Articles

Issue PDF

Essential Steps to Ensure Accurate Filter Scale-up

BioPharm International, September 2009 Issue (PDF)

Risk assessment exercises seem pretty simple. Make a list of everything that could go wrong. Quantify those risks, by assigning scores. Then decide if you need to do more to prevent the high-risk events from happening. Tedious, but not complicated. 

The recent CMC Strategy Forum on quality risk management, however, revealed the challenges involved. For example, how do you deal with the subjectivity of scoring? And how do you score a potential harm based on risk to patients, especially when you're analyzing upstream steps, where almost any risk would be mitigated before the product reached the patient? Group dynamics also can present challenges. You have to ensure that all voices are heard, and that the group really prioritizes risks, rather than just ranking everything high. That means finding a skilled facilitator, or training one. 

None of the challenges is insurmountable, of course, and the discussion at the Forum revealed many ways to handle them. But then one must consider the substantial resources involved. The process requires many hours of meeting time from expert staff, and someone has to manage it all. Many large companies are creating new risk management positions and even departments.  

Given the size of the task, I start to wonder whether it is truly necessary or worthwhile. Clearly, some sort of risk management is essential to ensure product safety and efficacy. But many would say that the industry has always done quality risk management, without calling it that. It was how one justified specifications.  

So if quality risk management is not really new, the question becomes, Is it necessary, or beneficial, to 

That's hard to say. Proponents cite various benefits. For example, the process documents institutional knowledge, as well as upper management's awareness of risks. It gets staff from different departments to talk to each other. You also may be able to use the data in regulatory filings for site and process changes. 

All that is nice, but the real value of formalizing risk management will be if it improves product quality. Yet most companies, I think, would assert that their drug product quality is already excellent; manufacturing processes are robust, and quality control measures ensure that work-in-process or final product that doesn't meet specifications is rejected.  

Assuming that's true, "improving quality" really translates to two things: a) minimizing waste by reducing deviations, lot rejections, and investigations, and b) averting a major—but so far unforseen—quality disasater that could harm patients. The latter, I have to assume, is the core goal of risk management.  

To truly achieve the goal of preventing disasters, risk assessment exercises will have to identify risks that previously had not been considered or taken seriously. Because as several people pointed out during the Forum, if these discussions only identify "the usual suspects," the process will not add much value.  

So if companies are going to take the time and effort to implement a new risk management program, they should focus on the dialogue—that conversation among experts and across departments that so many say doesn't otherwise happen, or happen often enough. And they should make sure that in those discussions, people really question their assumptions, and don't take a a "been there, done that" attitude.  

Otherwise, risk assessments will devolve into an exercise to please regulators, or just a paperwork drill, as many people say sometimes happened with process validation. In that case, we might as well continue doing things the old way. 

From the Editor

If risk assessments only identify "the usual suspects," the process will not add much value.

A True Assessment of Risk

A novel Protein A alternative stationary phase based on the variable heavy chain fragment of immune camelid antibody was evaluated and compared to commonly used commercial Protein A resins. The parameters evaluated were elution pH, equilibrium isotherm, dynamic binding capacity, and host cell protein clearance using a set of Chinese hamster ovary-derived monoclonal antibodies and Fc-fusion proteins. Linear retention experiments were used to compare the specificity of these resins for both non-IgG model proteins as well as antibodies and Fc-fusion proteins. The experimental results showed that the new camelid antibody resin behaved very similarly to Protein A resins in terms of retention of non-IgG model proteins and IgG-based molecules. Dynamic binding capacity was found to be comparable for Fc-fusion proteins and slightly lower for antibodies. Host cell protein clearance profiles were also similar under preparative conditions using complex biological feeds. Finally, the binding mechanism was explored by using different mobile-phase modifiers in linear pH gradient retention experiments. 

Protein A affinity chromatography has been widely used for antibody purification in the biopharmaceutical industry because of its excellent selectivity and product recovery.

  In recent years, it has been recognized as the industry standard for capture and purification of antibodies and Fc-fusion proteins. The use of this highly selective and robust capture step allows for faster process development and has enabled the use of a platform approach for monoclonal antibody (MAb) purification.

   Despite all of its advantages, Protein A chromatography suffers from the limitations of high cost, ligand leaching, and caustic instability.

In the last decade, several mixed mode and Protein A mimetic ligands have been developed as alternatives to Protein A chromatography. One is hydrophobic charge induction chromatography (HCIC), which uses heterocyclic ligands at high ligand densities that can get positively charged at low pH values.

  Similar to Protein A, adsorption on these resins can occur by hydrophobic interactions without high salt concentration, while elution can be controlled by lowering the pH to induce charge repulsion between the ionizable ligand and the bound protein.

  Protein A mimetic ligands were also developed based on the IgG binding domain of Protein A using techniques such as molecular modeling, protein engineering, phage display, and synthetic chemistry.

  Although initial studies from the resin manufacturers had shown some promise for the above-mentioned alternatives,

  recent and more comprehensive studies have shown that none of these resins possess the selectivity offered by Protein A chromatography.

Recently, a novel technique was developed for the rapid identification of affinity ligands against a diverse set of targets using the variable heavy-chain (VHH) region of single-chain antibodies found in the Camelidae family.

  These molecules possess good specificity because of their enlarged hypervariable region and have very high physical and thermal stability because of their single-domain nature.

  Recombinantly expressed VHH fragments can be used as affinity ligands and have applications in laboratory-scale immunoaffinity and immunoperfusion chromatography.

  Their potential application at industrial scale became more promising after it was proven that these antibody fragments can be expressed efficiently in microorganisms such as the yeast 

This technology has been commercialized by the Bio Affinity Company (Naarden, The Netherlands) to generate ligands (called CaptureSelect ligands) that can be customized for any purification challenge. Unlike other proteinaceous ligands (such as Protein A), these ligands have the distinct advantage of being stable in strongly alkaline solutions.

  One such ligand was generated against the Fc-region of human IgGs. This was shown to bind to all human IgG subclasses and no cross-reactivity was found with bovine or mouse IgG.

  This ligand has the additional advantage of being specific for human IgGs only and unlike Protein A, it can bind IgG3s as well. In 2007, the ligand was immobilized on a highly cross-linked agarose-based backbone through a long, hydrophilic spacer arm and marketed through GE Healthcare (Uppsala, Sweden)  as IgSelect affinity medium.

  Although this resin can potentially be an attractive and manufacturing-friendly alternative to Protein A chromatography, very limited data exists so far on the performance of this new resin.

This article provides the first comprehensive evaluation of this new ligand for MAb purification. Using several industrial MAbs and Fc-fusion proteins, the performance of this resin was compared to two of the most commonly used commercial Protein A resins: MabSelect from GE Healthcare and ProSep-vA High Capacity from Millipore (Billerica, MA). These two Protein A resins are on two different backbones (agarose versus controlled pore glass) and were specifically chosen to represent a wide spectrum of Protein A resins. The parameters evaluated were binding affinity, dynamic binding capacity, selectivity, and binding thermodynamics. Finally, the binding mechanism of IgSelect resin also was explored by linear gradient retention experiments with different mobile-phase modifiers. 

The three resins evaluated in this study (MabSelect, IgSelect, and Prosep-vA High Capacity) were purchased from their respective vendors. All resins were packed to 20 cm bed height in 1 cm I.D. columns made by GE Healthcare. The four test proteins—Fc-fusion proteins (A and B) and MAbs (C and D)—were expressed in Chinese hamster ovary (CHO) cells and produced at Bristol-Myers Squibb (Syracuse, NY). Model proteins such as horse cytochrome c and human serum albumin (HSA) samples were purchased from Sigma-Aldrich (St. Louis, MO). All other chemicals were purchased from Mallinckrodt Baker (Phillipsburg, NJ).  

All chromatography experiments were carried out on an AKTAexplorer chromatographic system from GE Healthcare. High Performance Liquid Chromatography (HPLC) analysis was performed using Waters (Milford, MA) 2695 Separation Module and Waters 2996 Photodiode Array Detector. An Orbital Shaker 100 from ArmaLab (Bethesda, MD) was used for batch-adsorption experiments.  

The elution pH of the various proteins was obtained by linear gradient experiments under analytical conditions using pulse injection of the samples. A gradient of pH was run from 6.5 to 2.5 over 10-column volumes in citrate buffer. The elution pH at peak maxima was calculated from the gradient and further verified from the effluent pH trace obtained from the online monitor pH/C-900 that is part of the AKTA system. In the binding mechanism exploration test, different mobile phase modifiers were added to citrate buffers at both pH values and an identical pH gradient was run for comparison with the original pH gradient. 

Dynamic binding capacities were determined by performing breakthrough experiments at six minutes residence time. The column was equilibrated and regenerated using typical Protein A process conditions.  Adsorption isotherms for the proteins on the various stationary phases were determined using batch experiments. 

Selectivity of the three resins was compared under preparative condition using cell culture harvest material. Each resin was loaded to ~80% of its dynamic binding capacity for the respective protein. Typical Protein A equilibration, elution, and regeneration conditions were used for these experiments. Sample protein concentration was determined using an analytical Protein A assay. Host cell protein levels in the samples from the preparative experiments were determined using an in-house host cell protein ELISA assay.  

A Langmuir Isotherm, as defined by the following equation, was used to compare the binding thermodynamics of the three resins in this study:  

 is the equilibrium concentration of the solute on the stationary phase (expressed in mg solute per mL column volume), 

 is the mobile phase solute concentration at equilibrium, 

 is the maximum static binding capacity, and 

 is the affinity binding constant. The binding constant is thermodynamically representative of the protein's affinity to the resin. 

Comparison of Binding Affinity and Elution pH

Both the ligands evaluated in this study (Protein A and the camelid antibody) use a similar operating scheme—the protein is bound at close to neutral pH and eluted with a lower pH. At lower pH values, the electrostatic repulsion between the ligand and the bound protein helps to overcome the attractive forces and hence cause elution.

  The pH of elution in a way can be used as a measure of the strength of interaction with the ligand and is a useful parameter to compare the affinity of various resins. Figure 1 plots the elution pH at peak maxima under linear gradient conditions using a wide variety of proteins. These proteins were carefully chosen to test a variety of possible protein–ligand interactions and included model proteins of varying hydrophobicity (Horse cytochrome c and HSA), MAbs (molecules C and D), as well as Fc-fusion proteins (molecules A and B). In an ideal case, a resin that is selective for IgG-based molecules should only bind those and not any other non-IgG–based model protein.  

Figure 1.  Elution pH of model proteins and IgGs on MabSelect, IgSelect, and ProSep-vA resins 

As shown in Figure 1, the model proteins cytochrome c and HSA had an elution pH of around 6.5 (binding condition as described in the experimental section) on all three resins clearly indicating that they were not retained under these conditions. This is in contrast to results shown for other Protein A alternatives such as HCIC and mimetic ligands that bound model proteins as well as Ig-based molecules.

  The non-binding of the model protein to the IgSelect resin (similar to what is seen for Protein A resins) is indicative of its selectivity. Moreover, the elution pHs for all the IgG-based molecules (antibodies and Fc-fusion proteins) were somewhat comparable across the three resins and fell in the pH range of 3.5–3.9. Thus, it can be said that the new camelid antibody ligand had very similar affinity to IgG molecules as a conventional Protein A ligand under the conditions tested.  

Dynamic binding capacity (DBC) is one of the most important performance parameters for chromatography resins. This is particularly true for Protein-A–based resins, which are very expensive, and are used as a capture step in antibody purification. DBC was calculated for all four antibodies and Fc-fusion proteins using the methodology described in the experimental section. Figure 2 plots the 1% breakthrough capacity on all three resins.  

Figure 2.  Dynamic biniding capacity (DBC, 1%) results of the four IgG molecules on MabSelect, IgSelect, and ProSep-vA resins. The tests were carried out at pH 7.0 and six minutes residence time. 

From Figure 2, it can be seen that the two Fc-fusion proteins (molecules A and B) had very similar capacities on all three resins. Furthermore, the Fc-fusion proteins had a lower binding capacity than the two antibodies (C and D), particularly on MabSelect. This has been noted in the literature and has been explained by the differences in the steric hindrance exerted by Fc-fusion proteins versus MAbs.

  On the other hand, a difference in DBC was seen amongst the three resins for the two antibodies, molecules C and D. Both antibodies showed the highest capacity on MabSelect. The capacity of the new resin IgSelect was less than MabSelect but comparable (or even slightly higher for D) to ProSep-vA. Controlled pore glass resins such as ProSep-vA have been shown to have lower binding capacities but improved mass transport properties.

  The performance of the new resin was in-between the two Protein A resins with respect to DBC. The dynamic binding capacity on all these resins were compared at 6 minutes residence time. These resins might have different responses with varying residence time based on the transport properties of their backbone. However, the response of MabSelect and IgSelect would be very similar because they are based on a similar cross-linked agarose backbone.  

It has been conjectured in the literature that the monomeric VHH fragments might have a lower binding capacity because they do not have the capability to bind multiple IgGs like Protein A, which has multiple binding domains.

  This was found to be true by our experimental results with antibodies, particularly on MabSelect and IgSelect because their backbone and ligand densities are comparable. Interestingly, the single-binding domain did not seem to have any negative effect on Fc-fusion proteins because they have a larger steric hindrance and thereby cannot optimally access the multiple binding domains of the Protein A ligand.  

Figure 3 shows the adsorption isotherms of Fc-fusion protein A at pH 7.0 for all the three resins. Table 1 lists the thermodynamic parameters (

). It shows that amongst the three resins, MabSelect had the highest 

value, which was consistent with DBC results. On the other hand, Q

for ProSep-vA was slightly lower than IgSelect even though their relative trend for dynamic binding capacity was opposite. This can be explained by improved flow properties for the glass-bead–based ProSep-vA resin, which causes its DBC to be closer to its equilibrium capacity. Furthermore, 

 values for all three resins were very similar, indicating that the binding strength for Protein A and this new ligand was very similar.  

Figure 3.  Comparison of adsorption isotherms on MabSelect, IgSelect, and ProSep-vA resins at pH 7.0 

As mentioned before, it has been shown in the literature that none of the small-molecule synthetic ligands developed as Protein A alternatives were able to provide the lock-and-key induced fit as Protein A.

  The non-specific binding of host cell proteins to these ligands have made their selectivity much lower than that of Protein A resins. On the other hand, the VHH fragment has the capability to recognize unique conformational epitope and hence has the promise of showing greater selectivity.  

Table 1. Langmuir Isotherm data at pH 7.0 

To compare the selectivity of the new resin to Protein A, host cell protein clearance was compared for the four antibodies and Fc-fusion proteins using complex cell-culture fluid. The details of these preparative experiments are outlined in the experimental section. Protein recoveries for all of these experiments were comparable and >90%. The elution pools from these experiments were collected and analyzed for host cell protein (Chinese hamster ovary protein, CHOP) levels. Figure 4 plots the CHOP log reduction values (LRVs) from these experiments. The higher the LRV, the more selective is the resin. Figure 4 shows that IgSelect demonstrated comparable host cell protein clearance to the Protein A resins. In fact, for molecules A, B, and D, the LRV values for IgSelect were slightly lower or comparable to MabSelect but higher than the other Protein A resin ProSep-vA. ProSep resins are known to give slightly lower CHOP clearance than the agarose-based Protein A resins because of non-specific interactions of CHOP with their silica backbone.

  On the whole, the selectivity shown by this new camelid antibody resin appears to be very promising.  

Figure 4. Log reduction value (LRV) of Chinese hamster ovary host cell proteins (CHOP) under preparative conditions on MabSelect, IgSelect, and ProSep-vA resins 

The binding mechanism of IgGs on Protein A ligand has been studied in detail by x-ray crystallography and it has been shown that the interactions consist of hydrophobic interactions along with some hydrogen bonding and two salt bridges.

  To explore the differences in the binding mechanisms of Protein A ligand and the IgSelect ligand, linear gradient experiments were conducted with different modifiers in the mobile phase. Molecule A and D (one Fc-fusion protein and one antibody) were used as the test proteins. Only one of the representative Protein A resins (MabSelect) was used for comparison.  

As summarized in Table 2, the results for both molecules with no additive in mobile-phase buffer showed similar elution pH on both MabSelect and IgSelect resins, which proved the reproducibility of the data previously included in Figure 1. Change in the elution pH on addition of mobile-phase modifiers was used as an indicator of the change in the strength of binding. An increase in the elution pH would suggest a decrease in binding strength and vice versa. Ethylene glycol (20%) was used in the mobile phase to test the role of hydrophobic interactions in IgG binding on IgSelect resin because this mobile-phase modifier can reduce the effect of hydrophobic interactions. Compared to the control (i.e., with no additives), both molecules A and D had higher elution pHs on MabSelect and IgSelect resins. This indicated that 20% ethylene glycol weakened bindings on both the ligands to a similar extent.  

Table 2. Elution pH values with different mobile phase modifiers 

A suppressant of surface charge interactions (300 mM NaCl) was added to the mobile phase to investigate the role of electrostatic interaction in IgG binding on the IgSelect resin. Compared to control, both proteins had lower elution pHs on IgSelect resin on adding salt to the mobile phase. In fact, molecule D did not elute from IgSelect even at pH 2.5 under this condition. This indicated that the binding between IgGs and the new ligand was significantly increased after the addition of a medium concentration of salt. In comparison, MabSelect actually showed slightly higher elution pH for both molecules under the same conditions. These results showed that a medium concentration of NaCl affected IgG binding on the two resins in different ways and it required further investigation to determine if the stronger binding on IgSelect was because of decreased electrostatic interactions or increased hydrophobic interactions or a combination of both. 

NaCl used in the previous set of experiments could have contributed to increased hydrophobic interactions while decreasing electrostatic interactions. To explore this effect further, a stronger kosmotropic salt (100 mM sodium citrate) was tested as the third mobile-phase modifier to explore the effect of increased hydrophobic interactions with minimum surface charge shielding. Compared to the data with no additives, 100 mM sodium citrate caused a significant binding increase on IgSelect resin while the effect on MabSelect resin was very small. This proved that salt played a much bigger role in the binding on IgSelect resin than MabSelect resin. When the mobile phase contains medium concentration of Kosmotropic salt or even NaCl, the binding of IgG molecules with CaptureSelect ligand can be significantly increased. In comparison, the Protein A ligand did not seem to be significantly affected by the tested salt concentration.  

Gaining a fundamental understanding of the interaction mechanism on this new ligand requires additional sophisticated analytical techniques (such as x-ray crystallography and docking calculations) and was beyond the scope of this study. The results shown here demonstrate a clear difference in the way Protein A and the new ligand interact with IgGs.  

As the newest alternative to Protein A chromatography, the camelid antibody ligand provided very good selectivity for IgG-based molecules. This resin showed similar binding affinity and somewhat comparable dynamic binding capacity to the commonly used commercial Protein A resins. For antibodies, it did have a lower capacity than the leading agarose-based Protein A resin; however, this shortcoming might be compensated by the vendor by further optimizing the ligand density and appropriate resin pricing. This resin does offer the distinct advantage of being base stable. Different from previously reported mixed mode and Protein A mimetic ligands, this ligand showed as good host cell protein clearance as the Protein A resins used for comparison. Although the new resin has been primarily marketed for IgG3 purification, the results presented in this study demonstrate that this can potentially be used very effectively for industrial MAb purification. Parameters that remain to be evaluated comprehensively before its widespread acceptance and use are column lifetime and viral clearance. Finally, the binding mechanism exploration studies indicated a difference in ligand–IgG interactions between the two ligands. For IgSelect, medium to high concentration of kosmotropic salt should be avoided in the background buffer to maximize antibody recovery during purification. 

Features

A stable alternative to Protein A chromatography.

Comparison of Camelid Antibody Ligand to Protein A for Monoclonal Antibody Purification

The primary reason for the pharmaceutical industry's hesitancy to adopt rapid micobiological methods (RMMs) for environmental monitoring is a lack of understanding of how to assess the cost of implementing such methods. This article explains how to apply a financial cost model to RMMs and presents a case study of three manufacturing facilities of different sizes. For the three facilities evaluated, the year 1 return on investment for implementing an RMM ranged from 183 to 365%, and the payback period ranged from 3.3 to 6.6 months. 

The FDA's Process Analytical Technology (PAT) guidance document describes a regulatory framework that encourages the development and implementation of innovative approaches in pharmaceutical development, manufacturing, and quality assurance.

  One of the key opportunities the document offers is the introduction of new analytical technologies intended to measure, control, or predict product quality and performance. The goal is to design and develop well-understood manufacturing processes and facilitate continuous improvement and operational excellence. Many in the industry regard in-line and on-line physical and chemical technologies as the primary enablers in the PAT toolbox. However, alternative microbiology platforms, such as rapid microbiological methods (RMMs), are now being implemented under the PAT umbrella for ensuring that a state of microbial control is maintained during manufacturing. Although a number of firms have already validated RMMs, it is unfortunate that the majority of the pharmaceutical industry continues to show apprehension about embracing RMMs. Concerns regarding the availability of validation strategies and regulatory acceptance have, for the most part, been satisfied through numerous guidance documents

  as well as by presentations and papers provided by regulatory authorities.

  It is this author's view that the primary reason for the industry's hesitancy to adopt RMMs is a lack of understanding of how to apply financial cost models to economically justify the implementation of these alternative technology platforms. This view is supported by the understanding that manufacturing site heads and production managers want to be convinced that the return on investment (ROI) is sufficient to substantiate the time and expense involved in qualifying and installing RMMs in their facilities. 

There are obvious costs involved with the purchase, qualification, and implementation of RMMs. Depending on the capital expense and the process required to adequately validate a system for its intended use, the costs associated with implementing an RMM can be significant. However, it is important to fully understand all of the financial components that should go into an economic analysis before a decision is made about whether to proceed with a formal qualification program. These components may include the costs associated with the existing method, the costs associated with the initial capital investment of the new method, and the long-term financial benefits (cost savings and avoidances) that the RMM may provide. Dollar amounts for each of these components can then be used to develop a comprehensive economic analysis and business case for introducing the new method.  

The three key steps for considering the implementation of an RMM are as follows: 

1. Review existing conventional methods and recognize potential technology, quality, and business opportunities for implementing an RMM. 

2. Understand the technical and business benefits of RMMs. 

Below, we explain these steps, and provide a case study to illustrate them. 

REVIEW CONVENTIONAL METHODS AND RECOGNIZE OPPORTUNITIES FOR IMPLEMENTING AN RMM

Manual sampling and resource-intensive operating procedures are responsible for much of the cost associated with conventional, growth-based microbiology methods. In addition, the time to result when turbidity is visually detected (in liquid media) or colony-forming units are enumerated (on agar plates) can be quite long. Furthermore, microorganisms that are stressed, damaged, or in a viable but nonculturable (VBNC) state may or may not replicate when cultured on artificial media. In the event of a positive result or an out-of-specification finding, which could occur anywhere from a few days to more than two weeks after the original sample was analyzed, the opportunity to respond to the excursion in a timely manner has been lost. In addition, the effect of an environmental monitoring (EM) excursion on the manufacturing process could be significant, including holding product, rejecting a batch, or shutting down a manufacturing line. In contrast, obtaining these results faster, or even in real time, through RMM, would allow corrective action to be taken quickly, thus reducing the number of lot rejections and plant shutdown time for EM investigations. For these reasons, companies should explore the use of RMMs that may decrease the overall costs associated with conducting microbiology testing while at the same time continuously improving the manufacturing process, finished product quality, and operational efficiency. 

UNDERSTAND THE TECHNICAL AND BUSINESS BENEFITS OF RMMS

When selecting an RMM, it is important to understand the technical and business benefits of implementing the new technology for its intended purpose. Technical benefits may include a significantly faster time to result (or even getting results in real-time); greater accuracy, precision, sensitivity, and reproducibility; single-cell detection; enhanced detection of stressed and VBNC organisms; increased sample throughput and automation; continuous sampling; and enhanced data handling and trend analysis. For RMMs that can provide results in real-time, the response time to an adverse trend or excursion can be immediate, which is not possible for conventional growth-based procedures. Implementing RMMs also may provide a company with numerous business benefits, including, but not limited to, reduced time and costs for product release testing; a reduction or elimination of off-line assays, laboratory overhead, resources, and equipment; a lower cost of product sold; decreased resampling, retests, and deviations; reductions in rework, reprocessing, and lot rejections; and a reduction in plant downtime. 

Once a review of available RMM technologies has been completed, and when one or more technologies are identified that meet the technical and business benefits for the intended application, a business case for implementing the new method should be developed. This will include the use of financial models that can compare the overall costs associated with the current microbiology method with the costs and savings associated with the purchase, qualification, and implementation of the RMM. The goal is to obtain the necessary economic justification that manufacturing site heads will need to initiate the RMM project. Listed below are examples of the types of operating costs and potential savings that can be incorporated into a financial model. 

Potential operating costs associated with a conventional growth-based method:

cost of consumables, regents, and supplies

total sampling, preparation, testing, data handling, and documentation resource time

media, reagents, and consumables disposal costs

laboratory equipment depreciation, calibration, and qualification

overhead for laboratory and storage space

preventive maintenance and service contracts

Potential investment and operating costs associated with an RMM:

The same as for conventional methods, plus:  

capital costs during the initial investment

system qualification and method validation costs

regulatory filing costs, if applicable (changes to in-process microbiological methods or methods that are not specified in an NDA or marketing authorization may not

require a formal regulatory submission to implement the change)

Potential cost savings and cost avoidances associated with an RMM:

reduced testing and finished product release cycle times

reduction or elimination of laboratory equipment and overhead

reduced repeat testing and investigations, lot rejection, reprocessing, and rework because EM excursions can be detected and acted on more quickly

increased yields and lower cost of product sold

reduced raw material, in-process, and finished goods inventory holdings

Each of the identified operating costs and cost savings should have an associated monetary value that can be added up and entered into an appropriate financial model. Two financial tools that can be used to calculate whether there is a financial advantage for implementing an RMM include the ROI and the payback period models. An overview of each model is provided below. 

ROI is the ratio of money gained or lost on an investment relative to the amount of money invested. For RMMs, the cost of performing the conventional method (CM) is compared with the cost (and savings) of using the new method. The information is reported as a percentage and usually represents an annual or annualized rate of return. The ROI is calculated using the following formula: 

The PP is the time required for the return on an investment to "repay" the sum of the original investment. In the context of implementing an RMM, this would be the time (usually in years) required to realize sufficient cost savings to pay for the initial investment of the RMM capital equipment as well as for qualification and implementation activities. The formula used to calculate the PP is the inverse of the ROI formula: 

A CASE STUDY IN THE USE OF ROI AND PP CALCULATIONS FOR JUSTIFYING AN RMM 

A pharmaceutical company operates three parenteral manufacturing facilities using isolators and conventional cleanrooms. The company is exploring ways to implement an RMM that will provide real-time, in-process monitoring capabilities while realizing significant cost savings. After a review of current microbiology methods, the company identified active air monitoring as its most expensive and time-consuming activity, and identified the BioVigilant IMD-A (Tucson, AZ) as a potential replacement for its existing agar-based air sampling procedure. The company initially assumed that it could realize significant cost savings with this particular RMM because the system eliminates manual sampling and laboratory testing, has no consumables, reagents, media, or supplies, and can run continuously during the entire manufacturing campaign.  

The company was also interested in the potential technical benefits of the system, because results are obtained in real-time for both viable and nonviable particles. The company gathered monetary values for each operating cost and cost savings component for use in its ROI and PP calculations, and applied these calculations to the three manufacturing facilities. A brief description of each facility and its EM program for active air samples is provided below, followed by a detailed overview of each financial component. 

1. A small fill–finish facility that processes 40,000 active air samples per year. Manufacturing is performed in conventional cleanrooms. This site rejects one $300K product lot per year because of EM excursions on conventional media and shuts down the line to conduct an EM investigation. 

Manufacturing Economics Series

A case study implementing rapid microbiological methods.

Breaking the Rapid Microbiological Method Financial Barrier: A Case Study in Return on Investment and Economic Justification

The isolation and sequencing of a human gene that encodes a particular domain of a protein is the basis for many inventions in biotechnology. In a recent decision, however, the US Court of Appeals for the Federal Circuit held that claims to a specific DNA sequence were obvious and that it was "obvious to try" known methods to obtain the sequence. 

the invention focuses on DNA molecules encoding a protein known as the natural killer cell activation inducing ligand (NAIL). The patent specification discloses an amino acid sequence of a NAIL polypeptide, the isolation and sequencing of a polynucleotide that encodes a NAIL polypeptide, and the binding relationship between NAIL and a protein known as CD48.    

The Board of Patent Appeals and Interferences rejected the claims over the combined prior art teachings of a patent and a laboratory manual. The prior art patent discloses a receptor protein called p38 that is the same protein as NAIL and further discloses, "[t]he DNA and protein sequences for the receptor p38 may be obtained by resort to conventional methodologies known to one of skill in the art." The Board concluded that one of ordinary skill in the art would have recognized the value of isolating NAIL cDNA and would have been motivated to apply conventional methodologies, such as those disclosed in the laboratory manual and used in the patent. 

The Federal Circuit affirmed the Board's determination that the claims were unpatentable as obvious over the prior art. The Court held that the record supported the Board's factual determination that a person of ordinary skill in the art would have been motivated by the prior art to make a cDNA encoding the protein at issue given the prior art's teaching that the protein played a role in the immune response. The Court further sided with the Board that there would have been a reasonable expectation of success given the prior art's teaching of how to use a monoclonal antibody specific to a protein to clone the corresponding gene.   

The Court rejected Kubin's argument that the prior art did not teach the claim's requirement that the "polypeptide binds CD48." The Federal Circuit effectively overruled, at least in part, its 1995 holding of 

, that claims to a genus of polynucleotide sequences coding for a known protein are not necessarily obvious. The Court concluded that "[i]nsofar as 

 implies the obviousness inquiry cannot consider that the combination of the claim's constituent elements was 'obvious to try,' the Supreme Court in 

unambiguously discredited that holding."   

, rendered just a week earlier, represents an ongoing trend that makes patenting classic biotechnology inventions more challenging. In 

, the Court held that a comprehensive reference listing of every relevant sense oligonucleotide in a known nucleic acid sequence anticipates claims to specific antisense sequences. In the opinion's dictum, the Court appears to suggest that the patentability of claims to specific antisense sequences, without more, may be precluded for any known gene. More than ever, successful patenting strategies require careful planning and taking into account nuanced legal standards to ensure creation of a strong specification and prosecution record. 

 is a partner at McDermott Will & Emery LLP, San Diego, CA, 858.720.3320, 

Final Word

Recent patent rulings raise significant patentability questions for DNA sequence inventions.

Patenting Biotechnology Inventions Will Become More Challenging

Filters of small areas are useful as models for larger operations such as batch-scale processing. Flat disc 47-mm microporous filters are often used for this purpose, especially in flow decline (throughput) studies aimed at sizing the effective filter area (EFA) required for large-scale production. 

In the flow decline method of filter sizing, also known as the 

method, the quantity of effluent produced using a given small filter area is extrapolated to determine the filter area needed to process the drug volume of an entire production batch. It has customarily been considered convenient and economically useful, both in terms of effort and material costs, to conduct such sizing studies with 47-mm flat discs, because their small size minimizes the quantity of fluid involved and the operational time required for the assay. 

If desired, however, discs of an even lower EFA may be used. The smaller the assay filter, the less product is consumed in the testing. This can be an important consideration if the fluid being tested is an expensive drug. Because pharmaceutical filtration is a technico-economic enterprise, the expenses involved are reflected in the drug's cost of goods. 

Figure 1.  Variation in unspecific adsorption (mg per 10-inch cartridge) when different filter designs are used with the same filter polymer 

Regrettably, extrapolating from data obtained by using smaller filter areas or over shorter test intervals tends to give less dependable results. The smaller the filter area and liquid volume used, the less accurate the extrapolation. Consequently, the use of 47-mm flat disc filters in flow decay studies sometimes produces results that are so inexact that they must be used with an extremely high safety margin. Such  allowances usually are set at 15–20%, but can be as high as 150%. Such over-sizing will result in high value losses because of unnecessary large hold-up volumes and unspecific adsorption. In such circumstances, the savings achieved from the flow decay studies are lost because of batch-to-batch running costs. For example, a 10-inch filter element could adsorb an average of 200 mg of drug (Figure 1), which commonly can be doubled to 400 mg for a 20-inch filter, which could be required if the filter were improperly scaled. The per-batch value of such waste can be calculated from the dosage and value of the drug product being filtered.

 The importance of proper filter sizing also has received attention in the revised PDA Technical Report #26, which describes filter choice and trials in detail.

When dealing with a "clean" liquid (i.e., a liquid devoid of suspended particles), the flow rate and the resulting throughput, per unit of time, at the applied differential pressure is straightforward to measure, based on the selected EFA. (Temperature must be kept constant, particularly because the liquid's viscosity is its reciprocal, and any increase in either factor will lead directly to a proportional rise in throughput.)

If the liquid contains suspended particulates, however, the filter's porosity will be decreased by particle retention, so additional EFA may be required to compensate for the filter area that is blocked or clogged by the particulates.

  In such cases, it will be necessary to conduct experiments to assess the EFA needed for batch processing.  

Alternatively, it may be possible to restore lost throughput by using higher inlet pressures rather than increasing EFA. If filtration is already in progress, that would provide a more manageable alternative. The effectiveness of such a measure would depend, however, on the total suspended solids (TSS) in the solution. Higher differential pressures could increase the pressure drop by compacting any filter cake that may have formed. In contrast, a larger EFA would be less likely to produce such compaction under the same conditions, as there would be little or no filter cake build-up. High differential pressure also commonly leads to high flow rates, which can cause elevated fouling or adsorptive effects. Those effects, in turn, can lead to losses in product yield or the need for larger EFA. 

Figure 2.  Coomassie-blue stained 47-mm filter discs 

If the preparation presented for filtration is relatively free from suspended particles, flat discs of even smaller diameters may suffice

because the EFA available for liquid permeation would not be diminished by particle deposits. An example of such an application would be filter sizing for deionized waters. These liquids contain so little suspended matter that the flow decline data can be secured fairly quickly from tests using a small filter. The fewer the solids in the suspension, the less demanding the mathematical extrapolation.  

Nevertheless, a sizeable inaccuracy inheres to the extrapolation of results from a 47-mm disc's EFA of 1.49 in

 cartridge. At best, the results indicate only hypothetical, non-committal values; hence the large margins allowed for error. An assessment method that requires an EFA overdesign of as much as 1 or 1.5 times the extrapolated value does not merit endorsement.  

It may well be that the extreme safety margins reported are exaggerations of the arithmetical uncertainties, reflecting the experimenter's strong fear of having to interrupt the filtration mid-stream to install new filters to allow processing of the batch to be completed. Indeed, the aseptic replacement of a filter is a risk-prone operation, so its avoidance is strongly recommended. 

The flow decline, or flow decay method, is used to determine the effective filtration area required to process an entire batch of any volume in an acceptable time frame under a given delta (Δ) pressure. It is managed by a simple arithmetical proportioning. A small filter with a known effective area is used to measure the volume of a drug preparation that can be processed at a selected ?P (differential pressure), given the rate of flow and the throughput it produces. The ratio of the filtration area to the sample volume is extrapolated to the filter area required to process an entire batch. The simplicity of the test ensures its easy mastery; its performance demands only a modest technical background.  

In a flow decline experiment, a small aliquot of the suspension is filtered through a small filter. The retained particulate load accumulates on the filter pores, successively diminishing filter porosity and the flow rate. At a certain point, the flow is judged to have ceased for all practical purposes; the effort and time that would be spent obtaining additional filtrate beyond that point is not economically feasible.  

In these experiments, one measures the flow rate (volume over time), the throughput, and the total filtrate volume obtained over the duration of the filtration.

  The ambient temperature is also noted. The results identify the ratio between the EFA and the volume of filtrate that is produced before the filtration is terminated. The numerical value has a significance beyond that inherent in the specific filter. Rather, it is taken to quantify the capacity of filters of that type to retain the maximum amount of particulate matter present in that unique fluid preparation under the given filtration conditions. Its importance is in the nature of a validation exercise. 

An arithmetical extrapolation of the experimentally obtained ratio can then be used to determine how large a filter area will be required to process a batch of product.

is the filter area required to process the batch.  

The required filter area, thus calculated, can then be translated into the number and lengths of cartridges or other filtration devices one wishes to use for the batch operation, by using additional equations. The area of a flat disc filter is:  

 ). A 10-in cartridge composed of the same filter polymer

will be assumed to contain an area of 6 ft

 . (The areas of other filter cartridges may be calculated in the same way.) Therefore, the throughput volume of the cartridge relative to that of a 47-mm filter of the same polymeric composition is: 

Thus, if the 47-mm disc yields a total throughput of 2,500 mL, then 2.5 L x 580 = 1,450 L (~383 gallons) will flow through the cartridge before it shuts down. The rate of flow measured on the 47-mm disc then indicates the number of 10-inch cartridges with an EFA of 6 ft

  that would be required to complete the batch filtration in a timely manner at the constant differential pressure, temperature being kept constant.  

From the foregoing, it should be self-evident that the accuracy of the arithmetical extrapolation is highly dependent on the accuracy of the EFA attributed to the 47-mm disc. Nevertheless, it is in this very matter that many incorrect assumptions are made. 

ACCURATELY MEASURING THE EFFECTIVE FILTER AREA OF A 47 MM FLAT DISC FILTER 

In the classical application of the flow decline method, a 47-mm membrane is removed from its package and inserted into a stainless steel holder, where it is held in position by the compressive action of an O-ring. The area of a 47-mm filter disc is 17.4 cm

 . In the holder, however, the O-ring pre-empts a certain quantity of the disc's peripheral space by clamping down on the filter disc to prevent edge-leakage. This reduces the filtration area available for filtration. Thus, the EFA of the inserted 47-mm disc is reduced to less than 17.4 cm

The actual EFA of such an assembly was measured by filtering a staining solution of acridine yellow or Coomassie blue (Figure 2). The stained area identified the sealed area in the confines of the O-ring. As can be seen, the staining solution did not extend beyond or under the O-ring. The stained area measured 41 mm in diameter, which has a total area of  13.2 cm

  were rendered unavailable for filtration by the O-ring's preemptive sealing action. 

USING PREASSEMBLED 47-MM FLAT DISC FILTERS

A pre-assembled, disposable unit containing a "47-mm" diameter flat disc filter is available. The pre-assembly offers the attractive advantage of disposability. It is likely that users assume that its EFA is that of a 47-mm disc. However, the membrane used for these devices is actually 50 mm in diameter, a common filter size. The staining technique using acridine yellow reveals its effective diameter to be 48 mm (Figure 3), which means it has an EFA of 18.4 cm

 . Its throughput, however, is being ascribed to that of a smaller EFA—the assumed EFA of the 47-mm disc ordinarily used, namely, 17.4 cm

 . The discrepancy in EFA is even larger if the stained 13.2 cm

  area of an actual 47-mm diameter filter disc is used in the comparison (Figure 4). 

Figure 3.  Disposable test filter with 48-mm diameter or 18.4 cm2 filtration area 

  EFA of the 48-mm diameter disc filter may mistakenly be ascribed to the actual 13.2 cm

  EFA of an O-ring sealed 47-mm diameter filter. If so, the extrapolation exercise will lead to a numerical multiplier that will indicate that a smaller EFA is needed for processing than would result if the multiplier were based on the smaller (stained) value of 13.2 cm

 . The result will be a low EFA that is insufficient for batch processing needs. This may cause a mid-process filter change-out, which the use of the flat disc filter sizing aims to avoid. 

Figure 4.  Typical 47-mm test filter used in a stainless steel holder  

Even though the use of 47-mm flat disc filters is widespread in flow decline work, the operational details of these studies may differ among users, because the method is not standardized. It is not known what EFA values the many users of 47-mm disc filters actually use in their sizing protocols, because these numbers are seldom reported. Measuring filter EFAs by acridine yellow or other staining does not seem to be a widely discussed or published procedure. Clearly, its application to the use of 47-mm discs in filter sizing studies is not universal. 

It is recommended that the staining technique be used to determine the exact diameter of the disc being used, so that its actual EFA can be calculated from that diameter. This practice would minimize the risk of underestimating the EFA needed for processing a batch operation, and reduce the possibility of needing mid-process filter replacements. 

Even if staining techniques are used to improve the accuracy of filter sizing studies, the extrapolations from 47-mm flat disc filters to 10-inch cartridges are beset by the uncertainties derived from suspended matter. At best, flow decline assessments based on 47-mm flat disc filters constitute indicator trials.  

Therefore, it is best to follow up these indicator trials with verification trials using larger-area pleated filter devices, (commonly 1.5 ft

  Indeed, when costly drug preparations are involved and properly defined filtration area scaling is needed, the use of full-scale filters in assurance trials is recommended. Full-scale trials ensure that the filtration system will be large enough to filter the required batch volume without being oversized, thus minimizing product yield losses. 

The function of the 47-mm flat disc filter is to serve as the model for the filter to be used in batch processing. Unfortunately, filterability trials that use 47-mm flat discs can only roughly indicate which filter combination might be optimal. To perform reliably in its pilot role, the model filter should be as similar as possible in all its structural details to the production filter. Moreover, it should be tested under the conditions and in a manner as similar as possible to those of the production operation.  

Batch filtration will most likely involve pleated filter cartridges. The details of pleated cartridge construction, however, are substantially different from those of flat filters. The measurement of the effective filtration area of flat disc filters is straightforward, whereas that of pleated filters is appreciably more complex. As a result, flow and retention data will not extrapolate well from flat stock to pleated filter cartridges, and EFA forecasts based on flat stock may overestimate the EFA available from pleated filter constructions. This would lead to a need for mid-process change-outs.  

The differences in the EFAs of flat disc and pleated filters can result from the pleat-pack constructions and density of pleated filters. These discrepancies may derive from the flow-attenuating influences of the cartridge's support and drainage layers. In pleated filters, the entire filter area may not be available for the filtration function; a portion may serve to satisfy structural or mechanical requirements. How much of the remainder is available to the EFA function depends on how the pleating operation is conducted.  

The details of the pleating operation are beyond the scope of this article.

  We can note, however, that the very nature of the pleat numbers, heights, and degrees of tightness affect the flow and retention properties of the cartridges composed of them.  

The flow-attenuating influences of the cartridge's support and drainage layers, and the retention-modifying effect of pleat construction features cannot dependably be assessed from pilot studies using flat stock. Therefore, if pleated filter cartridges will be used in batch processing, indicator trials using 47-mm flat disc filters must be followed up by verification trials using miniaturized pleated filter devices.

  Using pleated filters in these verification trials will mean that the test results will be much more relevant. Pleated filter cartridges are available on the market in sizes as small as two inches in length.  

When even more reliable EFA forecasts for batch processing are needed, assurance trials should be conducted using larger-scale cartridges. Conducting tests using filters with a larger EFA, even to the extent of using full 10-inch pleated cartridges, would proportionately increase the reliability of the extrapolated value. 

Cover

Why staining is crucial in flow decay studies.

BioPharm International-09-01-2009

BioPharm International, September 2009 Issue (PDF)

A True Assessment of Risk

Comparison of Camelid Antibody Ligand to Protein A for Monoclonal Antibody Purification

Breaking the Rapid Microbiological Method Financial Barrier: A Case Study in Return on Investment and Economic Justification

Patenting Biotechnology Inventions Will Become More Challenging

The Proper Use of 47-mm Flat Disc Filters in Filter Sizing Studies

Biopharmaceutical Manufacturers Seize Opportunities in the Global Health Arena

Quality Risk Management Demystified at CMC Strategy Forum

Strategic Outsourcing of Media Design and Cell Culture Media Manufacturing