Were you planning to attend the 29th International Forum for Process Analytical Chemistry (IFPAC) annual meeting in Arlington, Virginia, but something white and fluffy got in the way? You certainly weren’t alone, but darn it if you didn’t miss a good meeting this year.
Many would agree, I think, that effective process understanding and control through the practical and focused application of technologies like NIR and Raman is a trend that will only continue to gain momentum, especially in Generic Pharma circles.
Meanwhile, it’s clearly evident that the analytical chemists and others in Big Pharma have moved well-past piloting and studying the efficacy PAT in their processes and have moved on to confidently apply these technologies to improve operational excellence.
The presentation “Enterprise PAT Solutions for Commercial Manufacturing: A Case Study on Real-Time NIR Blend Monitoring and NIR Tablet Assay,” authored by Bristol-Myer Squibb’s (BMS’s) Elvin Varghese and Optimal Industrial Optimization’s Andy Sweeney, provided IFPAC attendees an unambiguous look at just how effective PAT, in combination with well integrated informatics and data management systems, can be at delivering advanced cGMP process control.
According to BMS, implementation of PAT and a toolbox of analytics to monitor blend uniformity and tablet consistency provides predictive intelligence, helps assure quality, mitigates risk and ultimately brings R&D and manufacturing into a more collaborative, constructive relationship. The company’s vision was to achieve real-time analytics in pursuit of process control and process understanding through 1) “Consistent user interface for [a] diverse collection of devices; 2) closed-loop feedback control; 3) Data management for control and analysis; and 4) Analysis tools and interfaces.”
At the heart of it, says BMS, is Emerson Process Management’s synTQ PAT data management package which integrates inputs and data flow to and from PAT instruments, multivariate analysis, lab operations, process control equipment, operator interfaces, a prediction engine, the BAS, the plant historian and archives. Bristol-Myer Squibb’s presenter explained that NIR spectra data is acquired from the four blending steps required for an unnamed OSD formulation including API and excipient blending and the addition of disintegrates and anti-adherents.
Blend monitoring, says BMS is managed via an endpoint detection control algorithm from MatLab. Applying MatLab’s PLS regression function for API concentration model prediction, operations personnel can calculate and understand run differences and torque averages taken from the blender shaft. The algorithm checks predictions against product-specific Acceptance Criteria, which determines end points in support of blend uniformity. If something goes astray, the algorithm detects it and provides closed-loop feedback through synTQ, which then can send a stop signal to the blender’s programmable logic controller. The presenter noted that data is also being delivered to the plant’s building automation system to adjust environmental conditions to suit blend uniformity.
Bristol-Myer Squibb showed a slide representing the company’s PAT architecture that revealed the platform-to-platform integration with synTQ connecting its NIR spectrometers in both electronic and paper-based batch record plants, to and through the plant LAN, to the Services LAN and ultimately to the Business LAN.
It occurred to me as I pondered the snowpocolypse that was looming over New York, that effective PAT implementations, like the one BMS put together, let process and operations people characterize and understand the attributes of individual snowflakes (i.e., the APIs and excipients) within the blizzard that is the batch in blend process. Perhaps if the forecasters had a chance to understand the form and moisture content of the snowflakes falling on NYC in real time, they might have been able to understand that conditions did not support shutting the city down for the sake of a foot of snow.