JOURNAL OF AEROSOL MEDICINE
`Volume 18, Number 3, 2005
`© Mary Ann Liebert, Inc.
`Pp. 311–324
`
`Comparison of Cascade Impaction and Laser Diffraction
`for Particle Size Distribution Measurements
`
`JOCHEN ZIEGLER, Ph.D., and HERBERT WACHTEL, Ph.D.
`
`ABSTRACT
`
`The Andersen cascade impactor (ACI) and laser diffraction (LD) can be correlated at ambient
`temperature for aqueous drug formulations atomized by Soft Mist™ inhalers. A comparison
`of the two particle size determination methods at different conditions (flow rate, relative hu-
`midity) was performed. Under well-defined conditions, the faster LD can substitute the time-
`consuming ACI at least for routine tests. The measurements were performed with three dif-
`ferent drug formulations. The aerosol was generated by Soft Mist™ inhalers, and the droplet
`distributions were measured simultaneously using a laser diffraction analyzer together with
`the eight-stage Andersen cascade impactor. The simultaneous measurements ensure that
`aerosol and air conditions are identical for both LD and ACI. In order to measure the scat-
`tered laser light intensity of the aerosol passing the induction port, glass windows were fit-
`ted to the induction port. The evaporation effect of the aqueous aerosols on the PSD was
`investigated at ambient humidity and high humidity (RH ⬎ 90%). The simultaneous deter-
`mination of the droplet size distribution leads to a good correlation between the ACI and LD
`method only if the measurements were performed at RH of ⬎90%. The humidity of the am-
`bient air had the strongest influence on PSD not only for ACI, but also for LD. In our set-up,
`the almost saturated air prevents aqueous droplets from drying. The influence of the flow
`rate on LD was negligible, whereas for ACI, a flow rate dependence is expected. The advan-
`tages of LD and the demonstrated compatibility to established EP/USP methods motivate the
`substitution of the ACI and the use of LD for routine measurements.
`
`Key words: soft mist inhaler, particle size distribution, impactor, laser diffraction
`
`INTRODUCTION
`
`IN THE PHARMACEUTICAL INDUSTRY, the determi-
`
`nation of particle size distributions (PSD) of at-
`omized aerosols is important for estimating the
`deposition characteristic in the lungs. In practice
`the common principle for measuring the PSD is
`the impaction method as described in the USP
`26.(1) The cascade impactor can be considered as
`
`a simplified model of the respiratory system of
`human beings. The aerosol is guided by means
`of an air stream at defined flow rate through the
`rectangular bend (model of the human throat)
`and the following impaction stages (modelling
`the particle size dependent deposition in differ-
`ent parts of the lung). Further information about
`the cascade impactor and the measurement prin-
`ciple can be found in a monograph series by
`
`Boehringer Ingelheim Pharma GmbH & Co. KG, Ingelheim, Germany.
`
`311
`
`Downloaded by Society - Active - International Society for Aerosols in Medicine (ISAM) from www.liebertpub.com at 11/09/20. For personal use only.
`
`Liquidia's Exhibit 1058
`Page 1
`
`

`

`312
`
`Lodge and Chan.(2) This method is well accepted
`by the national medical agencies due to its sim-
`plicity and robustness. The whole system is de-
`fined and can be described by only a few para-
`meters like the flow rate of the air stream and the
`geometry of the impactor, for example, the num-
`ber of nozzles, the jet diameters defined by the
`nozzle diameters of the nozzle plates, the dis-
`tances of the nozzles to the impaction plates and
`the length of the nozzles. However, the process
`of aerosol analysis is time consuming and there-
`fore not suitable for routine measurements with
`large batch numbers. Especially the analysis of
`the different mass fractions on the impaction
`stages is very labor intensive. Hence it is neces-
`sary to establish faster alternatives for particle
`size determinations based, for example, on laser
`diffraction (LD). A typical laser diffraction in-
`strument and further details are given in the In-
`ternational Standard ISO 13320-1,(3) for example.
`A laser is used to generate a monochromatic,
`coherent, parallel beam that illuminates the dis-
`persed particles after expansion by the beam pro-
`cessing unit. In many conventional systems, the
`measuring zone has ambient air conditions. En-
`closures are offered for light protection. The ef-
`fect of ambient air interacting with the aerosol is
`often neglected. The incident light is scattered by
`the ensemble of dispersed particles. The total an-
`gular intensity distribution (I[␪]), consisting of
`both direct and scattered light, is then focused
`by a lens system onto a multi-element detector,
`where a discrete spatial intensity distribution
`(I[r]) is recorded. By means of a computer the par-
`ticle size distribution can be calculated which best
`approximates (I[r]).
`In order to introduce and establish the laser dif-
`fraction method as a tool that may replace the cas-
`cade impactor for routine measurements on phar-
`maceutical inhalers, the equivalence of both
`methods must be proven. Using continuously op-
`erating nebulizers, Clark,(4) Kwong et al.,(5) and
`Vecellio None et al.(6) established a good corre-
`spondence between the methods regarding the
`aerodynamic diameters and the geometrical stan-
`dard deviations. However, only Clark(4) simulta-
`neously measured the PSD of a non-volatile
`aerosol (dibutyl phthalate) with both methods.
`Kwong et al.(5) used aqueous aerosols which
`are affected by evaporation. By laser diffraction,
`they investigated a free aerosol cloud. On the
`other hand, the standard set-up was used for the
`
`ZIEGLER AND WACHTEL
`
`Andersen impactor measurements and uncondi-
`tioned room air was entrained into the nebulizer
`chamber. The authors stressed the importance of
`humidity control during the cascade impactor
`measurement, and achieved this goal by cooling
`the cascade impactor in order to minimize evap-
`orative losses. However, Kwong et al. did not find
`any evidence suggesting a significant evaporative
`loss of fine particles using LD.
`Vecellio None et al.(6) have used a T piece sam-
`pling technique with LD in order to have the
`same experimental set up as cascade impactors
`used in European Standard EN 13544-1. The au-
`thors have demonstrated that it is important to
`use the same experimental set up to compare the
`different measurement methods; for example,
`when sampling at a 90-degree angle at 2 L/min
`air flow in accordance with EN 13544-1,(7) it was
`shown that LD used with T piece underestimated
`the MMAD of the aerosol produced by nebuliza-
`tion with respect to sampling at 0-degree angle at
`15 L/min. The tests were performed close to stan-
`dard conditions in the range of 23 ⫾ 2°C and
`40–75% RH.
`As far as metering inhalers are concerned,
`Ziegler and Wachtel(8,9) described the first suc-
`cessful attempts to establish a correlation be-
`tween laser diffraction and cascade impaction us-
`ing aqueous aerosols generated by soft mist
`inhalers. Dedicated equipment is required as the
`soft mist inhalers generate a high particle density
`(⬎106 particles/cm3 ) for a time span of ⱕ1.5 sec.
`The metered dose operation of the inhaler pre-
`vents the entrained air from establishing an equi-
`librium humidity at reduced temperature and
`motivates the need for assessment of individually
`delivered doses. A simultaneous measurement is
`the only way to assess one individual dose with
`both methods, LD and ACI, respectively. For that
`reason, the measurements were performed si-
`multaneously and evaporation was accounted for
`by a comparison between volatile aqueous liquid
`formulations and non-volatile aerosols. The
`aqueous aerosols were generated by a soft mist
`inhaler. Humidified air with RH of ⬎90% was
`passed through inhaler, induction port and ACI.
`The measurements were performed at ambient
`temperature (22 ⫾ 2°C). For the simultaneous
`measurement of the PSD with LD and ACI the
`induction port (also denoted USP-throat, see USP
`26(1)) was modified without changing the char-
`acteristic impactor geometry.
`
`Downloaded by Society - Active - International Society for Aerosols in Medicine (ISAM) from www.liebertpub.com at 11/09/20. For personal use only.
`
`Liquidia's Exhibit 1058
`Page 2
`
`

`

`COMPARISON OF CASCADE IMPACTION AND LASER DIFFRACTION
`
`313
`
`MATERIALS AND METHODS
`
`Prototype Respimat® soft mist inhalers were
`used to generate the aqueous aerosols. The in-
`haler uses the mechanical energy of a loaded
`spring which drives a piston. A metered amount
`of liquid is pressed through a micro-nozzle, pro-
`ducing an aerosol of the desired MMD, for ex-
`ample, MMD of ⬍5 ␮m. The investigated for-
`mulations were close to final formulations
`intended for market and contained different ac-
`tive drugs (active drug concentration c indicated)
`as well as excipients. They are called formulation
`A (c ⫽ 0.049%), B (c ⫽ 0.198%), and C (c ⫽
`0.833%). By this choice, the concentration c of
`drugs ranged from c ⫽ 0.049%, 0.198% to 0.833%.
`The density of the aqueous formulations was
`close to unity (1.0 g/cm3). A single actuation of
`the inhaler resulted in a spray duration of 1.5 sec.
`The non-volatile aerosol was generated with a
`Sinclair-LaMer type aerosol generator MAG-2010
`(PALAS® GmbH in D-76229 Karlsruhe, Ger-
`many). This aerosol was used for testing the reli-
`ability of the laser diffraction analyser. The gen-
`erator is capable to generate adjustable particle
`diameters between approximately 0.3 and 6 ␮m
`with a geometric standard deviation ␴g less than
`1.15 and a number concentration up to 106 cm⫺3.
`In the boiler where the aerosol material is vapor-
`ized, the temperature controls the particle diam-
`eter. The corresponding aerosol material is DEHS
`(Di-2-Ethylhexyl-Sebacate).
`
`Particle size measurement
`Aerosol droplet distributions were measured
`using the Sympatec HELOS laser diffraction
`analyser (Sympatec GmbH, D-38678 Clausthal-
`Zellerfeld, Germany) at ␭ ⫽ 632.8 nm (He-Ne
`laser) together with an Andersen Mark II 8-stage
`cascade impactor operated at 28.3 L/min with the
`corresponding cut-off points 0.4, 0.7, 1.1, 2.1, 3.3,
`4.7, 5.8, and 9.0 ␮m. To our knowledge, the cut-
`off diameter of the throat is not well defined in
`the range from 10 to 20 ␮m. We assumed 10 ␮m
`as a first approximation. As another experimen-
`tal restriction, particles with diameters below 1
`␮m are hardly detectable with the LD configura-
`tion used for the presented measurements. There-
`fore the comparison of the two methods is lim-
`ited to one decade of particle sizes from 1 to
`10 ␮m.
`
`The analysis of the drug was performed in the
`case of formulation C with an UV/VIS scanning
`spectrophotometer at the wavelength ␭ ⫽ 218 nm
`and sometimes additionally at the wavelength
`␭ ⫽ 276 nm. The detection of the other two for-
`mulations A and B was performed with stan-
`dardised HPLC because of their lower drug
`concentrations. The amount of DEHS was deter-
`mined by weight.
`
`Particle size calibration
`For the control of the reliability of the gener-
`ated data the laser diffraction apparatus was
`tested with a reference reticle. The reference ret-
`icle consists of silicon particles of defined sizes
`deposited onto a glass slide. The size distribution
`of the reticle was measured with the laser dif-
`fraction apparatus used for the measurements
`and with a laser diffraction apparatus of the same
`type as a reference. The results were compared
`with the nominal values given for the reference
`reticle. The laser diffraction analyser including
`the throat (configuration with windows before
`the bend; Fig. 1) was additionally tested with a
`monodisperse aerosol. The generation process of
`the test aerosol is based on the Sinclair-LaMer
`principle by condensation of the vaporized
`aerosol material at nuclei. The aerosol consisted
`
`FIG. 1. Front side view of the experimental set-up for
`simultaneous particle size distribution measurements
`with the cascade impactor and the laser diffraction
`method. The distance from the centre of the measurement
`cone to the lens is 4 cm. The cascade impactor is used in
`a turned position for technical reasons.
`
`Downloaded by Society - Active - International Society for Aerosols in Medicine (ISAM) from www.liebertpub.com at 11/09/20. For personal use only.
`
`Liquidia's Exhibit 1058
`Page 3
`
`

`

`314
`
`of DEHS (di-2-ethylhexyl-sebacate). Three differ-
`ent monodisperse particle size distributions with
`D50 values between 2 and 6 ␮m were generated
`and measured simultaneously with the laser dif-
`fraction analyser and the cascade impactor.
`
`Dedicated set-up
`We decided to stay as close as possible to the
`induction port described by the USP 26(1) and
`other pharmacopeia. Therefore, the sample in-
`duction port was used and adapted to the re-
`quirements of the LD method. In addition to mea-
`surements under ambient humidity (relative
`humidity RH ⬇ 30–45%) the particle size distri-
`bution was investigated under water vapor satu-
`rated air (RH ⬎ 90%) conditions to study the
`evaporation effect of the aqueous aerosols. The
`air inlet vents of the inhaler or the complete in-
`haler device were housed and flooded with wa-
`ter vapor saturated air which was produced by a
`humidifier operating ⬃2°C above room temper-
`ature. Excess humidified air escaped to the sur-
`rounding. The schematic experimental set-up is
`shown in Figure 1.
`In order to measure the scattered laser light in-
`tensity of the aerosol passing the induction port,
`two holes were drilled perpendicular to the air
`duct which were sealed with O-rings and glass
`windows. A three-dimensional view of the mod-
`ified USP throat is presented in Figure 2. Unless
`stated otherwise, the laser beam crossed the
`
`FIG. 2. Visualisation of the modified USP throat. In the
`direction of air flow: (a) Windows before the bend. (b)
`Windows after the bend. The inlet orifice for the laser
`beam is not visible.
`
`ZIEGLER AND WACHTEL
`
`aerosol exiting the inhaler before the bend of the
`USP throat (Fig. 2A), because the optional posi-
`tion “after the bend” (Fig. 2B) is expected to have
`a limited measurement range.
`This bend represents a first impaction stage for
`large particles, and therefore these particles can
`be detected neither by the laser diffraction nor by
`the cascade impactor. From the point of view of
`quality control of a spraying device, the windows
`positioned before the bend are preferred, because
`in this position all droplets can be detected by the
`laser system. Irrespectively of the window posi-
`tion it is possible with this set-up to measure the
`PSD with the cascade impactor and the laser dif-
`fraction method simultaneously. To ensure suffi-
`cient drug deposition on all the impactor plates
`to allow for UV spectrophotometric or HPLC
`analysis, four to eight actuations per measure-
`ment were collected. However, for the laser dif-
`fraction device one single shot would be suffi-
`cient. The laser diffraction data was analysed
`based on the Mie-theory which is applicable for
`transparent spheres. For that purpose the refrac-
`tion and absorption index of the droplets must be
`known. The refraction index of the aqueous
`aerosol particles was 1.33 and the absorption was
`0.0. For the DEHS particles, the refraction index
`was 1.45 and the absorption was 0.0. It is impor-
`tant to use the Mie correction to take into account
`the increased scattering of light from smaller
`droplets compared to the Fraunhofer theory.(11,12)
`
`Data and statistical analysis
`The PSD measured with laser diffraction was
`calculated automatically from the scattered light
`intensities striking the 31 detector elements. The
`Sympatec HELOS software used for the calcula-
`tion was WINDOX version 3.3.
`The basis for the calculation of the PSD mea-
`sured with the cascade impactor was the total
`mass detected with the photometer or HPLC; that
`is, the total mass is the sum of all masses recov-
`ered on the different impaction stages and in the
`USP throat for all measurements with LD before
`the bend (Fig. 2A). In the alternative position “af-
`ter the bend” (Fig. 2B) the mass deposited in the
`USP throat was excluded. The implicit assump-
`tions for the comparison of aerodynamic diame-
`ter (dae) measured by ACI and geometric diame-
`ter (dg) measured by LD are a constant density of
`the particles, for example, ␳p ⫽ 1 g/cm3 for the
`present aqueous formulations, and constant ho-
`
`Downloaded by Society - Active - International Society for Aerosols in Medicine (ISAM) from www.liebertpub.com at 11/09/20. For personal use only.
`
`Liquidia's Exhibit 1058
`Page 4
`
`

`

`COMPARISON OF CASCADE IMPACTION AND LASER DIFFRACTION
`
`315
`
`mogeneous concentration of drug among all
`droplets, the latter being required for the ACI
`analysis. In the Stokes regime, theory predicts
`dae ⫽ (␳p/␳ref)*dg, with ␳ref ⫽ 1 g/cm3. Therefore,
`in the present case of aqueous formulations, the
`diameters should be equal.
`All PSD data were converted in percentage of
`the cumulative undersize fraction CF with relation
`to the cut-off diameters of the cascade impactor,
`for example, CF(5.8 ␮m) means the fraction in per-
`centage of a particle ensemble with diameters less
`or equal than 5.8 ␮m. The PSD and the character-
`istic aerosol parameters D50, ␴g, and fine particle
`fraction (FPF) (⬍5.8 ␮m) measured with the two
`particle size detection methods were evaluated
`qualitatively (visual assessment) and if appropri-
`ate quantitatively by means of a significance analy-
`sis (t-test, confidence intervals(13)). The correlations
`between the different measurements were charac-
`terized by linear regressions between the cumu-
`lated fractions of the respective size distributions.
`The geometric standard deviation ␴g is given
`by the following:
`冱ni(ln di ⫺ ln dg)2
`ᎏᎏᎏ
`N ⫺ 1
`
`␴g ⫽冤
`
`冥1/2
`
`dg ⫽ (d1 . . . dN)1/N (1)
`
`RESULTS
`
`Reliability tests
`The results of the reticle measurements are
`shown in Table 1. In order to obtain representa-
`tive results, seven measurements per laser dif-
`fraction analyser at different reticle positions
`were performed. The results of the test analyser,
`which was used for all subsequent investigations,
`show excellent correspondence to the reference
`analyser results (t-test, n ⫽ 7, p ⬎ 0.05). However,
`all nominal values are slightly but significantly
`(NE, t-test, n ⫽ 7, p ⬍ 0.05) higher than the mea-
`sured ones.
`Since the reticle spot diameters are quite large
`it is reasonable to control the reliability of the
`laser analyser in a size range less than 10 ␮m. No
`reticle was available in this size interval. There-
`fore an aerosol generator was used. The charac-
`teristic parameters of the monodisperse PSD gen-
`erated by the MAG-2010 aerosol generator are
`presented in Table 2. Three different boiler tem-
`peratures and hence three PSD were investigated
`simultaneously with the laser diffraction appara-
`tus and the cascade impactor. The cascade im-
`pactor served as the reference test method.
`The D50 values show differences from 0.4 to 0.6
`␮m between the two detection methods (t-test,
`n ⫽ 8, p ⬎ 0.05). Differences of this order of mag-
`nitude are expected due to slightly different cal-
`ibrations and the completely different operating
`principles of LD and ACI. All geometric standard
`deviations (t-test, n ⫽ 8, p ⬎ 0.05) are statistically
`equal.
`
`INFLUENCE OF THE THROAT
`MODIFICATION ON THE PSD
`
`The original induction port was modified and
`the usual position of the impactor was changed
`during the simultaneous measurements with
`
`where ni number of particles with diameter di;
`N ⫽ total number of particles; and dg geometric
`particle diameter.
`Under the prerequisite of a log-normal distri-
`bution (the logarithm of the particle diameters is
`normal distributed) the geometric standard devi-
`ation is equal to the following:
`D84ᎏ
`D50ᎏ
`D84ᎏ
`D16
`D16
`D50
`Eq. 2 is used in the following for calculating ␴g.
`D50 is the median diameter, D16 and D84 are the
`diameters at which the cumulative size distribu-
`tion reaches 16% and 84%, respectively.
`
`␴g ⫽
`
`⫽
`
`⫽冤
`
`冥1/2
`
`(2)
`
`TABLE 1. PSD OF A RETICLE MEASURED WITH TWO LASER DIFFRACTION ANALYSERS
`OF THE SAME TYPE (TEST ANALYZER AND REFERENCE ANALYZER)
`
`Test analyzer (n ⫽ 7)
`
`Reference analyzer (n ⫽ 7)
`
`Nominal value
`
`D10 [␮m] ⫾ SD
`D50 [␮m] ⫾ SD
`D90 [␮m] ⫾ SD
`
`27.49 ⫾ 0.84
`36.85 ⫾ 1.58
`47.03 ⫾ 2.12
`
`27.61 ⫾ 0.47
`36.91 ⫾ 1.16
`47.54 ⫾ 2.48
`
`30.61
`39.05
`49.69
`
`The mean values of D10, D50, and D90 are compared with the nominal value. Measurements according to reticle
`manufacturer’s instructions without throat.
`
`Downloaded by Society - Active - International Society for Aerosols in Medicine (ISAM) from www.liebertpub.com at 11/09/20. For personal use only.
`
`Liquidia's Exhibit 1058
`Page 5
`
`

`

`316
`
`T ⫽ 180°C
`
`T ⫽ 210°C
`
`T ⫽ 240°C
`
`ZIEGLER AND WACHTEL
`
`TABLE 2. PSD OF A MONODISPERSE TEST AEROSOL OF DEHS
`
`D50 ⫾ SD [␮m]
`␴g ⫾ SD
`D50 ⫾ SD [␮m]
`␴g ⫾ SD
`D50 ⫾ SD [␮m]
`␴g ⫾ SD
`
`LD (n ⱖ 8)
`
`1.92 ⫾ 0.10
`1.17 ⫾ 0.32
`3.33 ⫾ 0.18
`1.16 ⫾ 0.08
`6.03 ⫾ 0.30
`1.19 ⫾ 0.07
`
`ACI (n ⱖ 8)
`
`2.29 ⫾ 0.38
`1.32 ⫾ 0.32
`3.90 ⫾ 0.06
`1.12 ⫾ 0.03
`5.60 ⫾ 0.17
`1.15 ⫾ 0.25
`
`Difference
`
`⫺0.37
`⫺0.15
`⫺0.57
`⫺0.04
`⫺0.43
`⫺0.04
`
`The particle size was tuned by the temperature T inside the Sinclair LaMer generator. For each temperature, at least
`eight measurements were performed. LD measurements before the bend (cf. Fig. 2a).
`
`laser diffraction and cascade impactor. These
`modifications do not distort the PSD, as shown
`in Figure 3. The linear correlation of the cu-
`fractions yields CF(modif) ⫽ 0.87 ⫹
`mulated
`0.98*CF(original) with R ⫽ 0.999 for n ⫽ 9 stages,
`including filter. The cumulative fraction curves
`overlap and justify the use of the modified throat
`for the correlation studies. For the experiment
`the formulation C with the highest concentration
`(c ⫽ 0.833%) was used and all measurements
`were performed under saturated air conditions
`(RH ⬎ 90%).
`
`INFLUENCE OF THE RELATIVE
`HUMIDITY ON THE PSD
`
`It is known that the humidity of the air affects
`the PSD of aqueous aerosols measured with the
`
`cascade impactor.(5) Due to evaporation, the size
`distribution is shifted to smaller particles if RH is
`reduced. Even if the laser diffraction method was
`used, where evaporation should not play such a
`dominant role as for the cascade impactor be-
`cause of shorter times of flight, the PSD depends
`also on the relative humidity of the ambient air.
`LD measurements at 28 L/min and two humid-
`ity levels are presented in Figure 4. For compar-
`ison, a measurement without air flow at an ini-
`tial RH of 30–45% has been added. However, it
`is not recommended to use the modified throat
`without air flow, because the aerosol fills the
`whole throat and deposits on the windows. The
`data relate to laser diffraction measurements on
`formulation C with the highest drug concentra-
`tion (c ⫽ 0.833%). The flow rate was 28.3 L/min.
`The linear correlation of the cumulated fractions
`yields CF(RH 30–40%) ⫽ ⫺3.07 ⫹ 0.92*CF(RH ⬎
`
`FIG. 3. Cumulative undersize fraction in dependence of the cut-off diameters for ACI (standard set up) and with
`modified throat (windows before the bend). The full lines are sigmoidal fits. Formulation C (c ⫽ 0.833%) was used.
`
`Downloaded by Society - Active - International Society for Aerosols in Medicine (ISAM) from www.liebertpub.com at 11/09/20. For personal use only.
`
`Liquidia's Exhibit 1058
`Page 6
`
`

`

`COMPARISON OF CASCADE IMPACTION AND LASER DIFFRACTION
`
`317
`
`FIG. 4. The RH of the air influences the laser diffraction results. The detected FPF(⬍5.8 ␮m) value increases and the
`D50 decreases with increasing humidity. For comparison, a measurement without air flow was included (dotted line).
`Formulation C (c ⫽ 0.833%) was used.
`
`90%) with R ⫽ 0.987 for n ⫽ 9 stages, including
`filter.
`
`FLOW RATE DEPENDENCE ON THE PSD
`
`The PSD was investigated by laser diffraction for
`different flow rates and under saturated air condi-
`tions (Fig. 5). The flow rate was varied between 18
`
`and 38 L/min. No systematic dependence was es-
`tablished between the flow rate and the D50 values
`or FPF respectively. The measurements were per-
`formed with the formulation C with concentration
`c ⫽ 0.833% under saturated air conditions. As
`shown in Figure 5, the D50 value varied between
`4.32 and 4.64 ␮m, which is well within an error
`band of ⫾0.2 ␮m which must be taken into account
`when performing LD measurements.
`
`FIG. 5. Cumulative Fraction (CF) versus particle diameter measured by LD. The flow rate was varied between 18
`and 38 L/min. Formulation C (c ⫽ 0.833%) under saturated air conditions (RH ⬎ 90%).
`
`Downloaded by Society - Active - International Society for Aerosols in Medicine (ISAM) from www.liebertpub.com at 11/09/20. For personal use only.
`
`Liquidia's Exhibit 1058
`Page 7
`
`

`

`318
`
`ZIEGLER AND WACHTEL
`
`TABLE 3. INFLUENCE OF THE MEDIAN PARTICLE SIZE ON THE DOSE TRANSMITTED BY
`THE USP SAMPLE INDUCTION PORT (NOT MODIFIED, ACCORDING TO USP(1))
`
`Device
`
`MMD ⫾ SD
`(␮m)
`
`GSD ⫾ SD
`(⫺)
`
`Delivered dose ⫾ SD
`(␮g)
`
`Residue in throat ⫾ SD
`(␮g)
`
`Dose after bend ⫾ SD
`(␮g)
`
`1
`2
`3
`4
`
`4.3 ⫾ 0.1
`4.7 ⫾ 0.1
`8.4 ⫾ 0.4
`11.6 ⫾ 0.3
`
`1.7 ⫾ 0.02
`1.7 ⫾ 0.02
`1.8 ⫾ 0.00
`1.7 ⫾ 0.03
`
`4.7 ⫾ 0.3
`5.4 ⫾ 0.3
`5.5 ⫾ 0.4
`5.8 ⫾ 0.2
`
`0.3 ⫾ 0.0
`0.9 ⫾ 0.2
`1.5 ⫾ 0.2
`3.9 ⫾ 0.02
`
`4.2 ⫾ 0.2
`4.3 ⫾ 0.1
`3.7 ⫾ 0.1
`2.1 ⫾ 0.03
`
`The results are based on three measurements each. Experimental inhalers with modified nozzles. Size data refer to
`values before the bend (cf. Fig. 2a). Air Flow ⫽ 28.3 L/min. Formulation A (c ⫽ 0.049%) was used at RH ⬎ 90%.
`
`EFFECT OF THE INDUCTION PORT
`
`In order to investigate the influence of the bend
`of the induction port, two induction ports were
`used. One port had the windows in front of the
`bend (Fig. 2A), and another port had the win-
`dows behind the bend (Fig. 2B). As long as
`aerosols with MMD of ⬍5 ␮m are investigated,
`there is only a minor influence of the impaction
`process at the bend in the induction port. How-
`ever, larger particles become impacted and the
`correspondence between LD in the preferred con-
`figuration (windows in front of the bend; Fig. 2A)
`and the CI will not hold, because the LD set-up
`is able to detect these large particles, which are
`not sizable by the CI. Table 3 gives an impression
`of the amount of drug retained in the USP throat
`as a function of the particle size MMD for a typ-
`ical aerosol with GSD of ⬃1.7. An air flow of 28.3
`L/min was applied.
`Table 4 gives the direct comparison between
`ACI and LD in the two different configurations.
`Depending on the position of the LD window, the
`effect of the throat is measured or not. This is
`taken into account for the impactor calculations,
`
`too, by considering (Fig. 2A) or excluding (Fig.
`2B) the dose deposited in the throat. Conse-
`quently, CF(CI) reached 100% at the throat cut-
`off size. In case of the window before the bend
`the linear correlation of the cumulated fractions
`yields CF(CI, before) ⫽ 4.2⫹0.939*CF(LD, before)
`with R ⫽ 0.998 for n ⫽ 9 stages, including filter.
`The configuration window after the bend yields
`CF(CI, after) ⫽ 1.8 ⫹ 0.998*CF(LD, after) with
`R ⫽ 0.997 for n ⫽ 9 stages, including filter. The
`correlation of data is presented in Figure 6.
`
`CORRELATION BETWEEN ACI AND LD
`
`The motivation for the present comparison
`between ACI and LD is illustrated by Figure 7.
`It shows the particle size distributions for for-
`mulation C, measured separately with the cas-
`cade impactor at RH of ⬎90% and the laser dif-
`fraction method under ambient conditions. The
`cumulative fractions differ significantly from
`each other for diameters less than 9 ␮m. The lin-
`ear correlation of the cumulated fractions yields
`CF(CI, RH 90%) ⫽ ⫺8.12 ⫹ 0.999*CF(LD, RH
`
`TABLE 4. D50, ␴G, AND FPF (⬍5.8 ␮M) FOR THE DIFFERENT CONFIGURATIONS
`
`LD BEFORE AND AFTER THE BEND (CF. FIG. 2)
`
`Formulation C
`(c ⫽ 0.833%)
`LD before bend
`
`Formulation C
`(c ⫽ 0.833%)
`LD after bend
`
`D50 ⫾ SD [␮m]
`␴g ⫾ SD
`FPF ⫾ SD [%]
`
`ACI (n ⫽ 13)
`
`4.43 ⫾ 0.19
`1.86 ⫾ 0.14
`68.5 ⫾ 2.3
`
`LD (n ⫽ 12)
`
`4.59 ⫾ 0.17
`1.76 ⫾ 0.04
`66.2 ⫾ 2.7
`
`ACI (n ⫽ 6)
`
`4.17 ⫾ 0.26
`1.61 ⫾ 0.04
`77.2 ⫾ 2.5
`
`LD (n ⫽ 6)
`
`4.12 ⫾ 0.15
`1.73 ⫾ 0.04
`74.2 ⫾ 1.9
`
`Formulation C (c ⫽ 0.833%) was used. Data of configuration “LD before the bend” (Fig. 2a) are taken from Table
`5. For the CI results in the configuration “LD after the bend” (Fig. 2b), the dose deposited in the throat was excluded
`from the calculation. RH ⬎ 90%.
`
`Downloaded by Society - Active - International Society for Aerosols in Medicine (ISAM) from www.liebertpub.com at 11/09/20. For personal use only.
`
`Liquidia's Exhibit 1058
`Page 8
`
`

`

`COMPARISON OF CASCADE IMPACTION AND LASER DIFFRACTION
`
`319
`
`FIG. 6. Correlation of Cumulative Fraction (CF) measured with the ACI as function of the Cumulative Fraction (CF)
`measured with LD. Two cases are compared: Window before the bend and window after the bend. Test substance is
`formulation C (c ⫽ 0.833%) under saturated air conditions (RH ⬎ 90%).
`
`ambient) with R ⫽ 0.987 for n ⫽ 9 stages in-
`cluding filter.
`The comparison is restricted to the range from
`1 ␮m (detection limit of LD in the present set-up
`is between 0.5 and 1 ␮m) up to 9 ␮m, which is
`the highest cut-off diameter of the cascade im-
`pactor at 28.3 L/min. Including the throat, the
`comparison may be extended to approximately
`10 ␮m.
`The most obvious way to investigate the cor-
`relation of two PSD analysers is the simultaneous
`
`measurement of the particle size distribution with
`both methods. The correlation studies were per-
`formed at RH of ⬎90% (measurement of RH be-
`hind the impactor) and at a flow rate of 28.3
`L/min for all drug formulations. The modified
`induction port having the inlet and outlet win-
`dows for the laser beam in front of the bend (Fig.
`2A) was used. The experimental set-up is de-
`picted in Figure 1. In Figure 8, the histograms il-
`lustrate the PSD correlation between the LD and
`ACI method.
`
`FIG. 7. Comparison of the Cumulative Fraction (CF) for different measurement conditions (ACI versus LD and
`RH ⬎ 90% versus RH ⬇ 30–45%). The distributions were not measured simultaneously. Formulation C (c ⫽ 0.833%)
`was used.
`
`Downloaded by Society - Active - International Society for Aerosols in Medicine (ISAM) from www.liebertpub.com at 11/09/20. For personal use only.
`
`Liquidia's Exhibit 1058
`Page 9
`
`

`

`320
`
`ZIEGLER AND WACHTEL
`
`FIG. 8. Cumulative Fraction (CF) versus the cut-off diameters of the ACI for the formulation A (c ⫽ 0.049%), for-
`mulation B (c ⫽ 0.198%), and formulation C (c ⫽ 0.833%). Simultaneous measurements under saturated air conditions
`(RH ⬎ 90%).
`
`Figure 9 shows an excellent correspondence be-
`tween the LD and the ACI results. The linear cor-
`relation of the cumulated fractions yields for for-
`mulation A: CF(ACI) ⫽ 0.56 ⫹ 1.03*CF(LD) with
`R ⫽ 0.997, for formulation B: CF(ACI) ⫽ ⫺0.5 ⫹
`
`0.97*CF(LD) with R ⫽ 0.997, and for formulation
`C: CF(ACI) ⫽ 4.68 ⫹ 0.95*CF(LD) with R ⫽ 0.998.
`All values refer to n ⫽ 9 stages including filter.
`Formulation C shows the strongest deviation
`from the ideal regression line which would read
`
`Downloaded by Society - Active - International Society for Aerosols in Medicine (ISAM) from www.liebertpub.com at 11/09/20. For personal use only.
`
`Liquidia's Exhibit 1058
`Page 10
`
`

`

`COMPARISON OF CASCADE IMPACTION AND LASER DIFFRACTION
`
`321
`
`FIG. 9. Cumulative fraction (CF) measured with the ACI in dependence of the cumulative fraction (CF) measured
`with LD. The experimental data represent the respective cut-off points of the ACI (i.e., the CF values for the 0.4, 0.7,
`1.1, 2.1, 3.3, 4.7, 5.8, 9.0 ␮m cut-off sizes). The largest value corresponds to the throat value. Each formulation is close
`to the ideal case (straight line) where CFACI and CFLD should be equal.
`
`CF(ACI) ⫽ 0 ⫹ 1.0*CF(LD) with R ⫽ 1. The mea-
`surable content of fine particles with diameter be-
`low 1 ␮m gives raise to this deviation, as our LD
`apparatus does not detect these fines.
`Table 5 summarizes the corresponding charac-
`teristic aerosol parameters D50, ␴g, and FPF(⬍5.8
`␮m). The D50 values determined by ACI and LD,
`respectively, coincide within intervals of the cor-
`responding standard deviations. The overall vari-
`ability as documented by the standard deviations
`is comparable for LD and ACI. The high numbers
`of independent experiments (12 ⬍ n ⬍ 18) result
`in t-tests indicating a statistically significant dif-
`ference below a difference of the mean D50 of ACI
`and LD of 0.2 ␮m. In our experiments, this is the
`case for formulation B, where the absolute dif-
`ference between the mean D50 values is 0.18 ␮m.
`In Table 6, the different cut-off points of the
`ACI are summarized in three size intervals [0 ␮m;
`
`1.1 ␮m], [1.1 ␮m; 4.7 ␮m] and [4.7 ␮m; ⬃10 ␮m].
`The corresponding cumulated fractions CF are
`compared for the ACI and LD method. Except for
`the [0 ␮m; 1.1 ␮m] interval good equivalence be-
`tween the ACI and LD method can be found. The
`higher CF values of the ACI evaluation in com-
`parison to the LD for the [0 ␮m; 1.1 ␮m] interval
`are caused by the detection limit of the LD. As
`shown in Figure 8, the quality of the correspon-
`dence between LD and ACI can be asses