withdraw air according to a known flow rate versus time profile.

A randomized experimental design was used to study the in vitro effects of different inspiratory maneuvers and airway geometries. Values for PIFR and V were chosen to correspond to the mean±2SD values used by Newman et al.⁽¹⁹⁾ in the clinic for use in the three airway geometries MT-TB_S, MT-TB_M, and MT-TB_L (Fig. 2). Precise values, selected to mimic the in vivo study, are shown, alongside the in vitro and clinical deposition results,⁽¹⁹⁾ in the Results and Discussion below.

To illustrate and clarify the selection of air flow rate versus time curves used in vitro, profiles chosen for the low flow rate arm of this study are shown in bold in Figure 4.

FIG. 4.

Simulated air flow rate versus time curves representing ∼95% of the range reported by Newman et al.⁽¹⁹⁾ for the low flow rate arm of Budelin testing. Reported values for PIFR and V (54±7 L/min and 2.77±0.97 L, respectively) were processed to yield in vitro flow rates at t_max of 40, 54, and 68 L/min (mean±2SD). Small, mean, and large values for V were calculated similarly (0.83, 2.77, 4.77 L) except V_large was held constant at 4 L, the maximum feasible value for in vitro testing with the breath simulator. Only bold profiles were used in vitro where small, medium, and large volumes were paired with small, medium, and large flow rates to provide an estimate of the expected deposition variations corresponding to 95% of Newman et al.'s⁽¹⁹⁾ clinical population.

In all experiments, the calibrated piston in the breath simulator was programmed to increase air flow rate through the apparatus over an acceleration phase to the chosen value for PIFR at t_max=0.45s according to \documentclass{aastex} \usepackage{amsbsy} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{bm} \usepackage{mathrsfs} \usepackage{pifont} \usepackage{stmaryrd} \usepackage{textcomp} \usepackage{portland, xspace} \usepackage{amsmath, amsxtra} \pagestyle{empty} \DeclareMathSizes {10} {9} {7} {6} \begin{document} \begin{align*}FR(t) = PIFR \times {\rm sin} {\left( \frac {\pi} {2} \frac {t} {t_ {max}} \right)} \tag {1}\end{align*}\end{document}

At t > t_max < 0.6 sec, FR was held constant at PIFR, after which flow rate was decreased to zero according to \documentclass{aastex} \usepackage{amsbsy} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{bm} \usepackage{mathrsfs} \usepackage{pifont} \usepackage{stmaryrd} \usepackage{textcomp} \usepackage{portland, xspace} \usepackage{amsmath, amsxtra} \pagestyle{empty} \DeclareMathSizes {10} {9} {7} {6} \begin{document} \begin{align*}FR(t) = PIFR \times {\rm cos} {\left( \frac {\pi} {2} \times \frac {t - (t_ {max} + 0.15)} {t_ {total} - (t_ {max} + 0.15)} \right)} \tag {2} \end{align*}\end{document}

The value for t_total, the time for completion of an inspiratory maneuver, was varied so that V, the total volume inhaled (or the area under each FR vs. time curve) (Fig. 4), corresponded to either the mean or mean±2SD reported in the clinical deposition study for Budelin by Newman et al.⁽¹⁹⁾

In vitro experiments were randomized with respect to the selection of paired values of PIFR and V (see Table 3) and the choice of model. Each in vitro experiment was performed five times and the average (±SD) and absolute range of data for mass of drug retained in different sections of the apparatus determined by chemical analysis.

Analysis

Budesonide was recovered from the different sections of the apparatus with a mixture of 30 parts 0.1% v/v acetic acid in water plus 70 parts methanol by volume. This solvent system was the same as the mobile phase used for budesonide analysis by HPLC. Drug amounts were calculated from the products of concentration×volume. Drug concentrations were quantified on a Waters HPLC (Waters 2690 separations module, Milford, MA) using an isocratic reverse phase separation system that employed a 3.5 μm, 4.6×100 mm C-18 column (Symmetry^™, Milford, MA), a mobile phase flow rate of 1.0 mL/min, an injection volume of 100 μL, and UV detection (Waters 2996 PDA, Waters Corporation) at 245 nm. Accuracy and imprecision were assessed according to ICH⁽²¹⁾ and found to be 97.1 and 0.56% in the linearity range 0.1–10 μg/mL.

IVIVC

The results for % drug deposition in each in vitro region were compared to those from in vivo gamma scintigraphic studies on Budelin Novolizers in normal adult volunteers of both genders by Newman et al.⁽¹⁹⁾ Inhalation profiles were chosen in vitro to represent both the average and extreme inhalation maneuvers used in the clinic to follow drug deposition of single 200-μg doses of radiolabeled budesonide.

Results and Discussion

Scaled physical models of the upper airways

A literature validation of the small-, medium-, and large-sized MT-TB models was performed by comparing their physical dimensions to data reported for normal human adults of both genders. Table 1 shows the internal dimensions of the MT used in this study alongside the means and relative standard deviations reported by Burnell et al.,⁽⁵⁾ following their analysis of the MRI scans of 20 adult volunteers using four different inhalers. Although small differences in the dimensions of the different regions were noted, the internal areas, volumes and angles of the MT_M used in this study were close to the mean values reported by Burnell et al.⁵ and the range of dimensions in our models fell within the reported 95% confidence intervals with the exception of two outliers (Table 1). Although the starting point for the “medium-sized” model employed here was a simplified elliptical version of the MT taken from a cast of an average-sized human,⁽¹⁰⁾ Burnell et al.⁽⁵⁾ used MRI scanning from subjects actually using inhalers. Furthermore, they reported that the single most influential variable affecting drug retention in their models was throat model volume. That feature (mean and variance in volume) showed close agreement to the scaled models described in the present study in large part because our volumetric dimensions were scaled to broadly correspond to the confidence limits described by Burnell et al.⁽⁵⁾ Until very recently Burnell's models were not publically available, although this situation has now been partly remedied by the publication of the details of a single model, purported to be one of those used in the study by Olsson et al.^(4,22) The geometric characteristics of the physical models described here can be downloaded as shown in Figure 2. Notably also, the MT models described by O'Callaghan's⁽²³⁾ group show a mean volume of 61.86 cm³, comparable to the value of 61.6 cm³ for our medium-sized model Thus, it appeared that our MT models were a reasonably accurate representation of the upper airway encountered by inhaled aerosols across both genders in a normal adult population, although in future we intend to perform more systemic comparison of our MT_M model dimensions with that made available recently by the industry consortium.⁽²²⁾ Notably, the length scale factors of 1.165 [e.g., (102.8 cm³/65 cm³)^0.333] and 0.748, respectively that we have used here differ from those advocated by Finlay et al.⁽²⁴⁾ of 1.3 and 0.7. Furthermore, although our models presently fail to account for dynamic changes that result from inhalation effort,^(3,25) it is not yet known how much such changes in volume actually affect drug deposition in vivo.

Table 2 shows the internal luminal dimensions of TB used in this study, alongside the values and variations reported in the literature for adults of both genders. All dimensions are for normal adults. As described earlier, the “medium-sized” TB geometry was scaled from Yeh and Schum⁽¹¹⁾ to a lung volume=3.5 L and paired with the “medium-sized MT” to create MT-TB_M. Small and large TB geometries, for pairing with the small and large MT models described above, were derived from TB_M using the same factors and methods used to produce the scaled versions of the MT. This resulted in models that showed close agreement with dimensions for normal adults in the literature; in particular, dimensional variations reported as two standard deviations by Montaudon et al.⁽⁷⁾ agreed well with luminal diameters of our MT-TB_L and MT-TB_S models (Table 2).

Table 2.

Luminal Diameters (mm) for Generation 0 Through 3 (Figs. 1 and 2) of MT-TB_M,S,L (Shown in Bold Type) in Comparison with Values Reported in the Literature

Generation	0 (trachea)	1	2	3
MT-TB_M	17.2	13.4	9.92	7.2
Montaudon: mean⁽⁷⁾	18.5	13.9	10.4	6.8
Nikiforov: mean⁽³²⁾	—	14.2	9.10	6.3
Weibel: mean⁽³³⁾	18	12	8.3	5.6
Horsfield and Cumming: mean⁽³⁴⁾	16	12.0(R), 11.1(L)	—	—
Raabe et al.: mean⁽³⁵⁾	20.1,	17.5(R), 13.8(L)	—	—
Raabe et al.: mean⁽³⁵⁾	23.5	18.5(R), 14.5(L)	—	—
MT-TB_L	20.0	15.6	11.5	8.4
Montaudon: mean+2SD⁽⁷⁾	21.7	16.7	12.2	8.4
MT-TB_S	12.9	10.0	7.4	5.4
Montaudon: mean−2SD⁽⁷⁾	15.3	11.1	8.6	5.2

R, L, right and left lung; SD, standard deviation.

We reviewed several additional dimensions in a similar way to that described above and in Table 2. For example, we compared values for the cross-sectional areas and the length and variability of different airway generations in our models to those from the airways of adults in the literature.^(6,7,26) Although dimensional values taken from the literature were comparable to those of our models, it became apparent that this approach was data heavy and unrealistic. Furthermore, because the literature used different methods and our models were simplified in several respects (e.g., tracheal rings are omitted and circular connecting segments are assumed in our models; Figs. 1 and 2) for the purpose of an overall comparison, internal volumetric comparisons, like that for MT, were thought to be best. Internal tracheal volumes for our small, medium, and large TB models were 9.5, 22.7, and 36.0 cm³, respectively. These were very similar to the mean tracheal volume (±2SD) as reported by Leader et al.⁽⁸⁾ of 22.6 (7.2 to 38.0) cm³. However, because luminal volume in vivo was difficult to define precisely (length depends on the way that the position of bifurcation is defined and variations in diameter and cross sectional shape are reported to occur with length and inspiratory flow rate,^(27,28) even these comparisons are challenging. Thus, because our aim was to relate variations in regional drug deposition to airway geometries seen across a population of normal adults, we hypothesized that our models were valid for this purpose. To test that hypothesis we built the models and sought to determine whether they were able to predict clinical variations seen in drug deposition. In short, only if the models described in Figure 2 failed to predict clinical deposition data did we plan to incorporate further physical details. If the models proved to be predictive, however, we planned to use them and vary, test, and report the effect of certain usage variables such as inhaler insertion angle, depth of insertion, etc.

In vitro deposition testing and IVIVC

Although several in vivo imaging methods are possible to define the deposition of radiolabeled drug aerosols,⁽²⁹⁾ two-dimensional gamma scintigraphy has become the most popular technique for studying this topic in vivo.⁽³⁰⁾ Although efforts continue to standardize the details of the method,^(29,30) Newman and his colleagues⁽¹⁹⁾ have led this field for some time and are an accepted source of inhaler scintigraphy data. Accordingly, we selected their study of Novolizer as a data-rich source of drug deposition information with which to compare our in vitro results; although many aerosol deposition studies can be criticized for providing only meager details of the method used, theirs is the one that offers some important details. Their study of budesonide deposition contained descriptions of the inspiratory maneuvers used by 13 trained adult volunteers. Each volunteer was trained to inhale at high, medium, and low flow rates through Novolizer containing ^99mTc labeled budesonide in a cross over study.⁽¹⁹⁾ Comparative in vitro particle size analyses showed that radioactive counting and drug assay produced statistically comparable data, showing that the ^99mTc label was a valid drug marker and that the labeling process did not perturb aerosol emissions from the inhaler.⁽¹⁹⁾ Total and regional lung deposition, oropharyngeal deposition, and inhaler mouthpiece retention were quantified as % total radioactive counts, following standard corrections for quenching and radioactive decay. Because of the overlay of the esophagus and the trachea in 2D scintigraphy, in vivo lung deposition is often expressed without including the trachea as part of the lung, and this was the method used in the scintigraphic evaluation of Budelin⁽¹⁹⁾ [S. P. Newman, personal communication, 2009]. Because of this anatomical inaccuracy, we assayed tracheal deposition separately in vitro and included it with drug deposited in the mouth–throat (Table 3).

Table 3.

In Vitro and In Vivo Results for % Budesonide (Mean±SD; n=5) Deposition from Budelin Novolizers in Small, Medium, and Large MT-TB Models Using Inhalation Profiles Selected to Represent the Mean and Extreme Values for PIFR and V of Newman et al. ¹⁹

MT-TB model ^a	Flow rate (L/min) ^b	Volume (L)^b	Device ^c	MT ^d	TLD (TB+P-trachea) ^e	In vivo^f
High flow rate
Small	73	1.11	48.36 (4.27)	40.08 (5.98)	11.55 (1.97) (9.72–13.99)	Lower limit=9.4
Medium	99	3.13	14.14 (2.47)	55.95 (2.92)	29.92 (1.35) (29.02–31.78)	Mean 32.1
Large	125	4	13.26 (1.65)	48.58 (2.83)	38.20 (1.73) (35.94–40.44)	Upper limit=41
Moderate flow rate
Small	59	1.3	49.48 (4.27)	40.09 (5.51)	10.43 (1.92) (8.03–13.15)	Lower limit=12.1
Medium	65	2.96	16.08 (4.97)	62.24 (5.84)	21.68 (1.32) (20.38–23.81)	Mean 25
Large	71	4	14.35 (1.81)	58.88 (1.95)	26.71 (1.93) (23.48–28.69)	Upper limit=37.4
Low flow rate
Small	40	0.83	55.43 (11.37)	39.12 (12.06)	5.45 (1.23) (4.82–7.51)	Lower limit=8.8
Medium	54	2.77	23.45 (5.60)	61.08 (4.93)	15.52 (1.96) (13.88–18.61)	Mean 19.9
Large	68	4	12.65 (1.33)	62.28 (1.41)	25.07 (2.46) (22.86– 27.97)	Upper limit=26.6

MT-TB_S, MT-TB_M, or MT-TB_L.

Mean±2SD reported by Newman as shown in Figure 4.

Mouthpiece and dosing chamber/air classifier.

MT includes trachea.

Total lung dose and experimental range (bold=in vitro values differ from clinical estimate).

% deposition values from Table 3 in Newman et al.

Results for total lung deposition in vivo and in vitro are compared head-to-head in Table 3 and Figure 5. Clearly, the in vitro results for TLD [the drug recovered from TB (minus trachea) and P (artificial thorax and filter)] were associated with the inhaler test conditions as well as the MT-TB models chosen to span 95% of the range of airway geometries. Although some of the clinical values fell outside of the in vitro range (shown in bold in Table 3), the overall similarity between lung deposition values reported by Newman et al.⁽¹⁹⁾ and the in vitro estimates was remarkable. With the exception of MT-TB_L at moderate flow and 4 L volume, low, mean, and high deposition values predicted in each model were close to the clinical results throughout,⁽¹⁹⁾ implying that our in vitro method produced meaningful results.

FIG. 5.

The total lung dose (TLD; Table 3) from Budelin following in vitro testing at the mean and extremes of flow and volume in each of the three MT-TB models following fast, moderate, and slow inhalation shown in comparison to median in vivo values reported by Newman et al.⁽¹⁹⁾ Error bars show the entire range in all cases.

The agreement shown in Table 3 between in vivo and in vitro results also appeared to support the way in which in vitro testing extremes were chosen in the present study, to minimize the number of in vitro tests required while still reflecting and predicting the overall deposition variations seen in the clinic. In practice, we first selected a powder inhaler that emptied reliably,^(2,9) avoiding the need to deal with dosage form variability as a significant source of additional variance in the present study. Then, we studied the effect of three separate test conditions and three major variables (model geometry, PIFR, and V). However, out of a possible 3³ experimental matrix, we studied the subset of cases described in Table 3. To reduce the size of the matrix we hypothesized that in each of Newman's crossover study arms [in each arm, the same 13 adults (with different geometries) were instructed to inhale at low, moderate, or high flow rates], upper and lower flow rate extremes (e.g., 40 and 68 L/min at the low flow condition) (Fig. 4) were coupled with the extreme small and large lung volumes (e.g., 0.83 and 4.77 L) (Fig. 4). We also coupled small, medium, and large profiles to S, M, and L models; the assumption, that the extremes of each inhalation maneuver, studied over the range of geometries seen across a normal male and female adult population, should describe the vast bulk of the variation seen in drug deposition in the clinic. The agreement between the in vitro results and the variations seen in budesonide deposition are shown most dramatically in Figure 5. In vitro variance in TLD for a given model under a given set of test conditions was small, reflecting the reproducibility of this inhaler when tested in vitro. However, the mean and the range of results in vivo was entirely predictable when the tested variations were created by coupling different breathing maneuvers and different airway geometries based on their ranges displayed in this mixed gender adult population; this, despite the in vitro models' inability to account for dynamic changes resulting from inhalation effort.^(3,25) Our findings were also consistant with the results reported by Olsson et al.⁽⁴⁾ in their IVIVC for inhaled budesonide. In that study, total drug dose in vitro was evaluated pharmacokinetically, after oral absorption was prevented using charcoal-block technique.⁽⁴⁾

Budesonide retention in the (Novolizer) device, MT (including trachea) and TB+P (TLD) for all in vitro test conditions (Table 3) is shown in Figure 6 in comparison with the in vivo results reported by Newman et al.⁽¹⁹⁾ Most deposition was either in the device, MT, or the peripheral in vitro compartment (Plexiglass container and filter). Although this statement was true for all models, and TB deposition from Budelin (in the absence of the model trachea) was <1% of the total recovered dose for all tested inhalation profiles, results for other inhalers⁽²⁾ to be reported elsewhere, shows that TB deposition in vitro depends on the choice of inhaler, drug, and formulation. Notably, and consistent with our in vitro results for Budelin, Newman⁽¹⁹⁾ reported significant peripheral deposition in vivo and no change in the central/peripheral distribution ratio as a function of slow, moderate, or fast inhalation. One significant disagreement between the in vitro and in vivo results in Figure 6 appeared to be for inhaler device retention in the case of the small model at low flow and volume extremes (Device; crosshatched bars; Fig. 6); deposition or retention in the inhaler in vitro appeared to overestimate the in vivo determination.⁽¹⁹⁾ Our current explanation for this discrepancy for Novolizer, a powder inhaler whose emptying is known to be affected by volume and flow rate (low volumes and low flow results in incomplete emptying)^(2,17) relates to the necessary but unrealistic test conditions used during in vivo investigations (radiolabeled powder is loaded and emptied, dose by dose to minimize risk); furthermore, validation of radiolabeling in the in vivo study was only performed at the “medium inhalation flow condition”; in short, we believe that our in vitro determinations (Fig. 6) for the commercial product are broadly correct, because these involve the device's self-metering capabilities, with its cartridge-packed powder reservoir in place.⁽¹⁷⁾

FIG. 6.

In vitro and in vivo deposition results for Budelin across the test conditions listed in Table 3. (A) Fast inhalation, (B) moderate inhalation, and (C) slow inhalation. MT includes the trachea; TLD excludes the trachea in vitro and in vivo.

In conclusion, we have described new physical models that, when partnered with appropriate inhalation flow rate versus time profiles provide excellent predictions of the median and range of lung deposition results in vivo for a trained normal population using a marketed powder inhaler. It is clear from the results that the bulk of in vivo variance in deposition across a trained population of normal volunteers was explained by variations in airway morphology and the way that the inhalation maneuver was performed. In the case of Budelin Novolizer, there was little additional variance in drug delivery due to the device or formulation and the present physical airway models appeared to offer the means to predict the in vivo results; in short, the reported IVIVC appears to be valid for this inhaler. With the goal of adding further weight to these test methods, we will report the results for other powder inhalers, for which in vivo results are also available in the literature, in a subsequent publication.

Footnotes

Acknowledgments

Guoguang Su and Ross Walenga (VCU School of Engineering) are gratefully acknowledged for their assistance in constructing and prototyping the MT-TB geometries used in this study.

Author Disclosure Statement

The authors are faculty and students of Virginia Commonwealth University. The work was supported by funds from the Medical College of Virginia Foundation. Supplies and inhalers were purchased from commercial sources. No conflicts of interest exist.

References

Borgstrom

, Olsson

, Thorsson

. Degree of throat deposition can explain the variability in lung deposition of inhaled drugs. J Aerosol Med, 2006; 19:473–483.

Byron

, Delvadia

, Longest

, Hindle

. Stepping into the trachea with realistic physical models: uncertainties in regional drug deposition from powder inhalers. Dalby

, Byron

, Peart

, Suman

, Farr

, Young

. Proceedings of Respiratory Drug Delivery 2010, I. Davis Healthcare International Publishing: River Grove, IL, 215–224. 2010. www.rddonline.com

Byron

, Hindle

, Lange

, Longest

, McRobbie

, Oldham

, Olsson

, Thiel

, Wachtel

, Finlay

. In vivo-in vitro correlations: predicting pulmonary drug deposition from pharmaceutical aerosols. J Aerosol Med Pulm Drug Deliv, 2010; 23,Suppl 2:S59–S69.

Olsson

, Borgstrom

, Svensson

, Lundback

. Modeling oropharyngeal cast deposition to predict lung delivery from powder inhalers. Dalby

, Byron

, Peart

, Suman

, Farr

, Young

. Proceedings of Respiratory Drug Delivery 2008, I. Davis Healthcare International Publishing: River Grove, IL, 197–206. 2008. www.rddonline.com

Burnell

, Asking

, Borgstrom

, Nichols

, Olsson

, Prime

, Shrubb

. Studies of the human oropharyngeal airspaces using magnetic resonance imaging IV—the oropharyngeal retention effect for four inhalation delivery systems. J Aerosol Med, 2007; 20:269–281.

Cherng

C-H

, Wong

C-S

, Hsu

C-H

, Ho

S-T

. Airway length in adults: estimation of the optimal endotracheal tube length for orotracheal intubation. J Clin Anesth, 2002; 14:271–274.

Montaudon

, Desbarats

, Berger

, De Dietrich

, Marthan

, Laurent

. Assessment of bronchial wall thickness and lumen diameter in human adults using multi-detector computed tomography: comparison with theoretical models. J Anat, 2007; 211:579–588.

Leader

, Rogers

, Fuhrman

, Sciurba

, Zheng

, Thompson

, Weissfeld

, Golla

, Gur

. Size and morphology of the trachea before and after lung volume reduction surgery. AJR Am J Roentgenol, 2004; 183:315–321.

Delvadia

, Byron

, Longest

, Hindle

. In vitro prediction of regional drug deposition from dry powder inhalers. Dalby

, Byron

, Peart

, Suman

, Farr

, Young

. Proceedings of Respiratory Drug Delivery 2010, III. Davis Healthcare International Publishing: River Grove, IL, 907–912. 2010. www.rddonline.com

10.

, Longest

. Transport and deposition of micro-aerosols in realistic and simplified models of the oral airway. Ann Biomed Eng, 2007; 35:560–581.

11.

Yeh

, Schum

. Models of human lung airways and their application to inhaled particle deposition. Bull Math Biol, 1980; 42:461–480.

12.

Tian

, Longest

, Su

, Hindle

. Characterization of respiratory drug delivery with enhanced condensational growth (ECG) using an individual path model of the entire tracheobronchial airways. Ann Biomed Eng, 2011; 39:1136–1153.

13.

Tian

, Longest

, Su

, Walenga

, Hindle

. Development of a stochastic individual path (SIP) model for predicting the tracheobronchial deposition of pharmaceutical aerosols: effects of transient inhalation, sampling the airways. J Aerosol Sci.(in press).

14.

Weibel

. Principles and methods for the morphometric study of the lung and other organs. Lab Invest, 1963; 12:131–155.

15.

Chapter 3 Physiological data for reference man. Ann ICRP/ICRP Publ, 1975; 23:335–365.

16.

Clark

, Hollingworth

. The relationship between powder inhaler resistance and peak inspiratory conditions in healthy volunteers—implications for in vitro testing. J Aerosol Med, 1993; 6:99–110.

17.

Weda

, Zanen

, Boer

AHd

, Barends

, Frijlink

. An investigation into the predictive value of cascade impactor results for side effects of inhaled salbutamol. Int J Pharm, 2004; 287:79–87.

18.

Fenton

, Keating

, Plosker

. Novolizer: a multidose dry powder inhaler. Drugs, 2003; 63:2437–2445.

19.

Newman

, Pitcairn

, Hirst

, Bacon

, O'Keefe

, Reiners

, Hermann

. Scintigraphic comparison of budesonide deposition from two dry powder inhalers. Eur Respir J, 2000; 16:178–183.

20.

USP: General Chapter <601> Aerosols, Nasal Sprays, Metered Dose Inhalers, and Dry Powder Inhalers. United States Pharmacopeia—National Formulary. United States Convention: Washington, DC, 2009.

21.

International Committee for Harmonization Q2B: Validation of analytical procedures: methodology. http://www.fda.gov/cder/guidance/index.htm. 2010 06 03

22.

Anatomical oropharyngeal model. http://www.isam.org. 2011 09 29.

23.

Ehtezazi

, Saleem

, Shrubb

, Allanson

, Jenkinson

, O'Callaghan

. The interaction between the oropharyngeal geometry and aerosols via pressurised metered dose inhalers. Pharm Res, 2010; 27:175–186.

24.

Finlay

, Golshahi

, Noga

, Flores-Mir

. Choosing 3-D mouth–throat dimensions: a rational merging of medical imaging, aerodynamics. Dalby

, Byron

, Peart

, Suman

, Farr

, Young

. Proceedings of Respiratory Drug Delivery 2010, I. Davis Healthcare International Publishing: River Grove, IL, 185–194. 2010. www.rddonline.com

25.

Ehtezazi

, Horsfield

, Barry

, O'Callaghan

. Dynamic change of the upper airway during inhalation via aerosol delivery devices. J Aerosol Med, 2004; 17:325–334.

26.

Vock

, Spiegel

, Fram

, Effmann

. CT Assessment of the adult intrathoracic cross section of the trachea. J Comput Assist Tomomgr, 1984; 8:1076–1082.

27.

Osmanliev

, Bowley

, Hunter

, Pride

. Relation between tracheal size and forced expiratory volume in one second in young men. Am Rev Respir Dis, 1982; 126:179–182.

28.

Fouke

, Pimmel

, Bromberg

. Direct dynamic measurements of tracheal diameter. J Appl Physiol, 1981; 51:767–771.

29.

Newman

. In vivo assessment of pulmonary drug delivery. Newman

, Anderson

, Byron

, Dalby

, Peart

. Respiratory Drug Delivery: Essential Theory and Practice. RDD Online/Virginia Commonwealth University: Richmond, VA, 97–133. 2009.

30.

Scheuch

, Bennett

, Borgström

, Clark

, Dalby

, Dolovich

, Fleming

, Gehr

, Gonda

, O'Callaghan

, Taylor

, Newman

. Deposition, imaging, and clearance: what remains to be done? J Aerosol Med Pulm Drug Deliv, 2010; 23:S39–S57.

31.

Longest

. Upper Airway Models. 2011. http://www.rddonline.com/resources/tools/models.php. 17 2011 August .

32.

Nikiforov

, Schlesinger

. Morphometric variability of the human upper bronchial tree. Respir Physiol, 1985; 59:289–299.

33.

Weibel

. Morphometrics of the lung. Fenn

, Rahn

. Handbook of Physiology. Section 3. Respiration: American Physiological Society: Washington, DC, 285–307. 1964.

34.

Horsfield

, Cumming

. Morphology of the bronchial tree in man. J Appl Physiol, 1968; 24:373–383.

35.

Tracheobronchial Geometry: human, dog, rat, hamster: Lovelace Foundation for Medical Education and Research. 1976. http://mae.ucdavis.edu/wexler/lungs/LF53-Raabe/. 2010 10 12.