At the Bioprocess International Conference in Raleigh this week, keynote speaker Paul McKenzie showed that he is taking some of the thinking that he had brought to BMS (read on for a 2007 cover story interview touching on this work) to Centocor, only now he is moving the effort from the lab to the plant level, rather than the other way around.
Mr. McKenzie is a chemical engineer who started his career at DuPont. Basically, he says, operations need to be reduced to their simplest form: the recipe, invoking S-88, which he has said needs to become the “DNA of manufacturing” in the past. Here is a quick recap of his presentation (we'll have more on this soon).
McKenzie noted that PAT concepts are anything but new, but that they are becoming more important to pharma. “Without change, we will become extinct,” he said. “We can only merge so much,” noting consolidations, slower pipeline, government actions and competitive pressures affecting the industry.
He envisions a life cycle management view of data over the entire precommercial and commercial time frame. “We need a common language from lab to plant . . . using simple approaches that industries, and sectors such as fine chemicals and additives, have been using since the 1980s,” he said.
As to Quality by Design, a term that he intentionally avoided during his presentation, “We need to be able to connect the clinical to be successful with QbD,” he said. “We need a base infrastructure to enable this, and we need to execute differently in order to succeed.”
Centocor recently merged large and small molecule development, an unusual approach, has expanded in India and has had two recent drugs approved by FDA.
Connecting CMC to Clinical
The essential part of QbD will be tying clinical to CMC, McKenzie said. “Regulators now expect to see that connection.” But a whole new approach to execution will be needed, and new technology will need to be applied to effectively capture, analyze and use process knowledge, he said.
As an example, he highlighted tech transfer, and mentioned analytical method transfer. Currently, this can take nine months for many companies. “Can we afford this, as an industry?” he asked.
One must ask whether the initial goal is to speed product to proof of concept or to commercial stages. Centocor, he said, is doing what many smaller bio companies doing by expending significant effort on early and full development. Now, these efforts must be carried forward to CMC and manufacturing.
The company is using the concept of the target product profile, a simple concept that is extremely difficult to manage, he said. This profile ensures cross-functional alignment and commitment on formulation, dosage form and drug delivery objectives and technical specs set forth to meet goals throughout drug development.
His team is also emphasizing Design for Manufacturing. In most companies, he says, there is a real disconnect between process groups in R&D, process groups in manufacturing, analytical groups in R&D and analytical groups in manufacturing.
“A lot of scientists at our company had never seen a P&ID of a plant . . . it’s hard to develop a process unless you know how it will work,” he said.
“We bring all scientists to the plant to ensure that they know how the process will be brought to the plant . . . and we get them thinking about issues very early in development."
Meanwhile, there is a “lab to patient” initiative focused on efficiency, consistency and quality and driving better ownership of process and science from the start.
Data Management Challenge
One of the greatest challenges is the global connectivity of data, processes and communication, he said. McKenzie sees S-88, still unknown to many in pharma, as the key. Using this framework gets everyone to define the manufacturing requirements for a product and to use similar, simple terms in common technical documents to express all lab activities, process plant, pilot, commercial plant and analytical testing data.
Recipe needn’t mean automation, he said. Simply break down every activity into unique blocks you can define once and manage throughout process.
“When grandma gave you recipe, did she include the serial number of the oven?“ he asked. “Why do we do that with batch records?” You have raw materials, quantities, process parameters, recipe procedures, equipment requirements. . . "
This approach may be known to engineers, but how do you get R&D people to embrace it? he asked. It’s important that everyone collect data in the same way, he said, so that tech transfer becomes ownership of that recipe.
He singled out Pfizer’s use of the IT tool Catalyst as an example. It allows you to express the process as recipe so that, over time, for a typical bio process you have a general recipe, deployed in series of different sites. Then you can compare recipes and know electronically what differences were between the way they were executed. This allows you to take a true QbD approach, he said.
“I don’t want people innovating based on what they call something in a MS word doc, but based on parameters . . . common global language to describe process and product,” he said. One recipe can be used across the world and at different facilities, becoming an easy way to link to regulatory filings. It also allows one to look at the continuum of visualization to the “golden batch,” McKenzie said.
How do we get to the golden batch today? “Tech transfer today is often a very complicated series of MS word docs . . . reports handed to operators on floor . . . and the likelihood that they actually saw them is zero . . . ”
Execution is a recipe, he said, and you don’t need to create bureaucracy and documents.
He mentioned the need to connect process specifics to clinical issues. “We need to put biomarkers in clinical trials that feed back to the process to make QbD happen.”
Process and analytical groups need to be brought together, and the balance between process and release testing needs to be reconsidered, he said. At his company, process scientists run analytical methods. Analysts develop the methods but process scientists run them.
Reducing analytical time scales will be essential, he noted.
In addition, McKenzie said, there’s a need for easier decisionmaking tools. “We’ve convinced ourselves that we need the fifth decimal point to make decisions, so we end up with large filings because we’re catching up after the fact on the analytical side.”
In one project, Centocor is using Raman spectroscopy to characterize cell culture medium. “Sometimes we get lots that don’t produce. We look at QA test results, and they don’t always tell why, so we would take every measurement of the media we could using FTIR, NIR and other methods.”
“When we started doing this, “ McKenzie went on,”we used unconventional tools but we started to see trends, then we did a multivariate analysis of medium power samples based on Raman spectroscopy, and were able to isolate lots that would not work in cell culture…
Thus, Raman became a cheap and easy screening tool for determining which media work. Now, he said, the company can make decisions and determine differences between the two media los. “For PAT, we want to get a lot quicker at getting this info, to make faster decisions rather than getting a PdD thesis,” he said.
Of all the challenges, he noted, people are the greatest one, but they will always work better when they have better access to data, shared best practices and deeper and broader organizational knowledge. VSM has been useful, he said, in eliminating non value added processes.
Progress is slow. He mentioned one situation where it takes 500 signatures to get from lab bench to the first pilot. “The industry has created a bureaucracy of GMP and compliance,” he said.
“Compliance is absolute necessity. But what are you being compliant too?“ he asked. “We’ve written these procedures and basically handcuffed ourselves.”
Ultimately, he said, pharma needs to streamline processes and introduce recipe thinking.