Although modeling and simulation are likely to become fixtures in drug development, they do come with risks. Just ask Duke University.
Enrollment has been suspended in three cancer treatment trials and investigators are now reviewing the data amassed from a project to test whether gene expression can help predict tumor response to drugs. The research is based on prediction models whose validity is now as much in question as the academic credentials of the lead researcher, Anil Potti of Duke.
The story goes back to 2006 when Potti and fellow researchers published a paper describing how they developed algorithms that predict tumor sensitivity to specific chemotherapies, based on genetic analysis of the tumor. But biostatisticians at the M.D. Anderson Cancer Center found "serious errors that question the validity of the prediction models," according to Nature. They were unable to reproduce Duke's results. and published their findings last year.
Duke halted enrollment then, but resumed after other experts affirmed the methodology. Although confident in the reaffirmation, Duke again halted enrollment earlier this month due to questions about Potti's credentials. He is now on administrative leave pending an investigation into his claim of being a Rhodes Scholar, which the Rhodes Trust denies.
The university also has called for a full review of the data and science.