Richard Smith over at The BMJ has a cute article that asks whether the time has come to assume medical research is frauduent until proven otherwise. I say “cute” because everyone paying attention knows that most published medical research is trash. A year ago, Irreal reported that fewer than 25% of articles in medical journals concerning cancer research were reproducible. Considering what’s at stake, those figures are very disturbing.
Smith’s article takes this a step further by considering not irreproducibility but outright fraud: Studies where the data was faked or where the study didn’t take place at all. According to Smith, 20% of the studies are fraudulent or what he calls Zombie studies, where the data are so suspect that the whole study is called into question.
This is fraud of course, and, moreover, fraud that all of us are paying for with our tax dollars. But, in a sense, that doesn’t matter. What matters is that people are suffering and dying because medical practitioners are relying on these studies.
The problem, as it usually is, is the inappropriate incentives that researchers labor under. They get promoted (and tenured) not for advancing knowledge but for publishing papers. When your career depends on the number of papers you publish, you do everything you can to publish papers, even if the results or worthless and even if the results are flat out fraudulent.
It’s hard to know what to do about this. One of the commenters suggest seeking a refund of grant money from authors of fraudulent studies. That and public shaming (and possibly dismissal from their universities or laboratories) would be a good start. The real problem, though, is how to detect fraudulent studies to begin with. Obviously the current system isn’t doing that.