Bridging the Gap Between Science and Clinical Efficacy: Physiology, Imaging, and Modeling of Aerosols in the Lung

Abstract

Development of a new drug for the treatment of lung disease is a complex and time consuming process involving numerous disciplines of basic and applied sciences. During the 2015 Congress of the International Society for Aerosols in Medicine, a group of experts including aerosol scientists, physiologists, modelers, imagers, and clinicians participated in a workshop aiming at bridging the gap between basic research and clinical efficacy of inhaled drugs. This publication summarizes the current consensus on the topic. It begins with a short description of basic concepts of aerosol transport and a discussion on targeting strategies of inhaled aerosols to the lungs. It is followed by a description of both computational and biological lung models, and the use of imaging techniques to determine aerosol deposition distribution (ADD) in the lung. Finally, the importance of ADD to clinical efficacy is discussed. Several gaps were identified between basic science and clinical efficacy. One gap between scientific research aimed at predicting, controlling, and measuring ADD and the clinical use of inhaled aerosols is the considerable challenge of obtaining, in a single study, accurate information describing the optimal lung regions to be targeted, the effectiveness of targeting determined from ADD, and some measure of the drug's effectiveness. Other identified gaps were the language and methodology barriers that exist among disciplines, along with the significant regulatory hurdles that need to be overcome for novel drugs and/or therapies to reach the marketplace and benefit the patient. Despite these gaps, much progress has been made in recent years to improve clinical efficacy of inhaled drugs. Also, the recent efforts by many funding agencies and industry to support multidisciplinary networks including basic science researchers, R&D scientists, and clinicians will go a long way to further reduce the gap between science and clinical efficacy.

Introduction

Aerosol inhalation is a well-established means of delivering drugs to the lungs of patients. For treatment of lung diseases, it requires a smaller dose than oral or intravenous administration; it minimizes systemic effects and has a more rapid onset of action than through other delivery routes. Aerosol inhalation has therefore the potential to maximize therapeutic effects and minimize side effects. However, there are several factors to consider for optimal treatment of the lungs.

These include aerosol characteristics, breathing patterns, geometrical factors (i.e., lung morphology), disease state, pharmacokinetics, and drug–cell interactions (pharmacodynamics) to name a few. The aerosol characteristics are mainly determined by the drug formulation and the inhalation device. All other aspects are—at least to some degree—patient-specific and may depend on age, sex, type of disease, and/or severity of disease.^(1,2) This has stimulated recent efforts towards more personalized therapeutic approaches aiming at optimized pulmonary drug delivery and selection of the most effective type of drug for any given patient.⁽³⁾

Delivering inhaled drugs preferentially to the diseased site of the lung is the prime objective of the science of pulmonary aerosol deposition. While the treatment of asthmatic patients requires drug delivery to the bronchial and bronchiolar airways,⁽⁴⁾ emphysema patients may benefit from drugs delivered to the alveolar region. Hence, not only the total pulmonary drug dose, but also the regional distribution, serial (proximal vs. peripheral) and parallel (among lobes or among apical, middle, and basal regions), of the lung-deposited aerosol, henceforth referred to as aerosol deposition distribution (ADD), is a key factor for the clinical success of an inhalation therapy. Optimized ADD is expected to increase drug effectiveness and reduce drug cost, side effects, and treatment times.

During the 2015 Congress of the International Society for Aerosols in Medicine (ISAM), a group of experts including aerosol scientists, physiologists, modelers, imagers, and clinicians participated in a workshop aimed at bridging the gap between basic research and clinical efficacy of inhaled drugs. This publication summarizes the current consensus on the topic as it was discussed at the workshop.

Basics of Aerosol Transport and Deposition in the Lungs

In the human lung, the airways form a dichotomous tree where each airway gives rise to two daughter branches. Because the airway tree needs to fill a space matching the shape of the chest cavity, the number of generations needed to reach the alveolar sacs from the trachea is not the same throughout the lungs but varies between 18 and 30.⁽⁵⁾ Even though the airways become gradually shorter and narrower with each generation, the increasing number of airways with each generation number leads to a large increase in total airway cross-section towards the lung periphery. An important consequence in terms of aerosol transport is that this cross-sectional increase results in a large decrease in airflow velocity from the first few airway generations down to the lung periphery. Accordingly, even with the decrease in airway length from proximal to peripheral airways, the residence time in each generation increases with each generation (Fig. 1).⁽⁶⁾

FIG. 1.

Mean airway velocity (solid line) and residence time (dashed line) in the airways as a function of generations of an idealized dichotomous branching lung model (Weibel model A⁽⁶⁾). As a result, the most relevant mechanism for aerosol deposition onto the lung epithelium changes from impaction in the upper to sedimentation and diffusion in the lower airway generations (see text for details). Data were calculated for an inspiratory flow rate of 500 mL/s.

The deposition of inhaled particles occurs mainly by the mechanisms of inertial impaction, gravitational sedimentation, and Brownian diffusion, and to a lesser extent by interception, turbulent mixing, and electrostatic precipitation. A detailed description of these mechanisms can be found elsewhere.^(1,7) Briefly, inertial impaction results from the inability of particles to follow sudden changes in gas flow direction and is a primary mechanism for particles larger than 5 μm; gravitational sedimentation results from the settling of the particles under the action of gravity and significantly affects the deposition of 1–8 μm diameter particles; and Brownian diffusion is the dominant mechanism of deposition for particles less than 0.5 μm in diameter.

As inertial impaction is a velocity-dependent mechanism, deposition by this mechanism occurs preferentially in the first few generations of airways where gas and particle velocity are high. Inversely, deposition by gravitational sedimentation and Brownian diffusion, which are time-dependent mechanisms, are most efficient in the lung periphery where airspace size is small and residence time high (Fig. 1).

Although often neglected in aerosol deposition analyses, turbulent flows, electrical charges, and particle shape can also significantly affect overall deposition. Turbulent flows generally occur at the glottic constriction and persist over several airway generations. They can be described in terms of mean flow over which are superimposed fluctuations. Deposition due to turbulent mixing results from the flow fluctuations as opposed to deposition by inertial impaction that is affected by the mean flow. Space charge forces (the repulsion between charged particles in an aerosol cloud) and/or image charge forces (the attraction between a charged particle and its image charge on an electrically conducting surface) lead to deposition by electrostatic precipitation. Finally, interception refers to particles coming into contact with the airway wall because of their shape and size. While negligible for spherical particles such as liquid droplets created by nebulizers, this mechanism becomes important for elongated particles such as fibers.

Targeting Approaches

An impressive body of work, spanning several decades, has been conducted to describe and predict regional deposition patterns (or ADD) for inhaled aerosols. These empirical, analytical, and computational models build on our fundamental knowledge describing various mechanisms through which aerosol particles come to deposit within upper, central, and/or peripheral airways of the respiratory tract. Naturally, such fundamental understanding of deposition mechanisms has been explored extensively for the purpose of optimizing delivery of pharmaceutical aerosols to the lungs, and to target sub-regions within the lungs, most often the central or peripheral airways.

Many improved aerosol formulations and delivery devices currently in development, or recently arrived on the market, offer much greater efficiency of delivery to the lungs than do traditional pressurized metered dose inhalers (pMDIs), dry powder inhalers (DPIs), and nebulizers. These improvements are the result of significant investment in research and development programs enabling the application of complex and interdisciplinary science, including aerosol mechanics underlying regional deposition patterns. However, as discussed below, there exist some important caveats that have limited the implementation of targeting approaches in commercial aerosol drug delivery systems and into clinical practice.

The accurate description of aerosol deposition processes through mechanistic in silico and experimental models permits the influence of various parameters on regional deposition patterns to be explored. Two parameters for which effects on regional deposition pattern are particularly well established are the particle aerodynamic diameter and inhalation maneuver. Manipulation of these two parameters for the purpose of regional targeting is discussed below, along with the potential use of aerosol bolus and of heliox (helium–oxygen gas mixture) for targeting peripheral and poorly-ventilated lung regions. Various additional targeting approaches, including use of intracorporeal nebulizing catheters,^(8,9) nonspherical particles,⁽¹⁰⁾ and application of external magnetic fields^(11–13) are not discussed herein, but have been recently reviewed elsewhere.⁽¹⁴⁾

Total lung deposition

Administering therapeutic agents directly to the lungs through inhalation is in itself an example of targeted drug delivery. Improved lung targeting may be achieved by reducing upper airways deposition, such that a greater fraction of the emitted dose reaches the lungs. Reducing upper airways deposition also reduces intersubject and intrasubject variability in lung dosing.⁽¹⁵⁾ Conventional pMDIs and DPIs typically deposit drug in high fraction in the mouth–throat airways, resulting in lung doses well below 50% of the nominal dose, with considerable variability. In contrast, introduction of new formulation and delivery technologies over the past decade has enabled more efficient targeting of the lungs from single-breath inhalers.^(16–18)

Existing empirical correlations to predict mouth–throat deposition (as described below) are not applicable to pMDI aerosols that enter the oral cavity as a plume of rapidly decelerating and evaporating propellant droplets. For pMDIs, the influence of inspiratory flow rate on mouth–throat and lung deposition has been shown to vary from one device/formulation to another.⁽¹⁹⁾ Empirical correlations must also be applied carefully for predicting mouth–throat deposition for DPIs, owing to flow rate dependence of fluidization and deagglomeration of many DPI aerosols, as well as to effects of unsteady inhalation flow rate on mouth–throat deposition.⁽²⁰⁾

The complex processes of aerosol formation and behavior within and immediately following emission from pMDIs and DPIs also render numerical simulation difficult and computationally demanding. For both pMDIs and DPIs, it is preferable to perform in vitro experiments in realistic or idealized mouth-throat geometries⁽²¹⁾ in order to estimate in vivo lung doses for a given device/formulation combination. For nebulizers, where fine droplets are well entrained in the inhaled airflow, deposition correlations and models tend to be more reliable,⁽²²⁾ provided hygroscopic behavior of the nebulized aerosol is included in the model, or may be safely neglected.⁽²³⁾

Regional lung deposition

Many inhaled therapies are developed under the assumption that, in addition to the total dose delivered to the lungs, the regional deposition pattern within the lungs is an important determinant of efficacy. For example, both asthma and COPD have been characterized by inflammation affecting the small, peripheral airways; therefore, targeting deposition of inhaled corticosteroids to small airways may improve treatment.⁽⁴⁾ In other circumstances, where the inhalation route is used for systemic delivery, targeting deposition to the alveolar lung region may be used to enhance drug absorption efficiency.

As discussed below, regional targeting that can be achieved using aerodynamic particle size alone is limited, with significant overlap in deposition efficiencies between lung regions. Furthermore, for any given particle size, regional deposition will additionally be influenced by an individual patient's lung volume, airway geometry, disease state, and breathing pattern (or inhalation maneuver). Accordingly, strategies to target central or peripheral lung regions with precision will generally need to incorporate some form of control over patient breathing in addition to particle size.

Parameters affecting ADD

Aerodynamic size

Inertial impaction is the primary mechanism of aerosol deposition in the upper and central airways (Fig. 1). Accordingly, those factors that influence impaction affect the fraction of aerosol that deposits in the upper and central airways versus that which penetrates to the peripheral lung regions (typically to deposit by sedimentation). The aerodynamic particle size is a key determinant of inertial impaction efficiency; targeting deposition to the central or peripheral airways using larger or smaller aerodynamic particle sizes, respectively, is well entrenched in aerosol medicine. Nevertheless, the refinement in regional targeting that can be achieved through variation of aerodynamic size alone is coarse, with considerable overlap in deposition between upper and central airways, and/or central and peripheral airways for a given aerodynamic size.⁽²⁴⁾

In addition, pharmaceutical aerosols are generally polydisperse, spanning a range of aerodynamic sizes, which tends to further spread deposition between upper, central, and peripheral airways. Variability in regional deposition efficiencies between patients with different underlying airway geometries and varying manifestation of disease, potentially inhaling at different flow rates, will also influence regional deposition patterns at a given aerodynamic particle size. Accordingly, while aerodynamic size is clearly a critical factor in determining regional deposition, targeting deposition by aerodynamic size alone has limitations, and may jeopardize the clinical success of inhaled therapies with steep dose-response relationship or narrow therapeutic index.

Controlled ventilation and bolus aerosol delivery

Along with particle size, the inhalation flow rate has long been identified as a primary determinant of inertial impaction in the upper and central airways. Instructing patients to inhale at low flow rates has been shown to shift deposition from more proximal to more distal airways.^(25–27) Unfortunately, DPIs and pMDIs use of low inhalation flow rates to reduce inertial impaction is typically compromised by the requirement of reasonably high flow rates needed to fluidize and disaggregate dry powders,^(28,29) or by complex interaction between high-speed droplet clouds emitted from pMDIs with enveloping air flow.^(19,30,31) Independent of the inhalation flow rate, patients are commonly instructed to perform a breath-hold for several seconds after inhaling, so as to allow aerosols additional time to settle under the influence of gravity in peripheral airways.⁽³²⁾

For nebulizers used to deliver aerosols during controlled tidal breathing, there exists greater opportunity to manipulate regional deposition patterns by varying the breathing pattern and employing the aerosol bolus technique. Commercial devices have been developed to guide or control patient breathing through nebulizers.^(33,34) It should be understood that both central and peripheral deposition fractions, as well as their uniformity among parallel lung regions, will depend on an individual subject's lung volume and airway geometry, thus the effectiveness of any prescribed inhalation pattern (or particle size) in achieving regional targeting will be variable between subjects.

Accordingly, definition of a single set of targeting parameters suitable for all subjects within a large population is not possible, at least with current methods.⁽³⁵⁾ However, enhanced delivered drug dose, reduced treatment times, and reduced variability in lung-delivered dose amongst patients have been observed for inhalation devices with controlled breathing.⁽³³⁾ A reduction in dose variability combined with enhanced pulmonary deposition may also decrease the number of patients required to assess efficacy of new pharmaceutical compounds in clinical trials.

To further improve upon regional targeting that can be achieved by varying aerodynamic particle size and inhalation pattern, bolus delivery of aerosol during only a portion of the total inhalation time has been employed.^(33,36) Limiting aerosol administration to the early portion of inhalation allows inhaled fine particles time to transit through the conducting upper and central airways to reach peripheral lung regions, but under severely constricted patchy conditions inhaled particles may preferentially reach well-ventilated regions. Indeed, poorly-ventilated regions tend to receive a greater fraction of ventilation late in inhalation, as the driving pressure gradient between the airway opening and other, better-ventilated regions decreases during filling.⁽³⁷⁾ Conversely, administering a bolus of larger particles later in the breath may be used to limit their penetration into the peripheral airways, thereby improving targeting to central airways and also preferentially reaching airways leading to poorly ventilated regions.⁽³⁸⁾

Hygroscopicity

Unlike stable aerosols commonly selected for use in laboratory deposition experiments, aerosol particles or droplets inhaled from drug delivery devices can undergo transient size changes that arise due to evaporation of propellants or solvents, and/or to condensation of water vapor from the warm, humid environment of the respiratory tract.^(23,39) Longest, Hindle, and colleagues^(40,41) have proposed to take advantage of such effects so as to allow aerosol sizes to adjust in transit through the respiratory tract, thereby promoting deposition in desired lung regions. Approaches to advantageously spur transient hygroscopic size changes have also been analyzed by Javaheri and Finlay.⁽⁴²⁾

In brief, inhaled submicrometer particles that grow by condensation as they travel from upper to central to peripheral airways will tend to deposit by sedimentation in distal regions while avoiding upstream filtering due to impaction in more proximal regions. Without condensational growth, these submicrometer particles would be exhaled in high fraction. Promoting condensational growth therefore has potential to improve the precision of targeting to peripheral lung regions over more conventional approaches relying on stable aerosols combined with controlled ventilation.

Carrier gases

Helium–oxygen gas mixtures (hereafter referred to as heliox) have been used as a respiratory therapy for decades, often including aerosol inhalation.⁽⁴³⁾ The differences in the physical properties of heliox compared to air (lower density, higher viscosity, and higher mean free path) have a direct effect on the mechanics of particle motion and deposition, but also on the fluid mechanics of the gas that can indirectly change the ADD in the respiratory tract.⁽⁴⁴⁾

When breathing heliox rather than air, particle deposition tends to occur deeper in the respiratory tract^(45,46) and ADD tends to be more homogeneous in particular in obstructed lungs.⁽⁴⁴⁾ On one hand, the lower density of heliox compared to air reduces the extent of turbulent flow, and as such reduces extrathoracic (ET) deposition due to turbulent mixing.⁽⁴⁵⁾ On the other hand, the higher viscosity of heliox compared to air tends to reduce impaction of >1 μm particles in the tracheobronchial (TB) region by keeping particle paths aligned with flow streamlines. For <1 μm particles, the larger mean free path of heliox results in a lower effective viscosity for the particles (i.e., particles are less likely to follow flow streamlines). As a result, deposition of these small particles might be increased, especially in the lung periphery where sedimentation and diffusion dominate. In terms of flow, the lower gas density results in lower inertial flow losses (as these are proportional to gas density) that can lead to a more homogenous flow in diseased lungs suggesting more uniform ADD.⁽⁴⁴⁾

The use of heliox has shown promising clinical results in some studies, even though physiological studies conducted to measure the effect of carrier gas (air versus heliox) on aerosol deposition have led to conflicting results.^(47–49) Kim and colleagues have shown that nebulized racemic epinephrine delivered with helium–oxygen inhalation therapy via high-flow nasal cannula was associated with a greater degree of clinical improvement compared with that delivered by oxygen among infants with bronchiolitis.⁽⁵⁰⁾ Kress et al. showed that, during acute asthma exacerbations, albuterol nebulized with heliox provided a larger improvement in lung function as measured by FEV₁ when compared to albuterol nebulized with oxygen.⁽⁵¹⁾

Despite these promising clinical results, some studies have also shown no significant differences in clinical outcome between heliox-driven and air-driven aerosol drug therapy. For example, deBoisblanc and colleagues⁽⁵²⁾ failed to demonstrate an improvement in FEV₁ when using heliox-driven bronchodilator therapy instead of air-driven therapy in COPD patients with acute exacerbations. Similarly, El-Khatib and colleagues⁽⁵³⁾ showed no significant improvement in pulmonary function test parameters in mild and moderate asthmatics between heliox-driven and air-driven bronchodilator therapy, although significant improvements were seen in severe asthmatcis (FEV₁ < 50%pred). Thus, to this date, clinical trials on the use of heliox in inhalation therapy have not been definitive.⁽⁵⁴⁾

While heliox-based inhalation therapies can modify regional aerosol deposition patterns, as noted above, consistent improvement in clinical outcomes has not been established. The inconsistent clinical results reported suggest that the use of heliox for inhalation therapy may lead to improvements for only selected medications that have dose responses sensitive to regional deposition patterns. Furthermore, the existence of non-responders [i.e., subjects for whom there are no significant differences in deposition patterns using heliox-driven or air-driven therapies (Fig. 2)], calls for the development of techniques (such as SPECT and other imaging technologies; see below) for easy and quick determination of patients, who should respond positively to treatment. These steps would tend to optimize the cost-benefit ratio, making helium–oxygen a more viable and widely used option for improving the effectiveness of inhalation therapies.

FIG. 2.

Coronal slices of 3D Single Photon Emission Computed Tomography (SPECT) deposition data superimposed on x-ray CT images of asthmatic subjects illustrating the effect of helium-oxygen carrier gases (right column) on ADD in a responder (subject A06, top row) and nonresponder (subject A05, bottom row) to helium-oxygen based aerosol therapy. For the same particle size and inhalation flow rate, the responder clearly shows a more homogenous ADD with helium–oxygen than air, whereas for the nonresponder, centrally located hotspots of deposited particles persist with helium–oxygen as the carrier gas. Data from Fleming et al.⁽⁵⁵⁾

Limitations, perspectives and relevance for bridging the gap between science and clinics

Extensive efforts have been made in research and development to improve lung targeting and regional targeting within the lungs. Generally speaking, formulations and devices become increasingly complex (and costly) as the level of refinement in targeting increases. Therefore, delivery technologies must be carefully selected and matched to a given active ingredient and intended use. Devices that guide or control patient breathing during inhalation drug delivery offer potential to target deposition to central or peripheral airways beyond what is possible by tuning aerodynamic particle size alone.

However, they come with a considerably higher price tag as compared with conventional inhalers, and require significant patient training and adherence monitoring. Further, intersubject variability in lung volume and airway anatomy cannot be controlled, such that any single prescribed inhalation pattern and aerodynamic size range will result in variable effectiveness of targeting between individual subjects. Perhaps a primary reason that a perceived gap exists between scientific research aimed at predicting, controlling, and measuring regional deposition and clinical use of inhaled aerosols lies in the considerable challenge of obtaining, in a single study, accurate information describing the optimal lung regions to be targeted, the effectiveness of targeting determined from regional distribution of deposited drug (and potentially subsequent dispersion), and some measure of the drug's effectiveness. Such scientific studies, with the potential to establish links between regional targeting and therapeutic effects, are necessary in order to further motivate development of commercially viable products that target inhaled drug to specific lung regions.

The Use of Models

Cell-free models for predicting ADD

In-vitro models

ADD depends on numerous parameters, including aerosol size distribution (median diameter and standard deviation), inhalation maneuver, and lung morphology. While in vivo imaging of ADD is possible (see section on Imaging Techniques), in vitro models for predicting ADD have been established for rigorous testing and quality control of inhalation devices and aerosol formulations. Considerable effort has been made over the past decade to improve in vivo–in vitro correlations.⁽²¹⁾ These in vitro models typically consist of an aerosol inlet, or induction port, followed by an aerosol-sizing device such as a multi-stage cascade impactor. While the standard European Pharmacopeia (Ph. Eur.)/United States Pharmacopoeia (USP) induction port⁽⁵⁶⁾ is well-adapted to routine, reproducible physical testing of aerosol therapy devices, it is known to be an inadequate representation of the mouth and throat airways when the goal is to mimic extrathoracic aerosol deposition.⁽²¹⁾

Given this limitation, several research groups have developed alternative, more anatomically representative models of the mouth and throat airways, which can be used in place of the standard induction port. These include use of realistic airway replicas derived from medical scans, carefully selected to represent high, medium, and low deposition fractions within the range of values typical of adults,⁽⁵⁷⁾ as well as geometric scaling of a single realistic airway replica to small and large sizes representing 95% confidence limits for mouth–throat deposition in adults.⁽⁵⁸⁾ An alternative approach is to develop idealized models of the upper airways that accurately mimic average aerosol deposition in realistic geometries, while preserving the ease of manufacture and robustness of standardized induction ports. This approach is exemplified by the Alberta Idealized Throat (Copley Scientific, Nottingham, UK).^(59,60)

It is also well known that not only aerosol size and upper airway geometry, but also the inhalation maneuver and handling of the inhaler device may have a strong impact on the pulmonary aerosol dose. These parameters can be highly variable depending on the patient, and contribute to variability in the ADD and hence the therapeutic outcome of an inhalation therapy may be highly variable.⁽⁶¹⁾ Numerous in vitro studies have focused on intersubject variability due to airway morphology (in adults and children) and inspiratory flow rate.^(58,62–65) Simulation of realistic patient inhalation maneuvers during in vitro testing, potentially including simulation of mis-use,⁽⁶⁶⁾ can also be important in establishing in vivo–in vitro correlations, in particular for DPIs where emitted doses and aerosol size distributions are flow-rate dependent.^(20,67,68)

Results of in vitro experiments have also been used to validate models or to derive analytical correlations that can be used in lieu of deposition models, especially for extrathoracic deposition.^{(62,63,65,69)} Such correlations must be applied with caution in predicting ADD or total aerosol deposition from single-breath inhalers, as discussed above in the previous section.

In silico models

In contrast to in vitro models, in silico models can provide information on the specific location of deposition (e.g., lung generation) and allow for fundamental understanding of the main parameters governing ADD. Repeated numerical experiments can be performed by changing only a single variable that may be extremely difficult or impossible to achieve with in vivo experiments. Thus in silico models have been used to assess the hazards associated with environmental aerosol toxicology and for the assessment and optimization of aerosolized drug delivery covering a wide range of conditions.⁽⁷⁰⁾ In particular, ADD optimization through modeling could increase efficacy, and reduce drug cost, side effects, and treatment time if more drug reached the intended lung target.

There are two broad categories of in silico models,^(70,71) analytical and numerical models derived from fundamental analysis of mechanisms, and those derived from empirical data fits of in vivo^(24,72) or in vitro experiments. While empirical models are based on actual data, they may not be applicable outside of the experimental conditions of the data upon which they were derived. In silico models based on fundamental mechanisms can simulate a wider range of applications, but must be validated with experimental data where appropriate.

Several scientific gaps may lead to difficulties in comparing results obtained from in silico models and in vivo imaging measurements.^(73,74) Models typically calculate deposition for a single breath, or equivalently several identical breaths, of a perfectly characterized monodisperse or lognormally dispersed aerosol. For models based on the application of physical deposition mechanisms (i.e., as opposed to empirical models fitted to deposition data), a morphological lung model of some kind must be employed. Even for the empirical models, some coarse correlations are needed to reflect intersubject variability of the respiratory tract.

Imaged data is spatially resolved, but not anatomically specific in terms of generation-specific ADD. Thus, direct comparison of spatially resolved aerosol distribution and lung morphology requires co-registration of the imaged deposition data and of anatomical images, a task that itself requires significant effort to reliably locate deposition to specific airways. This is likely why the deposition data sets most widely used for model validation are those that provide only total and/or regional (i.e., extrathoracic (ET), tracheobronchial (TB), and pulmonary (P)) deposition, where the differentiation between lung regions is accomplished by attributing the aerosol dose cleared rapidly from the lung (typically within 24 hours) to the TB region and the remainder to the P region.⁽⁷²⁾

Computational Fluid Dynamics (CFD) models

Predictive mathematical models of aerosol deposition are useful tools for the assessment and optimization of ADD in the lung. While whole lung models, empirical models, and other macro-scale models have the advantage of being able to predict deposition throughout the entire lung, they lack the ability to describe site-specific deposition within individual airways or specific locations of the lung. In addition, they may not be robust enough to simulate the complex particle dynamics emanating from inhalation delivery devices.

Computational fluid dynamics (CFD) models, on the other hand, allow for the prediction of microdosimetry patterns that can be useful for assessing ADD in specific regions of interest throughout the lung. However, models based on first principles such as CFD models that capture the underlying physics of airflow and aerosol transport are more difficult to implement, require extensive computing resources, and can presently only capture a small subset of the respiratory tract. Recent advances in imaging, computing power, and commercial software have made CFD modeling more available. Even with these advances, CFD models are typically limited to the upper respiratory tract and upper lung airways, but it is these regions of the respiratory tract where accurate predictions of airflow and particle dynamics are most needed due to the complex surface geometries, laminar/transitional/turbulent airflow profiles, and inertial effects.

CFD modeling involves the numerical solution of the governing airflow and transport equations to simulate the aerodynamics and deposition behavior of inhaled particles in the respiratory tract. The first step in the development of a CFD model of the respiratory system is to develop the surface geometry, which defines the extent of the model and forms the model domain. Earlier models of the oral airways and upper lung airways used approximate surface descriptions based on idealized shapes, such as connected cylinders for each airway bifurcation.^(59,75–79)

More recently, CFD applications have used surface models derived from imaging data from individual subjects.^(80–83) These models typically comprise the upper respiratory tract consisting of the oral airways, pharynx, and larynx, and upper lung airways from the trachea extending down through several generations of the lung. The limiting factor in determining how far down into the lung patient-based CFD models can go is driven by the resolution of the scanning procedure. Currently, CFD models extend to about 7–8 airway generations where airway diameters are about 1–2 mm, and can encompass on the order of 200–300 individual airways.

CFD models integrate aerosol characteristics, breathing patterns, and geometric features to predict ADD throughout the upper lung airways. Parameters such as particle sizes and flow rates can be easily varied to study deposition trends and sensitivity. CFD predictions in oral airways and upper lung airways can be compared with experimental data from replica casts for model validation. Ultimately we want to be assured that the CFD predictions can accurately simulate conditions found in vivo. One important aspect of CFD modeling is the selection of boundary conditions. When considering CFD models of the upper respiratory tract, boundary conditions are fairly straightforward because there is only one outlet, although inlet conditions can get complex when simulating drug delivery from a device such as an MDI.

However, CFD models of the lung contain many outlets if the model spans several airway generations. In this case, the simplest choice is to impart a uniform pressure condition on all model outlets. This is typically done in the absence of other physiological information concerning the lung airways. During inhalation, the lobes of the lung are known to expand at different rates, leading to non-uniform airflow profiles in each lobe. Uniform pressure conditions do not capture this phenomenon and will lead to erroneous airflow distribution.

Another option is to consider a weighted outflow boundary condition where the fraction of airflow at each outlet is apportioned according to the distal volume of the lung. This should lead to more accurate results, but it is not always possible to have anatomical information for distal lung volumes. Another avenue that is receiving recent interest is the integration of distal lung mechanics through coupling of the 3D CFD model with 1D or 0D models at each outlet.^(80,84,85) This should allow for a more realistic definition of airflow distribution at model outlets. DeBacker et al.⁽⁸⁶⁾ showed good agreement between CFD predictions of airflow distribution and those derived from SPECT/CT by accounting for airway resistance. It should however be noted that the airway resistance from the airway tree measurable by CT is only a very small fraction of the total pathway resistance, particularly in bronchoconstricted lungs.⁽⁸⁷⁾ Therefore to be quantitatively realistic in non-uniform diseased lung, either the distribution of regional ventilation should be determined experimentally, or the resistance of the peripheral airways needs to be accounted for.

Limitations, perspectives and relevance for bridging the gap between science and clinics

In a recent study, preliminary comparisons of in silico predictions from existing ET and lung deposition models with individualized 2D and 3D imaging measurements of aerosol deposition in healthy lungs⁽⁷³⁾ showed that predicted deposition in the ET models did not correspond well with each individual's experimental data.⁽⁸⁸⁾ However, there was rather good agreement between simulated and experimental results for regional lung deposition comparable to those previously found in the literature. Comparison of the model predictions to lung generational distributions was relatively poor.

These preliminary results suggest not only the need for further developments in deposition modeling such as individual lung morphology models, but also the need for better methods for analyzing experimentally determined 3D deposition distributions for comparison to simulated results. Furthermore, developments are also needed to account for ET variability including movement of the glottis⁽⁸⁹⁾ and to determine deposition patterns for individuals, especially heterogeneous 3D hotspots caused by disease. If progress is made in these areas modeling might become a tool to improve an individual patient's aerosol therapy.

While CFD models are very useful for providing site-specific deposition at distinct locations within the respiratory tract, two other important considerations are: 1) what happens to the drug particle after deposition, and 2) how we can obtain ADD predictions in distal lung airways beyond the CFD model. Comparing model predictions with experimental data from in vitro and in vivo studies is important, but ultimately we would like to use CFD predictions in lieu of experiments to assess the distribution of drug throughout the entire lung.

One option is to utilize the strengths of both CFD and whole-lung deposition models to predict site-specific deposition behavior using the CFD approach in upper lung airways, and have the model outlets connected to 1D models of airflow and deposition to predict deposition in remaining lung airways. The fate of the drug particle after deposition is an important gap to overcome to relate deposition predictions to clinical efficacy. To bridge this gap, CFD models can be linked with pharmacokinetic models that incorporate mucociliary clearance, dissolution and absorption of the drug in lung airways and with biological models discussed in the next section.

Biological models of the lung

Once the aerosolized drug has been deposited onto the lung epithelium, the fate of the drug is determined by the interaction of the drug with biological entities including the lining fluid, mucus, and the pulmonary cell barrier at the air-liquid interface. However, a major gap between ADD and clinical outcome relates to the lack of region-specific investigations of drug efficacy. While the bronchial region of the conducting airways presents a well-protected epithelial barrier consisting of various metabolically active cell types and a dense mucus layer (>50 μm thick), the alveolar epithelial barrier mainly consists of a single cell type (type I cells), which is metabolically inactive and very thin (<1 μm distance between air space and blood) to facilitate gas exchange with the blood.

The few available clinical studies relating ADD and clinical outcome suggest that matching ADD to the receptor distribution of a specific drug (e.g., β2-agonists) improves drug efficacy (see Section on Importance of ADD). Consequently, region-specific drug efficacy should already be investigated during preclinical drug development to provide guidance for the optimum ADD in clinical settings. This will allow formulation experts and engineers of inhalers to develop devices for optimum ADD (see Section on Targeting Approaches). This could reduce cost of clinical trials due to both more efficient drug use and reduced variability in clinical outcome, which relaxes the requirements on the number of enrolled patients.⁽⁹⁰⁾

Hence, the understanding of pathomechanisms and the development of novel therapeutic intervention strategies and their translation into clinically successful therapies depends on the availability of highly predictive physiological models of the human lung.⁽⁹¹⁾ The following sections provide a brief systematic overview of the currently available biological lung models and some examples on how these models can be used to study the relationship between ADD and biological/clinical outcome.

Top-down approach

Figure 3 presents an overview of preclinical biological models of the lung. These models can be stratified according to their level of complexity and to how these levels are attained (top-down or bottom-up approach). In general, advanced biological complexity is associated with a higher degree of similarity with clinical conditions. Four main levels of complexity can be distinguished, namely the levels of organism, organ (here the lung), tissue, and cell. Animal models such as rodents, pigs, and monkeys offer a means of studying biology in the complex context of higher organisms both under healthy and diseased conditions. It is well-known that animal models are limited in terms of their relevance for human disease due to species-specific differences in physiology, pathogenesis, and disease progression.⁽⁹²⁾

FIG. 3.

Available and proposed biological models of the lung stratified according to their level of complexity (or similarity to clinical settings). Different levels of complexity can be accomplished by top-down or bottom-up approaches. In the top-down approach, the model of lower complexity represents a section/part of a model with higher complexity. In the bottom-up approach, which is pursued by tissue engineering, lung models with higher complexity (e.g., tissue, lung, organism) are reconstituted from primary human cells. Some of these bottom-up models are already available as prototypes in miniaturized form (‘on-a-chip’).

As an alternative a whole suite of ex vivo and cell-based in vitro models of the human and animal lung are available for medical science studies such as whole lung explants, lung tissue slices, cell cultures, or reconstructed lung tissue from primary lung cells as graphically summarized in Figure 3.^(93,94) As all of these models (in vivo, ex vivo, in vitro) have their strengths and weaknesses,⁽⁹⁵⁾ scientists have to select the most suitable models depending on numerous aspects including relevance for the scientific issue to be addressed as well as model availability and cost of the study.⁽⁹⁶⁾ Reduced biological complexity of a model is typically associated with mitigated physiologic relevance, easier handling, and reduced economic burden. On the organ level, studies with perfused excised animal lungs can be performed for up to hours and days, the time period these models maintain many physiological functions such as cellular signaling, tissue homeostasis, and integrity of the air-blood barrier.⁽⁹⁷⁾ Direct access to both the epithelial (via tracheal) and the endothelial side (via perfusion liquid) of the air-blood barrier makes the perfused lung an excellent ex vivo model for pulmonary pharmacokinetics studies. In such a setup, aerosolized drug delivery during controlled ventilation of the lung allows investigation of the effect of different ADD patterns on pharmacokinetics and biological endpoints.⁽⁹⁸⁾

On the next lower level of biological complexity, slices of the lung may closely mimic the 3D microenvironment of the lung and capture organ-specific mechanisms of disease for days to weeks. A unique feature of organ slices is the combination of biomimetic tissue characteristics and direct tissue access for four-dimensional confocal live tissue imaging techniques allowing for simple assessment of molecular and functional characteristics of the sample.⁽⁹⁹⁾ As lung slices can be obtained from both animals and humans, they may serve as a direct link between animal studies (science) and clinics.⁽⁹⁹⁾

In spite of the increasing interest in advanced lung models such as perfused lungs and lung slices, most of the medical science research has been performed with biologically and technically more simple lung models, namely human-/animal-derived cell culture models (mainly single cell type models). An overview of the available single- and multi-cell models of the lung is given elsewhere.^(94,100,101) Recently, these cell culture models have matured into biomimetic models of the lung by incorporating previously neglected but crucial physiological aspects of the lung such as: 1) cultivation at the air–liquid interface with air on the apical side and liquid on the basal side of the epithelial barrier;^(102,103) 2) growing the cells on biocompatible 3D (or 2D) matrices mimicking the elasticity of the pulmonary extracellular matrix; 3) exerting cyclic stretch on the cell layer simulating the mechanical strain profile experienced by the alveolar tissue during breathing activity;⁽¹⁰⁴⁾ and 4) combinations thereof.⁽¹⁰⁵⁾

Moreover, the recently introduced ALICE-Cloud technology (licensed to VITROCELL Systems, Germany), for rapid (<5 min) and efficient (>15% substance efficiency) aerosol-to-cell delivery of drugs, may pave the way for dose-controlled delivery of aerosolized drugs to air–liquid interface tissue cultures mimicking the physiologic conditions encountered during inhalation therapy.^(106,107) In addition to the vast body of knowledge on biological endpoints and mechanisms obtained with in vivo, ex vivo, and in vitro models, to a limited extent these models have also been utilized for ADD measurements.^(79,93) For instance, the bioactivity of aerosolized Bortezomib, a proteasome inhibitor potentially suited for inhalation therapy, was verified with both bronchial and alveolar epithelial cell lines, suggesting that the efficacy of Bortezomib is independent of ADD.⁽¹⁰⁷⁾

Bottom-up approach

For decades the top-down approach dominated the development of lung models, i.e., models of lower complexity are obtained by sectioning models with higher complexity. For ethical reasons, patient-derived material is limited to cells and tissue sections. As animal models provide access to all biological complexity levels including whole organism and organ level (see Fig. 3), they represented the gold standard for pre-clinical drug testing and served as source for excised organs, tissue sections, and primary cells.

As an alternative route towards the design of human lung models with higher complexity, bioengineering has embraced the bottom-up approach to reconstitute biomimetic tissue, organ, and organism models from primary human cells.^(101,108) The bottom-up approach is still in its infancy, but tremendous progress has been made in the past decade and tissue engineering holds great promise for substantially improving the availability and clinical relevance of future human lung models.⁽¹⁰⁹⁾ There are already several commercial providers for bioengineered biomimetic bronchial lung tissue reconstituted from primary human cells (e.g., Epithelix, Switzerland; MatTek Corp., USA), which closely resembles the human epithelial tissue of the respiratory tract. This tissue remains fully differentiated and functional for several months.^(110,111). It is available from both healthy and diseased donors and it can be modified to incorporate disease-relevant features (e.g., lung cancer lesions).⁽¹¹²⁾ All of these characteristics facilitate the clinical relevance of medical research performed with these primary tissue models.

On the organ level, several proof-of-concept studies have already been performed. For instance, repopulation of a decellularized extracellular lung matrix (from rats) with endothelial and epithelial cells with subsequent incubation in a bioreactor resulted in an engineered lung with similar mechanical and functional properties to those of native lungs.⁽¹¹³⁾ This in vitro lung was able to contribute to gas exchange when implanted into rats in vivo for a short period of time (up to 2 h). While fully functional bioengineered human lungs are not available, yet they appear to be within reach.

Moreover, miniaturization of advanced 3D multi-cell culture models is under way, leveraging microfabrication and microfluidics technologies to create cell culture microenvironments conducive for differentiation and self-organization of human tissue.⁽¹⁰⁹⁾ These ‘organs-on-chips’ are expected to facilitate in vitro studies of human physiology in an organ-specific context. Ultimately, networks of these ‘organs-on-chips’ may permit the development of biomimetic in vitro organisms (‘organisms-on-chips’) resembling the hallmarks of healthy and diseased subjects offering an alternative and possibly superior (more predictive) methodology than the currently used animal experiments.⁽¹⁰⁹⁾

Limitations, perspectives, and relevance for bridging the gap between science and clinics

All lung models have their advantages and limitations. Scientists have to select the most suitable model depending on the physiological relevance for the scientific issue to be addressed and constraints due to availability and budget. In general, biologically less complex models (e.g., cell culture models) are more readily available, easier to handle, and more cost effective than models with higher biological complexity, but of course their clinical relevance is mitigated compared to more complex models.

In the field of pulmonary toxicology, there has already been a paradigm shift towards an evidence-based research discipline relying on advanced in vitro and ex vivo models instead of animal models.⁽¹⁰⁸⁾ These in vitro and ex vivo models should be able to mimic the complexity of an organ or multi-layered tissue section as much as possible, while maintaining the capability for standardization, high throughput, and reproducibility. Validation of the in vivo relevance of these models by comparison with animal models and clinical studies is currently under way with a focus on toxicological endpoints. The preclinical and clinical communities should embrace these efforts with emphasis on pathological and therapeutic endpoints to lead the way into modern animal-free, yet clinically more relevant and hence more reliable, preclinical drug testing.

It is evident that validated in vitro models of human organisms will not be available in the near future. Hence, animal experiments will continue to play an important role in preclinical drug testing for some time to come. In light of species-specific differences in physiology, pathogenesis, and disease progression, the predictive power of animal models for clinical outcome has been under scrutiny. Thus, complex biological models of the lung such as lung tissue slices, which are available from both animal and human tissue, are expected to play a pivotal role in closing the gap between (animal-based) biological science and clinics. Moreover, advances in tissue and organ engineering hold the promise for overcoming many of the shortcomings associated with therapeutic strategies developed from currently available biological models of the lung. Novel lung models combining innovative technologies for aerosolized drug delivery and tissue engineering (bottom-up approach) are expected to further reduce the gap between medical science and clinical relevance.

Imaging Techniques for the Characterization of ADD

Why do we want to image ADD?

In disease, the distribution of aerosol deposition within the lungs may be highly heterogeneous. As discussed above, since aerosol particles are carried along by gas, their local distribution and deposition are affected not only by the particle properties and local gas flow characteristics but also by the distribution of the inhaled gas into different regions of the lung. Depending on the subject's breathing pattern, aerosol and carrier gas characteristics, and pathophysiological differences within the lung, the local dose of a medication delivered by the aerosol could vary substantially in different parts of the lung.

Imaging provides a noninvasive way to measure in vivo the regional deposition of inhaled drugs along the airway tree and between different anatomical regions of the lung parenchyma. Hence, imaging provides a bridge between science and the clinic, by enabling the study of the relationship between aerosol characteristics, their mode of delivery, and the concentration and location of the medication in the lungs (ADD), which in turn affects its clinical efficacy and/or side effects.⁽¹⁾ Quantifying ADD variability among anatomical regions of the lung is the first step needed to study and understand the factors affecting such variability. For example, in asthmatic subjects the pattern of ventilation is patchy, with large and contiguous regions receiving very low ventilation and others receiving greater than normal levels.⁽¹¹⁴⁾ This and the remodeling of airway dimensions results in a highly heterogeneous ADD that includes both serial differences in the fraction of aerosol retained by the central airways feeding each lobe, as well as parallel differences in the aerosol that deposited distal to these airways.⁽¹¹⁵⁾

Positron Emission Tomography (PET) imaging of ADD combined with quantification of regional ventilation and detailed imaging of lung structure with computed tomography (CT) can be used to understand how serial differences in airway deposition couple with other factors to achieve the parallel heterogeneity in deposition of aerosol among lobes. Those detailed 3D data sets can also be used for advancing CFD models of ADD by providing patient-specific realistic boundary conditions and allowing their quantitative validation.

Estimation of local dosing in terms of the pharmacologically relevant parameters is necessary to bridge in vitro and animal model experiments to the human scale. Evaluation of local dose is also important to understand the regional and global effectiveness of an inhaled drug. Accurate assessment of aerosol deposition along the airway tree and the concentration of deposition on the airway surface may also be useful to evaluate airway functional features, such as mucocilliary clearance rate, a hallmark of some clinical conditions (e.g., cystic fibrosis).

Imaging fundamentals

Clinical aerosol imaging relies almost entirely on radionuclide imaging, which uses radioactively labeled aerosols.⁽¹¹⁶⁾ When inhaled, these allow the fate of the inhaled drug in the body to be monitored using an appropriate imaging device. Two types of radionuclide labels are used, gamma emitters and positron emitters.⁽¹¹⁷⁾ Imaging with gamma emitters is focused around technetium-99m, because of its near ideal physical properties. It uses a gamma camera, which can be used to obtain either planar 2D images (scintigraphy) or 3D images (Single Photon Emission Computed Tomography, SPECT). The principal positron emitters used are fluorine-18, carbon-11,⁽¹¹⁸⁾ and nitrogen-13.⁽¹¹⁵⁾ These are imaged using a PET scanner, which also provides 3D images with somewhat higher spatial detail than SPECT.

Radionuclide imaging is well suited to assessing the fate of aerosol in the body, as it provides accurate quantitative assessment of the amount and location of aerosol deposition. However, quantitative imaging requires careful attention to detail in the methodology and good quality control. A document providing a comprehensive description of recommended techniques for radionuclide aerosol imaging is available.⁽¹¹⁹⁾ This covers methods of labeling and delivering the aerosols, image acquisition procedures, and image analysis.

State of the art imaging uses scanners that combine SPECT or PET imaging with x-ray computed tomography (CT) (Fig. 2). The aligned scans improve the interpretation of the aerosol distribution images and also provide detailed information on airway structure.

Magnetic Resonance (MR) imaging has also been used to image aerosol deposition, but this technology is still very much in the development phase, as practical difficulties of using the technique are being tackled.^(120,121)

Applications and relative merits of techniques

All forms of radionuclide imaging provide measures of the percentage of inhaled aerosol deposited in the lung and of its regional distribution (ADD). Most work to date has used planar imaging. This provides reasonably accurate measurement of total lung deposition^(122,123) and has been very useful in monitoring the increase of lung deposition efficiency in inhalation therapy, as inhaler design has improved over the past 20 years. It has also contributed to elucidating the relationship between delivery of dose to the lung and clinical outcome, allowing assessment of the efficacy of the inhaled drug.⁽¹²⁴⁾

Planar imaging also provides information on the regional distribution of aerosol. This is usually determined by dividing the lung into inner and outer zones and comparing the amount of aerosol in these zones by calculating a penetration index.⁽¹²⁵⁾ This analysis is limited by the two-dimensional nature of the images, meaning that the inner and outer zones only correspond very approximately to central and peripheral airways. However it has been useful in demonstrating the influence of ADD on clinical efficacy.⁽¹²⁶⁾

3D imaging (SPECT and PET) is more complex to use than planar imaging, but provides better information on regional deposition,^(127,128) and has application in studies where this is the endpoint. Therefore it has a role in assessing the influence of regional deposition pattern on the clinical effectiveness of inhaled therapy. When combined with high resolution computed tomography (HRCT), 3D imaging provides excellent information on aerosol deposition pattern in relation to lung anatomy, and is useful in validating computer models of aerosol deposition, as described above in the section The Use of Models.⁽⁸⁸⁾

SPECT is easier to use and more widely available than PET but has some disadvantages, particularly in the inability to directly label aerosols from portable inhalers with technetium-99m.⁽¹²⁵⁾ PET radiotracers, by contrast, can be directly incorporated into drug molecules⁽¹¹⁸⁾ and also provide better image resolution.

HRCT imaging provides useful information on lung anatomy and is of value in providing individualized models of the airway tree useful in computer modeling of deposition. Radionuclide imaging can also provide information on lung ventilation, which is valuable in understanding the link between the aerosol characteristics and location of deposition in the airway tree.⁽¹²⁹⁾

Challenges

The standard approach to evaluate local concentration of a radiolabeled substance is to measure the average activity per unit volume within defined Regions of Interest (ROIs). ROIs are binary masks that serve to group voxels corresponding to specific anatomical regions. These ROIs are either defined based on generalizations of the anatomy,^{(118,125,130,131)}, or are segmented from detailed HRCT scans,^(129,132) when available. In the lungs, it is helpful to segment the lungs into anatomical regions that encompass portions of the central airway tree and the lung periphery. However, the number, size, and location of airways that can be accurately segmented from HRCT images can vary substantially among subjects and within lobes, and may depend (among other factors) on the lung volume at which the CT images are acquired. Therefore, a method to segment the lungs into consistent anatomical regions is required to compare aerosol deposition across subjects and among lobes.

Even after defining consistent anatomical regions, other challenges need to be confronted before aerosol deposition can be accurately evaluated. Regional deposition quantification with binary ROIs may be inaccurate due to partial volume and spill-over effects.^(133–135) These inaccuracies are caused by image blurring due to the limited spatial resolution of the nuclear imaging methods (∼6 and ∼15 mm, for PET and SPECT, respectively) and by the breathing motion of the lung during imaging. Also, although PET (or SPECT) images may be automatically co-registered with the CT images in combined imaging instruments (i.e., PET-CT or SPECT-CT), shifts in patient position, or differences in average lung volume between the PET and CT images, often require additional co-registration. The co-registration process includes errors that need to be considered when estimating regional activity.

Methods to evaluate aerosol deposition that accounts for the effects described above by expanding the concept of binary ROIs into the Grayscale domain have been described.⁽¹¹⁵⁾ This is accomplished by defining a Voxel Influence Matrix (VIM) to describe how activity originating from any anatomical region is sampled in each voxel of the 3D ADD image (Fig. 4). VIMs can include the combined effect of sources of blurring such as breathing motion, limited spatial resolution, registration and model uncertainties, and allows accurate estimation of activity originating from airways within anatomical regions with dimensions that may be smaller than the spatial resolution of the nuclear medicine method. Applied together, these methods leverage the exquisite anatomical detail provided by HRCT to consistently and accurately estimate the distribution of aerosol among specified anatomical regions, and to describe such a distribution in pharmacologically relevant terms such as airway inner surface concentration, and peripheral tissue dosing.⁽¹¹⁵⁾

FIG. 4.

Image processing steps to generate quantitative anatomical ADD images. HRCT images of the chest were collected during apnea at maximal lung inflation and at the average lung volume during spontaneous breathing. These CT images were rendered and segmented into five lobar regions and the proximal fraction of the airway tree up to subsegmental bronchi. The bronchial tree was further segmented into nine airway anatomical regions. The two 3D-rendered airway trees were also mapped to each other and the mapping function, combined with the spatial resolution characteristics of the PET scanner and co-registration uncertainties, were used to estimate the degree of blurring for each anatomical region. Such estimates were transformed into a “gray-scale” voxel of influence matrix, which specified the fraction of each anatomical region sampled by each voxel. Finally, the VIM and the ADD image were used to solve for the activity originating from each anatomical region.⁽¹¹⁵⁾

Limitations, perspectives, and relevance for bridging the gap between science and clinics

Limitations of imaging techniques

Producing adequate radiolabels is not easy. Ideally the aerosol needs to be robustly labeled so that the label remains attached to the therapeutic molecule during the imaging period, which may be several minutes. Many ^99mTc labeled aerosols do not fulfill this requirement and further work to produce more robust radiolabels is required. PET radionuclides can directly label the drugs and the short half-lives of the radionuclides may reduce the radiation exposure to the subject, but developing and validating radiolabels is expensive and requires an in-house cyclotron facility to generate the short half-life isotopes.

Radionuclide imaging provides an estimate of the spatial distribution of the aerosol, whereas what is more clinically relevant is the anatomical distribution in the different airways. This is complicated by the relatively poor resolution of the images (0.6–1.5 cm) and the complex structure of the lung airway, with nearly all of the airways being much smaller than the image resolution. Methods have been described for correcting for the limited image resolution and for transforming the spatial distribution to an anatomical distribution by airway generation,^(115,136) but most work to date has relied on simplified models of the lung airway. Nevertheless, 3D models combined with CT do provide a wealth of data on aerosol deposition related to anatomy, which would benefit from improved analysis techniques.

Radionuclide imaging has the disadvantage of exposing subjects to some health risk, due to the use of ionizing radiation. In the longer term, magnetic resonance imaging may develop as the method of choice for studying aerosol deposition, as it provides improved resolution images and without ionizing radiation dose. However, the practical difficulties in assessing aerosol deposition with MRI mentioned above, suggest that radionuclide imaging will continue to have a role for some time to come.

Despite its limitations radionuclide imaging provides very valuable and unique information on aerosol deposition and should continue to provide useful data in bridging the gap between science and the clinic.

Bridging the gap with a theoretical framework to describe the sources of ADD variability

Based on what is known about the physics of aerosol deposition, it is likely that most of the variability in peripheral aerosol deposition among lobes, sub-lobes, or any set of peripheral lung regions, can be attributed to one of four distinct factors: 1) differences in regional ventilation;^(1,137,138) 2) differences in how the aerosol and air distribute between branches in the series of bifurcations along the pathway feeding the region;⁽¹³⁸⁾ 3) variability in the amount of the aerosol that escapes the series of airways along that pathway;^(139–141) and 4) variability in the amount of aerosol that reaches the periphery and is not exhaled.⁽¹⁴²⁾

Using the concept of aerosol concentration as the average mass in suspension crossing any point of the bronchial tree, a theoretical framework can be defined to quantify each of the factors that lead to heterogeneous aerosol deposition among lobes:⁽¹⁴³⁾ differences in lobar ventilation per unit volume (specific ventilation), uneven splitting of aerosol and air at bifurcations (bifurcation factor), differences in the fraction of aerosol deposited along the feeding airways (escape factor), and differences in the fraction of aerosol that reaches the periphery but escapes via exhalation (retention factor).

Such a framework was used to quantify the contribution of these factors from ADD images in a group of bronchoconstricted subjects with asthma.⁽¹⁴³⁾ That analysis gave the following results: 1) Differences in lobar specific ventilation (measured from the turnover rate of ¹³N washout) and in net branching factors each accounted for more than a third of the variability in deposition among lobes and subjects. The remaining variability was due to differences in deposition along the feeding airways as characterized by their net escape fractions; 2) By exploring inter subject variations in aerosol deposition, it was found that subjects who were breathing slowly (<9 BPM) during nebulization had a strong relationship between regional lobar deposition and their respective regional ventilation, measured with PET. However, that relationship was weaker in the subjects who were breathing at increasing breathing rates. Also in subjects breathing faster, the more expanded lobes showed lower deposition per unit ventilation than less expanded lobes, while the opposite was true for subjects breathing slowly; 3) Differences in lobe expansion between HRCTs at two lung volumes, used by others as a surrogate of regional ventilation,⁽⁸⁶⁾ failed to explain the variability in regional deposition across subjects.

Importance of Aerosol Deposition Distribution to Clinical Efficacy

Currently there is much debate on whether it is advantageous to direct inhaled therapeutic aerosols to deeper lung regions than is currently achieved by conventional medical aerosols used in clinical practice.^(144,145) In order to achieve this very coarse differentiation in regional inhaled aerosol deposition within the lungs, many factors can be influential of which particle size and aerosol velocity are key aerosol attributes⁽²⁾ (see section on Basics of Aerosol Transport and Deposition in Lung).

Indeed, empirical correlations show that particle size influences not only the total but also the regional site of airway drug deposition.⁽⁷²⁾ Many investigators have attempted to determine the relationship between β2-agonist particle size and clinical effect using polydisperse and also monodisperse aerosols, but have reached differing conclusions on the optimal particle size. This probably relates to differences in experimental aerosol methodology and delivery, as well as differences in the variable clinical endpoints used and patient characteristics.

In a series of well-validated experiments utilizing monodisperse albuterol aerosols, Usmani and colleagues initially showed that larger 6 μm albuterol particles achieved better bronchodilation in asthmatic subjects than smaller 3 and 1.5 μm particles, when the clinical endpoint of FEV₁ was employed to assess efficacy.⁽¹⁴⁶⁾ The investigators hypothesized that as a result of the physical nature of aerosol monodispersity, there was relative selectivity in the airway tree that larger 6 μm particles were preferentially targeting the proximal airways, whereas smaller particle were being directed to the deeper lungs.

This group subsequently undertook a novel imaging study using a validated radioaerosol generation system⁽¹⁴⁷⁾ to assess the effect of different β2-agonist particle sizes and their influence on the aerosol deposition-clinical efficacy relationship.⁽¹²⁶⁾ Indeed, it was observed that larger particles visually deposited more proximally, whereas smaller particles distributed deeper into the lungs and had a greater penetration index.^(125,148) The paradox of overall less total lung deposition with the 6 μm particles but greater efficacy in FEV₁ was explained by better matching of the inhaled therapeutic to its receptor distribution.

However, the authors discussed the fact that spirometry was unable to differentiate between deep (smaller particle) and proximal (larger particle) airway effects. The authors concluded that perhaps the larger particles were showing better bronchodilation because the FEV₁ was assessing the proximal lung region. The smaller particles may also have ‘clinical efficacy’, but the appropriate clinical endpoint to assess deep airway effects had not been used. In this context over the last decade, technological advances have been made in imaging and physiological measures to probe the deep lung and are progressing the field to assess the so-called ‘quiet’ deep lung region⁽¹⁴⁹⁾ (see above section on Imaging Techniques).

Inhaled corticosteroids (ICS) remain the most important anti-inflammatory treatment for patients with asthma, and advances in drug formulations and device design have generated a variety of differing ICS particle sizes. These commercial inhalers have been investigated using scintigraphy to assess whether differences in aerosol deposition and regional lung distribution are observed due to differences in ICS particle size. Indeed, data show that small particles achieve higher total lung deposition than larger particles, and that smaller particles penetrate deeper into the lungs.^(150,151) With this fundamental difference observed in aerosol deposition and distribution, investigators have utilized different ICS particle sizes in patient studies to determine if there is a difference in clinical efficacy.⁽¹⁴⁵⁾

Hoshino⁽¹⁵²⁾ showed that smaller ICS particles (∼1.1 μm) significantly improved distal airway resistance R (as measured by impulse oscillometry (R5Hz minus R2OHz or 5–20 Hz)) compared to larger ICS particles (∼5.4 μm), but also showed that only the smaller particles significantly improved asthma control, most likely because better targeting was achieved with the smaller ICS.^(151,153) Importantly it was observed that there were no significant changes in spirometry in either group, but that distal airway physiological measures (i.e., impulse oscillometry) were able to pick up changes in smaller particle therapy that were not detectable with proximal airway measures such as spirometry.

Similar data have been shown using combination ICS and long-acting β2-agonist (LABA) inhalers, where smaller ICS-LABA particles (∼1.5 μm) achieved greater asthma control than separate standard larger particle (∼3 μm) inhalers.⁽¹⁵⁴⁾ Data are now accumulating that indicate treatment of the deeper lungs with smaller drug particles (whether monotherapy ICS or combination therapy ICS-LABA) in patients with asthma is as good as with larger particles, but additionally allows a reduction in the overall daily ICS dose with smaller particles; indeed, some studies show that smaller particles achieve greater clinical efficacy.⁽¹⁴⁵⁾

It has been questioned whether improving the efficiency of inhaled aerosol deposition within the lungs, by targeting drug delivery to the appropriate airway regions and sites, actually improves the therapeutic response. That is, is there a risk that deeper lung delivery improves efficacy but may detrimentally worsen adverse effects? Derom and colleagues⁽¹⁵⁵⁾ compared the systemic exposure of small versus large particle ICS monotherapy and observed that smaller ICS particles (∼1.1 μm) of ciclesonide did not significantly affect cortisol secretion, in contrast to larger ICS particles (∼4 μm) of fluticasone propionate.

Concerns have been raised about the clinical safety of aerosolized solution particles of tiotropium, a long acting muscarinic antagonist (LAMA),⁽¹⁵⁶⁾ that have recently been dispelled by the FDA at the recent American Thoracic Society Meeting. Indeed, there is interesting pharmacokinetic data comparing slow velocity aerosolized solution particles of tiotropium to faster velocity aerosolized powder particles, where blood exposure is lower following the slow velocity aerosol.⁽¹⁵⁷⁾ Part of this may be due to the lower nominal drug dose with the more efficient slower velocity tiotropium aerosol.

Conclusions and Outlook

Development of a new drug for the treatment of lung disease is a complex and time-consuming process. The identification of potential drug targets, the screening for bioactive substances with sufficient target specificity, the formulation of the drug, and the clinical validation require contributions from numerous basic and applied disciplines. Only substances with clinically proven safety and efficacy profile will receive regulatory approval and hence, only a very small fraction of the pre-clinically investigated substances will benefit the patient and successfully reach the marketplace.

The translation of scientific knowledge into clinical innovation is often delayed or even disregarded because of a significant gap between science and clinical relevance. For example, ventilation-controlled inhalation devices are only just starting to reach the clinics, even though the significance of the inhalation maneuver for optimized pulmonary drug delivery has been well-known since the mid 1980s.⁽⁷²⁾ Reasons for this gap include language and methodology barriers, as well as regulatory issues. The transfer of knowledge between disciplines is hindered or even made impossible by the use of different terminologies and methodological approaches. While this is a common feature of highly interdisciplinary ventures, the high regulatory hurdles for novel drugs/therapies are specific to the clinical setting. Undergoing all phases of clinical trials involves a high financial risk. Pharmaceutical companies prefer modification or repurposing of established drugs, therapies and technologies that requires limited budgetary risks to the introduction of completely new ones.

A key aspect for bridging the gap between science and clinic is the establishment of a more comprehensive approach to drug development. This should include regular interaction and knowledge exchange between the various disciplines of basic, applied, and medical/clinical science. This has been recognized by many funding agencies that now often require consortia to include clinical partners, even for calls geared towards basic research. Consequently, numerous interdisciplinary networks and comprehensive pulmonary centers have been established bringing experimental and clinical research units under one roof. As clinical trials have to be financed by industry, early involvement of industry representatives is also essential, posing another “language” barrier, as the scientific merit of new ideas/technologies has to be translated into budgetary terminology.

The workshop on Bridging the Gap between Basic Research and Clinical Efficacy of Inhaled Drugs, which took place during the 2015 ISAM Congress, also served to facilitate interdisciplinary communication. The main conclusions from this interdisciplinary group of experts can be summarized as follows:

• • Basic research towards identifying relevant parameters influencing patient-specific regional drug targeting has contributed to improved engineering of inhaler devices and drug formulations. Continued development of advanced products enabling patient-specific (precision) regional targeting will require modern 3D imaging and clinical pulmonology, so as to further establish relationships between regional doses and clinical outcomes.

• • Computational analysis of ADD is expected to play an important role in matching patient needs and device characteristics for disease-specific optimized drug delivery (precision medicine).

• • The availability of patient-specific, easy-to-use devices has to be accompanied by thorough training of patients in device handling for optimized compliance, a prerequisite for the therapeutic success of inhalation therapy.

• • The predictive power of preclinical substance testing for clinical efficacy should be improved by utilizing advanced preclinical in vitro and ex vivo lung models such as reconstructed/sliced tissue models from donors and patients combined with dose-controlled, aerosolized drug delivery under physiological conditions. This is expected to result in less false positive and less false negative drug candidates reducing the risk of conducting ill-fated clinical studies (reduced budgetary risk) or premature termination of the testing of clinically successful drug candidates, respectively.

• • In the intermediate future, embracing the currently available advanced preclinical in vitro and ex vivo models of lung tissue may lead to the development of human-based lung models at the organ and even organism level paving the way for more reliable, more cost-effective, and animal-free preclinical drug testing.

• • In the more distant future, linking computational ADD models with pharmacokinetic, pharmacodynamics and systems biology models may allow in silico drug testing.

These findings demonstrate the crucial role of bridging the gap between basic research and clinical efficacy for optimized translation of fundamental research from the bench to the bedside (clinical outcome).

Abbreviations Used:

ADD

aerosol deposition distribution (regional distribution of the deposited aerosol in the lung)

AR

anatomical region

CFD

computational fluid dynamics

CT

computed tomography

DPI

dry powder inhaler

ET

extrathoracic

FDA

Federal Drug Administration

FEV₁

forced expired volume in 1 second

FRC

functional residual capacity

GI

gastrointestinal

HRCT

high resolution computed tomography

ICS

inhaled corticosteroids

ISAM

International Society for Aerosols in Medicine

LABA

long-acting β2-agonist

LAMA

long acting muscarinic antagonist

¹³N

nitrogen-13

PET

positron emission tomography

pMDI

pressurized metered dose inhaler

P

pulmonary

ROI

region of interest

SPECT

single photon emission computed tomography

TB

tracheobronchial

^99mTc

technetium-99m

VIM

voxel influence matrix

Acknowledgments

Dr. Darquenne is partially supported by R01 HL073598 from the National Heart, Lung, and Blood Institute (NHLBI) at the National Institutes of Health (NIH). Professor Fleming would like to thank the NIH Research Biomedical Research Unit in Respiratory Disease of the University Hospital Southampton NHS Foundation Trust for their financial support. Dr. Martin acknowledges financial support from the Canadian Natural Sciences and Engineering Research Council (NSERC). Dr. Usmani is a recipient of a UK National Institute for Health Research (NIHR) Career Development Fellowship and supported by the NIHR Respiratory Disease Biomedical Research Unit at the Royal Brompton and Harefield NHS Foundation Trust and Imperial College London. Dr. Venegas like to acknowledge the substantial contribution of Elliot Greenblatt PhD on the analysis presented here that was part of his PhD dissertation and the referenced articles. Also acknowledged are the rest of the Pulmonary Imaging and Bioengineering Lab at MGH for their contributions in the imaging data collection and to Air Liquide scientists for their contribution in the design of the protocols and their analysis. Dr. Venegas also is indebted to Air Liquide Santé International and the US NIH who provided financial support for his work in this area.

Author Disclosure Statement

Professor Fleming has acted as a consultant to Air Liquide Santé International (France). He has also received funding from Astra Zeneca for his work in the imaging of aerosol deposition and clearance. Dr. Venegas is employed by the Massachusetts General Hospital, Boston Massachusetts USA, and has received financial support from Air Liquide Santé International.