1 April 2007
A big pill to swallow
Designing quality into processes helps manufacturers gain an edge in FDA audits, safer products
By Ellen Fussell Policastro
In March, federal inspectors found the strain of salmonella behind a recent food-poisoning outbreak at the ConAgra Foods Inc. plant that produced tainted peanut butter. In February, ConAgra recalled all Peter Pan and Great Value peanut butter made at the Sylvester, Ga., plant after federal health officials linked the product to an outbreak that began in August. The recall now includes all such products made since December 2005. The FDA said finding the salmonella in the plant environment further suggests the contamination likely took place before the product reached consumers.
In manufacturers' ongoing quest to be more proficient, quality-driven, and of course, compliant with federal regulations, more and more are initiating formal programs and technologies to avoid costly mistakes like the peanut butter SNAFU, and to stay ahead of the compliance curve. Using good automated manufacturing practices (GAMP) and process analytical technologies (PAT) are two ways manufacturers can make sure they not only shine in FDA audits, but keep consumers healthy with better, safer products.
The objective behind GAMP is to achieve "validated and compliant automated systems meeting all current healthcare regulatory expectations, by building upon existing industry good practice in an efficient and effective manner," said Dennis Brandl, president of BR&L Consulting in Cary, N.C. Some of the benefits behind the concept include a better understanding and a common language and terminology. Not only does it help manufacturers improve their expectations with compliance, it clarifies the division of responsibility between user and supplier, he said. It also reduces the cost and time manufacturers take to achieve compliant systems, and it provides better visibility of projects to ensure delivery on time, on budget, and to agreed quality standards.
Proof is in process
A significant way to practice GAMP is taking hold in unconventional industries such as pharmaceuticals. PAT is a strategy that uses the physics and physical characteristics of a process with process control technologies to maintain process control at the highest standard. (See related sidebar, "PAT means quality by design") "It's not really just process control, but a parameter multivariable based on process control," said Roddy Martin, general manager and vice president of chain research at AMR Research in Boston. It is well known in the brewing industry; wherein fermenting 10,000 liters of beer, different physical attributes of the process will tell you whether the process is under control-purity of the CO2, the layer temperature of the beer, and the turbidity of the product.
Before using PAT, manufacturers could have instruments measure each parameter independently, but they'd only know a process was going out of control when one of those instruments had a problem. "PAT gives you the capability of looking at each one of those parameters at each stage of the process and identifying them when you start to go out of control, not when you're already out of control," Martin said.
While PAT has existed in the chemical and brewing industry for years, the FDA has revitalized it for the compliance aspect in life sciences and pharmaceuticals. Yet innovation in pharmaceutical manufacturing in the past has been constrained "due to regulatory uncertainty," said Velumani Pillai, director of strategic architecture at Pfizer in New York. A paradigm shift now indicates quality has to be "designed in, not tested into products, which requires a comprehensive understanding of the process, the impact of product components on quality, process variability, and a mechanism to manage it," he said.
The petrochemical and brewing industry have been doing this for years because they operated on a low margin, needing to know before the process went out of control. "You can't brew 100 liters of beer and then find out only four out of 10 came out," Martin said. The process is new for the pharmaceutical industry because their margins have been historically high. "If a certain amount was not up to spec, then they'd just scrap it. But now, producing products with active ingredient that could be distributed unevenly, it's important for manufacturers to demonstrate they are in control at all stages of the process so they can move companies against parametric release" (only having to check several tablets instead of the entire batch).
Measuring medicine a team effort
In the pharmaceutical industry's tableting process, for instance, active ingredient comes in the form of powder ground up and uniformly batched and mixed to make a tablet. "But sometimes the active ingredient isn't spread evenly across the tablet," Martin said. "I've seen tablets in which all the active ingredient is spread on 75% of a tablet; that's not uniform. If someone wants to take half a tablet, they could get half with no active ingredient." PAT uses near infrared and other performance instrumentation to assess the distribution of the active ingredient, so when you actually form the tablet, you have a uniform distribution of the active ingredient.
Using instruments in process analytics to do a sophisticated version of process control has led to big opportunities for the instrumentation industry. "But just putting more instruments into process isn't the answer," Martin said. "The key issue is a fundamental understanding of the physics and the parameters of the process itself." The pharmaceutical R&D community will design a product and give it to manufacturing, which has to keep the process under control.
"Continuous improvement is a critical element in a sound quality system," Pillai said. "A systems approach, the PAT framework, includes continuous improvement, risk assessment, knowledge management, and online sensors." Process analytical technologies can help with "better design, monitoring, process control, and prediction of process performance," he said. PAT enables this process understanding at different levels, whether it is to monitor, control, or predict process performance. Today these capabilities are available as individual components in different systems. What's needed, Pillai said, is "an integrated environment that combines the modeling tools for design/analysis, process analyzers, process control/optimization, and knowledge to efficiently apply these technology innovations to pharmaceutical manufacturing."
An example of one of these analytical systems is an intelligent sensor developed with specific advantages for biotechnology and pharmaceutical industries, said Ken Queeney, product manager at Mettler-Toledo Ingold, Inc. in Bedford, Mass., in a paper on intelligent pH and oxygen sensors (www.isa.org/link/Phperformance). "Pre-calibration offers advantages in harsh working environments of the chemical industry, where sensor wear and adaptive calibration information is insightful as well," Queeney said. "Quality engineers appreciate the 21CFR Part 11 conformity, adaptive calibration timer, enhanced sensor diagnostics, and peak temperature reading to assure accurate readings and reliable control."
Smart sensor/transmitter systems can trigger alarms based on poor calibration sensitivity or unacceptable offset. Manufacturers can also measure response time. "You can establish sensor performance trends by tracking historical sensor information," he said, "and you can use these trends to forecast future performance. It's this forward-looking aspect which distinguishes smart systems from the more intelligent systems," Queeney said. And manufacturers are implementing these types of prognostics on plant-wide control systems as well in analytical transmitters.
Although technology is making its mark in improving pharmaceutical processes, collaboration is still key, Pillai said. Common PAT software takes committed suppliers and end-users to work together. Some examples of PAT spectroscopy methods in adaptation are near infrared, raman, UV-visible, fluorescence, and acoustic. "We anticipate that common functions will be identified first," and products will emerge later, he said. To enable common PAT software and make information exchange meaningful, Pillai proposes end users standardize these complex measurements.
R&D, engineers collaborate
Before PAT became commonplace, the most significant issue was R&D's isolation from engineers, Martin said. They designed a product and "threw it over the wall to manufacturing to interpret R&D requirements and to manufacture the process that controlled the product to the right production level." Now with PAT, engineers and R&D work closely together when designing a process and product so they can understand the process parameters they need to track to help manage a PAT capability. "It's like sticking a fork into cake batter before the cake is done to measure the consistency of the batter during baking," Martin said. "If the batter sticks to the fork, it isn't done." With a blending and crushing process (powdering) that has tableting as its outcome, engineers are working with R&D to make sure at all stages of mixing and blending of the process there is a uniform distribution of active ingredient. Now, when engineers work with R&D, and they're looking at various stages of grinding and tableting, there is a better opportunity to understand the key performance measures at each stage of the manufacturing process in order to ensure a conforming product at the end.
"We have heard about several efforts to build master plans for deployment of innovative measurement technologies," Pillai said. "But most of them are done in isolation without considering the impact on architecture, existing application, and infrastructure needs." Any master planning effort has to consider these impacts. The master planning effort has to align strategies and derive specific objectives for infrastructure, applications, integration, and PAT systems. "To do this effectively, we'd need to map all functions to domains. We should consider new functions such as process optimization and process improvement in manufacturing system architectures. For each of the functions, we need to consider impacts on levels and approaches to automation."
Building in quality
The dreaded salmonella is cropping up in other products pending recall, Martin said, because some manufacturers are doing quality control instead of quality assurance. "They're not looking at the process end of production," he said; "they're looking at testing at the end and finding out there are broken parts in the process. When R&D works with engineering, they can design more scalable products with better process control, and more effectiveness in product outcome.
Most pharmaceutical companies used to run with a level of 50% rework and 40% capacity utilization. "They could tolerate this in the past because of lower margins," Martin said. "There's a huge squeeze on margins, and they can no longer operate that way. The bottom line to the benefit is to improve capacity utilization to eliminate waste and rework and to produce safer products with higher effectiveness." It all boils down to safe and effective product. The pharmaceutical industry knows they have to do this at a lower cost.
"The problem we now face is one in which individual instrument vendors offer their own proprietary software, which results in unique PAT applications," Pillai said. While we tolerated this situation through the infancy of PAT, in the future, PAT applications need to operate and acquire data from the process in real time to support three PAT paradigms" (monitoring, controlling, and optimizing).
Some PAT systems are used not only for monitoring processes but to control and predict process performance, Pillai said. "We need to exchange data with systems that perform analysis, control, decision-making and reporting. We should start with standard ways of exchanging data across these existing systems." In this new environment, the process communicates about its performance. "Who really owns this information? What are the immediate uses for this data? Any exercise to determine this should involve process owners," he said.
ABOUT THE AUTHOR
Ellen Fussell Policastro is the associate editor of InTech magazine. Her e-mail is email@example.com.
PAT means quality by design
Process analytical technology (PAT) is about designing quality into a process. It is more of a guidance, rather than a regulation. In 2002, the FDA published a report explaining a new regulatory framework, called "Pharmaceutical CGMPs for the 21st Century - A Risk-Based Approach." As part of this report, in 2004, the FDA published a "Guidance for Industry" called "PAT - A Framework for Innovative Pharmaceutical Development, Manufacturing, and Quality Assurance." The guidance is mainly to reduce variability by gaining a better understanding of a process than manufacturers could get by a traditional approach.
In section 14 of the "Guidance for Industry," the FDA defines PAT as:
A system for designing, analyzing, and controlling manufacturing through timely measurements (i.e., during processing) of critical quality and performance attributes of raw and in-process materials and processes, with the goal of ensuring final product quality. It is important to note that the term analytical in PAT is viewed broadly to include chemical, physical, microbiological, mathematical, and risk analysis conducted in an integrated manner. The goal of PAT is to enhance understanding and control the manufacturing process, which is consistent with our current drug quality system: Quality cannot be tested into products; it should be built in or should be by design.
The focus of this new regulatory environment is on discovering process variation and controlling that variation when it might contribute to patient risk. The process variation is discovered by identifying and measuring crucial quality attributes in a timely fashion. In this way, processes can be developed and controlled in such a way that the quality or product is guaranteed.
Being able to analyze the sources of process variability and control that variability could provide engineers with a basis for making equipment changes and scaling up processes.
SOURCE: New Directions in Bioprocess Modeling and Control, by Michael A. Boudreau and Gregory K. McMillan.