Dispersion and Characterization of Pharmaceutical Dry Powder Aerosols

Dry powder inhalers may be employed for the delivery of drugs to the lungs for the treatment of pulmonary and systemic diseases. Dry powder inhalers consist of a formulation, metering system and delivery device. The formulation consists of small drug particles (median aerodynamic diameter <5 J.lm) usually combined with large excipient particles (30-125 J.lm). The physico-chemical properties of particles influence flow, fluidization and dispersion. An aerosol is passively inhaled by means of the inspiratory effort of the patient. Active devices use independent energy sources to deliver the aerosol. The aerodynamic particle size of an aerosol significantly effects lung deposition. Inertial deposition of particles, as a function of their aerodynamic diameter, allows measurement of the drug mass in a specific size range. The fine particle dose, or fine particle fraction (FPF) of the metered dose, is the simplest expression of the aerosol particle size. Passive inhalers have been reported to administer up to 50% FPF and active inhalers up to 70% FPF. The criteria for optimal inhaler performance must be defined to facilitate the rapid growth in research and development in this field.


Introduction
Epinephrine was the first drug to be incorporated in a chlorofluorocarbon (CFC) propellant driven metered dose inhaler in the 1950s [216).This aerosol delivery device was a forerunner of current inhalers used in the treatment of asthma [233).In recent times the success of CFC propellant driven metered dose inhalers has been overshadowed by their contribution to ozone depletion in the upper atmosphere and the concomitant health effects [169).Therefore, the international community has agreed to phase out CFC propellants by the year 2,000 [52).Pharmaceutical scientists have been responsive to the concerns over the environmental effects of propellants and have sought alternatives for the delivery of drugs to the lungs.The three major alternatives are the use of propellants that do not deplete ozone [179,235), aqueous droplet aerosols delivered by nebulizer [178) and dry powder inhalers [2,175).Each of these approaches has its merits and their uses have been discussed in detail in the literature [90,91).
The successful delivery of dry powders as aerosols depends upon a number of closely related factors.The intrinsic physico-chemical properties of the drug particles are the foundation for optimization of their delivery as aerosols [24).These properties such as particle crystallinity, size distribution, shape and density may be modified in the manufacturing process and must be characterized carefully before proceeding to their incorporation in the dosage form [32].The forces of interaction between particles in a powder bed are contingent on their physico-chemical properties and can be defined for specific powder forms [208).The dynamic properties of flow and dispersion, which are related to ease of particle separation, may be evaluated as the basis for modification of powder performance.Finally, the interaction of the powder metering system and the device intended to generate the aerosol must be evaluated to optimize the dose delivery to the lungs.Considering each of these topics at a fundamental level will lead to successful integration of the powder formulation, metering system and inhaler to facilitate aerosol dispersion.The following sections discuss aspects of the dispersion and characterization of pharmaceutical dry powder aerosols and integrate them to define appropriate inhaler quality and performance criteria.

Physicochemical Properties
Crystallinity, polymorphism, particle size distribution, shape, moisture content and hygroscopicity have an impact on the dispersibility and stability of dry powders intended for delivery as aerosols.It should be noted, when considering particles that are included in aerosol powder systems, that the drug is rarely administered alone.The majority of formulations include carrier particles to which the drug particles adhere and from which the drug is dispersed [45].

Crystallinity and Polymorphism
The crystallinity of the powder may be described in terms of the crystal system and crystal habit of the particles.The crystal system is defined by the interatomic or intermolecular spacing within the particle, i.e., bond lengths and angles [247].The crystal structure can be studied by X-ray diffraction [32].Determination of phase angles for each Bragg reflection allows the calculation of the electron density map from which the molecular structure can be calculated [77].The structure of cromolyn sodium is an example of pharmaceutical relevance [50].Crystallographic studies of this particle structure have evaluated the interaction of the drug with water [189].There are seven major crystal systems [171].The crystal habit describes the morphology of the particle which may differ from the unit cell defined by the crystal system if the growth of the particle has been restricted in any dimension [32].Figure 1 illustrates the characteristics of the major crystal systems and gives an example of crystal habit.
Different crystal systems of a single chemical entity give rise to polymorphism.Since the bond angles and lengths differ between polymorphs their physical properties also vary [245].Thus, polymorphs may exhibit different particle shape, solubility, hygroscopicity and moisture content which will have an impact on aggregation, through electrostatic charge, capillary and mechanical interlocking forces.Particles must be very small, <5 ~m, to be suitable for lung delivery [26,89,232].In order to manufacture small particles, a considerable energy expenditure is required, commonly in the form of attrition milling or spray drying [136].Much of this energy is expended in breaking the uniform bonds that exist in crystalline structures.Crystallinity must also be considered for carrier particles, notably lactose monohydrate, since this will have an impact on the nature of the mixing process and final distribution of drug particles in an aerosol powder.Elastic particles (low Young's modulus) are difficult to disperse due to deformation of interacting particles, generating large contact surface areas.Deformation increases interparticulate forces and reduces the porosity of the bulk powder.(a) The seven major crystal systems [171] and (b) an example of changes in crystal habit [32].

Particle Size, Distribution and Shape
The particle size and distribution of the drug in the bulk product are not usually controlled.The powder, which is most frequently prepared by crystallization or precipitation from solution [70], must be reduced in size to be suitable for inhalation [9,136].Size reduction frequently results in a powder which is log-normally distributed around a median diameter [6].This type of distribution is typified by a large number of small particles.Individual large particles when broken into small ones produce numbers which are proportional, maximally, to the cube of the diameter.This is not only true of attrition milling but also of spray drying where droplets disintegrate during atomization in a similar geometrical fashion.Particle size reduction often eliminates specific irregular morphologies, or shapes of particles.Most spray drying processes are optimized to produce spheroidal particles [35,212,239].Attrition milling results in particles which have a large number of planes of rotational symmetry [146] but are not spherical.Supercritical fluid manufacture is a relatively new process that offers control of the physical characteristics of the particle [212].This method has the advantage of generating respirable particles from heat labile compounds, e.g., peptides and proteins.
The size and shape of carrier particles must be considered when preparing blends with the drug.Lactose monohydrate occurs as a triclinic crystal and must be manufactured or sieved into size ranges suitable for drug delivery.In the original formulation of disodium cromoglycate the range of lactose particle size was 30-60 J..lm [119].Various sizes up to 125 J..lm [47] have been explored for this purpose.The range of particle size employed will effect the flow and packing of the powders [144].
Particle shape is known to influence powder dispersion [199] and lung deposition [142].In general, large deviations from sphericity are not considered useful in powder formulation, i.e. fibers or plates.Thus, any advantages that shape offers in terms of lung deposition are negated by their effects on reproducibility of dosing.Shape affects particle interactions with other particles in the powder and also with air during fluidization and deaggregation.Particles that have bluff body profiles (high drag coefficients) experience large drag forces and are therefore easier to fluidize.Elongated, fibrous particles have been shown to have unique transport properties that enhance deposition in the lung [33,68,99].The contact state between fibrous particles is extremely unstable, promoting fluidization and deaggregation [186].Rugosity balances the effects of interparticulate forces and mechanical interlocking.Rugose particles reduce interparticulate forces by the small effective diameters of the asperities but can promote mechanical interlocking as the surface irregularities increase, particulary for small particles in interactive mixtures.Reducing particle rugosity can significantly improve the respirable fraction of a dry powder inhaler [71].However, smooth particles tend to experience high interparticulate forces, e.g., potato starch.
Particle size may be described by a variety of dimensions.The most common term for expressing the particle size from microscopic examination is the projected area diameter.This equates the diameter of a circular disc with the projected area (A) of the particle resting in the plane of greatest stability [6].Dp = (A4/ rc) 1 1 3   (1) For particles which are near spherical, the projected area diameter approximates the diameter of a sphere that has the same volume as the particle, i.e., the equivalent volume diameter (Deq).Laser scattering and diffraction particle sizing methods allow the determination of equivalent volume diameters.
The most appropriate dimension for defining aerosol particles is the aerodynamic diameter [204].The aerodynamic diameter (Dae) of a particle is equivalent to the diameter of a unit density (po) sphere that has the same terminal velocity in still air: D _ D ( PP )o.s ae-eq poX (2) Eq.(2) shows that Dae is dependent on both particle shape (x) and density.
Particle size is statistically represented by two parameters: a representative diameter and a measure of the distribution of particles sizes.These parameters usually take the form of a median diameter and geometric standard deviation for inhalation aerosols, although many other representative diameters and distribution factors can be used [ 110,139].The Deq commonly calculated by impaction methods is the mass median aerodynamic diameter (MMAD).MMAD represents the aerodynamic particle size for which 50% of the total aerosol by mass have smaller aerodynamic diameters.MMAD is calculated assuming a log-normal distribution of particle sizes.Distribution of particle sizes is represented by the geometric standard deviation (0g) of the log-normal distribution (Dae.os4/MMAD).
A recent development in the physical structure of particles has been made with the production of large (5<Dp<30 J..lm), porous particles [60].Porous particles are less dense than solid particles with equivalent volume diameters, making it possible to increase the particle cross-sectional area (improved fluidization) without increasing the aerodynamic diameter (deposition) (see Eq. ( 2)).Administering large diameter particles (>3 J..lm) to the lung also increases bioavailability by reducing phagocytosis [211] and thereby extends the residence time of the particles in the lungs.
The powder packing consists of individual particles in mutual contact with voids filled with gaseous medium.The volume of gas to the total volume of the powder pack is known as porosity (£).This parameter provides a measure of the compactness of the packed powder bed and can quantify the uniformity of the powder blend.Porosity is effected by particle size distribution, interparticulate forces, mechanical interlocking and friction, random settling and boundary effects [208].These forces dictate the bulk and tap density.The bulk density, in contrast with the true density, is a measure of the interstitial space between particles, or the spontaneous packing geometry [32].The tap density indicates the extent to which the packing of particles can be re-arranged by imparting mechanical energy.This can be viewed as a measure of the stability of the spontaneous packing geometry.Both of these properties will be influenced by particle size, distribution, shape, density, charge and moisture content.

Moisture Content and Hygroscopicity
The moisture content of powders will have an impact on stability.The equilibrium moisture content at ambient temperature and relative humidity may be a good indicator of the suitability of a drug or excipient for use in an aerosol product.Figure 2 shows the equilibrium moisture contents for lactose (a) ""  and mannitol [244].The moisture content of most sugars is relatively low at a relative humidity below 90%, e.g., Glucose, lactose, maltodextrins, mannitol exhibit equilibrium moisture contents of <21%, <7%, <6%, =0.5%, respectively.The amount of free water in a system is closely related to the magnitude of the capillary forces which hold particles in aggregates.
The water confined between particles in narrow pores is highly structured relative to free water, giving rise to strong interactions [61].Moisture between particles also generates capillary forces which are a function of the surface tension, liquid density and pore radius [70].
The method of manufacture of particles also has an impact on their uptake of moisture.Micronized cromolyn sodium exhibits a decline in respirable particles (0.3-3.3 )lm) from 8 to 4%, whereas spray dried particles drop from 17 to below 1% for RH in the range 0-80% [2,241].This data was derived from a study in which both products were prepared as 1:1 blends with lactose [241] Hygroscopicity is the tendency of particles to take on moisture from the atmosphere.Susceptibility to hygroscopic growth can result in physical instability.Moisture uptake and loss over a narrow range of relative humidities can result in dissolution and recrystallization, leading to irreversible aggregation in the form of solid bridging.
Electrostatic charge occurs by a number of methods.The charge associated with certain ionic species at the surface of a particle will influence particle behavior.Contact with specific surfaces and charge transfer by motion will give rise to unique powder charge characteristics.The charge on a particle increases as it approaches the anhydrous state.The forces of interaction between hydrophobic surfaces are generally less charged than hydrophilic ones [5].The surface energetics may be investigated by inverse phase gas chromatography [116,117] from which the cohesive energy density of the surface of particles may be obtained.The chemical heterogeneity of the surface can also be studied by gas adsorption with fractal analyses [ 10 ].

Fundamentals
The fundamental aspects of powder aerosol dispersion for inhalation, illustrated in Figure 3, may be considered in terms of static powder properties, flow, fluidization and deaggregation.Static powder properties are the properties of the powder at rest and con- Static powder properties exist in the dry powder inhaler after the powder has been metered and before the powder is dispersed for inhalation.Static powder properties can be described in terms of microscopic (single particle) characteristics (physicochemical properties: section 1.1), interparticulate forces and macroscopic (bulk powder or powder packing) characteristics.
Interparticulate forces that influence particle interactions during the dispersion of powder aerosols include the cohesive/ adhesive van der Waals, electrostatic, capillary, mechanical interlocking and friction forces.Cohesive forces are due to the attraction of like materials, e.g., particle to particle.Adhesive forces are due to the attraction of unlike materials, e.g., particle to carrier or device.Particulate interactions are primarily dominated by van der Waals force, which are only effective over a short range ( < 10-3 Jlm), but can be very strong (> 103 F g).The van der Waals force of attraction between two spherical, rigid particles of different diameter is given by: ADz2 where A is the Hamaker constant, h is a characteristic length scale representing the shortest distance between particles and D12 is given by: Dz+D2 Application of Eq.( 3) is limited to qualitative calculations due to the difficulty in defining the Hamaker constant and the characteristic length scale [ 172).
Estimates of these parameters are available in the literature [64,140,120,208,243,254).Values for the Hamaker constant range from to 5x 10-20 -1 x 10-19 ], with values for the shortest distance between partie cles ranging from 0.35-10 nm.A range of van der Waals force of over 1600 is predicted from these values, highlighting the difficulty in quantifying this force.Eq.( 4) can be modified for non-spherical particles by defining an effective diameter of the asperity at the point of contact.The magnitude of the van der Waals force will be considerably reduced since the diameter of the asperities will be smaller than that of the particle.
Electrostatic charge is generated when particles are separated from the bulk material or contact other surfaces.This mechanism can significantly affect formulation homogeneity, especially when the powder contacts dissimilar materials [224].The force of electrostatic attraction between two particles with charges q1 and q2, respectively, is given by Coulomb's law: (5) q-q -;;zwhere K.q is the electrostatic force constant.General application of Coulomb's law is limited due to the difficulties in defining the respective particle charges, which change with contact surface, environment and flow conditions.Definition of a characteristic length of charge separation is also required.The magnitude and polarity of charge will be different for the static powder in contact with the device compared to a deaggregated particle.The net charge of the fine particle dose generated by commercial DPis has a dependency upon the type of inhaler and the method of deaggregation, although it is difficult to delineate these mechanisms [193].Electrostatic effects are negligible at relative humidities above 65% [100].
Capillary forces are formed by the condensation of water to form high tensile liquid bridges.Liquid bridges can occur at elevated relative humidities, i.e., RH>65% [214,242], and are dependent on particle material and surface characteristics.The capillary force between two spherical particles of different diameter is given as: Fe= 2rcaDz2 (6) where D12 is defined by Eq.( 4).This force will be significant during transport in the lung (RH-+ 100%) due to its dependence on relative humidity.
Mechanical interlocking and friction are significant effects when considering the relationships between irregular large carrier particles and small drug particles.The magnitude of the forces resulting from mechanical interlocking are dependent on surface characteristics.Small drug particles can be forced into cavities which require large separation forces.Mechanical interlocking can be enhanced with prolonged blending.Interactive mixtures form stable associations with other particles during blending.Particles continue to relocate to sites of higher stability /lower energy as the blending time is increased, promoting the effect of mechanical interlocking.Friction is generated between surfaces in mutual contact and is dependent on the contact force, area at the point of contact and surface roughness.The force of friction is more apparent for smaller particles, as both the coefficient and angle of friction increase with decreasing particle size [ 66].
The magnitude of the respective interparticulate forces (with the exception of mechanical interlocking and friction) can be approximated by considering an ideal binary system, i.e., a pair of spherical, rigid particles.Interparticulate forces are estimated in Table 1 for four binary systems: (I) carrier:carrier; (II) carrier:drug; (III) drug:drug and (IV) asperity:asperity.The model carrier is a 90 )1m lactose particle and the model drug is a 2 )1m albuterol sulphate particle.Capillary forces are significantly higher than both van der Waals and electrostatic forces but are only produced at high relative humidities (RH>65%).This suggests that capillary forces dominate in the lung and further deaggregation of agglomerates beyond the oropharynx is unlikely to occur.Ranking of interparticulate forces during dispersion within a DPI reveals that van der Waals force dominates with electrostatic forces being an approximate two orders of magnitude less.The magnitude of the forces estimated in Table 1 are dependent on particle diameter for a binary system, i.e., interparticulate forces decrease with decreasing particle size.However, it   [108] carrier=lactose, DJ=90 pm, p1= 1540 kg/m 3 IV 0.01 -23 a q approximated from specific charge measurements of Peart eta! [193] 12 becomes increasingly difficult to fluidize small particles since aerodynamic force is proportional to the cross-sectional area.
The magnitude of the forces for a 'real' D PI powder will be different to theory due to asperities and powder packing effects.Asperities depend on the rugosity of the particle and are typically <0.1 )1m [208].Thus, the van der Waals force between two particles, approximated in Table 1, would be reduced by a factor of nearly 900 if the particles were connected by asperities.The magnitude of the interparticulate forces within a powder packing is also affected by the number of mutual points of contact, i.e., the coordination number (k).Interparticulate forces for a binary system are approximately proportional to Dp (Table 1).Thus, small particles have smaller interparticulate forces at a single point of contact.Interparticulate forces acting on a particle within an agglomerate increase by a factor of the coordination number, e.g., a particle within a cubic packing (k=6) will experience interparticulate forces six times greater than that for a binary system.

Powder Characteristics
The dispersion of particles from a static powder bed may be considered in terms of dilation, flow, fluidization and ultimately aerosol production.These are integrated mechanisms during the dispersion process that can be evaluated individually to construct a model of powder dispersion.
Flow is the first step in the dispersion of aerosol powders, occuring as enough space is created between particles for them to pass over and around each other in the direction of an applied force.In a system of spherical, uniform, non-cohesive particles, this process might be expected to occur in a predictable fashion with the flow rate being proportional to applied force.Unfortunately, pharmaceutical powders are non-uniform in size and shape and are subject to various cohesive and adhesive forces.There are two scales of scrutiny of powder flow as it relates to aerosol dispersion.These are the bulk macroscopic and the discrete microscopic flow characteristics.There are many methods of measuring powder rheological properties [219].The angle of repose [31], Flodex (Hanson Research Corporation, Chatsworth, CA) and the shear cell [209] are examples of classical methods.The angle of repose indicates the ease of flow in terms of the angle that a pile of powder poured through a funnel makes with respect to the horizontal.A large angle of repose indicates poor flow properties.The Flodex, powder flowability index test instrument (US Patent #4, 27 4,286), is a simple method in which powder is poured through an orifice from a hopper.The orifice diameter may be reduced until the smallest orifice through which the powder will pass unassisted is identified.The diameter of this orifice becomes the measure of the ease of flow of the powder.The shear cell is analogous to a viscometer.The resistance in the form of torque experienced ~ when a force is applied to the surface of a powder is used to define its ease of flow.The latter technique is particularly useful in evaluating powders that will experience compression but may not be as relevant in the evaluation of aerosol powders.Each of these methods measures bulk flow properties which are known to be inversely proportional to particle size.
Small fluctuations in flow whieh arise from small stable aggregates within the powder are of additional interest.These fluctuations may be of great significance on the scale of a unit powder dose, i.e., :s25mg.Alternative methods have been developed for the evaluation of these small variations in flow.Flow from a vibrating spatula has been analyzed to allow derivation of fractal dimensions describing the irregularities in flow [94].This method measures the powder dispensed gravimetrically and is limited in the speed of data acquisition (~300ms intervals).This is insufficient to adequately distinguish between powders with differing flow properties, exhibiting a range of fractal dimension of 1.003 for smooth flow to 1.057 for irregular flow.A second method has been developed involving chaos data analysis of dynamic angle of repose measurements [95] yielding capacity dimensions, analogous to fractal dimensions, which have the sensitivity to discriminate between the performance of powders, ranging from 1.2 for periodic to 2.6 for random behavior [48].The phase space plots of the original data may be used to distinguish the effect of particle size distribution compared with particle interaction.Figure 4 shows a phase space plot of bulk lactose [95].The distance across the trajectories reflects the breadth of particle size distribution.The radius of the trajectories is inversely proportional to ease of flow.This is a relatively new method for studying powder behavior and it may be possible to derive more information from the data by alternative analytical methods, e.g., avalanching [132].The concept of self-similarity at all scales of scrutiny may ultimately be applied to many aspects of powder characterization, including particle morphology, powder mixing, flow and dispersion [93].
Particle detachment may be evaluated in a number of different ways.The most commonly used methods KONA No.16 (1998) Fig. 4 A phase space plot obtained for a bulk lactose sample [94].
include [ 172], ultracentrifugation, impact and vibrational forces.Ultracentrifugation is ideal for detachment of particles in small size ranges.However, without attachment to a classifier for the detached particles, or selection of an advanced system such as the spiral centrifuge [153], the method is incapable of distinguishing the nature of the particles, dispersed or aggregated, once they are detached.Impaction and vibration techniques were originally considered for large particles, i.e., Dp> 100 11m [186,187].The impact force method has recently been modified to allow characterization of particles down to 111m [ 46,48].This method also characterizes the particle size by microscopy and by quartz crystal microbalance inertial impaction, allowing the degree of dispersion of the particles to be studied.It is possible to characterize detachment in terms of device powder interactions, not just particle-particle interactions, by adopting support surfaces of materials with which the aerosol powder will come into contact.
Particles suitable for deposition in the lung require an approximate aerodynamic diameter of 5 11m.Powders consisting of such small particles are difficult to fluidize due to their small cross-sectional area.Dry powder aerosol formulations typically consist of an interacting or ordered mixture of small drug particles associated with large lactose carrier particles [74,88].The purpose of this binary system is to optimize interparticulate forces to obtain an interactive mixture with a high degree of homogeneity that will respond to the fluidization and deaggregation processes.Diluents can also be used to stabilize the active ingredient [23,40].Introduction of a third component into the mixture establishes competition between small particles for binding sites [222].This can result in a higher proportion of drug particles binding to weaker sites and improved control of the fluidization/ deaggregation mechanisms.Addition of tertiary particles has been shown to increase the respirable fraction of the doses emitted from DPis [131].
Carrier-free powders have the advantage of smaller dosage forms, which can be significant when using potent or expensive drugs.The disadvantage of drugonly formulations is the difficulty in fluidizing the bulk powder and deaggregating particle agglomerates.Difficulty in fluidizing/ deaggregating small particles is due to the low drag coefficient (Dp 2 ) of the individual particles relative to the interparticulate forces, requiring large aerodynamic forces for fluidization.

Fluidization
Fluidization is the mobilization of the bulk powder by air.This represents the first stage interaction with air in the dry powder inhaler.There are two types of fluidization mechanisms: hydrodynamic and mechanical, which can be further divided into sub-groups.These include gas-assist, capillary and shear force for hydrodynamic fluidization, and vibration and impaction for mechanical fluidization.Powder fluidization is dependent on the ratio of hydrodynamic to interparticulate forces.Hydrodynamic forces take the form of competing effects of viscous drag (Dp 2 ) and inertial (Dp 3 ) forces.Interparticulate forces are approximately proportional to Dp.Thus, the interparticulate forces will decrease as the particle size decreases (Dp).However, greater hydrodynamic force in the form of viscous drag (Dp 2 ) will be required to mobilize the particle, i.e., high air flow rate.The relationship between interparticulate force and inertial force during mechanical fluidization is complicated by the direction of the force vector.
Powders can be characterized by their fluidization properties [75].Geldhart proposed four distinguishable groups of powder behavior during gas fluidization, three of which can be applied to powder inhalation aerosols.Group A describes fluidized powder beds in which the hydrodynamic forces dominate relative to interparticulate forces, making the powder easy to fluidize.This group usually consists of intermediate size particles (Dp~40 11m) and/ or low particle density ( < 1.4 g/ cm 3 ).Group B powders maintain a balance between interparticulate and hydrodynamic forces (Dp: 40-500 11m) and group C powders are difficult to fluidize due to the predominance of interparticulate forces (Dp< 10 11m).Powder fluidization properties can be defined by the Hausner ratio which is the ratio of poured to tapped bulk density [75].Fluidization is also dependent on the elasticity of the powder particles and viscosity of the gas phase, limiting the general applicability of Geldhart's powder groups [208].
Gas-assist fluidization occurs when a gas is passed through the powder.This form of fluidization can be described by considering a Geldhart group A powder in a gas-assist fluidization bed.This consists of a powder contained in a cylindrical vessel that is sealed at the bottom by a gas permeable plate and open to atmosphere at the top.Gas is forced through the powder via the permeable plate, generating drag forces on the powder particles.Gas flows through the powder pores without disrupting the powder structure at low gas flow rates.Incipient fluidization occurs with increasing gas flow rates, shown in Figure 5, resulting in aerodynamic drag forces separating loosely bound particles.Further increases in gas flow rate causes bulk particle separation.Homogeneous fluidization is achieved with further increases in gas flow rate, producing bulk particle separation.The bed becomes unstable at a critical gas flow rate as bubbles begin to form in the powder.The bed height fluctuates as the bubbles are continually produced and collapsed.More complex phenomena occur with other powder groups (B-D), e.g., powder compression and channeling [75,208].Active dry powder inhalers are being developed that utilize gas-assist fluidization [42,59,69,82,122,191,215]. Capillary fluidization is a mechanism similar to gas-assist and occurs when a low pressure field is established close to the powder, e.g., recirculation zone.A pressure differential between the air contained in the powder voids and the motion of the air passing over the powder is developed and the powder bed is fluidized.
Gas Flow Rate Shear force fluidization (also known as saltation) occurs when a gas stream is passed over a powder source, contained in either a pocket or on an open surface.Particles on the surface of the powder source experience reduced interparticulate forces due to the smaller coordination number.Separation by shear force results in the transmission of both translational and rotational motion of the particles as they traverse the powder bed.Collisions with other particles force the particles to bounce, resulting in incipient fluidization.Particles separate from the bulk powder with high rotational velocities (up to 1000 rev/s [38]), generating Saffman lift forces that project the particles vertically.The high viscous shear stresses in the boundary layer close to the powder surface magnify the vertical projection due to the Magnus force.This form of fluidization primarily affects particles with Dp> 100 J.lm and is dependent on the velocity of the air flow around the particle.The critical velocity for shear force fluidization is given as (for Dp> 100 J.lm) [11]: Shear force fluidization predominates in the majority of passive dry powder inhalers.
Mechanical fluidization can be generated by vibration or impaction.Vibrational fluidization occurs when the powder is shaken from its source and falls through air by gravitational forces.The method used to measure the angle of repose exploits vibrational fluidization.The Spinhaler™ was one of the first dry powder devices used for inhalation therapy and employs the principal of vibrational fluidization [16].Fluidization by impaction when a tangential force is applied to the powder source such that the interparticulate forces forming the bulk powder can be overcome.The impaction force required to fluidize a powder is dependent on particle size and orientation of the powder source.The impaction force must overcome both interparticulate, friction and gravitational forces to fluidized the powder vertically upwards.Inverting the powder source orientation results in the gravitational forces assisting fluidization.
Investigations have been conducted on the effect of particle size and fluidization for dry powder aerosols.Fluidization within the Spinhaler™ occurs by the vibration of a rotating capsule that has been punctured for the powder to pass through.Figure 6 is a representation of the effect of particle size on fluidization within the Spinhaler™, with maximum fluidization occuring for a particle size of approximately 100 !J.m.The lower particle size limit is due to small drag forces relative to the interparticulate forces.The large size limit is due to blockage within the capsule hole.Experiments with a fluidized bed generator revealed that large particles (>90 !J.m) produced smooth and continuous fluidization that was primarily dependent on mean air velocity (shear force fluidization) [144].Fluidization of smaller particles was dependent on mean air velocity, tube diameter and pressure gradient.This was due to the complex interaction between interparticulate and aerodynamic drag forces.These results imply an optimal particle size for maximum fluidization.

Deaggregation
Deaggregation and deagglomeration are the processes by which particle aggregates (carrier-drug particle admixtures) and agglomerates (drug clusters) are dispersed into primary particles, :S5!J.m in median diameter, by interaction with an airflow.This is considered to be the primary aerosol dispersion mechanism affecting lung deposition [123].The interparticulate forces of a binary system or agglomerate must be overcome for deagglomeration to occur.Adhesive/ cohesive forces are generally dependent on Dp, whilst forces of separation are dependent on Dp 2 (aerodynamic) and Dp 3 (vibration, rotation and collision).
Deaggregation principally occurs in two regions: the device and the oropharynx.Particle transport beyond the oropharynx affects dispersion of the powder aerosol due to changes in flow geometry and environmental conditions.This ultimately affects lung deposition.Factors affecting lung deposition have been detailed in the literature [151,152,230].
Principal forces leading to deaggregation are unclear.The majority of literature points to turbulence being the principal factor without considering the detailed nature of turbulent fluid flow and its interaction with dispersed particles [72,131,138,234,238,246].The forces contributing to the transport and deaggregation of particle agglomerates travelling in an airstream can be broken down into the following closely coupled categories: (i) relative motion; (ii) turbulence; (iii) shear stress; (iv) electrostatics and (v) collision.
The predominating forces affecting the relative motion of an aerosolized particle can be presented as a supposition of forces: The term on the left-hand side of Eq.( 8) represents the inertial force of the particle, with the terms on the right-hand side of the equality respectively representing the drag force, the force due to static pressure gradients in the flow, the Magnus and Saffman forces, the force experienced in the presence of a strong electric field gradient and finally the body force term, taking the form of gravity.The predominating force acting on a particle is usually the drag force.The pressure gradient term becomes significant in turbulent flows typically experienced by dry powder aerosols.Magnus force accounts for translational motion of the particle due to rotation which is significant immediately after shear force fluidization.Saffman force results from free rotation within gas velocity gradients.This force would be experienced close to the powder source in the viscous sub-layer, boundary outer-layers and within swirling eddies.The electrophoresis term is significant for small ( <5 Jlm), charged particles in strong electric fields.
Rotation of carrier particles or agglomerates generates centrifugal forces on surface particles that may be large enough to overcome the adhesive forces.The centrifugal force acting on a small particle can be approximated by considering the particle as a point mass on the periphery of a large carrier particle, as follows: 2 F = 0.5 mpwpDc (9) where De is the diameter of the carrier particle.A binary particle system consisting of a 90 Jlm lactose carrier particle with a 2 Jlm albuterol sulphate drug particle attached to the surface requires an cop>840 revls to overcome the van der Waals force of attraction estimated in Table 1.Fluidized sand grains (Dp-200-300 Jlm) have been observed to rotate with high angular velocities (200-1000 rev Is) [38].The forces of attraction estimated in Table 1 are much higher than those experienced by a 'real' binary drugcarrier system due to surface asperities.For a binary system consisting of particles that have asperities with effective diameters of 0.1 Jlm, the van der Waals force is reduced by an order of 40, requiring an cop> 130 rev Is for centrifugal separation.
There are three types of fluid flow: laminar, turbulent and transition.Laminar flow usually occurs at low fluid flow velocities and is characterized by stable instantaneous velocities everywhere in the flow field.Turbulence can be defined as a spatially varying mean fluid flow with superimposed three-dimensional random fluctuations which are self-sustaining and enhance mixing, diffusion, entrainment and dissipation [250].Turbulent fluid flow consists of instabilities known as eddies that occur with various length scales.Large scale turbulence consists of large eddies that contribute to the convective transport of the fluid.Large scale eddies are unstable and break down to fine-scale turbulence that dissipate energy by heat transfer.Transition describes the unstable condition between laminar and turbulent fluid flow.The Reynolds number (Re) is the ratio of inertial to viscous forces: (10) where Uc and Lc are characteristic velocity and length scales of the fluid flow, respectively.The characteristic scales used to define the Re of fluid flow within a cylindrical pipe are the velocity of the gas and the diameter of the pipe.The characteristic scales for particle motion are the relative velocity of the particle and particle diameter.Re is commonly used to predict the onset of turbulence, e.g., turbulence occurs at Re""'2000 for fully developed flows in cylindrical pipes.The value of Re at which turbulence first occurs is dependent upon geometry, time scales and initial disturbances.The majority of DPis are composed of short tubes and complex geometries, suggesting the flow through such devices is in an intermediate state, consisting of a turbulent core surrounded by a laminar envelope.Neither flow type will dominate due to the short time scales of the fluid motion through the device.Definition of the type of flow containing suspended particles is further complicated by turbulence modulation.This occurs due to the conflicting effect of particles damping the turbulent flow field by their
inertia as well as instigating instabilities in the flow leading to turbulence production.This process has yet to be fully understood [63,164].Flow development around a spherical particle is summarized in Figure 7.The fluid flowing past the sphere exerts a drag force due to skin friction (low Rep) and pressure gradients along the surface.Viscous forces dominate at very low particle Reynolds numbers (Rep<0.1),known as creeping flow or Stokes' regime.As Rep increases ( ~ 10), separation of the boundary layer occurs and vortices (large scale eddies) are formed downstream of the particle that are eventually dissipated by the viscous action of the mean flow.Further increases in Rep result in unstable vortex generation and the production of a wake.At a critical Rep ( ~ 10 2 ), the eddies break off from alternate perimeters of the particle (vortex shedding) and form an oscillating vortex street.Vortex shedding results in alternating lateral forces and may cause forced vibration of the particle if the vortex shedding frequency approaches a resonant frequency of the particle.The limiting values of Rep, described above, are reduced by increased turbulence of the mean air flow, particle asymmetry, rugosity and the presence of other particles disturbing the mean flow.The Rep for turbulent flow around a particle is approximately 400 (Figure 7) .Shear stresses are established by velocity gradients, defined by: .. __ (JUi+ J~) Efforts have recently been made to relate the shear stress required to deaggregate particles in turbulent flow through a dry powder aerosol device [64] and the oropharynx [140].The volume flow rate (L/min) required to produce particle deagglomeration was derived by equating the viscous shear force based on a turbulent mixing length hypothesis to the adhesive van der Waals force [64]: The second approach estimated the viscous shear force in a three dimensional replicate oropharyngeal geometry using computational fluid dynamics.An expression for the minimum diameter of identical particles that can be separated was obtained by equating the viscous shear force to the adhesive van der Waals force: Both expressions for particle deaggregation highlight the relationship between gas flow rate and particle detachment.
A particle can acquire charge during deaggregation by two mechanisms: static electrification and diffusion charging.Each of these methods affects the transport and deaggregation of particles within a DPI.Static electrification causes particles to become charged as they are separated from the bulk material or other surfaces.This mechanism can significantly affect formulation homogeneity and fluidization, especially when the powder contacts dissimilar materials [193,225].A discharge occurs when particles are separated during deaggregation.The electrostatic charge acquired by a particle changes with time due to diffusion charging.Diffusion charging occurs due to random collisions with unipolar ions and particles.The rate of particle charging tends to zero as the particle produces an electrical field that will eventually be strong enough to repel additional ions.Diffusion charging is dependent on the physical and chemical characteristics of the particle, charge and charging time.As particles are charged by diffusion, some will deposit on the surface of the device and discharge.A sufficient number of particles depositing on the device surface could lead to the generation of an electrical field.This effects the transport of the fluidized powder as particles are attracted to (and repelled from) each other and the device surfaces.
Particle collision is an important phenomenon for aerosol dispersion.Particle collision is dependent on particle size, number concentration and particle to particle and gas to particle relative motion.The number concentration and relative velocities are highest during fluidization and therefore the rate of collision is most significant during this process.Deaggregation can occur during collisions of solid, aerosol particles, providing the particles have sufficient relative momentum.Small particles can also adhere to other particles if there is sufficient settling time for the interparticulate forces to overcome the kinetic energy of the particle.Particle collision is a very complex process and is difficult to quantify a priori, especially for polydisperse systems interacting with turbulent fluid flow.Although models of collision have been developed [181], more research in this area is required before the effect of particle collision can be quantified and exploited for dry powder aerosol dispersion.

Mechanisms of Dispersion by Inhalers
Dry powder inhalers are designed to generate an aerosol from a powder onto or via the inspiratory airflow.The powder formulation consists of the drug which is usually blended with a carrier (lactose or glucose) to facilitate flow and dispersion [71,72,234].The formulation is either contained as bulk material in a reservoir within the inhaler (multi-dose inhalers) or packed as unit doses in capsules (single dose inhalers) or blisters (multiple unit-dose inhalers).In the case of a powder reservoir, a single dose has to be metered from the bulk and transferred into the inhalation channel prior to inhalation.Pre-metered unit dose packages need to be opened to render the powder accessible for inhalation.
The powder in its static state is characterized by a number of cohesive and adhesive forces (section 3).These consist of van der Waals, electrostatic and capillary forces, mechanical interlocking and friction.Van der Waals forces dominate at low relative humidities (RH<65%) whilst capillary forces control cohesion/ adhesion at RH>65%.The extent of these forces is dependent on the powder properties (size, shape, geometry) [34, 93, 99, 100, 144, 242], the formulation (excipients) [30,64,224,225,228], and ambient conditions (humidity, electrostatics) [28,76,193,200,222,224].Generation of an aerosol in the respirable size range requires the fluidization and deaggregation of the static powder.The cohesive and adhesive forces within the powder have to be overcome by an external supply of energy.This is achieved in passive inhaler devices by the transfer of kinetic energy from the inhaled air stream to the powder, whereas active devices use an additional source of energy, i.e., compressed air or mechanical force.The extent and efficiency of this energy transfer depend upon the construction of the inhaler as well as the inspiratory flow of the patient and ultimately determine the particle size distribution of the aerosol generated [71,72,100].

Dispersion in Passive Dry Powder Inhalers
Passive dry powder inhalers rely on the inspiratory flow as the sole source of energy for powder fluidization and particle deagglomeration.The powder formulation is metered and/ or released from the site of storage and primarily fluidized by shear force.Particle deagglomeration can occur due to a variety of mechanisms that include relative motion (the different motion of the particles relative to the air stream), turbulence, shear stress and collision.Each mechanism occurs to a different extent in all inhalers.In the following section, devices used for the generation of dry powder aerosols are described and an attempt is made to pinpoint the dominant mechanisms of powder fluidization and particle deagglomeration (Table 2).Fine particle fractions as defined in section 3.3.5 are quoted from the literature to indicate

Spinhaler™
The Spinhaler™ (Fisons, UK) is a unit dose powder inhaler designed for the inhalation of drugs from gelatin capsules [Figure 8a].The capsules contain 20 mg of micronized drug without carrier and are mounted onto an impeller in the inhaler.A piercing mechanism punctures the inserted capsule.Inhaled air causes the impeller and thus the capsule to rotate at a speed that is dependent on the air flow rate.At sufficiently high flow rates the capsule vibrates leading to mechanical fluidization [16].This is supported by capillary fluidization due to the pressure drop across the capsule.The powder is conveyed along the capsule walls towards the perforations where it is

Rotahaler™
The Rotahaler™ (GlaxoWellcome, UK) is a single dose device for the inhalation of powder (lactose blends) from gelatin capsules [Figure 8b tion is mainly caused by turbulence promoted by the grid upstream of the mouthpiece.A fine particle fraction of 26% has been reported for this device actuated at 56.5 L/min [190].

Inhalator Ingelheim™
The single dose Inhalator Ingelheim TM (Boehringer Ingelheim, Germany) is designed for the inhalation of drug-glucose blends from gelatin capsules.A capsule is inserted into a narrow container in the inhaler.The capsule is pierced at both ends by pressing a button on the inhaler.In contrast to the Spinhaler™, the capsule is stationary and emptying occurs primarily by capillary fluidization due to the high pressure drop across the capsule.Deaggregation is caused by shear stress and collision.Fine particle fractions ( <6.4 pm) for oxitropium bromide have been reported to be approximately 27% at 30 and 60 L/min [228].

Inhalator MfM
The Inhalator MfM (Boehringer Ingelheim, Germany) is a multiple unit dose powder inhaler similar to the single dose Inhalator Ingelheim™ [Figure 8c].In contrast to the above, however, Inhalator M™ has a rotating drum magazine for the storage of six capsules.The mechanisms for capsule emptying, powder fluidization and particle deagglomeration are the same due to the construction similarities.Fine particle fractions of 28 and 36%, in the twin and multistage impinger, have been reported for ipratropium bromide [228].

Diskhaler™
The Diskhaler™ (GiaxoWellcome, UK) is a multiple unit dose inhaler.The inhaler housing contains a magazine and a rotating disk that hold the Rotadisl(fM, a disk-shaped blister pack containing four or eight single doses of powder [Figure 8d].The powder formulation consists of drug blended with lactose.When the hood of the inhaler housing is folded up, a blister is perforated by a pin and the powder becomes accessible for inhalation.As the inhaled air flows across the powder bed, the powder is aerosolized primarily by shear force fluidization and transported to the mouthpiece.The aerosol stream is then mixed with a bypass flow entering through two holes in the mouthpiece.The additional airflow and a grid in the mouthpiece give rise to turbulence that promote deagglomeration [229].Fine particle fractions (<6.4 pm) for the albuterol sulfate formulation Sultanol™ 200 Rotadisk™ were reported to be approximately 30% [114,228].A fine particle fraction (<7 pm) of 23% was measured for the Becotide™ Rotadisks™ [55].20

Diskus™
The Diskus™ (GlaxoWellcome, UK) is a multiple unit dose inhaler that contains a foil strip with 60 single dose blisters of drug blended with lactose [Figure 9a].A blister is transported towards the mouthpiece and the lid foil is stripped off prior to inhalation.Air is channeled through the inhaler and the powder is aerosolized by shear force fluidization.Additional air is drawn through two holes in the mouthpiece, similar to the Diskhaler™, to augment particle deaggregation by turbulence [21].The Flutide™ Diskus™ was reported to deliver a fine particle fraction ( <6.4 pm) of 25-30% fluticasone-17-propionate at 60 L/min [228].

Turbuhaler™
The Turbuhaler™ (Astra, Sweden) is a multiple dose inhaler with 200 doses of drug stored in a reservoir [Figure 9b].The micronized drug is spheronized into small pellets rather than blended with a carrier.The pellets are stable in terms of transport and handling but disintegrate into their primary particles during metering and inhalation [249].The dosing unit is caused to rotate by turning the grip at the bottom of the inhaler.Scrapers actively force drug from the reservoir into conical holes of the dosing unit.The pellets disintegrate and a single dose of drug is metered and ready for inhalation.Air enters the inhaler and passes through the dosing unit across the powder bed during inhalation, fluidizing the powder by shear force.Additional capillary fluidization occurs due to the direction of the air flow entering the inhaler.After fluidization, particle deagglomeration occurs by three mechanisms: turbulence in the narrow inhalation channel, impaction on the bottom of the mouthpiece, and predominantly high shear stress in the swirl nozzle of the mouthpiece [121,248].Different studies with the Aerodur™ Turbuhaler™ showed fractions of 47% [114], 50% [121], 32 and 35% [228] terbutaline sulfate below 6.4 pm .Fine particle fractions of 39 -45% budesonide have been reported for the Pulmicort™ Turbuhaler™ [55,182,228].

Easyhaler™
The Easyhaler™ (Orion Farmos, Finland) is a multiple dose powder inhaler with 200 doses of druglactose blend stored in a reservoir [Figure 9c].The metering system is based on gravitational flow of the powder from the reservoir into a cavity in the metering cylinder.Potential energy is supplied by shaking the inhaler to assist flow from the reservoir.Depression of the device overcap, similar to a pressurized metered-dose inhaler, causes the metering cylinder to rotate and a single dose of powder is transported towards the mouthpiece.Inhaled air passes along the powder bed in the metering cavity and fluidizes the powder by shear force.Turbulence in the long narrow inhaler mouthpiece causes further particle deaggregation.Fine particle fractions ( <6.4 \liD) of 32"/o [114], 36-39% [228] and 45% [240] at 60 L/min have been reported for a salbutamol powder blend.

Mechanical Aerosol Generator™
The multi-dose Mechanical Aerosol Generator™, or MAGhaler™, (Gesellschaft fur Gesundheits und Umweltforschung GmbH, Germany) is a multiple unit dose inhaler [Figure 9d].The principle of operation differs substantially from other passive devices.The KONA No. 16 (1998) drug is not supplied in the form of a powder, but as a ring-shaped tablet that contains 200 to 500 doses of micronized drug and lactose, compressed by high pressure isostatic compaction.The inner bore of the ring comprises the inhalation channel.Dose metering and powder dispersion are essentially one step, performed by a motor driven ceramic scraper.A fine disperse powder is scraped off the surface of the tablet (mechanical fluidization) into the inhalation channel during inhalation.The aerosol generated is very dilute and is inhaled in statu nascendi, before the particles can aggregate.Relative motion and turbulence promote deaggregation.Fine particle fractions (<6.4 J.Im) between 30 and 50% have been stated for a variety of drugs by the manufacturer.

Dispersion in Active Dry Powder Inhalers
Active dry powder inhalers have an additional source of energy than that provided by inhalation, using auxiliary means to fluidize and disperse the powder.The basic principles for powder fluidization and dispersion are the same as passive devices, however, the additional energy input decreases or eliminates the dependence of the aerosolization process on the patient's breathing pattern and reduces dose and particle size variabilities [47].A number of devices are currently under development though to date none have been approved [42,59,69,82,122,191,215]. Hence, no drug formulations for these devices are on the market but studies using both carrier blends and unblended drugs have been published [25,47,104,161,162,167].
Plunger Mixing Chamber

Spiros™ Dry Powder Inhalation System
Spiros™ (Dura Pharmaceuticals, San Diego, USA) is a power assisted single-unit or multiple-unit dose (blister, cassette) inhalation system [Figure lOa].A mechanical plunger ejects the drug from the reservoir into a mixing chamber (mechanical fluidization).A breath activated, motor driven impeller, rotating at 12000 to 14000 rpm, provides electromechanical energy to disperse the powder formulation by shear force, turbulence and impaction [59,161].This mechanism greatly reduces the effect of inspiratory flow on the fine particle fraction.Fine particle fractions between 26 and 32% have been reported using the albuterol sulfate lactose blend from Rotacaps™ at flow rates of 15 to 60 L/min [104].

Prohaler™
The Prohaler™ (Valois Pharm, France) is a multi dose powder inhaler where a built-in pump provides compressed air to facilitate dose metering and powder dispersion [Figure lOb].Air assisted powder fluidization reduces the influence of inspiratory flow on fine particle fraction and ensures effective powder dispersion [25].Fine particle fractions ( <6 pm) for a mannitol blend of the leukotriene antagonist SKF-104353 were reported to be 68% [167].

Dynamic Powder Disperser™
The Dynamic Powder Dispenser (Pfeiffer GmbH, Germany) is a multiple unit dose inhaler featuring an exchangable cartridge that holds 12 dosing chambers [Figure lOc].The inhaler is primed by an axially mobile piston that compresses air in the inhaler.The compressed air in the device is expelled through the dosing chamber in the cartridge upon manual actuation and an aerosol is generated by gas assist fluidization.Studies with a 2% albuterol sulfate blend showed fine particle fractions (<5 pm) of up to 40% [47].

Inhale Deep Pulmonary Drug Delivery
System The Inhale deep pulmonary drug delivery system (Inhale Therapeutic Systems, USA) is a gas-assisted dry powder inhaler designed mainly for the systemic delivery of insulin and other biomolecules through the lung [Figure lOd].Insulin is spray dried together with lactose acting as a stabilizing agent.The resulting fine powder (1-3 pm) is packaged in unit dose blisters [ 118,191].The base unit of the inhaler contains a mechanical air compression mechanism, a lever to transport and puncture the blister package and a transjector.Prior to inhalation, the inhaler is cocked and a small amount of air is compressed inside the inhaler.A blister is inserted, elevated to the transjector and punctured.When the inhaler is actuated, the compressed air is released through the transjector at sonic velocity.The powder is withdrawn from the blister during this process and aerosolized by gas-assist fluidization.T~rbulence and shear force are the main factors for further deagglomeration.The aerosol is then released into a holding chamber with a combined mouthpiece/ chamber cap on top from where it can be inhaled.

Inhaler Performance Criteria
Particles need to have aerodynamic diameters below 5 pm for deposition in the lung [26,89,232].The fine particle fraction delivered from a dry powder inhaler is determined by three independent factors: the powder formulation, the inhaler and the patient [72].Ideally, an optimal dry powder inhalation would reproducibly generate the same emitted dose and fine particle fraction in order to induce an equivalent pharmacological effect in all patients.In reality, both emitted dose and fine particle fraction vary, depending on patient specific parameters (inspiratory flow rate, handling) and ambient conditions (humidity, electrostatics).
Both powder and air flow have been characterized in a number of idealized systems.Static and dynamic angle of repose and flow from a vibrating spatula have been used to assess global flow parameters of powders [94], as described in section 2.2.Airflow has been characterized in detail in the engineering literature [154,250].However, the complex relationship of air flow and particle dispersion in dry powder inhalers is just beginning to be understood and a number of methods have been proposed to gather information that is relevant for these heterogeneous systems [39,93,95,132,133,144,200,224,256].In the following, the relevant criteria influencing the performance of dry powder inhalers are discussed.

Inhaler Flow Resistance
The specific flow resistance of a device is related to pressure difference and volumetric flow rate, as follows: R = ,j~p (14) where R is the specific flow resistance, Q is the air flow rate and boP is the pressure drop across the device [188].R is dependent on the internal geometry of the inhaler and the dimensions of the air channels.A wide range of values for a variety of commercially available dry powder inhalers has been reported and is summarized in Table 3 [39,228].The flow rate through an inhaler can be approximated as the ratio of the square root of the minimum inspiratory mouth pressure and the device resistance, the resistance of most inhalers exceeding that of the airways [ 41].Minimum inspiratory oropharyngeal pressures reported in the literature are generally in the range of 110-140 em H20 for normal, healthy individuals, slightly lower (90-115 em H20) for cystic fibrosis patients and still lower (60-70 em H20) for patients with chronic obstructive pulmonary disease [41].This has important consequences.The flow rate through an inhaler at a constant inspiratory effort (pressure differential) will be device dependent.Also, the flow rate through a specific inhaler will be dependent on the patients' inspiratory effort.The air flow generated by the patients' inspiratory effort is, however, the sole source of energy for powder fluidization and dispersion in passive dry powder inhalers.Hence, the accuracy of the dose and the fine particle fraction delivered by these devices depend on the specific resistance of the inhaler and the inspiratory flow rate generated by the patient [40,55,107,174,176,257].This has further implications in the selection of the appropriate in vitro testing methods.Dry powder inhalers have to be tested at flow rates that match their specific resistance to yield meaningful results.It has been proposed that high resistance inhalers (>0.12 em Hz0 1 12/L min-1 ) should be tested at a flow rate of 30 L/min, low resistance devices (<0.07 em H20 1 I2/L min-1 ) at 100 L/min and inhalers with intermediate resistances at 60 L/min [29,39,109].

Powder Formulation
Formulation of powders suitable for inhalation require that particles are of sufficiently small size to reach the target site in the lung (Dae<5 j.lm).However, the intrinsic cohesiveness of such fine powders opposes accurate metering, gives rise to the formation of agglomerates, and impedes the production of primary respirable particles.A variety of approaches have been taken to overcome this challenge and to optimize powder formulations for delivery to the lung in parallel to the development of active inhalers.
A coarse, inert carrier such as lactose or glucose is frequently blended with the drug to facilitate dispersion.In these interactive or ordered mixtures, the fine drug particles adhere to the surface of the carrier [88].The adhesion force of the drug to the carrier must be sufficient to avoid demixing, but small enough to allow detachment during inhalation [71].It has been shown that the adhesion between drug and carrier and the fine particle fraction generated from these mixtures can be influenced by a variety of factors.These include the surface properties of both drug and carrier, drug to carrier ratio, carrier particle size, mixing time and method, humidity and electrostatic behavior.Coating cromolyn sodium particles with fatty acids led to changes in the particle morphology and increased the fine particle fraction in the twin impinger approximately threefold [ 68].The surface roughness (rugosity) of carriers has a pronounced influence on the adhesion of small particles.Blends using recrystallized lactose with low rugosity gave significantly higher fine particle fractions than crystalline and spray-dried lactose with higher rugosity [127,226,258].Increasing the drug to carrier ratio has been shown to increase the respirable fraction [48,71,213], but the opposite effect has also been reported [228].Decreasing the carrier particle size has been shown to increase the fine particle fraction of a powder blend [ 48,128,129,228].The conditions under which a powder blend is processed influence the fine particle fraction.Blending time promotes mechanical interlocking and friction [135], whilst the blending procedure can strongly influence the fine particle fraction [44].The addition of a ternary component such as magnesium stearate, fine lactose or L-leucine to interactive mixtures can destabilize the mixture and increase the respirable fraction [ 44,128,145,192,226].
Spray drying of the drug has been used as an alternative to jet milling in the production of fine powders.The advantages of this method are control of the particle size, shape and surface properties [212].Vidgren et al. showed that a powder formulation of spray dried sodium cromoglycate produced a significantly higher fine particle fraction than micronized drug [239].The morphology of spray dried particles has been varied systematically by adding sodium chloride, the fine particle fraction changing proportionally with the salt content [34].
The formulation of carrier-free powder aerosols is a different approach to interactive mixing with carriers.Micronized particles can be agglomerated into soft pellets with small amounts of solvents using conventional granulation methods, which is applied in the Turbuhaler™.These pellets have good flow properties and can be engineered to disintegrate into respirable particles during inhalation [130,249].A novel approach to particle engineering has recently been reported by producing large, porous particles with mean geometric diameters up to 30 \Jffi.These particles have good flow properties and are easy to disperse, with an aerodynamic diameter that is small enough to be deposited in the lungs, due to the low particle density [60].

Humidity
Humidity can have a profound effect on the performance of a dry powder aerosol.This can be attributed to two factors: capillary condensation and hygroscopic growth.Capillary condensation between particles in close proximity can occur around or above 65% relative humidity [214,242].The resulting capillary force surpasses all other interparticulate forces and strongly adds to particle adhesion (Table 1, section 2.1) [19,200].Additionally, particles absorb moisture when the ambient humidity exceeds the critical relative humidity of hygroscopic compounds.Hygroscopic growth occurs until the particles start to dissolve and form droplets [96,123,184].Capillary condensation and hygroscopic growth strongly reduce the fine particle fraction of powder aerosols.Fine particle fractions of cromolyn sodium, albuterol and albuterol sulfate powder aerosols were reduced at high temperatures and humidities [25,123,124,220].Similar effects were reported for the Buventol Easyhaler™, the In tal Spinhaler™ and the Pulmicort Turbuhaler™, though not for the Ventolin Rotahaler™ [109,114].Salmeterol xinafoate lactose blends developed strong capillary forces during storage at relative humidities above 75%, that were irreversible after storage at low humidities [200].The metering and the dose emission of dry powder inhalers can also be affected by humidity, as the capillary forces impede the flow properties of the powder [109,114].
Protective measures have to be taken to minimize the detrimental effects of humidity on the performance of dry powder aerosols.This can be achieved by using desiccants that adsorb moisture from the air close to the powder formulation (Turbuhaler™).The access of humidity to the powder can be prevented by a protective housing (Turbuhaler™, Easyhaler™) or by hermetically sealing the powder in aluminum blisters (Rotadisk, Diskus™, Spiros™, Inhale device).Another approach is the development of formulations that are less sensitive to humidity.This has been achieved by modifying the particle surface with hydrophobic agents [97], the use of suitable carriers such as lactose [109] or by selecting hydrophobic salts of the drug.Fine particle fractions of free albuterol base and albuterol sulfate both decreased with increased humidity and temperature, whereas the hydrophobic salts, albuterol adipate diethanolate and albuterol stearate, remained largely unaffected [123].

Electrostatics
Two main mechanisms can lead to the formation of electrostatic charges in dry powders.When dry, non-KONA No. 16 (1998) conducting particles come into contact with other particles or surfaces, triboelectric charging occurs.When particles of different work functions are brought into contact, or are detached, a potential difference is formed.The resulting electrostatic forces can contribute significantly to the cohesion of particles in a gaseous environment.However, electrostatic charging is inhibited by humidity due to the adsorption of water on the particle surface which ultimately leads to the system decharging at humidities above 50-60% [125,242].Consequently electrostatic and capillary forces in a powder are mutually exclusive.This leaves a narrow window of optimal relative humiditiy (depending on the powder) where the effects of electrostatic charging and capillary condensation are minimized.
Electrostatic charges have influence on all aspects of powder formulation, affecting the flow properties, powder mixing and particle separation [222,224].Deposition of particles in the lung is also affected by aerosol charge [12,87,163] and drug retention in inhalers is related to the charge on the devices [14,15,180].The electrostatic properties of drugs and blends of albuterol sulphate with lactose in two inhalers have been analyzed.It was proposed that blending powders in metal vessels would aid powder homogeneity, while the plastic of the DPI devices should enhance the separation of drug particles from the carrier [193].The fine particle charge on budesonide and terbutaline sulphate aerosols emitted from the Turbuhaler™ and on aerosols generated with the Dryhaler™ has recently been determined [28].While no dependence of the fine particle fraction on the charge was reported, it was shown that charges depended on the powder investigated as well as on the inhaler device and its deagglomeration mechanism [28].

Emitted Dose Characteristics
The quality of dry powder aerosol dispersion is determined by the uniformity of the emitted dose and the fine particle dose.As discussed previously, these parameters depend on the inhaler device, its flow resistance, the powder formulation, humidity, electrostatic effects and last but not least, the patient.
The term "dose" requires clarification as there has been considerable debate and confusion regarding the terms metered-, delivered-, emitted-, respirable-, fine particle dose and fine particle fraction [Figure 11].The dose dispensed from the bulk powder by the metering mechanism of a multiple dose inhaler, or the dose contained in pre-packed units in single or multiple-unit dose inhalers is generally defined as the metered dose [73].The dose delivered by the inhaler to the patient is called the emitted dose or delivered dose.Hence, emitted dose equals the metered dose minus the fractions of the dose remaining in mouthpiece, inhaler, metering chamber and unit dose container.The emitted dose is composed of particles of varying size that will partially deposit in the airways depending on their aerodynamic characteristics, inhalation characteristics and airway geometry.The terms "respirable" versus "non-respirable" dose are often used to further characterize the emitted dose.This may lead to confusion as the term "respirable" neither defines a certain particle size nor indicates a specific site of deposition, but simply indicates inhalation and transport into the airways beyond the oropharyngeal region.Further consternation is added as the terms fine particle fraction and dose have specific definitions relating to inhalation hazards in the field of occupational and environmental medicine [110].Moreover, for development and quality control purposes, inhalers are generally characterized in vitro by methods based upon impaction, whereas a number of additional mechanisms contribute to particle deposition in vivo [89,141,232].Hence, the term "respirable" cannot legitimately be used for fractions determined by in vitro deposition.In contrast, the term fine particle aerosol dose, or fraction, appears superior, as it does not promote speculative interpretation, especially when stated with the respective particle size range.The fine particle dose is thus defined as the dose that can be inhaled and is theoretically available for pharmacological activity.It has become the convention to consider the fine particle dose (FPD) rather than the fine particle fraction (FPF), since it is the dose which will result in a pharmacological effect.FPF is useful to compare the dispersion properties of different inhalers or formulations, but does not allow conclusions to be drawn in terms of pharmacological activity unless quoted with the emitted dose.Few comparative studies characterizing the emitted dose of commercially available dry powder inhalers have been published.The dose accuracy and reproducibility of the Spinhaler™, Turbuhaler™ and Diskhaler™ at different flow rates as recommended by the USP advisory panel on aerosols [27] have been determined [107,109].Steckel and Muller focused on the fine particle dose of 11 commonly used dry powder inhalers at different flow rates using a twin and four stage impinger, listing the delivered and emitted doses found at 60 L/min [228].The results are summarized in Table 4. De Boer et at.determined the effect of inspiration time and flow rate on the dose discharge of Turbuhaler™, Diskhaler™ and Spinhaler™ (since no numbers are given, these data are not included in [Table 4]) [55].Generally, the emitted doses differed substantially from the label claim and increased with higher flow rates.Exceptions were the Easyhaler™ and the inhalers using cromolyn sodium capsules, for which label claims and emitted dose were in reasonable agreement (± 15%).The emitted dose for the Spinhaler™ decreased at a higher flow rate.Unit dose variations were generally below ± 15% [107] and thus meet the criteria presently advocated [29].with the exception of the Turbuhaler™.The metering of small doses of micronized powders is difficult, especially if the powders are unblended, due to their intrinsically poor flow properties.This is particularly relevant for the Turbuhaler™ which has a powder load of only 0.5 mg, leading to variations in dose uniformity.Nevertheless, the Turbuhaler™ has been an effective device in Europe and has recently been approved in the USA [51,185,218].It has been postulated that the dose variabilities of dry powder inhalers may be within the extreme variability of drug deposition in the lungs [107].
The fine particle fractions emitted from dry powder inhalers and their dependence on flow have been analyzed in numerous studies [40,55,104,114,115,121,160,174,176,228,240].Generally, the fine particle fraction increases with air flow rate.The strength of this effect depends on the formulation and the device.Air flow rate is of particular importance for medium  [107,228].Results determined at the flow rates proposed by the USP advisory panel on aerosols [27]  and high resistance inhalers.A reduction of the air flow rate from 60 to 30 L/min resulted in a decrease of the fine particle fraction of the Turbuhaler™ to approximately one third of the original value in vitro [55,160].In vivo studies for the same device with flow rates of 57 and 28 L/min showed reduction of pulmonary deposition of terbutaline sulfate to approximately 50%, though the pharmacological response was not significantly affected.A strong effect of the flow rate on the fine particle fraction in vitro has also been shown for the Spinhaler™ with unblended cromolyn sodium.The fine particle fraction approximately doubled when the flow rate was increased from 30 to 60 and from 60 to 100 L/min [55,228].In contrast, the effect for cromolyn sodium lactose blends was less pronounced [228].A strong increase in the fine particle fraction between 20 and 60 L/min was reported for the beclomethasone dipropionate Diskhaler™ [55], while in another study with the salbutamol Diskhaler™ no major change was observed between 60 and 90 L/min [228].This may be attributed to the different powder formulations.Only small variations of the fine particle dose with flow rate were reported for Inhalator M™, Inhalator Ingelheim™ and the EasyhalerTM [228].Few data for active dry powder inhalers are available, but it is anticipated that the fine particle fraction will be far less dependent on air flow rate, as shown for the Spiros™ inhaler [104].
Comparing these findings is difficult for a number of reasons.Firstly, the test conditions used vary widely in regard to impactor type, inlet port, onset of flow, flow rates, sampling time, environmental conditions etc., and it is known that each of these can influence the fine particle fractions determined.Secondly, the definition of the fine particle fraction itself may lead to confusion, as it depends on the impactor type and the flow rate.The mean cut-off diameter of an impactor is a function of the flow rate.Increasing the flow rate shifts the cut-off diameter of an impactor stage and the corresponding fine particle fractions towards smaller diameters [110,252], a factor that is often ignored in the pharmaceutical literature.
With an awareness of these limitations, Table 5 offers a comparison of the published data.A relative fine particle fraction (rFPF) is defined as the ratio of the fine particle fraction at the low flow rate to the respective fraction at the high flow rate, expressed in percent.The rFPF is a simplified measure of the effect of the flow rate on the fine particle dose of the inhaler, indicating the change in fine particle fraction when the flow rate is reduced.The cut-off diameters were recalculated for the impactors at the flow rates used, to take into account the effect of flow rate on the fine particle size range and the data included in the table [8].
The rFPF values at reduced flow rates for the different inhalers and powder formulations vary.Four groups of inhalers can be determined.The Spinhaler™, Table 5 Effect of flow rate on the fine particle fraction (FPF) in different dry powder inhalers.The relative fine particle fraction (rFPF) is the ratio of the fine particle fractions at the low and the high flow rate.expressed in percent.The size range of the fine particle fractions has been recalculated (in italics), unless stated by the authors.

Inhaler
SpinhalerH 1 !SF lnhaler the beclomethasone Diskhaler™ and the Turbuhaler™, produced rFPFs ::;50%.This implies that for these inhalers and respective powder formulations the fine particle dose is strongly reduced with decreasing flow rate.For a second group of inhalers, the beclomethasone Diskhaler™, the Easyhaler and Inhalator Ingelheim ™, the relative fine particle fractions were between 60-80%, indicating an intermediate effect of flow rate on fine particle fraction.The salbutamol Diskhaler™, Inhalator M and Inhalator Ingelheim have rFPF values of approximately 90%, indicating a weak effect of flow rate on particle dispersion.Finally, for the active inhalers Spiros™ and Prohaler™, little effect of flow on fine particle fraction can be seen, with the exception of Spiros™ using the Rotacaps™ blend.The rFPF values above 100% can be attributed to the increase in the impinger's cut-off diameter and thus do not imply a better dispersion at lower flow rates.It is intersting to note that rFPFs calculated from lung deposition data in vivo for the Turbuhaler™ [176] and the Spinhaler™ [176]  with the rFPF from in vitro deposition studies.Despite the inherent simplifications and limitations in deriving rFPF values from different studies, it is a useful parameter when comparing the susceptability of inhalers and powder formulations to changes in flow rate.

Inertial methods
The first inertial samplers were developed in the 1940s [155].These methods were quickly adopted for evaluation of environmental and occupational exposure to airborne hazardous materials [143].Inertial methods of particle size measurement are considered the most relevant for the description of pharmaceutical aerosols for a number of reasons.The estimation of particle size based on mass of the whole aerosol as determined by chemical analysis, such as spectrophotometry, is necesssary from a therapeutic and regula-tory perspective [1,27].In addition, determination of size based on aerodynamic behavior is of relevant to particles intended for drug delivery to the lung.Deposition in the lungs occurs predominantly by inertia, similar to cascade impaction.However, lung deposition also involves other mechanisms such as interception, sedimentation, diffusion and other less significant forces [ 173].

Theory
Inertial methods of aerosol particle sampling are conducted by passing the airflow and suspended aerosol through an orifice of known dimensions and impinging it onto a collection surface at a fixed distance from the orifice.The Stokes number (St) is the ratio of the persistence of the particle to travel in the direction of flow to the magnitude of the barrier [206].(15) where pp is the particle density, 11 is the air viscosity, <)> is the jet diameter, Dp is the particle diameter, Up is the linear velocity and C is Cunningham slip correction factor.For spherical orifices the cut-off diameters occur at a St 05 of 0.47.
All impactors are made to well-defined design specifications.Thus, the geometrical factors that influence the collection efficiency at a stage are constant, i.e., orifice diameter and orifice to surface distance.Flow rate is the only remaining variable influencing the collection efficiency of each stage of an impactor.The theoretical collection efficiency curve is a step function of percentage collection versus particle size [110].From this curve there is a discrete size at which a particle population is collected with 100% efficiency and below which particles are not collected.The actual curve is sigmoidal and fractional collection efficiencies occur over a range of particle sizes [110], as shown in Figures 12-17.The cut-off diameter of a stage is the particle size which is collected with 50% efficiency.Most inertial impactors were designed for sampling aerosols in an industrial hygiene, occupational and environmental medicine setting [143,147,149,221].These devices were designed to operate at fixed airflow rates such that the cut-off diameters were constant.Indeed, some impactors employed critical orifices to control flow which render the system operator independent [217].Many dry powder aerosols are delivered by the inspiratory effort of the patient, therefore, the patient's performance dictates the degree of aerosol dispersion.Variation of airflow rate affects the collection efficiency of each stage, requiring recalculation or, more appropriately, recalibration of the cut-off diameter.Impactor theory and design are based on the prerequisite of laminar air-.flow through the whole device.If the airflow is increased significantly, as has been suggested, then the Reynolds numbers must be recalculated and the suitability of each device to operate at a particular airflow rate must be established.
Laminar flow exists when no convective motion occurs perpendicular to the direction of airflow.The Reynolds number (Re) is based upon the kinematic viscosity of air and characteristic velocity and length of the system, e.g., diameter of the pipe/ orifice or particle (see section 2.4).This must be considered as it applies not only to the passage through pipes, where the length scale is the diameter of the pipe, but also to the motion of the particles within an air stream, where the length scale is the particle diameter.The motion of respirable particles is dominated by viscous forces, i.e., low Re (<400) [110].This is shown in Figure 7. Turbulence can be promoted within an impactor in regions of high fluid flow activity (Re <2-3,000 [80,236]), e.g., flow between stages, leading to "wall loss" errors.These observations are important to the validity of impactor theory, especially when operating at high flow rates.
Inertial impactors have been employed to characterize pharmaceutical aerosols for almost 30 years [201] and have continued to be used for evaluation of propellant driven metered dose inhalers (pMDis) throughout that period [85,235].The first inertial samplers to be used for pharmacopoeial standards were single stage devices [4].These devices were used to define the "respirable fraction" or "respirable dose" for analytical purposes.Multi-stage inertial impactors have been employed for characterization of pMDis for a number of years [ 103].In the last decade multi-stage devices have also been added to the United States Pharmacopoeia [4].However, the use of inertial samplers to characterize dry powder inhalers has a number of drawbacks.The method of sampling cannot be conducted independently of flow rate for passive inhalers [102].Initial attempts to conduct isokinetic sampling [29,131] were superceded by studies conducted at a number of flow rates [166].The latter methods have the advantage of sampling the whole aerosol and avoid the potential for sampling errors.Hard collection surfaces, such as glass or stainless steel, give rise to re-entrainment by bounce and blow-off [101].It is recommended that these surfaces are coated prior to sampling with a viscous fluid, such as silicon fluid or glycerol [78,81].The doses of aerosol administered from a dry powder inhaler are large (::o;25mg) in comparison with a pMDI (100jlg), consequently greater wall losses occur, particularly in the upper stages of the sampler [207].
Inertial impactors offer no intrinsic description of aerosol particle size.In order to be effective instruments for the measurement of particle size these devices must be calibrated.It is usual to employ monodisperse aerosols for the calibration of inertial sampling systems.Monodisperse aerosol generators must be capable of producing aerosols with a narrow particle size distribution, as signified by a geometric standard deviation <1.2 [65].The common methods of producing monodisperse aerosols includes the vibrating orifice aerosol generator [17], spinning top/ disk generator [36,54,156] and the evaporation/condensation aerosol generator [170,194,231].Suspensions of monodisperse polymeric lattices may be employed to calibrate inertial samplers if care is taken in the conditions of their generation [202,205].

Inertial Sampling Devices USP Apparatus A
This device is a glass system composed of interconnecting flasks [84].The device was designed to be operated at 60L/min and consists of one stage with an operational cut-off diameter of 6.4 jlm.Samples are collected into a suitable liquid.Figure 12 shows the device and its characteristic collection efficiency curve.This curve is very broad and indicates a relatively inefficient size selective sampling.However, this curve is remarkably similar to the oropharyngeal deposition curve [92].

USP Apparatus B
The USP Apparatus B is a two stage device with a cut-off diameter of 9.8jlm.The impinger consists of a metal body in which a right angled throat delivers the aerosol into the impactor [183].The inlet to this device is apparently size selective, acting as an elutriator [115].Thus, a proportion of the aerosol that enters the device for sizing is preclassified.Samples are collected on a solid surface and a filter which are washed to collect the drug.The standard operating flow rate of this system is 60L/min.Figure 13 shows a diagram of the apparatus and a plot of the collection efficiency curve.This collection efficiency curve is

Multistage Impinger
The Multistage impinger is a four stage impinger constructed of glass (Erweka, Milford, CT) with manufacturer's cut-off diameters of 0.8, 3.3, 5.5 and 10.5 ~m [210].The use of this apparatus has become more prominent as it provides information on the particle size distribution in the respirable range.The instrument is arranged vertically and samples are collected into an appropriate liquid medium at an operating airflow rate of 60L/min.Figure 14 shows a diagram of the device and a plot of the collection efficiency curves for a flow rate of 60L/min.This system has been calibrated at alternative flow rates of 30, 40, 50, 60, 80 and lOOL/min, making it suitable for the evaluation of dry powder inhalers [8]. .g

Andersen Impactor
The Andersen Non-Viable lACFM Sampler (Graseby-Andersen, Smyrna, GA) is an eight stage device with manufacturer's cut-off diameters of 9.9 (preseparator), 9.0, 5.8, 4.7, 3.3, 2.0, 1.0, 0.7, 0.4 and 0.22 (filter) ~m at operating flow rates of 28.3L/min.The sampler is constructed of aluminum stages and is vertically orientated with multiple orifices at each stage, as illustated in Figure 15.The collection surfaces are stainless steel but may be replaced with filter paper or aluminum foil.The collection efficiencies have been recalculated for two other flow rates of 60L/min and 90L/min [236].New stages have been manufactured to supplement or replace the original stages and maintain cut-off diameters at various flow rates.ough calibration of this impactor is reported in the literature [237].

Marple Miller Impactor
A vertically orientated zig-zag inertial sampler, Marple Miller impactor (MMI) is designed specifically for evaluating pharmaceutical aerosols (MSP, Inc., St. Paul, MN [166]).This impactor consists of five stages with manufacturers' calibrated cut-off diameters of 0.62, 1.25, 2.5, 5.0 and 10.0 IJ.m.Each stage has a single orifice and is made of anodized stainless steel.The system is arranged with collection cups which can be washed directly and was designed to operate at 30, 60 and 90L/min.Figure 16 shows a diagram of the device and collection efficiency curves expressed at each of the operating flow rates.Figure 16 shows a relatively narrow range of collection efficiences at each of the operating flow rates, indicating that each of these flow rates may be used providing the correct cut-off diameters are adopted

Delron Impactor
The Delron, or Batelle, impactor was one of the original vertical impactors employed for use in industrial hygiene.The DCI-6 model is a six stage stainlesssteel device with manufacturers' cut-off diameters of 16, 8, 4, 2, 1, 0.5 and 0.22 (filter) 11m for uncoated collection surfaces.The operating flow rate is fixed at 12.5L/min by the use of a critical orifice.The collection plates are glass and must be removed for washing.This plot indicates the deviation of the collection efficiency from the manufacturer's specifications if the collection surface is coated, in this case with silicon fluid [78].Table 6 is a summary of the calibrated cut-off diameters for the stages of each of the inertial sampling devices.The cut-off diameters for the multistage devices with solid collection surfaces, i.e., Andersen, MMI and Delran, were calibrated with coatings.The MSI was also calibrated.In each case the calibrated cut-off diameters differ from the manufacturers' specifications which highlights the need for calibration, particularly if different flow rates and collection surfaces are being employed.

Data Interpretation
It is sufficient from a quality control and regulatory perspective to plot the amount of drug sampled at each stage of the inertial impactor against the cut-off diameter for the stage.Specifications can be set on the batch to batch and through life reproducibility of inhaler performance based on this data.The multistage impactor data can be analyzed further.The cumulative percentage undersize, on a probability scale, can be plotted against the logarithm of cut-off diameter, so that the mass median aerodynamic diameter and geometric standard deviation can be derived [49,110].This information is useful in referring to data that conform to log-normality, i.e., a linear plot.The mass median aerodynamic diameter may also be derived for data that is not log-normal, i.e., a non-linear plot, but the geometric standard deviation may be misleading.It is possible to analyze the data by fitting to the calibration collection efficiency curves using non-linear methods [203].USP Apparatus A and B were designed with deliberately broad collection efficiency curves.This follows a tradition of impactor design to approximate oropharyngeal deposition for the purposes of contrasting invitro samples with the limits established for exposure to hazardous aerosols [58,150,253].The interpretation of estimates of "respirable fraction" derived by such methods has been debated in the pharmaceutical sciences and the current concensus favors the use of multi-stage devices which give more detailed information on the particle size distribution of the aerosol.

Optical Methods
Particle size characterization of dry powder aerosols by optical methods involves the interaction of light with individual particles (single particle counters) or with the aerosol (global characterization).The principal advantages of optical techniques are that they are generally non-intrusive, acquire rapid, real-time data and have high temporal and spatial resolution.In situ measurements obtained by optical methods can reveal detailed interactions between device parameters and the resultant aerosol characteristics.The main drawback of these techniques is that no discrimination between therapeutically active and inactive particles can be made.There are clear benefits of using optical particle sizing techniques to complement inertial impaction methods.
face.A light ray that undergoes specular reflection will have an equal angle of incidence and reflection.All surfaces exhibit, to some extent, both specular and diffuse reflection.Diffuse reflection occurs when light is incident on a rough, irregular surface and is scattered in many directions.More complex phenomena can occur depending on the material, e.g., selective absorption or scattering, fluorescence and luminescence.Refraction occurs when light is incident on the interface between two transparent media.Light will be reflected and refracted, as illustrated in Figure 18 for a spherical particle.The degree to which light is refracted is related to the refractive index of the material (TJ).Total internal reflection occurs when the angle of incidence exceeds the critical angle and T)p >Tjair.This affect is illustrated in Diffraction effects occur when incident light waves interact with an obstacle, causing the light to bend.When light interacts with a single obstacle, e.g., a particle, a diffraction pattern (concentric circles of light intensity, described by the Bessel function) is produced by the superposition of secondary Huygens' sources.This is known as Fraunhofer diffraction.The diffraction pattern is uniquely related to the cross-sectional area of the particle.

Imaging Methods
Imaging techniques provide direct visual data that are easily interpreted.They may be useful to obtain information regarding particle shape, size and structure.The traditional methods of microscopy can be combined with automated methods of image analysis to measure larger particle samples and derive information on particle shape.

Microscopy
Microscopy 1s a method that allows the direct 34 physical measurement of particle size and morphology.For dry powder aerosols this method may help to distinguish between drug and carrier particles and may help to detect particle agglomeration or changes in crystal form.However, limitations regarding the size resolution and the number of particles measured have to be taken into account.Optical microscopy uses the spectrum of visible light (400-700 nm).Although it is possible to measure particles down to 0.25 11m, the measurement of particles below 1 11m is limited by the optical resolution of the instrument.This drawback can be overcome by electron microscopy that has a much higher size resolution due to the shorter Compton wavelength of the electron.The resolution of scanning electron microscopy (SEM) and transmission electron microscopy (TEM) are approximately two and three orders of magnitude greater than the optical microscope, respectively.However, artifacts due to the coating process may occur, particularly for SEM, e.g., surfaces can appear smoother than in TEM, an effect that has been attributed to the coating process [98].
A general drawback of microscopy is the extensive time required to measure sufficient numbers of particles to obtain a representative sample.A representative sample is defined as a sample in which the concentration and size distribution of particles are equivalent to the entirety.In order to achieve that aim opinions vary on the numbers of particles that should be counted.The American Society for Testing and Materials recommends that the modal class of the size distribution should contain at least 100 particles and that at least 10 particles should be present in each size class [7].Another approach of "stratified sampling" recommends collecting data on at least 10 particles in every size class that has a significant influence on the size curve [255] Moreover, effects of the sample collection and prepararion on the particle size have to be taken into account, a~ •ell as observer error due to the arbitrary determination of particle size (Martin-, Feret-, or projected area diameter).Hence, microscopy is a useful method for the acquisition of visual powder characteristics, but not a substitute for other particle sizing methods.

Image Analysis
Image analysis is used to quantitatively count, measure or classify a variety of particulate system characteristics.The image is recorded by a camera, electronically captured and subsequently processed by suitable software.Features are discriminated on the basis of gray level values to appropriate thresh-olds.Data that can be obtained include length, diameter, area, aspect ratio and shape parameters.Image analysis can be automated, making it suitable for the analysis of large numbers of particles, so that frequency distributions, such as particle size distributions can be obtained.Automation can help to overcome the long process time of analyzing large samples when combined with microscopy.Limitations of image analysis are the difficulty to analyze aggregates and artifacts due to overlapping particles.A variety of advanced image analysis systems are currently on the market (e.g.Leica, Wetzlar, Germany; Leco, St. Joseph, MI, USA; Brinkmann, Westbury, NY, USA; Juliet, Iowa City, IA, USA).In addition to complete, automated image analysis systems, a microscope may be used to obtain images that can be scanned and analyzed using image analysis software [62], e.g., SigmaScan, Jande!Scientific, Sunnyvale, CA, USA

Particle Image Velocimetry
Particle image velocimetry (PIV) is a relatively new technique that provides particle velocity characteristics of the whole flow field being studied [3].PIV consists of a double-pulsed laser sheet that illuminates the flow at two discrete points in time.This generates a staggered pair of particle images that can be analyzed to yield particle velocity information.The velocity operating range of PIV is 0.1 <Up<lOOO m/s with a viewing area of 1-10 6 mm 2 .PIV offers the advantages of quantitative information on the particulate velocity vector field along with a detailed visualization of the fluid flow.This is of particular use in fluidization studies.

Light Scattering Methods
Light scattering and extinction of dispersed particles are widely used to gather information on the concentration and size of aerosols.Numerous techniques have been reviewed in the literature [251].This chapter focuses on techniques that are applicable to pharmaceutical dry powder aerosols.

Laser Doppler Methods
Laser-Doppler methods are based on the principal of the light scattered from a particle traversing a measurement volume (approximate diameter=0.ters light by reflection and refraction.The scattered light interferes and produces a Doppler burst interference fringe pattern.The frequency of the Doppler burst is proportional to the velocity of the particle, irrespective of shape.This is the principal of laser Doppler anemometry (LDA), also known as laser Doppler velocimetry (LDV).Phase shift, produced by viewing the Doppler burst via spatially separated photodetectors, is proportional to particle size for spherical particles only.This is the principal of phase-Doppler analysis (PDA), providing information on both particle size and velocity characteristics.The operational range of both LDA and PDA is Up<500 m/s with a concentration limit of 10 6 /cm3.The size range of particles obtained by PDA is 0.5<Dp< 10 4 1-1m.LDA has been used to characterize both dry powder aerosols and pressurized metered dose inhalers [137,138].Characterization of pMDis and nebulizers has been conducted using PDA [57,227].However, application of PDA to dry powder aerosols is restricted by the requirement of spherical particles to obtain size information.The shadow-Doppler particle analyzer (SDPA) is a PDA based method that is being developed to characterize irregular shaped particles [86].This method obtains particle velocity similar to LDA.Particle size is obtained by focusing the images of the particle projected by the light within the measurement volume onto a linear array of photodetectors.The two signals produced by the light projected from the individual laser beams is related to the equivalent particle diameter.This method overcomes the intrinsic problem of measurement volume particle size bias, as well as providing information on particle size, shape, orientation and velocity.This technique is currently being validated for a wide range of spherical and non-spherical particulate systems [ 148].The particle analyzing system (PAS) is a phase-Doppler based instrument that can measure both the size and velocity of nonspherical particles (Dantec Measurement Technology, Inc., Mahwah, NJ).The laser beams are modified to produce 'to!}-hat' intensity profiles rather than Gaussian beam profiles.Particle size is correlated with the intensity of the light scattered from the particle using generalized Lorenz-Mie theory (GLMT).PAS has yet to be rigorously validated.
The single-particle, aerodynamic relaxation time (SPART) analyzer is based on the phase lag produced when a particle is subjected to a given acoustic vibration within the measurement volume of an LDA [157].The aerodynamic diameter is determined by relating the response of the particle to the acoustic excitation.Electric SPART (E-SPART) is a modified version of SPART in which an electric field and acoustic vibration are both applied to the particle [158,83].This produces a migration velocity component, providing information on the magnitude and polarity of the particle electrostatic charge as well as aerodynamic diameter.The operational range of the instrument is 0.3<Dae<75 jlm, a data acquisition range of 10-2000 particles/ s and an electrostatic charge measurement range from zero to saturation.SPART and E-SPART have been used to characterize both dry powder aerosols and pMDis [105,106,196,197,198].
The advantages of LDA and PDA are that they are non-intrusive, provide rapid, real-time data acquisition in-situ and high spatial and temporal resolution.The disadvantages of the respective methods are that LDA provides information on particle velocity alone and PDA requires the particles to be spherical to obtain size information.The small measurement volume results in a small fraction of the aerosol being characterized that may not be representative of the whole aerosol.The advantage of SPART analysis is that it provides an aerodynamic diameter of the whole aerosol.E-SPART has the additional advantage of generating information on the electrostatic charge of the particles.However, the aerosol emitted from the device must be diluted to prevent particle coincidence, requiring intrusive characterization.Chemical identity of the particles being analyzed cannot be obtained using laser Doppler methods.

Time-of-Flight Methods
Two types of time-of-flight particle sizing systems are currently on the market: the Aerodynamic Particle Sizer™ (TSI, St. Paul, USA) and the Aerosizer™ (Amherst Process Instruments, Hadley, USA).Both systems are single particle analyzers that measure the time-of-flight of accelerated aerosol particles between two laser.
The principle of operation includes three steps: generation of an aerosol beam, determination of the parti- Barrel shock envelope cle time-of-flight and conversion of the data into a corresponding particle size distribution [ 13,37,177].The aerosol is drawn into the acceleration nozzle, along with additional sheath of air, by a pressure gradient between the partly evacuated measurement region and the ambient air.The aerosol beam is formed and stabilized by the surrounding air sheath.Due to their lower inertia, smaller particles are accelerated faster and reach higher velocities relative to large particles.Upon leaving the nozzle, the particles pass two laser beams [Figure 20 ].The light scattered by the particles generates start and stop signals from which the time-of-flight of the individual particles are calculated.The time-of-flight is then converted to particle size by internal calibration curves.
The Aerodynamic Particle Sizer™ (API) measures up to 1000 particles/ cm 3 in a size rage of 0.5 to 30 11m.The size range of the Aerosizer™ is 0.1 to 700 11m with an acquisition rate of up to 10 5 particles/s.A number of additional components for specialized applications are available for the Aerosizer™.These include the Aerodisperser™, for the measurement of powders, the AeroDiluter™, to control the volume flow rate of the aerosol into the Aerosizer™, and the Aero Breather™, to generate and analyze aerosols from dry powder inhalers.The latter is a microprocessor controlled sampling system that sits on top of the Aerosizer™ housing.Its inner plenum chamber is attached directly to the nozzle entry.A piston on the upper side of the chamber is moved upwards and draws air through the DPI into the chamber, generating the aerosol.Speed, acceleration and displacement of the piston are variable, so that different inspiratory profiles can be generated.The aerosol is subsequently drawn through the sonic nozzle into the Aerosizer™ and analyzed.
The advantage of time-of-flight methods is the rapid acquisition of data and calculation of the particle size distribution.Disadvantages are that information about the chemical identity or the mass of particles cannot be obtained.Thus it is not possible to differentiate between drug and carrier particles and the dose measured cannot be assessed.Also, coincidence effects due to particle overload may give rise to phantom particles and distorted size distributions [13].Studies using the Aerosizer™ I AeroBreatherTM were able to show the effect of air flow and acceleration on the particle size distribution of the Turbuhaler, but also gave rise to possible particle overload at higher drug doses [111].It has been proposed that concentrated aerosols need to be diluted for accurate measurement [168].Time-of flight methods are beneficial as an additional sizing method for the screening of formulations and during the development of dry powder aerosols [18,108,112,168,213].

Laser Diffraction
Laser diffraction provides a global measurement of the powder aerosol in situ.The method is based on Fraunhofer diffraction of particles in a beam of parallel, monochromatic light.Particles diffract light as a function of their diameter, with the angle of diffraction decreasing with increasing particle size [Figure 21].The diffraction pattern produced by a polydisperse population of particles consists of a number of concurrent, concentric diffraction rings, each of which is associated with a certain particle size.These diffraction patterns are focused onto a radial multi-element photodetector that measures the energy distribution of the diffracted light, which is subsequently transformed into a volume distribution by the software.
A number of laser diffraction systems suitable for the measurement of sprays and powders are available.However, only one of these can be equipped for the measurement of dry powder inhalers (Sympatec He los; Sympatec, Clausthal-Zellerfeld, Germany).The inhaler is attached to a special adapter, that allows air to be drawn through the inhaler.This generates the aerosol and passes the aerosol through the laser beam.The measurement is started by a trigger system as soon as particles are detected.The advantage of this technique is that a rapid measurement of the particle size distribution of dry powder aerosols in situ is possible.It is not possible, however, to differentiate between drug and carrier particles.Dry powder aerosol particle size distributions characterized by laser diffraction yielded similar results to those obtained using an Andersen impactor [159].This method appears to be a promising method to gain information on the particle size distribution of dry powder aerosols in a very short time.

Summary
A number of inertial impaction devices ar-e available for sampling and size classifying dry powder aerosols.The most useful of these can be operated at more than one airflow rate and have multiple stages.The information from these devices can be used to reconstruct the particle size and distribution of an aerosol under different flow conditions.This allows evaluation of ease of dispersion which is useful both in product development and quality control.Inertial impaction may be used to sample aerosols <15!-lm (Table 6).This method has the advantage of utilizing chemical analysis of the drug in specific particle size ranges.The major disadvantages of these methods are that they are tedious, time consuming and subject to operator error.
A variety of optical methods are available to characterize dry powder aerosols.These methods are capable of discriminating particles in a range of sizes.In general, the laser techniques described measure particles from 0.5-1000 11m.These techniques have the advantage of rapid in situ data acquisition.The major limitation of these methods is the inability to distinguish drug from excipient particles and to specify the dose in a particular size range.

Stability
Moisture has a significant impact on the physical and chemical stability of particles.Small particles have a greater tendency to take on moisture due to their large surface area to mass ratio and the presence of large interstitial space.Moisture may cause particles to associate and this physical instability results in aggregates which cannot be dispersed adequately.Individual particles may also undergo changes in crystal system or habit based on association with water.
Drugs which are subject to hydrolytic cleavage or undergo chemical reactions with the excipients or packaging may be unstable and therefore unsuitable for inclusion in dry powder formulations.

Contamination
Particulate contaminants must be minimized to avoid the introduction of foreign objects into the lungs where they might result in toxic side effects [83].The purity of the bulk chemicals is usually monitored and handling is controlled to avoid the introduction of contaminating species.Possible contaminants include impurities produced during manufacturing and particulates associated with the containers and closures.
Impurities and contaminants in a dry powder product can be identified and quantified by a number of methods.Any visually observable contaminant, as indicated by color or shape, will be >50 J.lm in diameter, the resolution limit of the human eye.This may be seen by direct examination of the powder or for insoluble contaminants by dissolving the powder and observing residual particulates in the solution, i.e., turbidity.Small insoluble contaminating particles may be observed by microscopy.Indirect methods such as electrical resistance, light obscuration or laser diffraction may also be used to quantify insoluble particulates in the product.Soluble materials may be identified by spectroscopic methods, such as UV /VIS spectrophotometry or Infra-Red Spectroscopy.The resolution of quantitative methods can be increased by coupling them with chromatography to allow species separation.
Micro-organisms have the capacity to degrade certain compounds in the presence of moisture.The production of esterases and proteases by micro-organisms may result in the degradation of the active ingredient, instability and problems in drug storage.It is very important, therefore, to monitor the powder for the presence of such micro-organisms.These may be bacterial or fungal contaminants.Pnosocomiosis, the introduction of infective micro-organisms as contaminants in aerosols, has been identified as a problem in the delivery of nebulized solutions [20,126].In addition, contamination in multidose solutions has been identified [134].There have been no cases reported in which normal patients have experienced this phenomenon when using dry powder aerosols.However, their use in the treatment of immunocompromised patients would require rigorous attention to potential microbial contamination.Products of biological origin may be more susceptible to degradation by microorganisms.Close attention to the microbial burden is required as the biotechnology industry considers this route of administration for its products.The Microbial Limits Test states that no more than ten colony forming units/gram of product should exist [165].There should be a complete absence of coliform bacteria.

Conclusion
Figure 22 illustrates the complexity of the issues surrounding dry powder aerosol delivery.The overlapping performance characteristics shown by the inner rings represent the true performance of the aerosol delivery device, its perfomance as estimated by an in vitro testing approach, most notably focused on particle size analysis and the ultimate in vivo performance.Efficacy is only one aspect of in vivo performance.The challenge is to develop or adopt in vitro techniques that measure the actual performance of the inhaler with respect to in vivo performance.Certain characteristics are known to have a direct impact on general performance and these are indicated as particle size, airflow and the operator (in the case of in-vitro evaluation) or patient.As technologies are developed to control the specific variables (itemized on the periphery of Figure 22), it is anticipated that the degree of overlap in actual physical, estimated physical and biological performance will increase to allow predictable therapeutic effects.
There is increasing interest in the administration of drugs to the lungs and a need for delivery systems to achieve this goal.Fundamental approaches to the development of dry powder inhalers may render them an increasingly attractive option.It is important when considering dry powders that their physicochemical, flow and dispersion properties are integrated with device performance and characterization.A framework has been presented indicating the major considerations in any development program based on an understanding of powder properties and behavior.There are a number of issues which still need to be addressed.These include stability, as influenced particularly by electrostatic charge, moisture temperature, active or passive mechanisms of powder dispersion, methods of analysis, presentation and interpretation of performance data.All of these issues are important in setting valid performance criteria.
It should be clear that not all drugs, excipients and packaging materials will be suitable as components of powder inhalers.These systems are developed as complementary dosage forms to aqueous and nonaqueous solution and suspension systems, delivered by nebulizers and propellant driven metered dose inhalers.However, this is a burgeoning scientific field and the rewards for success will include novel approaches to the treatment of pulmonary and systemic diseases.as examples should not be considered an endorsement by the authors.The omission of other devices was a matter of expediency and brevity in an already lengthy manuscript.We look forward to a future opportunity to thoroughly review the status of device development.

8
Fig.1(a) The seven major crystal systems[171] and (b) an example of changes in crystal habit[32].
Fig. 6 Relationship between fluidization and particle size for the SpinhalerTM [adapted from 16].
the airstream.Shear force and relative motion are the predominant mechanisms of powder deaggregation.The high interparticulate forces expected in micronized powders are reflected in low in vitro fine particle fractions of 4-12%[55,228].

Fig. 11
Fig. 11 Definition of terms used in describing the dose from a dry powder inhaler.

Figure 15
Fig. 15 (a) Andersen Non-Viable Mark II Sampler and its (b) collection efficiency curves (courtesy of Graseby-Andersen).

Figure 17
Shows a diagram of the impactor and a characteristic collection efficiency curve for stage 1.

Fig. 19
Fig. 19 Schematic of laser• Doppler method of particle size analysis.

Fig. 21
Fig. 21 Schematic of laser diffraction method of particle size analysis.

Fig. 22
Fig.22The influence of device.in-vitro and in-vivo performance on efficacy.

Table 2
Principal mechanisms of powder fluidization and dispersion in dry powder inhalers

Table 3
specific resistance and classification of dry powder inhalers

Table 4
Dose delivery of commercially available dry powder inhalers are printed in bold.mean value ± standard deviation (n=3) of delivered dose as percent of label claim determined in the Twin Impinger (TSI) and the Multistage Liquid Impinger at 60 L/min.Note: single doses were determined for Intal SpinhalerTM, Flui SCGTM and Cromolyn OrionTM inhalers, whereas 10 doses were determined in the other inhalers [Steckel and Muller 1997(1) ]. :> SultanoJTM and VentolinTM DiskhalerTM are identical salbutamol sulphate DiskhalersTM marketed under different names in Germany and the US 4 AerodurTM and Bricanyfi'M TurbuhalerTM are identical terbutalin sulphate Turbuhalers marketed under different names in Germany and the US.
[107]n value ± standard deviation (n=25) of emitted dose of five single doses from five different inhalers as percent of label claim determined in the unit dose sampling apparatus[107].' are in accordance