Data Analytics in Pharmaceutical Manufacturing
Shedding light on how robust data enables confident decision-making and drives operational efficiency was one goal of the “Emerging Trends & Innovation” track of the 2020 PDA/FDA Joint Regulatory Conference on Sept. 15.
Sharyl D. Hartsock, Senior Director Quality, Global Quality Systems, Eli Lilly and Company, moderated a session featuring expert presentations on using data analytics in pharmaceutical manufacturing.
Tara Gooen Bizjak, Director, Manufacturing Quality Guidance and Policy Staff, CDER, U.S. FDA, presented, “Trending Data to Drive Quality Improvements,” a topic she said, “needs a lot more discussion.”
Bizjak compared an Escher-esc graphic that appeared as three columns when seen from one direction and as four when viewed from another to data trending and looking at data from different sources. “It’s how it’s defined and how you look at it. It’s very challenging to companies to establish clear definitions, counting strategies [that are] consistent from site to site within a network.”
Bizjak reviewed various U.S. drug regulations relevant to trending at different stages in manufacturing and in the lab, touching on the design of process controls, trending variability, and investigations and improvements. She also discussed how data trending underlies important ICH quality guidelines. For example, ICH Q10, Pharmaceutical Quality System discusses elements that contribute to building an effective plan for continual improvement, while ICH Q9, Quality Risk Management, provides a powerful framework for looking at process.
Turning to inspection findings as case studies, Bizjak explained where trending could make a difference when conducting investigations, responding to complaints and identifying adverse trends. Design your periodic product review strategy, she suggested, less as a checklist exercise for regulators and more of a value-added tool.
In summary, Bizjak recapped the importance of measurement as the foundation of improving product and process quality, after first agreeing on how to measure and track the data. And she noted that there are many opportunities to use data more effectively to improve manufacturing while also demonstrating the effectiveness of a pharmaceutical quality system.
Bizjak encouraged participants to “look at your data, look at the different inputs you have and potential inputs into your tracking and trending program and look for ways to proactively improve your process and pharmacology.”
In his presentation, “Holistic Approach to Implementing the Digital Plant of the Future,” Wilfred J. Mascarenhas, Advisor, Data and Analytics Strategy for Manufacturing and Quality, Information and Digital Solutions, Eli Lilly & Co., described how his company has become a digital plant of the future and illustrated those steps in a relatable example.
The holistic approach, Mascarenhas said, is critical to digital plant transformation. He stressed the importance of following the strategic approach—integrating technology, processes and people—and balancing holistic architecture with use-case-driven models.
Lilly approached technology in two ways, both built on cloud architecture, which drove the decisions on how it could best be designed and implemented. From there, they developed a template that could be used for each of their manufacturing sites. “From a manufacturing standpoint,” Mascarenhas said, “you really cannot think of the cloud in isolation.” The cloud is not just a place to store data. It also integrates the devices on the shop floor, creating a system of edge devices and Internet of Things (IoT) sensors with edge-computing.
In developing a digital plant, the process strategy merges digital plant governance with innovation and architecture collaboration. Ideas are evaluated through the plant governance to determine their value and probability of success. If the proof of concept (POC) and pilot are successful, then the architecture and program are built and deployed to other sites. The plant governance process blends closely with architecture and long-term strategies. That is where POCs can help define the technical components that will be needed to make sure what is being implemented will actually work.
Mascarenhas emphasized the importance of the people strategy, saying “it is the people who make everything happen.” Lilly created a framework of “role families,” determined whether each role family was related to a business or an IT role and then mapped out various roles within each role family. For example, the data engineering family might include roles of database administrator, data wrangler or data analyst. The company also created a learning plan for each role family, allowing staff to “upscale” their skills from beginner to an advanced level through online courses, internships and on-the-job training. He used the data analytics group as an example, but the company has used the model across its sites.
In closing, Mascarenhas reiterated that the holistic approach to implementation is critical for long-term growth with a digital plant. To achieve that, technology should not be added just to add technology, but use it as part of a strategy that incorporates people and process.
Bizjak and Mascarenhas provided a wealth of valuable information, looking from both regulatory and industry viewpoints, on how pharma can use data analytics more effectively to predict trends, improve processes and overcome the challenges impeding innovation.