Scientists do not accept or reject findings as a function of probabilities, though they use probabilities every day. It gets somewhat technical, but Godel's incompleteness theorum is ultimately the rationale.
(Although, if you argue that there are no scientists at the EPA, I won't really object.)
To the contrary, they do it all the time, e.g. silicone breast implants not responible for mixed contective tissue disease, vaccination with thimerosal preserved vaccines not responsible for autism, etc.
What Does the Institute of Medicine Report Say About the Safety of Silicone Breast Implants?
When they do statistical meta-analyses, they are not responsible for the original studies, some of which they eliminate for being problematic.