`Aerodynamic Particle Size Analysis of Aerosols from Pressurized Metered-
`Dose Inhalers: Comparison of Andersen 8-Stage Cascade Impactor, Next
`Generation Pharmaceutical Impactor, and Model 3321 Aerodynamic Par-
`ticle Sizer Aerosol Spectrometer
`Submitted: July 14, 2003; Accepted: September 2, 2003
`Jolyon P. Mitchell,1 Mark W. Nagel,1 Kimberly J. Wiersema,1 and Cathy C. Doyle1
`1Trudell Medical International, 725 Third Street, London ON, Canada N5V 5G4
`
`
`ABSTRACT
`The purpose of this research was to compare three dif-
`ferent methods for the aerodynamic assessment of (1)
`chloroflurocarbon
`(CFC)
`-fluticasone propionate
`(Flovent), (2) CFC-sodium cromoglycate (Intal), and
`(3)
`hydrofluoroalkane
`(HFA)
`-beclomethasone
`dipropionate (Qvar) delivered by pressurized metered
`dose inhaler. Particle size distributions were compared
`determining mass median aerodynamic diameter
`(MMAD), geometric standard deviation (GSD), and
`fine particle fraction <4.7 µm aerodynamic diameter
`(FPF<4.7 µm). Next Generation Pharmaceutical Impactor
`(NGI)-size distributions for Flovent comprised finer
`particles than determined by Andersen 8-stage impac-
`tor (ACI) (MMAD = 2.0 ± 0.05 µm [NGI]; 2.8 ± 0.07
`µm [ACI]); however, FPF<4.7 µm by both impactors was
`in the narrow range 88% to 93%. Size distribution
`agreement for Intal was better (MMAD = 4.3 ± 0.19
`µm (NGI), 4.2 ± 0.13 µm (ACI), with FPF<4.7 µm rang-
`ing from 52% to 60%. The Aerodynamic Particle Sizer
`(APS) undersized aerosols produced with either formu-
`lation (MMAD = 1.8 ± 0.07 µm and 3.2 ± 0.02 µm for
`Flovent and Intal, respectively), but values of FPF<4.7 µm
`from the single-stage impactor (SSI) located at the inlet
`to the APS (82.9% ± 2.1% [Flovent], 46.4% ± 2.4%
`[Intal]) were fairly close to corresponding data from the
`multi-stage impactors. APS-measured size distributions
`for Qvar (MMAD = 1.0 ± 0.03 µm; FPF<4.7 µm = 96.4%
`± 2.5%), were in fair agreement with both NGI
`(MMAD = 0.9 ± 0.03 µm; FPF<4.7 µm = 96.7% ± 0.7%),
`and ACI (MMAD = 1.2 ± 0.02 µm, FPF<4.7 µm = 98% ±
`0.5%), but FPF<4.7 µm from the SSI (67.1% ± 4.1%) was
`
`Corresponding Author: Jolyon P. Mitchell, Trudell
`Medical International, 725 Third Street, London ON,
`Canada N5V 5G4. Tel: (519) 455-7060 ext.2206; Fax:
`(519) 455-9053;Email: jmitchell@trudellmed.com
`
`
`lower than expected, based on equivalent data obtained
`by the other techniques. Particle bounce, incomplete
`evaporation of volatile constituents and the presence of
`surfactant particles are factors that may be responsible
`for discrepancies between the techniques.
`
`KEYWORDS: pressurized metered-dose inhaler, im-
`pactor, time-of-flight, aerodynamic size distribution,
`aerosol measurement
`
`INTRODUCTION
`The particle size analysis of aerosols from pressurized
`metered-dose inhalers (pMDIs) by compendial proce-
`dures1,2 is typically undertaken using a multistage cas-
`cade impactor equipped with United States Pharma-
`copeia/European Pharmacopeia (USP/EP) induction
`port. This technique provides a direct link with the
`mass of therapeutically active pharmaceutical ingredi-
`ent (API) and particle aerodynamic size, which is ac-
`cepted as an indication of the likely deposition location
`within the respiratory tract.3 The recently introduced
`the Next Generation Pharmaceutical Impactor (NGI)
`(MSP, St Paul, MN4) was designed with the intent of
`improving the aerodynamic characteristics compared
`with the Andersen 8-Stage Cascade Impactor (ACI)
`(Thermo Andersen, Smyrna, GA) that is in widespread
`use for pMDI performance testing. The resulting im-
`paction-stage collection efficiency curves of the NGI at
`30 L/min5 are generally steeper than those obtained
`with the ACI,6 offering the prospect that the size frac-
`tionation process within the former will be more accu-
`rate. However, apart from a study involving prototype
`instruments,7 there is as yet almost no information to
`guide users as to the performance of the NGI with this
`class of inhaler. Cascade impaction is labor intensive
`whichever multistage impactor is used, even with aids
`
` 1
`
`R.J. Reynolds Vapor
`IPR2016-01268
`R.J. Reynolds Vapor v. Fontem
`Exhibit 1028-00001
`
`
`
`AAPS PharmSciTech 2003; 4 (4) Article 54 (http://www.aapspharmscitech.org).
`to speed up sample recovery.8 There is therefore a con-
`pactor. Each canister was shaken for 10 seconds and
`then primed by actuating 3 times to waste; then each of
`tinued interest in the development of more efficient
`5 actuations was delivered at 30-second intervals, with
`techniques that can be used particularly for early-stage
`product development.9 In the absence of a more rapid
`the mouthpiece of the inhaler coupled on axis with the
`entry to the induction port. Flow through the impactor
`multistage impactor-based technique, the use of so-
`was maintained until 30 seconds following the last ac-
`called ‘real-time’ aerodynamic particle size analyzers
`tuation. The impactor was subsequently disassembled
`based on the time-of-flight (TOF) principle has become
`quite commonplace.10 These instruments are capable of
`and the API recovered quantitatively from the induc-
`tion port, collection plates, and after filter, and then
`making a particle size measurement in typically less
`assayed by high performance liquid chromatography
`than a minute, depending on the concentration of the
`(HPLC)-UV spectrophotometry in accordance with
`aerosol that is sampled. However, TOF analyzers are
`established internal procedures. The size distribution
`susceptible to coincidence measurement problems
`from each of the canisters was determined using the
`when more than one particle is present in the measure-
`ment zone.11,12 Furthermore, the inability of at least one
`generic stage cut sizes supplied by the manufacturer, in
`accordance with compendial practice.1
`type of analyzer in this class, Aerosizer, (TSI, St Paul,
`MN) to discriminate between particles comprising API
`
`and those of excipient/surfactant has been shown to
`NGI
`result in significant bias when sizing the aerosol from a
`particular pMDI-produced suspension formulation.13
`The NGI measurements were made at 30.0 L/min ±
`More recently, however, studies with both
`the
`5%, also following the practice described for the ACI
`Aerosizer-LD14 and predecessor model 3320 Aerody-
`in the compendial method.1 The 304 stainless steel col-
`namic Particle Sizer (APS) aerosol spectrometer15
`lection cups were not coated with an adhesive agent,
`(TSI) have indicated that closer agreement with multi-
`based on previous experience using this impactor with
`stage impactor measurements may be possible for solu-
`pMDI-based aerosols.7 The NGI was used as supplied
`tion formulations where surfactant is absent. The APS
`for measurements with both chlorofluorocarbon
`is also supplied with the option of using a model 3306
`(CFC)-fluticasone propionate (Flovent, GSK Inc., Re-
`Single-Stage Impactor Inlet (SSI) (TSI), having a cut-
`search Triangle Park, NC) and CFC-sodium cromogly-
`point size of 4.7 µm aerodynamic diameter, to verify
`cate (Intal, Rhône-Poulenc Rorer Canada Inc., Mon-
`the magnitude of the so-called ‘respirable’ mass frac-
`tréal, QC, Canada), since the micro-orifice collector
`tion determined by the TOF analyzer.
`(MOC) acted as a substitute for a backup filter. How-
`ever, measurements made with a prototype instrument
`
`with Qvar had indicated that the MOC by itself might
`MATERIALS AND METHODS
`not have captured all of the extra-fine particles that
`penetrated beyond stage 7.7 An external filter unit
`Formulations
`(MSP) containing 2 layers of 934-AH glass microfiber
`Three pMDI-produced anti-asthmatic aerosols having
`was therefore connected to the outlet of the NGI for
`distinctly different particle size distribution properties
`measurements with this formulation. The operation of
`were evaluated (Table 1). Five canisters were chosen
`the pMDI canisters was as described for measurements
`at random from each of these formulations.
`by ACI.
`
`
`ACI
`APS and SSI
`Benchmark measurements were made using an alumi-
`The APS and SSI were operated together. The APS
`num ACI, sampling at 28.3 L/min ± 5%, following the
`counts particles as they pass individually through the
`procedure described in the USP.1 The ACI contained
`measurement zone where their aerodynamic size is de-
`uncoated glass collection plates with a backup glass
`termined, so it was necessary to transform the raw TOF
`microfiber filter (934-AH, Whatman, Clifton, NJ) lo-
`data to a mass-weighted size distribution using the pro-
`cated after the bottom impaction stage. In the case of
`prietary software provided (Aerosol Instrument Man-
`the measurements with hydrofluoroalkane (HFA)-
`ager, rev B [2002], TSI). The aerosol emitted from the
`beclomethasone dipropionate (Qvar) (3M Pharmaceu-
`inhaler was withdrawn at the nominal 28.3 L/min flow
`ticals, London, ON, Canada), 2 filters were used to-
`rate via a USP/EP induction port into the SSI, where
`gether in order to optimize collection of the small mass
`the incoming aerosol was sampled isokinetically at
`of extra-fine particles that penetrated beyond the im-
`
` 2
`
`R.J. Reynolds Vapor Exhibit 1028-00002
`
`
`
`AAPS PharmSciTech 2003; 4 (4) Article 54 (http://www.aapspharmscitech.org).
`Table 1. pMDI Produced Aerosols Evaluated by the Aerodynamic Particle Sizing Methods*
`Name
`Manufacturer
`Formulation Description
`Flovent-125
`GSK Inc (Canada)
`CFC-11/12 propellant mixture
`
`
`Lecithin surfactant
`
`
`125 µg/actuation fluticasone propionate†
`
`Intal-1 mg
`
`
`
`Rhône-Poulenc Rorer Inc (Canada) CFC-11/12 propellant mixture
`
`Sorbitan trioleate surfactant
`
`1000 µg/actuation sodium cromoglycate†
`
`Qvar-100
`
`
`
`
`
`3M Pharmaceuticals (Canada)
`
`
`
`HFA-134a propellant
`Ethanol cosolvent
`No surfactant
`100 µg/actuation beclomethasone dipropi-
`onate†
`*CFC indicates chlorofluorocarbon; HFA, hydrofluoroalkane; and pMDI, pressurized metered-dose inhaler.
`†Mass API/actuation expressed ex metering valve.
`
`similar [T. J. Beck, TSI Inc, November 2002 conversa-
`tion]. The operation of the pMDI canisters was again as
`described for the measurements using the multistage
`impactors.
`
`Interpretation of Data and Statistical Analysis
`The mass median aerodynamic diameter (MMAD) and
`geometric standard deviation (GSD), representing the
`measures of central tendency and spread, respectively,
`were used as metrics with which to compare the size
`distribution data. Since the central region (between
`16th and 84th percentiles) of the size distributions of all
`3 formulations obtained from the multistage impactors
`was in general well described by a log-normal distribu-
`tion function, the raw data were subjected to nonlinear
`regression analysis in accordance with the technique
`described by Thiel16 in order to establish values of
`MMAD without the need to interpolate. The APS pro-
`vides 43 size classes between 0.52 and 10.4 µm aero-
`dynamic diameter, so that error associated with interpo-
`lation between adjacent size classes to determine the
`MMAD was judged in this instance to be sufficiently
`small to be acceptable.
`FPF<4.7 µm, was also determined from the size distribu-
`tion data since this parameter is appropriate as a meas-
`ure of the therapeutically beneficial portion of the in-
`haled mass of anti-asthmatic medications capable of
`reaching the airways of the lower respiratory tract.17
`
`
`
`0.062 L/min (0.2% of the sample) directly to the APS
`(Figure 1).
`The remainder of the flow passed through the SSI. The
`portion of the mass entering this impactor contained in
`particles smaller than 4.7 µm aerodynamic diameter
`(defined as the fine particle or “respirable” fraction
`[FPF<4.7 µm]) was determined by HPLC-UV spectropho-
`tometric assay for the API collected on the after-filter
`of the impactor (containing 2 layers of 934-AH glass
`microfiber) and used to verify the equivalent result pre-
`sented by the TOF-based particle size measurements
`made using the APS. On this basis, FPF<4.7 µm could be
`determined as a percentage of the total mass entering
`the SSI in accordance with:
`
` 3
`
`(1)
`
`100
`
`
`
`
`
`)
`
`filter
`M
`+
`
`filter
`
`M
`
`M(
`
`stage
`
`
`
`FPF
`<
`
`mµ7.4
`
`=
`
`where Mstage and Mfilter are the masses of API that col-
`lect on the stage impaction plate and backup filter of
`this impactor, respectively.
`Where appropriate, a correction was applied to the
`APS-measured size distribution data to account for
`size-related losses in the sampling system. This correc-
`tion was based on the 100:1 size-efficiency relationship
`obtained for the Aerosol Diluter (model 3302A, TSI),
`also available for use with the APS, on the basis that
`the capillary dimensions and aerosol pathway from the
`isokinetic nozzle to the exit of the impactor inlet were
`
`R.J. Reynolds Vapor Exhibit 1028-00003
`
`
`
`AAPS PharmSciTech 2003; 4 (4) Article 54 (http://www.aapspharmscitech.org).
`
`Figure 1. Schematic of model 3306 showing flow pathways to the single-stage impactor and APS. (Cour-
`tesy TSI Inc).
`
`
`
`
`This parameter was obtained directly from the size dis-
`tributions measured by both ACI and APS as both in-
`struments have size class limits that correspond exactly
`to 4.7 µm aerodynamic diameter. FPF<4.7 µm could also
`be determined directly from the SSI, since its cut size is
`fixed at 4.7 µm aerodynamic diameter. FPF<4.7 µm was
`estimated by linear interpolation for the NGI since
`
`stages 2 and 3 have cut sizes of 6.4 and 4.0 µm aerody-
`namic diameter, respectively, at 30 L/min.
`Statistical interpretation of the data derived from the
`size distributions obtained by the various procedures
`was undertaken using appropriate tests of significance
`(SigmaStat, version 2.3, SPSS Science, Chicago, IL).
`Differences were deemed significant when P < .05.
`
` 4
`
`R.J. Reynolds Vapor Exhibit 1028-00004
`
`
`
`AAPS PharmSciTech 2003; 4 (4) Article 54 (http://www.aapspharmscitech.org).
`Values of the reported performance metrics represent
`mean ± SD based on 5 replicate measurements unless
`otherwise stated.
`
`RESULTS AND DISCUSSION
`The choice of the ACI as the benchmark device for the
`present study reflects the widespread use of this impac-
`tor for the measurement of pharmaceutical aerosols by
`the compendial procedure.1 The results from this study
`do not enable any claim to be made in terms of the ac-
`curacy of this impactor in comparison with the other
`techniques.
`Mass recovery of API was within ± 20% of label claim
`for the measurements with both ACI and NGI. The
`mean mass loading of the NGI, based on 5 actuations
`per measurement and considering only the mass that
`penetrated beyond the induction port to the impactor,
`was substantially greater for Intal (1443 µg) compared
`with either Qvar (227 µg) or Flovent (246 µg). Similar
`total mass loading data (not shown) were obtained for
`the ACI.
`Comparative size distributions for the ACI, NGI, and
`APS for Flovent, Intal, and Qvar are summarized on a
`cumulative mass-weighted basis in Figures 2, 3, and 4,
`respectively, using log-probability scaling.
`Only minor differences were observed in GSD values
`between the 3 measurement techniques for Qvar and
`Intal (Table 2), and GSDs for Flovent aerosols were
`equivalent (P = .87). No technique, therefore, consis-
`tently produced size-distribution data that were consis-
`tently less or more disperse than data obtained by the
`other 2 instruments. However, although values of
`MMAD for Intal determined by either of the multistage
`impactors (4.3 ± 0.19 µm [NGI], 4.2 ± 0.13 µm [ACI])
`were comparable (unpaired t test, P = .29), the NGI-
`measured MMAD for Flovent (2.0 ± 0.05 µm) was
`significantly finer than that obtained by ACI (2.8 ±
`0.07 µm) (P < .001). The ACI-based MMAD was,
`however, within the range from 2.4 to 2.8 µm reported
`by Cripps et al for this formulation, also using this type
`of impactor.18
`Overlap of the collection efficiency curves of neighbor-
`ing stages of either impactor is an unlikely cause of the
`observed differences between MMAD values obtained
`from the multistage impactors for Flovent, as the effect
`is reported to be small below stage 2 based on a previ-
`ously published calibration of an ACI,6 and should be
`even less apparent with the NGI in view of its sharp
`and well-separated stage collection efficiency curves.5
`
`
`Figure 4. Comparison of ACI-, NGI-, and APS-
`measured size distributions for Qvar.
`
`
` 5
`
`Figure 2. Comparison of ACI-, NGI-, and APS-
`measured size distributions for Flovent.
`
`
`
`Figure 3. Comparison of ACI-, NGI-, and APS-
`measured size distributions for Intal.
`
`
`R.J. Reynolds Vapor Exhibit 1028-00005
`
`
`
`AAPS PharmSciTech 2003; 4 (4) Article 54 (http://www.aapspharmscitech.org).
`Table 2. Size Distribution Parameters Obtained From the ACI, NGI, and APS*
`
`Flovent
`Intal
`
`Qvar
`
`
`
`MMAD (µm)
`
`GSD
`
`MMAD (µm)
`
`GSD
`
`MMAD (µm)
`
`GSD
`
`1.84 ± 0.06
`1.2 ± 0.02
`1.63 ± 0.04
`4.2 ± 0.13
`1.56 ± 0.03
`2.8 ± 0.07
`ACI†
`1.82 ± 0.03
`0.9 ± 0.03
`1.75 ± 0.04
`4.3 ± 0.19
`1.55 ± 0.07
`2.0 ± 0.05
`NGI†
`1.74 ± 0.05
`1.0 ± 0.03
`1.66 ± 0.04‡
`3.2 ± 0.02‡
`1.57 ± 0.02
`1.8 ± 0.07
`APS
`*ACI indicates Andersen 8-stage cascade impactor; APS, aerodynamic particle sizer; GSD, geometric standard deviation; MMAD, mass
`median aerodynamic diameter; and NGI, next generation pharmaceutical impactor. Except where indicated, n = 5 replicates/technique.
`†Size distribution parameters are based on mass penetrating the induction port and entering the impactor.
`‡4 replicate measurements.
`
`
`
`There are at least 2 other possibilities to consider, and
`both potential causes may have contributed to the ob-
`served behavior. Kamiya et al19 have recently reported
`that the NGI, if used with uncoated collection cups,
`undersized a CFC-suspension formulation (Vanceril,
`Schering-Plough, Kenilworth, NJ) that has similar size
`distribution characteristics to Flovent. They obtained
`close agreement between NGI- and ACI-measured data
`when their cups were precoated with silicone oil to im-
`prove particle adhesion. The NGI does appear to be at
`greater risk than the ACI of bias caused by this effect.
`Stage Reynolds numbers, which govern particle veloc-
`ity and therefore kinetic energy at the point of collec-
`tion in an impactor,20 are in the range from 324 to 2938
`at 30 L/min for the NGI, as a direct consequence of
`optimizing its aerodynamic performance,4 whereas the
`equivalent range for the ACI is from 110 to 782.20 At
`first sight, the agreement achieved between NGI- and
`ACI-measured size distributions for Intal is difficult to
`reconcile. However, although the sodium cromoglycate
`particles are in general larger than those formed from
`Flovent, they are hygroscopic,21 likely making their
`surfaces tacky under the conditions of the present
`study, and therefore more prone to adhere to an un-
`coated surface. In view of these observations, it may be
`prudent to coat the collection cups of the NGI with an
`adhesive agent to be sure that particle bounce does not
`occur, unless it can be demonstrated that this phe-
`nomenon is not occurring to a significant extent with a
`given formulation.
`It is also possible that propellant evaporation was not
`fully complete by the time that sampling of particles
`produced from Flovent took place. Under these cir-
`cumstances, the size distribution measured by the im-
`pactor having the larger internal volume (1000 mL,
`NGI; 450 mL, ACI) might be expected to be finer as a
`result of more complete evaporation having taken place
`by the time that the particles were collected. The boil-
`
`ing point of CFC-11 is 23.8°C at 101.3 kPa, so that at
`room ambient conditions close to 22°C, evaporation of
`this component would be expected to be relatively slow
`compared with that of CFC-12 (boiling point -26.1°C).
`However, published data on this effect are scant.
`Morén22 commented that the initial flashing of liquid
`propellant to vapor as pressure is relieved upon actua-
`tion is so rapid that heat required for the change of
`phase is taken from the liquid remaining, therefore re-
`sulting in cooling of the droplets. Further evaporation
`then occurs as energy is acquired from the surrounding
`air molecules; this evaporation is slow compared with
`the flashing process. Morén and Anderson,23 in a study
`involving a formulation containing terbutalene sulfate
`in a CFC-11:12:114 mixture, reported MMAD values
`that decreased from 43 µm immediately after actuation
`to 14 µm at a distance of 10 cm from the canister. On
`this basis, aerosol transport beyond the induction port
`might therefore have taken place before propellant
`evaporation was complete. However, precise evapora-
`tion behavior will depend on actuator orifice diameter,
`the mass of formulation metered per actuation, and
`aerosol pathway into the impactor, as well as physical
`properties of the propellant,24 so that data from studies,
`such as that of Morén and Anderson can only provide a
`general indication of the magnitude of this effect.
`The agreement between measurements by both impac-
`tors for Intal in the present study is also explicable in
`terms of propellant evaporation behavior if the kinetics
`of evaporation of propellant from droplets containing a
`large mass concentration of solids (as is the case with
`Intal) approached equilibrium more rapidly. Further-
`more, the relatively large particles produced by Intal
`compared with those formed from Flovent, travel a
`shorter distance into either impactor before being col-
`lected, so that differences in internal volumes between
`the NGI and ACI would be expected to be less impor-
`tant.
`
` 6
`
`R.J. Reynolds Vapor Exhibit 1028-00006
`
`
`
`AAPS PharmSciTech 2003; 4 (4) Article 54 (http://www.aapspharmscitech.org).
`The APS undersized aerosols from both Flovent and
`Intal, compared with either multistage impactor. Thus,
`the MMAD measured by the APS was 1.8 ± 0.07 µm
`for Flovent and 3.2 ± 0.02 µm for Intal (1-way analysis
`of variance [ANOVA] for each formulation, P < .05).
`The cause is believed to be the presence of surfactant
`particles that are formed together with particles com-
`prising API with both formulations. It is pertinent that
`in a previous study using an Aerosizer TOF aerody-
`namic particle size analyzer, the Aerosizer-measured
`MMAD of a suspension formulation containing
`budesonide with sorbitan trioleate surfactant was found
`to be 2.4 ± 0.2 µm, compared with 3.9 ± 0.1 µm by
`ACI.13 The surfactant particles were observed by mi-
`croscopy to have collected further into the impactor
`(finer sizes) compared with particles of API, and the
`Aerosizer-measured size distributions were shown to
`have included both surfactant and API particles. The
`APS, like the Aerosizer, cannot discriminate between
`particles of surfactant and API, since no assay for drug
`substance is undertaken.
`In the case of Flovent, the ACI-measured FPF<4.7 µm
`(88.8% ± 2.9%) was slightly lower than 93.4% ± 0.7%
`obtained with the NGI (P < .001) (Table 3), consistent
`with the larger MMAD obtained by the ACI. FPF<4.7 µm
`from the SSI (used in conjunction with the APS [82.9%
`± 2.1%]) was only slightly smaller than the ACI-
`measured value (P =
`.006). However, the SSI-
`determined FPF<4.7 µm was considerably less than the
`97.9% ± 1.2% obtained using the APS (P < .001), as a
`consequence of the tendency for the APS to undersize
`compared with the impactors. For Intal, FPF<4.7 µm
`measured by the SSI (46.4% ± 2.4%) was also mark-
`edly smaller than 78.8% ± 2.1% obtained by the APS
`(P <.001), and also slightly lower than corresponding
`values determined by the multistage impactors (60.2%
`± 2.8% [ACI]; 52.0% ± 2.9% [NGI]) (P < .05). This
`slight discrepancy between SSI- and multistage impac-
`tor data can be explained if the CFC-11 propellant
`evaporation was incomplete by the time that the parti-
`cles entered the SSI, whose internal volume is less than
`200 mL [G. Pence, TSI Inc, letter, February 2003].
`In contrast to the behavior with Flovent and Intal, the
`APS-measured MMAD (1.0 ± 0.03 µm) for Qvar,
`which is a solution formulation containing no surfac-
`tant, was located between the corresponding values
`from the NGI and ACI, and the difference between this
`value and the NGI-measured MMAD was barely sig-
`nificant (P = .011). Hence, FPF<4.7 µm from both the
`APS and the multistage impactors were in close
`agreement (98.0% ± 0.5% [ACI]; 96.7% ± 0.7%
`[NGI]; 96.4% ± 2.5% [APS]) (P = .18). The steeper
`
`
`Figure 5. Effect of correction for inlet sampling effi-
`ciency on APS-based size distribution measurements
`for Intal.
`
`NGI-measured particle size distributions for Qvar were
`also finer than those obtained by the ACI (MMAD =
`0.9 ± 0.03 [NGI]; 1.2 ± 0.02 [ACI]) (P < .001). Again
`the ACI-based MMAD was in close agreement with
`1.1 µm reported previously for this formulation, also
`using this type of impactor.25 In the case of this formu-
`lation, the finer MMAD measured by the NGI com-
`pared with ACI can also be explained in terms of dif-
`fering evaporation behavior. However, in this instance,
`the ethanol solubilizer (boiling point 78°C) is more
`likely than HFA-134a propellant (boiling point -
`26.5°C) to have been incompletely evaporated when
`the particles were collected. This explanation is sup-
`ported by data from a study by Gupta et al,26 who
`demonstrated that ethanol evaporation is incomplete at
`the distal end of a USP induction port, working with a
`range of solution formulations similar in composition
`to Qvar.
`It should be noted that the APS-measured size distribu-
`tions were not corrected for particle counting efficiency
`as a function of particle size, since Peters and Leith27
`recently indicated that counting efficiency is size inde-
`pendent in the range of 45% to 60% between 0.7- and
`4-µm aerodynamic diameter for the model 3321 APS.
`However, a correction was made for the sampling effi-
`ciency of the SSI (see Materials and Methods section).
`The effect of this correction on the reported size distri-
`bution data was found to be negligible, except for Intal
`(Figure 5), since the majority of the particles sampled
`by the APS for Flovent and Qvar were finer than 4-µm
`aerodynamic diameter, where the efficiency of the
`Aerosol Diluter is close to 100%.28
`
` 7
`
`R.J. Reynolds Vapor Exhibit 1028-00007
`
`
`
`AAPS PharmSciTech 2003; 4 (4) Article 54 (http://www.aapspharmscitech.org).
`Table 3. Values of FPF<4.7µm (%) Obtained From the ACI, NGI, and APS*
`Qvar
`
`Flovent
`Intal
`98.0 ± 0.5
`88.8 ± 2.9
`60.2 ± 2.8
`ACI†
`96.7 ± 0.7
`93.4 ± 0.7
`52.0 ± 2.9
`NGI†
`96.4 ± 2.5
`97.9 ± 1.2
`78.8 ± 2.1‡
`APS
`67.1 ± 4.1
`82.9 ± 2.1
`46.4 ± 2.4
`3306 Impactor
`*ACI indicates Andersen 8-stage cascade impactor; APS, aerodynamic particle sizer; FPF<4.7 µm, fine particle fraction <4.7 µm aerody-
`namic diameter; and NGI, next generation pharmaceutical impactor. Except where indicated, n = 5 replicates/technique.
`†Values are based on mass penetrating the induction port and entering the impactor.
`‡ Based on 4-replicates.
`
`
`
`slope of the APS-measured data for particles finer than
`0.8 µm aerodynamic diameter (Figure 4) can be ex-
`plained by loss of sensitivity of the TOF detection sys-
`tem, which reaches its limit of detection at about 0.5
`µm aerodynamic diameter.29 However, FPF<4.7 µm de-
`termined by the SSI (67.1% ± 4.1%) was surprisingly
`much smaller than any of the other values for this for-
`mulation, which ranged from 96% to 98% (P < .001)
`(Table 3). Similar behavior was reported by Gupta et
`al26 and explained in terms of incomplete ethanol
`evaporation in the SSI. A solution might be to extend
`the passageway to the SSI to match more closely the
`aerosol transit time with that for the APS, as was done
`by Gupta et al, who found that as much as 40 cm of
`inlet extension was needed to achieve good agreement
`between techniques. Such a change would also be ex-
`pected to improve the agreement between SSI- and
`multistage impactor-measured FPF<4.7 µm for formula-
`tions, such as Flovent and Intal that contain relatively
`low volatile CFC-11 propellant. However, it will be
`necessary to be careful not to introduce additional sur-
`faces for impaction if a design change of this nature is
`contemplated. An alternative strategy, avoiding such
`problems, might be to heat the existing transfer channel
`so that the ethanol/propellant is more rapidly evapo-
`rated. Whichever solution is implemented, some flexi-
`bility will be needed for the user to set up the aerosol
`transport conditions in the SSI and APS on a formula-
`tion-by-formulation basis, given the variety of volatile
`species that may be present within the range of pMDIs
`that are available.
`
`CONCLUSION
`These in vitro comparisons indicate how 3 pMDI-
`generated formulations behaved in laboratory-based
`sampling systems. Discrepancies in size distributions
`were identified in data from 2 different designs of cas-
`cade impactor operated under near equivalent condi-
`
`tions. Aerosols from Flovent and Intal were undersized
`by the APS, most likely because of its inability to dis-
`criminate between particles containing API and finer
`surfactant particles. Further work is merited to resolve
`whether apparent undersizing observed with the NGI
`with both Flovent and Qvar is caused either by particle
`bounce and/or incomplete evaporation of volatile con-
`stituents. The SSI should be modified to ensure compa-
`rable evaporation of volatiles to that obtained in the
`APS, if formulations such as Qvar are to be sized accu-
`rately. It is strongly recommended that both NGI and
`APS/SSI data be evaluated on a formulation-by-
`formulation basis in relation to the large database that
`already exists for ACI-based measurements.
`
`ACKNOWLEDGEMENTS
`The authors acknowledge the support of Dr. Daryl
`Roberts, MSP Corp, St Paul, MN, for advice and assis-
`tance with the modifications to the NGI, and Mr. Greg
`Pence and Mr. Tyler Beck, TSI Inc, St Paul, MN, for
`the loan of a model 3321 APS with model 3306 Impac-
`tor Inlet and advice on use and interpretation of data.
`
`
`REFERENCES
`1. USP 26-NF 21. Chapter 601-Physical tests and determinations:
`aerosols. United States Pharmacopeia. Rockville, MD: United
`States Pharmacopeial Convention; 2003:2105-2123.
`2. European Pharmacopeia. Section 2.9.18-Preparations for inha-
`lation: aerodynamic assessment of fine particles. European Phar-
`macopeia. 3rd ed. [Suppl 2001]. Strasbourg, France: Council of
`Europe; 2002:113-124.
`3. Rudolph G, Kobrich R, Stahlhofen W. Modeling and algebraic
`formulation of regional aerosol deposition in man. J Aerosol Sci.
`1990;21(suppl 1):306-406.
`4. Marple VA, Roberts DL, Romay FJ, Miller NC, Truman KG,
`Van Oort M, Olsson B, Holroyd MJ, Mitchell JP, Hochrainer D.
`Next generation pharmaceutical impactor. Part 1: Design. J Aero-
`sol Med. 2003;16:283-299.
`
` 8
`
`R.J. Reynolds Vapor Exhibit 1028-00008
`
`
`
`AAPS PharmSciTech 2003; 4 (4) Article 54 (http://www.aapspharmscitech.org).
`17. Dolovich MB. Aerosol delivery devices and airways/lung
`5. Marple VA, Olson BA, Santhanakrishnan K, Mitchell JP,
`Murray SC, Hudson-Curtis BL. Next generation pharmaceutical
`deposition. In: Schleimer RP, O’Byrne P, Szeffler S, Bratsand R,
`impactor. Part 2: Archival calibration. J Aerosol Med.
`eds. Inhaled Steroids in Asthma. New York, NY: Marcel Dekker;
`2003;16:301-324.
`2001:169-210.
`18. Cripps A, Riebe M, Schulze M, Woodhouse R. Pharmaceuti-
`6. Mitchell JP, Costa PA, Waters S. An assessment of an Ander-
`sen mark-II cascade impactor. J Aerosol Sci. 1988;19(2):213-231.
`cal transition to non-CFC pressurized metered dose inhalers. Res-
`pir Med. 2000;94(suppl B):3-9.
`7. Mitchell JP. The next generation impactor (NGI): results from
`the evaluation of prototype instruments with pressurized metered
`19. Kamiya A, Sakagami M, Hindle M, Byron PR. Particle sizing
`dose inhaler (pMDI)-based formulations. In: Drug Delivery to the
`with the next generation impactor: a study of Vanceril™ metered
`Lungs – XI. London, UK: Aerosol Society; 2000:223-226.
`dose inhaler. J Aerosol Med. 2003;16(2):216.
`8. Miller NC, Roberts DL, Marple VA. The ‘Service Head’ ap-
`20. Marple VA, Olson BA, Miller NC. The role of inertial particle
`proach to automating the next generation pharmaceutical impac-
`collectors in evaluating pharmaceutical aerosol delivery systems. J
`tor: proof of concept. In: Dalby RN, Byron PR, Peart J, Farr SJ,
`Aerosol Med. 1998;11(suppl 1):139-153.
`eds. Respiratory Drug Delivery VIII. Raleigh, NC: Davis Hor-
`21. Chew NYK, Bagster DF, Chan H-K. Effect of particle size, air
`wood International; 2002:521-523.
`flow and inhaler device on the aerosolisation of disodium cromo-
`