Absolute or Sterilizing Grade Filtration – What is Required?

K. Kawamura, Ph.D., M.W. Jornitz and T.H. Meltzer, Ph.D.

Background

There has been an evolution in our understanding of membranes and of membrane

filtration over the half-century of their use. Their early successes in producing sterile filtrates led

to the optimistic belief that such membranes were absolute; and that they unquestionably

removed from pharmaceutical preparations the organisms commonly suspended therein. The

filtrative action was believed to result from sieve retention, the mechanism whereby particles

(organisms) larger in size than the pores become spatially restrained from passage. Subsequent

findings showed that adsorptive sequestration can play a role in filtrative sterilizations.

Adsorptions can be influenced by the filtration conditions and by the drug’s properties.

Consequently, filter validation is required to determine the sterilizing performance of a filter.

({PDA Technical Report 26).

Brevundimonas diminuta

B. diminuta (ATCC 19146), previously classified as Pseudomonas diminuta, came to

serve as the model organism for pharmaceutical filtration. These microbes suspended in a

penicillinase solution were found to penetrate 0.45 m-rated membranes, the “sterilizing

membranes” at the time, but were restrained by the tighter 0.2 m-rated filters devised for that

very purpose. However, the invoking of the sieve retention mechanism was called into question

because the 0.45 m-rated membrane did remove these organisms from aqueous suspensions

absent penicillinase (Bowman et al. 1967). The rationalization was that the organisms were

removed by adsorptive arrests to the filter surfaces (as well as by sieving) unless protein

1

1

competitively preempted the adsorptive sites. As a consequence, the adsorptive sequestration

mechanism came to be recognized.

Adsorption is governed by filtration conditions, such as the challenge density, the ionic

strength of the solution, possibly the temperature/viscosity, and most importantly by the applied

differential pressure (Mittleman et al. 1998). Dependence upon such various influences, by

definition, signifies the non-absoluteness of filters and of filtration (Tanny et al. 1979).

Absoluteness would mean freedom from such dependencies. However, so reliable is the

sterilizing removal of B. diminuta by 0.2 (0.22) m-rated membranes that these membranes have

come to be designated as the “sterilizing filters”, although the use of 0.45 m-rated membrane

can also result in a sterile filtration, depending on the process conditions. Pore-size numbers are

not particle-size retention ratings, as formerly believed. They do not imply unique retentivities,

sterilizing or otherwise. Their designation serves chiefly as a part-number identification code for

the filter manufacturer and user. Therefore, only appropriate validation suffices to ensure the

required sterilizing performance of a filter, independent of its pore size designation.

Sterilizing Filter Designation

Filter manufacturers designate their filters 0.2 (0.22) m-rated if they are capable of

withstanding challenges of 1 x 107 B. diminuta organisms per square centimeter of membrane

surface (HIMA 1982, ASTM 1988). The challenge is normally performed using a suspension of

a suitable concentration of B. diminuta in from 2 to 20 liters (usually) of saline lactose broth,

employing about 2 bar (30 psi) differential pressure for the filtration. Withstanding the challenge

test does not mean that such filter membranes have 0.2 m diameters. As shown in Figure 1,

membranes designated 0.2 m can be of larger pore size and still result in a sterile effluent, when

2

2

challenged as stated above. Adsorptive capture may be the effective mechanism in such a case.

Given the possibility of adsorptive sequestration, the FDA requires that membranes designated

by their manufacturers as being 0.2 m-rated or “sterilizing filters”, be experimentally

demonstrated to have sterilizing capabilities under “worst case conditions”, under the severest

conditions of the processing operations, (PDA/FDA 1995). These conditions would be the least

likely to augment organism retentions by adsorption effects.

The FDA stipulation in itself recognized that a membrane that sterilizes under one set of

circumstances may not so perform under another even when the same organism is involved. As

stated, the model organism is the B. diminuta grown under specific conditions (Leahy and

Sullivan 1978, Fennington and Howard 1997).

Pore Size Distribution 0.2 micronDifferent Filter Manufacturers

min. average max.0

0,2

0,4

0,6

0,8Pore Size (µm)

F1 F2 F3 F4 F5 F6

Figure 1: Pore size distribution measured with Coulter Porometer of 0.2 m-rated filter

membrane.

Bubble Point Correlations to Retention

Efforts have been made to match the B. diminuta size to the 0.2 m-rated pore size; a

relationship expressive of the sieve retention mechanism. The results are clouded if only

3

3

because pore-size ratings are innocent of any measurement standards. However, correlations

have been developed between organism retention levels and bubble point integrity test values

that are indicative of a filter’s largest pores (Johnston and Meltzer 1979) (See Figure 2). On the

basis of this correlation, a particular filter, given a sufficiency of B. diminuta retention, may be

designated as a “sterilizing filter.” The bubble point will differ for each polymeric type, but for

any type the label 0.2 (0.22) would apply to membrane that completely retains the B. diminuta

challenge of 1 x 107 cfu/cm2. The numerical rating should not be entertained seriously as a

measurement of dimension. There is, however, an identity between the bubble point

measurement and the affixing of the 0.2 (0.22) m label based on the requisite B. diminuta

retention.

4

4

Figure 2. Correlation of bubble points and organism retention (from data reported in the

literature). [From Johnston and Meltzer (1979).]

It is evident that the appellation “sterilizing filter” is defined in terms of B. diminuta removals.

This organism’s arrest by filters depends upon its particle size/pore size relationship and upon its

capability to adsorb to the polymeric membrane surface under the selected filtration conditions.

B. diminuta is hardly likely to serve as a universal model for all other microbes, of whatever size,

and under all other filtration conditions.

Still there is often surprise when filters with the 0.2 or 0.22 m-rated designations fail to

retain organisms. If nothing else, pore-size distributions, an essentially unelucidated property, in

itself signifies that larger pores are likely to be involved. Nevertheless, improper imputations of

poor filter quality, or of too large a pore size, may be made, as if the action of the “sterilizing

filter” were absolute and independent of the organisms involved, of the nature of the suspending

fluid, and of the filtration conditions. Consequently, urging that broad dependence be placed

upon 0.1 m-rated filters for pharmaceutical sterilizations, sans experimentally determined

needs, is not a responsible advocacy. Overstatements regarding the need for tighter filters offer

poor advice, given the potential problems introduced. The situation underlines the need for good

science, for the performance of filter validations.

Smaller Organisms

There is legitimate concern that organisms smaller than B, diminuta may be present in

certain pharmaceutical preparations. Given reliance on sieve retention, the 0.2 m-rated

membranes may not perform as sterilizing filters for smaller organisms. Organism penetrations

5

5

of 0.2 m-rated filters have been experienced on specific occasions and for certain products.

Organisms present in Water for Injection may, on account of the poor nutritional environment,

become reduced in size (Gould 1997, Collentro 1999). It has long been known that P. cepacia

and P. fluorescens are able to adapt to a low nutrition environment by reducing their

surface/volume ratio. These smaller organisms, which are mainly waterborne, are able to

penetrate 0.2 m-rated filters (Carter and Levy, 1998). L-Form organisms, devoid of their more

rigid outer membranes may be capable of negotiating the tortuous pore paths of filters (Thomas

et al. 1991). They are not retained by 0.2 m-rated filters (Hargreaves 1993), as also

mycoplasma are not. In such instances 0.1 m-rated or tighter filters should be used. There are

assertions that “nanobacteria” have been found in sera (Kajander et al. 1996). (Their existence

was inconclusively debated at a special discussion at the PDA Meeting 1999.) To elevate the

scope of concern, there is the threat of the unknown as posed by viable but non-culturable

microbes, entities difficult to determine (Colwell and Huq, 1995).

Substitution of 0.1 for 0.2 Membranes

As stated, such smaller organisms are better restrained from passage by 0.1 m-rated

filters. Some filter manufacturers advocate therefore that the 0.1 m-rated designation should

replace the current 0.2 specification for “sterilizing filter.” This advocacy is too general in its

intentions to merit practical application. Its implications echo the elusive “absolute filter” and in

so doing misread the scope and importance of process validation.

As an alternative to the use of 0.1 m-rated membranes, the use of two 0.2 m-rated

filters in series is also suggested. Jornitz and Meltzer (1999) indicate, however, that double 0.2

6

6

m-rated membranes, aside from augmenting adsorption effects, are not likely to be as efficient

as the 0.1 m-rated membrane (Meltzer, Jornitz and Johnston, 1999).

Supplanting 0.1 m-rated filters by their 0.2 rated counterparts, however prudently

intended, may be inadvisable. The suggested interchange should not be implemented unless the

need is experimentally indicated. Such substitutions can involve filtration problems and would

likely compel revalidations.

Whether the replacement of 0.1 and 0.2 m-rated filters requires revalidation deserves

consideration. The finer pored membrane structures may be more promotive of adsorptive

sequestrations since they offer more pore surface area. Viscosity effects may become enhanced,

flow rates being reduced. Borderline incompatibilities may become exaggerated, added

solid/liquid interfaces manifesting themselves. Additionally, the narrower pores may undergo

wetting with greater reluctance, almost all the newer membrane polymeric materials of

construction being borderline in their hydrophilicity. This could occasion an increase in (false)

integrity test failures, usually caused by incomplete wetting of the pore surfaces. Steam-in-place

failures could increase. These could eventuate from the greater impediment to the penetration of

narrower apertures by steam (Young et al., 1994), and by its more ready condensation upon the

increased surface areas (Steere and Meltzer 1998).

Above all, the improved, hoped for organism retention would require documented

experimental confirmation. As Lindenblatt (1999) indicated, there are different retention

performances by 0.1 m-rated filters from different manufacturers (Sundaram et al., 1999).

Some are able to retain mycoplasma, others not. Due to the lack of an appropriate challenge

standard for 0.1 m-rated filters, the user must either rely on the manufacturer’s data or perform

proper validations. Validation studies are to be preferred. They would also reveal whether or

7

7

not a 0.1 m-rated filter is really necessary. It may well be that a 0.2 m-rated filter will have

sufficient retention capabilities. The validation exercise avoids the too general advice to use 0.1

m-rated membranes.

Flow-Rate Differences

1Differential Pressure (bar)

0

5

10

15

20

25Flow Rate (ml/cm²/min)

0.1 µm 0.2 µm 0.45 µm

Figure 3: Flow rate differences of various pore-sizes

There are, as stated, applications for which the 0.1 m-rated filters are clearly indicated.

Risk assessments for other applications are appropriate. There are, however, penalties in flow

rates and potentially in throughputs to be considered. Figure 3 illustrates the dramatic decrease

in flux, for at least one type of membrane, that results when a 0.45 m-rated filter is substituted

for by a 0.2 and then by a 0.1 m-rated membrane. The flux decreases by 75% from a 0.2 m-

rating to a 0.1 m-rating. This considerable loss in flux is the result of flow varying as the fourth

power of the pore radius. The practical cost considerations of additional production time or of

purchasing added filters necessary to maintain the same production rate would be significant.

Some of this loss, but hardly all, can perhaps be compensated for by improved filter

design. However, tighter filter design will ineluctably result in flow rate diminution. Reliance

upon thinner membranes should be made with caution. They are more prone to imperfections.

Additionally, the use of prefilters to minimize increased flow-blocking particle accretions on the

8

8

tighter membranes could be necessitated. The application of 0.1 m-rated membranes has its

uses. The exercise should not, however, receive cavalier endorsements; it has its costs.

Validation is required and revalidation also when changes are made to already validated and

approved processes.

Validation Efforts; Product – Wet Integrity Testing

Filter manufacturers ascertain and specify the proper minimum water bubble point value

for their filter that accords with the required 1 x 107 cfu retention per cm2 of filter surface.

However, it is common for the filter manufacturers to provide a safety margin by furnishing

customers with membranes of higher than the minimum water bubble points that are listed. The

cost imposed on the filter user by this insurance is a somewhat lower rate of flow.

The classic situation involves a filter user testing three filters of the type selected to

determine their minimum water bubble points, and the corresponding product bubble points.

The ratio of the water/product bubble points is then ascertained from the averaged integrity test

values. This ratio value will determine for the user whether the product bubble point in

production meets the minimum allowable test value or not (PDA Technical Report 26). Of the

three filter specimens used in the integrity testing, at least one should have a bubble point value

close to the manufacturer’s listed minimum. Otherwise, the ratio level that becomes set will

reflect an unnecessarily high value. This, in turn, will mean that the filter user will not

subsequently be entitled to employ filters with lower water bubble points in production. This

will result in the filter purveyor not being able to sell for that specific validated application filters

having water bubble points between the minimum listed and those of the averaged test values.

9

9

Filter Manufacturer’s Choice

The manufacturer’s minimum listed bubble point is that which correlates to the required

organism retention level. It would seem to be in the filter manufacturer’s interest to provide

membranes with as close a value to that minimum as possible so as not to exclude from use

filters possessed of a sufficiency of retention but having bubble points lower than that set in the

validation. However, there is a problem. Filters, in each pore size rating, are produced within a

range of integrity test values. As said, many filter manufacturers have until now elected to

supply membrane with higher than the listed minimum bubble points, both as a safety factor for

users and perhaps in connection with minimizing scrap rates. Filter manufacturers, therefore,

must make a choice. One option lies with furnishing membranes having lower but acceptable

bubble points consonant with the listed minimum water bubble point. This would avoid

unnecessarily precluding such membranes from production use.

A second possibility is for the filter manufacturer to list a new water minimum bubble

point specification for the membrane he now regularly produces, to move the listed value closer

to his customary commercial offering. To be sure, this would forego selling filters of lower

bubble points but still above those representing the correlation to the required organism

retentions. The safety factor of higher minimum values would automatically inhere. The cost to

the filter manufacturer would be minimal if the product remains consistent in its bubble point

values. The scrap rate, too, would reflect the constancy of manufacturer, but the lower cutoff

level would be that of the (higher) new minimum setting. Implications that this change in

catalogue specifications implies a corrective to previous improper lower bubble point values

would be inappropriate.

10

10

How the filter producers address their problem, essentially a business decision, need not

concern the user. Most prefer to offer membrane having bubble points closer to the minimum

value stipulated in their product descriptions. At least one filter manufacturer, however, has

elected to specify a new (higher) minimum water-wet bubble point for his product. The new

value, routinely manufactured, is sanctioned by its being above the filter manufacturer’s

minimum value, the level at which that filter type retains 1 x 107 cfu/cm2 of filter surface.

Raising the minimum bubble point of the membrane being offered to the drug industry, in

effect a change in the catalogue listing, in no way impugns the correctness of the value initially

determined, by the filter manufacturer. Even so, one should not take the new value for granted.

Membranes are produced within a specified bubble point range; between an upper (maximum)

and lower (minimum) value. If boosting the minimum water bubble point is accompanied by a

rise in the filter’s upper level, the upward shift of the specified range may excessively elevate the

upper limit, thereby changing filter performance. At some upper value this filter is no longer a

0.2 m but perhaps a 0.1 m. Evidence is required to show that the filter’s performance did not

significantly change. Besides the ability to supply a sterile effluent, certain other properties have

to be evaluated, as for example described in PDA Technical Report 26, e.g. adsorptivity, flow

rates, pressure conditions, extractable levels, etc. When the bubble point is changed, now to a

higher level, one has to ask whether this will influence the pore size and pore size distribution,

and with it the flow rate and therefore, contact time or time to filter the required batch volume.

If, because of increased contact time between the solution and the filter, the adsorptive effect is

heightened, the product could be changed in respect to its activity, potency and strength.

Changes in the process, of equipment settings or parameters have to be evaluated regarding their

potentially adverse effects on the process and/or drug product. Therefore, a change of integrity

11

11

test value of a sterilizing grade filter, even to a higher assurance level, requires testing to secure

documented evidence that it will not have adverse effects.

Either way, filter users will have at their disposal microporous membranes dedicated to

sterile filtrations from which minimum product bubble point can be arrived at with confidence.

Conclusion

One can clearly state that there is no absoluteness in filtration except where the smallest

particle is larger than the largest pore. The questions is, is absoluteness required? Aseptic

processing in biopharmaceutical contexts requires assured sterile filtrates, but not “absolute

filters”. Thorough filter and process validations, properly performed, will ensure sterile filtrates.

The filter user has to understand the process and microbial circumstances and validate the filter

according to these parameters. This will reliably produce a sterile fluid product. As stated, a

sterilizing filter is independent of its pore size. A sterilizing grade filter can be 0.45 m-rated in

certain applications, in other instances 0.2 m or 0.1 m-rated filters could be a suitable

alternative. Penalties in flow rates and throughputs should not be incurred by the needless use of

tighter membranes. The choice of a filter rating should not be made in response to unassessable

concerns or undefinable fears. Appropriate process and product validations performed by the

filter user, supported by the filter manufacturer, should be employed to determine which filter

rating will offer the best choice for the specific process.

12

12

References:

American Society for Testing and Materials (ASTM), Standard F838-83, Standard Test Method for Determining Bacterial Retention of Membrane Filters Utilized for Liquid Filtration. 1983, Revised 1988.

Bowman, F.W., Calhoun, M.P., and White, M. (1967). Microbiological methods for quality control of membrane filters, J. Pharm. Sci., 56:222-225.

Carter, J.R. and Levy, R.V., (1998). Microbial Retention Testing in the Validation of Sterilizing Filtration in Filtration in the Biopharmaceutical Industry. Meltzer, T.H. and Jornitz, M.W., Eds., Marcel Dekker, New York.

Collentro, W.V. (1999), “The Use of 0.1 m Membrane Filters in USP Purified Water Systems – Case Histories, Part I”, Pharm. Tech.: 176-192

Colwell, R.R. and Huq, A., (1995). “Viable but Non-Culturable Bacteria and Their Implications for Water Purification”. Ultrapure Water 12(3), pp. 67-74

Fennington, G.J. and Howard, G., Jr., (1997). Preparation and evaluation of bacterial stock for filter validation, PDA J. Pharm. Sci. Technol. 51(4): 153-155.

Hargreaves, P., (1993). “Paul Hargreaves Speaks Out on Pharmaceutical Manufacturing” PDA Letter 10-13.

HIMA, (1982). Microbial Evaluation of Filters for Sterilizing Liquids, Document No. 3, Volume 4. Health Industry Manufacturers Association, Washington, D.C.

Johnston, P.R. and Meltzer, T.H., (1979). “Comments on Organism Challenge Levels in Sterilizing Efficiency Testing”, Pharm. Tech. 3 (11): 66-70

Jornitz, M.W. and Meltzer, T.H., (1998). “Sterile Double Filtration”, Pharm. Tech. 22 (10): 92-100

Kajanda, E.O., Kuronen, K. and Ciftcioglu, N., (1996). “Fetal Bovine Serum: Discovery of Nanobacteria”, Mol. Biol. Cell. 7 (Suppl.), 517a.

Leahy, T.J. and Sullivan, M.J. (1978). “Validation of Bacterial Retention Capabilities of Membrane Filters”, Pharmaceutical Technology 2(11) pp. 64-75.

Lindenblatt, J. (1999). “0.1 µm versus 0.2 µm Final Filtration in BFS Operations”, Minutes from AGM, BFS News, Autum Edition 1999: 44-48

Meltzer, T.H., Jornitz, M.W. and Johnston, P.R. (1999). “Relative Efficiencies of Double Filters or Tighter Filters for Small-Organism Removal”, Pharm. Tech. 23 (9): 98-106

13

13

Meltzer, T.H., Jornitz, M.W., and Trotter, A.M., (1998). “Application-Directed Selection of 0.1 µm or 0.2 µm Rated Sterilizing Filters”, Pharm. Tech. 22 (9): 116-122

Mittleman, M.W., Jornitz, M.W. and Meltzer, T.H.. (1998). “bacterial Cell Size and Surface Charge Characteristics Relevant to Filter Validation Studies”, PDA J. Pharm. Sci. Tech. 52 (1): 37-42.

PDA Special Scientific Forum, Bethesda, MD, Validation of Microbial Retention of Sterilizing Filters, July 12 – 13, 1995

Steere, W.C. and Meltzer, T.H. (1998). “Operational Considerations in the Steam Sterilization of Cartridge Filters, “ Pharm. Tech., 17(9): 98-110.

Sundaram, S., Eisenhuth, J., Auriemma, M., Howard, G. Jr. and Brandwein, H. (1999), “Retention of Diminutive Water-borne Bacteria by Membrane Filtration”, PDA Annual Meeting, Dec. 1., 1999, Wash., D.C., PDA Jour. Pharm. Sci. Tech. (in press)

Tanny, G.B., Strong, D.K., Presswood, W.G., and Meltzer, T.H., (1979). The Adsorptive Retention of Pseudomonas diminuta by Membrane Filters,” J. Parenteral Drug Assoc., 33(1): 40-51.

Technical Report No. 26, Sterilizing Filtration of Liquids. PDA Journal of Pharmaceutical Science and Technology, Vol. 52 No. S1, 1998.

Thomas, A.J., Durkheim, H.H., Alpark, M.J. and Evers, P., (1991). “Dection of L-forms of Pseudomonas aeruginosa during Microbiological Validation of Filters”. Pharm. Tech. 15 (1): 74-80.

Young, J.H., Ferko, B.L. and Gaber, R.P., (1994). “Parameters Covering Steam Sterilization of Deadlegs”, J. Pharm Sci & Tech 48(3), pp. 140-147.

14

14