By Rob Wright, Chief Editor, Life Science Leader
Follow Me On Twitter @RfwrightLSL
When Kevin Nepveux joined Pfizer, his first job out of college was working as a production engineer manufacturing penicillin in one of the company’s original plants in Groton, CT. Today, Nepveux is the VP of global technology services within Pfizer Global Supply. “Actually making the product and seeing it go into drums every day was very fulfilling,” he states. Being close to where the action is may have been rewarding as a new employee, but as a senior leader he has found it to be even more important to understand how to best improve Pfizer’s manufacturing processes.
Nepveux sat down with me to share some of his insights about continuous manufacturing (CM) and how Pfizer is deploying continuous and batch processes as a hybrid approach toward improving manufacturing efficiencies.
Life Science Leader (LSL): Describe how Pfizer went about deploying CM technologies.
Kevin Nepveux (KN): The Global Technology Services (GTS) group initially started on these projects back in the late 1990s. The pharmaceutical industry environment was very different from today. The cost pressures were building, but weren’t as severe, and our model was still that of the blockbuster. As a result, it was a more receptive time for developing and deploying these technologies than where we are today.
When we started to look at CM technologies, we first looked outside of pharma. A lot of the continuous technology had been developed and deployed in other industries that had experienced cost pressures long before pharma. We weren’t necessarily inventing anything new, but applying technologies from other industries to pharmaceuticals, recognizing, of course, that we typically have much tighter specifications and control ranges. The food processing and specialty chemical processing industries were good sources of ideas, as they use similar unit operations. The significant challenge was in adapting some of those technologies to pharma, as these industries have a different regulatory and cost environment and operate in much higher volumes.
We would pick some of the unit operations that were the most broadly used throughout Pfizer that we felt most comfortable making continuous via adapting similar technologies from other industries. These were typically the common unit operations, such as roller compaction within a dry granulation unit operation. The philosophy was, once you get that working for one product, it should be readily applicable to others, though it would usually require some additional customization or development depending upon the product. We would then pick a specific product we thought served as a good example of that unit operation. For the most part, these were usually high-volume, capital-, and labor-intensive products, with a fixed set of unit operations performed in a batch mode — the low-hanging fruit, if you will. We would develop the continuous option at a small laboratory or pilot plant, along with the site that was manufacturing the product to verify that it worked and to uncover the various challenges that would have to be overcome from either an engineering or a technology perspective prior to scaling up. In three or four cases, we built CM commercial-scale setups or production lines in those facilities. Two products with different unit operation processes were picked as part of our initial deployments — Lipitor (a wet granulation process) and Geodon (a dry granulation process). We also looked at some of the active pharmaceutical ingredients (API) processes for Lipitor, Lyrica, and Celebrex (which are all blockbuster drugs).
LSL: What Were Some Of The Lessons Learned From These Deployments?
KN: For starters, the business cases changed over time. It takes some time to adapt and develop these applications. You start in a pilot environment to develop data that convinces you it will work. Then you develop data to convince regulators so they will approve it as an alternate process. These activities take several years to work through. In our case, some of the business cases changed during this time frame. Though the deployments were technically very successful and effective when installed for commercial manufacturing, some of the manufacturing sites are no longer in our network, and the products have gone generic. When we look at the costs to relocate that technology to another site, it really didn’t have a good ROI. The cost pressures from generic competition changed the landscape. Today, when we approach a deployment, we do a much more rigorous job of identifying and making sure, up front, it has an enduring business case. We also look for projects that can be rapidly deployed on a number of products. It is difficult to have a project tied to one individual product, because as it moves through its life cycle, and generic competition begins, volumes drop and the cost environment changes. We look for opportunities with broader applicability that can be applied fairly quickly.
Another lesson learned involved analysis. For example, when we did Lipitor and Geodon, we did business cases comparing what we believed the technology would look like in five years, relative to how it was at the time, in terms of production costs, demand, margins, etc. In conducting our analysis, as we were the only manufacturer of the drugs, we used ourselves as the comparator with which to benchmark. Now, when a drug loses exclusivity, you are no longer competing with yourself, but with generic manufacturers, which often have much lower labor and capital costs. As a result, shortly after we had developed and implemented the CM process for these drugs, our analysis was no longer valid because the benchmark had shifted from internal to external.
Pfizer’s value proposition hinges on quality and supply assurance. We excel in manufacturing the highest quality products and assuring their supply globally. A lot of the customers have found that not always to be the case with generics. The ability to produce variable quantities of medicines, as well as the ability to ramp up and produce more medicines very quickly, has enabled us to compete effectively in the generic area. While it is challenging to provide both quality and supply assurance while also having the lowest cost, we have found that customers are willing to pay a bit more for the quality and supply assurance we offer.
LSL: What Particular Type Of Data Are You Evaluating In This Decision-Making Process?
KN: You are generally looking for a comparison to the baseline for the product or unit operation. We usually have quite a bit of data, because we have been running these unit operations for a long time. The most compelling data is situations where continuous processing can do something significantly better or something that a batch unit operation can’t do. One example is the synthesis part of small molecule API. You can do some exothermic reactions or hazardous reactions in a continuous-mode plug-flow reactor that you could not do in a batch reactor for either safety reasons or that the chemistry may be self-limiting. A second way to look at it is if you are going to be incrementally better from either a labor or productivity perspective. Those are still strong business cases, but you have to look beyond just the current situation with that product. You probably need to look at least 5 to 10 years down the road to determine whether the benefit of implementation will still exist or accumulate. In situations where you need additional capacity for a product or a suite of products, and you need to build more capacity, it is important to consider developing continuous processes, as I believe these applications are going to be significantly less capital-intensive, have a smaller footprint, and be more productive than building a large, traditional batch facility. At present, this is not generally the case, since we have excess capacity for most of the traditional unit operations. Where we are finding opportunities is in some of the smaller emerging markets where, either for tactical or political reasons, you are looking to establish a local manufacturing presence.
LSL: What Are Some Of The Advantages Of Continuous Versus Batch Manufacturing Processes?
KN: In CM you can operate at a steady state, which is optimal for that process or unit operation. Batch, by definition, is a transition from a starting point, through a process, to an ending. There may be a point along the way where it is optimal: a “sweet spot” for that specific process (i.e. a chemical reaction, formulation process, a cell-growth profile). But because it is in a batch, the sweet spot isn’t maintained. With CM you have the possibility of being able to maintain the process at that sweet spot for a long period of time, which can be more effective than batching. However, you need to be concerned with how long it takes you to get to that sweet spot, since there can be quite a lot of auxiliary equipment and processes necessary to maintain the environment around one processing step. That’s one of the challenges with a CM process. The other challenge that we have seen concerns linking multiple unit operations together in one CM process. This is the approach the Novartis MIT project has taken with great results. Using creative engineering and sophisticated control systems technology, they were able to match each process step, each unit operation, to the rest of the process to avoid bottlenecks or hold points where inventory might build up. Building in buffers is a technical challenge that can be overcome, but it takes a lot more work on the unit operations and is a potential downside to CM.
One of the big upsides to CM is that you don’t have a fixed batch size, so you can make as little or as much as you need. As the market dynamics change, and we try to get into a supply model that is more responsive to the customer demands (especially as we start to look at some of the emerging markets with smaller and variable demands that need to be satisfied very quickly in order to be effective from a business standpoint), CM gives you a lot more flexibility than batch. CM lead times (from purchase of raw materials to delivery of finished products) are typically a fraction of batch lead times — this can substantially reduce inventory carrying costs. Another basic advantage of CM is that you can get a higher throughput on a smaller footprint for less of a capital investment as compared to a batch process.
The other area where CM is really effective is when your process cannot tolerate much variability. In these cases, you can be more consistent and more robust using CM. This is also an advantage when introducing new products, since there is no “scale-up” — you can manufacture development, clinical, and commercial product on the same equipment by running longer.
LSL: What About A Hybrid Manufacturing Process That Utilizes The Best Of Both Batch And Continuous As A Means Of Improving Efficiencies And Managing Bottlenecks?
KN: That is what we’ve evolved to. I don’t know that this was an intentional, planned evolution, but it’s where we are now. We have evaluated CM for most unit operations that we execute in both large and small molecule areas of pharmaceutical manufacturing. There are some unit operations that are really effective and carry a very strong business case. As the business case justifies it, we are deploying CM for those products or for multiple products for that unit operation within our manufacturing network. As of yet, we don’t have any situations that are true end-toend continuous whereby you start with a raw material on one end and finish with a final drug product on the other with zero interruptions. We term our manufacturing approach as being a hybrid, which utilizes some elements of batch and continuous. We may have one unit operation or one step in a four- or five-step process that is continuous, or maybe a couple that are back-to-back, but there will also be some batch processes or batch steps within that process. The hybrid approach has proven effective for us, especially when you have one unit operation that might be the real bottleneck in a process. For example, tablet coating is a bottleneck, and can be a very long process in a batch mode. It is often the rate-limiting step for a drug product manufacturing operation. We have looked at and deployed platforms of continuous coating in those situations, which can de-bottleneck a process and allow your granulation and tableting, which is in front, and your packaging, which is in back, to be better utilized overall. Another example in the small molecule API area is crystallization, which can be a variable process. This variability can result in changes to physical properties of your API, which in turn can cause you problems in filtration, drying, and formulation. There are some cases where we’ve looked at continuous crystallization as a way to get more consistent physical properties, which improves other components of the manufacturing process.
LSL: What Are Some Of The Challenges To Implementing CM Given The Absence Of FDA Regulatory Guidance?
KN: There are two issues. One is how big is a lot size, and how do you define that with CM. I think there is some good guidance that has been developed here. The other issue involves the use of online analytics to monitor and control your process. These are not the traditional, single-loop feedback control systems (i.e. controlling temperature or pressure on a reactor). They require much higher-level control strategies, and understanding those is a challenge. These provide more data than that of a batch environment (i.e. doing military-standard sampling and testing for release). When you gather more data, you are going to see more variability than when working with a smaller sample size.
Now, the variability was always there; you just weren’t picking it up in the sampling technique you were using. That has been somewhat of a challenge — to talk about how you manage this much larger data set, and what is acceptable in terms of adequately controlling your process. This is a learning process for both the manufacturers and the regulators. We have found the FDA to be relatively open and cooperative to the manufacturing approaches we want to take at Pfizer. Some of the new approaches should improve overall robustness of manufacturing processes over time. In many cases, multiproduct CM platforms will make it easier to respond to surges in demand, allowing companies to better manage drug shortage situations.
I think one of the best ways to go about implementing CM processes is to develop the analytics in line with the application. Once we get to the point where we are satisfied with a control strategy, it’s really good to be open and suggest it to the regulators as part of the data exchange around the technology. Engage the regulatory agencies as early as you are comfortable, sharing with them where you are going. You need to suggest and propose the appropriate control strategy around that technology to the regulators, because you are the experts.
LSL: What Are Some Of The Important Considerations For Creating Control Strategies For Hybrid Manufacturing?
KN: Quite often, a lot of the data you’ve been capturing all along for other reasons can really be helpful in monitoring the health of your total manufacturing process. That’s where you start to get into some of the more subtle and less obvious relationships that can give you information on your process. This is a big challenge, because you have so much information now that you’ve got to be careful in terms of how you use it versus what is actually needed specifically for process control.
Take tablet coating, for example. Occasionally, we would have coatings that were uneven, which might result in having to reprocess or reject a batch that wasn’t cosmetically acceptable, even though the active and the tablet were fine. For years we have been striving to get better control of the inherent challenges with tablet coating. To that end, we placed some better technology to monitor temperature in the tablet coater. We found that we were collecting a lot of other data, like motor speed on the drum of the coater, flow rate of the coating solution going in, air flow rate and temperature, as well as the routine types of monitoring metrics. We then conducted a fairly sophisticated analysis that provided a “fingerprint” of the tablet-coating process. This allowed us to watch for trends and anticipate flaws before they showed up in the physical product.
LSL: What Are Some Of The Metrics Employed To Determine Hybrid Manufacturing As Being Effective?
KN: You can evaluate those traditional measures — quality, yield, productivity, labor costs, etc. — as a straight comparison. Certainly, we do see improvements in those unit operations that can benefit significantly from CM. From an expense perspective or in an ongoing cost per unit, you’ll typically see a lower cost for continuous than you would for batch. But if you’ve got a batch facility that is largely depreciated, and you’ve got to build a new CM facility, you need to compare capital and operating costs in your analysis. The other thing we noticed is that the processing costs are typically incremental. You need high volume so that you’re multiplying by a big number of units to realize value. This is something to pay close attention to as a product approaches the end of its life cycle. Generic incursion can drop volumes, substantially reducing the total benefit.
LSL: Any Pleasant Surprises From Implementation Of CM Processes?
KN: Yes, we have seen some examples. I can’t go into detail on the products, but they were typically in situations where you could actually do something a different way with continuous than you could with batch. I’ll use an example of a chemical process step where you have an exothermic reaction that you couldn’t do in batch because it was unsafe to have that much material in a reactor generating that much heat. Because you couldn’t control it safely, you had to use different chemistry. In a flow reactor, because you’re dealing with microscopic amounts of the material, even though the heat is generated, you can remove it much more effectively. Because you are using a small amount of material, there’s not as much of a risk. We’ve had a couple of cases that have produced substantial ROI compared to a batch process.