Having clinicians and modeling experts work together on simulated trials can save candidate drugs, as well as time and money.

Gina Shaw 
Shaw is a freelance writer based in Montclair, N.J.

The innovation of in silico development, or the use of computer-based modeling and simulation tools and techniques for drug development, can replace costly and complicated elements of the development process, such as certain aspects of clinical trials. For more than a decade, analysts have predicted that clinical trial modeling and simulation would soon come to the forefront of the drug discovery and development process, just as simulation has for so long been a cornerstone of industries such as aerospace and automobile manufacture.

But it hasn't exactly played out that way. The use of clinical trial simulation and modeling in the pharmaceutical industry is still far from consistent, and the main barrier isn't one of technology. "The field has grown in leaps and bounds primarily because the hardware has really evolved, and we have a much more accessible environment to high-

click the image to enlarge  Strategic modeling and simulation (M&S) can facilitate dose selection, minimize patient risk, and increase the probability of clinical trial success. (Source: Marc Pfister, MD)

performance computing. Techniques have also matured, as well as our software solutions," says Jeffrey Barrett, PhD, director of the laboratory for applied pharmacokinetics and pharmacodynamics at the Children's Hospital of Philadelphia and former head of global biopharmaceuticals for Aventis, now Sanofi-Aventis, Paris, France.

What still hasn't caught up to the leapfrogging improvements in the data, hardware, software, and computational tools that facilitate clinical trial simulation, experts like Barrett say, is the training of industry scientists, along with the overall mindset of the pharmaceutical industry.

"It's difficult to get scientists with the skill set needed to actually perform clinical trial simulations. All pharmaceutical companies draw from the same pool of scientists, and it's a very small world. We do not have many universities offering classes or programs in modeling and simulation," says Marc Pfister, MD, director of strategic modeling and simulation for clinical discovery with Bristol-Myers Squibb.

Another challenge, Pfister says, is that modeling and simulation experts need to learn the language of clinicians. "Traditionally, clinicians do not have a strong statistical background, so it's very important that we as the modeling and simulation community learn to communicate complex methods and complex results to our stakeholders in the industry."

A limited multidisciplinary talent pool isn't the only problem, says Sriram Krishnaswami, PhD, associate director of clinical pharmacology with Pfizer Inc., New York, and chair-elect of the modeling and simulation focus group of the American Association of Pharmaceutical Scientists. "To me, the single biggest hurdle in terms of more widespread application of clinical trials simulation would be buy-in, or the hierarchical structure that exists in companies." Clinicians, statisticians, and clinical pharmacology modelers are not on a par with each other in most pharmaceutical companies, he says. "The clinicians are a little higher, and the others are almost like service providers."

Krishnaswami and other simulation experts also believe that these three sets of players need to be more fully integrated in multidisciplinary teams, something that is now happening at Pfizer after a reorganization five years ago. "The pharmaco-metrics department has people from the statistics department as well as clinical pharmacology sitting together. This is an unusual approach," he says. "In fact, the [US Food and Drug Administration (FDA)] came to Pfizer last year to have a look at how we're organized in terms of this activity."

That is the approach at Bristol-Myers Squibb as well, says Pfister. "We recently did a survey, and most major companies now have a separate modeling and simulation group. At Bristol-Myers Squibb, we formed ours about two years ago. Although it's a separate group, we integrated this more technical modeling and simulation group in a clinical department. We're not isolated; we have close interactions with our stakeholders as part of a clinical team."

But that's not true everywhere. "Every company takes different approaches to its modeling and simulation group," says Barrett. "Some will remove it from the project team in order to focus on the simulation aspect, but I'm not a fan of that. Simulation has more influence it it's integrated into the project team from the beginning."

Killing and saving compounds
What can clinical trials simulation, if fully adopted and integrated, do for the $30 billion-plus research budget of the pharmaceutical industry? The possibilities are virtually limitless. At the phase I and phase II stages, clinical trial simulation can help fine-tune dose selection and study design. In phase IIb to phase III trials, simulation can help yield an understanding of the study population. "You can use simulation to identify responders from nonresponders, and this helps you to fine-tune, for instance, your inclusion and exclusion criteria. By identifying the population that is most likely to respond to a drug, you increase the probability of success," says Pfister. "At the same time, modeling and simulation allows you to identify patients that may have adverse events. In summary, it helps you identify the right drug, at the right dose, for the right population."

This isn't just theoretical. Five years ago, Aventis killed a candidate anti-osteoporosis compound aimed at competing with Eli Lilly's selective estrogen receptor modulator drug Evista (raloxifene hydrochloride). That was after clinical trial models using the popular Trial Simulator program from Pharsight Corp., Mountain View, Calif., showed that the drug's therapeutic benefit didn't match that of Evista.

But trial simulation doesn't just kill drugs. Sometimes it can save them. Recently, Pfizer had a new product in development for treatment of acute pain. "The biggest problem with pain drugs is that the models they use for measuring pain can be very misleading. The dose response is very shallow, and people tend to assume that greater doses will not be as effective. That's not always true," says Krishnaswami.

The new product had been scrapped after phase II testing found its pain relief wasn't adequate, but a few months later, the pharmacometrics group suggested that the drug's demise was premature. Clinical trial simulation, they said, suggested that a certain level of dose increase of the drug, administered in a particular way, should show a clinically meaningful difference compared to currently approved competitors.

"This was all done using literature data, pooling a multi-competitor analysis within the same class of drugs to figure out a dose-response profile," says Krishnaswami. "The

Clinical modeling and simulation can save time and money, but more importantly, it can get new drugs into the clinic sooner. (Source: National Cancer Institute)

study was done and the results were almost unbelievable: exactly as predicted, there was a dose-staging issue. The trial was successful and now the drug is moving forward. That, to me, is the Mount Everest of clinical trial simulation impact—resurrecting a compound."

This and other successful simulations have helped lead to a new mandate at Pfizer, implemented just within the last year. "It's written in our company divisional goals: 80% of all compounds undergoing clinical testing will undergo enhanced clinical modeling," says Krishnaswami.

Modeling and simulation can also be used not just during the trial design process, but mid-study as well to adapt the trial design as new data comes in. "Instead of conducting a phase IIb study and then waiting six months to do phase III, what we can do is combine studies into a phase IIb/III study," Pfister says.

For example, a study may start with five doses of a drug. After six months, ongoing computer models that are updated as data is collected may tell investigators that the lowest-dose arm of the trial yields results no better than placebo. So the low-dose arm could be eliminated, with patients only enrolled in the remaining four dose arms. "Maybe after 100 patients, we would find that the next lowest dose arm is also not superior to placebo," says Pfister. "So we could eliminate that arm and then extend the higher dose arms by one year to collect the safety data, so it's a seamless phase II/phase III study."

Regulatory buy-in
The FDA is actually somewhat ahead of industry in endorsing clinical trial simulation, says Pfister. "There are examples now where certain statements in drug labeling are based on pure simulation. Ten years ago, you had to conduct the study; today, once you have a large data set and a robust model, you can simulate a trial and indicate a label claim. The trend is that the FDA is more and more accepting and supportive of the idea of using novel approaches to streamline drug development."

Ultimately, says Pfister, it's an investment that will pay off. "Clinical trial simulation adds value, improves decision-making, and saves time and money and lives. If we can demonstrate that trial simulation allows patients to get important drugs earlier and reduces risks for patients because we can identify the right target population, then this field will really take off."