During my last week in Chicago I attended the iBio event on Understanding Quality by Design - a panel discussion between Venkat Venkatasubramanian (Purdue), Dan Heighway (Eli Lilly), and Sam Venugopal (Conformia).
The goal of the FDA's Quality by Design initiative is to create a "maximally efficient, agile, flexible manufacturing sector that reliably produces high-quality products without extensive regulatory oversight." QbD is part of the FDA's Process Analytical Technologies (PAT) effort. Another way to describe this is that if a company can demonstrate its understanding of the science behind its processes, the FDA will reduce the amount of regulatory oversight on the company. There are some huge implications here, both for the company and for the FDA.
For the company, they must demonstrate a history of being in control of their processes. Not just one process for one new drug, but all the processes that support the drug. Depending on how you read things, this also implies that the company understand any similar processes. What is "understand?" The panelists all seemed to agree that this is model-based understanding: statistical models and first-principles models. This is as opposed to the current methods which are experiment-based and do a lot of "quality by testing." One of the elements of discussion was referencing "prior art" (other scientific research) in submissions to the FDA. In 2007, there were only 27 references to prior knowledge in regulatory filings - this has to change significantly if companies are going to get anywhere near the capability to demonstrate that they are relying on demonstrated knowledge and experience.
Along with this statistic was a discussion of the cultural implications of doing this right. The low references to prior art is a demonstration of the way things have always been done, and of the not-invented-here syndrome common to knowledge management efforts. There is also a simple time element: it takes time to build internal and external access to that prior art. And it will take a long time before the pharma company will see the benefit of these efforts. That said, the expected benefits in process development (less rework, developing a platform of knowledge), manufacturing (optimal processes, lower costs) and ultimately for the patient (higher quality) are quite high.
There were some discussions of the information technology implications of QbD as well. Accessing prior art and making the connection to your current research is going to be supported through efforts around semantic analysis and automated text analysis. There are challenges around communicated what the organization (the pharma company) has learned about their process from other sources, when the traditional mechanism is to show experiment after experiment that supports their claims. I also heard suggestions of expert systems and companies that might be savvy enough to sell their intellectual property around this kind of design knowledge.
And for the FDA? There are big changes in how it might review submissions and how it will need to process information about prior art - either previously submitted or other widely-accepted scientific research. And the question that Venugopal repeated: will the FDA accept prior art arguments and reduce the regulatory burden, as the Quality by Design initiative suggests.