there might be multiple (duplicate) submissions
Another possibility, and there are probably others, but without transparency, doesn't seem useful to spend a lot of time coming up with yet more.
I'll be interested to see if anything changes once there is enough data to look at probable failure rates for a few more years. These two years were "early days" in terms of a learning curve for both industry and CDRH, the draft guidance having been only issued in August 2014. It will also be interesting to finally see failure rates for De novos submitted after user fees were imposed. I'm not clear how user fees might impact De novo failure rates, but it seems like an obvious question to ask.
Of course, if you are planning to submit a De novo any time soon, these later data won't be of much help.
it is another reason why it'd be nice if FDA could maintain some public data on rejected applications.
I would like to work towards that, but it would take people from industry stepping up to say that posting basic information, like DEN#, date of submission, date of resolution, and the nature of the resolution (withdrawn, rejected, subsequently reviewed as 510(k), PMA, etc.) will not threaten their "confidential and proprietary" information, and at times would even be helpful.
But this seems to be an industry that doesn't often speak up for itself, leaving a void that others rush in to fill. Which might be okay, except that it's not clear to me that these others actually know that much about the industry or what is likely to best serve its interests.
Perhaps I'm also only "look[ing] from my little tide pool", but I'm skeptical
I'm thinking there are not a lot of "innovative" startups in your tide pool? Definitely no shortage of those that do not heed and/or lack the capacity to understand. Except I wouldn't normally expect this crowd to pursue a De novo, more like run away from one as fast as they can. Still, the total number of "failed" De novos in 2015 and 2016 was only around 70, so perhaps some number of these were exceptions that prove the rule.