Map of life expectancy at birth from Global Education Project.

Monday, August 15, 2005

I have a bone to pick . . .

with Boston Globe reporter Michael Kranish, who keeps writing articles about stuff just before I intended to. Fortunately, I'm not a reporter, which means I don't have to be fair and balanced, which means I can tell you the truth.

Actually, in my view, Kranish somewhat conflates two issues. He writes here about peer review of medical journal articles (link only good for 24 hours), but his hook is the two famous, and subsequently discredited studies, that found that Hormone Replacement Therapy and vitamin E reduce the risk of heart disease. Kranish discusses a study by John Ioannidis who found that about 1/3 of widely cited studies in medical journals are found, after subsequent research, to have incorrect or overstated conclusions. He suggests that these may have been flawed or somehow incompetently done, and that the peer review process failed to catch the errors.

I don't think there was necessarily anything wrong with these studies, such as they were. The Ioannidis study actually comes to a very different conclusion: that associations found in ecological or cohort studies, or small-scale randomized trials, don't always hold up after large-scale randomized controlled trials. This is scarcely news, it's epidemiology 101. However, it can be very confusing to the general public.

Although Ioannidis's article in JAMA is available by subscription only, he has written an essay for the Public Library of Science entitled "Why most published research findings are false," which you can read here. And no, it isn't because of faulty peer review.

So in my next two posts, I'm going to take on peer review, and then Ioannidis's provocative thesis -- separately.

No comments: