The article from Nature yesterday came up with 26% of the peer reviewed published papers they examined (all RCT) were untrustworthy based on close examination of their data. They could only invalidate 2% without data.

I personally believe this is an underestimate.