At Certara, we are not afraid to think big. In fact, solving the hardest problems in pharmaceutical R&D is our passion. You might say that some of our ambitions could be described as “moonshots.” After all, they meet the criteria put forth by the Google X moonshot program.
- We seek to address the huge problem of getting safe and effective medications to patients while containing or reducing the burgeoning costs and time lines that pharma is currently incurring.
- Biosimulation— the use of modeling and simulation for drug development— is a radical solution for an industry that still largely relies on empiricism.
- Realizing the benefits of biosimulation will require the use of breakthrough technologies that integrate our understanding of biological systems with the power of computer modeling.
In this blog post, I will discuss the human and economic investments required to develop the field of quantitative systems pharmacology (QSP) to the point that it becomes an integral part of pharma R&D.
QSP: The next piece in the puzzle for addressing Phase 2 attrition
Decades of organizational and other changes have not improved Phase 2 attrition. The root cause of this attrition is failure to show efficacy. Our inadequate understanding of biological systems presents the major challenge in addressing drug efficacy. At Certara, we believe the solution to the problem of Phase 2 attrition lies in mathematical modeling of human system behavior. Biosimulation is a logical and relatively low cost way to tackle this key drug development issue.
The original moon shot proceeded in careful stages. First we built rockets that could escape earth’s atmosphere, then low orbits of earth, trips to the moon without landing, and finally Apollo 11 landed and returned. The biological moon shot is already well underway. A great example is the field of physiologically-based pharmacokinetics (PBPK). The FDA and many managers in pharma and biotech R&D now rely on mechanistic modeling to improve understanding of PK and therefore improve PK-dependent decisions.
Our chief scientific officer, Amin Rostami, defines a mature field not by whether the field can answer every interesting question, but by whether people take it seriously. He deems a field with at least 60 peer-reviewed publications to be mature. By that standard, PBPK is definitely mature.
Yet, PBPK addresses only about half of variability in drug response. The rest is tied up in epigenetics driven by lifestyle and environment, and, critically, variability in a patient’s “PD system”—their biological responses to drugs. Addressing the question of pharmacodynamic variability is where QSP comes in. This relatively new discipline seeks to answer two critical questions:
- How do drugs work (mechanism)?
- How will a patient respond when a drug perturbs her biological pathway(s)?
Once we know 1) how much drug arrives at the site of action, and 2) how that drug will engage the target, and 3) how the engaged target modulates the cell signaling pathway that leads to a pharmacological effect, we have the three pillars of understanding to guide rational drug development. We know where to perturb a network, and where not to bother. Phase 2 attrition will fall. The economics of drug R&D will be transformed. New tools will become available at the bedside to guide personalized care.
Realizing the potential of QSP
You say, that’s a great vision, but it’s too big a problem. We’re not NASA. We don’t have the resources President Kennedy had when he initiated the real moon shot. However, the firmly established field of PBPK provides real data that suggests the scale of the task. We can ask how much it would cost, and how long it would take, in relation to the PBPK effort, to generate predictive and reliable QSP models for all 1,000 or so diseases of interest to pharma.
Simcyp was founded in 2001. Since then, Simcyp has devoted approximately 500 man-years to create the excellent Simcyp Simulator that is used by all top pharma and the FDA. Humor me for a moment. If 1,000 QSP models all need 500 man-years to become as robust as the Simcyp Simulator, and a man-year costs about $100,000 in the West, then 1,000 indications X 500 man-years/indication X $100,000/man-year = $50 billion investment.
Some people think an estimate of 500 man-years to develop a single, robust QSP model is too high. They say that useful models can be built with 30 to 100 man-years of effort. Simcyp reached “escape velocity” in a few years after founding, well before investing 500 man-years of effort. If these lower estimates are reasonable, this would cut the costs to about 10-20% of the $50B. The total man-years needed to make 1,000 robust QSP models might be as low as 30,000.
Whether the costs of developing a QSP computational infrastructure are $5B, $10B, or $50B, it’s a lot to pay. But, consider that the pharma and biotech industries spend $150B on R&D in a year. The US National Institutes of Health (NIH) annual budget is another $30B, although that is spent on many research topics besides drug therapies.
Robust QSP models that reliably inform the strategies for dosing or choosing drug combinations should be able to significantly improve Phase 2 success rates. In 2010, Steven Paul and colleagues published a widely cited paper on the economic benefits of increasing the efficiency of drug development, “How to improve R&D productivity: The pharmaceutical industry’s grand challenge.” If Paul’s figures are correct, then improving Phase 2 success rates from 34% to 50% will save $400 million per new chemical entity (NCE). The industry launches about 35 NCEs per year. So the saved Phase 2 costs would be 35 X $400M = $14B savings per year. Not a bad return on a $5B, $10B, or even $50B investment.
In addition to financial considerations, there is also the question of the trained brainpower needed to fuel this effort. The currently available manpower resources are a serious constraint. There are about 2,000 pharmacometricians in the world. Today, probably only about 500 are competent in QSP modeling. At 30 man-years per model, the 500 can produce 17 models per year, or put another way, the universe of mechanistically competent pharmacometricians will take 58 years (1,000 models/17 models per year = 58 years) to create all the QSP models we need.
To solve the problem in a year, we would need to increase the number of mechanistic modelers by 58 times. Furthermore, there isn’t the requisite data to build the models in many indications today. However, as a 10-year project, the problem of establishing a robust QSP model in each of 1,000 major diseases becomes much more tractable. A moonshot program consisting of 5,800 QSP modelers linked with common software and tools would conceivably have the wherewithal to tackle this issue. And in terms of funding, the cost would be a relatively modest $600M per year to pay for the 5,800 mechanistic modelers. For perspective, there are 277,000 mechanical engineers in the US.
I quote President Kennedy liberally:
We possess all the resources and talents necessary. But the facts of the matter are that we have never made the decisions or marshaled the resources required for such leadership. We have never specified long-range goals on an urgent time schedule, or managed our resources and our time so as to insure their fulfillment. This enterprise demands a major commitment of scientific and technical manpower, materiel and facilities, and the possibility of their diversion from other important activities where they are already thinly spread. It means a degree of dedication, organization and discipline which have not always characterized our research and development efforts.
New objectives and new money cannot solve these problems. They could in fact, aggravate them further—unless every scientist, every technician, and manager gives his personal pledge that we will move forward, with the full speed of freedom, in the exciting adventure of human physiology.1
 May 25, 1961: JFK’s Moon Shot Speech to Congress
A thought leader discusses QSP
My colleague, Dr. Piet van der Graaf, was recently interviewed by Center Watch Weekly. Read the article to learn more about QSP’s potential to increase pharma R&D productivity. Is QSP worthy of a moonshot program? Let me know what you think in the comments section!