QbD Takes Biopharma by Storm

Aug. 21, 2006
Quality by Design has lost its buzzword status as it becomes a reality for more biopharma companies. Contributing Editor Angelo De Palma examines how Wyeth, Schering and other firms are advancing the concept, and interviews FDA’s Moheb Nasr for his perspective.

Quality by design (QbD), risk-based manufacturing and process analytical technology (PAT) have become three cornerstones of the U.S. Food and Drug Administration’s effort to modernize pharmaceutical and biotech manufacturing. While FDA has not promulgated an official definition of QbD, there is a general understanding at the Agency of the major elements. According to Moheb Nasr, Ph.D., Director of CDER’s Office of New Drug Quality Assessment, QbD incorporates five ideas:

    1. A “system” approach, where drug products are designed to meet patient needs and performance requirements.
    1. Design of a process that is consistently capable of meeting critical quality attributes.
    1. Thorough understanding of the impact of the starting materials and process parameters on product quality.
    1. A strategy for identifying, understanding and controlling critical sources of variability are identified, understood, and controlled.
  1. Process monitoring that enables continuous, consistent quality.

According to this roadmap, the QbD philosophy is identical for biologics and small molecule drugs, although the individual unit operations vary considerably.

Some of these ideas have been in place at top companies for years while some, particularly the “system” approach, is new. The prevailing tack today is empirical rather than system-oriented, with process optimization as the trump card. Through the system approach, the process is designed to meet product criteria, while the product is designed to address clinical needs. Other novelties here: a control strategy for understanding variability and assuring consistency over time, and the acknowledgement of risk and variability. “In traditional processing we avoid variability,” Nasr told Pharmaceutical Manufacturing. “Here we face it as a fact of life.”

To support industry’s adoption of the system approach, FDA has instituted a CMC Pilot Program, managed by the Office of Drug Quality Assessment. The program allows companies to share their QbD experiences through their New Drug Application (NDAs). Most major worldwide pharmaceutical manufacturers are participating. But this program is inherently limited, Nasr says. For industry to realize the full benefit of QbD, it must develop and adopt novel manufacturing platforms, for example continuous vs. batch processing.

Industry is enthusiastically but cautiously testing FDA’s position, attempting to gauge what level of regulatory flexibility will accrue through information-sharing. FDA, meanwhile, promises that if companies are forthright, they will be freer to introduce manufacturing changes without resubmitting the process for regulatory review.

Sufficient examples in pharmaceutical industry lore suggest that industry “talks the talk” with respect to regulatory initiatives to keep FDA happy, rather than adopting the new program on its merits. QbD is apparently not one of these. “The evidence is in their investment in new technologies and platforms. They won’t do that just to make us happy, there has to be a strong business case,” Nasr says.

FDA has already approved an application through its CMC pilot program, but the long-term benefits of QbD are still unknown. “It looks promising, but until we have a number of applications approved this way, it will be difficult to measure the financial rewards.”

In the Beginning

The starting point for QbD in biopharm is the realization that products are heterogeneous mixtures consisting of multiple active and inactive isoforms, plus numerous other proteins and impurities. Bioprocessors tend to worry more about productivity and yield, especially when a product’s heterogeneity is part of its regulatory application. Yield is a more critical factor in supply than isoforms, but some products, notably Immunex’s Enbrel monoclonal antibody for rheumatoid arthritis, depend on very narrow glycosylation patterns for their activity.

Marta Czupryn, Ph.D., Senior Director for Bioprocess Development at Wyeth Biopharma (Andover, Mass.) defines QbD as “achieving acceptable mixtures of product isoforms, such that the product meets purity and other acceptance criteria.”

A Wyeth BioPharma scientist performs mass spectrometric analysis of a biopharmaceutical product candidate.

Wyeth QbD efforts begin with a defined process beginning with cell line selection and continuing through cell culture and purification. Quality is enhanced by streamlining and defining standardized production methods, all the while applying analytical methods that indicate if the design has led to products that meet specifications. “We’re sometimes surprised,” says Czupryn, “but at least we can detect problems early on.”

One method used at Wyeth is high-resolution mass spectrometry (MS), which easily discriminates among isoforms differing by a single sugar residue. Peptide mapping plus MS, which is applied to digests of protein mixtures, is even more sensitve, according to Czupryn. Where glycosylations represent the only significant differences, it is possible to utilize an orthogonal technique, releasing the sugars and analyzing the peptide backbones for mass and sequence.

MS analysis has been automated in laboratories, to the point where hundreds of samples are run untouched by human hands. But this is still a far-off dream for cell culture, where everything is done offline, some distance from the fermenter. “In a PAT world, this type of analysis would be done inline or at-line,” Czupryn notes, “and if certain isoforms can be correlated with process conditions, one could change bioreactor conditions appropriately. But that’s in future.”

If QbD and PAT are about process understanding, tying measurements to process parameters to quality is clearly assisted by platform processes that are broadly applicable to product classes (e.g. monoclonals). By reducing development timelines and assuring a built-in, baseline quality level, process development can better correlate measurements with quality and apply that understanding across the platform to other products. This has been Wyeth’s strategy for some time. “We know we can design a platform-level process such that it delivers dependable product, which gives us more time to spend unique quality issues,” Czupryn says.

For example, Wyeth’s platform process for antibodies and fusion proteins has been standardized to harvesting, clarification, filtration or centrifugation (for smaller or larger batches, respectively), protein A capture, polishing and viral clearance.

Not by Analytics Alone

Product quality is not simply a function of fancy analytics, but the degree of alignment among processes, technologies, and people, says Doug May, director for strategy, life sciences at RWD Technologies (Baltimore, Md). “It’s possible to get very high-quality product from a low-tech manufacturing environment.”

RWD, a lean manufacturing engineering consultancy, tends to view quality more holistically or organically than the pharmaceutical industry, which tends to compartmentalize competencies and initiatives. For example, where lean practices have taken hold in most manufacturing industries, not everyone in pharma/biotech is convinced they apply to everyday quality issues.

May disagrees. “Lean can significantly and positively impact even nitty gritty operations. It’s about arriving at a desired output in the fastest, most efficient, and cost-effective manner by zooming in as far down into a process as is necessary, down to what happens in a fraction of a second,” he says. For example, once a process is optimized and validated, lean can be applied to ensure that variables never fall out of specified ranges, which is exactly what PAT as a mechanism for QbD is all about. Moreover, PAT works seamlessly with lean, since the two form the bookends of in-process problem resolution. Whenever process variables drift out of specification, lean kicks in with root cause analysis to determine the proper fix.

“PAT is really data analysis across multiple data sources,” notes Mark Gallant, enterprise life sciences and pharmaceutical executive at OSI Soft (San Leandro, Calif.). The company, which has monitored FDA’s PAT initiative since the beginning, offers software to perform multivariate analysis within the varied contexts of pharmaceutical and bioprocessing.

To a lean practitioner, the FDA’s positions on quality focuses too narrowly on operations. “I was taken aback by the fact that nothing FDA says about quality, or quality by design, accounts for the human dimension.” The focus on high- vs. low-risk variables, for example through utilization of quality maps, is a good start. But lean, May asserts, supports a more comprehensive approach to quality.

Old Meets New

QbD is not a new concept, having entered pharmaceutical parlance through its general use in other industries. As Robert Seltzer, compliance manager for GMP at Schering (Berkeley Heights, N.J.) notes, water systems have employed real-time, fully online conductivity and total organic carbon detectors for years. These analyzers, which are usually built into the water purification system, provide instant feedback with no need to validate or perform offline testing.

Tools such as QbD "don’t get around from company to company that easily" in pharma, notes ASQ's Singer. Because there’s such a steep learning curve, staffs at each drug firm have to discover everything on their own, he says, whereas in other industries, companies share such information.

At a high level, QbD refers to overall quality and tools to help achieve it. The term eventually worked its way down into nitty-gritty activities like drug and process development, even research. According to Donald Singer, a director at the American Society for Quality (ASQ; Collegeville, Pa.) and an employee at a top (but alas, un-nameable) pharmaceutical company, the key element of QbD is measuring process parameters up front to reduce risks and quality-related issues later on.

“In manufacturing, the idea goes back to scientists and engineers learning and utilizing quality tools at the product design phase. Whether it’s a drug, a process for manipulating or formulating drugs, or an IT system that controls how drugs are manufactured, it’s QbD,” he says.

QbD in pharma/biotech and other tightly regulated industries retains its historical connection with statistical methods through such tools as FMEA (Failure Modes and Effects Analysis), formal risk assessment programs, HACCP (Hazard Analysis of Critical Control Points, originally used in the food industry) and statistical process control.

FMEA is utilized at the lowest level, not only by top-down designers off site by process engineers on the plant floor. One of the first steps of FMEA is defining and detailing a process map: What steps in the process or formulation lead to the finished product? Once determined, those steps are broken down further, for example into operations that are adequately controlled and those that are not, and key measurement points. At this stage, engineers are interested in improving control through chemistry, equipment, or instrumentation, to build in greater consistency.

Issues and Hurdles

The major issue the biotechnology faces in its quest to relate process to quality is its industry’s historic reliance on “testing in” quality – demonstrating product quality by analyzing samples at the end of, instead of during, the process. “What this means is that in much of pharmaceutical processing the quality is merely inferred,” observes Ian Storr, managing consultant at PA Consulting Group (London, UK). Process control is also assumed inferentially, the reasoning goes, since if the product meets specifications the process must have been okay.

Given the consequences of a batch falling out of spec, Storr questions whether this exercise, which is standard operating procedure in biotechnology, is related to quality at all. A QbD program and its attendant tools, he says, achieves not only process control and continuous improvement, but in-process cost control and greater confidence in the product.

If that is the vision of QbD, then development engineers must first identify which parameters are critical to product quality, and then determine a means of measuring them in real time. This is never easy, given short timelines and biotech’s heavy burden of history and tradition. “It comes down to learning what things really matter, as opposed to things you have simply measured in past,” Storr says.

The idea behind QbD is to engineer robustness into a process by exploring parameters critical to the synthetic or biological route through multivariate analysis and other statistical techniques. None of these approaches is particularly novel, nor is the idea of measuring process parameters during manufacturing. Integrating these tools end-to-end, for process understanding, from cell culture to fill/finish, is somewhat revolutionary, at least for pharmaceuticals and biotechnology. The question is, how deeply will engineers delve into their processes, and for how long, given routine bioprocess timelines?

If the mutual goal of process improvement and QbD is a coherent process to turn over to the manufacturing group, then it’s impossible to overlook the importance of technology transfer. Ian Storr believes the transfer is best done when information moves both ways, with production groups providing feedback during process parameter identification and validation. “The experience of manufacturing in determining critical parameters is indispensable,” he says.

Another significant hurdle to adopting a QbD culture is familiarizing a critical mass of project participants with quality methods, and using those tools to best advantage for a specific product within a particular company. According to ASQ’s Donald Singer, quality infiltrates biotech operations more slowly than it should because of the industry’s competitiveness and secrecy. “These tools don’t get around from company to company that easily. And because there’s such a steep learning curve, they have to discover everything on their own. Other industries share this information,” he says.

Another roadblock, especially for early-stage companies, is the tradeoff between quality’s long-term benefits and shorter-term goals of meeting development milestones. “Some companies have great difficulty building in quality [over the] long term while focusing on short-term operations,” Singer remarks. While quality tools can generate significant cost savings after a very short implementation period, the most significant effects of QbD are long-term.

It’s a Living Thing

The complexity of bioprocesses make them far more difficult to analyze and control than chemical reactions. Within reasonable temperatures and pressures, it is quite difficult to “kill” a nucleophilic aromatic substitution. One might obtain more side products than is desirable, but something can almost always be salvaged from the reaction. With biologics, the slightest perturbation can affect quality. Serious breaches or excursions result in cells dying or becoming infected with adventitious pathogens, at which point the batch is lost.

That this occurs infrequently is a testament to the soundness of biological manufacturing practices, but paradoxically, it has also been applied to rationalize not using PAT or even rudimentary real-time analytics to control a bioprocess. Bioprocessors traditionally monitor temperature, pH, nutrients, and sometimes waste materials, but the consensus is that messing with too many process parameters too frequently will lead to trouble. “Since it’s a living production system, there are probably fewer opportunities to interrupt and control it aside from those simple parameters,” observes PA Consulting’s Storr. “Which is what makes application of some of these tools in biologics manufacturing so complex, and potentially so rewarding.”

About the Author

Angelo De Palma | Ph.D.