The pharmaceutical industry responded to the FDA's cGMP report with Process Analytical Technology, or PAT, a system to design, analyze, and control manufacturing and packaging processes through timely measurements of critical quality and performance attributes. The goal of PAT is to understand and control the processes with the assumption that quality cannot be tested into products, but rather should be built into systems by design.
The PAT framework aims to apply the "quality-by-design" tenet to ensure a predefined quality at the end of the manufacturing process, improving efficiencies while simultaneously reducing risks to quality. In-line measurements and controls will reduce cycle times, prevent rejects and scrap, improve operator safety and overall efficiency. The FDA has since created several subcommittees to provide recommendations on how PAT could be adopted throughout the industry.
"PAT fundamentally allows you to release your product without any additional release testing," explains Bikash Chatterjee, COO of consultancy Pharmatech Associates. "Currently you sample it, it goes to a lab, you test it, and then you can ship it. We've learned that release testing alone isn't always a good indicator of how well the process is working. PAT says that if you can demonstrate that you're controlling and monitoring the critical attributes associated with the process—the number of pouches, verification of that number, the proper label, the right country, legible printing, et cetera—the packaging machine can guarantee that the product will meet your quality standards. It's the ultimate quality assurance."
PAT is taking off slowly—it requires a high level of collaboration between the customer and the supplier during the equipment design and development process. It also requires smart machines and sensors that can communicate not only the state of the process, but the state of the sensor as well. This has led to an entirely new generation of integrated controls.
In the past, equipment simply controlled its own functions, and three or four other systems tracked what was being fed to it, or measured downtime, or tracked performance. But now the machine "knows" when it's functioning properly or when it's not.
--By Brian Pelletier, Contributing Editor
