Astrazeneca's Biologics Veteran: Applying A Risk-Based Approach To Plan For Capacity
Imagine you are seated at a table preparing to discuss the manufacturing of biologics. The person across from you possesses nearly 40 years’ worth of wisdom on the topic. You, on the other hand, have zero experience in this field. Kind of like a rookie stepping into the batter’s box against Nolan Ryan and understanding that if a 95 mph baseball is coming at his head he has less than .4 seconds to get out of the way.
When I recently sat down with Andrew Skibo, regional VP of supply biologics, global engineering, and real estate at AstraZeneca (AZ), I was the guy with no experience. Sure, I have 20+ years of pharmaceutical industry experience, but after a few minutes of conversation, I learned that this exchange could only be described as one between a veteran and a rookie — and I was the latter.
Skibo has an impressive list of industry accomplishments, including overseeing large-scale capital projects that garnered two International Society for Pharmaceutical Engineering (ISPE) facility of the year awards (FOYA) and two Leadership in Energy & Environmental Design (LEED) Gold awards. But what you can’t conclude from someone’s CV is the skill with which they are able to communicate their wisdom. For instance, I found Skibo to be a patient and skilled communicator, putting me at ease by stating, “If I answer any of these questions in too much detail, stop and fine-tune me as to the level you need.” What follows are his insights on applying a risk-based approach to modeling and planning for biologics manufacturing capacity. Of course, he knows a thing or two about this topic, considering nearly half of AstraZeneca’s 2014 development pipeline falls into the large molecule (biologic) category.
PLANNING FOR BIOLOGICS CAPACITY IS NO FIELD OF DREAMS
On the surface, Andrew Skibo’s decision to join AstraZeneca (December 2007) may seem somewhat odd, especially when you consider his expertise is in the manufacture of biologics. He had spent the previous four years restructuring and rebuilding the corporate engineering function on a global level for Amgen — one of the world’s oldest and biggest biotechs. AstraZeneca, on the other hand, with a small molecule blockbuster bullpen which included the likes of Crestor, Nexium, and Seroquel, was most certainly still perceived as Big Pharma. But Skibo arrived in the role of VP of global engineering at MedImmune, a biotech AstraZeneca had acquired seven months earlier to bolster its biologics product portfolio and position the company for long term growth. (He didn’t achieve his current title until a few years later.)
But executing AstraZeneca’s biologics strategy required more than the acquisition of a pipeline. Because, unlike the Field of Dreams film where a baseball field is built on the prophecy, “If you build it, they will come,” our industry requires more than a premonition, as the FDA’s mantra is not, “If you build it, we will approve,” and the stakes for miscalculation are significant. “To plan for the capacity you need,” says Skibo, “you’re looking at very large investments which can be on the scale of hundreds of millions of dollars.” For example, Novartis eclipsed the $1 billion mark with the completion of its Holly Springs, NC vaccine manufacturing facility. MedImmune invested more than 2.3 million man-hours in the building of its large-scale mammalian cell culturebased production facility in Frederick, MD (ISPE FOYA, 2011). In the cost-conscious pharmaceutical industry, these kinds of investments of time and money are rare. And when so many dollars are spent on one project that means there are fewer for others. “You don’t like to make these investments any larger than you have to,” Skibo states. “But if this capacity is the only way you can provide enough material to launch a new product, you must make the investment.”
For the folks at AstraZeneca, the challenge was developing detailed models that took the guesswork out of what kind, how much, and when that capacity was needed. According to Skibo, a good risk-based biologics manufacturing capacity planning model should take into account not only how big an investment is needed, but how late you can possibly postpone the investment to minimize risk given the distinct possibilities of a drug failing to be approved or underestimating market need. Ironically, Skibo admits that the model’s planning process, which began in 2009, came at a time when the company had a great deal of excess capacity, so much so, that in 2012 the company announced the development of a trusted partner network. The 15-year agreement with Merck allowed the two companies to use each other’s biologic facilities in production areas where there is a capacity shortfall. “We are selling the excess capacity we have and all the while modeling for when we will have to add more, because the timelines in our industry demand it.”
Regional VP Of Supply Biologics, Global Engineering, And Real Estate At AstraZeneca (AZ)
SOMETIMES IT PAYS TO DISCARD CONVENTIONAL WISDOM
When Oakland A’s major league baseball manager Billy Beane boldly discarded conventional wisdom and embraced advanced statistical analysis, he was soundly criticized by baseball purists until he demonstrated that it worked. Similarly, the pharmaceutical industry is filled with traditionalists and is often — very — slow — to — change. Take the simple biologics capacity planning model, for example. It would say that a plant the size of AstraZeneca’s MedImmune facility in Frederick running a legacy biologic process and producing one product straight for a year has 65 lots of annual capacity. “You have to factor in a twoweek shutdown for quality and maintenance and running at a rate of 85 percent of full capacity to allow for some slop,” notes Skibo. But the reality is — this model is based purely on the knowledge gained through experience. Companies like Amgen and Genentech have plants very similar to the MedImmune facility in Frederick and have been in operation for the better part of two decades or more. “It’s a fairly standard four pack plant,” he states. “We all have 4x15 or 6x15 thousand liter bioreactor plants.” Tell an industry veteran like Skibo at what rate of capacity a plant is running, how many bioreactors and their size, and the number of legacy biologics being produced, and he can tell you how many lots the facility is capable of producing. According to Skibo, this is why the simple model is unduly biased and flawed. “In a year when we have seven tech transfers taking place,” he affirms, “we were operating more in the 30-lot-per-year production range.” Had Skibo applied the simplistic model for capacity planning, he would have thought he had 35 more lots available than actually existed, a real problem. In 2015, Skibo anticipates that this kind of planning will get even more complex, such as conducting two commercial runs, four clinical runs, and as many as three process validation runs in the same plant. “In a year like that I’ll be lucky to reach 30 lots,” he admits. Consequently, the MedImmune team knew it needed to throw out conventional wisdom and develop a more accurate model.
One of Skibo’s first tasks was to get the biologics R&D, commercial, and clinical ops divisions operating as a fully integrated enterprise that was assessing the range of risk. He says this was an “ah-ha” moment for a lot of them because operations wasn’t looking for an exact answer. “I wanted their best guess of what the ranges were so I could go away and play with those risk ranges,” he says. “So we could determine the probability of making it to the end. Early stage, you have a 12 percent chance; in Phase 3 you have a 58 percent chance. What gets complicated is when you are within two or three years, and you have to start making a binary risk assessment because there isn’t such a thing as 58 percent success.”
In addition to getting these various risk ranges, Skibo was seeking to create a model that allowed input of other details. For example, a process development (PD) clinical run cycle is typically 11 days. A “flatlined” production run is five to five-and-a-half days. “The old model didn’t allow you to put that difference in,” he states. “When it comes to turnaround times, are they 17, 11, or 7 days?” he asks. “If you only turn it around once every 18 months, it doesn’t matter.” However, if you have to turn around five times a year, the difference between 17 days (5 x 17 = 85 days) and seven days (5 x 7 = 35 days) is 50 days, which equates to nearly two months of production you think you have but in reality, don’t. These are the things Skibo says you need to understand when planning for capacity with a risk-based approach. You also need to know where your benchmarks are now, how long it will take you to change them, and if the investment is worth the risk. “For example, a long-established product with an eight day process cycle could be improved,” he explains. “But you would have to change the license to do so.” Skibo suggests asking yourself how much of the biologic you need in any given year before you start trying to change the process (see sidebar — Shire Helps Skibo In Assessing Single- Use Risk).
In addition to having a deep understanding of what you need and how your plant runs, he advocates using a multilayered planning approach, especially on the large molecule side. Skibo believes senior leadership needs to drive the process for what the model needs to measure. Your manufacturing, science, and technology (MS&T) group — scientists who focus upon full commercial-scale process improvement rather than initial product development — should then develop the model. “As we concluded cases, factoring in risk data, the two ends of the team [senior leadership and MS&T] fine-tuned the model together to really get it to do what we needed,” he adds.
KEY LEARNINGS OF IMPLEMENTATION
Only after creating your model and analyzing some of the data will you really understand what your plant can and cannot do. For Skibo, there were four key lessons. First, many of the products in your pipeline aren’t going to impact overall plant capacity. “The ranges related to most small products are going to fall within the error (i.e., .7 to 4 lots per year) of the model,” he explains. Much like a baseball player knows that getting walked won’t hurt his batting average, managers can place these small products in the “don’t need to worry about impacting capacity” category.
Second, you learn which products do impact capacity. At the Frederick facility there are three products that drive the need for capacity. “These I really had to worry about. I needed to get the numbers right going forward because the outcomes related to how many of them are successful will drive the need for potential and substantial capacity beyond what we have,” he reveals. Third, and something Skibo says many bio-firms have now learned, “You need both the battleship [i.e., equipped with large bioreactors], and a plant with 2,500-liter bioreactors for the small products.” This led to the fourth lesson. Small products consumed a disproportionate amount of “battleship-wasted capacity” due to turnaround time. “It was worth getting them out of the battleship and into the 4x2, 500-liter bioreactor plant that we decided to build,” he affirms. There was yet another lesson that even an old industry veteran like Skibo admits he never even thought of.
ARE YOU COMMUNICATING IN THE SAME LANGUAGE?
If you have ever watched a baseball game, you will see the catcher flash signs to the pitcher, using his fingers. Sometimes, managers give signals to players on the field from the dugout. You might notice a first or third base coach using a combination of signs, such as touching his hat or pulling his ear to tell the runner or batter what to do. What makes it interesting is that each team can be using different signs to communicate the same message. But unless you know the sign, you won’t understand. The same thing happens within our industry. And though we may be using words and communicating in English, this doesn’t mean we are speaking the same language or being understood — something Skibo discovered when he met with the CEO and commercial team to discuss their findings. “Everyone has their own yardstick,” he explains. “If I’m speaking to clin-ops, we speak in terms of doses. If I speak to PD (process development), it’s in grams of protein. With clinical and research, we think of numbers of patients. Each one has their own metrics related to what this product means in their space.” Skibo says he thinks in doses, then dosage-perdose, then how much titer needs to run, and then he works his way through the math to arrive at the number of lots to be produced. To him this process seemed fairly straightforward. But when he was explaining the model’s results, the CEO asked, “Is this product constrained in capacity? It’s one of the earliest launches we might have.” Skibo’s answer, “Don’t worry about it, because it’s .7 to 4 lots.” The next question revealed that Skibo wasn’t speaking in the same language. “What are 0.7 lots in revenue?” the CEO asked. Skibo didn’t have that math in his head, and he realized that the model needed to be further fine-tuned so it could generate results in the internal languages that were meaningful to everyone. “We added to the model what the lots per billion dollars in revenue are because, at that very top level, they think in terms of how much we will sell,” he states. “That’s the math they deal with on a daily basis.” Regarding the commercial side of the business, Skibo learned they too have their own language. “If I’m speaking to commercial, it’s not just patients, but number of patients served,” he shares. His advice: When developing a model using a risk-based approach to plan for capacity, make sure you know whom you are speaking to and that the model generates results in the appropriate language. Doing so will result in a better plan for manufacturing capacity, and ultimately, patient access to your company’s biologics.
Great Communication Is A Learned Skill
There are certain voices that possess such a distinctive quality that upon hearing them you can immediately discern their owner. However, it is not just the tonal quality that makes the voice unique but a variety of characteristics, such as inflection, pace of delivery, and the use of simple words to communicate complex subjects. Great communicators of my youth, Paul Harvey, Walter Cronkite, and Ronald Reagan, owned such vocal tools and the intellect with which to use them, backed by character traits of honesty and humility that made the message easy to understand and implicitly believable. Some might mistakenly believe that they were simply “naturals.” But this is a fallacy. All of the people mentioned above worked very hard at their craft to become great at communicating their message. So too does Andrew Skibo, regional VP of supply biologics, global engineering, and real estate at AstraZeneca (AZ). He advises others to do so as well.
“We use a lot of executive coaching around change management and even presentation skills,” admits Skibo. “I’m really big on that, especially for our younger employees.” According to Skibo, you can have folks who are brilliant at what they do. However, if they can’t take that brilliant solution out of their heads and communicate it to 50 other people so as to bring the team along with them, they’re ineffective. “We have staff that ranges from people with 35 years of experience to people new to the industry,” he states. “We invest a fair amount in presentation and communication skills so we can get them all communicating the same language to one another.”
Shire Helps Skibo In Assessing Single-Use Risk
“For every decision we make, we go back to the risk of certainty that we can deliver,” shares Andrew Skibo, regional VP of supply biologics, global engineering, and real estate at AstraZeneca (AZ). “One of the most enlightening visits I’ve ever had was a lengthy tour of the Lexington Shire facility,” he states. “They gave us a complete debrief of what it took to design and bring it online and what they learned in a year and a half of operation.” What was illuminating to Skibo from this experience was the sourcing of single-use suppliers, something he thought seemed too risky when you have three new product launches depending upon available capacity. This is why when the company built its ISPE 2011 facility of the year plant in Frederick, MD it was decided to make it smaller and more flexible and to do so with stainless steel as opposed to single-use.
What Would You Do Differently?
Toward the end of my conversation with Andrew Skibo, regional VP of supply biologics, global engineering, and real estate at AstraZeneca (AZ), I asked if he would do anything differently. His reply was immediate. “Yes. There are always learning curves. I wish we had seen the need to have a more detailed model earlier because we wasted six months trying to figure out why we were stuck at 28 lots a year of manufacturing capacity.” Skibo believes the implementation of the new model for manufacturing capacity was such a game changer for the facility that he wishes he had responded to his instincts sooner. “It probably took us nine months to realize that most of our frustration was coming from the ‘This is the way we always used to do it,’ mindset.”
When Skibo first focused on the actual capacity of the new Frederick, MD plant, the company was running at 28 lots a year — and it was accepted that this was the best it was going to do. “I came from Amgen,” he states. “I know what a flatline plant can run. Talk to Genentech, GSK, and Biogen Idec, and they will tell you that a standard four-pack ought to be in the 60 to 70 lots a year range. At some point I realized, we’re going to be having this debate a year from now, two years from now, three years from now. If folks don’t understand why you can’t get out of where you are to get to where you need to go, you’ll never get there.” The typical response in such a situation is brute force, which is dangerous in a biologics manufacturing plant because it’s a quality risk. “If I had to do it over again,” Skibo affirms, “I would have infused more new team members sooner.”