`
`Particulate Matter in Parenteral Products: A Review
`
`~VEN J. BORCHERT..c., AMY ABe•, D. SCOTT ALDRICH, LLOYD E. FOX, JAMES I:. FREEMAN,
`and ROBERT D, WHITE
`
`The Quallty C:OntroJ Dirisioa, The Upjobn Company, Kalamazoo, Micblgaa
`
`ABSTRACT: Particulate matter in parenteral products is a complex subject. This article contaiw a
`discussion of several aspects of this topic including the use and limitations of the inspection and counting
`techniques for visible and subvisible particles, the identification of particles, and the elucidation of sources,
`mechanisft!S of formation, and particulate reduction techniques. Two significantly different approaches,
`human and machine inspection, have been used 10 detecl visible particulate matter in parenteral products. A
`description of both methods is given along with a discussion of their typical performance characteristics.
`Criteria for comparison of different visual inspection systems are also presented. A variety of methods have
`been utilized f()r the measurement of mbvisible particulate matter, including microscopic, electrical zone(cid:173)
`sensing, light blockage, light scattering, and holographic techniques. Each of these particle counting methods
`is described. In addition, the factors that affect the measurement of subvisible particulate matter are
`discussed. An approach to particle identification is outlined. General comments concerning the analysis of
`particulate matter in parenteral products are discussed along with a description of various particle identifi(cid:173)
`cation techniques and several examples illustrating how the methods have been applied. In particular, the
`techniques that are presented i,wlude light microscopy, atomic spectroscopic methods (SEM/EDXRA,
`electron microprobe, ESCA, and Auger electron spectroscopy), molecular spectroscopic techniques (jnfrared
`spectroscopy, Raman spectroscopy, and mass spectrometry), and chromatography. Finally, sources of
`particulace matter tncluding packaging macerials, manufacturing variables, formulation components, and
`miscellaneous factors are reviewed. The different mechanisms of particle formation, namely, direct contami(cid:173)
`nation, precipitation and agglomeration are discussed. Representative examples of particulate reduction
`steps are presented.
`
`Table of Contents
`
`Page
`Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212
`1. Inspection and Counting Techniques-Visible and
`214
`Subvisible Particles . . . . . . . . . . . . . . . . . . . . . . . . • .
`214
`A. Visible Particles . . . . . . . . . . . . . . . . . . . . . . . . . .
`Human Visual Inspection . . . . . . . . . . . . . . . • • . . 214
`Machine Visual Inspection . . . . . . . . . . . . . . • . .
`217
`Comparison Criteria . . . . . . . . . . . . . . . . . . . • . .
`218
`Use oflnspection Systems . . . . . . . . . . . . . . . • . .
`219
`B. Subvisiblc Particulate Matter . . . . . . . . . . . . • . .
`220
`!>article Counting "Techniques . . . . . • . . . . . . . . .
`220
`Characteristics . . . . . . . . . . . . . . . . . . . . . . . . . . .
`221
`Use of Detection Methods . . . . . . . . . . . . . . . . . .
`223
`l . Identification of Particulate Matter . . . . . . . . . . . . • . 223
`A. General Comments, Sample Selection and Sam•
`pie
`. . . . . . . . . . . . . . . . . . . . . . . . . . . . 223
`Preparation
`B. Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
`225
`C. Atomic Spectroscopic Techniques . . . . . . . . . . . .
`227
`D. Molecular Spectroscopic Methods
`230
`. . . . . . . . . . .
`
`,. Author to whom inquiries should be addressed.
`~ The Sherwin-Williams Company, Chicago, IL60628.
`
`232
`232
`
`E. Chromatography . . . . . . . . . . . . . . . . . . . . . . . . .
`F. A Case History
`. . . . . . . . . . . . . . • . . . . . . . . . . .
`3. Sources of Particles. Mechanisms ofTheir Formation
`234
`and Particulate Reduction Steps . . . . . . . . . . . . . . . .
`234
`A. Sources of Particles . . . . . . . . . . . . . . . . . . . . . . .
`235
`B. Mechanisms of Their Formation . . . . . . . . . . . . .
`235
`C. Particulate Reduction Ste~ . . . . . . . . . . . . . . . .
`Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
`Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 237
`References . . . . . . . . . . . . . . . . . . . . . . . . • . . . . . . . . . . . 237
`
`Introduction
`Control of the key features of a product and the process(cid:173)
`es by which it is manufactured is essential to the assurance
`of quality for that product. With parenteral formulations
`there are several important variables including potency,
`pH, sterility, pyrogenicity, and particulate matter. Of
`these, control of the particulate quality can represent a
`significant challenge.
`There are at least three reasons for focusing attention
`on particulate quality of parenteral products. These are
`
`212
`
`Journal of Parenteral Science & Technology
`
`FRESENIUS EXHIBIT 1058
`Page 1 of 30
`
`
`
`safety concerns, legal requirements, and process evalua(cid:173)
`tion.
`Concerns for patient safety were first addressed in the
`works of Gatvan and Gunner (1, 2). Since their initial
`reports, numerous papers have been published on this
`topic and a few review articles have been written concern(cid:173)
`ing the clinical significance of particulate matter (3-5,
`234).
`Although there is some controversy about this subject,
`it has been concluded that injectables should not contain
`an excessive number of particles. The primary evidence
`for this can be found in the literature on drug abuse ( 6-
`12). Injections of crushed tablets, capsules, and other solid
`dosage products have often resulted in serious conse(cid:173)
`quences. For example, one drug user died after i.v. injec(cid:173)
`tion of Darvon capsules (8). In another report, Douglas
`and coworkers found that three of seven addicts who in(cid:173)
`jected drugs had rontgenographic, pathologic, and func(cid:173)
`tional manifestations of pulmonary foreign body emboli
`and granulomas; the remaining had abnormal pulmonary
`function (9).
`The results of numerous animal experiments have also
`been reported (13-28). In many of these studies massive
`doses of particulate matter were injected into several spe(cid:173)
`cies including dogs, rabbits, rats, mice, and hamsters. The
`particles consisted of glass beads, cotton fibers, polysty(cid:173)
`rene latex spheres. paper fragments, and other insoluble
`substances. In addition to differences in the type and
`shape of the particulate matter, the particle sizes ranged
`from a few microns to several hundred microns. Although
`these studies provide less direct evidence than drug abuse
`studies, they do indicate that excessive levels of particu(cid:173)
`late matter in i.v. solutions can be harmful.
`However, the clinical ramifications are much less clear
`when we are concerned about the use of parenterals con(cid:173)
`taming levels of particles that are typical of current prod(cid:173)
`ucts. As T11rco and Davis suggested in a previous review
`article (3), lack of definition of these effects and, there(cid:173)
`fore, the significance of particulate matter is caused pri(cid:173)
`marily by the absence of controlled studies on humans.
`Some studies have been reported (29-38), but their con(cid:173)
`clusions are qualitative and nonspe<:ific. In particular,
`control human tissue samples are difficult to obtain be-(cid:173)
`cause of disease, environmental pollutants, and social
`habits. Finally, controls are difficult since tissue samples
`are taken from people who may have received an undeter(cid:173)
`mined number of parenteral solutions (3).
`The regulations pertaining to particulate matter in par(cid:173)
`enterals vary considerably among the different compendia
`(39-42). For example, the specifications for the British
`Pharmacopeia, United States Pharmacopeia, and The
`Pharmacopeia of Japan are shown in Tables I-III.
`It is noteworthy that the compendial requirements de(cid:173)
`pend upon the size of the particulate matter and upon
`whether the injectable is a Large--Volume Piuenteral
`(LVP) or a Small-Volume Parenteral (SVP). In the case
`of visible particles, there are regulations for both L VP and
`SVPproducts. The exact wording varies among the differ(cid:173)
`ent compendia, but the specifications are very similar.
`lnjectables are supposed to be clear and essentially free of
`
`TABLE I. USP XXI
`l"arlicle Size Parenteral
`
`Visible
`
`LVP
`SVP
`
`Subvisible
`
`L VP
`
`Subvisible
`
`SVP
`
`Requlremeat
`
`Good pharmaceutical practice
`requires that each final con(cid:173)
`tainer of Injection be subject(cid:173)
`ed individually to a physical
`inspection, whenever the na(cid:173)
`ture of the container permits,
`and that every container
`whose contents show evidence
`of contamination with visible
`foreign material be reject~.
`Microscopic:
`Not more than 50 particles per
`ml that are e(}Ual to or larger
`than 10 µm and not more than
`5 particles per ml that are
`equal to or larger than 25 pm
`in effective linear dimension.
`Light-obscuration:
`Not more than I 0,000 particles
`per container that are equal to
`or greater than IO µm in effec(cid:173)
`tive spherical diameter and/or
`1000 particles per container
`equal to or greater than 25 µm
`in effective spherical diameter.
`
`particles that can be seen by the unaided eye. However.
`the situation is different for subvisible particulate matter.
`Currently, most compendia have specifications for LVP
`solutions, but few have requirements for SVP products.
`Even if one restricts a comparison of the various regula(cid:173)
`tions to L VP solutions, there are significant differences in
`the size of particles that are measured and the methods by
`which they are detected. It is not possible to determine
`which compendia) specification is more stringent without
`making several assumptions concerning the size distribu(cid:173)
`tion of particles and their shape ( 4 ).
`Several rationales have been presented to justify the·
`guidelines concerning particulate matter. Some have ar-
`
`TABLE 11. British Pharmacopeia 1980
`Requirement
`Particle Size Parenteral
`
`Visible
`
`LVP
`SVP
`
`Subvisible
`
`L VP
`
`Injectable Preparations which
`are solutions. when examined
`under suitable conditions of
`visibility, are clear and practi(cid:173)
`cally free of particles.
`Electrical zone-sensing:
`Does not e:xceed 1000 per ml
`greater than 2.0 #lm and does
`not e:xoeed 100 per ml greater
`than 5.0 ,um.
`or
`Light blockage:
`Does not exceed 500 per ml
`greater than 2.0 µm and does
`not exceed 80 per ml greater
`than 5.0 µm.
`
`Vol. 40, No. 5/September-October 1986
`
`213
`
`FRESENIUS EXHIBIT 1058
`Page 2 of 30
`
`
`
`TABLE Ill. Japanese Pharmacopcia. Tenth Edition
`Partirle Size Parenteral
`Requirement
`LVP
`SVP
`
`Visible
`
`When the outer surface of the
`container is cleaned, injectable
`solutions or solvents for drugs
`to be dissolved before use•
`must be clear and free from
`foreign insoluble matter that is
`readily noticeable when in(cid:173)
`spected with unaided eyes at a
`position of luminous intensity
`of about I 000 lu:tcs [ 93 foot(cid:173)
`candles}. right under an incan•
`descent electric bulb. As for
`injections contained in plastic
`containers. the inspection is
`performed with unaided eyes
`at a position of luminous in(cid:173)
`tensity of 8000 to IOOO0 luxes
`1740 to 930 footcandles) with
`incandescent electric bulb
`placed at appropriate dis(cid:173)
`tances above and below the
`container.
`Microscopic:
`The limits arc not more than 50
`particles per ml that are equal
`to or larger than to pm and
`not more than 5 particles per
`ml that are equal to or larger
`than 25 µm.
`• There is an analogous requirement for preparations that are to be
`dissolved before use.
`
`Subvisible
`
`LVP
`
`gued that the particle standards are consistent with the
`capabilities of existing technology and, hence, are a mea(cid:173)
`sure of good manufacturing practice (4, 40, 43-45). Oth(cid:173)
`ers have stated that they can be justified on the basis of
`cumulative particulate insult the patient receives ( 40, 46).
`For example, the differences in L VP and SVP regulations
`have been rationalized on both accounts (46, 47).
`Numerous articles have been published concerning the
`level and sizes of particulate matter in LVP and SVP
`injectables (43, 48-7&). These studies have utilized a vari(cid:173)
`ety of methods for measuring particle counts including
`microscopic, light blockage, light scattering, and electri(cid:173)
`cal zone-sensing techniques. Furthermore, a wide variety
`of products, packaging types and dosage forms have been
`examined.
`There has also been a considerable interest in particu(cid:173)
`late matter for the purpose of process evaluation. Others
`have suggested that a significant increase in the level of
`particles for a parenteral could be used as an indication
`that the product or the process by which it is manufac(cid:173)
`tured may not be well-controlled (4, 43, 46, 49, 65, 79).
`The levels of both visible and subvisible particles have
`been considered as useful measures for process control
`requirements. For example, Brownley reported on the use
`of process control charts for rejection rates in visual in(cid:173)
`spection (79). In addition to data for the total number of
`rejects, he discussed the use of charts for specific types of
`visible particulate matter such as lint and glass. Brownley
`
`214
`
`I
`
`argued that every parenteral manufacturer could benefit
`from this type of information in assessing the process
`capabilities of an op~ation to produce a high quality
`product.
`As an inde~ of quality, others have suggested that the
`data on the level of subvisible particles could be even more
`helpful than the results of visual inspection ( 4, 43, 46, 49).
`Particle size distributions have been reported for numer•
`ous parenterals (49- 52, 54, 59, 63, 64, 67, 73). In the case
`of subvisible particles, it has been observed that there is a
`log- log relationship between the size and the number of
`particles ( 4, 43, 46, 49) and most workers have decided to
`summarize their data using the following equation
`In N = In N1.o- Min D
`where N is she cumulative number of particles at the
`threshold corresponding to diameter D, N 1 _0 is the value of
`N for D = 1.0 pm and Mis the slope of the Jog-log plot.
`Based on the results of these size distributions, a variety of
`limits have been suggested for both L VP and SVJ> solu(cid:173)
`tions ( 4, 43, 46, 49, 65).
`The detection and quantilation, identification and ulti(cid:173)
`mate reduction of particulate matter in parenteral prod•
`ucts represent a complex subject. This paper addresses
`particulate matter in parenteral products in three sec(cid:173)
`tions: ( /) inspection and counting techniques-visible and
`subvisible particles; (]) identification of particulate mat(cid:173)
`ter; and (3) sources of particles, mechanisms of their
`formation, and particulate reduction steps.
`1. Inspection and Counting Techniques-Visible a■d
`Subvisible Particles
`Inspection for visible particulate matter and the enu(cid:173)
`meration of subvisible particles provide a quantitative
`assessment of product quality. This section is separated
`into two parts: (A) Visible Particles and (B) Subvisible
`Particulate Matter. Each part will contain a description of
`the various techniques that are utilized as well as a discus(cid:173)
`sion of their performance capabilities and limitations.
`(A) Visible Particles: As shown in Table 1, the USP
`stipulates 100% impection of injectables for visible for(cid:173)
`eign material. Not all compendia require 100% inspection
`of parenterals, but most state that the injectables are
`supposed to be practically free of particles which can be
`seen by the unaided eye. Two significantly different meth(cid:173)
`ods have been used to detect the presence of visible parti(cid:173)
`cles. One utilizes people and the other uses machine detec(cid:173)
`tion. For each method a general description will be fol(cid:173)
`lowed by a discussion of the typical performance
`characteristics of the various techniques. Next, criteria
`that can be used to compare different inspection syste~
`will be presented. Finally, in view of the current knowl(cid:173)
`edge of visual inspection methods some general comments
`will be presented.
`Human Visual lmpeczion: A review of the literature
`indicates that human inspections have been carried out in
`a variety of ways (80-86). General guidelines for this
`process were developed by a Parenteral Drug Association
`(PDA) Task Force (85). In particular, the normal inspec(cid:173)
`tion apparatus is comprised of a box containing a lamp
`
`Journal of Parenteral Science & Technology
`
`FRESENIUS EXHIBIT 1058
`Page 3 of 30
`
`
`
`with sufficient light intensity and suitable lighting condi(cid:173)
`tions. The lighting may be fluorescent, incandescent, spot,
`and/or polarized. Also, a combination of light sources
`may be employed and the light source(s) may be posi(cid:173)
`tioned above, below or behind the units to be inspected.
`Magnification (2X-3X) is used by some but not all manu(cid:173)
`facturers. In general, the background consists of both
`black and white sections, permitting inspection under
`both conditions. In addition, pacing methodology is often
`utilized in order to provide an effective rate of inspection
`while maintaining acceptable quality levels. Finally, those
`factors which can affect the human component such as
`training, visual acuity, and operator fatigue are usually
`controlled.
`Besides manual inspection systems, numerous semi(cid:173)
`automated machines have been developed which also use
`people for the detection of particles (80, 82-84, 87-92).
`These systems have a significantly higher throughput
`than manual processes because they perform most of the
`mechanical manipulations normally done by humans.
`These include such operations as swirling containers, in(cid:173)
`verting samples, stopping containers, and_ removing de(cid:173)
`fects. Several people have daimed that these machines
`reduce eye strain for the operators and provide improved
`inspection quality by using significantly better imaging
`capabilities than exist for manual systems (80, 82-84, 87-
`91 ).
`Whether one uses a oompletely manual system or one of
`the semi-automated processes, the decision to accept or
`reject a container is still made by a person. Therefore, it is
`important to review what is known about the human visual
`inspection process. From the broad range of literature on
`the subject (80-106), the articles published by Knapp and
`coworkers (80, 93-97) stand out as key references.
`The USP specifications for visible particles suggest that
`human visual inspection is a deterministic process. For a
`deterministic process, if the same set of containers is ex(cid:173)
`amined under the same inspection conditions several
`times, then the same containers will always be rejected.
`The rejection probability can be only one of two values, 0
`for good and I for bad containers. In contrast, for a
`probabilistic process, each container has a rejection prob(cid:173)
`ability associated with it, and the rejection probability can
`be any value between O and 1.
`Knapp and Kushner carried out some experiments to
`determine if human visual inspection is deterministic or
`probabilistic (80). In their studies a set of 1000 uninspect(cid:173)
`ed vials was examined by each of five inspectors ten times
`each for a total of fifty inspections. Rejection records were
`maintained for each vial; any rejection score from O to 50
`was possible. A summary of these results, shown in Table
`IV, indicates that containers were found in every rejection
`probability group. Only 2 samples were rejected all the
`time and approximately 20% of the containers were reject(cid:173)
`ed at least 10% of the time. These experiments confirmed
`that the inspection process is probabilistic.
`In addition, Knapp and coworkers found there is a
`relationship between rejection probability and the size of
`the particle (80, 96-97). They observed that the vials with
`the smallest particulate matter were in the lowest rejec-
`
`TABLE IV. Results of Knapp and Kushner Experiments
`(Data Taken from Ref. 80)
`Rejection Probai.ill,y
`
`Number of Vials
`
`0.0
`0.1
`0.2
`0.3
`0.4
`o.s
`0.6
`0.7
`0.8
`0.9
`1.0
`
`805
`98
`33
`17
`11
`10
`8
`6
`5
`5
`2
`
`tion probability groups and the samples with the largest
`partides were in the highest rejection probability groups.
`They concluded that the larger the panicle (all else con(cid:173)
`stant) the more certain its detection. Therefore, it is no
`longer adequate to state that particles have been observed
`in a parenteral, but the probability with which they can be
`detected is also essential information (80, 93-97).
`The findings summarized above are consistent with the
`biopbysicalliterature on human vision (96, 107-109). An
`objective description of visual inspection contains several
`essential elements, including the capability of the viewer,
`the size of the target, the total background illumination,
`and the contrast of the target against its background.
`The concept of rejection probability zones, introduced
`by Knapp and coworkers, is very useful for assessing visu(cid:173)
`al inspection systems (80, 93-97). The range of the rejec(cid:173)
`tion probability, p, can be conveniently divided into three
`regions:
`
`0.0 ::i p < 0.3
`Accept Zone
`0.3 :s; p < 0. 7
`Gray Zone
`Reject Zone
`O. 7 ~ p :S 1.0
`The region of low rejection probability, to which most
`containers in a well-controlled process will belong, is
`termed the "Accept Zone." The region of moderate rejec(cid:173)
`tion probability, the .. Gray Zone," is most sensitive to any
`changes in the visual inspection process. This zone is a
`buff er region between the truly bad containers, which
`should be rejected, and good containers that should be
`accepted. The remaining region of high rejection proba(cid:173)
`bility is termed the .. Reject Zone." This group of samples
`is especially interesting from a quality assurance stand(cid:173)
`point and the inspection process should be very efficient in
`rejecting these containers.
`A more thorough understanding of this subject can be
`obtained by reviewing the references cited above. Howev(cid:173)
`er, for the purposes of this article it will be sufficient to
`discuss several general observations concerning human
`visual inspection.
`First, one of the most important characteristics of any
`visual inspection system is its detection limit. Several
`workers have reported that particles larger than 50 µmare
`usually detected by the naked eye (87, 110-113). Al(cid:173)
`though these claims are not necessarily inaccurate, they
`
`Vol. 40, No. 5/September-October 1986
`
`%15
`
`FRESENIUS EXHIBIT 1058
`Page 4 of 30
`
`
`
`can be very misleading if they are applied without qualifi(cid:173)
`cation to human visual inspection processes.
`In any analytical measurement there are several factors
`(sample matrix, experimental conditions, signal/ noise,
`etc.) which must be specified in order to determine the
`detection limit ( 114). For example, it is usually signifi(cid:173)
`cantly more difficult to measure low levels of a species in a
`solution containing many components than it is in a medi(cid:173)
`um with only a few species. The actual experimental con(cid:173)
`ditions are defined because methods of concentrating
`samples, time-averaging, and other procedures can dra(cid:173)
`matically affect the detection limit. Finally, the ratio of
`the measured signal to the response in the absence of the
`species of interest is specified at the limit of detection.
`An analogous situation exists for a human visual in(cid:173)
`spection process. The set of samples used to determine the
`detection limit should be well-characteri:i:ed. In particu(cid:173)
`lar, one needs to specify the fraction of that set which is
`defective, the solution volume, and the nature of the de(cid:173)
`fe<:ts, including the number of particles per container,
`their size, their shape, and their reflectivity. Also, the
`conditions under which the inspections are carried out
`should be described. The inspection rate, the amount of
`magnification, the visual acuity of the inspectors, and the
`type of illumination and background that are used can
`have a significant effect on the detection of particulate
`matter. Finally, the rej~tion probability at the detection
`limit should be defined.
`The work of Knapp and c<iworkers provides useful in(cid:173)
`formation concerning detection limits for human visual
`inspection (97). The ampoules that were impected were
`thoroughly characterized using a nondestructive tech(cid:173)
`nique, transmission holography. The inspection condi•
`tions were also weU·defined. The Schering standard 10·
`sec paced inspection (two am!)Oules) with a 3X magnify(cid:173)
`ing lens, a diffuse light source, and a white/black
`background were utilized. The light intensity at the posi(cid:173)
`tion of the samples was approximately 225 foot-candles.
`In addition, the inspectors chosen for the study were se(cid:173)
`lected on the basis of measurements of their visual acuity.
`and the results of 70 inspeclions provided an accurate
`estimate of the rejection probability for each of the am(cid:173)
`ponles. In these studies a 70% deteclion probability was
`obtained for a spherical particle with a diameter of 65 µm.
`The equivalent rejection probability using the same condi•
`tions without magnification would be seen for a spherical
`particle approximately 100 ,,min diameter.
`Using a slightly different protocol than that reported by
`Knapp et al. (97), we have also studied the visual inspec•
`tion pro4-ess. In our experiments a set of 1000 ten-ml
`ampoules having the composition shown in Table V was
`used. The defectives were randomly distributed and all the
`ampoules had been thoroughly characterized by nonde(cid:173)
`structive techniques. The particles were fluorescent-dyed
`polystyrene divinylbenzene spheres and the sizes of these
`beads were measured in-situ using an inverted microscope
`procedure. Each inspector examined the entire group
`without magnification in an inspection booth with typical
`lighting and background conditions. Paced inspection was
`utilized with a clip of l O ampoules being examined every
`
`TABLE V. Ampoule Particulate Set
`Number of
`Number of
`AmpoultS
`Particles/ Ampoule
`
`50
`'75
`875
`
`Size of
`Particles (I.cm)
`
`165
`100
`
`TABLE VI. Average Results for 14 Inspectors at One Facility
`Mean Rejection
`Probability (%)
`
`Category
`
`Gooo
`One 100-µm particle
`per 10-ml ampoule
`One 165-µm particle
`per I 0-ml ampoule
`
`1.1
`59
`
`82
`
`38 sec. The people chosen for the studies included both
`quality assurance and production inspectors at a few of
`our manufacturing sites. The results of one study with 14
`inspectors at one fa.cility is shown in Table VI. Based on
`these data, the 70% rejection probability would occur for a
`spherical particle with a diameter between 100 and 165
`µm. In view of the differences in inspection rates, magnifi(cid:173)
`eation and other conditions, these results arc comparable
`with the findings of Knapp and coworkers (97).
`Most of the above discussion has assumed that there is
`one visible nonreflccting particle per container. As expect(cid:173)
`ed, for the same type and size of particle the dete<;tion
`probability increases as the number of particles increases.
`Also, the human rejection probability is strongly affected
`by the optical charact.elistics of the particulate matter
`(97).
`A second important characteristic of a visual inspection
`system is its reproducibility (84, 93, 94, 97, 105). If the
`same set of samples is examined several times by several
`people under identical inspection conditions, one would
`like to know the consistency of both the rejection rate and
`the defectives for an individual inspector as well as for the
`entire group of inspectors. Moreover, it would be desirable
`to have this information as a function oftime. Although a
`number of articles have been published on this subject (84,
`93, 94, 97, 106). it is difficult to summarize the observa(cid:173)
`tions. In particular, this topic is similar to the previous
`subject since an informative discussion cannot be given
`without defining the specific range of rejection probabili(cid:173)
`ties for the samples of interest.
`In general, the performance of a human visual inspec(cid:173)
`tion system is only moderately reproducible. There is a
`wide variability in the capabilities of individual inspectors
`and the performance of each inspector can change signifi(cid:173)
`cantly over the course of time. For example, when the set
`of ampoules described in Table V was examined by in(cid:173)
`spectors at one facility, the results in Table VII were
`observed. Among these inspectors the rejection probabili(cid:173)
`ty varied from 19 to 84% for a 100.µ.m sized particle and
`from 64 to 96% for a 165-µm sized particle. Similar results
`
`216
`
`Journal of Parenteral Science & Technology
`
`FRESENIUS EXHIBIT 1058
`Page 5 of 30
`
`
`
`TABLE VII. Rej«:tion Probability
`
`lnspeet•r
`
`Reject Rate(%)
`
`Good(%)
`
`One JOO-pm
`Particle/ A11poule (%)
`
`One 165-,im
`Particle/ Alllp&ule (%)
`
`I
`2
`3
`4
`5
`6
`7
`8
`9
`10
`11
`12
`13
`14
`
`10.3
`8.0
`6.9
`8.4
`9.9
`8.4
`6.0
`9.4
`12.l
`10.0
`12.1
`12.1
`8.3
`8.7
`
`0.6
`2.1
`1.3
`0.1
`LS
`0.2
`0.3
`0.8
`2.5
`1.4
`1.4
`1.7
`0.2
`0.8
`
`71
`40
`19
`53
`60
`58
`35
`.59
`81
`68
`84
`80
`58
`56
`
`94
`66
`88
`90
`84
`82
`64
`86
`78
`78
`96
`94
`78
`76
`
`have been seen by Knapp et al. (93) when they studied the
`inspector variability in 10 inspections using a reject-seed(cid:173)
`ed test batch. They found that among 23 inspectors the
`Reject Zone Efficiency varied from 57 to 99% and the
`reject rate varied from 13.7 to 49.3%. Also, the variation
`in the performance of individual inspectors over time can
`be seen by the data shown in Table VIII. For this study
`several inspectors at a different site examined the set
`described in Table V two or three times during a 24-month
`period.
`Finally, human visual inspection systems usually have a
`small, but detectable false reject rate. Several factors can
`lead to the rejection of good containers, i.e., those which
`do not contain any particles of sufficient size to be detect(cid:173)
`ed by an inspector. A very small fraction of the containers
`in a batch are removed by people because the inspectors
`occasionally interpret an air bubble as a pa,ticle {88).
`Also, when several samples are examined at the same time
`and particulate matter is observed in one of the containe,s,
`human error could re.suit in the removal of a good sample
`instead of the defoctive (88). In practice, these containers
`arc included in the group of samples having a very low
`rejection probability (80, 93-97). Nevertheless, because
`all measurement techniques have noise, it is probably
`
`more a~urate to classify them as good containers which
`are rejected as a consequence of noise in the inspection
`system. The results in Tables VII and VIII for the good
`samples suggests that the false reject rate for hum.an
`visual inspection systems is approximately 0.5-1.0%.
`Machine Visual Inspection: An alternative to hum.an
`visual methods is machine inspection. Several machines,
`using a variety of particle detection methods, are commer(cid:173)
`cially available (93, 94, 96, 97, 99- 102, 105, 106).
`Although the machines are based on different princi(cid:173)
`ples, there are several common features which all of the
`systems have. First, the containers are spun at a high rate
`of speed and the movement of the container is stopped just
`prior to the time the sample is viewed by the detector. The
`primary purpose of this step is to place the particulate
`matter in motion. On most systems the spin speed can be
`varied (93, 94, 99-101). In the case of the Eisai system,
`this parameter can be varied from 800 to 3500 rpm (Table
`IX).
`Second, the interval between the deceleration of the
`container and the observation time is usually very small.
`This is done in order to achieve reliable detection of heavy
`particles such as glass fragments (93). On the Eisai sys(cid:173)
`tema brake is used to stop the containers, and the location
`
`TABLE VIII. Rejection Probability (%)"
`
`Impector
`
`Omo.
`
`Good
`llmo.
`
`24mo.
`
`Omo.
`
`One 100-pm
`Partlcle{.Q1J1tainer
`llmo.
`24mo.
`
`One 1'5-1tm
`PardcleL Contaiaer
`24uo.
`1l me>.
`
`Omo.
`
`0.2
`
`0.2
`0.2
`0.4
`1.8
`0.0
`
`0.1
`3.4
`0.0
`0.2
`0.1
`
`25
`
`27
`21
`31
`30
`48
`
`I
`33
`0.3
`2
`1.3
`13
`41
`3
`0.1
`0.0
`4
`26
`26
`5
`0.0
`5