Optimization, scale-up, and validation ISSUES in FILTRATION of Biopharmaceuticals, Part II

September 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-09-01-2004, Volume 17, Issue 9
Page Number: 40–45

Filtration is one of the most commonly used unit operations in the manufacturing of biopharmaceuticals. This is the second part of the fourth article in the "Elements of Biopharmaceutical Production" series. In this second segment, Manoj Menon and Frank Riske present an approach for the development and optimization of a TFF application, followed by a contribution from Jennifer Campbell and Elizabeth Goodrich reviewing key issues involved in validation of a TFF step.

Filtration is one of the most commonly used unit operations in the manufacturing of biopharmaceuticals. This is the second part of the fourth article in the "Elements of Biopharmaceutical Production" series. In this second segment, Manoj Menon and Frank Riske present an approach for the development and optimization of a TFF application, followed by a contribution from Jennifer Campbell and Elizabeth Goodrich reviewing key issues involved in validation of a TFF step.

Anurag S. Rathore

Manoj K. Menon and Frank J. Riske, Genzyme Corporation

Process Development for a Tangential Flow Filtration Step

Tangential flow filtration (TFF) is used extensively in the biopharmaceutical industry for harvest clarification, protein concentration and diafiltration, and viral clearance filtration. A typical TFF system is shown in Figure 1. The feed stream to be processed is recirculated across the upstream surface of the porous membrane by the recirculation pump. The flow generates a pressure difference across the membrane, with the average trans-membrane pressure (TMP) given by:

where P1 and P2 are respectively the pressure values at the inlet and outlet of the retentate stream, while P3 and P4 are pressure values at the corresponding points on the permeate side. TMP can be increased by partially closing the retentate valve (V1), which results in an increase in the retentate pressure (P2), and a corresponding increase in feed pressure (P1). Increasing recirculation flow rate raises the feed pressure (P1) and also results in an increase in the TMP.

As the fluid is forced through the membrane pores, particles retained by the membrane accumulate at the membrane surface, a phenomenon known as concentration polarization (Figure 2). The accumulated particles may form a dense particle cake or "gel layer" on the membrane surface thereby increasing resistance to flow. The high concentration of particles may also cause membrane fouling and reduce permeate flow. The shearing action of the flow across the membrane in TFF sweeps the retained particles away from the membrane surface, reducing the extent of concentration polarization and increasing the permeate flow. Increasing the shear rate (by increasing recirculation flow or reducing the channel height) can further reduce the concentration polarization (Figure 2). However, the higher shear rates cause a larger pressure-drop across the retentate channel — P1-P2 in Figure 1 becomes larger — increases the power required for recirculation, and may cause lysis of cells or denaturation of proteins. A key aspect in the design of TFF systems is choosing a recirculation flow rate that provides high product recovery at an acceptable operational performance, as characterized by step recovery, process time, pool quality, filter capacity, and minimal overall cost.

Membrane Screening

Due to the complex interaction of various factors that determine performance, it is difficult to predict which membrane type will perform best for a given application. Scale-down experiments with a representative feed stream will need to be performed before the membrane choice can be made. Select candidate membranes for evaluation based on the following guidelines:

  • Limit choices to products from manufacturers with proven manufacturing expertise and reliable quality control systems.

  • Flat-sheet cassettes with open channels and wide ports are preferred for feed streams with high levels of suspended solids — for harvest clarification — or a shear-sensitive product, while modules with narrow channels and turbulence promoting screens or hollow-fiber systems are preferred for applications with low or no suspended solids — ultrafiltration for protein concentration.

  • Pick membranes with an average pore size significantly smaller than the size of the retained species in order to minimize membrane fouling by pore plugging and to ensure the retained species (in many cases your product) remains on the upstream side of the membrane.1 Test different pore sizes and select the largest possible pore size that does not result in significant fouling during the process to maximize process flux and does not allow target retained species to pass through the membrane.

Membrane Evaluation

The experimental set-up should mimic the large-scale installation in its essential elements (Figure 1). Choose the smallest membrane module that is a true linear scale-down version of the full-scale modules with identical flow-path length, channel height, and ports.


Use a lab-scale rotary lobe pump as the recirculation pump since these pumps produce relatively low shear and provide constant flow rate against varying pressure and are typically used in production-scale systems. Peristaltic pumps, however, produce flow with significant pressure fluctuations, especially when operated at high pressure, and the results obtained in the lab may not be readily scalable. It is useful to install flow meters on the retentate and permeate streams. Simple in-line rotameter-type flow meters (with a suspended float and a transparent tube with a variable area of cross-section) are effective for this purpose except for very turbid feed streams.

Figure 1. Schematic Illustration of a Tangential Flow Filtration System

Determine the volume of product required per experiment based on estimates of membrane flux (see "Calculation of Product Volume") and obtain the estimated volume of representative feedstream. Assemble the membrane cartridge and measure the initial water permeability as recommended by the manufacturer. The membrane manufacturer usually provides a range of recirculation flow rate for a given type of cartridge. Run the recirculation pump to generate a flow rate in the middle of the range recommended by the manufacturer. Partially close the retentate valve (V1 in Figure 1) to generate a retentate back-pressure (P2 of 5-15 psig for microfiltration, 20-40 psig for ultrafiltration). Keep the permeate valve (V2) open for the initial experiment. Monitor permeate flow rate during the experiment by collecting permeate in a vessel placed on a balance. Collect samples of permeate and retentate at intervals to assay for product concentration. As the operation proceeds, the volume of material in the feed tank will decrease, and the concentration of retained species (in this case, cells and cell debris) will increase. Continue processing until the feed volume is close to the system hold-up volume or until the permeate flow drops to less than 25% of the initial value. Obtain a sample of the permeate pool and the retentate at the end of the process. Drain the system and rinse and clean the membrane as recommended by the manufacturer. Measure the water permeability after cleaning. Calculate the average permeate flux. Evaluate product transmission (ratio of product concentration in permeate to that in the retentate) at different points in the run and plot against product load (L/ft2) as shown in Figure 3.

Significant fouling of the membrane will be indicated by a sharp drop in product transmission and permeate flux. Figure 3 shows the product transmission during the clarification of cell harvest for two types of membranes. Product transmission drops off more quickly for membrane A compared to that for membrane B, suggesting that the product may be binding to membrane A or that significant membrane fouling occurs with membrane A .

Calculate overall product recovery as:

Measure key quality attributes of the product in the permeate pool. For example, if the product is known to aggregate, fragment, or lose biological activity under high shear, compare these attributes in the feed and permeate samples to ensure that the product is not adversely affected.

Compare the water permeability of the membrane after cleaning to the initial water permeability. The permeability typically drops 10 to 20% after the first run and levels off at 60 to 80% of the initial value after subsequent runs. A drop in the permeability of more than 40% or a continued drop in permeability after every run suggests that the cleaning procedure is not adequate.

Figure 2. Effect of Shear Rate on Concentration Polarization in TFF

If the initial experiment results are promising, repeat the experiment with a permeate pump to control the permeate flow 10 to 20% higher than the average permeate flow obtained in the previous experiment. Operation at constant permeate flux typically results in a higher average permeate flux and better performance compared to a constant TMP operation, since in the latter case, the high initial flux causes a large concentration polarization that may cause increased membrane fouling.

For constant permeate flux operation, the retentate side pressures (P1, P2) stay nearly constant, while the permeate side pressures (P3, P4) decrease as the operation proceeds and the membrane becomes fouled. This results in an increase in TMP over the course of the process. Perform the constant permeate flux experiment at the low and high ends of the recirculation flow recommended by the manufacturer. At higher recirculation flow rates, rise in TMP is slower due to reduced concentration polarization. Figure 4 shows the effect of recirculation flow rate on the TMP profile for a scale-down microfiltration process for cell harvest clarification. The results show TMP rises rapidly for operation at recirculation flow rates below 2.3 L/min. The overall recovery also dropped significantly for operation at recirculation flow less than 2.3 L/min due to increased membrane fouling at lower recirculation flow.

Figure 3. Effect of Membrane Type on Product Transmission

Perform similar experiments with other candidate membrane systems using the same feed stream and product loading used in the experiments above. Compare product transmission, product quality attributes, average permeate flux, recirculation flow rates, cleaning procedures, and recovery in water permeability after cleaning. The critical deciding factors are product recovery and quality. If these factors are similar for two types of membranes, consider the operational parameters. Calculate the membrane area and recirculation flow rates required at scale based on the average permeate flux, and product loading determined from the above experiments. Choose the system that requires lower capital costs — lower membrane area, and smaller system size — and lower operating costs — lower recirculation flow, more membrane cycles before replacement.

Figure 4. Effect of Recirculation Flow Rate on TMP Profile for Microfiltration of Cell Culture Harvest Using a 4 ft2 TFF Membrane

Once the membrane and operating parameters are chosen, repeat the experiments with several lots of feed material to ensure that the process is robust and produces a high quality product with overall recovery at or above the target value.

Jennifer Campbell and Elizabeth Goodrich, Millipore Corporation

Validation of a Tangential Flow Filtration Step

This section describes how validation is linked back to the choices made during process development and scale-up and will highlight important validation considerations for a large-scale TFF process. The items that must be validated in a TFF process can be divided into three separate categories — process characterization, process validation, and cleaning validation. In addition, the suitability of the industrial-scale system to manufacture product must be verified. This article will highlight the important considerations within each of these four segments of a TFF validation plan.

Process Development

The output of process development is a proposal for a system containing a particular membrane and membrane area, controlled to specific setpoints for crossflow, pressure, and temperature (within some reasonable ranges), for processing of a particular volume of a product stream with fixed buffers, cleaned per a set protocol and re-used up to a specified number of times. During the development phase, this particular combination of physical components and operating strategy has been shown to result in a process that meets the success criteria for yield, cycle time, robustness, and economics and final product that meets the success criteria for quality and purity. It is this completely defined process, once it has been accurately scaled up and the suitability of the system design has been verified, that must be validated through an overall validation scheme that includes process characterization, process validation, and cleaning validation.

Process Scale-Up and Verification of System Design Suitability

Although not explicitly part of validation, proper scale-up and design of the manufacturing system is critical to ensuring that the process results achieved at small scale can be reproduced at industrial scale without encountering problems. Again, the success criteria must be considered. Not only must the manufacturing system be able to provide the flow rate, pressure, temperature, and volume handling capability as specified by the development work, but it also must not have any adverse effects on the product yield, quality, purity, or consistency. Specific areas of system design that should be carefully considered for the potential for adverse effects are: minimization of working volume, thorough drainability, proper flow distribution, the choice of recirculation pump, any points of air-liquid interface, adequate mixing, minimization of deadlegs, and compatibility of materials and extractables from all fluid contacting materials with any product, buffer, or cleaning stream.

One area where system design explicitly falls under the validation umbrella is in the validation of extractables from any wetted components in the product stream. All fluid contacting materials from every component in the large-scale system must be compatible with all product, buffer, and cleaning solution streams and must not add contaminants to the product. This is of particular importance if the TFF unit operation is near the end of the purification process. Equipment vendors for the biopharmaceutical industry generally ensure that their materials of construction meet USP requirements for Class VI Biological Tests for Plastics and are non-toxic per USP General Mouse Safety Test.4 In addition, effluent from any filters must test negative for USP oxidizable substances after an appropriate flush volume. In cases where vendors do not have appropriate test results available, extractables validation is most easily carried out on individual components as opposed to the full-scale system. However, depending on the design of carryover studies, which will be discussed later, it is possible to include any component extractables testing with carryover results.

Process Characterization

Process characterization work is typically carried out at small scale in order for the labor and feedstock requirements to be reasonable. This requires that a small-scale system that is scaleable to the full-scale process system is available and gives reliably representative results. These experiments are done to establish a window of acceptable operation for each critical process control setpoint. During process development, setpoints are determined for certain parameters that give acceptable process results. For a TFF process, these can include, but are not limited to, crossflow rate or pressure drop, transmembrane pressure or retentate pressure, process temperature, filtrate flux (more common in microfiltration than ultrafiltration), total volume of feedstock processed per area of membrane, number of diavolumes, and diafiltration buffer composition. In full-scale manufacturing, it is unlikely that all parameters will remain precisely at their setpoint value throughout the course of a process run or from one run to the next. Therefore, in order to create a manufacturing procedure that is both feasible to operate in a plant environment and one that results in good product, an appropriate range for the process parameters must be determined.

The characterization experiments should focus on the parameters that are most critical to process performance and most difficult to control to a precise setpoint at industrial scale. For parameters where it is determined that process characterization is required, the range of a particular parameter to investigate should be set based on the range expected during full-scale operations (bearing in mind manufacturing capability). It is also based on the range that is desirable to claim in a license application that will allow for any future process changes to be implemented.

Calculation of Product Volume Required for Scale-Down TFF Experiments

Process Validation

Process validation, a second segment of the validation strategy, demonstrates that the process that was developed and defined at small scale operates successfully and reproducibly at manufacturing scale. Typically, process validation is achieved by performing several (often three) qualification lots, using feedstock that runs sequentially through all steps of the process and, ideally, using the equipment intended for manufacturing. Product material and process data from the runs are analyzed to demonstrate that all critical process parameters are consistently controlled to their setpoint, that processing times are consistent and within specification, that bioburden is consistently low and controlled at all sampling points, and that the final material is consistent from run to run and meets all yield, quality, and purity specifications. In addition to the actual intended process, there are other factors associated with a TFF step that are important in a manufacturing environment, such as allowable hold times and temperatures for the feedstock and product pools and any allowable re-work steps. Although it adds extra work to a validation strategy, validating a range of hold times or rework possibilities allows additional processing freedom at large scale.

Cleaning Validation

Cleaning validation is the third section of an overall validation plan and is the section that historically has received the most attention. The overall goals for cleaning validation are to demonstrate that 1) the membranes are returned to a state where the process will perform reproducibly each run, 2) contaminants are adequately removed to prevent run-to-run carryover, and 3) bioburden level is controlled. These goals must be met for the full-scale system, including membranes and all associated hardware. In addition, cleaning validation must establish the number of times the membranes can be cleaned and reused in a given process. While much of the cleaning validation is dependent on the specific equipment used for manufacturing and therefore must be carried out at full scale, anything related to membrane re-use is primarily dependent on the membranes and can be performed on a small-scale, representative system.

The first goal of cleaning validation — demonstrating that the membranes are returned to a state where the process will perform reproducibly each run — can be met by tracking process performance as well as membrane data from run to run. Typical measures of process performance are process flux, process time, product yield, and product purity. While these measurements tie most directly to process success and are the most relevant, they require that product feedstock be committed to the cleaned membranes without knowing whether the membranes have yet reached their reuse limit. Therefore, in addition to process data, it is desirable to also track clean membrane data — most typically membrane integrity and clean membrane permeability. Membrane integrity should be monitored before and after each use, and the vendor's specification for membrane failure should be used as a criterion for replacing the membranes. Clean membrane permeability — water flux divided by transmembrane pressure (TMP) — should be measured and trended from run to run. Although a drop in clean membrane permeability may not be as directly related to process performance as the actual process flux, it is a simple measurement that can be an early indication of degradation of membrane performance. The reduction of clean membrane permeability to a certain percentage of the initial value may be used as a membrane change-out criteria. However, it is important to note that measured permeability values are often related to the system on which the measurement is taken, especially for high permeability membranes (ultrafiltration > 30 kDa and all microfiltration membranes) where measured TMPs are very low and often within the measurement error. Differences in the placement of pressure transmitters in relation to the membranes and differences in pressure loss in piping also add to variations in permeability measurements from system to system. For these reasons, a change-out criterion based on a permeability measurement should only be set when the data is collected on the same system on which it will be applied.

The second goal of cleaning validation is ensuring that contaminants are adequately removed to prevent run-to-run carryover. Carryover studies can either be performed between batches using placebo product, or by testing rinse solutions or extraction solutions. The placebo product, rinse solution, or extraction solution is analyzed for residual product protein, other host cell proteins, DNA, excipients, and residual cleaning and storage reagents. Care should be taken not to set the specification for any contaminant level as the limit of detection (LOD) of the assay, as LODs often decrease as assays improve with new technology. Carryover studies must be performed out to the established limit of membrane reuses. Membrane re-use validation and carryover validation are often done concurrently at full-scale during production in order to minimize cost and effort. However, care must be taken to verify that product quality is not compromised while running a process for which the validation is still open. This can be done by putting in place appropriate sampling, analysis, and quarantining of final product until it is shown to be acceptable. In addition to run-to-run carryover, users of TFF membranes must also demonstrate that the preservative (shipping) solution is completely removed from the devices after a specified flushing protocol prior to the first use with product. The assay showing clearance of the preservative or storage solution should be easy to use, and validation of an assay which can be run on the manufacturing floor will save processing time versus having to submit samples for QC testing and waiting for results before beginning a process run.

Finally, the third goal of cleaning validation is to demonstrate that bioburden and endotoxin levels are kept under control using the specified cleaning protocol. This is of particular importance if the TFF unit operation is near the end of the purification process. Sterility requirements and maximum endotoxin levels should be specified — often not just for the final processed product pool but for incoming buffers and product as well. Bioburden and endotoxin elimination from the membranes and system using the specified cleaning reagents (concentration, temperature, and exposure time) must be documented. In addition, storage solutions should be evaluated for bacteriostatic capability over the intended storage time.

Special considerations surround the cleaning validation of a microfiltration system due to exposure to cells. Whether bacterial, yeast, or mammalian, the cleaning of the system must be adequate to show inactivation and removal of all cells and cell debris between runs. The biggest validation issue surrounding cell harvest is the containment of recombinant organisms and equipment decontamination. If a claim is made that the filtrate from the TFF system will be free of recombinant organisms, then protocols should specify testing of the filtrate for such.


Prior to validating a TFF process, it is important to ensure that the process has been adequately defined during process development and accurately scaled up onto a well-designed system. If these criteria have been met, the task of executing the overall validation strategy — which should include process characterization, process validation, and cleaning validation segments — becomes much more straightforward with a high probability of success.


1. Russotti G, Goklen KE. Crossflow membrane filtration of fermentation broth. In Wang WK, editor. Membrane separations in biotechnology. New York: Marcel Dekker, Inc; 2001.

2. van Reis R, Goodrich EM, Yson CL, Frautshchy LN, Dzengeleski S, Lutz H. Linear scale ultrafiltration. Biotechnology and Bioengineering 1997; 55:737-746.

3. Rudolph EA, MacDonald JH.Tangential flow filtration systems for clarification and concentration. In Lydersen BK, D'Elia NA, Nelson KM, editors. Bioprocess engineering: systems, equipment and facilities. New York: John Wiley and Sons; 1994.

4. USP. USP XXIII; Section 88: Biological reactivity tests, in vivo. Rockville (MD): USP.