Last month, Novartis and the Massachusetts Institute of Technology (MIT) announced the formation of a research partnership dedicated to continuous manufacturing. Novartis will give MIT $65 million over the next 10 years to establish the program; according to The Boston Globe, $40 million of this funding will come in the first five years. The Novartis-MIT Center for Continuous Manufacturing (CCM) will be staffed by seven to 10 MIT faculty members, in addition to students, postdoctoral fellows and staff scientists. Early research efforts will be conducted at MIT and Novartis will take the techniques developed at MIT and scale them up in the future. The CCM is in its earliest planning stages and is preparing for a wide range of research projects. Karl Hornung, director of Global Quality Operations at Novartis Pharma AG, said that the decisions for specific projects have not yet been made by the CCM Steering Committee. However, on a general level, as Charles Cooney, professor of chemical engineering at MIT and a member of the CCM, explained, some areas of research are known: “We’re looking at methods of chemical reaction, we’re looking at methods of formulation and we’re certainly doing work on API [active pharmaceutical ingredient] and excipient selection.”
In general, the use of process analytical technology (PAT) in the pharmaceutical industry is beginning to increase since the FDA’s publication of its first PAT Framework draft in September 2003. Professor Cooney acknowledged that the pharmaceutical industry may be slow to use the newest technologies for process analysis, but he was optimistic about the general adoption of PAT by the industry: “I think there’s still some hesitation on the industry’s part to adapt . . . newer concepts because there hasn’t been a lot of application of them. There haven’t been a lot of successes. There have not been failures, either: it’s just that there isn’t a lot of success on which to build experience. But certainly the number of examples, the number of people working in applying PAT is certainly increasing.”
For example, Novartis, which has collaborated before with MIT on individual PAT projects through the Consortium on Advanced Manufacturing for Pharmaceuticals, has been working on a Collaborative Research and Development Agreement (CRADA) with the FDA since 2005 for Quality by Design (QbD). QbD is a principle of PAT requiring that quality be built into—or designed into—a product, rather than testing the product for quality after it has been manufactured. For this project, Novartis developed a PAT system for the entire manufacturing process of one of its pharmaceutical products and is sharing its results with the FDA.
Professor Cooney was clear about the definition of PAT: “if you think about PAT as simply adding instruments to a process, you end up collecting a lot of data and then correlating the data with performance, and that’s exactly what PAT is not.” Dr. Cooney explained that the purpose of PAT for the CCM is “to explore novel methods of pharmaceutical manufacturing that are suitable for continuous flow and develop a deep enough understanding of those processes, such that we can deploy whatever relevant analytical methods are necessary to both understand and better operate the process.”
One of the most challenging elements of developing continuous manufacturing methods is that manufacturing processes for drugs may need to be drastically changed. As Dr. Hornung attested, “Continuous manufacturing does not mean that you simply connect all your dots” from an earlier, batch-based manufacturing method whose steps may not have been fully understood in the first place. James Cheney, global director of PAT for Novartis AG, gave one example of a potential analytical challenge in continuous manufacturing: “Say you’re using connected microreactors; there’s still going to be reaction chemistry going on in some of those reactors and knowing when you reach that endpoint [in the manufacturing process] will be critical to move the material on to the next phase, or the next reactor or the next section of the process.”
Although PAT may not mean simply adding instruments to the process stream, there are technologies that are likely to play an important role in the CCM’s research. Molecular spectroscopic techniques were discussed by all of the researchers as important tools for the CCM. As Professor Cooney elaborated, “[Spectroscopic techniques] have shown to be very useful because they relate specific chemical information or allow you to monitor specific chemical content during the process. . . . And that allows you to think and operate more mechanistically.” Professor Bernhardt Trout, associate professor of chemical engineering at MIT and director of the CCM, explained that this kind of information will be crucial for the creation of the mathematical models that will be used in continuous manufacturing. Other techniques that could be widely used include laser diffraction for particle size analysis and electrochemical techniques, such as pH meters. All of the interviewees acknowledged that the usefulness of other instruments would be evaluated, and that process needs would dictate the methods used in the coming research.
Regardless of the kind of analysis being performed, there are two general challenges that face researchers working on continuous manufacturing PAT. The first is a question of the capability of a given instrument to accurately reflect information about the whole of a substance from which a sample is taken. As Professor Cooney explained, “the scale of scrutiny . . . is defined by the instrument that you use, and you’re going to take the results of that method and use it to interpret what is happening in the complete system.” He added that scale of scrutiny issues can potentially yield misleading results: “It could be that the mixing within the system is generating heterogeneity at the scale of scrutiny that you’re looking at. If I were to put a window to look at everything that’s in there, obviously that would be representative. But when I only look at a small segment, then I have to be careful about how I interpret those results. . . . I think that’s certainly as important here as it is in batch processing, and possibly even more important.”
Professor Trout addressed the second general challenge in using analytical instruments for continuous manufacturing: “Clearly, issues with continuous are that we’re going to want to get more rapid sampling time—when I say sampling time, I mean response from the time we get a sample to the time we get data.” Mr. Cheney related Novartis’s success in reducing this delay from approximately 20–25 seconds to five seconds for some of the analytical techniques used in conjunction with the CRADA for PAT with the FDA, but believed that the lag time still needs to be shortened: “When you’re talking about a continuous process, where things are going to be happening faster, we need to be able to collect the data and present it to models faster.” Mr. Cheney explained that this lag is generated by both instruments and software. “What I feel we’re going to need is a clear standard for data communication between vendors and our higher-level systems, which would be our data collection and analysis systems. We don’t have the time to send data through filters . . . to align and augment the data, so we’re going to have to come up with standards. We’re going to have to work with our vendors and say, ‘Look, this is going to be our standard for data interface, and you’re going to have to meet it’ in order to keep the speed at the levels I think we’re going to want moving forward.”
As mentioned above, an important component of continuous monitoring PAT is a variety of software, which is used for data collection, multivariate data analysis, process modeling and process control. The research efforts of the CCM will also address the specific challenges that continuous manufacturing entails for PAT software. Professor Cooney described one particular data management challenge within the context of particle analysis: “[W]hen you’re doing mathematical modeling of powder processes, one of the techniques that you use is called DEM—discrete element method—modeling.” DEM assigns physical properties, such as size, density and friction coefficients, to each particle in a model. Professor Cooney explained, “[I]f you want to model a powder flow system, you’re talking about tens of thousands, or a million particles. That’s a lot of numbers, and for each particle you have multiple properties, and you want to keep track of where those particles are.” In regards to modeling software, Mr. Cheney believed that current modeling programs used for continuous manufacturing in other industries might be adapted to pharmaceutical continuous manufacturing without much difficulty. And Professor Trout pointed out that the data from continuous monitoring PAT will make for changes in the process control software: “We’re going to have more quantitative models of the various processes, so . . . that’s going to play into the control systems which are effectively part of PAT,” adding, “we intend to develop new algorithms, particularly model-based algorithms for control systems.”
All of the CCM participants IBO interviewed agreed that it is still too early to describe even hypothetical layouts of future pharmaceutical continuous manufacturing facilities, but they believed that the results from the work of the CCM will be markedly different from current production processes. As Professor Trout stated, “We have some ideas that we’re going to pursue, and some of them could be pretty far out in terms of what it’s going to look like 10 years from now as opposed to what it looks like now.”