Next Generation Pharmaceutical Impactor (A New Impactor ... · such as materials, clamping mechanisms and rel-ative location of the stages. These characteristics have negligible effect

JOURNAL OF AEROSOL MEDICINEVolume 16, Number 3, 2003© Mary Ann Liebert, Inc.Pp. 283–299

Next Generation Pharmaceutical Impactor (A New Impactor for Pharmaceutical Inhaler Testing).

Part I: Design

VIRGIL A. MARPLE, Ph.D.,1 DARYL L. ROBERTS, Ph.D.,2 FRANCISCO J. ROMAY, Ph.D.,2

NICHOLAS C. MILLER, Ph.D.,3 KEITH G. TRUMAN, B.Sc.(hons),4

MICHIEL VAN OORT, Ph.D.,4 BO OLSSON, Ph.D.,5

MICHAEL J. HOLROYD, M.A.(hons)Cantab.,6 JOLYON P. MITCHELL, Ph.D.,7

and DIETER HOCHRAINER, D.Phil.8

ABSTRACT

A new cascade impactor has been designed specifically for pharmaceutical inhaler testing. Thisimpactor, called the Next Generation Pharmaceutical Impactor (NGI), has seven stages and is in-tended to operate at any inlet flow rate between 30 and 100 L/min. It spans a cut size (D50) rangefrom 0.54-mm to 11.7-mm aerodynamic diameter at 30 L/min and 0.24 mm to 6.12 mm at 100 L/min.The aerodynamics of the impactor follow established scientific principles, giving confident par-ticle size fractionation behavior over the design flow range. The NGI has several features to en-hance its utility for inhaler testing. One such feature is that particles are deposited on collectioncups that are held in a tray. This tray is removed from the impactor as a single unit, facilitatingquick sample turn-around times if multiple trays are used. For accomplishing drug recovery, theuser can add up to approximately 40 mL of an appropriate solvent directly to the cups. Anotherunique feature is a micro-orifice collector (MOC) that captures in a collection cup extremely smallparticles normally collected on the final filter in other impactors. The particles captured in theMOC cup can be analyzed in the same manner as the particles collected in the other impactorstage cups. The user-friendly features and the aerodynamic design principles together providean impactor well suited to the needs of the inhaler testing community.

Key words: impactor, pharmaceutical inhaler, calibration, design

283

INTRODUCTION

THE PHARMACEUTICAL INDUSTRY uses pressur-ized metered-dose inhalers (MDIs) and dry-

powder inhalers (DPIs) as a means to aerosolizemedicines for inhalation. The regional depositionin the lung is a strong function of the aerody-namic diameter of the particles. In effect, the lung

1University of Minnesota, Mechanical Engineering Department, Minneapolis, Minnesota.2MSP Corporation, Shoreview, Minnesota.3Nephele Enterprises, White Bear Lake, Minnesota.4GlaxoSmithKline, Ware, Hertfordshire, United Kingdom.5AstraZeneca R&D Lund, Lund, Sweden.6Phoqus Pharmaceuticals Ltd., West Malling, Kent, United Kingdom.7Trudell Medical International, London, Ontario, Canada.8Boehringer Ingelheim Pharma AG, Ingelheim am Rhein, Germany.

is a type of aerosol classifier, with the larger par-ticles collected in the upper airways, mainly at bi-furcations in the flow path, and the smaller par-ticles depositing deeper in the lung. Impactors arethe instruments of choice for the in vitro assess-ment of delivery efficiency of inhalation productsfor three predominant reasons. First, by using aspecific drug assay, the size distribution of drugparticles that is obtained is drug specific, not con-founded with any non-drug material that may bein the sample. Second, the size distribution ismeasured on all of the drug that is delivered,rather than on a sub-sample that may or may notbe representative. Third, impactors classify parti-cles according to aerodynamic diameter (Appen-dix). Thus, the cascade impactor is a naturalchoice for an instrument to evaluate the aerosolemitted from an MDI or DPI. However, it is im-portant to recognize that the cascade impactor isnot a lung simulator because of many features,including the geometry at the point of impact, col-lection surface hardness and coating, and opera-tion at constant flow rate. In particular, collectionstages in the impactor do not correspond to anyspecific deposition sites in the lung.

The first MDIs were commercially producedin 1955.1 Shortly thereafter, a method was pub-lished for sizing aerosols using the light scatterdecay technique2 and was put into use in RikerLaboratories at about that time.3 This techniquespecified that the test aerosol be fired into a darkcubic box, the scatter of a beam of light insidethe box was monitored, and the mean size ofaerosol could be calculated. The use of a cascadeimpactor for pharmaceutical aerosols was firstdescribed in 1969 by Polli et al.4 from MerckScharp & Dohme, using the impactor laterknown as the Delron. This impactor had beenfirst described 10 years earlier.5 The Andersencascade impactor was developed in 1958 as a six-stage design to sample airborne viable particlesonto agar gel held in a Petri dish.6 In 1971, aneight-stage Andersen Mark I impactor was in-troduced in Riker Laboratories.3 The Multi-Stage Liquid Impinger was originally developedin 1966 by May to sample viable particles into aliquid.7 The use of this instrument for pharma-ceutical applications was subsequently de-scribed by Bell et al.,8 and the impactor was latermodified by adding one stage to obtain addi-tional resolution for the smaller particle sizes.9

A feature of this instrument is that particles arecollected on a glass frit that is in contact with

liquid, so the problem of particle bounce and re-entrainment is thought to be eliminated.

Two other devices to size particles by inertialimpaction were developed as quality controltools within the laboratories of pharmaceuticalcompanies. These single impaction stage instru-ments, termed “twin impingers,” have been usedwith pharmaceutical aerosols because they per-mit significant labor savings compared to the con-siderable time required to analyze samples froma multistage impactor. A glass twin impinger de-scribed by Hallworth and Westmoreland10 al-luded to an earlier version described by Hall-worth et al.11 This instrument, along with a metaltwin impinger of different design, first appearedin the British Pharmacopoeia.12 Comparisonswere reported by Aiache et al.13 The two devices,although convenient to use, had inherent limita-tions in the adequacy of size description,14 andthe measurements were not comparable betweenthe instruments. Nevertheless, they remain in usein some countries because of their operationaleconomy and their historical use in registeringsome products.

Figure 1 shows the instruments recommendedfor use by the current standards of the U.S. Phar-macopeia15 and European Pharmacopoeia.16

The first cascade impactor designed specifi-cally for pharmaceutical aerosol applications wasthe Marple-Miller Impactor17 (MMI). Althoughthe MMI did not find favor within industry, itdemonstrated that a cascade impactor could bedesigned specifically for sampling MDI and DPIaerosols and that its unique external samplingcups could offer productivity improvements overexisting impactors. Discussions among pharma-ceutical industry scientists about the inadequa-cies of the various available impactors led to theestablishment in the mid-1990s of the Next Gen-eration Impactor (NGI) consortium. In time, theconsortium recruited members representing all ofthe major participants of the worldwide inhala-tion product testing community. The specific pur-pose of this consortium was to define the re-quirements for a new impactor and to fund itsdevelopment and testing. The consortium con-sidered proposals from several parties to under-take this development and awarded the projectto MSP Corporation (Minneapolis, MN). The firstproject meeting was held in December 1998.

The design of the NGI was a cooperative effortbetween the NGI Consortium, spearheaded by itsExecutive Committee, working with the impactor

MARPLE ET AL.284

design team. The Executive Committee fielded acontinuing stream of design questions and issuesthroughout the project, giving regular input andguidance. At major points of design decisions, thedesign team would present sound options to the

entire consortium, and the consortium memberswould select the option that they felt best servedthe end-user needs. MSP interviewed users insideconsortium member companies to get first-handinput. Twice in the design process, prototype

NEXT GENERATION PHARMACEUTICAL IMPACTOR DESIGN 285

FIG. 1. Instruments recommended for use by the U.S. Pharmacopeia and European Pharmacopoeia. (A) Andersen-type. (B) Andersen-type with pre-separator. (C) Multistage liquid impinger. (D) Marple-Miller impactor. (E) Twinimpinger (metal impinger). (F) Twin impinger (glass impinger).

NGIs were built and tested by the consortiummembers in their laboratories. The qualitativeand quantitative information from the prototypetesting resulted in important improvements tothe design. This development process was in-tended to ensure that the NGI would be an in-strument that would serve the pharmaceutical in-dustry well for many years.

This paper describes the NGI design and thereasons for its particular configuration. The cali-bration of the NGI, defining the particle collec-tion efficiency curves, is described in a compan-ion paper in this issue.18

USER REQUIREMENTS

The NGI consortium met several times to de-fine items that it considered to be important fea-tures of the NGI. The final result was a list of userrequirements that were divided in two groups: 23“must” features (Table 1) and 12 “want” features(Table 2). The 35 items listed in Tables 1 and 2 be-came the guide for all design decisions for the de-velopment of the NGI.

From a designer’s viewpoint, the user require-ments can be divided into two categories: (1)

those that affect the design of the impactor stagesand (2) those that affect the overall impactor lay-out. The first category relates to the aerodynamicdesign of the stage nozzles and the flow passagesbetween stages. The particular geometric config-uration in these regions determines the collectionefficiency characteristics of the impactor. The sec-ond category relates more to mechanical features,such as materials, clamping mechanisms and rel-ative location of the stages. These characteristicshave negligible effect on the aerodynamic per-formance of the impactor but have a large effecton chemical compatibility, reliability, and oper-ating convenience. Particular focus was given toensuring that the ergonomics of the new impactorwere optimized from a user perspective and alsothat the impactor could be automated relativelyeasily.

STAGE DEFINITION

The first step in designing a cascade impactoris to define the number of stages and the particlecut size of the stages at the desired flow rate orflow rate range. The requirements M-2, M-4, M-5, M-7, and W-2 (Tables 1 and 2) most closely

MARPLE ET AL.286

TABLE 1. “MUST” NGI FEATURES

M-1 Automatable, but suitable for manual operationM-2 Operates over a range of flow rates 30–100 L/minM-3 Calibration data for the flow rates of 30–100 L/min are availableM-4 Capable of fully characterizing the “less than 10 mm” cloud size rangeM-5 Needs right number of stages at right cut-offs (independent of flow rate); minimum of 5 stages 0.5–

5 mm, one stage between 5 and 10 mm plus high capacity stage at 10 mm or higher (pre-separator)M-6 The stages will be followed by a micro-orifice collector that is 90% efficient for particles larger than or

equal to 0.2-mm aerodynamic diameter (or smaller particles; see W-6 in Table 2)M-7 Efficiency curves for stages must be appropriate; GSD for all stages similar; overlap between stages is

minimizedM-8 Deposition profile unaffected by stage loadings (up to 10 mg)M-9 Operates using defined entry conditionsM-10 The empty volume of the impactor must be no more than 1.5 L, measured from the inlet of the

USP/EP inlet to the exit of the impactor body, including the pre-separatorM-11 Low wall losses; not more than 5% on any stage and not more than 5% total on all stagesM-12 Good drug recovery (mass balance)M-13 No bounce/re-entrainmentM-14 Capable of being grounded; unaffected by staticM-15 Physically robustM-16 Constructed using inert/robust materials (can use common solvents)M-17 Good accuracy and precision (65%)M-18 Operator independent (no statistical difference between results between independent laboratories)M-19 Acceptable to regulators/PharmacopoeiaM-20 Designed and manufactured to ISO 9000 or equivalentM-21 Applicable to all single shot inhaled delivery systems (MDIs, DPIs, aqueous inhalers)M-22 Easy to qualify and validate in the laboratoryM-23 Fast; cycle time of less than 30 min for manual determinations

relate to the choices that must be made duringthis first design step. Requirement M-2, whichspecifies the flow rate range, is the most quanti-tatively explicit of these requirements. Require-ment M-5 puts a lower bound on the number ofstages (namely six). Requirement M-7 puts aqualitative upper bound on the number of stages(“overlap is minimized”).

Within these constraints, the design teamconsidered several cascade impactor configura-tions having six or seven stages. These consid-erations included cases wherein the user wouldphysically substitute nozzle pieces at will, sothat the cut sizes could be nearly unchanged re-gardless of changes in the flow rate (inter-changeable nozzles). Consortium members se-lected seven stages with no interchangeablenozzles so as to best achieve five stages in therange of 0.5–5 mm over the entire flow range (M-5, Table 1) and to eliminate possibilities for usererror that might arise if interchangeable nozzleswere available. Further, the constraint of mini-mization of stage overlap was met by insistingon a logarithmic spacing of the particle cut sizes(meaning that the ratio of the cut sizes for anytwo neighboring stages is a constant through-out the impactor). The logarithmic spacing ofcut sizes also aids in the intuitive understand-ing of mass distribution bar charts, a qualitativeissue of value to users.

The aerosol dose can be introduced to theNGI through an induction port having the sameinternal dimensions as the port described in theUSP and EP, to meet requirement M-9 and M-19.

OVERALL IMPACTOR LAYOUT

The next step in designing the impactor was tochoose a physical arrangement of the stages andthe associated means of flowing air from onestage to the next. The requirements most relatedto the overall impactor layout were M-1, M-9, M-10, M-11, M-14, M-15, M-16, and M-23. Of theserequirements, “ability to automate, yet suitablefor manual operation” (M-1) was considered themost important.

Most of the requirements were qualitative, andtherefore the design team was free to pose a widerange of options for the impactor layout. At leastnine separate arrangements of the impactor bodywere considered. Both for automation and man-ual use, designs that incorporated external cupsof some type appeared to have an advantage overlayouts where impactor stages must be disas-sembled to get to the internal impaction plates.[The Andersen eight-stage impactor (USP appa-ratus 1) and Marple-Miller (USP apparatus 2) im-pactor, are both shown in Figure 1 and are ex-amples of internal and external impaction platelayouts, respectively.]

Nevertheless, internal impaction plate layoutswere considered. One of these was a “stacked”layout, much the same as the Andersen impactor.However, in this configuration, shown in Figure2, each nozzle plate incorporated the impactionplate for the previous stage. On one stage, thenozzles would be in the central portion of thestage, and the impaction plate would be in theannular area around the nozzles. The nozzle platebefore, and after, this plate would have the noz-


TABLE 2. “WANT” NGI FEATURES

W-1 Cheap; manual version to cost less than UK £5,000W-2 Flow rates able to be continuously varied (see M-2 in Table 1)W-3 Although the “must” specification has a lower flow rate limit of 30 L/min, it is desirable for the lower

flow rate limit to be 5 L/minW-4 Capable of simplification for QC use, e.g., able to remove one or more stagesW-5 Works on single actuation by minimizing wash volumesW-6 The stages will be followed by a micro-orifice collector that is 90% efficient for particles larger than or

equal to 0.1-mm aerodynamic diameter (up to 0.2 mm is acceptable, see M-6 in Table 1)W-7 Capable of characterizing particle cloud greater than 10 mm; pre-separator has to be well characterized

and have a sharp cut offW-8 For automated systems, in-built diagnostics (obvious when instrument is functioning correctly)W-9 Capable of being operated over a range of temperature/humidity (20–30°C; 25–75% RH)W-10 Suitable for nebulizersW-11 The impactor must be easy for technicians to use, including handling, training, and overall ergonomicsW-12 The empty volume of the impactor should be no more than 1 L, measured from the inlet of the

USP/EP inlet to the exit of the impactor body, including the pre-separator (see M-10 in Table 1)

zles in the annular area and the impaction sur-face in the center area. This layout made for avery compact cascade impactor.

Several layouts with external impaction plateswere also considered. Some of these layouts wererejected for reasons of anticipated machining dif-ficulties, and the surviving layouts evolved to onewith all of the impaction plates in one plane. Thisconfiguration resulted in what amounts to a two-piece cascade impactor, the lower part compris-ing the base that retains the collection cups, theupper part consisting of the lid and seal body thatcontains the stage nozzle assembles and inter-stage passageways.

A study was made to determine whether au-tomation was easier with the “stacked” impactor orthe “all impaction plates on one plane (planar)” im-pactor. The steps that would have to be taken to re-move the impaction plates and wash the depositsfrom the impaction plates for each of the layoutswere enumerated. From this analysis, it was deter-mined that the least number of steps would be takenfor one sampling cycle with the external impaction

plate layout of the two-piece impactor. Mock-upsof the competing stack and planar designs were pre-sented to the full consortium along with the analy-sis of the speed and ease of use. The consortiumvoted heavily in favor of the planar, two-piece im-pactor. This layout was then further refined and be-came the basic layout for the NGI.

DETAILED AERODYNAMIC DESIGN

With the number of stages and the design cutpoints decided upon in the early project discus-sions, the next step was to define the exact sizesand numbers of nozzles on each impactor stage.The selection of the nozzle diameters and thenumber of nozzles followed the guidelines de-veloped for impactor design initially described byMarple,19 Marple and Willeke,20 and later refinedby both Rader and Marple,21 and by Fang et al.22

These guidelines are as follows:

1. The Reynolds number (see equation 1) of theflow through the nozzles should be in therange of 500 to 3000.

MARPLE ET AL.288

FIG. 2. Stacked plate impactor (first three stages).

FIG. 3. Schematic diagram of a typical nozzle/impaction plate stage.

2. The nozzle-to-plate distance should be at leastone nozzle diameter and no larger than 10 forround-nozzle impactors.

3. The cross-flow parameter (see equation 2)should be less than 1.2.

If an impactor stage is constructed accordingto these guidelines, experience has shown that itwill have a steep collection efficiency curve andthat its cut point, defined as the particle size thatis collected with 50% efficiency (d50), will be veryclose to the calculated value.

Figure 3 is a schematic diagram of a typicalnozzle plate and impaction plate with the rele-vant dimensions labeled. Conventional theoryfor single-nozzle impactors has shown thatthere are two dimensionless parameters thatare important in defining the flow field throughsuch a nozzle,19 and it is the flow field that isimportant in defining the sharpness of cut ofthe stage.20 These parameters are the ratio ofthe nozzle-to-plate distance to the nozzle di-ameter, S/W, and the Reynolds number, Re, de-fined as:

Re 5 5 (1)4rQ}npmW

rWVo}m

For a specific flow rate, n and W can be selectedto keep Re in the desired range.

The above statements hold true for single-nozzle impactors, and in most cases, for multiplenozzle impactors. However, in multiple-nozzleimpactor stages, there can be a “cross-flow” prob-lem caused by “spent” air from the nozzles nearthe center of the nozzle plate flowing outwardpast the air jets located near the edge of the noz-zle cluster. In some cases this cross-flow can pre-vent the air jets near the edge of the cluster fromreaching the impaction plate. This phenomenonwas studied by Fang et al.,22 who found that across-flow parameter, Xc, could be defined as:

Xc 5 (2)

If the value of this parameter is less than 1.2, thejets from the outer nozzles are not affected.

Thus, the three parameters—Re, S/W, andXc—are important in defining the correct aero-dynamics in an impactor stage. Another dimen-sionless parameter important in determining thecut size of the impactor stage is the Stokes num-ber, St:

St 5 5 (3)4raeQCaedae2}}

9npmW3

4rpQCpdp2}}9npmW3

nW}4Dc


FIG. 4. NGI seal body (top view).

To calculate Cp or Cae, one uses the followingexpression for C:

C 5 1 1 Kn 31.142 1 0.558 exp12 24 (4)

with the Knudsen number, Kn, calculated fromequation 5, using dp or dae as appropriate, de-pending on whether Cp or Cae is to be calculatedfrom equation 4:

Kn 5 or Kn 5 (5)

When the value of dae in equation 3 equals d50,the Stokes number is denoted as St50, and the gov-erning (implicit) equation for the cut size of animpactor stage becomes:

ÏC50 d50 5 !§ ÏSt50 (6)

The value of C50 is computed from equation 4using the diameter d50. Consequently, for each se-lected impactor stage design and flow rate (m, n,W, rae, Q), equations 4 and 6 must be solved si-multaneously, either numerically or iteratively.However, for particles greater than 1 mm the Cun-ningham slip correction is equal to 1.0 for practi-cal purposes, permitting a direct solution of d50from equation 6.

Ideally, the value of ÏSt50 should be near to0.495 if Re is 500–3000, and S/W is at least 1.0 butnot greater than about 10. However, if factorsother than inertia are significant in the collectionmechanism, such as gravitational effects for large

9pmnW3}}4raeQ

2l}dae

2l}dp

0.999}Kn

or very dense particles, or significant deviation inC from unity associated with particles havingdp , 0.1 mm, then ÏSt50 will have a value otherthan 0.495.

Thus, when designing a stage with a specificvalue of d50, n and W are selected so that equa-tion 6 is satisfied,20 while keeping Re in the rangeof 500–3000. By using this method, the values ofn and W in Figure 4 were determined so that thetheoretical cut size values at 60 L/min of stages1 to 7 of the NGI ranged between 0.33 mm and7.8 mm aerodynamic diameter, with equal spac-ing between the values on a logarithmic scale.

Figure 5 depicts the resulting expected perfor-mance based on the above theoretical considera-tions for the NGI within the entire flow range forwhich the impactor was designed. Note that therequirement that there should be five stages withcuts between 0.5 and 5 mm is not strictly met.There are, however, five cut sizes between 0.5 and6.5 mm at all flow rates. This result was judgedto be acceptable, since five stages strictly in the0.5 to 5 mm size range at all flow rates between30 and 100 L/min would have required eightstages, leading to overlapping of the stage effi-ciency curves.

The critical design parameters for the NGIstages including the MOC, including their asso-ciated dimensionless constants—Re, S/W, andXc—are listed in Table 3 for volumetric flow ratesof 30, 60 and 100 L/min. The Re values for theseven impactor stages are in the generally desir-able range of 500–3000, with several stages belowthe range at low flow rates and small cut sizes,

MARPLE ET AL.290

FIG. 5. Theoretical NGI stage cut size versus flow rate.

and above the range at large cut sizes at high flowrates. This outcome is a consequence of the de-sign choice that the NGI should not have inter-changeable nozzles while operating over a largeflow rate range. The high Re for stage one is un-avoidable with a single nozzle design, which inturn was forced by the finding that only a singlenozzle design could accept the airflow directlyfrom the USP/EP induction port without unac-ceptable upstream losses. The values of S/W areall in the desired range of greater than 1.0 andless than 10. Also, the Xc values are all less than1.2. Values of D50 calculated from the archival cal-ibration in the companion publication18 are sum-marized in Table 4 for inlet flow rates of 30, 60,and 100 L/min.

The aerodynamic design of the flow passageswas expected to result in very small interstagelosses. Investigations of losses for a range of typ-ical pharmaceutical aerosols, including 14 MDIformulations and 16 DPI formulations, were con-ducted on prototype NGIs. The sum of the de-position on all surfaces other than collection cups

was measured to be in the range 1–5% of the to-tal delivered sample mass, depending on the nature of the formulation.23,24 For dry-powderaerosols, of course, the stages in the NGI, as forall cascade impactors, must be coated with an ad-herent material unless specific tests show that thecoating is not needed.

Consequently, particles should need to be re-moved only from collection surfaces following aparticle size determination, and the non-collec-tion surfaces in the impactor should need to bewashed only after multiple determinations. Allnon-collection surfaces are either on the lid or theseal body; these two parts disassemble easily andoptionally can be placed in a dishwasher forcleaning.

DETAILED LAYOUT DESIGN

After the basic layout of a planar impactor wasdecided upon, there were several iterations madeto establish the appearance of its components,how the impactor should open and close, whetherthe flow direction should be from left to right orright to left, and to determine the user-friendlyfeatures that could be incorporated. Much of theinput specifying these features came from theuser comments during the testing of the first pro-totypes. The final latching mechanism of the im-pactor was one of the more challenging designproblems, involving careful setting of tolerancesfor the various components to achieve both agood seal (leak rate smaller than 100 Pa/sec) andthe proper jet-to-plate distances.

Figure 6 is a cross-section through a represen-tative stage of the final NGI design. The nozzleassemblies are all held in one plate called the


TABLE 3. VALUES OF RE, S/W, AND XC FOR THE NGI STAGES AT 30, 60, AND 100 L/MIN

Reynolds number at eachflow rate (L/min)

Stage W, mm n S, mm S/W Dc, mm Xc 30 60 100

1 14.3 1 15 1.0 na na 2938 5876 97932 4.88 6 9.764 2.0 na na 1435 2870 47833 2.185 24 6.555 3.0 38 0.35 801 1602 26714 1.207 52 3.621 3.0 38 0.41 669 1339 22315 0.608 152 1.824 3.0 38 0.61 455 909 15156 0.323 396 1.001 3.1 38 0.84 328 657 10957 0.206 630 1.000 4.9 38 0.85 324 647 1079MOC 70 mm 4032 0.500 7.1 75 0.94 149 298 496

TABLE 4. STAGE CUT SIZES FOR THE NGI AT 30-, 60-, AND 100-L/MIN IMPACTOR INLET FLOW RATE

CALCULATED FROM THE ARCHIVAL NGI CALIBRATION18

D50 (microns) at eachflow rate (L/min)

Stage 30 60 100

1 11.7 8.06 6.122 6.40 4.46 3.423 3.99 2.82 2.184 2.30 1.66 1.315 1.36 0.94 0.726 0.83 0.55 0.407 0.54 0.34 0.24

“seal body.” The impaction plates are tear-shapedcups located below the stage nozzles. The largeend of the impaction cup is located directly be-low the nozzles and is where particle collection

occurs. Air flows from the impaction region of thecup to its small end, where the flow is withdrawnupward into a cavity in the lid of the NGI. Thecavity in the lid directs the air to the next stage.

MARPLE ET AL.292

FIG. 6. Cross section of one NGI stage.

FIG. 7. The optional “false” lid for stage 1.

The nozzles of the stages are staggered back-and-forth in the seal body so that the tear shaped cupscan be nested in a compact configuration.

Although Figure 6 shows three pieces (lid, sealbody, and bottom frame) instead of two, as out-lined in the original concept, the lid and the sealbody are secured as one assembly when the NGIis used for routine sampling. To change the im-paction cups between sampling runs, only thecups and tray must be separated from the sealbody/lid assembly.

Two options were considered for the impactioncups. One option was to have all of the cups ma-

chined into a one-piece manifold. The other op-tion was to have individual cups held in a rackthat can be inserted into a frame below the sealbody. In practice, both options can be used in-terchangeably, since the internal airflow pathconfiguration would be the same. For manual op-eration the use of individual cups makes thewashing of the deposits from the cups more man-ageable and is the option shown in this paper.However, for automation, the manifold optionmay be more satisfactory.

The roughness of the surface of the collectioncups can affect wetting of drug recovery solvents


FIG. 8. NGI components. (A) Lid with internal air passages (bottom view). (B) Seal body with nozzles (bottom view).(C) cup tray with cups (top view). (D) Bottom frame with locking handle (top view).

but has no effect on the particle capture efficiencyso long as the surface roughness measure is muchsmaller than the particle stop distance.25 The stopdistance is approximately 25% of the nozzle di-ameter for round jets.19 Applying this criterion tothe smallest nozzles (0.206 mm on stage 7), weconclude that approximately 2.5 microns is theupper limit of acceptable surface roughness (fivepercent of the stop distance on stage 7). Cups of-fered commercially by MSP Corporation have asurface roughness between 0.5 micron and 2 mi-cron. Because native steel typically has a surfaceroughness of 0.3 micron to 0.5 micron, the choiceof 0.5 microns to 2 microns for the surface rough-ness represents a “slight” roughening of the sur-face to improve wetting without affecting theaerodynamics.

The cup for stage 1 is larger than for the otherstages to minimize impaction of larger particlesnear the cup’s vertical wall, identified in earlyprototype testing. This issue is not important forthe other stages since these stages collect smallerparticles that are less prone to this so-called “sec-ondary” impaction. So, for stages 2–7, the di-mension of the nozzle holder determines the cupsize.

For stage 1 only, the option is provided for alid to retain particles within the cup of stage 1, ifsignificant secondary impaction is encounteredwith a specific formulation. This lid rests on theedge of the stage-one cup (Fig. 7). Deposits on theinside surface of this lid can be rinsed easily intothe stage-one cup. Without this “false” lid, thesedeposits would be on the underside of the sealbody of the impactor and would therefore becumbersome to recover. In prototype testing, asmuch as five percent of the mass of material onstage one was found on the false lid for a formu-lation that had a mass median diameter approx-imately equal to the d50 value for stage one. Ingeneral, however, this false lid is not needed.

The cup size for the MOC is also larger thanfor the cups for stages 2 to 7, and is the same sizeas the cup for stage 1. The larger diameter wasnecessary to accommodate the 4032 nozzles in theMOC, and conveniently permits a standard 75-mm filter to be used in an internal filter configu-ration.

Figure 8 shows the individual parts of the NGI.The lid contains the air passages between stages,the seal body contains the stage nozzle plates, andthe bottom frame holds the collection cups in acup tray. The bottom frame also contains hingepins and a past-center clamping handle. The lidcontains the receivers for the hinge pins and theclamping handle.

Figures 9 and 10 shows the NGI in the “closed”and “open” positions, respectively. In Figure 9,the USP/EP induction port is shown on the inletof the first stage. The NGI pivots open on twohinges located at the back of the NGI between thebottom frame and the lid. The seal body is heldto the lid by two limited torque screws (the blackknurled knobs on the upper surface of the lid).When closed, the lid is locked to the bottom frameby a past-center cam lock handle mechanism lo-cated at the front of the NGI. The cup tray andthe seal body are clamped between the lid andbottom frame when the handle is pushed down-ward into a locked position. Tolerances on allparts from the hinges to the clamping mechanismare such that no adjustment screws or springcomponents are required to ensure a leak-tightseal when the NGI is clamped.

Pre-separator design

Design criterion M-5 states that the NGI musthave a high-capacity pre-separator stage to pre-vent oversized particles from depositing at un-wanted places. Because the need for a pre-sepa-

MARPLE ET AL.294

FIG. 9. Closed NGI with USP/EP induction port on theinlet.

FIG. 10. NGI in the “open” position.

rator is formulation dependent, it was designedas an “add-on” unit, typically used for DPIs, thatcan be positioned on the NGI between theUSP/EP induction port and the first stage, asshown in Figure 11. This pre-separator incorpo-rates two collection surfaces working in tandem,as shown in Figure 12.

The first (scalper) collection surface is a circu-lar cup, containing solvent, beneath the inlet ofthe pre-separator and is intended to remove verylarge particles (e.g., lactose carrier particles). Typ-ically, this cup is filled half-way with solvent be-fore testing to reduce particle re-entrainmentfrom the cup. The liquid reservoir gives the unita high solids capacity along with a fairly low in-ternal volume, but by itself does not provide asatisfactorily sharp cut. The volume of this firstcup is approximately 30 mL, so most users loadthe cup with 15 mL of solvent. The pre-separatorperformance should be insensitive to the exactvolume of solvent in the cup.

The scalper is immediately followed by the sec-ond collection surface comprising a more con-ventional impaction stage. This stage has a muchsharper cut than the scalper to eliminate mostparticles that are larger than the cut size of thefirst stage of the impactor itself, while not re-moving particles that are of a size to be capturedon the second impactor stage.

Sealing between the USP/EP induction portand the pre-separator, and between the pre-sep-arator and the first stage is accomplished by 10°tapers at the interfaces, to reduce the number of“O” rings in the inlet of the NGI. If the pre-sep-arator is not used, the tapered exit of the USP/EPinduction port seals to the tapered inlet of the firststage. Whereas the outside dimensions of the in-duction port are customized for these tapers, theinternal dimensions of the induction port arethose described in the USP and EP.

Micro-orifice collector (MOC)

One unique feature of the NGI is that an im-pactor stage of 4032 nozzles, each nominally 70mm in diameter, is used to replace the final filterused in other cascade impactors. There are tworeasons for this feature. First, the collection cupbeneath the micro-orifice nozzles can be handled,and particles analyzed, in the same manner asfrom any of the other collection cups. This aspectis especially important for automation. Second,particles are much more easily dissolved from asolid collecting surface than from the fiber net-work in a filter. The large number of nozzles inthe MOC is necessary to keep the pressure dropwithin a range that can be managed by pumpsnormally found within aerosol laboratories. To


FIG. 12. Exploded view of NGI pre-separator.

FIG. 11. NGI with pre-separator.

keep the crossflow parameter below the criticallimit, the diameter of the nozzle cluster was setto 75 mm (Table 3). The size of the collection cupfor this stage is the same as for stage 1. Four dim-ples on the downstream face of the nozzle platesupport the plate against the cup and maintain aminimum nozzle-to-plate distance.

Some samples may contain particles that are sofine that they are not collected by the MOC. How-ever, since the particle size distributions gener-ated by most inhalers are larger than the cut sizeof the MOC, its efficiency is satisfactory for themajority of drug product formulations. Never-theless, its effectiveness must be evaluated forany new formulation, or inhaler device, duringmethod development by placing a filter down-stream of the MOC and determining the magni-tude of the portion of the dose (if any), that pen-etrates the MOC. Figure 13 shows an externalfilter designed especially for this purpose. Thisfilter holder has a cassette into which the filter isplaced. In this manner the filters are easy to han-dle and the cassette with filter can be transportedwith the cup tray for analysis. An internal filterholder has also been designed for the NGI, in theevent that it is routinely found that a significantproportion of particles penetrates the MOC. Theinternal filter holder is contained in a special cup(Fig. 14) that replaces the cup normally used withthe MOC. It is not necessary to remove the MOCnozzle plate, since any particles passing throughthe MOC nozzles impact directly on the filter up-per surface.

Although the MOC may appear to be an im-paction stage, this is specifically not the intent forits incorporation into the design, and judgmentsabout the size of material collected on the stageare not recommended. The MOC permits signif-icant simplification of analytical procedures forthe many samples for which it is shown, by theprocedures above, to be applicable, but it will not

MARPLE ET AL.296

A.

A.

B.

B.

C.

C.

FIG. 13. External filter holder attached to exit of the NGI.

FIG. 14. Internal filter holder (for insertion in place ofthe MOC cup).

be useful for all formulations, especially thosecontaining a significant portion of very fine ma-terial.

Special features of the NGI

Several special features have been imple-mented that make the NGI easy to use:

1. The “O” rings are held in the seal body bydovetail “O” ring grooves. That is, the groovesare slightly narrower at the surface. Once the“O” ring is inserted in the groove, the taperwill retain it in the groove until removed bythe operator.

2. The seal body can quickly be separated fromthe lid by two thumbscrews. This feature al-lows convenient inspection of the air passagesin the lid or washing of the stage nozzles,which is performed periodically.

3. The lid can be removed from the bottom framehinge pins by opening the lid all the way andsliding the lid off of the pins. The lid can alsobe washed periodically. The bottom framenever has to be washed routinely as no com-ponents on it are exposed to any formulation.

4. The cups can all be changed in one action bysimply removing the cup tray. This featuremakes changing collection cups fast and keepsthe cups in the correct location until analysis.

5. The sides of the cups are tapered so that theycan be easily stacked during storage.

6. The lid support at the rear of the NGI alsoserves as a stand for storage by rotating thefront of the NGI upward. In this position theNGI has a smaller footprint than when the im-pactor is positioned for normal use with thecollection cups oriented horizontally.

PERFORMANCE

The design objectives related to stage perfor-mance were satisfied. The stages have beenshown to have sharp cut off characteristics, withcut sizes at the desired places, and very low in-ter-stage losses so that the impactor body andnozzles do not have to be cleaned between everyrun. The collection efficiency curve for each stagehas been determined at three flow rates and theresults are reported in the companion paper.18 Acollaborative inter-company study has been con-ducted by members of the European Pharma-

ceutical Aerosol Group (EPAG) to establish in-terlaboratory variability,26 and results of thisstudy are expected to be submitted for publica-tion in 2003.

CONCLUSION

The NGI is the result of a collaborative process,with the sponsoring pharmaceutical companiesinvolved in all major decisions throughout de-velopment. Input was gathered from as broad aconstituency as could be solicited. Design deci-sions were made in consideration of mechanical,physical, regulatory, and end-use viewpoints.The result is a seven-stage impactor with an op-tional pre-separator at the inlet for over-size par-ticles, and a micro-orifice collector at the exit in-stead of a final filter to collect the finest particles.A unique external particle collection cup designwas used so that all of the collection cups are inone plane and can be removed from the impactoras a single unit. This feature eases the ability toautomate the NGI, which was a primary designrequirement.

The NGI is also unique in that it is not designedwith stages having specific cut sizes at a specificflow rate. Instead, the NGI operates at any flowrate from 30 to 100 L/min with the cut sizes span-ning a particle size range nominally from 0.25-mm to 11-mm aerodynamic diameter. The NGI isdesigned so that the spacing of the stages variesbetween four or five cut points per decade of par-ticle size, depending on the flow rate chosen forthe measurement.

Efforts are underway to incorporate the NGIinto the European Pharmacopoeia and the U.S.Pharmacopeia.

APPENDIX

Aerodynamic diameter is a parameter definedas the diameter of a hypothetical spherical parti-cle of unit density (i.e., rae 5 1.00 g/cm3) that set-tles in air at the same falling velocity as the phys-ical particle. In the most general case fornon-spherical particles:

dae 5 dp !§ (A-1)

If the particles are spherical, x, the dynamic shapefactor is unity and equation A-2 applies:

rpCp}xraeCae


dae 5 dp!§ (A-2)

The aerodynamic diameter, rather than physicaldiameter, is a useful concept for analysis of sys-tems of aerosols because many fundamentalequations describing particle motion can be ex-pressed with aerodynamic diameter as a para-meter. Aerodynamic diameter is a parameter thatcan be used, along with geometric descriptionand air flow profile, to correlate with(A1) and pre-dict(A2) regional deposition in the lung.

Appendix references

A1. Rudolf, G., R. Kobrich, and W. Stahlhofen.1990. Modeling and algebraic formulation ofregional aerosol deposition in man. J. AerosolSci. 21:s403–s406.

A2:. Martonen, T.B. 1993. Mathematical modelfor the selective deposition of inhaled phar-maceuticals. J. Pharm. Sci. 82:1191–1199.

ABBREVIATIONS

C Cunningham slip correction factorC50 Cunningham slip correction factor for a par-

ticle of size d50Cae Cunningham slip correction factor for a par-

ticle of size daeCp Cunningham slip correction factor for a par-

ticle of size dpD50 calculated stage cut size from cut size-flow

rate equationdp diameter of a spherical particledae aerodynamic diameter of a particled50 aerodynamic diameter of particle collected

with 50% efficiencyDc diameter of the cluster of nozzles on a stageKn Knudsen numbern number of nozzles in a stageQ total volumetric flow rate through an im-

pactor stageRe Reynolds numberS jet-to-plate distanceSt Stokes numberSt50 Stokes number at 50% collection efficiencyVo average velocity of air in a nozzleW nozzle diameterXc cross-flow parameterl mean-free path of airm air viscosity

rpCp}raeCae

r air densityrp particle densityrae unit density (i.e., 1 g/cm3)x dynamic shape factor (x 5 1.0 for spherical

particles)

ACKNOWLEDGMENTS

Considerable technical input was obtainedfrom representatives from all the sponsoringcompanies; their input is gratefully acknowl-edged. The representatives and their affiliatedcompanies are Dave Warren (3M HealthcareLtd.), Beatrix Fyrnys (Asta Medica—now SofotecGmbH), Bo Olsson and Lars Asking (AstraZenecaR&D Lund), Steve Nichols (Aventis Pharma), Di-eter Hochrainer (Boehringer Ingelheim PharmaKG), Professor David Ganderton (CoordinatedDrug Development Ltd), Bernie Greenspan (DuraPharmaceuticals), Keith Truman and Mike VanOort (GlaxoSmithKline), Mike Holroyd (NortonHealthcare—now IVAX Pharmaceuticals), DilrajSingh (Novartis Pharma AG), Terhi Mattila andJari Kovalainen (Orion Pharma), Paul Miller(Pfizer), Tön Forch and Hans Keegstra (Pharma-chemie), Bruce Wyka (Schering-Plough ResearchInstitute), Jolyon Mitchell (Trudell Medical Inter-national), and Jeremy Clarke (Vectura Ltd.).

REFERENCES

1. Thiel, C.G. 1996. From Suzie’s question to CFC free:an inventor’s perspective on forty years of MDI de-velopment and regulation. In R.N. Dalby, P.R. Byron,and S.J. Farr, eds. Respiratory Drug Delivery V. In-terpharm Press, Buffalo Grove, IL, 115–123.

2. Dimmick, R.L., M.T. Hatch, and J. Ng. 1958. A parti-cle-sizing method for aerosols and fine powders.A.M.A. Arch. Indust. Health. 18:23–29.

3. Thiel, C.G. 2002. Private communication, 3M Phar-maceuticals, St. Paul, MN.

4. Polli, G.P., W.M. Grim, F.A. Bacher, et al. 1969. Influ-ence of formulation on aerosol particle size. J. Pharm.Sci. 58:484–486.

5. Mitchell, R. I., and J.M. Pilcher. 1959. Improved cas-cade impactor for measuring aerosol particle sizes inair pollutants, commercial aerosols and cigarettesmoke. Indust. Eng. Chem. 51:1039–1042.

6. Andersen, A. A. 1958. New sampler for the collection,sizing, and enumeration of viable airborne particles.J. Bacteriol. 76:471–484.

7. May, K. R. 1966. Multistage liquid impinger. Am. Soc.Microbiol. 30:559–560.

MARPLE ET AL.298

8. Bell, J. H., K. Brown, and J. Glasby. 1973. Variation indelivery of isoprenaline from various pressurized in-halers. J. Pharm. Pharmacol. 25(suppl.):32–36.

9. Asking, L, and B. Olsson. 1997. Calibration at differ-ent flow rates of a multistage liquid impinger. AerosolSci. Technol. 27:39–49.

10. Hallworth, G. W., and D. G. Westmoreland. 1987. Thetwin impinger; a simple device for assessing the de-livery of drugs form metered dose pressurizedaerosol inhalers. J. Pharm. Pharmacol. 39:966–972.

11. Hallworth, G.W., D. Clough, T. Newnham, et al. 1978.A simple impinger device for rapid quality control ofthe particle size of inhalation aerosols delivered bypressurized aerosols and powder inhalers. J. Pharm.Pharmacol. 30(Suppl):39.

12. British Pharmacopoeia. (1988). C. Pressurized inhala-tions: deposition of the emitted dose. In British Phar-macopoeia, Vol. II. Department of Health, London,A204–A207.

13. Aiache, J. M., H. Bull, D. Ganderton, et al. 1993. Col-laborative study on the measurement of fine particledose using inertial impactors. Pharmeuropa 5:386–389.

14. Miller, N.C., V. A. Marple, R. K. Schultz, et al. 1992.Assessment of the twin impinger for measurement ofMDI sprays. Pharm. Res. 9:1123–1127.

15. U.S. Pharmacopeia. 2000. ,601. Aerosols, metereddose inhalers, and dry powder inhalers. In U.S. Phar-macopeia 25/National Formulary 20, U.S. Pharma-copeial Convention, Inc, Rockville, MD, 1964–1980.

16. European Pharmacopoeia. 2002. 2.9.18. Preparationsfor inhalation: aerodynamic assessment of fine parti-cles. In European Pharmacopoeia, 3rd ed., Suppl.2001, Council of Europe, Strasbourg, France, 113–124.

17. Marple, V.A., B.A. Olson, and N.C. Miller. 1995. Alow-loss cascade impactor with stage collection cups:calibration and pharmaceutical inhaler applications.Aerosol Sci. Tech. 22:124–134.

18. Marple, V.A., B.A. Olson, K. Santhanakrishnan, et al.2003. Next generation pharmaceutical impactor (anew impactor for pharmaceutical inhaler testing).Part II: archival calibration. J. Aerosol Med. 16:301–324.

19. Marple, V.A. 1970. A fundamental study of inertial

impactors [Ph.D. dissertation]. University of Min-nesota, Minneapolis.

20. Marple, V. A., and K. Willeke. 1976. Impactor design.Atmos. Environ. 10:891.

21. Rader, D. J., and V. A. Marple. 1985. Effect of Ultra-Stokesian drag and particle interception on impactorcharacteristics. Aerosol Sci. Technol. 4:141–156.

22. Fang, C. P., V. A. Marple, and K. L. Rubow. 1991. In-fluence of cross-flow on particle collection character-istics of multi-nozzle impactors. J. Aerosol Sci.22:403–415.

23. Mitchell, J.P. 2000. The next generation impactor(NGI): results from the evaluation of prototype in-struments with pressurized metered dose inhaler(pMDI)–based formulations. In Drug Delivery to theLungs XI. Aerosol Society (UK), London, 223–226.

24. Shrubb, I. 2000. The next generation impactor (NGI):results from the evaluation of prototype instrumentswith dry powder inhaler (DPI)–based formulations.In Drug Delivery to the Lungs XI. Aerosol Society(UK), London, 227–230.

25. Marple, V. A., 2003. Private communication, Univer-sity of Minnesota.

26. Nichols, S. 2003. Private communication, AventisPharma UK.

Received on January 31, 2003in final form, May 6, 2003

Reviewed by:Peter R. Byron, Ph.D.John E. Agnew, Ph.D.

Address reprint requests to:Jolyon Mitchell, Ph.D.

Trudell Medical International725 Third Street

London, Ontario, Canada N5V 5G4

E-mail: [email protected]


Documents

Next Generation Pharmaceutical Impactor (A New Impactor ... · such as materials, clamping mechanisms and rel-ative location of the stages. These characteristics have negligible effect