Pharmaceutical companies can build more robust manufacturing processes with the use of modeling, according to speakers in the session “Utilizing Modeling in Manufacturing” of the 2020 PDA/FDA Joint Regulatory Conference. The Sept. 15 session was moderated by Ellen A. Huang, Senior Advisor, ORA, U.S. FDA.
The first speaker, Alina A. Alexeenko, PhD, Professor and Co-Director of the Advanced Lyophilization Technology Hub at Purdue University, talked about the growing importance of modeling as lyophilized use of lyophilization in injectable drugs expands.
During her presentation titled, “Modeling Lyophilization,” Alexeenko explained the three stages of lyophilization and how modeling applies to each: freezing, primary drying and secondary drying. Its use is widely accepted for the primary drying stage; it is in initial development for the freezing stage; and it is still under development for secondary drying.
Alexeenko said the boundaries of design space, historically measured experimentally, can now be calculated using modeling and more limited experiments.
Modeling allows for improved design and development of the lyophilization process, supports ongoing manufacturing in cases of deviation and enhances the ability to transfer technology to different products and equipment. Overall, she said, modeling could significantly accelerate the lyophilization process development, optimization, transfer and scale-up. It can also assist with process validation, post-approval monitoring of manufacturing and assessment of process deviations. Alexeenko noted other industries are currently harmonizing modeling approaches and sharing best practices on applications.
Up next was Julia C. O’Neill, who had just left her post as Principal Consultant for Direxa Consulting to join Moderna as Distinguished Fellow, CMC Statistics Lead, to deliver, “Control Strategy Grounded in Process Modeling.”
O’Neill shared her thoughts on control strategies—how much more powerful they can be when grounded in process modeling and how those strategies can be used without compromising safety, efficacy and quality.
The goal of a control strategy, as expressed in ICH Q10 and Q11, is to assure process performance and product quality. O’Neill noted, however, that control strategy and validation and process modeling share the same goals: identify key causes for variation in results, test and define process controls, demonstrate that the controls work and update the control strategy based on monitored results. In process modeling, the data and results observed through development experiments and commercial production are used to build a robust model that allows predictive results, reducing the level of uncertainty through knowledge-building.
O’Neill illustrated how drastically drug development has changed over the past decade.
The traditional sequential development timeline that averaged 12 years in 2015 may now be compressed to fewer than 12 months, as seen with COVID-19. The accelerated timeline between discovery and distribution-to-patients often forces concurrent development, making control strategies especially important in keeping track of the various pieces.
Many activities can be done in parallel, but not all. Long-term stability studies, for one, cannot be compressed. However, O’Neill noted that modeling might be used to project or extrapolate, for instance, a full 24 months of data from only nine actual months of study, particularly for products intended to be consumed in six months. Having a plausible scientific stability model and using sound statistical modeling may successfully bridge that gap.
O’Neill summarized: The use of statistical and mechanistic process modeling to forecast future results and define linkages between input and outputs will produce a much more robust control strategy.
During the Q&A session that followed the presentations, Huang introduced Francisco Vicenty, CDRH, U.S. FDA. He said he has been working to engage more companies to leverage modeling but has not received much in submissions; most include it only for directionality. He encourages sponsors to submit all data, even if it is not perfect.
Vicenty said sharing data, good and bad, demonstrates a company is showing control, and that is a way to build trust. “Speaking for any regulator, not just FDA, some of that mindset shift [from relying solely on traditional test data] needs to come from our side. Industry should challenge agencies. Show how you used modeling and make a case for it.”
Huang asked all three panelists if they had any thoughts on how to change the mindset of industry. O’Neill jumped in quickly, “Never waste a good crisis!” On a more serious note, she continued, “Any disease that does not have a cure or treatment, where no other options are available, gives us the opportunity to try [modeling] … Success there will spread the use of modeling more broadly.”
Regarding how many companies are using modeling, Alexeenko conveyed survey results that showed most are only using modeling for lyophilizing, many use it for product development and, echoing Vicenty, only one used it for agency submission. She sees evidence that the use of modeling is more advanced in other industries, noting that trust is more necessary in the pharma industry. However, she remains optimistic it will happen soon.
Alexeenko praised the collaboration between industry and academia for using modeling to transfer knowledge to design a ventilatory system during Covid. She commented, “Lots of engineers have been hired from Purdue. I hope they bring modeling into the manufacturing industry.”
The speakers concurred that the use of modeling, which can be applied in many ways in many stages of drug development and monitoring, could help build more robust manufacturing processes. The broader use of modeling may provide the control needed to ensure safety in accelerated development.