August 06, 2007
The math of adaptive trials is hairy. But the appeal is simple. Adaptive trials can reveal the optimal dose in a single trial—not six. They can identify dud drugs early. The approach can require significantly smaller sample sizes and save big companies billions of dollars.
There’s just one mathematical fly in the clinical ointment. Adaptive trials will, by design, have wildly variable needs for drug supply that are difficult to forecast by either of the industry’s preferred tools, the time-honored seat-of-one’s-trousers method or Microsoft Excel.
With six or seven arms of an adaptive trial starting up (or not), much less running the length of the study (or just a week) it’s possible the sponsor of such a study would (under improbable but not impossible assumptions) need several times as much drug supply as a traditional design. Or less. It all depends.
Multiply the costs and uncertainties of one adaptive study across a portfolio of trials, and it’s no wonder some sponsors are hesitating to explore adaptive approaches just because of the supply forecasting issue.
The companies have integrated two different programs for simulating entire trials and drug supply needs. Cytel’s DoseSim and Tourtellotte’s tcVisualize will be able to generate both a set of trial scenarios—and an estimate of drug supply requirements tailored to those assumptions.
After Statistics, Logistics
“You are trying to trade off having a huge amount of quantity against the risk of running out of the right dose,” explains Nitin Patel, Cytel’s founder, chairman and chief technology officer. Patel admits that enthusiasm for adaptive techniques may not be equal across all areas of domain expertise. “The statisticians love an adaptive trial,” he says. “But the drug supply people tear their hair out.”
Cytel’s DoseSim software isn’t mentioned on the Cytel website. It is ready for prime time, however, with giant Merck already using the software and helping Cytel to sharpen its capabilities. By adjusting key study parameters, DoseSim can help sponsors zero in on the study outcomes that clinicians think are most likely. Some sponsors simulate a trial a hundred times. Others, seeking more knowledge about various scenarios, simulate ten times that number of trials.
The objective of using the right software and the right partner, says Patel, should be to make operational choices about a trial that make sense both statistically and logistically: “You could do some things to the design that don’t give up statistical power but can reduce the complexity and cost of supply.”
Patel believes there is a growing recognition across the industry that biostatisticians and others will have to collaborate to solve drug supply issues. “It’s not a throw-it-over-the-wall kind of thing,” he says. “You have to work together.”
Once the most probable scenarios for a trial have been generated, the data can be passed in a file to Tourtellotte’s tcVisualize. That program’s role is to estimate supply needs, but the Cytel trial design tool makes its predictions all that much more robust.
Company president and CEO Ed Tourtellotte says drug supply needs in adaptive designs are significantly more complicated than traditional approaches. “With a regular trial, if you have a single center, you can figure it out. You can do it out on the back of a napkin if you want to,” Tourtellotte says. “Adaptive is just so much more complex.”
The combination of the two software tools, he believes, is something genuinely new under the operational sun. The companies are promising a way to create a mathematically rigorous estimate of a trial’s drug supply needs, one that minimizes overage but comes close to assuring the sponsor that an empty-handed stock-out situation will be avoided. Doing that without breaking the bank was the topic of last month’s forward-looking ExL Pharma conference on trial design where the Cytel-Tourtellotte arrangement was announced.
“You can actually get to somewhat of a scientific answer,” Tourtellotte says. “Before, you really couldn’t. You could way, way over package or do a forced randomization. We’ve finally gotten around to scientifically calculate what’s needed for an adaptive trial.”
Over time, Tourtellotte says he hopes to be able to help clients re-simulate drug supply needs in midcourse of a trial, just to be sure the ship isn’t headed into the rocks.d9A2t49mkex