However, for the biomedical sciences, you should assume on the order of half of the scholarly, published in refereed journals papers are incorrect. See e.g. that essay by a Lancet editor, and the general history of this, which I've been watching since the late '70s.

Why? I think it's because this stuff is hard, so many variables, publish or perish, and the fact that while referees for any field can filter out a lot of guff, they have strict limits in their ability to determine correctness. That comes when the results of a paper are used as the basis of further research in other labs.