P

Ic

LUa

b

Ac

Gd

a

ARR1AA

KDFANCC

1

htiapcieBttaa

0h

International Journal of Pharmaceutics 457 (2013) 143– 149

Contents lists available at ScienceDirect

International Journal of Pharmaceutics

j o ur nal ho me page: www.elsev ier .com/ locate / i jpharm

ersonalised Medicine

nvestigation of dry powder aerosolization mechanisms in differenthannel designs

an Chena,b, Rui-Lin Henga, Mulugeta Admasu Deleleb,c, Jian Caia, Di-Zheng Dud,mezuruike Linus Oparab,∗

School of Medical Instrument and Food Engineering, University of Shanghai for Science and Technology, Shanghai 200093, ChinaPostharvest Technology Research Laboratory, South African Research Chair in Postharvest Technology, Stellenbosch University, Stellenbosch 7602, SouthfricaDepartment of Post-harvest Technology, Leibniz Institute of Agricultural Engineering (ATB) Potsdam-Bornim, Max-Eyth-Allee 100, Potsdam 14469,ermanyShanghai Chenpon Pharmaceutical Technology Co., Ltd., Shanghai 201203, China

r t i c l e i n f o

rticle history:eceived 17 August 2013eceived in revised form1 September 2013ccepted 15 September 2013vailable online xxx

eywords:

a b s t r a c t

Aerosolization efficiency is the key characteristic of dry powder inhaler (DPI). However, lack of knowledgeabout powder dispersion and deposition is still a major obstacle to further improve inhaler. In the currentwork, both the in vitro deposition experiments and numerical simulations were employed to investigatethe performance of three different DPI channel designs. The powder model was commercially availableSeretide® Accuhaler®, which contains carrier lactose and drug mixture of fluticasone propionate (FP) andsalmeterol xinafoate (SX). The in vitro results, such as the mass mediate aerodynamic diameter (MMAD),fine particle fraction (FPF) and fine particle dose (FPD), were obtained by the Next Generation Impactor

ry powder inhalerine particle fractionerosolizationext generation impactoromputational fluid dynamicsarrier-based dry powder

(NGI). The values of MMAD were significantly (p < 0.05) affected by channel design. However, based onthe FPF result, the three channel designs had similar capabilities of aerosolization. It was demonstratedthat particle–wall collision was the dominant mechanism for the detachment and de-agglomeration atthese conditions. Furthermore, good linear correlations were found between the FPD values on the first4 stages of NGI and the outlet velocities of their corresponding particles, which would be used for apotential on-line approach to the evaluation of DPI efficiency.

© 2013 Elsevier B.V. All rights reserved.

. Introduction

Pulmonary drug delivery to treat chronic respiratory diseasesas been proven as a potential delivery route to complex drugshat cannot be delivered orally (Stegemann et al., 2013). Pressur-zed metered-dose inhalers (pMDIs) and dry powder inhalers (DPIs)re the main techniques for dispersing and aerosolizing solid drugarticles, which is paramount for pulmonary delivery. Due to theoncern about ozone depleting effects of chlorofluorocarbons usedn pMDIs, the DPIs are emerging as an important noninvasive deliv-ry approach in the new decade and beyond (Calvert et al., 2009;ehara et al., 2011; Heng et al., 2012). The need to utilize drug par-icles smaller than 5 �m in order to obtain a local deposition within

he lower parts of the lung leads to a variety of challenges (Cordtsnd Steckel, 2012). For DPI formulations, micronized drug powdersre commonly mixed with relatively larger coarse lactose carriers

∗ Corresponding author. Tel.: +27 21 808 4064; fax: +27 21 808 3743.E-mail addresses: opara@sun.ac.za, umunam@yahoo.co.uk (U.L. Opara).

378-5173/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.ijpharm.2013.09.012

to facilitate powder handling during the manufacturing and pow-der aerosol delivery (Zhou and Morton, 2012; Le et al., 2012). Uponaerosolization of inhaler, the carrier particles would deposit in themouth and throat regions. It is essential for the drug particles thatare attached to the carrier particle surface to be able to detach fromit so that they do not deposit together with the carrier particles, butinstead deposit in the targeted lower respiratory airways (Kho andHadinoto, 2013).

DPI efficiency is decided by three factors: properties of particles(carriers and drug powders), device design (geometrical structure)and inspired flow rate. Young et al. (2011) studied the influence ofdrug loading and carrier size on drug aerosol performance usinghomogeneous spherical model carriers. The results showed that ascarrier size increased, fine particle fraction (FPF) decreased, whileas drug loading increased, there was no change in FPF until a criticalthreshold was exceeded. Adi et al. (2011) obtained an inverse linear

relationship between the agglomerate strength and the dispersionperformance which provided direct information on them. Das et al.(2012) calculated distributions of powder strength of a cohesivebed by Monte Carlo simulations and analyzed the de-agglomeration

144 L. Chen et al. / International Journal of Pharmaceutics 457 (2013) 143– 149

Fig. 1. Scanning electron micrographs of model powders. The lower scale bar is5

pfw(iiflme2(aaaaoadi

cab(Eeppa

2

2

S2iad(st(u

cess continued till 4 L air was drawn. Then the pump was turned

0 �m and the upper one in the zooming picture is 5 �m.

roperties of lactose particles. Dispersion by acceleration in a uni-orm flow field was also studied, which seemed to be effectivehen the relative particle–fluid velocity is beyond a certain value

Calvert et al., 2011). Coates et al. (2004, 2005, 2006, 2007) stud-ed the design of Aerolizer®, such as mouthpiece geometry, airnlet size and grid structure, by the combination of computationaluid dynamics (CFD) and experiments, which indicated that eveninor modifications of the DPI features could have a significant

ffect on the overall inhaler performance. Tong et al. (2009, 2010,011, 2012) used CFD or CFD coupled with discrete element methodDEM) to study the dispersion performance of mannitol loosegglomerates, where the results indicated that the breakage of thegglomerate was mainly attributed to the mechanical impactionnd less affected by the shear effect from the flow–particle inter-ction. These results may be more useful for dispersion analysisf carrier-free powders, while DPIs are often made up of carriernd drug particles. Lack of knowledge about carrier based particleispersion mechanisms is still a major obstacle to further improve

nhaler performance.The aim of this study was to explore the dispersion behaviors of

arrier–drug particles from different DPI channel designs by in vitroerodynamic deposition experiments of the model drug, com-ination of fluticasone propionate (FP) and salmeterol xinafoateSX), using next generation impactor (NGI). At the same time, anulerian–Lagrangian particle tracking mutiphase CFD model wasmployed to further investigate the potential correlations betweenarticle-wall compaction frequency, turbulent intensity and drugarticle outlet velocity and the in vitro experimental results, suchs the FPF and fine particle dose (FPD).

. Materials and methods

.1. Materials

The model dry powders were from commercially availableeretide® Accuhaler® (GlaxoSmithKline, UK), which contains50 �g FP, 50 �g SX and 12.5 mg lactose per blister as the spec-

fication states. The densities of each ingredient were 1.37, 1.11nd 1.53 g/cm3, respectively. Particle morphology of model pow-ers was visualized (Fig. 1) using a Scanning Electron MicroscopeJSM-6360LV, JEOL, Japan). On the lactose surface, there were somemaller agglomerates adhered, which were FP and SX (Fig. 1). Par-icle size of lactose was obtained using laser diffraction analyzer

MASTERIZER 2000, Malvern Instruments, Malvern, UK). The vol-me average diameter of lactose was 83.19 �m.

Fig. 2. Top view of the three channels, channel a, b and c.

2.2. Experimental studies

2.2.1. Quantitative sample analysis by high pressure liquidchromatography (HPLC)

Quantitative analysis of the sample was done using HPLC sys-tem (Agilent Technologies Inc., CA, USA). The column used wasa Hypersil BDS (4.6 mm × 150 mm) which was packed with 5 �mC18 stationary phase (Elite Analytical Instruments Co., Ltd., Dalian,Liaoning, China). The mobile phase was a mixture of methanol and0.3% (w/v) ammonium acetate buffer in a ratio of 60:40. The bufferwas made by dissolving ammonium acetate (AR grade, SinopharmaChemical Reagent Co., Ltd., China) in reverse osmosis water. Themobile phase was freshly made before each analysis, which wasfiltered through 0.45 �m nylon filter and degassed. The flow ratewas 1.00 mL/min at 40 ◦C. An ultra violet (UV) detector set at awavelength of 228 nm was employed. The injection volume of thesample was 20 �L that was determined by means of a loop. Eachsample was analyzed in triplicate using a run time of 15 min. All sol-vents used were HPLC grade. SX standard was obtained from ApeloaJiayuan Pharmaceutical, Co., Ltd. (Dongyang, Zhejiang, China) andthat of FP was obtained from Auriso Pharma. Co., Ltd. (Tiantai, Zhe-jiang, China).

2.2.2. Deposition testThree DPI channel designs were used in the in vitro deposition

tests. The only geometrical differences among them were betweenthe two dashed lines in Fig. 2, where the main turbulence mixingeffect was expected. Each blister powders of Seretide® Accuhaler®

were filled in the drug feeders before experiment. The channelswere installed on a specially designed holder and, via an adaptor,connected to NGI set including the throat and pre-separator (MSPCorporation, Minneapolis, America), testing unit (TPK2, ERWEKA,Heusenstamm, Germany) and vacuum pump (HVP1000, ERWEKA,Heusenstamm, Germany) in sequence (Fig. 3). Each of the connec-tion port was properly sealed. The testing unit is a control systemfor this setup, which has several functions such as pressure dropmeasurement, airflow rate adjustment and duration timing.

All the seven NGI stage collection cups were coated using ethylalcohol solution with 1% Tween80. During the experiments, thepressure drop of each channel was controlled at 4 kPa. The pro-

off. The next dose was loaded and the pump restarted until 10doses were sampled per experiment. The deposits of the throat,pre-separator and each NGI collection cup were recovered by HPLC

L. Chen et al. / International Journal of Pharmaceutics 457 (2013) 143– 149 145

Fd

mpatEdt

2t

imNnmda3

2

hc

2

(ewttptcacsm

∇

ig. 3. Schematic diagram showing the apparatus employed to measure the drugeposition across the dry powder inhaler (DPI) channel models.

obile phase solution and analyzed corresponding to each com-artment. All the experiments had been done in 6 replicates. Inddition, the airflow rates were measured in triplicates followinghe method of Taki et al. (2010). The Digital Flow Meter (DFM 2,RWEKA, Heusenstamm, Germany) was adopted. Throughout theispersion experiments, the temperature and relative humidity ofhe laboratory were maintained at 18 ◦C and 50%, respectively.

.2.3. Deposition parameters of FP and SX determined by usinghe NGI

The aerosolization efficiency is characterized by five parameters.e., (1) delivered dose (DD), which was calculated as the total drug

ass recovered from all parts of the NGI, (2) FPD depositing on eachGI stage, (3) FPF, the mass of particles, which is below an aerody-amic diameter of approximately 5 �m, against the DD; (4) massedian aerodynamic diameter (MMAD); (5) geometric standard

eviation (GSD). The deposition parameters were calculated andnalyzed by inhaler testing data analysis software (CITDAS) version.10 (Copley Scientific, UK).

.2.4. Statistical analysisAnalysis of variance (ANOVA) tests followed by Tukey’s post

oc analysis were carried out with a probability of less than 0.05onsidered statistically significant (STATISTICA 11.0, Statsoft, USA).

.3. Computational methodology

The CFD code used for this work was ANSYS FLUENT 13.0ANSYS, Inc., Canonsburg, PA, USA). The air inlet boundary wasxpanded to a relatively large area (Fig. 4). Thus, the inlet pressureas assumed as atmosphere (0 Pa) and the outlet pressure was con-

rolled at −4 kPa. In the current research, the dispersion process wasreated as fluid–particle flow. The motion of the continuous fluidhase was governed by the conservations of mass and momen-um in terms of the local mean variables over a computationalell. Thus, the Navier–Stokes equations were adopted. Assumingir as incompressible, viscous, isothermal and Newtonian fluid withonstant properties, the steady state three-dimensional mass con-ervation equation and time averaged Navier–Stokes equations foromentum in Cartesian coordinates are given as:

· u = 0 (1)

∂(�f u)∂t

+ ∇ · (�f uu) = −∇P + ∇� + �f g (2)

Fig. 4. Expanded inlet area of DPI channel.

where u, �f, P, �, t and g are fluid velocity (m s−1), fluid density(kg m−3), pressure (Pa), fluid viscous stress tensor (Pa), time (s) andgravitational acceleration (m s−2). The k-� model was used to solvethe momentum conservation equations. The discrete particles werestudied by using Lagrangian particle tracking model which is givenas:

−dupi

dt= FD + FG + FB +

∑

j

Fj (3)

where upi

is the velocity of particle i (m s−1), FD, FG, FB and∑Fj are drag force, gravity, buoyancy and other forces (N),

respectively (Zhang et al., 2000). The governing equations werenumerically solved using the finite volume method. The computa-tional domain was discretized by a tetrahedral hybrid mesh using552,435, 545,714 and 800,704 cells for channel a, b and c, respec-tively. More fine meshes, which are at least four times smaller thanthe other part meshes, were adopted in the vicinity of the bend-ing parts. The mesh quality was evaluated using its aspect ratio. Allthe equations were discretized using QUICK scheme. The solutionwas converged to a normalized scaled residual below 10−3 in thesesimulations. The calculation was done using a 64-bit, AMD AthlonIITM × 4 CPU, 3.1 GHz, 4 Gb RAM, Windows 7 computer. The CUPtime of calculation was more than 20 h. Validation of the CFD modelwas performed by comparing the airflow rates obtained from thecomputational models with the experimental data.

As the concentration of disperse phase (solid particles) was low(about 1% volume), one-way coupled simulation was performed,which only concerned the influence of fluid on the small particle.Tong et al. (2010) compared one-way and two-way coupling meth-ods and the results showed that the one-way coupling methodcould speed up the simulations by 5 times in a certain condition.In this research, Eulerian–Lagrangian particle tracking multiphaseflow model was adopted to track the carrier particles, lactose, withthe average diameter of 83.19 �m, through the channels after beingentrained from the drug feeder and subjected to drag and turbulentdispersion forces. Sensitivity tests were carried out to ensure thatthe results were independent of the number of particles.

Simulations were run in which the particle number was var-ied from 300 to 100,000, which did not show significant differencein simulation results such as particle velocity at the outlet. Addi-tionally, since the particle deposition in NGI is likely to have

146 L. Chen et al. / International Journal of Pharmaceutics 457 (2013) 143– 149

rlabmio2tTp

3

3

(iwrcwtass

tSwaar(cccHai

words, it showed that higher airflow resistance resulted in greaterFPF value. In the current study, the differences of airflow resistancewere mainly due to the geometrical structures of channels, whichwould cause different magnitudes of fluid drag forces acting on the

Table 1Deposition results (n = 6) of fluticasone propionate (FP) and salmeterol xinafoate(SX) from commercial formulations via the three channels using the next generationimpactor (NGI). Mean ± SD values are shown.

Drug Parameter Channel a Channel b Channel c

FP MMAD (�m)* 3.04 ± 0.10b 3.23 ± 0.10a 2.87 ± 0.12c

GSD* 2.16 ± 0.20a 2.11 ± 0.10b 2.26 ± 0.04a

DD (�g)* 216.57 ± 10.28b 236.47 ± 15.78a 216.52 ± 8.18b

FPF (% Wt.)* 24.80 ± 1.62a 26.64 ± 1.64a 27.25 ± 1.38a

SX MMAD (�m)* 3.27 ± 0.21a 3.45 ± 0.12a 3.00 ± 0.15b

Fig. 5. Air flow rates at the outlets of the three channels, channel a, b and c.

elationship with the particle size and velocity, the channel out-et velocities of several typical size micro-particles, which weressumed to be already separated from carrier surfaces at the veryeginning of experiments, were calculated. Here, the typical sizes oficro-particles were followed “Quality Solutions for Inhaler Test-

ng” (Specification, Copley Scientific, UK): at volumetric flow ratef 60 L/min, the cut-off diameters for NGI stages 1–7 are 8.06, 4.46,.82, 1.66, 0.94, 0.55 and 0.34 �m, respectively, which are closeo the real cut-off diameters for the experiments of this research.herefore, the 7 different particle diameters were used as the modelarticle sizes in the simulation.

. Results and discussion

.1. Airflow analysis of the three channels

In this research, three slightly different channels were designedFig. 2). The airflow rate producing a pressure drop of 4 kPa var-ed among the 3 channels. The average experimental airflow rates

ere 72.5, 69.2 and 61.0 L/min, respectively. Therefore, the airflowesistance is channel a < channel b < channel c. Airflow resistanceould be described as the energy dissipation caused by frictionsithin the fluid and between the fluid and solid surfaces. The higher

urbulence level is, the greater airflow resistance will be. ANOVAnalysis indicated that the differences among the 3 flow rates wereignificant. Fig. 5 shows that the agreement between predicted andimulated results was good (r2 = 0.9608).

In each of the channels, the stream direction was changed twoimes. Therefore the carrier lactose with bigger sizes and highertokes numbers may have more chances to collide on the channelalls in the bend area. However, the “angles” of the bending parts

re different for the 3 channels. For instance, channel c has a 90◦

ngle which is a typical structure to provoke turbulence and sepa-ated flow. This behavior was confirmed by the simulation resultsFig. 6) which show that there are more vortices at the corner ofhannel c. Thus, larger energy dissipation could happen in thishannel comparing to the other two channels. Fig. 6 also shows

hannel b has more even flow dispersion, especially at the outlet.owever, at this moment, there is no much knowledge about theirflow pattern in DPI that is better for drug transfer and depositionn human airways.

Fig. 6. Contours of airflow velocity magnitudes in the three channels, channel a, band c.

3.2. Effect of channel design on drug aerosolization anddeposition

3.2.1. Deposition resultsThe deposition results, including the MMAD, GSD, DD and FPF,

are shown in Table 1. By using ANOVA analysis, it was known thatthe MMAD, GSD and DD had significant differences between someof them. For example, the MMAD values of channel c were signif-icantly lower than the other two channels, while channel b hadthe lowest GSD values for both FP and SX. Usually, FPF is one ofthe most important parameters, which is used to evaluate the DPIaerosolization efficiency. However, no significant difference of FPFwas found. Therefore, regarding to the FPF values of the two drugs,all the channels tested had the similar capabilities of aerosolizationat these conditions.

An inverse relationship between the trends of FPF values andthe airflow rates of the 3 channels were observed. In another

GSD* 2.03 ± 0.14b 2.03 ± 0.11b 2.41 ± 0.06a

DD (�g)* 51.41 ± 2.22a 50.70 ± 2.35a 53.78 ± 1.81a

FPF (% Wt.)* 20.33 ± 1.98a 20.44 ± 1.75a 22.29 ± 1.36a

* For each row, similar lower case letters are not significantly different at p < 0.05.

L. Chen et al. / International Journal of Pharmaceutics 457 (2013) 143– 149 147

Fa

pttaecd

3d

bsdgrIpvectct

F

wiotmeatbb

3

l

Fig. 8. Three hundred particle trajectories extracted randomly from 10,000 movingparticles in channel a, b and c (the colors refer to the particle velocity magnitudes,

ig. 7. Integral scale strain rate (ISSR) distributions in the three channels, channel, b and c.

articles and particle-wall collision effects. Nevertheless, owingo the complexity of aerosolization process which is con-rolled by interparticle cohesion, powder–fluid interactionnd powder–device impaction (Daniher and Zhu, 2008; Tongt al., 2011), both the turbulent kinetic energy and mechanicalollisions are likely to contribute to particle detachment ande-agglomeration.

.2.2. Effect of integral scale strain rate (ISSR) on particleispersion

The turbulence kinetic energy is a measure of the absolute tur-ulence level generated in the device, whereas the integral scaletrain rate (ISSR) defined as the turbulence eddy dissipation rateivided by the turbulence kinetic energy is a measure of the velocityradient across the integral scale eddies (the most energetic occur-ing in a turbulent flow) (Coates et al., 2006; Tong et al., 2011).n Fig. 7, for each of the channels, ISSR is larger at the bendingart where the flow keeps changing direction and the maximumalue is observed near the wall. Among the 3 channels, the high-st average value of ISSR was obtained in channel b, while nolear relationship between ISSR and FPF could be found. Base onhe investigation of several different DPIs, Mendes et al. (2007)onstructed a non-dimensional functional relationship for charac-erizing the FPF, which was given as

PF = f (FP1/2 · Q 1/4 · T1/10) (4)

here FP is the ratio of viscous air flow resistance forces in thenhaler device to gravitational sedimentation forces, Q is the ratiof inertial impaction forces to viscous particle drag forces and T ishe ratio of turbulent kinetic energy to mean kinetic energy, i.e. a

easure of the turbulence intensity, which is similar to ISSR. How-ver, the equation indicates that T has the least effect on FPF valuemong the three variables. Wong et al. (2012) also demonstratedhat internal shearing by airflow was unimportant to agglomeratereak-up. The shear force caused by the velocity gradient may note high enough to compete with interparticle forces.

.2.3. Effect of particle–wall collision on FPFThe channels studied in this research were a kind of angu-

ar design. Simulated particle tracks are given in Fig. 8. The flow

e.g., the red means the highest and the blue means the lowest). (For interpretation ofthe references to color in this figure legend, the reader is referred to the web versionof this article.)

characteristics of the particles were highly influence by the chan-nel design. Particle–wall colliding velocity (or energy) in channela > channel b > channel c, while the frequency of collision in channela < channel b < channel c. Based on the FPF values of FP and SX ofthe 3 channels, particle–wall collision was found to be a dominantmechanism for powder aerosolization. Additionally, collision fre-quency was more important than colliding velocity for the processin this study. It may be because most of the collision forces werestill below a threshold for detaching or de-agglomerating of parti-cles by one collision. Thus, at the beginning, plastic deformationwas initiated and some micro-fractures were also possibly gen-erated in the agglomerate. During the subsequent collisions, theseparations would happen gradually. Adi et al. (2010) examinedthe role of impact angles on the de-agglomeration performance ofpowders for inhalation, which implied the potential importance ofusing angular design features for multiple impactions to improveDPI performance.

3.3. Relationship between the particle velocity and the FPD

The outlet velocity of particles with 7 different cut-off diameterswas calculated using the Lagrangian particle tracking method. Sinceconcentration of the aerosol particles is dilute, it was assumed thateach of the particles was ‘freely’ flowing with the air stream withoutparticle to particle interactions. The relationship between the outletvelocity and the FPD on each stage of NGI is shown in Fig. 9. Inter-estingly, from stages 1–5, the outlet velocities had similar changingtrends with the FPD, especially for the first 4 stages, which wasconfirmed by statistical analysis. Fig. 10 illustrates the good linearcorrelation between FP particle outlet velocities through channela and the FPD data of the first four stages. Similar good correla-tions were obtained as well: for FP particles, the r2 was 0.9714,0.9636 and 0.6786, respectively; for SX, the r2 was 0.9936, 0.9978and 0.6110, respectively. Regarding to the smaller particles, whichare less than 0.5 �m in diameter, corresponding to the last twostages, they are moving by Brownian motion, settle very slowly

and, hence, may not deposit in the human airway at all (Telko andHickey, 2005). So it is more useful to focus on the bigger particlesranging from 1 to 5 �m which are relevant to stages 2, 3 and 4. Thisresults demonstrated the potential of online measurement of DPI

148 L. Chen et al. / International Journal of Pharmaceutics 457 (2013) 143– 149

Fig. 9. Relationships between particle velocities at the outlets of channel a, b and c and

Impactor (NGI).

FF

epa

4

Ae

ig. 10. Correlation between FP particle outlet velocities through channel a and thePDs on the first four stages.

fficiency, especially when there are more ‘free’ fine particles in theowder, where if the particle outlet velocities and their diametersre measured, the FPD on each stages may be predicted.

. Conclusions

In this study, the aerosolization performance of Seretide®

ccuhaler® through three DPI channel designs, which were differ-nt at the bending parts, were examined by the NGI. Although the

the fine particle dose (FPD) depositing on each stage (1–7) of the Next Generation

airflow rates and the values of MMAD, GSD and DD had significantdifferences, the FPF values of both FP and SX were not significantlyaffected by channel design. This indicated that based on the FPF,channels a, b and c had similar capabilities of aerosolization perfor-mance.

CFD simulation of the aerosolization processes in the 3 channelswas also employed. By comparing the results of experiments andsimulations, particle–wall collision was the dominant mechanismfor the particle detachment and de-agglomeration. In addition,the collision frequency was more important than the velocity atthese conditions. Furthermore, good linear correlations were foundbetween the FPD values on the first 4 stages of NGI and the outletvelocities of their corresponding particles, which would be used fora potential on-line approach to the evaluation of DPI efficiency.

Future studies will pay more attention on the relationshipbetween particle motion and deposition characteristics.

Acknowledgements

This work is based upon research supported by Senior VisitingScholar Program of the Shanghai Educational Committee, China,and the South African Research Chairs Initiative of the Departmentof Science and Technology and National Research Foundation.

References

Adi, S., Tong, Z., Chan, H.-K., Yang, R., Yu, A., 2010. Impact angles as an alternative

way to improve aerosolisation of powders for inhalation? Eur. J. Pharm. Sci. 41,320–327.

Adi, S., Adi, H., Chan, H.-K., Finlay, W.H., Tong, Z., Yang, R., Yu, A., 2011. Agglomeratestrength and dispersion of pharmaceutical powders. J. Aerosol Sci. 42, 285–294.

ANSYS, 2010. ANSYS FLUENT 13.0 Users Guide. ANSYS, Inc., Canonsburg, PA, USA.

l of Ph

B

C

C

C

C

C

C

C

D

D

H

K

L. Chen et al. / International Journa

ehara, S.R.B., Larson, I., Kippax, P., Morton, D.A.V., Stewart, P., 2011. The kineticsof cohesive powder de-agglomeration from three inhaler devices. Int. J. Pharm.421, 72–81.

alvert, G., Ghadiri, M., Tweedie, R., 2009. Aerodynamic dispersion of cohesivepowders: a review of understanding and technology. Adv. Powder Technol. 20,4–16.

alvert, G., Hassanpour, A., Ghadiri, M., 2011. Mechanistic analysis and computersimulation of the aerodynamic dispersion of loose aggregates. Chem. Eng. Res.Des. 89, 519–525.

oates, M.S., Fletcher, D.F., Chan, H.-K., Raper, J.A., 2004. Effect of designon the performance of a dry powder inhaler using computational fluiddynamics. Part 1. Grid structure and mouthpiece length. J. Pharm. Sci. 93,2863–2876.

oates, M.S., Chan, H.-K., Fletcher, D.F., Raper, J.A., 2005. Influence of air flow on theperformance of a dry powder inhaler using computational and experimentalanalyses. Pharm. Res. 22, 1445–1453.

oates, M.S., Chan, H.-K., Fletcher, D.F., Raper, J.A., 2006. Effect of design on theperformance of a dry powder inhaler using computational fluid dynamics. Part2. Air inlet size. J. Pharm. Sci. 95, 1382–1392.

oates, M.S., Chan, H.-K., Fletcher, D.F., Chiou, H., 2007. Influence of mouthpiecegeometry on the aerosol delivery performance of a dry powder inhaler. Pharm.Res. 24, 1450–1456.

ordts, E., Steckel, H., 2012. Capabilities and limitations of using powder rheologyand permeability to predict dry powder inhaler performance. Eur. J. Pharm.Biopharm. 2, 417–423.

aniher, D.I., Zhu, J., 2008. Dry powder platform for pulmonary drug delivery. Par-ticuology 6, 225–238.

as, S.C., Behara, S.R.B., Bulitta, J.B., Morton, D.A.V., Larson, I., Stewart, P.J., 2012.Powder strength distributions for understanding de-agglomeration of lactosepowders. Pharm. Res. 29, 2926–2935.

eng, D., Lee, S.H., Kwek, J.W., Ng, W.K., Chan, H.-K., Tan, R.B.H., 2012. Assessingthe combinatorial influence of climate, formulation and device on powder

aerosolization using the Taguchi experimental design. Powder Technol. 226,253–260.

ho, K., Hadinoto, K., 2013. Dry powder inhaler delivery of amorphous drug nanopar-ticles: effects of the lactose carrier particle shape and size, powder technology.Powder Technol. 233, 303–311.

armaceutics 457 (2013) 143– 149 149

Le, V.N.P., Robins, E., Flament, M.P., 2012. Agglomerate behaviour of fluticasone pro-pionate within dry powder inhaler formulations. Eur. J. Pharm. Biopharm. 80,596–603.

Mendes, P.J., Pinto, J.F., Sousa, J.M.M., 2007. A non-dimensional functional relation-ship for the fine particle fraction produced by dry powder inhalers. J. AerosolSci. 38, 612–624.

Stegemann, S., Kopp, S., Borchard, G., Shah, V.P., Senel, S., Dubey, R., Urbanetz, N., Cit-tero, M., Schoubben, A., Hippchen, C., Cade, D., Fuglsang, A., Morais, J., Borgström,L., Farshi, F., Seyfang, K.-H., Hermann, R., van de Putte, A., Klebovich, I., Hincal,A., 2013. Developing and advancing dry powder inhalation towards enhancedtherapeutics. Eur. J. Pharm. Sci. 48, 181–194.

Taki, M., Marriott, C., Zeng, X.-M., Martin, G.P., 2010. Aerodynamic deposition ofcombination dry powder inhaler formulations in vitro: a comparison of threeimpactors. Int. J. Pharm. 388, 40–51.

Telko, M.J., Hickey, A.J., 2005. Dry powder inhaler formulation. Respir. Care 50,1209–1227.

Tong, Z.B., Yang, R.Y., Yu, A.B., Adi, S., Chan, H.K., 2009. Numerical modelling of thebreakage of loose agglomerates of fine particles. Powder Technol. 196, 213–221.

Tong, Z.B., Yang, R.Y., Chu, K.W., Yu, A.B., Adib, S., Chan, H.K., 2010. Numerical studyof the effects of particle size and polydispersity on the agglomerate dispersionin a cyclonic flow. Chem. Eng. J. 164, 432–441.

Tong, Z.B., Adi, S., Yang, R.Y., Chan, H.K., Yu, A.B., 2011. Numerical investigation ofthe de-agglomeration mechanisms of fine powders on mechanical impaction. J.Aerosol Sci. 42, 811–819.

Tong, Z.B., Zheng, B., Yang, R.Y., Yu, A.B., Chan, H.K., 2012. CFD-DEM investigation ofthe dispersion mechanisms in commercial dry powder inhalers. Powder Tech-nol., http://dx.doi.org/10.1016/j.powtec.2012.07.012.

Wong, W., Fletcher, D.F., Traini, D., Chan, H.-K., Young, P.M., 2012. The use of compu-tational approaches in inhaler development. Adv. Drug Deliv. Rev. 64, 312–322.

Young, P.M., Wood, O., Ooi, J., Traini, D., 2011. The influence of drug loading on formu-lation structure and aerosol performance in carrier based dry powder inhalers.Int. J. Pharm. 416, 129–135.

Zhang, J., Li, Y., Fan, L.-S., 2000. Discrete phase simulation of gas–liquid–solid flidiza-tion systems: single bubble rising behavior. Powder Technol. 113, 310–326.

Zhou, Q.T., Morton, D.A.V., 2012. Drug–lactose binding aspects in adhesive mixtures:controlling performance in dry powder inhaler formulations by altering lactosecarrier surfaces. Adv. Drug Deliv. Rev. 64, 275–284.