Maik W. Jornitz

Integrity testing multiround filter housings is difficult and inaccurate because of  the multitude of filter elements involved. The diffusive flow sum of all of the filter elements can make it difficult to detect a single flawed filter or determine the bubble point which is bulk flow point. 

The single-point diffusion test reflects the overall diffusion rate of the cartridges being tested, thus increasing the risk for a minor failure (e.g., one single filter element with a loss in integrity) being potentially masked by the overall low diffusion values of the surrounding cartridges. A statistical approach with a reduced maximum allowable diffusion test limit would reduce this potential risk. 

The bubble point will reflect the biggest pore-size area. Its accuracy depends on the capacity of the integrity test unit to detect the over-proportional increase of gas flow when passing from the diffusive flow region to the bulk flow region. For a large filtration area this transition phase potentially will be masked because of the overall high diffusion rate. In addition, the bulk flow through possible flaws does not increase exponentially as the test pressure increases, thus making bubble-point determination difficult. 

The multipoint diffusion test will create a diffusive flow profile (diffusion versus applied pressure) of the filter system over a broad range of test pressures all the way up to the characteristic bulk flow of the bubble point. The combination of multipoint diffusion test and bubble point leverages the advantages of the combination and increases a higher accuracy of detection. 

Some automated multipoint diffusion-test systems allow programming of a specific maximum allowable diffusion value for every pressure step, thus setting a maximum diffusion profile. 

The following materials were used: automated integrity-test unit (Sartocheck 4, 

); five pieces of 30-in. PESU 0.2-µm cartridges (Sartopore 2, Sartorius Stedim Biotech GmbH), referred to as A, B, C, D, and E, respectively; one piece of 20-in. PESU 0.2-µm cartridge (Sartopore 2), referred to as F; one piece of 10-in. PESU 0.2-µm cartridge (Sartopore 2) failing cartridge (low bubble point), referred to as G; one piece of 10-in. PESU 0.2-µm cartridge (Sartopore 2) failing cartridge (high diffusion), referred to as H; one multiround (5 x 30 in.) housing; one single-round 30-in. housing; and one single-round 10-in. housing. 

All filter cartridges were individually tested three times with a multipoint diffusion test over a range of pressure points. The lowest point was at 2500 mbar (36.25 psi), which is the traditional single-point diffusion test pressure for the filter used. The highest point was 3250 mbar (47.13 psi), which is 50 mbar (0.73 psi) above the specified minimum allowable bubble point of the filter tested. The filter cartridges were thoroughly rinsed with demineralized water at room temperature before each integrity test. 

After the diffusion profile curves (from the multipoint diffusion test) and the individual bubble point values were generated, the data were analyzed. The individual cartridge values were used to mathematically simulate different filter combinations in a multiround housing (see Figure 1). 

 Figure 1: Individual values of cartridges A, B, C, D, F, and H and their sum. (ALL FIGURES ARE COURTESY OF THE AUTHORS.)  

To verify the mathematically simulated test results, the 30-in. filter elements (i.e., A, B, C, D, and E) with no break in integrity were installed into the multiround housing and tested with the bubble point and multipoint diffusion tests. All cartridges were tested three times and rinsed between each test (see Figure 2). 

Figure 2: Multiround housing 5 x 30-in. cartridges A, B, C, D, and E.  (ALL FIGURES ARE COURTESY OF THE AUTHORS.)  

One 30-in. cartridge (cartridge E) was then replaced by a 20-in. cartridge. A 10-in. filter housing, containing the bubble-point failing cartridge, was connected in parallel with the multiround housing, consequently representing a five-round 30-in. setup in terms of membrane surface. The system containing cartridges A, B, C, D, F, and G was tested with bubble-point and multipoint diffusion tests. All cartridges were rinsed between each of the three tests (see Figure 3). 

Figure 3: Multiround housing and 10-in. housing, 4 x 30-in. cartridges A, B, C, and D; 1 x 20-in. cartridge F; and 1 x 10-in. cartridge G or H. ( (ALL FIGURES ARE COURTESY OF THE AUTHORS.)  

The bubble-point failing cartridge (cartridge G) in the 10-in. housing was then replaced by the high-diffusion cartridge (cartridge H), and the same tests and rinses were performed. Instead of performing a separate single-point diffusion test at 2500 mbar (36.25 psi), the authors considered the diffusion value at the same pressure during the multipoint diffusion test was representative. 

The expected difference in diffusion between five 30-in. cartridges (i.e., A, B, C, D, and E) (137.9 mL/min) without loss of integrity and an equivalent setup including the bubble-point failing 10-in. cartridge (i.e., A, B, C, D, F, and G) (139.4 mL/min), at the standard test pressure of 2500 mbar (36.25 psi) was <2 mL/min, thus being not possible to detect with a high degree of confidence. 

An equivalent setup including the failing 10-in. high-diffusion cartridge (i.e., A, B, C, D, F, and H) would require the maximum allowable diffusion value to be lowered by more than 40% as compared with the cumulated individual maximum diffusion values to be detected at all. Using such a high safety margin creates a high potential risk of getting false-failure test results with filters that otherwise have no break of integrity.  

Based on the typical diffusion profile of a cartridge with no loss of integrity, it was possible to define a maximum allowable diffusion curve, which made it possible to get a clear multipoint-diffusion test failure when the bubble-point failing 10-in. cartridge was included in a multiround setup (cartridges A, B, C, D, F, and G). The same approach showed that it should be possible to get a clear test failure when having the 10-in. high-diffusion cartridge in a multiround set-up (cartridges A, B, C, D, F, and H). 

To simulate the expected bubble-point value of various multiround setups the automated integrity test bubble-point test algorithm was programmed in an Excel (Microsoft) spreadsheet, and the diffusion profiles of the individual cartridges were entered. The authors found that it should be possible to get a clear bubble-point test failure involving the 10-in. bubble-point failing cartridge in a multiround setup (cartridges A, B, C, D, F, and G). In addition, the authors found that it should not be possible to get a bubble-point test failure when including the failing 10-in. high-diffusion cartridge in a multiround set-up (cartridges A, B, C, D, F, and H), because no exponential increase would be expected to take place before the minimum bubble point. 

The diffusion profile of a filter cartridge can be plotted in two distinct areas, with the applied differential pressure on the 

-axis and the resulting diffusion on the 

-axis.The first area is the so called diffusive flow region, where the diffusion rate can be considered linear as a function of the applied test pressure. In this test-pressure region, the pores are still filled with the wetting liquid. 

The second region is the so-called bulk flow region where the applied pressure gradually blows out an increasing number of pores, giving a free-gas flow through the membrane. The flow graph turns exponential. 

Between the linear and exponential area is the so-called inflection point or transition phase. The sharpness of the inflection point is directly related to pore-size distribution, pore design, and the size of the filter surface. The smaller the filter surface and the more homogeneous the pore-size distribution, the sharper the inflection point (see Figure 4). 

Figure 4: Small filter surface inflection point versus large filter surface inflection point. (ALL FIGURES ARE COURTESY OF THE AUTHORS.)  

The multipoint diffusion limit curve is based on the diffusive flow and the bulk flow and is defined by three set points. It is a straight line from the first test pressure point, which is the common single-point diffusion-test pressure (in this case 2500 mbar or 36.25 psi), up to a point at 100 mbar before the minimum bubble point (in this case 3100 mbar or 44.95 psi). The last point at 50 mbar after the minimum bubble point (in this case 3250 mbar or 47.15 psi). This much higher maximum allowable diffusion rate allows one cartridge to reach the bubble point at the minimum bubble point without giving a false failed test result (see Figure 5). 

Figure 5: Multipoint diffusion test limit curve. (ALL FIGURES ARE COURTESY OF THE AUTHORS.)  

The maximum allowable diffusion value for the two first pressure steps is calculated taking into account the number of cartridges, their individual maximum allowable diffusion value, the applied test pressure, and a safety margin. The maximum allowable diffusion value for the last test-pressure point is calculated taking into account the same variables as the two previous set points but allowing one cartridge to reach bubble point. This is achieved by adding a cartridge size-related diffusion value representing the over-proportional increase generated by the bulk flow at the bubble point. In the integrity-test unit bubble point test algorithm, this value is called the 

2-value (c.f. Functional Design Specification of Sartocheck 3, Sartocheck 3+, Sartocheck 4). The set point equations thus become: 

 is the number of cartridges in the multiround housing, 

minus the cartridge eventually reaching the bubble point, Diff

 is the validated maximum allowable diffusion value for the concerned cartridge type (here 54 mL/min), α = (1 – safety margin), 

1 is the standard diffusion test pressure (here 2500 mbar or 36.25 psi), 

2 is the second set point (here 3100 mbar or 44.95 psi), 

3 is the third and last set point (here 3250 mbar or 47.15 psi), and 

2 is the cartridge size–related diffusion value (here 240 mL/min). 

During out-of-box trials using new 30-in. Sartopore 2 filter cartridges, the typical diffusion value was only slightly higher than half of the maximum allowable diffusion value given by the manufacturer. During the simulation trials, a maximum allowable diffusive flow limit one-third lower than the standard manufacturer's limit created a robust test result without causing false failures. The a-value was thus set to two-thirds, and the programmed multipoint diffusion limit line became: 

  5 x 54 x (2/3) x (3100/2500) = 223.2 mL/min  

(5-1) x 54 x (2/3) x (3250/2500) + 240 = 427.2 mL/min 

The trials showed that the traditional single-point diffusion test was not accurate enough for multiround housing integrity testing. Subtracting more than 40% from the cumulative individual maximum diffusion values would have detected the failing high-diffusion cartridge (cartridge H), but it also would have increased the risk for getting false-failure test results. Single-point diffusion testing also was incapable of detecting the bubble-point failing cartridge even when the maximum value was reduced by almost 50%. 

Figure 6: Multipoint diffusion test on cartridges A, B, C, D, and E. (ALL FIGURES ARE COURTESY OF THE AUTHORS.)  

The bubble-point test, however, was capable of detecting the bubble-point failing cartridge. If the multiround setup contained the high-diffusion cartridge as the only failing cartridge, the bubble point was incapable of detecting this flawed filter because the high diffusion cartridge did not have a clearly defined bubble-point value below the minimum bubble point. The difference between the expected bubble-point value and the measured bubble point was ≤50 mbar (0.73 psi). 

Figure 7: Multipoint diffusion test on cartridges A, B, C, D, F, and G. (ALL FIGURES ARE COURTESY OF THE AUTHORS.)  

The multipoint diffusion test was capable of detecting the flawed high diffusive flow filter element respectively low bubble-point filter element. The difference between expected diffusion values and measured diffusion values on the multiround housing was less than ±5% (see Figures 6–8). 

Figure 8: Multipoint diffusion test on cartridges A,B, C, D, F, and H. (ALL FIGURES ARE COURTESY OF THE AUTHORS.)  

Articles

The authors describe a novel approach for the integrity testing of large sterile filter systems such as multiround housings and describe a multipoint diffusion test capable of detecting minor failures.

Improving the Integrity Test Assurance of Multiround Housings Assessments

A septically produced liquid pharmaceutical and biopharmaceutical products are usually sterilized by filtration. The filtration process must be validated to ensure that it is capable of removing all microorganisms from the product. Validation consists of challenging the filter with a suspension of 

 and analyzing the filtrate for microorganisms. The filtrate must be sterile. 

Filters used in production should be equivalent to the filters used in the bacterial-challenge validation studies. Because actual production filters cannot undergo bacterial-challenge testing, integrity testing is performed to demonstrate bacterial-retention equivalence. If the integrity-test values obtained for the production filters are equivalent to those obtained for the filters successfully passing bacterial-challenge testing, then it is assumed that the filters have the same bacterial-retention properties and that the filtered pharmaceutical product is, therefore, sterile. 

Integrity testing for the hydrophilic filters used in pharmaceutical production relies on the measurement of gas flow through wetted membranes. This flow can be classified as diffusive or bulk and is sometimes a combination of both. Fick's Law of Diffusion shows that diffusion of the test gas through the liquid-filled pores in the membrane is a function of the diffusion constant and the solubility of the test gas in the liquid at the test temperature, the pressure differential of the test gas across the membrane, the thickness of the liquid layer, and the area and porosity of the membrane (1). Diffusion is not directly related to pore size although, as will be shown later, there is an indirect correlation. Bulk flow occurs when the test gas flows through the nonwetted or empty pores of the membrane. Open pores occur because the filter membrane has been incompletely wetted or because the bubble point of the membrane has been exceeded. Bulk flow primarily is a function of the size and number of the open pores, the thickness of the membrane, and the pressure differential of the test gas across the membrane at the test temperature. 

As indicated previously, gas flow through wetted membranes is diffusive, bulk, or a combination of both. Figure 1 shows the relationship between gas flow and differential pressure for a typical membrane filter. Similarly, because it is an arithmetic function, pure bulk flow is a straight line whose slope is determined by the size and number of open pores, assuming all other variables are held constant.  

The knee area of the curve is where the influence of the bubble point is manifested. Here, the largest pores of the filter become unblocked as the applied pressure overcomes the capillary forces within those pores, and bulk flow begins to increase the slope of the curve. It is also in this region that diffusive flow begins to increase because the thickness of the liquid trapped in the largest pores begins to decrease as a result of the increasing pressure differential. Therefore, within the knee area of the curve a complex relationship exists between diffusive and bulk flow, both influenced by the pore structure and pore-size distribution. 

Validating the bacterial-retention capabilities of sterilizing-grade filters involves challenging the filters with a suspension of 

 (ATCC 19146) at a level of at least 1 × 10

  of filter area, resulting in a sterile filtrate. Specific details and options may be found in "PDA Technical Report No. 26

(2). The integrity-test values of the filters used in the bacterial challenge studies must be known to form the basis for ensuring the filters used in production have equivalent bacterial-retention capabilities. In addition to the integrity-test values, the manufacturing processes and quality-assurance systems of the filter manufacturer must be adequate to ensure consistency of the filters within and between each manufactured lot with respect to nominal pore size, pore-size distribution, and membrane thickness and area.  

Bacterial retention is assessed by challenging filters of the intended type having a range of integrity-test values and analyzing the bacterial-passage results. Ideally, the retention pattern observed will allow identification of physical integrity-test values above or below which, depending on the type of integrity test, there is virtually no probability of bacterial passage. For example, with a series of filters having a range of bubble points and equivalently challenged, there would be a value below which bacterial passage always occurred, a range where passage sometimes occurred, and a value above which passage never occurred.  

It is important to identify the integrity values above and below which bacterial retention and bacterial passage can be expected. This ensures that the correlation between the integrity-test value and bacterial retention is valid and there are not other factors such as pore-size distribution and anisotropy that are potentially influencing the perceived correlation. 

During development of what he referred to as the "forward flow" test in 1973, Dr. Pall performed bacterial-challenge testing on a series of 43 cartridges having water-wet forward air flow values ranging from 16 to 900 cm

/h at a pressure differential of 10 in of mercury (4.9 psi) (3). The results of this testing showed that no filter with a flow of less than 199 cm

/h allowed bacterial passage. On the basis of those results, he established a maximum flow limit of 100 cm

/h for an acceptable production filter element, building in a safety factor. However, the testing also showed that there were filters with forward-flow values of 210, 258, 450, and 645 cm

/h that retained the challenge organism and filters with forward-flow values of 199, 238, 471, and 600 cm

/h that allowed bacterial passage. The study did not establish a point above which bacterial passage always occurred; therefore the correlation between the integrity-test value and bacterial retention was not fully elucidated.  

Because the forward-flow test as described relies on diffusive flow, which in terms of the filter membrane is a function of porosity, area, and thickness and is not directly related to pore size, the bacterial-retention results obtained for the higher forward-flow values could be indicative of pore-size distribution differences (or possibly thickness variations) among the tested membranes. Had these tests been carried out at pressures closer to the bubble point as is done today, differences in pore-size distribution might have been revealed as a result of the pressure-induced thinning of the liquid layer in the larger pores and better correlation obtained with bacterial retention at the higher diffusive-flow values. Nonetheless, Pall's efforts served as the basis for developing the automated filter-integrity tests in use today. These tests operate at considerably higher pressures, capturing the influence of the knee area and providing effective correlation between bacterial retention and the integrity-test value. 

It is important to note that the integrity-test values were obtained on clean filters, before commencing the bacterial-retention tests, eliminating the influences of filter loading on filter porosity and diffusive-flow values. 

Another validation consideration is to test the filters after sterilization and bacterial-challenge testing to evaluate the influences of filter loading and the sterilization process on the integrity-test values. 

The sterilizing filtration process is designed to remove viable and nonviable particles from the fluid passing through the filter. The particles removed from the fluid either remain on the surface of the filter membrane or are trapped within the membrane matrix. 

It has been shown that hydrodynamic flow is preferentially directed to the larger pores in a filter membrane (4). This means that hydrodynamic flow directs the suspended particles to the largest pores. However, because the largest pores are far less numerous than the average-sized pores, particles initially tend to be captured by the many smaller pores. As the smaller pores become blocked, the hydrodynamic flow directs more of the particles to the larger pores. The larger pores now may become blocked or partially blocked because of bridging, which is promoted by high levels of particulates competing for the pore openings. This phenomenon is corroborated by Grant and Zahka (5). They showed that during the filtration of a dilute suspension of silica particles there was complete retention at the beginning of the filtration, followed by particle passage. The retention then increased as the filtration continued. Similar results were obtained by Roberts and Velazquez while filtering latex particles (6). Thomas 

 found that the successive changes affecting a filter's pore-size distribution as a filtration of 

progressed showed the larger pores to be less immediately engaged in the retention of particles (7). 

Effect of filter loading on integrity-test results

The effects of filter loading on diffusive flow and the bubble point have been evaluated; however, comprehensive studies remain to be performed, and understanding of these effects is currently incomplete. Trotter 

 showed that the diffusive-flow values at 45 psi began to change at a bacterial cell loading of 1 × 10

, while the bubble point remained essentially unchanged until the bacterial cell loading approached 1 x 10

Particle capture inherently reduces the porosity of the filter, thus decreasing the diffusive flow, and it increases the bubble point as the largest pores become blocked. Also, the effective thickness of the membrane increases as material accretes on its surface. These effects have implications with respect to the integrity-test values and their correlation to bacterial retention. Certainly, porosity reduction and the increase in thickness depend on the feed stream and the properties of the particles (gel or solid) removed from it. 

Filter loading influences diffusive-flow and bubble-point values, although not to the same degree and not in the same sequence according to published studies. Postfiltration integrity testing evaluates filters that have been subject to an unknown level of particulate loading; therefore the influence of the particulate loading on the integrity-test results is also unknown. It is reasonable to assume, however, that as particles are arrested by the membrane, the porosity of the membrane decreases and the bubble point increases. This assumption has been confirmed experimentally (8). Testing under different conditions or with different membrane types and configurations could produce different results, however. 

Extensive validation testing using multiple lots of product representative of normal production is necessary to determine the potential effects of filter loading on the postfiltration integrity-test values. In any case, the precise effect of the filter loading on the postfiltration integrity test cannot be known with certainty; therefore, it is not possible to ensure that filter loading has not masked a defect that initially may have been present in the filter. 

Regulatory agencies differ in their positions regarding the need for prefiltration integrity testing of membrane filters used to sterilize liquid pharmaceutical and biopharmaceutical products. Annex 1 of the 

 specifies integrity testing of the sterilized filter before and after use. FDA's 2004 

Guidance for Industry, Sterile Drug Products Produced by Aseptic Processing

 suggests that integrity testing can be performed before processing and that integrity testing should be routinely performed post-use to detect leaks or defects that may have occurred during the filtration process. The FDA guidance does not take into account potential changes in the validated prefiltration integrity-test values as a result of filter loading. 

Because the validation studies correlating bacterial retention with integrity-test results use prefiltration integrity-test values, albeit with a safety factor, the only reliable way to ensure that the filters used in production are equivalent to those used to establish the integrity-test specification is to perform a prefiltration integrity test. The postfiltration integrity test then serves its intended function: to ensure the filter has not been damaged during the sterilization or filtration process.  

 is president of the Williamsburg Group, LLC, 18907 Lindenhouse Rd., Gaithersburg, MD 20879, tel. 301.938.4266, 

 is group vice-president of global product management, bioprocess of Sartorius Biotech Inc. 

 is a consultant for Capitola Consulting Company. Madsen and Meltzer are also members of 

*To whom all correspondence should be addressed. 

Keywords: filtration, integrity testing. sterilization 

 1. F. Hofmann,"Integrity Testing of Microfiltration Membranes." 

 "PDA Technical Report No. 26: Sterilizing Filtration of Liquids," 

 3. D.B. Pall, "Quality Control of Absolute Bacteria Removal Filters," 

, "Integrity Test Measurements—The Effects of Bacterial Cell Loading," 

 5. D.C. Grant. and J. G. Zahka, "Sieving Capture of Particles by Microporous Membrane Filters from Clean Liquids," 

 6. K.L. Roberts and D.J. Velazquez, "Characterizing the Rating and Performance of Membrane Filters for Liqid Applications Using Latex Spheres," 

 during Microbiological Validation of Filters." 

 "Investigation of a Filter Structure by Microbial Retention Studies: A Synthesis and Elaboration of Prior Findings," 

Pre-use integrity testing of sterilizing-grade filters eliminates the potential adverse effects of filter loading on the integrity-test results, allowing unambiguous correlation with the integrity-test specification established during filter-validation studies.

The Importance of Pre-Use Integrity Testing in Sterilizing Filtration

Microporous membrane filtration is the technique often applied to the aseptic processing of drug preparations. This is especially appropriate when the ingredients are heat labile. Certain filter performance qualities are of specific interest in such filtrations, namely, the extent of organism removal, the rate of liquid flow, and the total throughput volume. Essentially, it is the numbers and sizes of the pores relative to the number and sizes of suspended particles that determine the filter retention performance, although there are other factors that also affect organism removals (1). The total aggregate space of the pores within the solid filter matrix represents the membrane's porosity. It is constituted of the total number of pores of whatever dimensions. The importance of pore size is obvious. The filter pores are the sites through which the liquid flows. Simultaneously, its suspended particles, such as organisms, are retained at the pores. Ultimately, the throughput is limited by the accumulation of the particles at the pores causing their blockage.  

Microporous membranes prepared by the reverse-phase casting process are available from several manufacturers in a series of pore sizes ranging from 0.04 to 8 μm. Curiously, the rating values are not determined by direct pore-size measurements, although several different methods for sizing have been proposed.  

Tests such as bubble-point measurements and bacteria challenges are used to assign pore-size ratings. The propriety of translating bubble-point values into pore-size ratings will be discussed later in this article. The assigned numbers are meant to imply particles-size retentions, not dimensional mensurations from which flow properties might also be derived. The bubble points are indicators of the largest size pores present in a membrane; the focus in filtrations being chiefly upon particle retentions. 

The significance of pore sizes lies in their implication to particle-size retentions. The quantitative characterization of pore sizes derives from bubble-point measurements coupled with bacterial challenges. These tests are correlative. The former does not compromise subsequent use of the filter. It is a nondestructive test. The latter, although more direct in its diagnosis, is a destructive test in that it contaminates the filter with the test organisms; thus obviating its later application to process filtration. 

The organisms used in the microbial challenges differ in accordance with the filter's presumed pore size. The log reduction values necessitated by FDA's definition of a sterilizing filter is the retention of  1 × 10

  colony forming units (cfu) per square centimeter of effective filter area (EFA) (2). Such a retention produced against a 

 ATCC-19146 confrontation characterizes the 0.2/0.22 μm-rated membranes. 

 is the organism commonly used in challenging 0.45 μm-rated membranes. Because Bowman 

 determined that the 0.45 μm-rated membranes retain 

  level, that test is sometimes used for testing the 0.45-μm membranes (3). The 0.1 μm-rated pore size is mostly tested against 

 It is fair to say that the testing of  0.1 μm-rated membranes is not yet fully in place within the industry. 

Intermittent bubble-point measurement also is used as a process guide during the membrane-casting operation to ensure that the intended pore-size ratings result. The point being made is that the relevant measurements being performed are of bubble points, not of pore sizes. From these, as will be discussed, pore-size ratings and log reduction values (LRVs) are derived. 

The pore-size ratings are the remnants of early efforts at membrane classifications that developments in pharmaceutical filtrations have reduced to rather insignificant utility. Pore-size ratings are administrative devices. They serve only as parts numbers or catalogue listings for filter manufacturers. The selection of filters for sterilizing applications is made primarily on the basis of the LRVs that the various filter types impose upon 

 challenges in accord with FDA's definition of a sterilizing filter (2). The relationship of LRVs to experimentally determined bubble-point values enables use of the latter to identify potential sterilizing membranes subject to verification by microbiological assays performed by the filter manufacturer. The putative pore sizes are affixed accordingly. The membranes that meet FDA's stipulation of retaining the 1 ×10

  challenge are listed as being rated as 0.2/0.22 μm. 

Microporous membranes can be prepared by various methods. Among these is the track-etch process wherein a polymeric, dielectric film is bombarded with heavy-mass fission fragments, followed by an alkali etch of the radiation-damaged pathway. This process creates pores that are straight through and regularly cylindrical in shape. Their diameters can be accurately measured with scanning electron microscopy. Despite the relative regularity of their pore structures, these membranes do not usually find application in pharmaceutical processing. The membranes used in pharmaceutical processes are almost exclusively prepared by the phase-inversion technique, generally referred to as the "casting method" (4). 

Little is known about the numbers, sizes, and shapes of the pores of microporous membrane so prepared. The membrane structure usually is pictured as being analogous to that of a polymeric sponge. A hypothesized oversimplification of the pore passageways is that of irregular and tortuous capillaries that are, therefore, more extended in length than the filter's surface-to-surface thickness. The pores are marked by irregularly restricted diameters that provide the choke-points that interfere with particle passage. However complex, the pores are pictured as being essentially cylindrical and composed of interconnected spaces extending through the depth of the polymer matrix.  

The very concept of a definable "pore" is an artificiality when applied to microporous membranes other than the straight-through columnar pores that characterize the track-etched variety. The complex geometry of the sponge-like membrane results in the pores having ratios of cross-sectional areas to perimeters, called the "hydraulic parameters." These vary over the entire thickness of the membrane (5). A membrane's depth can be constructed of several superimposed unit planes that in their aggregate impose their effect on retention and flow rates (6). The "pores" so considered are presumably connected throughout the unit planes to constitute pathways for fluid flows. However, where particle retentions interfere flow redistributions may result through new "pore" alignments. The "pore" concept arises as a hypothetical construct useful in understanding filter performance. Unlike the track-etched pores, they are not integral, structural pathways for fluid flow.  

The pore structure derives from a cast polymer solution wherein the polymer chain segments are separated from one another by distances that reflect the degree of dilution. It is the inter-segmental distances among the polymeric chains that in their interconnections prefigure the "pores" of the finished membrane. Formulae of various polymer concentrations give rise to different intersegmental separations, ultimately to different porosities, when by proper manipulations the polymer is precipitated as a gel, to be washed and dried to its solid, microporous membrane state. There is inevitably a pore-size distribution and some anisotropic pore shape formation (4). 

 Figure 1: Free-floating soap bubbles.			

It is hypothesized that the formation of the microporous membrane structure accords with the known phenomenon of "soap bubble clustering" (7). The reasons for this resemblance is that in both cases there is the coming together of spheres whose spatial clustering is under the influence of area-minimizing forces. The geometric consequences of these forces is known from the study of soap bubbles (8). Polygonal facets characterize the resulting spaces of a free-floating cluster of soap bubbles (see Figure 1). In the pore formation, the nonsolvent of the casting solution takes the place of the air of the bubbles. In support of such structures, Figure 2 is that of a detergent foam confined between glass plates. The polyhedral spatial structures are obvious. Figure 3 is of a reticulated polyurethane (polymeric) foam. The cellular pores can be seen, in fact, to be polygonal in shape. The phenomenon of clustering through polyhedral spatial arrangements is manifested in other settings as well. It is a trait of zeolitic molecular sieves whose interconnection is through the open panels common to contiguous polyhedra, albeit caused by crystal-packing rather than area-minimizing forces (7). There is, therefore, technical support for the concept of polyhedral microporous structures. 

Figure 2: Detergent foam between two glass plates.			

In the case of the microporous membranes, the pores, as stated, are formed from the open intersegmental areas prefigured in the casting solution. They are hypothesized to be of various polygonal shapes, framed by polymeric struts and walls, and to be, like the zeolites, interconnected by openings in their common walls. It is these openings that are seen to be the metering and retaining pores of the membranes. As stated, casting solutions of various polymer concentrations give rise to separate but rather similar intersegmental distances. It is these spaces that are ultimately transformed into the membrane pores that, in consequence, are also similar in their dimensions—except for the pore-size distribution that results from the casting solutions' deviation from an ideal homogeneity. 

Figure 3: Reticulated polyurethane foam.			

Filtration is not a simple sieving process, except perhaps in the case where the filter is a track-etched membrane with pores passing straight through a much thinner film (~10 μm). As stated however, for the membranes prepared by the casting process (thinness ~150 μm), a fluid on passing through encounters pores of different diameters and surface areas. Particles are separated from the fluid by adsorptive attachments to the pore surfaces as well as by the size exclusion mechanism. Thus, the meaning of the average pore size as reflecting the size of a restrictive diameter within a pore passageway is necessarily an oversimplification. Nevertheless, the events taking place at the pores are depicted as if the pores were continuous and integral paths through the depth of the membrane. 

Membrane characteristics are assessed because of their pertinence to aseptic processing. However, the pore-size distribution, despite being an important structural feature that influences both flow and retention, is not among them. It is seldom known or investigated although an ASTM method  based on airflow rates enables its assay (9). The reason for not assaying the pore-size distributions of filters is that complete organism removal is dependent upon the largest pores of the pore-size distribution retaining the smallest particle of the particle-size distribution. This is the singular circumstance wherein an absolute filtration can eventuate. Given this felicitous situation, only the size of the largest pore has pertinence. In fact, however, the distributions of neither the pore nor organism sizes are likely to be known by the filtration practitioner. Depending on the relative numbers of pores and particles and their sizes, particles may encounter the larger pores of the distribution to escape removal. These larger pores of a distribution are assayed by bubble-point measurement recorded in pressure units (psi or bar) rather than in dimensional units (μm). The accompanying smaller size pores are ordinarily not quantified or measured because they are not seen to influence retentions.  

 Numerous methods have been used to assess pore size (10). Early on, mercury porosimetry was the method of choice (11–13). In this procedure, mercury is forced under pressure to penetrate into membrane pores. This process is performed at increasing differential pressures, Δ

, the smaller the pores that the fluid metal can intrude. Quantitation of the different pore sizes relates to the volume of mercury that is intruded in filling the pores at each of the progressively increasing differential-pressure stages. Airflow porosimetry studies also have been performed (14, 15).  

The porosimetry method is flawed by the assumptions necessary to its application. It is not suited to polymeric materials that are at temperatures above their glass-transition points and whose pores are, therefore, liable to stretching and distortion under the pressures of the intruding fluid. Also, the averaging of volume changes required by the technique may mask the true dimensions of the "pores." 

Pore sizing based on the retention of organisms of presumably known sizes has been widely investigated (14, 16–18). Pore sizing attempts also were made using latex beads of very narrow particle-size distributions (19, 20) (see Table I). These latter methods, in conjunction with the use of surfactant, have the advantage of eliminating adsorptive effects from those ascribed to sieving (21) (see Table II). Various surfactants were found to differ in their influencing the sizing results (22). The assigned pore sizes assumed the particle to be spherical and the area of pore restriction to be circular in shape. Simplifying assumptions were necessitated to finesse the shape factor that is operative in retentions. The exclusivity of sieve retentions also was assumed, ignoring adsorptive arrests, and at best allowed inadequately for pore-size distributions. Reductions in organism sizes resulting from contact with given liquid preparations, shown to be possible by Sundaram 

 (23) were not investigated. Nor have the possible plasticizing effects of the suspending liquids on membrane pore-size alterations reported by Lukaszewicz and Meltzer in 1980 been investigated (24). 

Table I: Percent retention of various size latex particles for 0.2 Î¼m-rated membranes.			

The results of these trials usually were judged to be of limited value. This is typical of efforts where simplifying assumptions are used to support a hypothesis, and the very conclusions are limited by the inherent arbitrariness of the necessary premises. They are complicated by the fact that the definitions of the particles themselves may depend on the particle-measuring methodologies, on the procedural protocols, and on the measuring devices used (16). 

Table II: Retention (%) of 0.198-Î¼m spheres by various 0.2 Î¼m-rated membranes.			

 The engineering principle enabling an orifice's dimensions to be sized using the rates of flow it permits can be applied to determining pore sizes. The rate of flow through a pore at a given differential pressure is essentially inversely proportional to its length and directly so to its width or diameter. Thus, the flow characteristics of single pores are described by the Hagen–Poiseuille Law: 

wherein a fluid of viscosity η with an average velocity 

 of the fluid is related to the tube diameter

At the site of the pore's most restricted diameter its retention and flow properties are defined. Nonetheless, flow rate is influenced by more than the pore's area of constriction. It reflects also the length of the pore. Conceivably, a relatively open pore, less constricted and, therefore, less retentive, may be rendered slower-flowing by its extended length. The area of a filter, however, contains numerous pores. Its rate of flow is defined as its flux. It is the total and simultaneous flow through the many pores of the given filter area that is measured. Flux has the dimensions of volume of flow per unit time, under unit pressure differentials, per unit area of filter. Because several pores are involved, the concept of porosity comes into play. 

 one means the percentage of a filter that is its voids or pores; that is, the ratio of space to solid in the membrane matrix. A given porosity can be constituted of a random combination of various numbers of pores, nonuniform in their lengths and diameters and of various shapes. The more numerous the pores, the more likely a higher flux. However, the lengthier the pores, the lower their individual flow rates. Pores of less-restricted diameters would flow faster but would tend to be less retentive. The filter's sieve retention would be decided by the size of its widest restrictive diameters. It is this dimension that will decide the largest particle to escape capture by penetrating the pore.  

As a foreseeable consequence of random pore arrangements, the same flux rates could characterize two membranes differing from one another in their mix of longer and shorter pores and of wider or narrower restrictive diameters, as modified by more or fewer pores. Likewise, the random balancing of numbers, lengths, and widths of pores could produce filters of the same porosity but differing in their flux and retention. As said, the pore-size ratings are indicative of particle retentions. Flow rates are not a function of retention ratings. Individual filters of different pore-size ratings could be characterized by similar rates of flow or even by flow rates opposite from those expected on the basis of the pore-size (retention) rating.  Membranes, if any, based on a randomness in pore formation could be of a same pore-size rating, indicative of retentions, but of higher porosities and flow rates. This would depend upon the random mix of pore numbers, lengths, and widths. Longer pores, the accompaniment of thicker membranes produced by filter layering, exhibit reduced flows but tend to greater retentions. Pall and Kirnbauer showed that layering in increasing the membrane thickness, elevated the bubble point to some leveling value (25). 

Figure 4: Modelling a fibrous filter by a system of lines drawn at random (n = 25).			

Certain filters, especially depth filters composed of fibrous structures, are manufactured by a technology that produces filters of randomly arranged pore structures; hence, their broad pore-size distribution (see Figure 4). Membranes fabricated using the casting technique, however, are characterized by the narrow pore-size distribution that is the result of the physical laws governing solutions. Within the volume of any solution, an equilibration by diffusion, and hastened by stirring, results in the solute molecules becoming evenly spaced from one another. The casting formula is such a solution of polymer dissolved in solvent. Within this solution the polymeric segments are, therefore, essentially equally spaced from one another. It is this relative regularity of intersegmental spacing that prefigures the similarly sized pores of the finished membranes. The resulting pore structure is, thus, not the product of random mixes of various pore lengths and diameters but rather the consequence of influences directed to pore homogeneity. It bespeaks a greater regularity of the pore structures constituting the porosity. The pores of cast microporous membranes should, therefore, be of a rather uniform size. This is, indeed, the case as shown by their narrow pore-size distributions (see Figure 5). 

Figure 5: Automated flow pore measurements of two different 1.2 Î¼m membranes.			

The possibilities are unlikely of pores of such regularity in size exhibiting identical retention ratings but differing in flow rates. Except for the irregularities introduced by pore-size distributions, organism retention, flux, and porosity should be parallel expressions of the membrane's pore size. This assumes, however, the use of identical polymers, casting solutions, and converting techniques, along with standard systems of measurements. None of these conditions are likely met by competitive filter manufacturers using proprietary practices. Therefore, comparisons, as of flux or of retentions, between filters of various manufacture may not usefully be extended to speculations concerning their structural features or performances. 

Within any given pore-size designation available in the market, there exists some range to its quantified properties. In its manufacture, a membrane lot may, and indeed will, incur some variation in the in-process bubble-point measurements that are translated into pore size. Each pore-size rating is prepared using a different casting formula. Nonetheless, the preparations do not yield distinct quanta, although the variations within a class are less than those between classes. Each batch or even each filter within a batch of membranes is classified to a single pore-size value, although each of the individual filters comprising the group may have its own bubble point within the range of values that define the given rating.  

As Schroeder points out, the pore size–retention correlation is not a step function (26). In its manufacture, a membrane lot will incur some variation. In using nonuniform standards, filter manufacturers might assign their pore-size labels somewhat ambiguously. One fabricator may, on the basis of flux, consider a membrane to be an "open" 0.1 μm. Another filter producer, using a somewhat different rating system might classify it as being a "tight" 0.2 μm. This could give rise to a labeled 0.1 μm, not so handicapped by reduced flows, being compared with a labeled 0.2 μm, not so advantaged by enhanced flows. The more "open" 0.1 μm may not flow faster than an average 0.2 μm but may do so against a "tight" 0.2 μm. Consequently, it becomes an unrewarding exercise to try to compare competitive membranes each rated by their own individual catalogue descriptions as perhaps being of the same pore size, yet exhibiting different flux rates or retentions under test.  

As is usually the case in the physical sciences, reliable evaluations and meaningful comparisons must be derived by users from performance data obtained through experimental investigation founded on suitable experimental designs. Another possibility for initial comparisons is the standard ASTM organism challenge test designed for 0.2/0.22 μm-rated filters but extendible to 0.1- or 0.45 μm-rated membranes as well (27). 

Figure 6: Different flow rates of 0.2- and 0.1-Î¼m rated filters.			

As shown in Table II, Tolliver and Schroeder found that the retention of 0.198 Âµm latex particles by five commercially available 0.2/0.22 μm-rated membranes differed significantly (21). The particle-retention mechanism was not complicated by adsorptive sequestrations because the solution contained surfactant. Given the size uniformity of latex particles, the variation in retentions based on sieving indicates different reference standards for the like-rated pore sizes. Nonstandard ratings also explain the Lindenblatt 

 (28) report that the flow rates exhibited by four 0.2/0.22 μm-rated membranes, as also of four 0.1 μm-rated filters differed markedly (see Figure 6) (28). Moreover, as indicated in Figure 7, the throughput volumes measured for the nonstandardized similarly rated pores also differed. Comparable data are offered by Jornitz 

Pore-size ratings are so unrelated to actual dimensions and so subject to anomalous interpretations as to make substantial dependency upon their values an unwise choice. Moreover, the means of measuring them are questionable. The pore-size rating system at best provides a qualitative differentiation.

The Relationship among Pore-Size Ratings, Bubble Points, and Porosity

There are advantages and disadvantages to using model microorganisms in the validation of sterilizing filtration. Although filtration validation studies may demonstrate the removal of the challenge organisms (whatever they may be), those studies can only imply that any particular filter used to filter a batch of drug product will yield a sterile filtrate. Successfully passing a bubble-point test (1) (or multipoint diffusive-flow integrity test), the results of which have been correlated to successful microbial challenge testing and validation studies, provides additional assurance that a sterile filtrate has been obtained. The filter manufacturer's quality-management system and production-control program further support the validity of the sterilizing filtration process. All three elements must be present to predict the successful outcome of a particular sterile filtration event (2). 

Perhaps the best method of determining whether a filter can yield sterile effluent is to subject the drug product to a proper microbial challenge and to analyze the filtrate for its microorganism content. This method of direct measurement would seem to be the most reliable way of characterizing a filter's sterilizing ability. Even though an individual filter can be characterized by a direct microbial challenge, however, the tested filter may not then be used in a processing operation because it has been contaminated by the test organisms. 

In practice, the measurement of a filter's sterilizing capability is derived from a correlation of its degree of bacterial retention with a characterization of its largest pores, essentially in terms of their width. The latter is obtained by means of the bubble-point integrity test method (3). It is necessary to confirm this correlative relationship because to assume it exists could prove a mistake. For instance, two qualities or properties of a filter may so constantly occur together as to be assumed to be correlated. Actually, they may in fact be two independent but parallel events that are not correlated (1). For example, single-point air diffusion rates of a filter measured at 5 psi had formerly been thought to correlate with a bubble point of some 50 psi for that filter, but it was discovered that it was necessary to determine the bubble point by measurements made at the actual bubble-point pressure. It did not necessarily correlate with the extension of the 5 psi single-point air diffusion test reading in all cases. So what had seemed to be a correlation turned out to be a case of parallel events. As a result, the use of single-point diffusive airflows measured at 5 psi was abandoned. In the present instance, extensive examination by many investigators has established the validity of correlating the bubble point with microbial retention (4). 

A filter can be characterized by direct microbial challenge as being capable of producing a sterile effluent. Extrapolating from the properties of the tested filter to other "identical" filters would, however, involve assumptions related to the uniformity of the filters' manufacture. To investigate this aspect of the filtration picture to a statistically meaningful level would involve a large number of correlated studies. Moreover, it might well necessitate the disclosure of proprietary information in a competitive market. This would prove highly impractical. From this one consideration alone, it becomes evident that a fundamental assumption must be made, namely, that the filter manufacturers can be depended upon to control their manufacturing processes so that "identical" filters achieve the same level of retention performance. 

Reliance upon the filter manufacturer is one necessary factor in this action. Identifying the requirements does not ensure the attainment of sterile effluent. 

Until rather recently, it was believed that the sterilization of liquids could be achieved by their filtration through a sterilizing membrane whose proper and pertinent identity was confirmed by its pore-size rating, which was itself determined by integrity testing. The pore-size rating, plotted against organism size, indicated an inverse correlation. However arranged, organism removal by a filter was seen as resulting from the mechanism of size exclusion, also known as sieve retention, wherein the organism is retained because it is larger in size and shape than the filter's pores. The sizes of various organisms are available from listings in the literature, and the size of a filter's largest pores derives from its bubble-point value. Therefore, it was possible to accommodate the size of the restraining pore to the size of the organism to be retained. In the interest of obtaining a maximum flow rate consonant with complete organism removal, the 0.2/0.22-μm-rated membranes were widely accepted as being sterilizing filters. 

The sterilizing filter was defined in 1987 by the US Food and Drug Administration on the basis of its retaining a minimum of 1 × 10

(at that time taxonomically identified as 

) per square centimeter of effective filtration area (5). 

The situation is complicated in its application by the absence of pore-size rating standards. The pore sizes may not be assumed to have been rated in the same fashion by the various filter manufacturers. Thus, while filters may bear the same pore-size designation, they may not be identical in this regard. Nevertheless, such filters may prove interchangeable in filtration processing operations. Manipulative adaptations often make it possible to substitute one such filter for another. 

Although the individual rating methods differ, bubble-point measurements performed in as similar a manner as possible enable comparisons to be made among the various types of filters. Given the importance of gauging the sizes of the largest pores, the integrity testing should be as accurate and reproducible as possible. Accordingly, using automated integrity test instruments is recommended, because their use eliminates the subjectivity inherent in manual integrity testing (6). For the same reason, the bubble point should be identified by the straight line plotted from multipoint diffusion measurements extended through and beyond it to a robust flow of air. Single-point diffusive airflow measurements are not sufficiently reliable for this purpose (7). 

There are numerous types and sizes of microorganisms. To explore the sterilizing capabilities of the available filters with each of these different organisms would not be practical and might not even be meaningful from the standpoint of assessing patient risk. Given the large number of organism types and their differences in size, it would be advantageous to select for testing an organism that could serve as a model for all the others, or at least for those commonly encountered in pharmaceutical operations. The smaller the test organism, the more its removal by a filter would ensure the sieve retention of larger organisms. Properties other than size, however, also are important in the selection of the model microbe. Ease and safety in cultivation and handling are real considerations. The likelihood of the selected organisms' being encountered in pharmaceutical operations is another influence on its selection. 

In the mid-1960s, Frances Bowman of FDA identified an organism that would penetrate the 0.45-μm-rated membranes that typically were used for sterilizing filtration at that time (8). The organism, currently known as 

 American Type Culture Collection 19146, had the ability to distinguish between 0.22- and 0.45-μm-rated membranes (it readily penetrated the latter) and, at concentrations higher than about 1 × 10

, it penetrated 0.22-μm-rated membranes, demonstrating that because penetration was concentration-dependent, mechanisms other than sieve retention were active. Although not the smallest organism known (smaller organisms had been known for decades), 

generally was considered small enough to represent whatever smaller organisms were likely to be present in pharmaceutical preparations. ASTM standard F 838, "Standard Test Method for Determining Bacterial Retention of Membrane Filters Utilized for Liquid Filtration," is based on the use of 

 as the challenge organism. The standard was originally approved in 1983 and was designed to determine the bacterial retention characteristics of membrane filters for liquid filtration. 

Since then, however, some 25 cases have been noted wherein the 0.2/0.22-μm-rated membranes qualified by withstanding the requisite 

 challenge did not yield sterile effluent (9, 10). Obviously, 

 does not serve as a universal model for all organisms. This finding has led some to advocate that the choice of the model organism whereby a filter will be qualified for use as a sterilizing filter should be an organism native to the drug preparation. In this view, 

 suffers from the demerit of not being part of the drug preparation's bioburden. Be that as it may, selecting a component of the bioburden as the model organism is not enough. The test organism must be amenable to identification and cultivation. Its life stages must allow amply for its management in the necessary test manipulations. Above all, its susceptibility to size diminution by contact with a given drug preparation will have to be investigated, and the kinetics and direction of the morphological changes, if any, will require elucidation. 

As previously stated, it had been believed that a sterilizing filter could be defined by way of its pore-size designation as identified by integrity testing. Developments in filtration practices showed this belief to be too simplistic. What had once seemed simple now is recognized as being quite complex. It was discovered that the conclusions based on pore-size ratings were subject to modification by the physicochemical specificity of the organism-suspending fluid, the individuality of the organism type in its size-changing response to the fluid, the possible change in pore size induced by the fluid, and the adsorptive qualities of the filter resulting from its particular polymeric composition, all influenced by the filtration conditions in their numerous varieties, but especially by the transmembrane pressure (11, 12). 

				A filter may not sterilize the same preparation under different filtration conditions, especially under dissimilar differential pressures (13). A given membrane may or may not retain a particular organism type suspended in a different drug vehicle (8). The organism type need not remain constant in size, but may alter in response to its suspending fluid (14–16). The effect of the vehicle upon the polymeric membrane may cause a change in its pore size (17). 			

The certainty of obtaining sterile effluent requires far more than the identification of a sterilizing filter by a pore-size rating. The variety of influences governing the outcome of an intended sterilizing filtration necessitates a careful validation of the process, including the filter. The very drug preparation of interest, the exact membrane type, the precise filtration conditions, and the prefiltration bioburden, including specific organism type(s) and number, should be used in the necessary validation. 

Given the complexity of the organism removal operation, it is doubtful whether a universal sterilizing filter can be devised. Certainly, there is no known absolute filter, one that will retain all organisms under all conditions, especially if viruses are included. Therefore, the successful attainment of a sterile filtration with regard to specified organisms of interest must in every individual filtration sterilization be corroborated by the documented experimental evidence that constitutes validation. The specifics of such operations are discussed in the Parenteral Drug Association's "Technical Report No. 26." Adherence to the teachings of this report is a second factor that is essential to attaining sterile effluent. 

Obviously, the many organisms whose presence in pharmaceutical contexts deserves consideration may differ from one another so greatly as to compel individual techniques for their identification and cultivation. Conclusions cannot be made regarding the sterile filtration of microorganisms unless methods of quantifying them by culturing and counting are available. Organisms such as the L-forms, nanobacteria, and "viable but nonculturable" entities may not be amenable to such analyses. Concerns about their presence may be justified, but without the means to cultivate and count them, it is impossible to attest to their complete absence. It follows that a sterilizing filter can be judged only by its performance in the removal of identifiable and culturable organisms known to be present in the drug preparation (18). 

Model organisms, although perhaps necessitated at least as an opening gambit, will not truly serve their intended, more general use as surrogates for most other organisms. In the present circumstance, however, 

 can serve as a model as well as any bioburden organism. Albeit with surprise to some and consternation to others, 

 did fail to serve as a filtration model in some specific circumstances. Nevertheless, the failures were relatively few. Moreover, they were the result of organism size reductions occasioned for particular organisms by contact with particular drug preparations. Many involved in pharmaceutical filtration were unaware that an organism's morphology could reflect its manner of cultivation, as well as the nutritional value of its suspending fluid, but those facts are known to microbiologists. 

It was not, nor is it as yet, sufficiently recognized within the pharmaceutical industry that organisms penetrating the 0.2/0.22-μm-rated membranes were likely to have undergone size reductions occasioned by exposures to particular drug preparations (

Ralstonia pickettii, Burkholderia cepacia,

 in particular). The dependability of the exclusion mechanism was predicated on the stability of the organism and pore sizes as measured in certain specific situations. It does require modification when organisms or pore sizes do not remain constant. Each may undergo change upon contact with particular drugs. At present, little is known about the kinetics of these size alterations. 

There is no universal sterilizing filter. Sterile effluent results from a combination of many influences, not solely from the use of filters of any particular pore size. Each filtration is an independent event that arises from a proper combination of several stated requirements.

Sterility must be defined in terms of specific organisms that are targeted in a given filtration. The ability to identify and cultivate the organism of interest is indispensable.

 as the model organism likely to represent the bioburden of the drug preparations presented for aseptic processing by sterilizing filtration is limited to those microbes that do not undergo shrinkage upon contact with the drug preparations. Also, it is clear from considerable experience that risks arising from the use of the 

 model are small enough in the vast majority of cases to be considered completely insignificant to public health.

Choosing an actual bioburden organism to replace 

 as a filtration model is unlikely to improve the situation. The substituting organism will replace the present problems with those peculiar to itself. It is questionable whether a universal model is at all possible.

 has served its model role adequately, except for those organisms like itself that undergo size diminution during their cultivation or that manifest size shrinkage through contact with drug preparations that are nutritionally poor for them. Thus far, these have been limited to relatively few, 

 These Gram-negative rods can be kept to extremely low numbers in prefiltration bioburden with reasonable and effective bioburden control measures.

Bacterial challenge studies should be conducted under process conditions by means of inoculating the product to be evaluated with 

  and evaluating the filtrate for the presence of the challenge organism. In the event that 

 is not viable in the product formulation, precondition the filter by recirculating product through it to simulate process conditions and follow this with a microbial challenge by modifying the process to ensure the viability of the challenge organism (

 change temperature), modifying the formulation to ensure the viability of the challenge organism (

 adjust pH, remove bactericidal component), reducing the exposure time to ensure the challenge organism remains viable, or change from 

 to a microorganism that has been isolated from the formulation. It is important to use the product formulation if possible, because there have been instances in which 

 has penetrated a sterilizing filter in contact with the product formulation but has been retained by the same filter when inoculated into a surrogate fluid. 

The ability to predict the successful outcome of a particular sterile filtration event requires a filter manufacturing enterprise that is highly capable and consistent, a dependable and accurate method to conduct integrity testing, and possession of the understanding and ability to implement effective filtration validation strategies. 

Model organisms are useful when validating sterile filtration, but successful retention of the model organism does not always guarantee that effluent is sterile. The authors explore the various factors that influence sterile filtration. 