An industry perspective on the monitoring of subvisible particles as a quality attribute for protein therapeutics

PERSPECTIVES

An Industry Perspective on the Monitoring of SubvisibleParticles as a Quality Attribute for Protein Therapeutics

SATISH K. SINGH,1 NATALIYA AFONINA,2 MICHEL AWWAD,3 KAROLINE BECHTOLD-PETERS,4 JEFFREY T. BLUE,5

DANNY CHOU,6 MARY CROMWELL,7 HANS-JUERGEN KRAUSE,8 HANNS-CHRISTIAN MAHLER,9 BRIAN K. MEYER,5

LINDA NARHI,10 DOUG P. NESTA,11 THOMAS SPITZNAGEL12

1Biotherapeutics Pharmaceutical Sciences, Pfizer Inc., 700 Chesterfield Parkway West, Chesterfield, Missouri 63017

2Biologics Process and Product Development—Analytical Development and Testing, Bristol-Myers Squibb, Hopewell,New Jersey 08534

3Pharmacokinetics, Dynamics and Metabolism, Pfizer Inc, Andover, Massachusetts 01810

4Process Science Department/Biopharmaceuticals, Boehringer Ingelheim Pharma GmbH & Co. KG, D-88397 Biberach, Germany

5Bioprocess Analytical and Formulation Sciences, Merck & Co., West Point, Pennsylvania 19486

6Technology Development, Genzyme, Framingham, Massachusetts 01701

7Analytical R&D, Genentech, South San Francisco, California 94080

8Manufacturing Science and Technology, Abbott GmbH & Co. KG, D-67061 Ludwigshafen, Germany

9Pharmaceutical and Analytical R&D, F. Hoffmann-LaRoche, CH-4070 Basel, Switzerland

10Formulation and Analytical Resources, Amgen, Thousand Oaks, California 93012

11Biopharmaceutical Technology, GlaxoSmithKline, King of Prussia, Pennsylvania 19406

12Biopharmaceutical Development, Human Genome Sciences, Rockville, Maryland 20850

Received 14 August 2009; revised 9 December 2009; accepted 11 January 2010

Published online 22 March 2010 in Wiley InterScience (www.interscience.wiley.com). DOI 10.1002/jps.22097

Corresponde9979; Fax: þ1-8

Journal of Pharm

� 2010 Wiley-Liss

3302 JOURN

ABSTRACT: Concern around the lack of monitoring of proteinaceous subvisible particulates inthe 0.1–10mm range has been heightened (Carpenter et al., 2009, J Pharm Sci 98: 1202–1205),primarily due to uncertainty around the potential immunogenicity risk from these particles.This article, representing the opinions of a number of industry scientists, aims to further thediscussion by developing a common understanding around the technical capabilities, limita-tions, as well as utility of monitoring this size range; reiterating that the link betweenaggregation and clinical immunogenicity has not been unequivocally established; and empha-sizing that such particles are present in marketed products which remain safe and efficaciousdespite the lack of monitoring. Measurement of subvisible particulates in the <10mm size rangehas value as an aid in product development and characterization. Limitations in measurementtechnologies, variability from container/closure, concentration, viscosity, history, and inherentbatch heterogeneity, make these measurements unsuitable as specification for release andstability or for comparability, at the present time. Such particles constitute microgram levels ofprotein with currently monitored sizes �10mm representing the largest fraction. These levelsare well below what is detected or reported for other product quality attributes. Subvisibleparticles remain a product quality attribute that is also qualified in clinical trials. � 2010 Wiley-

Liss, Inc. and the American Pharmacists Association J Pharm Sci 99:3302–3321, 2010

Keywords: biotechnology; particle size; prot
ein aggregation; protein formulation; immunology
nce to: Satish K. Singh (Telephone: þ1-636-247-60-686-7768; E-mail: [email protected])

aceutical Sciences, Vol. 99, 3302–3321 (2010)

, Inc. and the American Pharmacists Association

AL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 8, AUGUST

INTRODUCTION

Therapeutic proteins have made an immensecontribution to treatment of human diseases and

2010

MONITORING SUBVISIBLE PARTICLES IN THERAPEUTIC PROTEIN PRODUCTS 3303

represent an increasingly important part of thearmamentarium available for this purpose. Thelife-saving benefits of products derived from re-combinant protein technology, starting from the veryfirst product insulin in 1982, have never been indispute. However, concerns have always been presentabout the potential adverse consequences of aggrega-tion of the protein in the product being dosed.Aggregation in biotherapeutic products is oftendiscussed in conjunction with risk for immunogeni-city,1 although a clear connection between the two forprotein therapeutics has not been demonstrated.Nevertheless, this concern has led the industry andregulatory authorities to use aggregation as a criticalquality attribute for biologics.

Forming aggregates is to some extent an inherentproperty of a protein. While significant advances havebeen made in the understanding of the pathways ofchemical degradation of proteins, the same does notcompletely hold for aggregation pathways or mechan-isms. In light of this, judicious process and formula-tion development research is performed to control thedevelopment of aggregates during the production andstorage of the product. The success of this effort isevident from the number of biotherapeutic productsthat have been commercialized and found to havefavorable safety and efficacy profiles. It is quitereasonable to assume that these commercial products

Table 1. Subvisible Particles (�2, �5, �10, �25mm) in someUsing Small-Volume Methods But Without Dilution

Product Info.

Subvisible Particles ((Mean�SD Whe

�2mm �5mm

Product 1 Lyo/vial 2091�1453 352�225Product 2 Liquid/PFS 4477�486 1220� 91Product 3 Lyo/vial 664�191 59�12Product 4 Lyo/vial 13081�3035 1122�581Product 5 Lyo/vial 4692�725 385�73

Product 6Lot 1 Liquid/vial 5240�117 1312� 33Lot 2 Liquid/vial 8826�205 2016� 58

Product 7 (CC¼ container/closure)CC2, Frmln 1, Lot 1 Liquid/PFS 2745� 46 823�38CC2, Frmln 1, Lot 2 Liquid/PFS 6740� 57 1585� 53CC2, Frmln 1, Lot 3 Liquid/PFS 1191� 25 403�19CC2, Frmln 1, Lot 4 Liquid/PFS 3679� 86 1108� 20CC3, Frmln 2, Lot 1 Liquid/vial 12�3 <10CC3, Frmln 2, Lot 2 Liquid/vial 88�12 <10

Product 8 Liquid/vial 1054 150Product 9b Liquid/vial 17145 4842

aProtein mass in the form of subvisible particles has been estimated assuming sworst-case size and number assumption. All particles in range �2mm were assume25mm; and �25mm were taken to be 100mm in size. Counts for each of the abocumulative but this was not corrected for in the calculations. Counts shown as ‘‘<1constitutes the largest fraction of the total mass estimated.

bProduct may contain proteinaceous visible particles.

DOI 10.1002/jps

contain a range of aggregate levels and associatedsubvisible particulates, (see Tab. 1 for some exam-ples), the latter being the focus of this article. Note,however, that the data in Table 1 do not distinguishbetween proteinaceous and extrinsic subvisibleparticles due to limitations of the technology, asdiscussed later.

The term aggregates covers a large variety ofheterogeneous species from reversible to irreversible,native and nonnative, and from dimers to multimersthat range in size from a few nanometers to visibleparticles in the hundreds of microns.2–4 The technicalability to measure and quantify aggregates through-out this entire size range with one single method doesnot exist.5,6 Carpenter et al.7 have recently publisheda commentary highlighting concerns with proteinaggregates in the subvisible size range between 0.1and 10mm. This article has been prepared in responseto the commentary to provide an industry perspectivearound this topic.

Statement of Issue to be Discussed

The primary basis of the commentary by Carpenter etal.7 lies in the assertion that the presence of protein ina product aggregates is a risk factor for immuno-genicity. Under this umbrella concern, the commen-tary raises the following issues:

Marketed Biologics, Measured by Light Obscuration/HIAC

in Counts per mL)re Available)

Estimated Protein Mass Representedby Subvisible Particles (g/mL)a�10mm �25mm

29� 13 3� 0 2.95e�6118� 26 2� 1 5.15e�6

7� 7 0� 0 3.44e�799�128 3� 5 1.56e�516� 4 0� 1 1.45e�6

227� 27 <10 8.46e�6350� 17 <10 9.82e�6

229� 22 21�6 2.25e�5343�6 24�7 2.66e�5163�6 24�8 2.57e�5284� 12 22�3 1.88e�5<10 <10 5.08e�6<10 <10 5.09e�613 1 3.05e�7590 30 9.98e�6

pherical particles of density 1 g/cm3. Mass estimation has been made using ad to be 5mm; �5mm were assumed to be 10mm; �10mm were assumed to beve sizes was taken as being equal to (meanþ3�SD). Counts as given are0’’ were counted as 10. In all cases, it is the largest size range (�25mm) that

JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 8, AUGUST 2010

3304 SINGH ET AL.

(1) P

JOURNA

roteinaceous subvisible particles (i.e., subvi-sible particles of protein-origin) have the poten-tial to negatively impact clinical performance,since they could represent aggregates of thou-sands to millions of protein molecules.

(2) C
urrently, there is a gap in routinely measur-ing and controlling subvisible particles smallerthan 10mm in biotherapeutic products, withno pharmacopeial requirements for this sizerange.
(3) R
ecent studies have indicated the existenceof proteinaceous particles in the subvisiblesize range below 10mm (0.1–10mm) in biother-apeutics.
(4) U
ncertainty created by the lack of monitoringof (proteinaceous) particulates in this sizerange and the general risk for immunogenicityassociated with aggregates implies that (a)further development of technologies that canmeasure particles in this size range is critical,and (b) more research is needed to investigatethe relationship between aggregation andimmunogenicity.
Objective

As with any new technology, collection of data acrossa broad set of samples is critical prior to interpretingthe significance of the results. The industry has ademonstrated history of developing, adopting, andapplying new technologies to characterizing products,and when appropriate, adding additional controlseither in-process or at release. While we acknowledgethat the published commentary will serve to accel-erate this discussion, we feel it is important tothoroughly understand the methodologies prior toimplementing any required controls. In addition,understanding the causes and sources of subvisibleparticles is critical to interpreting the data.

In this response, we want to:

(1) E
xamine the current state of knowledge in theconnection between aggregation and immuno-genicity, including the state of current thera-peutics. We want to emphasize that availablebiotherapeutics have a strong record of safety,and there are no immunogenicity issues orother clinical findings that can be directlyrelated to aggregates or to proteinaceous sub-visible particles of any size, despite the gap inmonitoring.
(2) R
eview currently available analytical tools andtheir capabilities for the measurements andidentification of subvisible particles in the sizerange above 0.1mm, and share current experi-ence from the industry in the monitoring ofsubvisible particles of this size.
L OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 8, AUGUST 2010

(3) D
evelop a common understanding between allinterested parties, including academic labora-tories, the industry and regulatory agencies,around the capabilities for monitoring particlesin this size range using currently availabletechnologies, and set realistic expectationsaround the value such measurements wouldprovide.
There are no consistent or formalized definitions ofsubvisible versus visible particles. Visible particlesare usually defined as having a lower size limitaround 100mm, based on an analysis of the sizethreshold at which a majority of inspectors consis-tently detected the particles under appropriatelighting and test conditions.8 Light obscurationmethods designed for subvisible size ranges candetect particles up to approximately 150mm. Thus,although there is some overlap between the visibleand subvisible size ranges, we will broadly referto the subvisibles as between 0.1 and 150mm, withthe understanding that particles represent a con-tinuum in size range, and the classification assubvisible versus visible is fluid and not veryinformative. Subvisible particles, for the purpose ofthis communication, therefore comprise the submi-cron and micron size ranges, spanning three orders ofmagnitude.

AGGREGATION, SUBVISIBLE PARTICULATES,AND IMMUNOGENICITY

Multiple publications have implied that proteinaggregates may be one of the factors, if not the majorfactor, for tolerance reversal and induction of immuneresponse against protein biotherapeutics.1,9,10 Thisconclusion is based on theoretical considerations,circumstantial evidence obtained from investigationsconducted on clinical products, as well as datafrom animal studies using artificially induced andstabilized aggregates.11–14 Despite the fact that allbiotherapeutics contain some level of aggregates (andsubvisible particles as discussed later), there has beenno direct evidence from clinical studies demonstrat-ing aggregates as the true dominant risk factorcontributing to the immunogenicity. The principalreason for this lack of a clear demonstrated connec-tion is that clinical observations of immunogenicityare a consequence of combination of numerous(patient, therapy, product, dosing) factors.15 Thismakes the task of deconvoluting the specific impact ofaggregation very difficult. Furthermore, in manycases the incidence of immunogenicity is so low that todetermine the true cause would be impossible (see,e.g., the survey by Hwang and Foote16). The broaderconcern with aggregates and the more specific

DOI 10.1002/jps


concern with proteinaceous subvisible particles, istherefore primarily due to the uncertainty around therisk they entail.

The primary risk with aggregates evolves from theconcept of pathogen associated molecular patterningwherein repetitive arrays of proteins are found to bepotent modulators of immune response. This raisesthe possibility that aggregates of human biothera-peutics, including those in the micron range (sub-visible proteinaceous particles) could beimmunogenic in patients through the same mechan-ism.1 To the best of our knowledge, there are nopublished reports examining the structure of natu-rally occurring protein aggregates which either proveor discount the presence of such repetitive motifs intherapeutic proteins. Moreover, aggregates producedby different methodologies can have very differentbiochemical and biophysical characteristics andmorphologies,17 and could produce completely differ-ent immunological effects when injected into thesame animal model.13,14,18

Due to the difficulty in clearly delineating theimpact of aggregates in clinical studies, animalmodels, including transgenic mice, are being devel-oped and employed.12–14,18 The use of animal studiesto assess immunogenicity of aggregates and/orparticulates comes with caveats. As already notedabove, aggregates produced by different methods canlead to different immune response in the same animalmodel. Immunogenicity observed in animal studies ofhuman protein drugs may not be predictive of whatcould happen in humans.19–21 Jahn and Schneider20

report that a range of scenarios have been observedduring dossier evaluation, ranging from an almostabsent immune response in animals but highimmunogenicity in humans, to (the more usuallyobserved) high immunogenicity in animals but lowimmunogenicity in humans. The utility of animalmodels would therefore lie in assessing the relativeimmunogenicity risk of aggregates or particles,although the translation of an observation of‘‘increased’’ or ‘‘decreased’’ immunogenicity in theanimal model, to the human clinical response wouldstill be difficult. This relative ranking of risk wouldlikely be dependent on the biotherapeutic class and onthe construct of the animal model.

In summary, the immunogenic potential of aggre-gates in general, and of proteinaceous subvisibleparticles in particular, is complicated to assess.Animal models have limited utility and clinicalstudies specifically designed to test such product-related impurities would not be ethically justifiable.Ultimately, the assessment of subvisible particlesmay have to be similar to the approach currentlyapplied to other product-related impurities: anacceptable product safety and immunogenicity profileis demonstrated in clinical studies with drug product

DOI 10.1002/jps

lots containing a certain historical range of product-related impurities (in this case subvisible particles).However, the ability to do this from a QC orspecifications perspective has some practical andtechnical limitations as discussed in the subsequentsections of this article.

STATE OF CURRENT PROTEIN PRODUCTS

The specification for protein oligomers (generallymeasured by SEC) in the therapeutic proteins on themarket are set prior to clinical studies and qualified inthese studies. Subvisible particles larger than 10mmare routinely measured for these products, while therange below 10mm is not. Despite the lack of acomplete picture about subvisible particles/aggre-gates, it is likely that such particulates/aggregates(including those below 10mm) existed during clinicalstudies and are present in the marketed product.There are no reports to suggest that these currentlymarketed products are in anyway unsafe as aconsequence of these subvisible particles.

An example showing subvisible particle counts innine marketed, efficacious and safe biotherapeuticproducts is given in Table 1. It should be noted that forparticles in the �2 and �5mm ranges, the standarddeviation in a number of cases are quite high, andthus precision is low. Variability between differentlots of the same product and between differentpresentations of the same protein is significant, andis likely a consequence of product, including thedevice and presentation of the drug, as well as methodof measurement (see discussion on Product-RelatedFactors Leading to Variability in Subvisible Particu-lates Data Section; all measurements were carriedout with small-volume methods, see discussion onSample Requirements Section). Although limited, thedata show that a wide range of particle counts existsin marketed products in the size range below 10mm,even within the same product. However, a largersurvey would be useful, as discussed later in the RiskAssessment Section.

Using the particle count data in Table 1, anestimate of protein mass contained in the subvisibleparticles has also been made under some worst-caseassumptions about size and number, and is likely anoverestimation by at least an order of magnitude. Theresults demonstrate that under this worst-caseestimation, and assuming that all these particleswere proteinaceous, they comprise less than 30mg ofmass, and in most cases less than 5mg. Theproteinaceous particles thus constitute a negligiblefraction of the total dose of most products. Further-more, in the above calculations, >90% of the particlemass actually arises from the �25mm particles(which were all assumed to be of 100mm size for


3306 SINGH ET AL.

the calculation), with the next highest contributionarising from the �10mm particles (assumed to be25mm). Thus, the size ranges that are currentlyalready monitored represent the greatest mass ofproteinaceous subvisible particles. Although it is notclear if it is the mass of antigens or the number ofantigens that is critical to an immune response, Whiteet al.22 suggested that micron-sized particles ofovalbumin were more immunogenic in mice thansubmicron-size particles at the same total mass dose,as measured by an in vivo CTL activity assay. Theseaggregated particles were also found to generate abetter anti-ovalbumin IgG response in the micecompared to solubilized aggregates, suggesting thatsize is more important than number. Whether thissize versus number translates to a human clinicalsituation with a protein that is not inherentlyimmunogenic (unlike ovalbumin to mice) is howevernot known.

Package inserts for a number of marketed bio-therapeutic products also mention the possibility ofthe presence of visible proteinaceous particles (see,e.g., asparaginase, laronidase, imiglucerase, algluco-sidase, cetuximab, infliximab, etanercept, orthocloneOKT3, ofatumumab, panitumumab). These solutionsare also likely to contain subvisible particles in awide size range. However, such products continue tobe safe and efficacious.

SUBVISIBLE PARTICLES AND THEMEASUREMENT GAP

The current standard limits or specifications forsubvisible particles and information about methodsin the compendia are summarized in Table 2. Thecurrent standards arose to mitigate the risk asso-ciated with the presence of extraneous particles inintravenous injection solutions. The particle sizesmonitored (�10 and �25mm) were set based on therisk for blood vessel occlusion by intravenous admin-istration of small particles. Modern aseptic produc-tion technologies have reduced the occurrence of highextraneous particle counts such that there are movesto tighten the standards.23

The general consensus is that IM or SC routes ofadministration are more immunogenic than IV. Thecurrent USP 32 <1> (as well as the forthcoming USP33 <1>) exempts SC/IM products explicitly from theexisting subvisible particulate requirement of USP<788>; the Ph.Eur.6.0 and JP 14th Ed. do not(see Tab. 2). From a protein therapeutic perspective,this was an unfortunate gap as pointed out inthe commentary.7 However, these exemptions inUSP<1> are proposed to be removed per USPPF35.3. Interestingly, USP PF35.3 also states that‘‘parenteral products for which the labeling specifies


use of a final filter prior to administration areexempted from the requirements of USP <788>,provided that scientific data are available to justifythis exemption.’’ This harmonizes the USP withPh.Eur. The Ph.Eur. also states that ‘‘in the case ofproducts for SC and IM injection, higher limits may beappropriate.’’

Current Measurement Techniques: Performanceand Limitations

Aggregates and particulates can be looked upon as acontinuum in size, ranging from dimers to visiblemultimers. In this context, subvisible proteinaceousparticles refer to aggregates that have grown in sizeto fall in the subvisible size range. The ability todetect and quantitate these aggregates/particles inprotein solutions is a function of their size, amount/number, and the capability of the technique beingemployed. Figure 1 illustrates some analyticaltechniques and the approximate size ranges overwhich they are viable. Although there may bedifferences in the ranges proposed by differentauthors, it is generally accepted that no singletechnique can cover the whole range of interest.2–6

The capture, detection, and quantitation of particu-lates is a stochastic process and the accuracy of theresults are impacted by their concentration/numberand size. The particle measurement landscape as afunction of number and size of aggregates/particles isillustrated in Figure 2. The straight lines in thisfigure represent constant mass of particles (under theassumption of spherical particles of density 1 g/cm3).The detection and quantitation ability of somemethods is mapped on this chart. Sizes above �1 or2mm can be detected and quantified by countingmethods while those below �0.1mm can be detectedand quantified by concentration-based methods.Intermediate sizes can be detected by light scatteringmethods but obtaining an accurate quantification isnot feasible, as discussed later. The ability to bothdetect and accurately quantify the number ofparticles in solution in the 0.1–1mm size rangerepresents the true measurement gap in currentlyutilized techniques, and is discussed in more detailbelow.

The various techniques summarized in Figure 1also report results in different ways, depending on theprinciples of measurement. The counting techniques(e.g., light obscuration) quantify numbers in a sizerange, while SEC or AUC quantify by mass fractionsof the various species. The results reflect differencesin the sensitivity of the techniques/detection techno-logy used. Conversion between various modes ofreporting data requires making several assumptionsabout the properties of the particles. This makesstudies exploring the conversion from one species

DOI 10.1002/jps

Ta

ble

2.

Su

mm

ary

ofC

omp

end

ial

Req

uir

emen

tsfo

rS

ubvis

ible

an

dV

isib

leP

art

icu

late

sin

Pare

nte

rals

Att

ribu

tes

US

P33

Ph

.Eu

r.6.0

JP

15th

Ed

Un

itP

rod

uct

Vol

um

eD

efin

itio

n�

100

mL¼

Sm

all

Vol

um

eP

are

nte

ral

(SV

P)

>100

mL¼

Larg

eV

olu

me

Pare

nte

ral

(LV

P)

Un

itP

rod

uct

Vol

um

eD

efin

itio

n<

100

mL¼

Sm

all

Vol

um

eP

are

nte

ral(

SV

P)

�100

mL¼

Larg

eV

olu

me

Pare

nte

ral(L

VP

)

Sp

ecifi

cati

ons:

ligh

tob

scu

rati

on(L

O)

(pre

ferr

edm

eth

od)

SV

P�

10m

m:�

6000

cou

nts

/con

tain

er�

10m

m:�

6000

cou

nts

/con

tain

er�

10m

m:�

6000

cou

nts

/con

tain

er�

25m

m:�

600

cou

nts

/con

tain

er�

25m

m:�

600

cou

nts

/con

tain

er�

25m

m:�

600

cou

nts

/con

tain

erL

VP

�10m

m:�

25

cou

nts

/mL

�10m

m:�

25

cou

nts

/mL

�10m

m:�

25

cou

nts

/mL

�25m

m:�

3co

un

ts/m

L�

25m

m:�

3co

un

ts/m

L�

25m

m:�

3co

un

ts/m

LS

pec

ifica

tion

s:m

icro

scop

y(M

)(s

econ

dst

age

toL

Oor

inca

seL

Oca

nn

otbe

use

d)

SV

P�

10m

m:�

3000

cou

nts

/con

tain

er�

10m

m:�

3000

cou

nts

/con

tain

er�

10m

m:�

3000

cou

nts

/con

tain

er�

25m

m:�

300

cou

nts

/con

tain

er�

25m

m:�

300

cou

nts

/con

tain

er�

25m

m:�

300

cou

nts

/con

tain

erL

VP

�10m

m:�

12

cou

nts

/mL

�10m

m:�

12

cou

nts

/mL

�10m

m:�

12

cou

nts

/mL

�25m

m:�

2co

un

ts/m

L�

25m

m:�

2co

un

ts/m

L�

25m

m:�

2co

un

ts/m

L

Intr

aven

eou

sin

ject

ion

sL

imit

sap

ply

Lim

its

ap

ply

Lim

its

ap

ply

Inje

ctio

ns

sole

lyfo

rin

tram

usc

ula

r(I

M)

orsu

bcu

tan

eou

s(S

C)

dos

ing

Lim

its

do

not

ap

ply

per

US

P33<

1>

;U

SP

PF

35.3

pro

pos

al:

Lim

its

toap

ply

for

IMan

dS

Cp

rod

uct

sals

o

Lim

its

ap

ply

.H

igh

erli

mit

sm

ay

be

ap

pro

pri

ate

Not

men

tion

edse

para

tely

Non

IM/S

Cd

isp

erse

dsy

stem

sM

(no

spec

ial

pro

ced

ure

des

crib

ed)

M(n

osp

ecia

lp

roce

du

red

escr

ibed

)M

(no

spec

ial

pro

ced

ure

des

crib

ed)

Su

spen

sion

part

icle

�150m

mE

mu

lsio

nd

rop�

7m

mN

onIM

/SC

solu

tion

sor

pow

der

sfo

rin

ject

ion

Wit

hfi

nal

filt

erbef

ore

inje

ctio

nE

xcl

ud

edfr

omre

qu

irem

ents

per

US

P33<

1>

;U

SP

PF

35.3

pro

pos

al:

pare

nte

ral

pro

du

cts

for

wh

ich

the

label

ing

spec

ifies

the

use

ofa

fin

alfi

lter

pri

orto

ad

min

istr

ati

onare

exem

pt

from

the

requ

irem

ents

pro

vid

edth

at

scie

nti

fic

data

are

avail

able

toju

stif

yth

eex

emp

tion

Exem

pt

from

requ

irem

ents

,p

rovid

ing

ith

as

bee

nd

emon

stra

ted

that

the

filt

erd

eliv

ers

aso

luti

onth

at

com

pli

es

Not

men

tion

ed

Hig

hvis

cosi

tyD

ilu

tion

foll

owed

by

LO

Dil

uti

onfo

llow

edby

LO

Dil

uti

onfo

llow

edby

LO

Oth

erso

luti

ons

LO

alo

ne

orfo

llow

edby

ML

Oalo

ne

orfo

llow

edby

ML

Oalo

ne

orfo

llow

edby

M

Tes

tp

roto

col

Sta

tist

icall

yso

un

dsa

mp

lin

gp

lan

Req

uir

edfo

r<

25

mL

/un

itR

equ

ired

for<

25

mL

/un

itR

equ

ired

for<

25

mL

/un

itF

or�

25

mL

/un

it,

10

un

its

acc

epta

ble

For

�25

mL

/un

it,

10

un

its

acc

epta

ble

For

�25

mL

/un

it,

10

un

its

acc

epta

ble

<25

mL

/un

itP

ool�

10

un

its

toob

tain

>25

mL

,te

st4�

NL

T5

mL

ali

qu

ots,

dis

card

firs

tre

sult

Poo

l�

10

un

its

toob

tain

>25

mL

,te

st4�

NL

T5

mL

ali

qu

ots,

dis

card

firs

tre

sult

Poo

l�

10

un

its

toob

tain

>25

mL

,te

st4�

NL

T5

mL

ali

qu

ots,

dis

card

firs

tre

sult

�25

mL

/un

itN

op

ooli

ng,

test

edin

div

idu

all

y,

test

4�

NL

T5

mL

ali

qu

ots,

dis

card

firs

tre

sult

No

poo

lin

g,

test

edin

div

idu

all

y,

test

4�

NL

T5

mL

ali

qu

ots,

dis

card

firs

tre

sult

No

poo

lin

g,

test

edin

div

idu

all

y,

test

4�

NL

T5

mL

ali

qu

ots,

dis

card

firs

tre

sult

Vis

ible

part

icu

late

s

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 8, AUGUST 2010


Vis

ual

orm

ach

ine

insp

ecti

onE

ach

fin

al

con

tain

erof

all

pare

nte

ral

pre

para

tion

ssh

all

be

insp

ecte

dto

the

exte

nt

pos

sible

for

the

pre

sen

ceof

obse

rvable

fore

ign

an

dp

art

icu

late

matt

er(h

erea

fter

term

ed‘‘v

isib

lep

art

icu

late

s’’)

init

sco

nte

nts

.T

he

insp

ecti

onp

roce

sssh

all

be

des

ign

edan

dqu

ali

fied

toen

sure

that

ever

ylo

tof

all

pare

nte

ral

pre

para

tion

sis

esse

nti

all

yfr

eefr

omvis

ible

part

icu

late

s.Q

uali

fica

tion

ofth

ein

spec

tion

pro

cess

shall

be

per

form

edw

ith

refe

ren

ceto

part

icu

late

sin

the

vis

ible

ran

ge

ofa

typ

eth

at

mig

ht

eman

ate

from

the

man

ufa

ctu

rin

gor

fill

ing

pro

cess

US

P<

1>

Lig

ht

box

obse

rvati

onfo

r5

s.S

olu

tion

sfo

rin

ject

ion

exam

ined

un

der

suit

able

con

dit

ion

sare

clea

ran

dp

ract

icall

yfr

eefr

omp

art

icle

s(E

P2.9

.20)

Mon

oclo

nal

an

tibod

ies:

free

ofvis

ible

part

icu

late

s(E

P6.6

2031E

)

JP

6.0

6:

Insp

ect

wit

hth

eu

naid

edey

esat

ap

osit

ion

ofli

gh

tin

ten

sity

ofap

pro

x.

1000

lux

un

der

an

inca

nd

esce

nt

lam

pA

ccep

tan

cecr

iter

ia:

Inje

ctio

ns

eith

erin

solu

tion

s,or

inso

luti

onco

nst

itu

ted

from

ster

ile

dru

gso

lid

s(M

eth

od1):

Cle

ar

an

dfr

eefr

omre

ad

ily

det

ecta

ble

fore

ign

inso

luble

matt

ers

Inje

ctio

ns

wit

hco

nst

itu

ted

solu

tion

(Met

hod

2)

Cle

ar

an

dfr

eefr

omfo

reig

nin

solu

ble

matt

ers

that

iscl

earl

yd

etec

table

Cla

rity

Tu

rbid

ity

com

pare

dto

refe

ren

cesu

spen

sion

Not

men

tion

edT

urb

idit

yis

the

sam

eas

that

ofw

ate

rR

orof

the

solv

ent

use

d,

orn

otm

ore

than

that

ofre

fere

nce

susp

ensi

on1

Not

men

tion

ed

Vis

ual

exam

inati

onco

mp

are

dto

refe

ren

cesu

spen

sion

Not

men

tion

edT

urb

idit

yis

the

sam

eas

that

ofw

ate

rR

orof

the

solv

ent

use

d,

orn

otm

ore

than

that

ofre

fere

nce

susp

ensi

on1

Not

men

tion

ed


3308 SINGH ET AL.

(size) to another over time, or comparing products/samples, difficult to interpret.

Detection and Measurement of SubmicronParticles (0.1–1mmmm)

As discussed earlier and shown in Figure 1, dynamiclight scattering (DLS) techniques can in principlecover a wide size range (0.1 nm to �3mm; Philo3)but have poor size resolution as well as report ahydrodynamic size instead of a direct size value.Geometry or shape therefore plays a big role in theresults obtained. Small particles can be masked ifsome large particles are present.5 DLS response ishighly dependent on the solution conditions andprotein concentration, and also cannot distinguishbetween particle types. Furthermore, DLS cannotgive absolute quantitative counts of particles. Staticlight scattering techniques also suffer from thelimitation of lack of absolute counting/quantifyingability common to all light scattering systems.

Field flow fractionation (FFF)-based separationand analytical ultracentrifugation (AUC) are cur-rently used as alternative (to SEC) techniques toquantitate oligomers. AUC can be applied in the�0.01–0.1mm range, while FFF offers a broaderdynamic range from 0.01mm to several mm dependenton the separation mode and detection.4,24 Thesensitivity of these techniques is limited by thesensitivity of the detectors used. A limitation ofthese methods is the difficulty of use of instrumenta-tion as well as complexity of data analysis. Thesetechniques, especially AUC, are therefore very usefulfor development and characterization, but not forbatch release or stability. There are no commerciallyavailable techniques or instruments that can be usedfor quantitation of particles in this size range on aroutine basis for QC applications.

Detection and Measurement of Micron-RangeParticles (>1mmmm)

The primary methods for ‘‘subvisible particles’’analysis in the micron range include (i) lightobscuration and (ii) microscopy-based methods.

Theoretically, light obscuration can quantify parti-cles in the size range of approximately 1–150mm orlarger, however, with varying degrees of precisionand reproducibility.25–27 The HIAC instrument is theindustry standard and can be applied to sizes between2 and 150mm. This method has been used to quantifyparticles in biotech formulations in the size range of2mm and larger.17 Particle concentrations between�10 and 18,000 counts/mL can be quantified. Theinstrument is commonly available and is robust andsimple to use. There are certain limitations when it isused for analyzing protein-based products. A majorlimitation is that the instrument cannot differentiate

DOI 10.1002/jps

Figure 1. Schematic representation of particulate size detection ranges for varioustechniques (adapted from Mahler et al.2).


between particles from protein aggregation, particlesfrom extraneous sources, silicone oil and air bubbles.This can potentially lead to false high counts. Falselow counts may result if the transparency of theparticle is high, that is, some very transparentparticles might not obscure enough light and there-fore might not be detected. Artifacts in the resultsmay also be seen at high particle concentrations if thechannel becomes physically blocked or if the limit ofdetection is reached. Dilution of samples can reducethe probability of simultaneous blockage of theincident laser beam, but could also cause dissociationof protein aggregates thereby changing the particledistribution. Most importantly, the sample prepara-tion and handling procedure is critical to obtaininggood quality results by light obscuration (as well asother techniques). For example, reconstitution oflyophilized protein samples can result in the genera-tion of microscopic air bubbles as a consequence of thediluent distribution and cake dissolution process (see,e.g., Tab. 3). Removal of air bubbles can be difficult,especially in high concentration protein formulations,and can have a significant impact on results obtained.Vacuum degassing has been found to be extremelyuseful in this respect (in contrast to sonication asrecommended in USP <788>). In general, analyzinghigh-concentration protein formulations with lightobscuration is a challenge, due to their higher

DOI 10.1002/jps

turbidity and viscosity. Dilutions can be used but ifthe results are not linear, impact on particulates bydilution must be suspected, as mentioned above. Thebottom-line however is that this technique, whenproperly used, can be relied upon to give consistentrelative values between different samples, and hashelped to assure the quality of products on themarket. One aspect that hinders improved reliabilityof this (and other such techniques) in quantitatingproteinaceous particles is the lack of an appropriatecalibration standard mimicking the small differencesin refractive index between protein particles and themedium.

The microscopic method for determination ofsubvisible particles is also featured in USP<788>and Ph.Eur.2.9.19, based on a manual or automaticcount of particles with the use of a binocularmicroscope. The sample is vacuum filtered onto agrid-lined filter and, once dried, is placed under themicroscope for counting. For this to reflect theactual numbers of particles present in the sample,the particles on the filter should not arise fromthe preparation environment, equipment or from thepersonnel, and should not be altered by the samplepreparation procedure. For soft proteinaceous parti-cles, these requirements may be difficult to meet.The wash step can alter the particle distribution. Forprotein particles below 10mm, isolation is very


Figure 2. The particulate detection and measurementlandscape. The straight lines represent boundaries for con-stant mass of protein particles (under the assumption ofspherical particles of density 1 g/cm3). The 0.1mg/mL linehas been taken as a detection limit for UV/RI detectors.Approximate operating regions for SEC, AUC, and FFF areshown. Note that the upper size limit for SEC is a conse-quence of ‘‘filtering’’ or entrapment of larger aggregates/particles on the column. The standard use of AUC is foraggregates from dimer to 20 mers representing an upperrange of around 0.01–0.02mm. Gravitational sweep sedi-mentation can be used to work in the 0.01–0.1mm range.FFF is currently used to detect aggregates or particles up to1mm. The USP and Ph.Eur. limits for �10 and �25mm sizesare also shown (figure adapted courtesy Gabrielson andKendrick, unpublished).

3310 SINGH ET AL.

difficult. It is possible for the amorphous proteinparticles to pass through the filter, or they can alsospread out on the filter surface such that they becomevery difficult to visualize. Calibration is performed;however this can be difficult and tedious, and whenperformed with particle standards, is not represen-tative of the results obtained for protein particles.Similar to light obscuration, counting is better whenthe particle contrast/translucency is high compared tobackground. The specification limits in the pharma-copeias by the microscopic method are much tightercompared to light obscuration (Tab. 2).

The manual optical microscopy technique has beenenhanced with the development of flow-imaging.Four instruments are available that take images ofparticles as they flow through the microscope. Theinstruments [Micro-Flow Imaging (MFI) by Bright-well Technologies Inc. Ottowa, Canada; Flow ParticleImaging Analyzer (FPIA) from Malvern InstrumentsLtd., Malvern, UK; FlowCam from Fluid ImagingTechnologies, Yarmouth, ME; ParticleInsight fromMicromeritics Inc., Norcorss, GA] employ a combina-tion of microscopy in a flow-through mode coupled toan image processing system to enable automatedanalysis and counting of particles in liquid formula-


tions. In this case the manipulations involved inparticle filtration and isolation are avoided, andrecent developments in digital imaging are leveraged.Besides number and size, the flow microscopytechniques also claim to assess parameters such astransparency and circularity/shape, thus potentiallyhelping to differentiate between, for example, siliconedroplets or air bubbles and extraneous particles.28,29

However, assessing some of these parameters requireparticles to be of a certain minimum size so as toimage an adequate numbers of pixel for the analysisto be carried out. Figure 3 shows an example of theimages obtained with two of these instruments (MFIand FPIA). The images of a particle countingstandard used to calibrate the instruments, of asilicone oil droplet, and of a protein particle are quitedifferent. Some protein particles look like agglomer-ates of smaller particles, with uneven irregulardensity and borders. The quality of the optics andthe fraction of total volume analyzed vary in these twoinstruments (based on magnification used), but bothare destructive, low throughput instruments usefulfor research and development but not suitable forroutine quality testing. The picture also illustratesthe wide range of particle morphologies and opticalproperties possible in a sample. This makes it difficultto make a proper estimation of the size of theseparticles, with each instrument using a differentalgorithm to measure the same object. Results aretherefore dependent on the algorithm used to select,classify, and ‘‘size’’ particles, with the same (sample)image potentially yielding different results if repro-cessed.

The Coulter principle is another method that canquantify subvisible particles. The technique providesabsolute particle counts in dilute conducting liquids.Recent instruments utilizing this principle (COUL-TER COUNTER1 Multisizer 4, Beckman Coulter,Fullerton, CA, and the Micromeritics Elzone II53901), have good sensitivity and can measureparticles from 0.4 to about 50mm, though differentapertures and instrument settings are needed for inorder to span the entire range. Depending on theformulation buffer properties and experimental con-ditions used, modification of the ionic strength of thesample (either by dilution or addition of salt) may benecessary, which in turn could either create or breakaggregates/particulates. It is likely that the actualimpact of this manipulation is dependent on theprotein and the nature of the aggregate, thus creatingsome uncertainty about the results reported. Looseaggregates carrying a large fraction of enclosedsolution, and/or particulate geometries differingsignificantly from spherical, can also lead to difficul-ties in sizing by the Coulter principle. Table 3 shows acomparison of the experimental characteristics oflight obscuration, flow imaging and the Coulter

DOI 10.1002/jps

Ta

ble

3.

Com

pari

son

ofC

urr

ent

Part

icle

Cou

nti

ng

Tec

hn

olog

ies

Att

ribu

tes

Lig

ht

Obsc

ura

tion

Flo

wM

icro

scop

y(M

FI)

Flo

wM

icro

scop

y(F

PIA

)F

low

Mic

rosc

opy

(Flo

wC

AM

)C

oult

erP

rin

cip

le(M

ult

isiz

er4)

Mea

sure

men

tp

rin

cip

leL

igh

tob

scu

rati

onM

icro

scop

icim

agin

gM

icro

scop

icim

agin

gM

icro

scop

icim

agin

gC

oult

erp

rin

cip

le(m

ay

requ

ire

mod

ifica

tion

ofio

nic

stre

ngth

)M

easu

rem

ent

outc

ome

Part

icle

con

cen

trati

on(c

oun

ts/m

L)

Part

icle

con

cen

trati

on(c

oun

ts/m

L)

an

dsh

ap

e/m

orp

hol

ogy

a(i

mage

qu

ali

tym

ediu

m)

Part

icle

con

cen

trati

on(c

oun

ts/m

L)

an

dsh

ap

e/m

orp

hol

ogy

a(i

mage

qu

ali

tyh

igh

)

Part

icle

con

cen

trati

on(c

oun

ts/m

L)

an

dsh

ap

e/m

orp

hol

ogy

a(i

mage

qu

ali

tym

ediu

m)

Part

icle

con

cen

trati

on(c

oun

ts/m

L)

Part

icle

size

sli

mit

sfo

rb

Det

ecti

onvs.

iden

tifi

cati

on/

class

ifica

tion

5X

:d

etec

t1m

m,

id4m

m4X

:d

etec

t4m

m,

id15m

m;

10X

:d

etec

t1.5m

m,id

5m

m;

20X

:d

etec

t1m

m,

id:

4m

mM

inim

um

part

icle

size

sd

etec

ted

�2m

m5X

:�

1m

m20X

:�

1.5m

m20X

:�

1m

m�

0.4m

m

Sam

ple

han

dli

ng

Deg

as

ifre

qu

ired

by

all

owin

gto

stan

dor

ap

ply

vacu

um

To

be

defi

ned

by

use

rT

obe

defi

ned

by

use

rT

obe

defi

ned

by

use

rT

obe

defi

ned

by

use

r

Min

imu

msa

mp

levol

um

ere

qu

ired

�25

mL

(US

P/P

h.E

ur.

)�

3–5

mL

(sca

led

dow

n)

5X

:0.5

mL

�1

mL

4X

:>

1m

L;

10X

:>

0.2

5m

L;

20X

:>

0.2

5m

LT

otal

cell

vol

um

e10

mL

(aft

erd

ilu

tion

into

con

du

ctin

gbu

ffer

)%

Tot

al

sam

ple

vol

um

ean

aly

zed

�60%

5X

:�

65%

�1%

4X

:�

65%

;10X

:20–40%

;20X

:5–10%

Set

by

use

r

Part

icle

con

cen

trati

onra

nge

�18,0

00

cou

nts

/mL

5X

:�

850,0

00

cou

nts

/mL

at

2.5m

msi

ze20X

:�

500,0

00

cou

nts

/mL

at

1.5m

msi

ze10X

:1E

6/m

Lat

2.5m

msi

ze�

1E

6–2E

6co

un

ts/m

L

Det

ecti

onli

mit

edby

bP

art

icle

tran

spare

ncy

Part

icle

con

trast

an

dp

ixel

savail

able

Part

icle

con

trast

an

dp

ixel

savail

able

Part

icle

con

trast

an

dp

ixel

savail

able

Sol

uti

onco

nd

uct

ivit

y

aS

hap

e/m

orp

hol

ogy

info

rmati

onm

ay

hel

pto

dif

fere

nti

ate

part

icle

typ

es(e

.g.,

sili

con

eoi

l),

for

part

icle

sabov

e4m

m.

abD

etec

tion

an

dcl

ass

ifica

tion

inim

agin

gsy

stem

sis

dep

end

ent

onp

ixel

savail

able

at

sam

ple

pla

ne.

DOI 10.1002/jps JOURNAL OF PHARMACEUTICAL SCIENCES, VOL. 99, NO. 8, AUGUST 2010


Figure 3. Subvisible particle images obtained with two different flow microscopyinstruments. Difference in clarity of images reflects differences in the optical cap-abilities of the two instruments. (a) Representative particle images by MFI; (b)representative particle images by FPIA.

3312 SINGH ET AL.

principle techniques. Some illustrative data gener-ated by these methods is presented and discussedbelow to allow a comparison of the performance of themethods. It is clear that absolute number of countsvaries between techniques but are consistent within atechnique across test samples.

Table 4 shows particle count results obtained for twolyophilized recombinant drug product presentations(rP1, Lots 1 and 2) and (rP2, Lot 1) by three differenttechniques. For HIAC measurements, each sample wasanalyzed on 3 days by two analysts on two differentinstruments essentially following the current USP

Table 4. Intermediate Precision Results of Two Lyophilized DrLight Obscuration Measurement (HIAC), Membrane Microscop

Analyst and Instrument Particle Size (mm)

Sub

rP

Light obscuration (HIAC)1 �2 35

�5 7�10 1�25

2 �2 43�5 10

�10 1�25

Membrane microscopy1 �2 Not m

�5 5�10 3�25 1

2 �2 Not m�5 5

�10 3�25 1

Flow imaging (MFI)1 �2 342

�5 48�10 8�25 1


<788> and Ph.Eur.2.9.19 to obtain an estimate ofintermediate precision. Ten individual containers werepooled for each analysis performed by each analyst. Themean values (Day 1, Day 2, and Day 3) are in agreementbetween analysts with similar distribution ranges.However, the largest variability in the particle countand the largest variability range of the results areobserved for the �2 and �5mm sizes. In this study,sample preparation (reconstitution) technique andwaiting time prior to analysis had a significant impacton the results. Higher counts were seen if analysis wasperformed immediately after reconstitution (data not

ug Product Presentations Obtained by Two Analysts Usingy and Flow Imaging (MFI)

visible Particles (Counts per mL) Mean Day 1 to Day 3þDeviationFrom the Mean to the Maximum

1, Lot 1 rP 1, Lot 2 rP 2, Lot 1

0þ 156 4338 þ2735 966þ 7853þ 44 838þ 587 257þ 2194þ9 84þ 47 46þ 281þ 1 3þ 4 3þ 20þ 205 4247þ 3962 1179þ 9550þ65 710þ 854 359þ 3228þ8 67þ 50 57þ 271þ 1 1þ 1 3þ 1

easurable Not measurable Not measurable4þ6 45þ 9 72þ 36þ6 29þ 5 46þ 34þ2 11þ 4 17þ 2easurable Not measurable Not measurable4þ8 48þ 9 77þ 26þ5 29þ 4 51þ 45þ3 12þ 4 21þ 2

4þ 1047 3741þ 2128 2862þ 8179þ 291 597þ 713 593þ 1178þ 17 88þ 45 206þ 585þ5 15þ1 40þ 18

DOI 10.1002/jps

Figure 4. Comparison of particle counts for a MAb for-mulation by HIAC and MFI. The samples were exposed tothree different stress conditions.


shown). Submicroscopic bubbles formed after reconsti-tution that persisted in solution, were the main cause ofthe variability and could not be controlled even thoughan established preparation procedure was followed.Note that while the absolute numbers of subvisibleparticles are different, the relative numbers or trendsbetween samples, are the same. This also demonstratesthe inadequacy of the compendial sample handlingprocedure for many products.

The reconstituted samples, when assessed bymembrane microscopy, show significantly lowercounts in the size range �5mm (Tab. 4). The datawere obtained from 10 different individual con-tainers. The microscopy was performed using a grayfilter to enable the analysts to accurately observe andcount different types of particles.

MFI data for the same product were obtained from10 different individual containers in three replicatessimilar to HIAC analysis (Tab. 4). Subvisible particlecounts are generally higher with the MFI compared tothe HIAC results, although not always. The deviationfrom the mean to the maximum is also higher,suggesting that the MFI data are subject to a similarrange of variability and errors as the HIAC.

The significantly lower counts by the membranemicroscopy method suggests that the HIAC and MFIresults are either (1) impacted by sample preparationprocedure and are likely to be counting the micro-scopic bubbles resulting from the reconstitutionprocedure, or (2) the optical microscopic method ismissing a large number of true particles due toinherent difficulties in their isolation, or (3) acombination of both these effects.

Other comparisons of results by light obscurationand flow-imaging have been published. Huang et al.28

noted much larger differences in particle counts whencomparing measurements by light obscuration andMFI. Measurements in a number of participatingauthors’ laboratories also confirms the observationthat MFI reports significantly higher counts thanHIAC, particularly in the smaller size ranges. Onesuch set of observations is shown in Figure 4 below.Although particles below 10mm were not measuredon the HIAC in this case, the trend for all sizes issimilar in both instruments, leading to the conclusionthat the subvisible particulate results correlate wellwith the applied stress conditions. Similar observa-tions have been made in other laboratories (data notpresented).

Huang et al.28 attribute the differences betweenlight obscuration and flow microscopy results pri-marily to the refractive index of the smaller particles,although the morphology of the particles is alsoexpected to play a role. The hypothesis proposed isthat the smaller the particles, the closer theirrefractive indices are to that of the formulationbuffer. In this situation, the MFI detector, which

DOI 10.1002/jps

captures an ‘‘image’’ as opposed to the HIAC detectorwhich captures a ‘‘shadow,’’ is purported to give amore accurate count. Furthermore, it is proposedthat particles with noncircular morphologies are notaccurately counted by light obscuration. Differencescould also arise due to the different algorithmsused to calculate size in these different techniques.The flow microscopy instruments all use algorithmsthat include assuming the protein particle is asphere, or at best a cylinder. At present, it is notpossible to say which technique (including isolationand microscopy, or the newer flow imaging techni-ques) is giving the most accurate quantitation of theseparticles.

Finally, some illustrative data are given in Table 5for a test system with bovine serum albumin (BSA),where the results by Coulter principle (Multisizer 4)and flow imaging (MFI) are compared in the range 1–10mm. The results are mostly comparable, except inthe high count sample. Here, the MFI gives almost 10times as many counts for the �2mm range, and evengreater factors for larger sizes. The difference couldarise because either the aggregates are reversiblewhen diluted in saline for Multisizer 4 (conductivemedia) or the result is simply a reflection of thedifference in measurement principles. However,measurements in other laboratories have found thereverse trend between flow imaging and Coulterprinciple measurements of the same sample, suggest-ing that the results are product specific.

The difference in numbers of subvisible particlesobtained by the various techniques shown above,highlights that a single value for subvisible particu-lates is difficult to interpret, especially when dealingwith nonspherical, ill-defined amorphous proteinac-eous particles. None of the results are incorrect,


Table 5. Data from a Coulter Counter1 (Multisizer 4) Compared to Flow Microscopy (MFI)

Multisizer 4 �1mm �2mm �3mm �4mm �5mm �6mm �7mm �8mm �9mm �10mm

BSA in PETG bottle 1399 466 166 133 33 33 33 33 0 0BSA with glass beads unshaken 1834 1434 434 134 67 67 33 33 0 0PBS shaken with glass beads 2433 1266 533 233 133 33 33 0 0 0BSA shaken with glass beads 333149 12265 1332 399 99 66 66 33 33 0BSA shaken and filtered 332 299 166 133 66 33 0 0 0 0

MFI �1mm �2mm �5mm �10mm

BSA in PETG bottle 322 187 57 18BSA with glass beads unshaken 1267 363 84 11PBS shaken with glass beads 2351 572 118 18BSA shaken with glass beads 605769 128816 10684 678BSA shaken and filtered 127 53 15 3

3314 SINGH ET AL.

but accurately determining the absolute number ofproteinaceous particles and distinguishing themfrom extraneous particles, is currently not possible.Results therefore need to be placed in context of themethod used for the measurement, and should not becompared across techniques. Differences in reportedresults for the same sample, between the techniquesdiscussed above (and other techniques) will arisedue to the different measurement principles, the widerange of optical and morphological properties of theparticles being counted, and also the individualalgorithms to estimate size and count. The varioustechniques detect particles differently, measuredifferent characteristics and therefore provide differ-ing counts. Furthermore, sample preparation andhandling to adapt the sample to the instrument, is acritical step that adds to the variability in reportedvalue.

A limitation when comparing results from varioustechniques is the lack of appropriate countingreference standards for proteins, as mentionedearlier. The polystyrene particle counting stan-dards differ considerably from the proteinaceousparticles in all their characteristics (morphology,refractive index, shape, texture, deformability,etc.). Interestingly, the absolute numbers obtainedon polystyrene standards by the different techniquesagree well with each other, in contrast to thedata obtained on protein samples above. Figure 3also illustrates the difficulty in defining appro-priate standards for proteinaceous particles giventheir morphological heterogeneity. It also illus-trates the inappropriateness of the current poly-styrene standards which differ significantly fromthe protein particles in both optical propertiesand morphology. Hence, creation of such a standardor of a standardized procedure to create a labo-ratory standard, would be of great value, allow-ing comparisons across techniques/laboratories/products.


Product-Related Factors Leading to Variability inSubvisible Particulates Data

The data (in Tabs. 1, 4, and 5) show that variabilityin levels of subvisible particles can be quite high,especially in the smaller size ranges. We now examinesome of the product-related factors that impact thesecounts.

Some level of particulates is unavoidable inparenteral products. For a given product, actuallevels of particles can conceivably be impacted byprocessing and postproduction handling includingfreeze–thaw, transportation, shock-drop and theoverall time-temperature history. Container/closuresystems can contribute significantly to the back-ground level of (extraneous) particulates in productsmaking it difficult to monitor proteinaceous particles.For products in prefilled syringes (PFS) or cartridges,monitoring can be even more problematic dueto silicone-oil-based particles potentially making upa large part of the total population of subvisibleparticles. Furthermore, the content, distribution andrelease of silicone-oil from the individual PFS varieswidely. Use of surfactant in the formulation mayenhance the dispersion of silicone-oil, while shock/vibration history, storage orientation, stopper move-ment and head-space volume can impact the dis-tribution into the product solution. The variabilityintroduced by container/closure is demonstrated inFigure 5a and b which compare the subvisibleparticles in two lots of the same protein in vialsand PFS. Particle counts in the PFS are almost anorder of magnitude higher. Inability to distinguishbetween silicone-oil and proteinaceous particles is alimitation of current techniques (discussed later).Indeed, it was found in one of the participatingauthor’s laboratories that the particle counts inplacebo from PFS were often indistinguishable fromthat of product, especially below 10mm. Experiencesuggests that as a consequence, a greater degree of

DOI 10.1002/jps

Figure 5. Subvisible particles measured by HIAC on twolots of the same protein filled into (a) vials or (b) prefilledsyringes.


variation will be seen in the particle counts below10mm.

MFI and FPIA claim to be able to distinguishbetween silicone oil and proteinaceous particlesbased on the geometry/morphology as illustrated inFigure 3. However, attempts to develop mathematicalmodels/algorithms to differentiate between the popu-lations can result in significant over/under countingof the protein particles.

The variability due to microscopic bubbles createdduring reconstitution of lyophilized products hasalready been illustrated in Table 3. Proper removalof microbubbles becomes more difficult for higherconcentration products due to the increased viscosity.Surfactants, a very common ingredient in theformulations, add to this difficulty of sample pre-paration.

Another factor for consideration when analyzingreconstituted products is the contribution to parti-culates from the diluent itself. Saline, WFI, or otherdiluents including those packaged in infusion bags

DOI 10.1002/jps

or vials are a possible contributor to thousandsof particles in the lower (�5mm) size ranges, asmeasured in laboratories of a number of thecontributing authors (see, e.g., Mahler et al.30).

In summary, subvisible particulate measurementresults can be impacted by a number of product andprocess factors. A consequence of the variabilityintroduced by such factors is that it might notbe possible to generate reliable baseline data forall product/formulation/container combinations. Thevariability also makes the setting of specifications, orestablishing comparability criteria very problematic.For example, changing from a frozen solution in a vialto a refrigerated liquid in a PFS will affect the numberof subvisible particles, making comparability difficultto establish between materials made at differentstages of a project. Furthermore, it may lead to thesituation that the same biologic in different packa-ging presentations and/or different concentrationshas different subvisible particle limits (even on anormalized basis). An example is shown by the datafor Product 7 in Table 1 where the same product intwo different formulations and two different con-tainer closures has dramatically different particlecounts. A single specification based on a comparisonof these numbers would not be justified. Anyspecification would need to be context (formulation,strength, package, volume) dependent and not simplymolecule dependent. This is unlike other productattributes. Oxidation, for example, is an attributethat can also be potentially impacted by the abovementioned factors, but the mechanism is well under-stood, it is readily measurable, it is not impactedby transportation, it can generally be preservedunchanged by freezing, and it does not varysignificantly within a batch. Molecule specific attri-butes such as glycosylation do not change once amolecule has been defined. In contrast, proteinaceoussubvisible particulates vary as a consequence of allthe above factors and are also impacted by producthistory while a mechanistic understanding of thecause of their formation is seldom very clear.

Sample Requirements

In order to have the possibility of understanding theimpact of subvisible particles on safety, their countsand identity must be known. With this objective andin light of the large differences seen in resultsreported by the techniques (e.g., HIAC and MFI), itwould be important that the same method is usedduring development, for release, and throughout thecommercial lifecycle of a product.

The standard USP <788>/Ph.Eur.2.9.19 metho-dology places a heavy demand on sample required fortesting (summarized in Tab. 2), although smallervolume tests can be qualified as discussed below.However, once the HIAC is calibrated from 2 to


3316 SINGH ET AL.

10mm, no extra sample volume is required beyondwhat is needed for standard pharmacopeia measure-ments. Thus, in formal stability programs, this datacan be gathered. However, if high numbers ofparticles in the 2–10mm range are present, theresultant saturation of the instrument and errormessage can lead to assay failure. Dilution intoformulation buffer might be required if the compen-dial method is used, which would entail methoddevelopment and qualification to ensure that theresults are not impacted.

The use of HIAC in development studies has somelimitations. Generally, the full sample volumes assuggested in USP/Ph.Eur. are not available. Smallermeasurement volumes down to 0.5 mL can bequalified in the HIAC. However, smaller measure-ment volumes are also inherently more prone tocounting errors, further increasing the variability inthe data. MFI requires approximately 0.8 mL; how-ever, replicates are necessary for reliable counting,which raises the volume requirements to about 3 mLafter allowing for accurate sample removal. Studies inwhich such measurements are to be made thereforehave to be selected and designed judiciously.

Development samples also have another weak-ness—these are generally not manufactured in aClass A environment. Environmental contaminationcan confound the results, causing high variabilitybetween vials and artificially inflated counts that canbe a distraction from development. The questionremains as to how to interpret the data generatedduring development. Current regulatory opinionbased on informal discussions is that the data shouldbe evaluated for ‘‘trends.’’ The accuracy and precisionof measurements in this size range, as well asconsistency of measurement techniques used overthe course of product development, should be kept inmind when evaluating the data.

In summary, the appropriate number of samples(sample size) to test depends on the number ofsubvisible particles present, the sample-to-samplevariability associated with the particular product,and the level of acceptable risk, among otherfactors. Confidence in the reliability of the resultsobtained from testing a limited number of samples torepresent the entire lot is significantly affected by thesample size tested. Subvisible particles measure-ments during development and characterization (asillustrated later) will therefore emphasize trendingand relative differences. In these applications, asmaller sample size with the greater inherentvariability in results may be acceptable.

Identification of Particulates—Under WhatCircumstances and How?

The ability to identify the source/nature of particles isimportant in order to be able to eliminate them.


Identification of particulates in the lower subvisiblesize ranges is especially difficult. Thus, if an aberranthigh count is seen, the question becomes ‘‘what shouldwe do with this.’’ Isolation of these particulates is notsimple. The particles can go through the filter, thehandling itself can break them apart, or generatefibrils, and it can be very difficult to detect afew particles of small size and uncertain edges on aneutral background under the microscope. Some newtechniques have been developed that propose to helpaddress this issue, but require further validation.Most identification technologies also focus on a fewparticulates, having been designed to look at foreignmatter, leaving open the question whether the subsetis representative of the sample.

Brightwell’s MFI technology suggests that themorphology is indicative of whether a particulate isproteinaceous or extraneous. However, morphologi-cal characterization of particles is limited by currentcamera technology and is useful only on particles ofsize greater than 4mm.31 Various algorithmic filter-ing techniques can be created to distinguish bubblesor silicone oil droplets,29 although validation of thesealgorithms for quality control purposes is consideredproblematic, and can also result in mis-assignment ofa large fraction of the protein particles as silicone oil(up to 20%, data not shown). More experience will berequired to confirm these capabilities, and whether itcan distinguish between a protofibril and a cellulosefiber or other material such as rubber/plastic (con-sidering the size limitation given above). Thetechnology will be useful as an investigative tool incase of a ‘‘deviation’’ and as a developmental tool.

Rap.ID (Berlin, Germany) using Raman spectro-scopy may also provide some clues. However, in theauthors’ experience, it is not always unequivocal inidentification of particles as proteinaceous or not. Thequality of the database and statistics of scannedparticles as well as quality of spectral overlaysimpacts the analytical output significantly. Particlesin protein solutions can be heterogeneous, and thisalso confounds the spectral identification.

If the difficulty of isolation is overcome, attemptscan be made to further characterize and analyzeaggregates on a structural level. Techniques used forthe structural analysis such as circular dichroism(CD), Fourier-transformed infrared spectroscopy(FT-IR), nuclear magnetic resonance spectroscopy(NMR) or intrinsic fluorescence might be considered.Other techniques that can be used to identifyparticulates, that is, to differentiate whether theyare proteinaceous or from an external source, areRaman or FT-IR (micro)spectroscopy, Electron micro-scopy with elementary analysis may also be utilized.

Data analysis (e.g., spectral comparison and over-lays) should be considered with care. For examplespectra overlays may be difficult to interpret due to

DOI 10.1002/jps

Figure 6. Subvisible particles measured after reconstitu-tion for a lyophilized product stored under stressed storageconditions, using a small volume HIAC method.


differences in signal intensities and signal broad-ening of aggregated and isolated species compared tothe signal of the species in solution, differentcontributions from buffers, etc. Additionally, thesensitivity of structural analysis methods such asFT-IR to small conformational changes (<5% changein signal) is low. Furthermore, even extraneousparticles may be of proteinaceous nature (e.g., humandander). Dissecting and identifying individualcomponents in heterogeneous particles containingmore than two species is very difficult. Therefore, anunambiguous composition of particles cannot alwaysbe given by any available method to date. Due to thelimitations of the methods used, any resultant datashould be interpreted with care and on a case-by-casebasis.

Similar to particulate counting, the primarylimitation in all identification methods is samplepreparation and handling, and must be kept in mindwhen interpreting data. Well-trained operators mustpreferably be used. The use of such techniques cannotbe left to every development scientist. A highly skilledforensic laboratory would be required to isolate andidentify particles of 20mm and larger, and this wouldnot be a routine task. Identification of particles below20mm may sometimes be possible but with significantchallenges in unequivocal identity assignment.

From a risk management perspective, given thelack of ability to identify the particle populationinvolved, the assumption may have to be made thatall particles in the size range below �10–20mm areproteinaceous, adding to the error and variability ofthis parameter. The only situation where an alter-native assignment could be made is if a placebomanufactured in parallel shows similar amounts ofsubvisible particles, in which case they could beassessed to be ‘‘most likely not protein.’’ However,such a possibility is not always available and alsoassumes similarity in the amounts of contaminatingextrinsic particulates, including the silicone oildroplets that are present in any product from PFS.

Figure 7. Subvisible particles measured by (a smallvolume) HIAC method on untreated and stressed monoclo-nal antibody.

VALUE OF MONITORING SUBVISIBLEPARTICLES OVER A BROAD RANGE

Available observations, albeit limited, suggest thatparticle counts within the various size ranges arepositively correlated. The data presented below(Fig. 6) on a lyophilized product stored under stressedconditions shows that subvisible particles in allsize ranges increased concurrently. The number ofparticles in size range �10mm correlated with thecounts for the smaller sizes.

A similar correlation between the subvisibleparticle counts in various size ranges was detectedfor another antibody when a stressed and an

DOI 10.1002/jps

unstressed sample were compared (see Fig. 7). Thenumbers of subvisible particles increased due to lightand heat stress, but the size distribution remainedconstant as indicated by the parallel lines, especiallyabove the LOD/LOQ of the techniques (10 particlesper mL).

Although the positive correlation between particlesizes shown in the above two examples may notalways hold, it is apparent from these and theprevious examples that the numbers of particles inthe �2 and �5mm ranges represent a good tool forproduct development. These numbers tend to be highand are therefore a sensitive measure to assessimpact of variables of interest during developmentalstudies. However, as discussed above, the highsensitivity is also associated with higher variability


Table 6. Impact of Filtration on Subvisible Particle Populations Measured by Light Obscuration (Narhi et al.6)

Sample Type Particles/mL, �2mm Particles/mL, �5mm Particles/mL, �10mm Particles/mL, �25mm

Unfiltered, Lot 1 339�78 29� 2 <10 <10After 1� filtration Lot 1 29�6 11� 5 <10 <10After 2� filtration Lot 1 19�9 14� 9 <10 <10After 5� filtration Lot 1 50�14 12� 9 <10 <10Unfiltered, Lot 2 680�51 36� 4 <10 <10After 1� filtration Lot 2 20�3 5� 2 <10 <10

3318 SINGH ET AL.

or noise. Data interpretation must be made with careto prevent time and effort being expended on falsetrails. Keeping these caveats in mind, this type ofdata can be useful in designing optimal formulationsand in troubleshooting.

The utility of measuring particles under 10mmsizes for examining a process step is illustrated by thedata in Table 6 where the efficacy of the filtration stepin removing these particle is shown.6 The number ofparticles in the �10mm size range was below LOQ(about 10 counts/mL) and therefore not useful forprocess development. Results for the �2mm size showthat a single filtration step is adequate and subse-quent filtration steps do not provide added benefit inparticle clearance.6

To be useful, these types of comparative analysesmust be performed under similar conditions (con-centration, container/closure) and by the samemethodology (sample preparation procedure, technol-ogy/instrument).

Setting Specifications for Sizes <10mm?

When performed properly, measurements of sub-visible particles below the 10mm range, can haveutility in guiding product development and inestablishing clinical experience with the product.However, the ability and utility of setting specifica-tions for particles in this size range requires muchmore critical thought. The variability of the methodsfor measuring particles under 10mm and the otherdifficulties in analysis discussed earlier (silicone oil inPFS/cartridges, bubbles, high concentration/viscositysolutions, environmental contaminants, impact ofcontainer/closure, handling, time/temperature his-tory, etc.) need to be considered for the specificproduct before it can be determined whether mean-ingful specifications can be established. Heterogene-ity within a batch is quite likely, increasing thesampling requirements to get a statistically relevantnumber. This is quite unlike other quality attributes.In order to accommodate the variability, producthistory will need to be acquired over multiple drugsubstance and drug product batches. In the case ofPFS or cartridge products, this need will be magnifiedto cover various silicone loads. If such a path is taken,it is likely that specifications for each size range will


be driven by the product/container configuration andnot by any characteristic of the biologic itself.

Subvisible particles assessment as part of compar-ability exercises may present several logisticalchallenges. For most quality attributes, the preferredapproach to establishing comparability is to perform aside-by-side analysis of the two processes/formula-tions/configurations to be compared, to minimizediscrepancies contributed by day-to-day assay varia-bility. Material (samples) retained from variouscritical (toxicology, clinical) batches would need tobe available in sufficient number and volume to bepart of this study. This would require significantlymore retained samples than is current practice asthese samples would be required to support anychange in process, container/closure, strength,volume, etc. The above-mentioned heterogeneitywithin a batch will also significantly increase thenumber of samples required to get a statisticallyrelevant comparison. Furthermore, it is very possiblethat the subvisible particle profile would changeduring storage of these samples, making the trueassessment of comparability difficult. This is all themore likely if the retained samples are stored frozen,as is the practice so that other quality attributes donot change. These limitations may require thatcomparability be assessed by evaluating the particu-late profile in relation to historical informationavailable for the product, thus necessitating a goodunderstanding of the variability of the method usedfor measuring the subvisible particulates as well asthe normal variability observed for the product.Modifications in container/closure and/or formulationduring the course of development (e.g., going fromvial to syringe or from lyophilized to liquid) couldresult in differences that have no clinical impact, suchas an increase in reported particulates due to siliconeoil droplets. With these constraints, arriving at ameaningful conclusion would be challenging, espe-cially given the generally limited history andexperience with this attribute during development.

Risk Assessment

The mass represented by proteinaceous subvisibleparticles over a wide size range is shown in Figure 2.Particle sizes that are currently already specifiedrepresent the greatest mass of protein in particulate

DOI 10.1002/jps


form. On the other hand, particles between 1 and10mm, even if present at 10,000 counts/mL, representabout 0.1mg of material, while the protein aggregatesat sizes below 1mm comprise material in the ng to pglevels (see Fig. 2). The risk of an immunologic reactionwith this amount of material is low considering thateven vaccine doses lie in the mg–mg range, andfurther require the inclusion of adjuvants. Theauthors realize that the comparison with vaccinesis not perfect since unlike vaccines, certain proteintherapeutics are dosed chronically, while vaccines arepurposely dosed at levels high enough to create animmune response. However, given the lack of clarityaround the impact of aggregates on immunogenicity,transferring learnings from related fields may help toput the risk from aggregates in perspective.

Another aspect to assess is the level of theproteinaceous subvisible particles relative to otherproduct-related impurities, variants or degradants.Subvisible particles comprising 0.1mg of proteinrepresent about 0.0002% of a 50 mg/mL monoclonalantibody product and about 0.01% of a 1 mg/mLtherapeutic protein product. For most other productquality attributes (e.g., oxidation, deamidation, charge-variants, glycosylation variants, etc.), these levels areneither detected nor reported, and are considered to bewell below the safety concern threshold.

More experience with the measurement andmonitoring of subvisible particles, similar to otherproduct quality attributes is therefore required todevelop thresholds for control and monitoring ofproteinaceous subvisible particulates. A step in thisdirection would be to assess the current stateof affairs in more detail. A survey of subvisibleparticulate levels in current marketed products(along the lines in Tab. 1) could help set possiblebenchmarks to judge individual products duringdevelopment, especially prior to accumulatingclinical experience. For instance, questions like ‘‘is10,000 counts/mL of size 2–10mm high, low or aboutaverage?’’ would be very helpful in decision making.Given the lack of experience as well lack of clarityaround the clinical impact of aggregates, this mayprovide an avenue to approach risk assessmentaround this question. Such an assessment can putthe subvisible particle data in perspective, especiallyif the risk is broken down by therapeutic area,functional modality, biologic type (monoclonal anti-body or therapeutic protein), duration of therapy(acute or chronic), dosing site (IV, IM, SC), disease,and other patient-related factors.

CONCLUSIONS AND RECOMMENDATIONS

(1) C

DOI 10.

oncerns about aggregation as a potentialcause of immunogenicity in human biothera-

1002/jps JOURNA

peutics have been around for a long time. Themultiplicity of determinants of immunogenicityin a clinical setting makes the deconvolution offactors very difficult, and no direct evidencefrom clinical studies has been forthcoming todemonstrate that aggregates are a true domi-nant risk factor. It is also difficult to determinethe immunogenicity potential of naturallyoccurring aggregates of a particular humanprotein in human patients, based on the resultsof animal studies using artificially createdaggregates due to differences in the morphol-ogy and structure of the aggregates. However,we concur with Carpenter et al.7 in their sug-gestion that this remains an important area forcontinued research.

(2) T
he authors agree that research is needed andnew technology must be developed for accurateparticulate (and aggregate) characterizationand quantitation. Development of proteinac-eous subvisible particle standards would beof immense value in improving the resultsfrom the current techniques. Simultaneously,research into the clinical immunogenic poten-tial of aggregates in general and proteinaceous(subvisible) particles in particular must be pro-gressed, as stated above. However, it is alsolikely that such a link may never be unequi-vocally demonstrated. An effort to assess of thecurrent state of affairs among marketed pro-ducts, broken down by therapeutic area andclass of biologic among other factors, may bean approach to understand the true risk poten-tial.
(3) P
roteinaceous subvisible particles representlarge assemblies of protein molecules, and there-fore are subject to the same concerns as aggre-gates in general. Measurements on a limited setof marketed products suggest that these (andpresumably most if not all marketed products)contain large but varying numbers of subvisibleparticles in the size range below 10mm. Thesesmall particles represent protein content in frac-tions of mg or lower, with the largest portion ofthis mass represented by the size rangesabove 10mm, the population that is currentlymonitored. To place this amount in perspective,other product quality attributes such as oxida-tion, deamidation, charge-variants or glycosyla-tion-variants, are neither detected nor reportedat these levels, and are considered well belowthe safety concern threshold. Human biothera-peutics currently on the market continue to besafe and efficacious from an immunologic per-spective, despite the gap in the monitoring iden-tified in the commentary and further detailed inthis article.

3320 SINGH ET AL.

(4) M

JOURNA

easurement of subvisible particulates smal-ler than 10mm size range has value as an aid inproduct development and characterization.Data can be generated as a ‘‘for informationonly’’ test, assisting in the understanding of theproduct behavior. However, lack of ability todifferentiate proteinaceous from extrinsic par-ticles between 0.1 and 10mm particles, andvariability introduced as a consequence of fac-tors such as container/closure, product concen-tration, viscosity, product history, and inherentheterogeneity within a batch, imply that theresults would differ considerably betweenphases of the development cycle. Furthermore,considering the limitations of the methodolo-gies (namely low precision and reproducibility)demonstrated here, as well as limited experi-ence and the lack of understanding of theimpact of this attribute, it is not appropriateto add specifications for subvisible particles inthe size range <10mm at release and stability,or for comparability at the present time.

(5) C
haracterization of product quality attributesincluding subvisible particulates, is performedon all products in R & D during clinical develop-ment. Preclinical studies are not currentlycapable of assessing the clinically relevantimmunogenicity potential of product attributessuch as aggregates and/or particles smallerthan 10mm. Preclinical studies are also notcurrently capable of assessing the clinicallyrelevant immunogenicity potential of any typeof product variants or degradants present atfraction ofmg levels or less. Thus, the safety andefficacy profile of biotech products is estab-lished during research and development byappropriately designed nonclinical and clinicalstudies. Immunogenicity is monitored andevaluated during clinical studies using the clin-ical study material, and thereby product qual-ity attributes which could be potentiallyrelevant for safety and/or efficacy are qualified.Subvisible particles, of which the currentlymonitored range �10mm population wouldrepresent the largest fraction of protein aggre-gate mass (if proteinaceous), remain a productquality attribute that is also qualified in thisfashion.
ACKNOWLEDGMENTS

The authors would like to acknowledge a number ofpeople from various organizations who helped to gen-erate the data and/or shape the discussions: HatsukiAsahara, Nancy Barbour, Shawn Cao, Art Chirino,


Adriana Chynoweth, Lisa Erickson, Erwin Freund,Michael Grace, Gino Grampp, Richard Harms,Wenchang Ji, Yijia Jiang, Nancy Jiao, Chulani Kar-unatilake, Brent Kendrick, Bruce A. Kerwin, HansKoll, Ian Langdon, Tony Mire-Sluis, David Palmlund,Matt Rhyner, Thomas Schreitmuller, Deepak Sharma,Li Shi, Debra Shnek, Steve Swanson, Julie Trulson,Ed Walls, Greg Walsh.

REFERENCES

1. Rosenberg AS. 2006. Effect of protein aggregates: An immu-nologic perspective. AAPS J 8:E501–E507.

2. Mahler H-C, Freiss W, Grauschopf U, Keise S. 2009. Proteinaggregation: Pathways, induction factors and analysis. JPharm Sci 98:2909–2934.

3. Philo JS. 2006. Is any measurement method optimal for allaggregate sizes and types? AAPS J 8:E564–E571.

4. Philo JS. 2009. A critical review of methods for size character-ization of non-particulate protein aggregates. Curr PharmBiotechnol 10:359–372.

5. Das T, Nema S. 2008. Protein particulate issues in biologicsdevelopment. Am Pharm Rev 11(4):52–57.

6. Narhi LO, Jiang Y, Cao S, Benedek K, Shnek D. 2009. A criticalreview of analytical methods for subvisible and visible parti-cles. Curr Pharm Biotechnol 10:378–381.

7. Carpenter JF, Randolph TW, Jiskoot W, Crommelin DJA,Middaugh CR, Winter G, Fan YX, Kirshner S, Verthelyi D,Kozlowski S, Clouse KA, Swann PG, Rosenberg A, Cherney B.2009. Overlooking subvisible particles in therapeutic proteinproducts: Gaps that may compromise product quality. J PharmSci 98:1202–1205.

8. Knapp JZ. 1999. The scientific basis for visible particle inspec-tion. PDA J Pharm Sci Tech 53:291–302.

9. Rosenberg AS, Worobec A. 2004. A risk-based approach toimmunogenicity concerns of therapeutic protein products. PartII: Considering host-specific and product-specific factorsimpacting immunogenicity. BioPharm Int 17(12):34–42.

10. Schellekens H. 2005. Factors influencing the immunogenicityof therapeutic proteins. Nephrol Dial Transplant 20:vi3–vi9.

11. Braun A, Kwee L, Labow MA, Alsenz J. 1997. Protein aggre-gates seem to play a key role among the parameters influencingthe antigenicity of Interferon alpha (IFN-a) in normal andtransgenic mice. Pharm Res 14:1472–1478.

12. Hermeling S, Jiskoot W, Crommelin DJA, Bornaes C, Schelle-kens H. 2005. Development of a transgenic mouse modelimmune tolerant for human Interferon beta. Pharm Res 22:847–851.

13. Hermeling S, Schellekens H, Maas C, Gebbink MFBG, Crom-melin DJA, Jiskoot W. 2006. Antibody response to aggregatedhuman Interferon alpha2b in wild-type and transgenic immunetolerant mice depends on type and level of aggregation. JPharm Sci 95:1084–1096.

14. Purohit VS, Middaugh CR, Balasubramanian SV. 2006. Influ-ence of aggregation on immunogenicity of recombinant humanFactor VIII in hemophilia A mice. J Pharm Sci 95:358–371.

15. Schellekens H. 2008. How to predict and prevent the immu-nogenicity of therapeutic proteins. Biotechnol Annu Rev 14:191–202.

16. Hwang WYK, Foote J. 2005. Immunogenicity of engineeredantibodies. Methods 36:3–10.

17. Kiese S, Pappenberger A, Friess W, Mahler H-C. 2008. Shaken,not stirred—Mechanical stress testing of an IgG1 antibody. JPharm Sci 97:4347–4366.

DOI 10.1002/jps


18. Fradkin AH, Carpenter JF, Randolph TW. 2009. Immunogeni-city of aggregates of recombinant human growth hormone inmouse models. J Pharm Sci 98:3247–3264.

19. Bugelski PJ, Treacy G. 2004. Predictive power of preclinicalstudies in animals for the immunogenicity of recombinanttherapeutic proteins in humans. Curr Opin Mol Ther 6:10–16.

20. Jahn E-M, Schneider C. 2009. How to systematically evaluateimmunogenicity of therapeutic proteins—Regulatory consider-ations. New Biotechnol 25:280–286. 10.1016/j.nbt.2009.03.012.

21. Rosenberg AS, Worobec A. 2005. A risk-based approach toimmunogenicity concerns of therapeutic protein products. PartIII: Effects of manufacturing changes on immunogenicity andthe utility of animal immunogenicity studies. BioPharm Int18(1):32–36.

22. White PJ, Anastasopoulos F, Church JE, Kuo C-Y, Boyd BJ,Hickey PLC, Sze Tu L, Burns P, Lew AM, Heath WR, DaveyGM, Pouton CW. 2008. Generic construction of single compo-nent particles that elicit humoral and cellular responses with-out the need for adjuvants. Vaccine 26:6824–6831.

23. Nath N, McNeal E, Obenhuber D, Pillari B, Shelton L, Stevens-Riley M, Sweeney N. 2004. Particulate contaminants of intra-veneous medication and the limits set by USP General Chapter<788>. Pharmacopeial Forum 30:2272–2280.

24. Cao S, Pollastrini J, Jiang Y. 2009. Separation and character-ization of protein aggregates and particles by field flow fractio-nation. Curr Pharm Biotechnol 10:382–390.

DOI 10.1002/jps

25. Chrai S, Clayton R, Mestrandrea L, Myers T, Raskin R, SokolM, Willis C. 1987. Limitations in the use of HIAC for productparticle counting. PDA J Parenteral Sci Technol 41:209–214.

26. Etzler FM, Sanderson MS. 1997. Particle size analysis. Acomparative study of various methods I. Part Part Syst Charact12:217–224.

27. Etzler FM, Sanderson MS. 1997. Particle size analysis. Acomparative study of various methods II. Part Part Syst Char-act 12:278–282.

28. Huang CT, Sharma D, Oma P, Krishnamurthy R. 2008. Quan-titation of protein particles in parenteral solutions using Micro-Flow Imaging. J Pharm Sci 98:3058–3071.

29. Sharma D, King D, Moore P, Oma P, Thomas D. 2007.Flow microscopy for particulate analysis in parenteral andpharmaceutical fluids. Eur J Parenteral Pharm Sci 12:97–101.

30. Mahler H-C, Fischer S, Randolph TW, Carpenter JF. 2010 (inpress). Protein aggregation and particle formation: Effects offormulation, interfaces and drug product manufacturing opera-tions. In: Wang W, Roberts C, editors. Formulation and processdevelopment strategies for manufacturing of a biopharmaceu-tical. New York: John Wiley.

31. Brown L. 2009. Imaging particle analysis: Resolution and sam-pling considerations. http://www.fluidimaging.com/imaging-particle-analysis-white-papers.aspx (accessed June 2009).


Documents

An industry perspective on the monitoring of subvisible particles as a quality attribute for protein therapeutics