Optimization, scale-up, and validation issues in Filtration of Biopharmaceuticals, Part 1

August 1, 2004
Anurag S. Rathore
Anurag S. Rathore

Anurag S. Rathore is a professor in the Department of Chemical Engineering at the Indian Institute of Technology Delhi and a member of BioPharm International's Editorial Advisory Board, Tel. +91.9650770650, asrathore@biotechcmz.com.

Jerold M. Martin

Jerold Martin was the senior vice president of global scientific affairs at Pall Life Sciences and chairman of the Board and Technology Committee at Bio-Process Systems Alliance.

BioPharm International, BioPharm International-08-01-2004, Volume 17, Issue 8
Page Number: 40–45

Filtration is one of the most commonly used unit operations in biopharmaceutical manufacturing. Available formats include direct or normal flow filtration (NFF) and cross or tangential flow filtration (TFF). These methods are used for sterilization and virus filtration, depth filtration or ultrafiltration, and diafiltration applications. Some common objectives include:

Filtration is one of the most commonly used unit operations in biopharmaceutical manufacturing. Available formats include direct or normal flow filtration (NFF) and cross or tangential flow filtration (TFF). These methods are used for sterilization and virus filtration, depth filtration or ultrafiltration, and diafiltration applications. Some common objectives include:

  • separation of particulates from the process stream (for example, to separate host cell components and cell debris in the upstream or midstream portion of the process or before a chromatography step to prevent particles in the feed or buffer from clogging the column)

  • separation of impurities from the process stream (for viral clearance, for sterile filtration, or for host cell protein or DNA removal by charged filters)

  • concentration of the feed stream using ultrafiltration to reduce the process stream's total volume or to reach a targeted product concentration

  • buffer exchange via the use of a diafiltration step to deliver the product to the desired buffer system before chromatography or final formulation.

This article is the fourth in the "Elements of Biopharmaceutical Production" series and will be published in two segments. In this first segment, Anurag Rathore and Alice Wang present development and scale-up of a depth filtration step, and Jerold Martin discusses development and validation issues for a viral clearance filtration step.

Anurag S. Rathore and Alice Wang, Amgen Inc.

Process Development and Scale-up for Depth Filtration

The following shows the development of a robust depth filtration step for clarification.


We wanted to design a robust and scaleable depth filtration step that could maximize product recovery (in this case, a protein expressed in

Pichia pastoris

) and that could handle varying % solids content in the feed stream from the preceding centrifugation step. First, we performed a screening study using various depth filters available on the market that met our manufacturing process criteria. Second, we chose filters based on the screening data; we directly and thoroughly compared those filters to determine the best performing filter. Then, we performed characterization studies on the chosen filter to ensure robust performance at scale. Finally, we scaled the filtration step up to pilot scale and compared its performance under these conditions.


As listed in Table 1, we chose ten different filters from three manufacturers based on their suitability for manufacturing. We used disposable filter disks for filter screening experiments. We used filter format, surface areas, and recommended flow rates from the various manufacturers as guidelines for the screening study (Table 2). We designed the depth filter train in stages with a more open filter (such as Cuno 10SP, Pall Supra 80P) ahead of a tighter grade to allow for robust operation with high capacity. Some of the depth filters in Table 1, including the Millipore Millistak+A1HC, Millistak+B1HC, and the Cuno 90M08, combine sequential grades of media in one filter.

We evaluated filter performance by a constant flow experiment (Pmax study).2 The study's screening criteria were filter capacity, filtrate quality determined by filtrate turbidity, target protein mass balance, and robustness regarding variation in feed characteristics such as % solids and feed turbidity. Depth filters were first flushed with sufficient buffer to wet filter media thoroughly and remove extractables. Then the feed, which was agitated and maintained cold (2 to 8°C), was pumped through the filters at a constant flow rate (manufacturers' recommended rates are given in Table 2) until reaching a differential pressure of 25 to 30 psi. We plotted the differential pressure versus normalized throughput at different time intervals. Throughput at 50% of maximum differential pressure for all the filters is summarized in Table 3. While most filters exhibited good capacity with feed containing 0% solids, filter capacity declined significantly when using feed containing 0.7% solids. The only exceptions to this observation were the performance of the Millistak+A1HC and B1HC, which have open diatomaceous earth (DE) media, tighter DE media, as well as 0.1 mm cellulosic membrane. Both filters provided adequate capacity for both feeds. The recovery across the filters was measured by a product specific assay and was found to be 80 to 95%. Instant filtrate turbidity and filtrate pool turbidity were measured by a Hach turbidity meter. There was no turbidity increase, and the filtrate pool turbidity with feed containing 0.7% solids was about 3 Nephelometric Turbidity Units (NTU) for all the filters. For feed containing 0% solids, an appreciable turbidity breakthrough was observed, and the pool turbidity was 10 to 15 NTU for most filters, 8 NTU for the Millistak+ A1HC, and 11.3 NTU for the Millistak+B1HC.

Table 1. Depth Filters Used in the Screening Study

Filter screening results indicated that pressure breakthrough is the primary limitation for this application, and the Millistak A1HC and B1HC appeared to be the most suitable filters due to their high capacity with respect to variation in feedstock characteristics.


Table 3 presents data from experiments with different filters and feeds with different proportions of solids. We chose the feed (0% versus 0.7% solids) based on the manufacturer's recommendation for each filter. In most cases, if the filter performed poorly with feed containing 0.7% solids, we did not test it with the 0% solids feed. As seen in Table 3, the A1HC filter has a higher capacity than the B1HC filter with 0% solids feed, but a lower capacity with 0.7% solids feed. We believe this is due to the different particle size distribution of the two feeds; the 0% solids feed contains more smaller particles that lead to a faster plugging of the B1HC filter, which is structurally more open in comparison to the A1HC. We compared the Millistak+A1HC and B1HC further, using a different lot of feed (Figure 1). For this lot, the A1HC demonstrated higher capacity than the B1HC for both low- and high-solids content feed. Also, while there was no turbidity breakthrough with 1% solids feed, with the 0% solids feed, breakthrough occurred earlier with the B1HC filter than with the A1HC. Based on these results, Millistak+A1HC was chosen due to its superior filter capacity.

Table 2. Recommended Flow Rate for Pmax Study


To ensure robust operation at large scale, we evaluated the effect of key operating parameters on the performance of this unit. These parameters included % solids in the feed, lot-to-lot variation in feed, batch-to-batch variation in filters, filtration flow rate, and temperature. Figure 2 illustrates the performance of the Milistak+A1HC filter using two different lots of feed material. While both feeds contained 0% solids, the turbidity was 58 NTU and 129 NTU, respectively. Lot-to-lot variation in feed resulted in earlier pressure and turbidity breakthrough in one case, indicating the need for a safety factor during scale up.

Further, the effect of the percentage of solids on filter capacity was evaluated (Table 4). As expected, the required filter area increases significantly with the increase in percentage of solids in the feed. Based on this data, we decided that specifying the percentage of solids for the preceding centrifugation step is necessary for a consistent and validatable process.

Table 3. Throughput Comparison (L/m2 at 15 psi)


To test the feasibility of large-scale operation of the depth filtration step, we used an 8 cell 16 in. A1HC filter with 1.8 m


surface area to filter 300 L of fermentation broth. We evaluated filtrate turbidity and the pressure drop generated at this scale and compared the results to those at bench scale. As seen in Figure 3, under the conditions investigated, data from the two scales compare favorably, demonstrating robust performance and scaleability of this step. Product recovery across the step was 95%, same as that seen in the small-scale experiments.

Figure 1. Comparison of A1HC with B1HC Using Different Feeds


We found the Millistak+A1HC to be the most effective depth filter for harvest of this protein expressed in Pichia pastoris. It was observed that pressure breakthrough is the dominant factor while processing high % solids feed (>0%). Turbidity breakthrough was only observed while processing lower percentage solids feed (~0%). We found the characteristics of feedstock significantly impacted filter performance.

Table 4. Millistak+A1HC Capacity vs. % Solids

Jerold M. Martin, Pall Life Sciences

Validation of Viral Clearance by Direct Flow Membrane Filtration

The use of cell culture and plasma-derived proteins to treat a wide range of disorders continues to expand. Coupled with these developments is the need to ensure the safety of newly developed and established biopharmaceutical drugs. While biological raw materials contaminated with infectious or cytopathogenic viruses are preferably excluded from use in biopharmaceutical manufacturing, a variety of circumstances can result in viral contamination of final products. The use of appropriate viral clearance methods can significantly reduce the risk of accidental viral transmission and thus enhance safety.


An ideal viral clearance method should embody several key characteristics, including: a well-defined mode of action, high viral titer reduction, and high product recovery. Also, it should not interfere with the biological integrity and reactivity of the product; it should not contaminate the product; it should have no additional stabilizers or other additives; and it should be scaleable, possible to validate, and applicable to a wide range of products.


The characteristics of a product dictate the suitability of a particular viral clearance method. The size of the protein, its conformation, and its lability to heat or other inactivation methods are important considerations. Likewise, viral size, lability, and the presence or absence of particular macromolecules should be evaluated as characteristics of potential viral contaminants. Process evaluation techniques also guide the suitability of a particular viral clearance method. Understanding how the chosen viral clearance method affects other variables in the process is a critical component of the elimination and it should be documented as part of a viral clearance strategy.


Of the available viral clearance (inactivation and removal) strategies, size exclusion filtration is preferred because it is less dependent on product or process conditions and, therefore, more robust. Membrane surface chemistry and interactive forces between contaminants and the membrane can affect the retention efficiency and capacity of membrane filters. Adsorptive retention by a membrane is subject to pore surface chemistry and resultant electrokinetic or hydrophobic interactions between viral particles and the membrane surface. Virus retentive membrane filters typically rely primarily on size exclusion with minimal adsorptive properties. Ion exchange chromatography membranes, however, feature significantly larger, non-size-retentive pores with active membrane surfaces able to adsorb viruses from process fluids.

Figure 2. Impact of Lot to Lot Variation on Performance of Millistak + A1HC - Feed 1 vs. Feed 2, both 0% solids. Pressure drop curves show data from duplicate experiments for each set of conditions.


Users should understand the basis of a filter's rating. In general, virus filters can be classified into two groups, higher rating filters (nominally 35 to 80 nm) providing ≥4 log titer reduction (LTR) for ≥40 to 50 nm viruses (for example, retroviruses, SV40, and BVDV) and lower rating filters (nominally 15 to 25 nm) providing ≥3 LTR for 18 to 25 nm viruses (that is, parvoviruses and hepatitis A. Where a specific virus size, LTR, and test condition are required, typical virus retention for a specific filter generally can be predicted through a combination of the manufacturer's validation data — often generated with bacterial viruses (bacteriophage) of known size in model fluids — and published data on retention of various mammalian viruses in process fluids. Manufacturers' ratings are useful in membrane selection for further performance evaluation under product and process-specific conditions.


Close examination of a filtration system's performance requirements, among other criteria, should determine selection. High removal efficiency (>5 to 6 LTR) has been documented with virus filters for numerous mammalian viruses >40 to 50 nm in size, including: vaccinia virus (250 nm x 300 to 450 nm), herpes simplex virus (120 to 300 nm), influenza virus (80 to 120 nm), murine leukemia virus (80 to 128 nm), HIV (80 to 100 nm), Sindbis virus (40 to 70 nm) and SV40 (40 to 55 nm). Filters with finer pores (nominally 15 to 20 nm) have demonstrated removal efficiencies >3 to 4 LTR for small viruses (18 to 30 nm) such as porcine parvovirus, human parvovirus ±9, minute virus of mouse MVM (or minute mouse virus, MMV) and poliovirus. Table 5 shows typical viral clearance data in various protein solutions for virus filters nominally rated at 50 and 20 nm. Using two filters in series (double filtration) can further enhance clearance efficiency for small viruses.


Maximized protein recovery is a desired goal of any viral clearance strategy. Many viral retention filters are inherently hydrophilic or are hydrophilized through surface modifications to decrease protein binding and enhance product transmission. Transmission of over 95% of plasma-derived and monoclonal immunoglobulin G (IgG, ~ 160 kDa) has been reported through various virus filters, using concentrations up to 5%. Protein transmission effectiveness depends on the degree of protein aggregation, solution purity, and whether a protein gel layer forms on the membrane surface. As some product loss may be attributable to membrane adsorption or holdup within the filter assembly, flushing the filtration system with a buffer may enhance product recovery.


The validation of filtration processes requires special attention, as both the filter manufacturer and the end user play a vital role in assuring specified performance. The filter manufacturer must initially qualify the membrane and filter construction, then ensure each filter will meet that specification. The filter user must demonstrate that each filter satisfies the process needs and their validated viral clearance claims.

Viral filter retention validation studies provide two kinds of clearance. Virus-specific clearance is the direct evidence that the production process will effectively remove viruses that are either known to contaminate the starting materials or that can conceivably do so. General virus clearance is indirect evidence that the production process can remove potential novel or unpredictable viruses. Proper design of the validation study is critical to ensure its success. Usually, the retention study uses a scaled- down version of the full-scale process with small area membrane discs or modules. Some important factors in the study design include: choice of viruses; target reduction factor; virus titer; comparability of test feedstock to process feedstock with respect to concentration, temperature, chemistry, purity, and accuracy of scale-down model; volume-to-surface area ratio for the test and process filter; and inclusion of proper study controls.

Figure 3. Scale-up of the Depth Filtration Step Using Feed Containing 0% Solids


Several product and process parameters may affect microbial retention, including viral retention by filtration. Product parameters include pH, viscosity, surface tension, ionic strength, and osmolarity. Process parameters include batch size, temperature, time, pressure differential, and flux (flow rate per unit area). Details of these product and process parameters are outlined in PDA Technical Report No. 26, "Sterilizing Filtration of Liquids."


Although this report focuses on retention of bacteria by "sterilizing grade" filters, many of the same considerations apply to viral retention by virus filters. When designing the validation protocol, it is important to address the effect of extreme processing factors on the filter capability; the filter validation should be conducted using worst-case conditions. In a virus filter retention validation study:

  • Maximum pressure differentials should be incorporated in model challenge conditions. A high-pressure differential might reduce viral retention. This should be the actual maximum differential across the test filter, not the total available system pressure. The pressure differential across the test filter during the validation challenge should meet or exceed the maximum pressure differential observed during processing.

  • Maximum flux (flow/area) should be used for the model challenge condition. High flux might reduce viral retention. It may not be possible to mimic pressure differential and flux simultaneously during validation studies. The user should determine which is more relevant to the specific process and develop a rationale to support the decision.

  • The longest potential processing time should be used for filter viral validation testing. Several factors related to process time may affect retention by membrane filters. These include filter compatibility, maintenance of integrity, any changes in bulk fluid during challenge, hold times, and time-dependent penetration.

  • The volume/area throughput should be at least as high as planned for the scaled-up process.

  • The lowest adsorption conditions should be used. If several versions of the same basic product are planned, virus adsorption might be affected by product composition. Product specific assays may be needed to test the lowest adsorption conditions. In general, formulations with higher ion strength and higher protein concentrations tend to reduce adsorption of viruses to membranes. However, in some cases, higher protein concentrations can enhance virus/protein interactions leading to virus aggregation.

  • The viral filter retention testing temperature should be close to the process temperature. Excessive high or low temperature variance might affect filter performance and, therefore, viral retention.

It may not be justified to test each of these variables independently. A parametric approach considers possible interactions, testing a true worst-case condition. In addition to validating viral retention, the filter's capacity, compatibility, absorption, and extractables must also be validated. These concerns are comparable to validation of sterilizing filters, as outlined in PDA Technical Report No. 26, "Sterilizing Filtration of Liquids."4

Table 5. Typical Viral Clearance with Virus Membrane Filters


FDA's 1987 "Guideline on Sterile Drug Products Produced by Aseptic Processing" states, "After a filtration process is properly validated for a given product, process, and filter, it is important to assure that identical filter replacements (membrane or cartridge) used in production runs will perform in the same manner. One way of achieving this is to correlate filter performance data with filter integrity testing data."


There are two types of physical integrity tests in use today: destructive and nondestructive. Destructive tests involve challenges with particulates (for example, particulate gold colloids) and are therefore limited to post-use application. In contrast, nondestructive tests based on liquid porosimetry or gas diffusion can be used both pre- and postfiltration. A nondestructive test recommended by several virus filter manufacturers is the "diffusive," or forward-flow test that quantitatively measures the diffusive flow of pressurized air or nitrogen (plus the bulk flow through any open or nonwetted pores) across a wetted membrane filter. Test sensitivity is enhanced with elevated gas pressure and reduced wetting liquid surface tension. Figure 4 shows data correlating virus retention to a forward flow diffusion limit value for a virus filter. Forward flow measurements lend themselves ideally to automation and can be performed readily with test instruments commonly used in the biopharmaceutical industry for similar integrity tests on 0.2 and 0.1 µm rated "sterilizing grade" filters. Liquid porosimetry can also be conducted as a nondestructive test but requires cleaning after pre-use testing and before post-use testing and is less easily automated.

Figure 4. Virus Filter Integrity Test Correlation Data*

While there is common agreement that filters should be tested both before and after filtration, postfiltration integrity testing is often a regulatory requirement for product release. Because filter integrity tests must correlate with virus removal claims, performing such tests is a critical safeguard for biopharmaceutical manufacturers. In contrast to the product-specific viral clearance data generated by the user with scale-down filter models, the filter manufacturer develops conditions and data showing a correlation of virus retention by process-scale filter elements to a particular integrity test method and limit value. Often, this data is not considered by regulatory reviewers or inspectors, so users must understand the basis of a filter manufacturer's integrity test and its correlation claims, and they must determine if the test's sensitivity is adequate to detect even a marginal lack of filter integrity. Once installed in a process, the filter integrity test is the sole means of confirming validated virus clearance performance.


1. A Wang, L Russell, T Tressel, A S Rathore, ACS Poster # 263, 227th ACS National Meeting, Anaheim, CA, 2004.

2. Millipore Application Note, Filter Sizing Methods- for Normal Flow Filtration Applications, Document # AN1512EN00.

3. D Uavorsky, S McGee, Selection and Sizing of Clarification Depth Filters, Genetic Engineering News, Vol. 22, No. 9, May 1, 2002.

4. Sterilizing Filtration of Liquids, Technical Report No. 26, PDA, 1998.

5. Note of Guidance on Virus Validation Studies: The design, contribution and interpretation of studies validating the inactivation and removal of viruses, CPMP EMEA, 1996.

6. Viral Safety Evaluation of Biotechnology Products Derived from Cell Lines of Human or Animal Origin, ICH Harmonized Tripartite Guideline, 1997.

7. Guideline on Sterile Drug Products Produced by Aseptic Processing, FDA, 1987.