How many devices do you need for approval?

Creo Quality recently asked about the possibility of making molds quickly with electron beam melting, he later found out that it does not make a good mold. How many medical devices do you actually need to make for approval? Planning for this is very important for start up companies, you don't want to slow down to make samples, but you also don't want to overproduce and waste money (although I can always find uses for extra samples). Assuming you have a medium size disposable device and are shooting for a 510(k), you don't need that many.

You'll need to validate the following:

• Biocompatibility- this really doesn't take that many samples, I'll call it 30 samples required, which is probably more than you need, but I'll play it safe.

• Packaging- but you can use dummy samples consisting of approximately the same shape, weight and density. I've never found it that difficult to come up with reasonable substitutes for packaging validation. Zero samples required.

• Sterilization/sterility- even with ethylene oxide, which requires the most intensive testing, you need less than 100 samples for this.

• Safety and effectivity- the number of samples required for this depends on how good you are and the type of product. If you're making something absolutely unusable after use, then you'll need many more samples. If you're making something more durable, you can play around and get your methods and procedures down with much fewer samples. Some of the tests can be done on incomplete samples as well, if you're testing the safety of one portion of the device, it can be isolated. I'll say it takes 50 samples to get through this.

• Shelf life- which requires do overs of the product's safety, effectivity and sterility. Usually you run an artificially aged batch and a real time batch at the same time. This time I'll assume you have your tests down and don't need samples to practice with, and you can get away with 50 samples and various dummy samples to validate the packaging to prove this. I've seen people argue that all you have to validate for shelf life testing is package sterility, but I think its worth it to rerun the safety and effectivity tests just to be sure.
  

That is a total of 230 samples assuming all goes well. Just to be on the safe side and cover giveaways (FDA, testing labs), showoffs (doctors and hospitals, you won't sell any without a few freebees spread around) and the almost inevitable problems, 400 is probably a good number. If you really know what you're doing or have a fairly simple device, I'm betting you could squeak by with less than 200. That is for ETO sterilization, which is fresh in my head, other types require less testing and you would need significantly less samples. Also, the more expensive your device is, the more reasons people come up with to justify less testing.