PAT, APC, and beyond
The journey toward a culture of innovation in life science process automation
- Can process innovation and regulation co-exist?
- What is holding PAT back?
- Inline data collection is useless unless you can ultimately feed it back to an optimization loop.
By Jonathon Thompson and Larry Balcom
While the idea of inline data collection and real-time product quality analysis has been common in less-regulated continuous process industries for many years, regulatory constraints have limited adoption in the life sciences. In 2004, to loosen the regulatory stranglehold on process innovation that had for so long slowed the industry in the technological dark ages, the U.S. Food and Drug Administration (FDA) launched a process analytical technology (PAT) campaign for the life sciences. This included a combination of standards, best practices, and strategies for minimizing innovation-related risk in the pharmaceutical, biotechnology, medical device, and animal health industries.
Two FDA guidance documents defined the PAT campaign: "Pharmaceutical cGMPs for the 21st Century-a Risk Based Approach" and "PAT-A Framework for Innovative Pharmaceutical Development, Manufacturing, and Quality Assurance." Both guidance documents were well-received by the industry and sparked a great deal of conversation about the future of manufacturing. But while the future prospects were exciting, the enormity of such a transformation stifled early optimism. The industry had too long been regulated into a culture in which a static process was the only safe process. Patient safety is, after all, the number one driver in drug manufacturing, and no one wanted to interfere with that. The wheels of change moved slowly.
Over the next several years, industry organizations offered their input on the best path forward, issuing standards such as the International Society for Pharmaceutical Engineering's GAMP V and ASTM's E 2500-07, all aimed at fostering innovation in regulated industries. These standards helped to define what a "risk based approach" would mean in the development and validation of systems and processes. The ASTM standard in particular also offered PAT as a data collection strategy for providing risk management. These standards helped to form a framework by which the industry could begin to develop products.
Following these new standards, many companies embarked on a journey towards a culture of process innovation. This has largely begun with the practice of designing quality in at the new product development stage, rather than testing it at the end of the process. Drugs take many years to move from the lab to commercial production, however, so drugs manufactured today show little impact from those 2004 guidance documents. This leaves us still with the challenge of taking already operational, validated, and, most importantly, FDA-approved processes and inserting principles of process innovation, which brings us back to PAT.
PAT, according to the FDA, is a system for designing, analyzing, and controlling manufacturing through timely measurements of critical quality and performance attributes of raw and in-process materials and processes, with the goal of ensuring final product quality. The implementation of PAT depends on the ability to identify the critical-to-quality variables of a product and building inline measures of those variables into a system. These became the initial constraints for PAT implementation and presented many challenges.
First, PAT requires a deep understanding of the impact of the process on the product. More specifically, there must be an understanding of the impact of the process on the critical quality attributes. The attributes themselves were easy enough to determine because they were already being tested in the finished product release, and the extensive drug development process forced the company to understand every aspect of these attributes. But the disconnect was in determining the impact of the process on these attributes and in isolating the process conditions under which they were optimized. Process knowledge from various groups involved in the development, scale-up, and validation ultimately helped to determine where and when measurements should be taken. Once the when and where were resolved, the next step was the how: building effective inline measurements meant understanding the real-time state of the product.
Real-time measurement then, is the second constraint to PAT implementation. The market for inline instrumentation to measure physical attributes (conductivity, pH, temperature, pressure, etc.) was established, and these instruments were widely available. When life science companies began to understand the critical quality attributes, they realized that the spectroscopic and other complex methods being used in final product release testing would need to be duplicated inline. This was a challenge, but one quickly met by the instrument market. Today, due in large part to PAT, the number of spectroscopic and other complex instruments has grown to take these measurements online. They have been designed to be non-invasive to the process and therefore can be implemented with less downtime and revalidation. This gave companies the ability to begin inserting instrumentation into their existing processes and collecting an immense amount of data about the impact of process variables on the critical quality attributes.
Now, eight years removed from those original guidance documents, we have seen that most companies have begun a level of cultural change, which will help move the industry to a culture of process innovation. New development is being approached by building quality into the process. Current processes are being measured, and the data is being used to generate process models and understanding. So, where does the industry go from here?
To date, the PAT initiative has been largely data-driven. Data may be turned to information, but after the process and much too late to have any impact on the quality of the product being measured. So, the next step in this evolution is to follow the example of the process industries and turn this data to information in real time. Furthermore, the information must lead to action. The next step beyond PAT, then, is Advanced Process Control (APC).
The industry has learned since implementing PAT that data collection alone has minimal impact on product quality or the bottom line. Off-line data evaluation gives information, but too late. The best path to derive the benefits of PAT is to build feedback loops to turn the data to information and act on that information in real time. Such a system enables higher quality product, reduced rework and scrap, reduced risk of product recall, and real-time release or release by exception. APC allows information to be immediately fed back into the process to optimize the critical quality attributes. APC in the continuous process industries works independently to optimize a process. For the regulated, batch-based processes in the life sciences, a modified view of APC will be required.
Much like the less-regulated process industries, an APC system that provides process optimization capability can be the cornerstone of a real-time information system. For batch-based processes, a batch control system is also required along with an APC system to communicate with it. The impact of the process parameters on the critical quality attributes varies over time during a batch process. This variance must be understood by the APC system if optimization is to minimize batch variance. Of course, an APC system will also need to communicate with the automated control systems, which is absolutely necessary when the APC system is performing closed loop feedback control and online process optimization. However, online optimization by the APC may not be desired, nor be the best solution for every situation.
An APC system should also allow for open loop feedback control, where data can be fed directly to people who can then use it for real-time decisions. To aid the people in these decisions, automated process analysis can be used to digest process conditions into recommend actions. An online statistical process control system will have the capability to take the data from an APC or control system, model it against a "Golden Batch" profile for the product, and create real-time information about the batch process. This information is fed into a visualization system that can display run charts, alarms, trends, etc. An advanced visualization system should also include an expert advice system to give operators direction in using the information generated. All of this, of course, must meet 21 CFR Part 11 requirements for electronic records.
Electronic records requirements can be managed in a number of ways. Independent databases can be created and validated for each component of the solution. But the simplest approach is to utilize a centralized historian for all critical quality records. This can be an existing plant historian or a new system implemented just for the APC solution. Either way, a centralized record storage location will reduce the amount of initial and ongoing validation required.
An example system architecture configuration.
The figure illustrates a PAT, or Quality by Design (QbD) architecture, that can facilitate understanding and controlling variation in a life sciences process. In this architecture, basic process control functions are provided by a programmable automation control system (PAC), which integrates with an APC system and dynamic optimization software.
Monitoring the illustrated process in real time is a process-monitoring software platform, which provides rigorous process modeling, soft sensor capabilities, and real-time SPC analysis. All of the measured and calculated process data is stored uniformly in the process historian that supports 21 CFR Part 11 commonly deployed in the life sciences industry. Process visualization is provided by HMI software, as well as real-time SPC software.
APC is the next natural step in the evolution of PAT, but it is certainly not the last. After all, the intent of process innovation is to continue to innovate and optimize. If the life sciences are to be successful, the solutions they choose will need to be open platform, modular, expandable, and flexible enough to meet whatever innovations are ahead. There must be a close partnership between the life sciences companies and the vendors providing these solutions to eliminate as many limitations as possible to process innovation. Because, like process innovation, PAT is not a destination but a never-ending journey to higher quality and greater cost savings.
ABOUT THE AUTHORS
Jonathon Thompson (firstname.lastname@example.org) is the Senior Manager of Regulatory Compliance Consulting at Invensys Operations Management. He has a BS in Chemistry from Indiana University and an MBA from Indiana Institute of Technology. Jonathon has more than 15 years of experience in the life sciences with management roles in operations, validation, automation, and software. He specializes in the application of technology to optimize regulated processes.
Larry Balcom is a Senior Principal Technical Sales Consultant at Invensys Operations Management with more than 25 years of experience in manufacturing support, control system design and integration, technology development, reactor design, and pilot plant development for the specialty chemicals industries. Larry has been creating advanced applications solutions for clients for six years and has expertise in software integration and communications. He has provided technical leadership in emerging markets, such as biofuels, power, life sciences, and specialty chemicals. Larry is a life member of the Phi Kappa Phi engineering honor society.