By Deborah Borfitz
August 25, 2008 | Tessella’s soon-to-launch Adaptive Trials Simulator will be the first opportunity for most sponsor companies to “interact directly” with a Bayesian design of Berry Consultants. The simulator is built around a flexible core design developed by the statistical consultancy and ready for deployment, out of the box, for about half of all Phase II dose-finding adaptive clinical trials (ACTs), says Tom Parke, head of clinical trials solutions at Tessella.
For other trials having “novel characteristics,” the software will need only a few minor tweaks by Berry Consultants before it can be used to make adaptations, says Parke. Up to now, companies typically hired the consultancy to produce designs from scratch for specific trials.
Tessella first attempted to launch the Adaptive Trials Simulator last year, but ran into unexpected legal concerns over intellectual property rights with some of its partners. While lawyers were ironing that out, work on the simulator progressed and it grew “considerably more sophisticated,” says Parke.
Previously, the tool consisted of a single statistical model suitable for dose-response studies where the endpoint could be measured on a continuous scale. It has since evolved to accommodate ACTs where the outcome is dichotomous (yes/no) or measures time to an event, such as a second heart attack, says Parke.
Final patient outcomes can take many months to obtain. But predictions and adaptations can be made by the Adaptive Trials Simulator early in a trial using preliminary data, as was the case with Pfizer’s oft-mentioned ASTIN (Acute Stroke Therapy by Inhibition of Neutrophils) study, says Parke. This “longitudinal” model, used strictly when a study is in progress, has been a component of about 50 percent of ACTs run over the last few years.
On a few trials, Tessella has supplemented sponsors’ existing electronic data capture (EDC) systems to feed “headline numbers” into the probability model during the lag time between data capture and data cleaning, says Parke. In one trial currently under way, Tessella has given investigators the wherewithal to fax in raw patient data at the point of capture.
Also enriching this year’s launch is the fact that the simulator includes an option to model dose response using a logistic curve, for occasions where there is strong evidence that level of dose response will move in parallel with dose amount, says Parke. It’s a simpler, more efficient modeling tool than Normal Dynamic Linear Model (NDLM) for dose finding that allows for the possibility of patient response worsening at higher doses.
Whichever model is used, the Adaptive Trials Simulator estimates whether any dose will fit the target criteria and which dose it is likely to be, thereby recommending whether or not to continue with a trial and which dose to favor in randomization, says Parke. Seamless Phase II/III simulations, which select treatment arms to retain and calculate sample size, can likewise be done with the tool.
The Adaptive Trials Simulator made its debut at the 3rd Annual Trial Design Innovation Conference in Washington, D.C., last month and will launch commercially in the fourth quarter of this year, says Parke. Three top 10 pharmaceutical companies are already actively working with Tessella in a handful of dose-finding ACTs designed by Berry Consultants. In one central nervous system study, the sponsor figured it saved $1.5 million by stopping the trial early upon learning the compound wasn’t working as planned.
Earlier this year, Tessella launched its drug supply forecasting tool that is useful for any trial, adaptive or not, involving large volumes of drug. Most importantly for ACTs, Tessella handles all the necessary integration between disparate EDC, randomization, and drug supply systems, says Parke.
________________________________
This story first appeared in eCliniqua,one of Bio-IT World’s free e-newsletters. Subscribe here.