Before all this talk of “Snowmaggedon”, a beaming headline posted late last week stating that, “FDA Issues Guidance to Help Streamline Medical Device Clinical Trials”.  If you were hoping to learn how it works, you were in for a statistics lesson on Bayesian methods for designing studies and analyzing clinical data.  (Wait, don’t leave!)  Good news is you don’t need a statistics lesson, to simplify the main point. FDA is now advocating the use of prior clinical data and even post-market data from a prior device to justify shorter clinical studies and/or smaller sample sizes. This is a huge change from the same Agency that has demanded more data and more time to review device approval applications.  How can this be?  Why does Bayesian probability allow us to do this? The answer is not rocket science.

Beyes’ Law basically views probability as just “degree of belief” that an event will occur today, given knowledge of prior events.  Think of it this way…Have you ever “tried” to flip heads on a coin, or roll an 8 with dice?  How is this possible, if the probability of coins and dice are already known?  Since we know the physical act of flipping the coin can affect its outcome, it may not be a simple 50-50 probability.   The Bayesian idea of prior beliefs, allow us to set constraints and make conclusions beyond the actual data.  Humans are quite good at intuitively making predictions on very little information.  Bayes just put this into mathematical notation. (not to diminish the significance of Bayes’ Law in any way.)

Now, before this starts to sound like a “free pass” to shorter clinicals, be sure to finish reading the FDA guidance.   Once again for the sake of time, we can summarize… It makes sense that statistical theory cannot replace sound clinical science.  The FDA is quick to caution that patient data from prior studies rarely are 1-to-1 exchangeable with the patients in the current study. Instead, a test can be applied to find their  “borrow strength” from the previous studies.  Also, FDA will require that you still submit to their reviewers, your rationale for considering such prior clinical data. All of the same requirements apply when filing for an Investigational Device Exemption (IDE), and all methods and assumptions will need to be reviewed before the study can begin.

That said, the fact that good prior information on clinical device use exists, and that a Bayesian approach may enable smaller-sized or shorter-duration trials is welcome news to the industry.  In the official press release, FDA Commissioner, Dr. Margaret Hamburg says, “This is a terrific example of regulatory science in practice at FDA.”  We at my510k.com hope to see these kinds of developments continue, balancing regulation enforcement with sound scientific practices.