Bayesian Gets A Boost With Guidance, But Technique Still Not For Everyone

Article preview reprinted from "The Gray Sheet"- February 22, 2010

The complex calculations of Bayesian statistics are not likely to become standard for FDA device submissions anytime soon, but they could help streamline clinical trials in many more instances, CDRH's top statistics official says. Read more...

Bayesian Gets A Boost With Guidance, But Technique Still Not For Everyone

Article preview reprinted from "The Gray Sheet"- February 22, 2010

The complex calculations of Bayesian statistics are not likely to become standard for FDA device submissions anytime soon, but they could help streamline clinical trials in many more instances, CDRH's top statistics official says.

Greg Campbell, who heads the biostatistics division in the device center's Office of Surveillance and Biometrics, has long been a proponent of the advanced techniques, which potentially allow for smaller and shorter-term studies by "borrowing strength" from prior data (The Gray Sheet' May 31, 2004).

"The beauty of [the technique] is you do not end up doing a trial that is too big or too small; you end up doing a trial that is just right," Campbell said in an interview.

His team reached an important milestone towards increased adoption of Bayesian approaches earlier this month with the release of a long-awaited final guidance document for using the tools in device trials (The Gray Sheet' Feb. 8, 2010).

The guidelines attempt to provide companies with a roadmap for employing Bayesian statistics appropriately and to their advantage.

"We area actually pretty excited about the adaptive nature of Bayesian trials," Campbell said. "It could be used in a lot of settings that currently it is not used in."

Big Advantages, But Also Complications

Currently, Bayesian approaches support about 5%-10% of pre-market device submissions to FDA, according to Campbell.

The first applications applying the methods started to trickle in at the end of the 1990s and the early 2000s. And there was some rise in interest in the techniques by firms after the 2006 release of the draft version of the Bayesian guidance (The Gray Sheet' July 31, 2006).

Recent examples of devices with pre-market applications relying on data from Bayesian-designed trials are TherOx's SSO2 oxygen therapy for treating heart attack patients, which received a negative panel vote last March (and is awaiting a final decision from FDA) (The Gray Sheet' March 23, 2009), and Johnson & Johnson's Navistar ThermoCool ablation catheter, which last year became the first catheter approved to treat atrial fibrillation (The Gray Sheet' Feb. 9, 2009).

Also, "There continues to be a stream of orthopedic products that regularly use Bayesian methodology," Campbell said.

Nonetheless, he does not anticipate much of an increase in its use in the near future.

"I don't think it will become a primary approach, partly because it is much more complicated from a statistical point of view," he explained. "Companies would need to make sure that they have people that could design and analyze a Bayesian trial. That could be a limiting factor."

But there are serious potential advantages to the approach that companies should be aware of, including cost savings, Campbell said.

Hence the guidance, which comes almost four years after the draft version was published. During the interval, Campbell and his team analyzed stakeholder comments and focused on improving the explanations about how Bayesians statistics can be applied.

Sign up for a complimentary 30-day trial of "The Gray Sheet" and get immediate, complimentary online access to our best-selling Special Report: "When is an iPhone a Medical Device?" There is no obligation and no credit card needed.  Click here to sign up and get your Special Report now

Plus: Meet the promising new breed of oncology technologies in Medtech Insight's new 58-page Special Report: Oncology Outlook: How Cutting-Edge Medical Technology Is Improving the Odds for Cancer Patients. Available for purchase online  

To find out about more about more about Elsevier Business Intelligence's medical device publications and databases, multi-user access and/or advertising with Medical Devices Today, please contact Julia Uspaskikh at 240-221-4538 or j.uspaskikh@elsevier.com