June 13, 2007 | In the early 1990s, as a senior biostatistician at the Genetics Institute (later acquired by Wyeth), Jerald Schindler found himself pining for statistical gear to bridge all the pesky stops and starts of drug development. He envisioned a continuum of data crunching from the earliest dose-finding studies to the final dose-testing trials that count toward product registration. Assumptions would be checked while a trial was in progress, rather than at its conclusion a year or two later, and the protocol would be altered accordingly.
Though Schindler didn’t know it at the time, he was describing what has since been termed adaptive clinical trials. He was in fact responsible for establishing the vision for the underlying eClinical infrastructure that has allowed Wyeth to become one of the pharma trail-blazers in the adaptive trial field (see “Biting the Adaptive Trials Bullet,” Bio•IT World, May 2007). In 2005, he became a big advocate of the concept as a consultant presiding over the pharmaceutical research division of Cytel.
But in March, Schindler left Cytel to return to big pharma, this time as the new vice president of late-stage clinical development statistics at Merck, where Schindler will be giving the adaptive approach its most serious exposure yet. “Part of the reason I’m here is to...look for opportunities to do dose ranging and Phase 2/3 trials,” he says. Currently, Merck has a “relatively small number” of adaptive trials completed and in progress in different phases.
“In my view, adaptive clinical trials make sense everywhere,” says Schindler. “There are different opportunities at different stages. Early on, you want to identify products that are likely to continue and find the right dose regimen and patient population. You start with a blank slate and over time explore a lot of ideas and eliminate ones that don’t look promising. In a later stage, you...start with a few doses and identify the best of the few.”
As with any major shift, the key challenge will be overcoming inertia issues. Adaptive trials require an investment in upfront planning “to do simulations and understand the different options and their implications,” says Schindler. Merck will continue to develop its “worldwide common infrastructure so all [trial-related] data flows into similarly structured databases,” thereby allowing the re-use of simulation and data analysis tools. Adaptive trials demand “a lot of data analysis, and quickly. You may need to do three or four small analyses of the data. If it’s painful to do, you will run out of resources quickly.”
One type of frequentist adaptive trial, known as a “group sequential trial,” has gained wide acceptance over the past 30 years. “These trials usually contain a small number of interim analyses with the option to stop the trial, or discontinue a treatment arm, based on how various patient groups, on different drug doses, are faring,” says Schindler.
The alternative Bayesian statistical approach deals with new data as it comes in and estimates the probability that the drug is effective. The blending of data with prior ideas makes the methodology a bit controversial. The required computing power is also overwhelming, although tools to do Bayesian analysis have improved substantially over the past two decades, he says.
“Adaptive clinical trials can be done in a frequentist way, a Bayesian way, or a mix of the two,” says Schindler. But the new adaptive approach to clinical development establishes a process for making interim decisions during the drug development process that “takes advantage of all the information you have all the time.” Conceivably, a drug could journey through clinical development based on a combined adaptive proof-of-concept/dose response trial and then an adaptive, multiple dose Phase III trial that “ends enrollment into groups that don’t do as well and focuses on the dose or doses that will make it to market.”
Adaptive drug development across the entire portfolio “enables companies to allocate resources among all products based on real-time results from the ongoing clinical trials,” says Schindler. “The combination of the adaptive process with portfolio management will likely be the next step in the evolution of the adaptive approach.”
The New Era of Drug Development
A main concern of United States and European regulators is the potential for bias to creep into an adaptively designed Phase III trial via communications with investigators and patients about whether a particular dose is working, says Schindler. “It’s hard to prove you didn’t influence [future data], so you need to lock data down and keep it separate from those who conduct the trial.”
Regulators are clear that, with early-stage trials where blinding is not a big issue, “companies can look at almost anything they want within reason,” says Schindler. Consequently, it pays to “learn a lot about a drug early, when the bar is lower.”
Just about every pharmaceutical company today is doing adaptive trials, including those that involve Bayesian statistics, or talking about it, says Schindler. “Some companies will not do well and give up. The others — and presumably Merck will be among them — will figure it out and zoom ahead.”
This new level of skill required for successful adaptive trials will give companies a major competitive advantage over companies that fail to develop the necessary skills, says Schindler. “At Merck, we have been adding to our already strong statistical capabilities by selecting talented statisticians to join our statistical staff with the desire for close collaboration with the clinical research departments.”
That translates into a lot of positives in the years ahead, and not just in terms of adding efficiencies to the business of drug development, says Schindler. “The fact that we have better information earlier is a big deal.” Adaptive trials can lower the risk of clinical development by “identifying losers early.” They can improve public health. Effective treatments will get to market sooner. Companies also won’t give up on useful compounds simply because of bad guesses on dosing.
“One of the side benefits [of adaptive approaches] is that they require close collaboration among everyone who helps run a trial...making the whole process more efficient,” says Schindler. “Information should be shared when people need to know it.”
Schindler adds, “What I’d like to do at Merck is help influence how drugs are developed. [The company] has a pipeline filled with potential products. All of us here need to find a way to develop good products in the most efficient way and get them to market. The adaptive approach will go a long ways toward doing that.”
It also makes the job a whole lot more interesting. “As the decision making process is driven more and more by understanding emerging signals in the clinical data, the role of the statistician becomes even more important.”
Subscribe to Bio-IT World magazine.