A review and critique of multifilter arrangements

The myriad of successful sterile drug preparations by way of filtrations attests to the pharmaceutical industry's confident application of this purification technique. The constant search for improvements in purification has resulted in a growing interest in multifilter arrangements, which offer greater assurances regarding organism retentions. Multifilter arrangements are often over-designed and the most successful are based on solid scientific principles.

Committee for Proprietary Medicinal Products (CPMP) guidelines from the EU and FDA, and FDA's Aseptic Guide (2004) recommend the use of redundant or serial 0.2/0.22 µm-rated membranes. FDA's new aseptic guideline states: "Use of redundant filtration should be considered in many cases." This guidance is unspecific and reveals the level of uncertainty surrounding the subject.^1–5

This article will review and critique multifilter arrangements, and examine filter combinations in terms of flow rates and throughputs. Flow rates are important for their very practical significance with respect to filtration time, filling speed and filter contact time. Additionally, they connect to other factors that relate to time and durations, for example, organism proliferation and an increase in endotoxin. The utilitarian value of throughput is self-evident in terms of yield. It will also measure what has so far been attained (i.e., a 50–75% increase in throughput of filters in the last 15 years) within the operational limits set by the industry's unswerving drive for sterility.

Multifilter arrangements

Two (or more) filters can be arranged to

• provide a prefilter/final filter effect

• enhance flow rates and throughputs

• ensure against an imperfect final filter by using repetitive filters.

The latter is more common in Europe than in the US. However, there is a distinct difference between two 0.2 μm filters in series and redundant 0.2 μm filtration, meaning the two filters are directly connected to each other (Figure 1). No other equipment is in-between the two filters.

Figure 1 Schematic of redundant and serial filtration.

Double-layered filters

Two types of double-layered filters are available. The heterogeneous form consists of layers of dissimilar pore-size ratings. The homogeneous type is composed of two membranes of the same pore-size rating. The two different constructions are intended for different purposes.

Heterogeneous filters

An example of a heterogeneous construction would be a 0.45 µm-rated membrane upstream from a 0.2/0.22 µm-rated final filter. The more open filter, in its position upstream of the tighter version, serves in the protective role of prefilter; accepting a portion of the particulate load, which prolongs the service life of the final downstream membrane (Figure 2). The flow rate is slowed by the longer overall flow path of the double construction. The slowing effect is greater than that of either single membrane.

Figure 2 Schematic of a heterogeneous filter layer construction.

When serial filtration is used, each filter is in its own housing and separated so can be tested independently. When the filter arrangement is redundant, neither filter can be tested without invading the space separating them unless some very unusual arrangements are devised. This is a disadvantage of redundant filter system designs.

Key points

Homogeneous filters

Homogeneous double constructions contain two membranes of similar pore size ratings, for example, both 0.2 µm-rated (Figure 3). This double-layer construction reduces the probability of particle passage. The intention is to ensure that even if an inappropriate or flawed membrane was somehow included, the second filter would serve as a safeguard; likewise, when single membranes cannot be cast to thicknesses sufficient to retain the target organism. All too often, the situation is that of a filter of an unknown pore size distribution being confronted by organisms of an unknown particle size distribution. The possibility of selecting an improper filter is conceivable.

Figure 3 Schematic of a homogeneous filter layer construction.

Unfortunately, a homogeneous double-layer configuration does not allow the detection of an integrity test flaw of only one membrane when the bubble point test is used. Therefore, there could be retention risks attached to the homogeneous double-layer design, when both membranes are required to be integral to ensure a sterile effluent.

The flow rate through these double-filter constructions is less than through single membrane layers because the overall pore lengths are longer. The flow rate through homogeneous arrangements is less than through heterogeneous double filters because in the former combination both membranes have the tighter restrictions of the final filter.

Repetitive filtration — in close contact design

In this case, the pore is conceived as being a convoluted, irregular capillary. It is, therefore, necessarily longer than the depth of the filter because of its tortuosity. Its restrictive dimensions may occur anywhere along its length. This assumption explains certain observations regarding repetitive filters. Thus, it would be expected that lengthening a pore — as by layering membranes — might proportionally increase the number of its restrictive regions. This should promote particle trapping. In effect, prolonging the pore length is the technological achievement that results when membranes are superimposed on one another in intimate contact. A second consequence is a homogenization of the pore size distribution. A pore with larger openings will likely juxtapose one with smaller openings because there are many more smaller pores. The result is a narrower — but longer — overall pore passageway. This occurrence assists sieve retentions by decreasing the number of larger (wider) pores. Moreover, the pore path length is doubled. This encourages adsorptive sequestrations.

Repetitive filtration — space-separated layers

When two membranes separated from one another are used, a different result is obtained. Between the two membranes there exists a space where the fluid exiting the first membrane forms a pool from which it is hydrodynamically directed preferentially to the larger pores of the second membrane.

Higher differential pressures can have the same effect as the direct and intimate superimposition of one membrane on the other. At the higher differential pressures, the effluent is more immediately directed perpendicularly from the pores of the top filter to those, of whatever sizes, of the second filter. There may not be enough time to permit the hydrodynamic flow to assert itself. This limits the influence of the hydrodynamic preference.

This situation favours sieve retentions because the larger particles are denied time within the between-the-filters pool to orient towards the scarcer, larger pores of the second filter. Adsorptive retentions would seem to be marginally favoured given the preponderance of smaller pores (Figure 4).¹⁰ In these cases, filter efficiency is increased. Where the total suspended solids (TSS) content is large enough, albeit an unlikely condition, the system would incline towards lower throughputs.

Figure 4 Adsorptive sequestration and bridging of particles.

Nonadditivity of LRVs

The common term used to describe the organism-reducing power of a filter is to state its log reduction value (LRV). In effect, FDA defines a sterilizing filter as one that can cause a seven log reduction in the titre of a suspension of unspecified magnitude. The LRV is log to the base 10 of the ratio of the organisms in the influent stream to those that emerge in the filtrate.^1,6

The LRVs of a redundant filter combination are not necessarily multiples of the single filter. The exact sum depends upon the probabilities of a given size organism encountering an appropriately sized pore within each membrane. In any case, the likelihood is that the organisms penetrating the upstream filter will be smaller. This shifts the organism size distribution confronting the second filter in the direction of smaller sizes. The smaller organisms are more elusive, making a lower LRV for the downstream filter more likely. In a case study the LRV of the second filters has been found to be 2–4 logs lower than the first filter.¹² The situation where particles of an unknown size distribution challenge pores of an unknown size distribution is complex enough to permit numerous (but profitless!) hypothetical combinations to be considered. Nevertheless, perhaps the following question should be asked: If a given filter cannot completely retain the organism challenge, why should an identical second filter be expected to do so? Once again, the filter industry's neglect of the important area of quantifying membrane pore size distributions is evident. In any case, the literature attests to the nonadditivity of the LRVs of individual filters.^10–12

Redundant filtration

This term is often used to designate a repetitive or double-filter arrangement where the two filters are housed separately, but connected to each other. The term 'redundant' is usually associated with being wasteful or unnecessary. Its intended meaning in filtration is to imply 'enough and to spare'. Redundant filtration incurs the cost of two housings, but the separate housings permit the integrity testing of each filter, although the risks inherent to asepsis in prefiltration integrity testing still apply. Within a single housing the upstream testing of the downstream filter is manageable, but will require the isolation of the upstream filter. The housing modifications necessary to this action is rarely undertaken.

The first filter is expected to fulfil the retention requirement.⁹ In the event of its failure, the second filter ensures this function. This arrangement is regarded as prudent by its practitioners. However, given the dependable performance of stipulated integrity testing, the endorsement of the use of backup (repetitive) membranes may also reflect some degree of personal insecurity. While understandable, this subjective reaction would be an intrusion into what ought to be an entirely technical decision.

A filter's integrity can undergo damage during use; most likely (but not only) from in-line steaming. Hence, the requirement for final integrity testing.^4,7,8 However, in the actual event, both filters would likely be affected similarly given their mutually shared exposures. Even where employed for reasons of prudence, the use of the repetitive filters is not necessarily justified economically.

Economics

Calculations can be made to approximate the economics of enlisting repetitive filtrations to safeguard the quality (sterility) of an effluent. The question is: In the event of a failed filter, what would be the value of the drug product that had to be discarded, and what would be the cost of the additional filters needed to prevent the loss? Hypothetically, if a filter failure happens every 10000 batches, and the filtered batch has a value of $100000, the expense of saving the batch by using repetitive filters for every batch would be: 10000 batches×filter unit cost of $150=$1.5 million. The total gained by the saved batch is its value of $100000 plus $50000 in downtime and overheads that replacing the batch would have cost; a total of $150000.

The expense of saving the batch by using repetitive filters for every batch would be the cost of $1.5 million less savings of $150000: a total loss of $1.35 million. The costs involved from a risk management standpoint do not substantiate a redundant filtration. However, some biologic drug products easily reach a value of $1.5 million a batch. In these cases an 'insurance' filter might be useful. However, a second filter adds to unspecific adsorption and consequently yield losses and possibly higher hold-up volumes. These costs need to be evaluated, before a decision can be made. The question that needs to be addressed is why an insurance filter might be required? When the process is appropriately validated, retentivity and performance of the filter should be ensured.

Repetitive filter use

Regarding the EU and FDA, both the European Agency for the Evaluation of Medicinal Products' (EMEA's) guideline (CPMP, April 1996)³ and the EC Guide to Good Manufacturing Practice Annex 1,² as well as FDA's Aseptic Guide (2004)⁵ recommend the use of redundant 0.2/0.22 µm-rated membranes. FDA's new aseptic guideline states: "Use of redundant filtration should be considered in many cases." The placement of the filters is to be "as close as possible" (or practical) to the filling needles. This will reduce the likelihood of contamination resulting from random biofilm shedding. The usual location is just before the filling needles or before the reservoir that feeds them. In practice, the 'recommendations' are enforced as if they were law. Thus, in Europe they are becoming common practice.

The recommendation is not based on any known survey or experimental data, nor have the regulatory authorities explained their reasoning. As stated, overdesign is a normal response to uncertainties. The technically oriented individual will resist making unsupported decisions. He/she may insist on waiting for data. Those responsible for public safety may be obliged to make decisions when adequate supporting data are not yet available.

So what is required regarding the integrity testing of the repetitive filters? If the downstream filter is intended as the final filter, then it alone needs testing. If the filter pair is being relied on to ensure against the likelihood of filter failure, both membranes should be tested in the hope that one will remain unflawed. From an engineering point of view this is manageable if the redundant membranes are housed separately.

The employment of automated integrity test machines is especially advantageous as it avoids invading the system downstream of the filters that are housed separately. The threat of asepsis is, therefore, eliminated from the integrity testing activity. As previously said, placing both filters in the same housing makes it difficult to perform upstream integrity testing of the downstream filter. It is manageable, but would require the isolation of the upstream filter.

Conclusion

The regulating authorities have, it seems, a preference for the application of multimembrane combinations to maximize organism retentions. The reasons for their views, as usual, are not disclosed. Perhaps, therefore, it may not be too great a heresy to suggest that there are no known compelling reasons to support the recommended practice. Reducing risk is dependent on overdesign. Nevertheless, however subjective, the recommended practice may be useful. There does, however, remain the problem of how to integrity test each of the membranes of a redundant filter arrangement.

Maik W. Jornitz is group vice president, global product management bioprocess, Sartorius North America Inc., Edgewood, NY, USA.

Theodore H. Meltzer is president, Capitola Consulting Co., Bethesda, MD, USA.

References

1. ASTM F-838-83 "Determining the Bacterial Retention of Membrane Filters Used for Liquid Filtration," American Society for Testing and Materials, Philadelphia, PA, USA (1983).

2. EC Guide to Good Manufacturing Practice Annex 1, Brussels, Belgium (2003).

3. EMEA-CPMP Note for Guidance on Manufacture of the Finished Dosage Form, Page 5/6 (April 1996; Reissue).

4. Food and Drug Administration, "Guideline on Sterile Products Produced by Aseptic Processing", Division of Manufacturing and Product Quality, Office of Compliance, Center for Drugs and Biologics, Rockville, MD, USA (1987).

5. Food and Drug Administration, "Guidance for Industry, Sterile Drug Products Produced by Aseptic Processing — Current Good Manufacturing Practice", Office of Training and Communication, Division of Drug Information, HFD-240, CDER, Rockville, MD, USA (2004).

6. HIMA Document #3, Vol. 4, "Microbiological Evaluation of Filters for Sterilizing Liquids", Health Industry Manufacturing Association, Washington D.C., USA (1982).

7. M.W. Jornitz and T.H. Meltzer, "Filtrative Particle Removal From Liquids", in Filtration in the Biopharmaceutical Industry (Marcel Dekker, New York, NY, USA, 1998).

8. T.H.Meltzer, Filtration in the Pharmaceutical Industry (Marcel Dekker, New York, NY, USA, 1987) pp 255–259.

9. PDA, Technical Report #26, PDA J. Pharm,. Sci. Technol. Suppl. 52(S-1) pp 1–31 (1998).

10. A.R. Reti, T.J. Leahey and P.M. Meier, "The Retention Mechanism of Sterilizing and Other Submicron High Efficiency Filter Structures", Proceedings of the Second World Filtration Congress , London, UK, pp 427–436 (1979).

11. A.M. Trotter, P.J. Rodrigues and L.A Thoma, Pharm. Tech. 26(4), 60–71 (2002).

12. S. Sundaram et al., "Bacterial Retention by Double (Serial) 0.2/0.22 Micron Rated 'Sterilizing Grade' Filters", PDA Annual Meeting, Washington D.C., USA (2001).