Clinical trial study design: 'one size does not fit all'
Andrew Lo, Ph.D., the study's senior author, told Outsourcing-Pharma.com the study’s goal was to show that "one size does not fit all" when it comes to interpreting clinical trial results.
“For patients with terminal illnesses and no existing therapies, taking a bigger chance of false positives might be a perfectly acceptable trade if there's a reasonable chance that a new therapy will help,” Lo said.
The study, recently published in JAMA Oncology, examined the application of Bayesian decision analysis (BDA) to a data set of 10 clinical trials from the Alliance for Clinical Trials in Oncology.
As Lo explained, “Bayesian decision analysis is a well-known technique developed by statisticians to balance the potential harm of false positives (approving ineffective drugs) against the potential harm of false negatives (rejecting effective drugs).”
According to Lo, the key takeaway is that for certain types of hard-to-treat cancers with no current therapies, the rate of false positives should be set “much higher” than 2.5% or 5%.
“By adapting the rate of false positives to the specific clinical setting – higher rates of false positives for more dire circumstances – more therapies will be approved, thereby incentivizing biopharma companies to produce more innovative products for patients,” Lo explained.
Additionally, he explained the Bayesian decision analysis (BDA) framework “allows all stakeholders to weigh in while providing the FDA with a rational, transparent, systematic, objective, and repeatable process for making decisions that incorporate patient preferences.”
The 21st Century Cures Act requires the FDA to consider patient preferences and “BDA provides a straightforward platform for doing so,” added Lo.
“Our results show that incorporating patient preferences into the drug approval process is feasible and can greatly increase innovation in biomedicine.”
According to the researchers, the framework is flexible enough to include various stakeholder perspectives. Additionally, the authors have provided open-source software so their analysis can be re-run under a different assumption set for risk preferences, disease burden and prevalence, and value delivered to patients.
Publisher: JAMA Oncol.
doi: 10.1001/jamaoncol.2017.0123
