By Evan Boswell, Senior Principal Scientist, Pfizer CentreOne Contract Manufacturing
The “recipe” for manufacturing an active pharmaceutical ingredient (API) usually evolves during the journey from flask to commercial manufacture. As you scale up, the efficiency of your API synthesis becomes increasingly important. In the end, the goal is to have a viable process that produces the same results day after day, creating a reliable, cost-effective supply chain for your drug product. For smaller pharmaceutical companies, an efficient process can also provide a “sweetener” for attracting a Big Pharma buyer.
What are practical steps you can take to build efficiency into your API synthesis?
Life Science Connect interviewed chemist Evan Boswell for his perspective. Boswell is a senior principal scientist for Pfizer CentreOne Contract Manufacturing, where he is currently responsible for assessing API manufacturing costs for incoming requests for proposal (RFPs). He works at Pfizer’s U.S. site in Kalamazoo, Michigan, supporting the CDMO’s operations. Over his career, Boswell has served as a medicinal and rapid scale-up chemist and has participated in the API development process from discovery through commercialization.
Q: What part does route selection play in synthesis efficiency?
A: Route selection is the core of an efficient API synthesis. A fully optimized bad chemistry route will almost always be worse than a partially optimized good chemistry route. For example, liquid chromatography steps in a small-molecule chemistry route are useful at lab scale to purify your compound and are often used by the medicinal and rapid scale-up chemist. However, when you move to large-scale manufacturing, liquid chromatography becomes costly. The throughput is usually slow, it uses a lot of solvents that should be recycled, and the column packing can get expensive very quickly. It’s not always possible to avoid liquid chromatography in an API plant, but it should be used only when all other avenues have been exhausted.
The order of the reactions is also important. In one of my former roles, I worked on a process that had an oxidation step and a hydrogenation step. At the time, knowing very little else about the reaction, it made sense to do the oxidation first: Those reactions are typically dirtier and tend to create more impurities, and I could eliminate the impurities downstream. So I proceeded. The oxidation went perfectly; but the hydrogenation step took 14 days! While workable at a small scale, this would have been completely cost-prohibitive at a large scale.
Through further analysis, I found it was possible to do the hydrogenation step first, which took only two days in this order, and the oxidation step was still successful. This is just one example of how your route selection—in this case, the order of the reactions—is a critical piece of the puzzle. Determining the best way to assemble your molecule is a good use of time and resources.
Q: Once you select the best route, when is the appropriate time to start optimizing your process?
A: The quest for a lower cost of goods never ends when it comes to your API, since it’s typically the most expensive raw material in your drug product. Process optimization should begin when you start to need material for clinical trials, and you should never stop trying to optimize, even after you launch your drug.
Nonetheless, the “sweet spot” for process optimization is typically during the transition from Phase 2b to Phase 3 clinical trials. That’s when your need for API is growing significantly and you usually need to scale up to meet those requirements. With scale-up come other challenges, such as making sure your chemistry works and is efficient in large-scale equipment. You also have to consider that deviations at large scale are more costly, so the robustness of your process needs to be evaluated at that point. So immediately prior to scale-up is the ideal time to consider process optimization.
Q: Do you have tips for reducing manufacturing time, which is a big cost driver?
A: First, you want to minimize the time spent in the tank. You pay for the reactor as well as the labor and overhead by the hour. The longer your reaction takes, the more money you have to spend, especially at a commercial scale. Consider your potential downstream consequences, too. For example, if a reaction takes 18 hours but the resulting cake takes seven days to dry, you could create a bottleneck in your process. An idle tank costs nearly as much to operate as a tank in use. Ideally, the time on the filter should be no longer than the time it takes to finish the next reaction.
You also want to optimize the time spent on the filter. Chemical engineers should understand the crystal morphology in solution and look for ways to improve the characteristics to ensure the resulting crystal will filter nicely. I once worked on a process that yielded a crystal with the consistency of shaving cream, which took many days to filter. Just a change in crystallizing solvent eliminated the issue.
Cake washing techniques—how you dry your cake—and your dryness endpoint can also affect filter time. Cake washing helps drive impurities out of your cake, but it can also be used to drive out a hard-to-dry solvent. There are many options in the plant for cake drying, from heated recycled nitrogen to single-pass nitrogen or wet nitrogen for hydrates. It’s crucial to understand which is needed at what time.
Finally, your endpoint is of critical importance. If your next reaction can handle five percent of the solvent used in the previous reaction, why dry to 0.1 percent? The caveat to this is with water. It is not a good idea to keep a water wet cake in the warehouse for an extended period of time, as it may be a good breeding ground for mold or bacteria.
The best way to reduce time on filter, though, is to avoid using the filter in the first place.
Q: How can you avoid using a filter?
A: During early process development, you may have isolated every intermediate along the path to your API to understand which impurities were created at each step and how they purged. However, at large scale, crystallizations are expensive. I always look for opportunities to telescope steps. When you try to telescope a process, consider which compounds in the reaction sequence are effective crystallizers, so you know how to exclude as many impurities as possible during the crystallization. These experiments are often very time-consuming and can be tedious, but they generally yield significant cost reductions through reduced crystallizations.
It’s also a good idea to run a series of experiments to understand the limits of how concentrated your reaction can be. The more concentrated your reaction runs, the more throughput you have in your tank, which helps save money. Other ways to concentrate your reaction might be to consider a slurry-to-slurry reaction where only a portion of your compound is in solution at any given time. Driving these types of reactions to completion presents challenges, but little you cannot overcome. And in the end, doing so can yield great savings.
Q: How does equipment factor into process optimization?
A: When you move to equipment at a large scale, it is often difficult to maintain the conditions created at a laboratory scale. For example, because of their high reaction rate, certain reaction types are almost always done at very cold temperatures in the lab using acetone and dry ice. This also helps reduce cost and adds convenience. However, a reaction requiring dry ice and acetone in the lab translates into using a cryogenic-capable work center in a commercial plant. While available at a large scale, this type of equipment requires additional engineering and tends to be highly utilized so it’s usually less available. You may save yourself money and increase flexibility by trying to run your reaction at -15° to -30° Celsius instead of in cryogenic conditions. Surprisingly, I’ve found cryogenic reactions often can be converted to just cold or very cold reactions.
You might also consider using glass-lined versus stainless steel tanks. Glass-lined tanks are great because they mimic your glassware in the lab and are easy to clean. Yet, they are somewhat less flexible for rapid changes in temperatures and very cold reactions, such as the cryogenic conditions I mentioned previously. But they can chip, which could contaminate your compound. Sometimes stainless steel is clearly the best option, such as when operating at high pH. When in doubt, consider coupon-testing your reaction in the lab: Put a piece of stainless steel of known weight into your glassware during the reaction and weigh it again after the compound is crystallized. If the coupon weighs less than when you started, you know the reaction conditions are too harsh for stainless steel.
Q: What is an ideal chemical process?
A: The ideal chemical process would be where you can weigh your inputs of a known quality—reagents, starting material, solvents, etc.—and put them in a tank under a specific set of criteria for temperature, pressure, and stirring speed. After a set amount of time, you filter your crystallized API, and it is the exact quality and yield you expected. At this point, all critical sources of variability have been identified and are managed by the process.
Q: What about safety?
A: High-energy reactions at small scale are dangerous nuisances. At plant scale, high-energy reactions are dangerous and deadly. It’s the difference between a firecracker and a stick of dynamite. High-energy reactions can be done at a large scale, but they require highly trained operators along with proper engineering controls, and such additional safety requirements will undoubtedly cost more. In some cases, there are safer reactions that can be substituted for a dangerous high-energy reaction.
For example, in the lab, medicinal chemists will sometimes use an azido group because it can be easily inserted in the compound and then converted to the desired amine with catalytic hydrogenation. However azides are high-energy explosives, which are problematic to handle at larger scale. For large-scale synthesis, instead of using an azido group, a dibenzylamine can be used as a masked amine, which is then unmasked with catalytic hydrogenation. Other non-explosive masked amine options are amides and nitriles.
Q: When do you know you’re done with process optimization?
A: When you’re making limited quantities of API for clinical trials, you optimize your process as much as you can before you need to stop and make more. At this stage, you optimize the low-hanging fruit that gets you the most bang for your buck. For a commercial process that’s been ongoing for a few years and run more than a dozen times, opportunities for process optimization often show themselves. If you couple that with regular data collection, you can identify more tweaks. Even small changes can result in big efficiencies when you’re using 8,000- or 16,000-liter tanks. So, in effect, process optimization should never end.