Moving Drug Manufacturing from Good to Great
Pankaj Mohan, U.S.-based management leader for global process engineering at Eli Lilly, shares strategies that can help more drug manufacturers move from good to great.
By Agnes M. Shanley, Editor in Chief
The approach taken to optimize technology at any given facility or operation will vary from case to case, but success generally requires a marriage between a strong business case and a strong technical project layout.
A.S.: We’ve seen that pharma manufacturing remains hampered by “siloed” information and lack of integration between different databases and IT platforms. Although some companies are investing the time and effort in building bridges between these platforms, more are not. Will this problem only be resolved by the availability of more commercial “off-the-shelf” solutions?
P.M.: It’s true that pharma companies are collecting a lot of data, but on an “FYI” basis. The industry needs to convert data into knowledge, and I have yet to find one comprehensive commercial solution that will make it easier to move from data on the one end to knowledge at the other. There are many intermediate steps. First is data filtering, to ensure that you are not looking at “outliers,” and so that you can distinguish between noise and a true process signal as you analyze the data. This is very important.
Applying process analytical technologies (PAT) is the next step. One must ask: “What kind of technology will be used to collect the data?" Then you must determine how to analyze the data, such as how often to average, etc.
The drug industry collects a lot of data and spends a lot of money to get information, but it still doesn’t have a true knowledge engine that would allow it to transform data into knowledge.
I see various commercial process control packages that can handle some of the tasks required, but using them requires pulling various packages together, from data analysis to knowledge extraction, storage and artificial intelligence. So far I haven’t seen one company provide a comprehensive solution that integrates various layers between data and knowledge. The industry is not getting maximum value out of the data it collects, for process optimization, or proactive troubleshooting.
A.S.: So, companies must be prepared to do the work themselves? What suggestions could you offer?
P.M.: Well, the first thing to realize is that, as information moves to knowledge, using it becomes multifunctional activity. Instead of each different department looking at information differently, a very focused, multifunctional “task force” should look at data collectively and on a regular basis, with the specific intent of generating knowledge.
Secondly, it’s important to avoid data overload. Companies should not record data that won’t be used. If you have a lot of data coming through in a GMP setting, FDA might start looking at that and ask, “Why aren’t you putting these data to use?” If you don’t intend to use data for extraction or real-time batch info analysis, don’t just make every data point available.
Finally, integrate technology and methodology that is available. Multivariate statistical process control can serve as a knowledge engine, and as a tool to help you understand which factors, among the hundreds of existing variables at play in your operations, will have the greatest impact on profit margins.
A.S.: What skill sets will be required to improve pharma manufacturing operations? The industry still seems to be comprised mainly of chemists who may not have the knowledge of control or modeling required to do all this.
P.M.: Well, to be fair, pharma does employ some ChemEs. I am one myself. But, typically, what can add horsepower to any pharmaceutical manufacturing site or development unit is the involvement of statisticians.
Some companies are now developing strategies in which scientists and engineers become “secondary loop” specialists. Primary loop people ensure that the process works as it should. More-senior people at this second level take a step back, and look at the data that are coming out of the plant. They provide systemic or breakthrough improvements, or suggest changes that need to be made at the laboratory level. These people need specific knowledge, training and experience, and must be connected, internally and externally, to a network of subject matter experts.
A.S.: What is needed to ensure success with PAT?
P.M.: The way I look at it, PAT must be integrated with an overall control strategy. Suppose you have a process for which you’ve identified critical process parameters. Ideally PAT should be applied before the process hits manufacturing, so that, once the process is “locked down,” PAT becomes an integral part of the control strategy. Integrate it with various control philosophies to ensure your process is capable and in control.
A.S.: Will PAT eliminate the need for old-fashioned quality testing?
P.M.: FDA is pushing the use of PAT for “in-process” control, so that critical process parameters and quality indicators are measured and controlled during the manufacturing process. This approach would reduce lab and end quality testing, and give more confidence in parametric batch release. The question for many companies today is, “Is PAT an add-on to what we already do, or will it make things simpler and easier?”Some people today are using it as an add-on, which is a good intermediate step, before eliminating any end-testing procedures.