Pads38, that's a great list, thanks.
Since posting my question, I have been thinking about the difference between searching for data on your device versus data on "equivalent" devices. I've worked mostly with Class III devices supported by a clinical trial. For these devices, published literature takes a bit of a back seat, where it is a key source of clinical data for devices not supported by a clinical trial. You can usually limit the literature search to publications of data on your device, rather than including equivalent devices. In addition, the devices were on the market in the US, so they were also subject to FDA oversight and reporting.
For this situation, MAUDE does not seem worth searching, because all of the reports for your device in MAUDE are also in your complaint files (and vice versa), and therefore those events will be included in the evaluation of your complaint data. More than that, if you included both the MAUDE data and your complaint data, you would be reporting many of the same adverse events twice. The same applies to FDA's recalls. Hard for FDA to recall your device without you knowing about it and any adverse events that inspired it. So it seems to me that FDA data is useful for finding adverse events for equivalent devices, but not for your own. Or am I missing something here?
Embase is described as "a biomedical and pharmacological bibliographic database of published literature designed to support information managers and pharmacovigilance in complying with the regulatory requirements of a licensed drug." When I first started writing CERs, I searched it several times and and never found anything in it, so I finally stopped searching it. Has anyone found device data in Embase?
I'm not a big fan of clinical trial databases as a source for CER data. Like MAUDE reports, they strike me as unconfirmed rumors, not "valid scientific evidence." I am skeptical enough of data that makes it into a peer reviewed publication. Also, if the publication reports data in your clinical trial, then you are once again reporting the same data twice.