Articles

Issue PDF

BioPharm International, August 2009 (PDF)

Change-point analysis is an effective and powerful statistical tool for determining if and when a change in a data set has occurred. The tool provides a confidence level that indicates the likelihood of the change. Change-point analysis can be used in three distinct applications: 1) determining if improvements or process changes may have led to a shift in an output, 2) problem solving, and 3) trend analysis. This paper describes how the tool can be used in the pharmaceutical industry for the three applications. Two case studies are presented to show how change-point analysis was used to verify the potential effect of process changes. 

One question that is commonly asked during the analysis of time-ordered data is "Has there been a shift in the mean of the data?" One technique for assessing if and when a shift has occurred is a cumulative sum chart (CUSUM chart). CUSUM charts rely on a visual assessment of whether there is a change in the slope of the CUSUM plot. This technique works well with large changes because they produce an obvious change in slope. Subtle changes in slope are more difficult to detect and may be missed. Also, it is difficult to determine if a less pronounced change in slope represents a significant change. Potential change points identified with CUSUM charts can be confirmed by running a 

-test of the data before and after the suspected change point. However, a 

-test should only be used if the data is normally distributed and the change point is known. Bandurek provides a review of the use of CUSUM charts.

Another statistical tool that uses CUSUM charts to look for shifts in data is change-point analysis. This approach goes one step further by assigning a confidence level for each change detected. The confidence level is determined using a technique known as bootstrapping, which takes the subjectivity out of a CUSUM analysis.  

This paper describes how to conduct a change-point analysis and discusses three applications for the tool in pharmaceutical process monitoring and control: demonstrating improvements, problem solving, and trend analysis. The use of change-point analysis to demonstrate an improvement is highlighted in two examples. 

Change-point analysis uses cumulative sum and bootstrapping techniques to identify changes and assign a confidence level to the change.

  First, a CUSUM chart is generated, which displays the cumulative sum of the differences between individual data values and the mean. If there is no shift in the mean of the data, the chart will be relatively flat with no pronounced changes in slope. Also, the range (the difference between the highest and lowest data points) will be small. A data set with a shift in the mean will have a slope change at the data point where the change occurred, and the range will be relatively large. 

Figure 1. Example of a CUSUM chart. Each point plotted represents the difference between an individual data point and the mean, which is added to or subtracted from the previous point on the graph (depending on whether the difference between the individual data point is positive or negative). The data for this CUSUM chart is shown in Figure 2. 

Figure 1 shows an example of a CUSUM chart. Data for this chart is listed in the Excel table in Figure 2. Column B contains the raw data and column C contains the cumulative sum. The cumulative sum at each data point is calculated by adding the difference between the current value and mean to the previous sum [i.e., 

 is the mean]. A CUSUM chart starting at zero will always end with 

 = 0. If a CUSUM chart slopes down, it doesn't necessarily mean that the data are trending down. Rather, it indicates a period in time when most of the data are below the mean. A sudden change in direction of a CUSUM indicates a shift in the average. From the CUSUM chart in Figure 1, it appears that a change may have occurred at data point 20. At this point, the slope changes direction and increases, indicating that most of the data points are now greater than the average. The point at which the CUSUM chart is furthest from the baseline of zero is the estimated point of change. Interpreting the CUSUM chart would lead one to the conclusion that a shift in the mean occurred at data point 20. This interpretation relies on a subjective assessment as to whether there is a change in the slope.  

Figure 2. Excel table with data used to generate the CUSUM chart in Figure 1 

Change-point analysis builds on a CUSUM chart by determining a confidence level for each change. Confidence levels are calculated using a technique known as bootstrapping, whereby many random iterations of the data set are generated. For each randomized data set, the corresponding cumulative sums are determined, along with the ranges for the sums. The percent of times that the cumulative sum range for the original data exceeds the cumulative sum range for the randomized bootstrap data is the confidence level. The idea behind this algorithm is that if a pronounced change has occurred, the range on the CUSUM chart for the original data will be large, and randomizing the data will not lead to data sets with larger ranges, or will do so only infrequently. 

Figure 3. Bootstraps of original data from Figure 2. The average of each bootstrap data set is shown at the bottom. Bootstraps are data sets in which the original data has been randomly reordered. For this example, 100 bootstraps were conducted, the first 15 of which are shown here. 

In the sample data set from Figure 2, 100 bootstraps were produced using the Excel formula =INDEX($B$3:$B$32,1+30*RAND(),1). The first 15 bootstraps are shown in Figure 3. For each bootstrap data set, the corresponding CUSUM data are generated along with the range (difference between the highest and lowest CUSUM values). The bootstrap CUSUM values are shown in Figure 4. All the formulas used for the analysis are in Table 1. 

Table 1.  Microsoft Excel formulas for conducting a change-point analysis 

The final step in determining the confidence level is to calculate the percent of times that the range for the original CUSUM data exceeds the range for the bootstrap CUSUM data. The CUSUM data range for the bootstraps is in row 115 of Figure 4, whereas the CUSUM range for the original data is in cell F3 of Figure 2. In this example, the confidence level was 99%. It is appropriate to set a predetermined threshold confidence level beyond which a change is considered significant. Typically, 90% or 95% is selected. 

Figure 4. CUSUM values for the bootstrap data. The range for each data set is shown at the bottom in row 115. 

The change at data point 20 that was indicated on the original CUSUM chart has been shown to have a 99% confidence level based on 100 bootstraps. A histogram of the CUSUM ranges for the 100 bootstraps is shown in Figure 5. As shown in the histogram, only one of the bootstrap ranges was greater than 9.8, which was the CUSUM range of the original data. Thus, the confidence level was 99%.  

Figure 5. Histogram of bootstrap CUSUM ranges. The CUSUM range for the original data is indicated by the red line at 9.8. 

The original CUSUM chart hints at other, more subtle changes. These potential changes can now be assigned a confidence level by dividing the data into two subsets: data points 1 to 19 and data points 20 to 30. Each data subset can then subjected to its own change-point analysis to see if the threshold confidence level is exceeded. 

Potential changes in variation also can be assessed using the change-point analysis technique. Because biologics manufacturers produce a relatively small number of batches each week, it is not always practical to perform a change-point analysis on standard deviation. As an alternative, a change-point analysis is conducted on the difference between consecutive data points. For the sample data from Figure 2, let X

 represent the 30 data points. From this, 15 consecutive differences, D

 = 1 to 15. The change-point analysis is then performed on D

Microsoft Excel was used to perform the analysis described above. A commercially available software package known as Change-Point Analyzer (Taylor Enterprises, Inc.) greatly simplifies the analysis. The remaining examples in this paper used Change-Point Analyzer, version 2.2. 

APPLYING CHANGE-POINT ANALYSIS TO PROCESS MONITORING AND CONTROL

In the two case studies shown here, change-point analysis using Change-Point Analyzer was used to show that a process change correlated with an improvement in a measured output.  

Manufacturers of active pharmaceutical ingredients typically monitor process recovery to ensure that the process is performing as expected and that yield targets are being met. The historical record of recovery also provides a baseline against which potential improvements can be measured. Change-point analysis is an effective tool to verify whether a process change has led to measurable improvements, such as an increase in recovery.  

In the first example, a process change was initiated in an attempt to improve recovery of a target protein, and a change-point analysis was conducted on lots bracketing the time of the change (Figure 6). This analysis revealed that an increase in recovery occurred at lot 28 (indicated by a shift upward in the green zone and from the summary table below the plot). The actual process change was initiated at lot 24. An increase in the percent recovery was detected only a few lots after the actual process change occurred and within the 95% confidence interval indicated in the summary table for Figure 6. The change was from 48.6% to 53.7% and had a confidence level of 96%.  

Figure 6.  Change-point analysis of percent recovery for a plasma protein. The analysis was done using Change-Point Analyzer. The data show an upward shift in the percent recovery. The shift is centered at lot 28 and represents a shift from 48.6% to 53.7%. This change has a 96% confidence level. An additional output from the software is a confidence interval, which in this example is from lots 10 to 39. The red lines indicate control limits. The level is an indication of the importance of the change. A level 1 change is the first change detected and the one most visually apparent in the plot. A level 2 change is detected on a second pass through the data after the data are subdivided into two subsets. 

In the second case study, the parameter that was being monitored was potency targeting. Different potencies of active ingredient are produced depending on the needs of the customer. To meet these needs, it is important that the actual potency be as close as possible to the potency level required by the customer (referred to as 

). The effectiveness of potency targeting is measured by how far the actual potency is from the target potency (percent from target). A recent improvement project sought to improve potency targeting. Process- and assay-related factors are both responsible for the variation in targeting accuracy. In this project, assay improvements were implemented and a change-point analysis was conducted on lots that bracketed the time of implementation of the improvements (Figure 7). Actual improvements were implemented at lot 17, and the results of the change-point analysis indicate that a shift in the mean from 9.9% to 5.8% occurred at lot 31. The confidence interval for this change is from lots 15 to 45, which encompasses when the actual change occurred. 

Figure 7.  Change-point analysis of potency targeting. Data shows a decrease in the percent from target at lot 31. The potency targeting improved from 9.9% to 5.8%. This change has a confidence level of 97% and a confidence interval of 15â45. 

The statistical significance of changes identified by change-point analysis also can be verified by conducting a two-sample 

-test. This was done for the two examples described above, and the results confirmed the statistical significance of the change (Tables 2 and 3). For each example, two 

-tests were conducted. One compared data before and after the change point identified by change-point analysis, and the other compared data before and after the actual lot when the process change occurred. The 

-tests was <0.05, indicating that with at least 95% confidence, the means are not equal. The 

-tests confirmed the change-point analysis results that a shift in the mean occurred. 

Table 2.  T-test results from the percent recovery data 

Because a normal distribution is an assumption for a 

-test, it should only be used to verify the statistical significance of changes when data are normally distributed. Also, the 

-test by itself does not identify when changes occur. It can only be used to confirm a hypothesized point of change, and as such is not a substitute for change-point analysis when used for problem solving and trending. 

Table 3.  T-test results from potency targeting data 

If the timeframe of a process change corresponds to the timeframe of a shift in the output of a measured parameter, one can conclude that the process change may have caused the shift. Correspondence doesn't necessarily prove a cause-and-effect relationship between the process change and the shift in output, however. The analysis should be supplemented with process knowledge, other statistical analysis, or scaled-down experimentation to more definitively demonstrate cause and effect. 

In the problem-solving mode, a retrospective change-point analysis is conducted on selected attributes to see if the timeframe of a change in one of the attributes corresponds to the occurrence of a manufacturing problem. As a hypothetical example, let's assume that there was an increase in out-of-range results for sodium concentration in a processing buffer. A change-point analysis should be run on the sodium concentration of buffer lots over time. This allows us to determine if the out-of-range batches are isolated instances or if there has been a shift in the sodium concentration of the buffer that is causing more batches to be out-of-range. The analysis should encompass enough lots before the increase in out-of-range results to provide a good representation of the variation in the data. If the change-point analysis showed that a change in the sodium concentration occurred in the same timeframe as the onset of the out-of-range results, the investigation will focus on the cause of the shift.  

As part of the investigation, many potential causes may be identified, including changes in:  

Change-point analysis can be conducted on any of the time-ordered data such as mixing speed for each lot, mixing time for each lot, or assay control values. If any of the results from a change-point analysis indicate a shift in the data that corresponds to the sodium concentration shift, one can narrow the scope of the investigation. Alternatively, root causes can be eliminated if a measured parameter did not change over the timeframe examined. 

The hypothetical example described above shows how change-point analysis can be used to provide a retrospective analysis of a data set to see if shifts in the data correspond to the onset of a manufacturing issue. In this context, change-point analysis serves as a powerful problem-solving tool. 

Control charts are commonly used to evaluate data on a lot-by-lot basis or sample-by-sample basis, and can indicate when a particular lot or sample is not part of the same population of data as the data that was used to generate the control limits (special cause variation). Furthermore, trend analysis can be done with control charts if the data is normally distributed. Likewise, change-point analysis also can be used on a real-time basis to look for trends or shifts in a process. Change-point analysis provides some advantages over control charts when looking for shifts in data.

Features

A better method for trend analysis than CUSUM and control charts.

Use of Change-Point Analysis for Process Monitoring and Control

The Quality by Design paradigm demands an enhanced understanding of processes and products. Biotech therapeutic products, being complex molecules, require a robust analytical platform to serve as the foundation for commercialization activities such as the definition of critical quality attributes. This platform is generally a combination of orthogonal, high-resolution techniques that together provide definition to the product and the process. This article presents recent developments in four analytical applications, namely disulfide linkage analysis, glycan analysis, analytical ultracentrifugation, and flow injection protein analysis.

Prroduct and process characterization are critical components of the overall commercialization activities for protein therapeutics. Product characterization reveals the biochemical and biophysical nature of the product as well as the nature of product-related substances and impurities. Thorough product characterization is a necessary precursor to determine critical quality attributes (CQAs) and the associated analytical methods that in turn can be used as in-process controls and specifications, and for stability testing. Process characterization focuses on understanding and defining the operating and design spaces for the process to achieve a product with consistent CQAs.

 series and presents recent developments in four analytical applications, namely disulfide linkage analysis, glycan analysis, analytical ultracentrifugation, and flow injection protein analysis. 

Determining CQAs and developing methods suitable for their measurement is an intricate and evolving science. This stems from the inherent complexity of biological macromolecules. A perfectly pure biological product is in itself challenging to characterize, typically comprising a chain of hundreds of amino acids and associated glycan subunits. This chain is sometimes linked within itself or to other chains through disulfide bonds, and then folded into a discrete secondary and tertiary structure.

 Quaternery structures also exist, from dimerization up to potentially large numbers of noncovalently or covalently linked subunits. Furthermore, manufacturing processes never achieve 100% purity, resulting in the need to measure and control product-related variants and product- and process-related impurities.  

Product-related variants closely resemble the desired product and have a potency and safety profile equivalent to those of the product itself.

 Typical examples include minor post-translational modifications such as C-terminal processing, N-terminal variants, and deamidation. Product-related impurities, however, do not resemble the desired product with respect to safety and efficacy. Examples include aggregates and highly truncated forms. Process-related impurities include host cell DNA, host cell proteins, and raw materials from the process. A complex array of analytical methods is necessary to adequately characterize both the product and process.

 These methods must be developed and then qualified and validated for their intended use.

A NEW METHODOLOGY FOR DETERMINING DISULFIDE BONDING PATTERNS

Regulatory agencies expect disulfide connectivity to be determined as part of product characterization.

 Contemporary methods for elucidating bonding patterns between closely-spaced cysteine residues include cyanylation-induced cleavage after limited reduction and limited reduction and alkylation followed by mass spectrometric analysis.

 The latter approach was successfully applied to a recombinant IgG2, revealing subspecies of disulfide variants.

 The main limitation of the reduction step inherent in both these approaches is the difficulty in finding conditions that allow for controlled reduction of specific disulfide bonds.  

Recently, the analysis of mass spectral fragmentation patterns of disulfide-linked peptides has been used to elucidate disulfide linkages.

 However, the applicability of this approach has been limited to fairly simple systems, and not to the highly complex linkage patterns of the recently reported IgG2 variants.

To overcome the above limitations, we have recently developed an innovative methodology that permits rapid determination of the connectivity between closely-spaced or adjacent cysteine residues in disulfide-linked molecules, and can be readily applied to complex monoclonal antibodies (MAbs) such as those in the IgG2 class. This methodology uses a unique combination of traditional Edman chemistry with mass spectrometric detection. Optimization of the sample processing for Edman chemistry played a crucial role and provided predictable structures for facile interpretation of mass data.  

With many disulfide-linked proteins, even after proteolytic digest, there are often multiple potential disulfide connectivities that cannot be distinguished by mass alone because they merely represent different assemblies of the same peptides. However, the differences in the linkage geometries will result in characteristic "residual" and "leaving" groups after repetitive Edman cleavage of the cysteine residues from the N-terminus of each chain. An example of this scheme is presented in Figure 1, using insulin as a model substrate. The peptide presented in Figure 1 is derived from the Glu-C digestion of human insulin. 

Figure 1.  Differentiation of disulfide variants through sequential manual Edman cycles with  characteristic residual and leaving groups 

The characteristic residual and leaving species may be readily differentiated by mass spectrometry, and their identities may be confirmed by standard fragmentation analysis. If a mixture of disulfide structures for a given variant is present, it may not be possible to differentiate them from the observed leaving groups. However, the residual groups are diagnostic, and may even provide a means for a semiquantitative measure of their relative abundance, based on their relative signal in either the mass spectrometry (MS) or ultraviolet (UV) traces. Additional rounds of coupling, cleavage, and liquid chromatography–MS (LC–MS) analysis provide a further means for confirming structures to a greater level of detail, and also demonstrate the absence of scrambling during the chemical processing steps. 

We have successfully applied this approach to the well-characterized insulin molecule and to a MAb where we identified new substructures of IgG2 variants, including the presence of intra-chain disulfide linkages in the hinge region.

 Details of this approach will be the focus of upcoming publications. 

N-linked oligosaccharide mapping is routinely performed as an in-process test for recombinant glycoproteins derived from mammalian cell lines. Reasons include the complexities associated with processing Asparagine-linked (N-linked) oligosaccharides and the sensitivity of the enzymes involved, to even subtle changes in cell culture conditions during manufacturing. In addition, oligosaccharide mapping as a potential component of product lot release is increasingly requested from regulatory agencies for consideration.

Standard methodologies contain several bottlenecks with respect to throughput. First, the sample preparation required for analysis is both laborious and time-consuming, requiring multiple steps of hands-on manipulation by the analyst. Second, the typical separation methods used have cycle times ranging from 90 to 180 minutes.

 In instances where multiple cell culture conditions are being screened or large numbers of in-process samples are being tested, the resulting analysis sequence can take days. In addition, the mobile phases typically used in the separation of oligosaccharides are not compatible with online MS analysis, requiring the collection of multiple fractions and additional sample manipulation before MS analysis, providing another bottleneck in the characterization stage. 

We have implemented methodologies to address each of these throughput constraints. Through the use of robotic liquid handling, automated sample preparation, and rapid resolution reverse phase chromatography (RRRP–HPLC), we are able to completely process 30 samples per 24 hour period for oligosaccharide analysis, from the point of initial enzymatic digestion through full MS characterization of species accounting for as little as 0.1% of the oligosaccharide moiety.

 A traditional sample preparation scheme involves removal of deglycosylated protein by porous graphitized carbon (PGC) following PNGase F digestion, vacuum centrifugation before labeling with the fluorophore, and finally, removal of excess fluorophore and labeling reaction components using a cellulose phase matrix (S-cartridge) with subsequent vacuum centrifugation. We have implemented methodologies that replace these manually operated PGC and S-cartridges with PhyTip columns (Phynexus, Inc.) packed with Carbopack B and DPA-6S resins that are compatible with robotic platforms.  

A total of six identical PNGase F digestions were performed on 500 ug aliquots of a recombinant IgG. Three of the digests were processed following traditional protocols and the remaining samples were processed following the automated method. Samples analysis was performed by traditional high pH anion exchange chromatography (HPAEC). An overlay of the resulting chromatograms is shown in Figure 2. No significant differences were observed in the chromatograms obtained by the two preparation methods. In addition to requiring limited analyst manipulation during preparation, the samples prepared following the automated protocol were ready for analysis at the end of Day 1, whereas samples prepared following the standard protocol were not ready for analysis until the start of Day 3. 

Figure 2.  Comparison of samples prepared by automated and standard methods and separated by high pH anion exchange chromatography 

The most commonly used separation methods for oligosaccharide analysis are HPAEC and normal phase chromatography (NP–HPLC). The mobile phases used for these separations are not compatible with online MS characterization. Reverse phase (RP–HPLC) separations have been described previously, and provide superior resolution of species compared to HPAEC and NP–HPLC methods.

 However, with cycle times of three hours, they are not suitable for routine analysis. As seen in Figure 4, we have developed a method that takes advantage of new small particle-size resins available from column manufacturers. A batch of 30 samples from cell culture screening conditions can be prepared, separated, and characterized in 24 h using this approach with the chromatographic cycle time dramatically reduced relative to traditional RP–HPLC. This RRRP–HPLC method has a cycle time of 35 minutes, provides comparable resolution to standard RP–HPLC, and is compatible with online ESI–MS/MS detection with a limit of detection (LOD) of <20 fmol by fluorescence and ~1 pmol by MS. Using this approach, full characterization of the oligosaccharide map, including species accounting for as little as 0.1% of the total moiety, can be achieved in a timeframe that is competitive with basic profile fingerprinting achieved by standard capillary zone electrophoresis (CZE) methods. An example separation for an rIgG is shown in Figure 4. For this particular rIgG, a total of 36 unique species were identified from a single sample injection. 

Figure 3.  Comparison of flow schemes for standard and automated methodologies for oligosaccharide analysis 

Through the combination of improvements in sample preparation and chromatographic cycle time, we are now able to perform complete characterization of an oligosaccharide map in a single day, providing a minimum of a five-fold reduction in process time. In addition, because of the short cycle time required by the RRRP–HPLC and compatibility with online MS detection, thorough MS analysis can be incorporated as part of routine sample analysis. 

Figure 4.  Typical separation by reverse-phase of an N-linked oligosaccharide pool from an rIgG 

A QUANTITATIVE TOOL CHARACTERIZATION AGGREGATES

Size exclusion chromatography (SEC) is the standard method for quantifying levels of aggregation in protein products.

 However, aggregate characterization by SEC can perturb the delicate equilibrium of protein self-association because of resin–protein interactions and undesirable mobile-phase effects inherent in chromatographic-based separations.

 Sedimentation velocity–analytical ultracentrifugation (SV–AUC) can be applied in biopharmaceutical development as an orthogonal technique to SEC. SV–AUC is free from many limitations intrinsic to SEC but exhibits much lower precision, compromising its usefulness as a quantitative technique.

The level of aggregation in a protein product is a CQA of the product, which must be monitored during development and controlled during manufacture. Accurate and precise quantitation of protein aggregation in biopharmaceutical products therefore meets a crucial business need. We have developed an improved SV–AUC methodology that enables enhanced precision and accuracy of protein aggregate measurements directly in product formulation buffer. The improved methodology allows successful implementation of quantitative SV–AUC in biopharmaceutical applications requiring high-precision aggregate measurements. 

SV–AUC can be applied in protein biopharmaceutical development for many diverse characterization purposes.

 For example, SV–AUC is useful for process characterization, product structural characterization, product impurity profile determination, protein-receptor binding studies, comparability assessment, and evaluation of SEC method accuracy. Some of these applications permit qualitative assessment whereas others require quantitative results with a high degree of precision and accuracy. Unfortunately, the current precision of SV–AUC may preclude or limit its use in characterization studies that require quantitative aggregate measurements. The technique's use is particularly limited in cases of low aggregate levels (<1%), or when attempting to quantify subtle differences in aggregate levels.  

The standard deviation of repeated protein aggregate measurements determined by SV–AUC is typically between 0.3 and 0.5% aggregate, but can be as high as 1% aggregate.

 Measurement precision is affected by common laboratory variations, including the instrument used, analyst technique, centerpiece quality, sample characteristics, and the data-fitting approach.

 Although the analyst has limited control over many factors, others can be deliberately controlled to improve the consistency of results. Two specific aspects of SV–AUC experiments are considered here: data analysis approaches for experiments conducted in non-ideal solutions containing excipients, and alignment of centerpieces. Proper control of these factors can dramatically improve both accuracy and precision.  

Therapeutic protein formulations often contain excipients such as sugars or sugar alcohols. Typical angular velocities used during SV–AUC experiments (i.e., 40,000 to 60,000 rpm) cause excipient concentration gradients to form over experimentally relevant time scales. If excipient concentration gradients are not properly modeled during data analysis, the capability of SV–AUC to measure protein aggregation is dramatically impaired.

 For example, measured aggregate levels for a MAb declined from 3.0 to 0.7% after 10% sorbitol was added to the solution.

 Sophisticated data analysis approaches, including modeling excipient gradients, can mitigate the quantitation problems caused by the sedimentation of excipients.

 In this case, modeling the 10% sorbitol gradient substantially improved accuracy (2.3% aggregate measured).

 SV–AUC measurements of protein aggregate in product formulation buffer exhibit poor accuracy unless the effects of excipients are adequately taken into account during data analysis. 

Poor analyst technique and inadequate experimental control can significantly increase variability in SV–AUC results.

 SV–AUC experimentation is complex and requires extensive analyst training compared to SEC experiments. One experimental aspect that depends on analyst technique is alignment of the sample-containing AUC cells to the center of rotation. Misalignment of the centerpieces can cause convective disturbances resulting in sample mixing and inaccurate results.

 To better control cell alignment to the center of rotation, we designed a custom alignment tool capable of controlling cell alignment within 0.25°. The standard cell alignment approach relies on visual alignment of score marks on the rotor and cell assembly. This visual alignment technique can result in up to 0.5° misalignment from the center of rotation. When AUC cells were deliberately misaligned by 0.5, 1.0, 2.0, and 4.0°, measured levels of aggregate increased linearly with increasing angle of misalignment. Slight cell misalignment (≤0.5°) outside analyst control can yield variability as high as 1% aggregate. This variability is reduced to less than 0.5% aggregate by using an alignment tool.

The best way to improve the quantitative use of SV–AUC in biopharmaceutical development is to reduce the intrinsic variability of the technique. Many factors affect SV–AUC measurement variability; some are more difficult to control than others. For instance, inherent instrument differences (e.g., lamp intensity) and problematic sample characteristics (e.g., reversible self-association) may be unavoidable. But improved analyst technique (e.g., careful alignment of centerpieces) and advanced data analysis can dramatically increase the overall precision of aggregate measurements. Two methods to improve precision and accuracy were discussed: accounting for excipients in product formulation and alignment of cells to the center of rotation. Table 1 demonstrates how precision and accuracy can be improved by properly accounting for these effects. These methodology improvements can yield a 50% increase in precision and allow accurate measurements in product formulation. 

Table 1.  Accuracy and precision of aggregate measurements by sedimentation velocityâanalytical ultracentrifugation for a model monoclonal antibody under different solution and experimental conditions 

SV–AUC is routinely applied in many aspects of biopharmaceutical development, including process and product characterization. It is used as an orthogonal complement to SEC, with important advantages but also some limitations. SV–AUC allows analysis in product formulation buffer and separates without a stationary phase. However, SV–AUC has limited throughput and exhibits poor method performance characteristics compared to SEC. The specific examples of SV–AUC method improvements highlighted here demonstrate that measurement accuracy and precision can be substantially improved, thus enabling SV–AUC to be a quantitatively meaningful characterization technique. Further, methodology improvements are critical to ensure the continued application of SV–AUC in the biopharmaceutical industry. 

DEVELOPMENT OF A RAPID ANALYTICAL METHOD TO EVALUATE COLUMN AND MEMBRANE CLEANING

Chromatography columns and ultrafiltration membranes are commonly used during the downstream purification of therapeutic proteins or MAbs. Column or membrane cleaning is typically performed to remove any protein or impurity carryover from previous runs and to restore performance between campaigns.

 The effectiveness of cleaning is evaluated by measuring the low protein concentration (in μg/mL range) in a mock run following cleaning using the same elution or ultrafiltration buffers and conditions.   

The most commonly used methods are classic colorimetry-based assays such as Bradford, BCA (bicinchoninic acid), and Lowry methods for total protein assay. However, a variety of buffer systems with different pH levels, components, and salt strengths are used in purification processes, and many of them are not compatible with these colorimetric assays. For example, BCA assays are not compatible with reducing sugars or chelating agents, and Bradford is not compatible with detergents above certain levels. Furthermore, the high-salt strengths and buffer capacity of certain in-process pools (e.g., ion exchange elution buffer) may precipitate assay reagents or dominate the pH of an assay mixture, making these pools incompatible with the assays. Interfering substances can be removed by sample preparation steps such as buffer exchange, blocking reagents for reducing agents, or protein precipitation with organic solvents. These sample preparation steps, however, can potentially introduce assay artifacts, especially at a very low level of protein concentrations. In addition, the response factor for the same protein in different buffer systems can vary significantly because of these interfering factors, making it necessary to make protein standards in different buffers for different column pools. This in turn makes the assay more time-consuming and labor-intensive. A simple, sensitive, and robust method that does not need sample preparation steps would be beneficial for the analysis of column or membrane cleaning samples. 

We have developed a flow injection protein assay (FIPA) using a regular HPLC system with a fluorescence detector. The intrinsic fluorescence of tryptophan is used to monitor and quantitate protein concentration against a standard curve obtained from the same protein of interest spiked into Dulbecco's phosphate buffered saline (DPBS) containing 0.005% polysorbate solution. Samples are directly transferred into glass HPLC vials pre-coated with Sigmacote to prevent protein loss to the glass surface. Samples are then injected onto an HPLC system with no column connected, using DPBS as the isocratic mobile phase, which effectively prevents nonspecific interaction of low concentration protein analytes with instrument surfaces and minimizes run-to-run carryover to <0.5%. The flowthrough peak is monitored by a fluorescence detector (exc 280 nm, emi 335 nm) and integrated to obtain the peak area, which is compared against a standard curve to obtain assay results. The method has been shown to be linear in the range of 1 to 16 μg/mL of protein and no interference was observed from chelating reagents, reducing sugar, or detergents because of the selectivity of fluorescence detection. The limit of detection is observed to be about 0.5–1 μg/mL (depending on the sample matrix) and is equivalent to traditional colorimetric methods. Response factors of the same protein in different commonly used inorganic buffers are relatively constant. Intermediate precision of FIPA across laboratories was determined to be around 10% at the 6 μg/mL protein level and is comparable to traditional colorimetric methods.  

FIPA can be used specifically to quantify the protein of interest. It is not a total protein assay and it requires that the protein of interest have tryptophan residues in its sequence. Wavelengths that are specific to tyrosine or phenylalanine also can be used for this purpose with decreased sensitivity because of the lower molar absorptivity and quantum yield of these two amino acids.  

FIPA provides a simple, sensitive, and robust protein quantitation method to demonstrate the effectiveness of column or membrane cleaning. Compared with other methods (Table 2), FIPA is quicker and has the least matrix interference. Another significant operational advantage of FIPA is that a large set of samples can be analyzed quickly by HPLC without as much analyst hands-on time as needed with the other methods. 

Table 2.  Comparison of the flow injection protein assay with colorimetric methods commonly used to evaluate column and membrane cleaning 

 is senior associate scientist of Analytical Sciences, all at Amgen, Inc., Longmont, CO.  

, is biotech CMC consultant and a faculty member at the Indian Institute of Delhi, India, 805.744.8986, 

Select the best approach to determine critical quality attributes.

Analytical Tools for Process and Product Characterization

Whenever a new FDA commissioner is named, the pharmaceutical industry wonders how drug regulation will change. The big question is, Will the pendulum swing toward increased scrutiny of safety concerns, or toward faster approvals? The current commissioner, however, is being asked to address a more prosaic issue first, relating to the everyday management of the agency's work, particularly the people and resources involved. 

In recent years, a few lawmakers have loudly criticized the agency for failing to meet all its responsibilities. Now, in a new report, the General Accountability Office (GAO) has made a calm and comprehensive analysis, rightly pointing out the gaps in the agency's own knowledge of what it has, does, and needs. 

In the report, the GAO said the FDA should develop a comprehensive assessment of the agency's staffing resources, particularly contractors; an assessment of the extent to which the agency is meeting its responsibilities; and an evidence-based estimate of the resources needed to fulfill its mission. Currently, the agency lacks data in many of these areas. 

Given the scope of the agency's work, it is not surprising that the FDA cannot easily trace every dollar or every contractor hour worked. As the report succinctly captures, the agency's workload, and even its mission, keep getting larger. To wit:  

Since 1999, 11 laws have expressly expanded FDA's medical product oversight responsibilities. Other laws indirectly increase the agency's work.

The number of drug, biologic, and device applications grew 30% from 1999 to 2008, and the number of products approved or cleared for marketing grew 55%.

Adverse event reports increased 228% from 1999 to 2008. 

Companies requested 885 more meetings about drugs and biologics in development (an increase of 56%). 

The number of foreign establishments registered to produce medical products increased by 1,921 (up 23%). 

Meanwhile, rapid advances in science and technology have increased the complexity of the medical products submitted to FDA for approval. 

Yet clearly, it makes sense for any organization to have a good handle on how well it is fulfilling its mission, and how it is getting it done. If the FDA does not know where the gaps lie, it cannot ensure that it is using its resources well, or even make accurate budget requests. 

In its response to the report, the FDA agreed to gather the information requested. 

 also reported recently that FDA Commissioner Margaret Hamburg has proposed an agency reorganization that includes creating a new office of budget. "I intend to make the development and presentation of the agency's funding needs and performance goals among my highest personal priorities," she is quoted as saying. 

People often disagree about whether the best way to regulate drugs is to put greater emphasis on safety, on the one hand, or on speed, on the other (to get new treatments to patients faster). But perhaps the best way to regulate drugs is to regulate them not conservatively or liberally, but 

. A key step to doing that is to accurately assess how, and how well, the work is getting done.  

By carrying out the analysis the GAO has requested, the FDA will set a solid foundation for addressing its gaps, defending its funding needs, and then tackling the issues before it. Because certainly, the FDA has plenty of work on its hands. 

From the Editor

Perhaps the best way to regulate drugs is to regulate them not conservatively or liberally, but effectively.

Laying the Foundation

Current biotech manufacturing plants are capital intensive, with yearly depreciation costs sometimes as high as raw materials costs. In such an environment where indirect costs dominate direct charges, a single metric becomes important: throughput of sale-able product. In this article, we examine a return-on-investment case study for processing a high titer product in a large-scale biopharmaceutical plant. In this case, modeling the altered unit operation in the context of the existing unit operations was essential to establish accurate throughput metrics and overall valuation. We argue that such process-focused economics models are essential in the biopharmaceutical industry.

The key goal of any manufacturing economics study is the accurate evaluation of the processing costs for a particular manufacturing facility against the projected demand stream for the products it will produce. Typical economic analysis in the 1980s and 1990s was based around the value of custom-built plants for a single or a small number of products. Such economic evaluations were relatively simple, because they were based on building a "greenfield" facility that matched the required process specifications and projected demand with the minimum capital outlay. However, with a proliferation of capacity in biopharmaceutical production, manufacturers have shifted focus away from building large-scale single-use facilities and more toward retrofitting existing facilities to produce new or process-improved products. Economic evaluations are more difficult in this context for a number of reasons. 

First, retrofit options typically involve value destruction as well as value creation. Although value destruction is typically thought of as the damage caused by construction, in biotech the most significant modes of value destruction are the downtime associated with plant shutdown and requalification.  

Second, greenfield facilities are inherently right-sized, whereas retrofitted facilities are not. Retrofitting requires the comparison of legacy equipment in the plant, such as, steam and clean in place (SIP/CIP) skids and existing media, and buffer preparation tanks, with new technology. Such decisions inevitably involve design compromises, and these can have unintended effects on throughput and equipment utilization. 

Finally, cost-saving retrofits are not typically justified solely on direct cost reductions, but on metrics like flexibility or capacity increase. As an example, the adoption of disposables in many existing plants is either based around process standardization, decreased contamination risk, or increased flexibility (either in changeover or in utilities). Given existing tanks and CIP or SIP handlers, there is little or no motivation in direct cost-savings to move to new technologies, and thus retrofits typically use one or more indirect cost savings metrics in their justification. Such metrics also may be more qualitative than quantitative, such as the flexibility of a plant to produce multiple products in the future. 

Figure 1 shows the cost allocations in a typical large-scale mammalian cell culture production facility. Raw materials, direct support costs, and direct production labor costs account for around 33–50% of the total yearly operating cost of such a facility. Indirect labor, quality, corporate overheads, and depreciation account for more than 50% of the total costs. Such indirect costs are semi-variable in the sense that they are not directly run-rate–dependent and are difficult to control in the short-term. Indirect labor, for example, includes project work that may be an integral part of the plant even if its run-rate is zero. 

Figure 1.  Yearly operating cost breakdown for a typical mammalian cell-based biotech plant1 

Such a fixed-cost infrastructure is uncommon in most other manufacturing sectors, where raw materials costs and direct-support costs far outweigh indirect cost considerations. One of the few exceptions to this is in semiconductor manufacturing, where significant infrastructure costs and the requirement for a controlled environment create higher indirect costs on a scale similar to biopharmaceuticals.  

In such an environment where indirect costs dominate the direct cost of manufacturing, only relatively small improvements in performance can be achieved through direct cost reduction. For example, a 15% reduction in raw materials cost—roughly equivalent to completely removing the most expensive recovery step in most biopharmaceutical process operations, Protein A—reduces overall operating expenses by less than 2%. Such direct cost reductions are typically outweighed by the need to shut down the plant to perform installation and testing of retrofit options: an expensive proposition because most costs are not run-rate dependent. Far more important in most economic studies is the ability for biopharmaceutical manufacturers to maximize the number of kilograms of material they can manufacture (and subsequently sell) in a year.  

Figure 2 shows the outcome of an analysis done for a retrofit project at a large biotech manufacturer interested in installing a new perfusion-based downstream processing technology. The net present value (NPV) of the scenario was $35 million over five years.

  Figure 2 compares the three most significant causes of error in economic evaluations: incorrect raw materials costing, construction and installation costs, and incorrect estimation of the quantity produced by that scenario. Each of these three categories is altered by 5%, 10%, and 20% respectively, to see their relative effect on the NPV. 

Figure 2.  A comparison of the economic effect (NPV: net present value) of calculation errors2 

As shown, the cost category that is most sensitive to errors is the quantity produced, or supply of material to the production network. Small inaccuracies in this metric produce significant negative impact on the NPV of the scenario in the case of a 20% inaccuracy, actually causing the project to have a negative return. Accurate supply-based models are essential to establish accurate metrics for the value of retrofit scenarios. In the next section, we discuss this type of modeling in a case study involving a major biopharmaceutical manufacturer. 

RETROFITTING PLANTS FOR HIGHER TITER PRODUCTS

Supply-based planning can be seen in retrofit projects that aim to allow higher titer products to be produced in existing biopharmaceutical manufacturing plants. With the rapid advances in cell culture and fermentation technology, many plants suffer from a bottleneck in downstream (purification) operations. A project with a major biopharmaceutical company was undertaken to establish the most cost-effective means of producing 4 g/L products in a facility designed for the 1–2 g/L range. Although a number of alternatives were considered, the two most practical options were determined to be split batching, in which the fermentation volume is split in two at harvest and processed in two separate lots and partial batching, in which a reduced fermentation volume is prepared, exactly enough for the downstream capacity. Both required similar infrastructure investments and resulted in similar downtimes for retrofit. 

The economic comparison of these two scenarios was therefore predicated on their comparative run rates and manufacturing production profiles. However, there was considerable disagreement from subject matter experts on which of the two scenarios was better. Batch-splitting could be performed after the Protein A recovery step, a possible plant bottleneck, but it was unclear whether the material would exceed time at ambient specifications if required to wait for the other half of the "split" batch to complete processing. Partial batching produced less material per batch, but seemed to maximize the downstream plant's capacity for each batch.  

The University of California at Berkeley and the Bioproduction Group, Inc., a company specializing in quantitative biotech process models, worked in conjunction with the manufacturer to produce a process-based analysis. Rather than focusing on data mining accurate costing for raw materials or construction costs, the approach was to build highly accurate virtual-plant models of both scenarios.  

Figure 3 shows indicative data of what Bioproduction Group calls "characteristic variability" in operating times of biopharmaceutical plant operations.

  Such variability is common in biopharmaceutical manufacturing and confounds process improvement efforts: reductions in the variability of a processing time are often more important than altering the processing time itself. Accurately modeling such variability is critical to establish accurate metrics around what a plant can produce. 

Figure 3.  Variability in unit operation processing times3 

One of the key issues with variability in operating times is that changes to one or more manufacturing unit operations may have unexpected changes in the performance of other (untouched) areas of the plant. A unit operation requiring additional cleaning, for example, may exhaust existing CIP/SIP capability, reducing the total capacity of the facility. These unforeseen bottlenecks are common in almost all biopharmaceutical processing plants, and to get an accurate estimate of the run-rate possible with a new unit operation, a detailed analysis of the facility must be performed that incorporates the variability data seen above. 

The production-based analysis performed used a technology known as process simulation. This technique produces detailed process models that incorporate variability in unit operations, media, and buffer preparation activities, CIP and SIP activities, pre-use and post-use operations as well as quality testing. Process simulations mirror plant automation systems in the sense that they will not start an operation until all the required resources are present, which is important in the case of time-sensitive protein substances. This technique confirmed that expiration times were not a risk for the split-batch scenario, and made it possible to quantify the exact profile of the quantity and timing of the material to be produced in the altered plants.  

Figure 4 shows a comparison of the split versus partial batch outputs over time, for varying campaign lengths between 0 and 160 days. (This output also includes the fermentation time, e.g., it takes nearly 30 days to produce the first batch.) Note that the lines cross at small campaign lengths, because the partial batch scenario requires shorter fermentation times. However, for significant campaign lengths of 40 days or more, the split batch produces 10–15% more output for an equivalent campaign. The return on investment (ROI) of this scenario was nearly $200 million over four years.

Figure 4.  Comparison of outputs over time in a 4 g/L process using split and partial batches. 

One of the final questions in an economic analysis is the value of increasing throughput at a plant. As we have discussed, most large-scale retrofit projects do not justify themselves on the basis of direct cost savings alone because indirect costs make up such a high percentage of total plant operating costs. As such, one of the key issues for biopharmaceutical manufacturers is how to increase operating throughput or to configure plants in such a way to make them more flexible. This is only the first part of the analysis, however: the increased production must be balanced against the manufacturer's ability to sell the additional material it produces.  

Figure 5 shows a traditional valuation of a retrofit scenario, producing additional material at a lower cost per gram in an upgraded facility. A typical ROI calculation will use the blue line (linear valuation) to calculate return; a capital investment delivers positive return if sufficient additional material can be produced. However, in a manufacturing environment where the additional grams produced cannot be sold, there is actually no direct economic benefit to investing in additional plant capacity or flexibility projects. Benefit is only derived when the flexibility or capacity allows the plant to be used for other purposes (such as production as a contract manufacturer) or where capacity increases allow savings in other places in the network (such as the divestment of a more costly plant). This non-linear tradeoff makes manufacturing economic analysis in the biopharmaceutical industry's current environment an even more difficult proposition. Valuation of a new scenario must therefore be carefully aligned with the strategy of a company. 

Figure 5.  A traditional valuation of a retrofit scenario, producing additional material at a lower cost per gram in an upgraded facility.  

In conclusion, biopharmaceutical manufacturing's current proliferation of capacity and the trend toward an increasing number of small-volume products creates unique challenges in manufacturing economics studies. The high indirect costs of biopharmaceutical manufacturing means that any plant downtime for retrofitting raise the bar for performance improvement efforts to overcome. Typically, improvements with significant benefit to the organization focus on creating additional capacity (or additional effective capacity, such as reduced changeover times). In such analyses, however, producing accurate process models is of vital importance in establishing correct metrics around the economic benefits of any scenarios considered. 

 is co-director at the Center for Biopharmaceutical Operations, University of California, Berkeley, and  

 is principal at Bioproduction Group, Inc. Berkeley, CA, 650.823.7533,  

 1. The University of California at Berkeley. Center for Biopharmaceutical Operations; 2009. Available frome: 

 3. Bioproduction Group. Indicative times only, not from any specific single biopharmaceutical manufacturer. Comparative durations for times  retained, X and Y scales normalized. Processing step includes inline testing. Rejected batches removed. 

 4. Yang A. Genentech Simulation System. CBO Workshop; 21 May 2008. Available from: 

Manufacturing Economics Series

A case study in capturing indirect costs and benefits.

Garbage In, Garbage Out: The Case for More Accurate Process Modeling in Manufacturing Economics

Pharmaceuticals came to the continuous improvement table later than some other manufacturing-centric industries, but noteworthy results achieved in the relatively short time since have influenced the industry's approach to manufacturing and allied business processes. Operational excellence and Six Sigma programs were rolling out across the pharmaceutical industry by the late 1990s, but only in the last five to six years have similar initiatives and programs, collectively referenced as continuous improvement, achieved critical mass and industry imperative status. 

The pharmaceutical industry was built, and has prospered, on a "hurry up and test" manufacturing model that valued quality but failed to optimize the processes by which quality was achieved. Today's global pharmaceutical industry is the product of rapid and dramatic changes, and continues to evolve in response to market and other shifts. For the long-staid, risk-averse, and change-resistant pharmaceutical industry, embracing the concept of continuous improvement meant radically rethinking the same old approaches.  

Albert Einstein wasn't referring to implementing a culture of continuous improvement across an organization as large as Pfizer Global Manufacturing (PGM) when he said, "problems cannot be solved at the same level of awareness that created them," yet this often referenced quote captures the overriding shift in thinking that success in such an endeavor demands.  

Historically, quality was the gauge by which pharmaceutical manufacturing processes were measured. For example, PGM had plenty of checks and balances for ensuring quality, but checks and balances alone were not enough to ensure that our manufacturing processes were optimally robust. As a result, we were spending too much time and resources solving problems and not enough time on improvement activities and proactively resolving issues. 

A scientific approach designed to foster systemic, ongoing, and value-added change, Pfizer's continuous improvement path began in 2003. Initially focused on understanding and improving the manufacturing processes that directly affected our products, Pfizer's continuous improvement program (incorporating Six Sigma, Lean, and Agile tools, among others) is now being expanded across PGM and into the other divisions and enabling functions of Pfizer.  

At Pfizer, all continuous improvement activities are built on five key pillars. 

Pfizer's Principles of Continuous Improvement

1. Continuous improvement is a journey of relentlessly seeking perfection. 

 2. Everything is a process; continuous improvement is not simply limited to manufacturing. 

 3. Improvement is not judged internally, but by our customers. 

 4. You must measure performance (cost, quality, speed) to improve the process.

 5. Those closest to the work know the issues and opportunities best; get the right tools in the hands of the right people. 

What follows is an exploration of how these principles were applied to develop and implement "Right First Time" and Lean initiatives across PGM to instill a culture of innovation and continuous improvement.  

The "Right First Time" (RFT) strategy, like subsequent Lean initiatives, is a core component of PGM's strategies for ensuring supply and continuing to meet customers' needs. The goal of RFT is to reduce the variation in processes and make them stable and capable of being further optimized. The output of a process cannot be predicted unless the process is understood, and only when a process is understood can risk mitigating or minimizing strategies be devised and optimization achieved. Process understanding, in turn, enables value stream mapping (VSM), a method (incorporated into Pfizer's Lean training) for evaluating the effectiveness of processes (and their effect in a larger system) in a value-added framework. 

Pfizer's continuous improvement program is now being expanded across PGM and into other divisions and enabling functions of Pfizer.  

To implement RFT across PGM, the perception that it was just another corporate initiative had to be overcome. Getting the right tools in the hands of the right people was essential, as was convincing PGM colleagues of the benefits of using those tools. At opposite ends of the colleague reaction spectrum, small groups of early adopters and skeptics book-ended the silent majority of PGM colleagues in the middle. 

Focused on winning that silent majority's confidence and buy-in, the roll-out strategy was three-fold: connect the RFT concept and mission to those of PGM and of Pfizer; demonstrate the active buy-in and support of PGM executive and site-level leadership; and, focus on the core processes of manufacturing, documentation, and laboratory testing. 

RFT gives all PGM colleagues, not just a handful of first-line supervisors and operators, tools for boosting process understanding, and developing and implementing strategies for optimizing process capability. Appointing a credible RFT champion at site-level (to date, more than 40 RFT champions have been appointed across PGM's global site network) has proved enormously successful, as did encouraging managers to motivate and inspire operators and other colleagues to use their brains as well as their hands. 

As RFT rolled out, all PGM colleagues were given basic Six Sigma knowledge through method 1, or yellow belt, training. Selected candidates with clearly defined projects subsequently received method 2, or green belt, and method 3, or black belt, training. The coaching provided to colleagues during these projects was critically important to success. Because process improvement is impossible without process measurement, new process capability metrics were introduced. As the rollout continued, we continually monitored progress and made adjustments as needed to optimize buy-in and support. 

Supported by Six Sigma-based (yellow, green, and black belt) training, colleagues were empowered to present managers with their ideas for improving processes. Site management recognized and applauded the success of colleague-devised RFT strategies, and, as more colleagues took notice, the impact rippled across sites and plants. RFT represented a real and significant departure from the old way of doing things, and this awareness combined with early successes to fuel colleague buy-in and support.  

Currently, some 5% of PGM colleagues have received green belt training, and approximately 1% of Pfizer colleagues have received black belt training. Applying RFT principles and Six Sigma techniques, PGM colleagues are completing more than one project per day (in a calendar year, more than 365 projects will be completed). It is also worth noting that 23,000 PGM colleagues have received one-day yellow belt training and, with these enhanced problem-solving skills, are on a monthly basis resolving thousands of issues across PGM.  

Yellow belt techniques are routinely applied to standard root cause analysis (RCA) for quality, and environmental health and safety (EH&S) investigations, and PGM colleagues use these skills to improve many other business processes.  

As the RFT roll out continued, the results of our optimization efforts drove variability from our core processes, which had previously recorded levels of variability that  ranged from 2–3 Sigma, whereas today, they consistently deliver a level of performance between 4 and 5 Sigma. As a result, we are experiencing significantly more consistent process output and fewer recurring problems. Our costs were decreasing while the speed and reliability of our processes were increasing. 

As its name implies, continuous improvement is organic and continuously evolving, and PGM's journey had just begun. As a result of RFT, our processes were capable, but were they efficient?  

BioPharm International-08-01-2009

BioPharm International, August 2009 (PDF)

Use of Change-Point Analysis for Process Monitoring and Control

Analytical Tools for Process and Product Characterization

Laying the Foundation

Garbage In, Garbage Out: The Case for More Accurate Process Modeling in Manufacturing Economics

Embedding a Culture of Continuous Improvement & Lean Manufacturing Across Pfizer Global Manufacturing

Recent Patent Rulings Can Impede the Progress of Personalized Medicine

Waiting for Godot: Will Venture Funding Return?

Health Reform Challenges Biotech Industry

Avoiding the Regulatory Catch 22