Replicating Medical Studies

The NIH, tired of paying for studies that are trumpeted in major journals but then can't be replicated, is considering a requirement that all grant recipients pay somebody to try to reproduce their studies before publication. I think this is a great idea. It would be expensive, but the NIH has a budget of $29 billion and I say better fewer studies that we know are right than more uncertainty. This is, after all, a huge problem:

In biomedical science, at least one thing is apparently reproducible: a steady stream of studies that show the irreproducibility of many important experiments. In a 2011 internal survey, pharmaceutical firm Bayer HealthCare of Leverkusen, Germany, was unable to validate the relevant preclinical research for almost two-thirds of 67 in-house projects. Then, in 2012, scientists at Amgen, a drug company based in Thousand Oaks, California, reported their failure to replicate 89% of the findings from 53 landmark cancer papers.

If reproducing all experiments is not practical, the NIH should certainly insist that early-stage studies of new treatments be independently reproduced before proceeding to expensive clinical trials. This seems so obvious to me that I was shocked to discover it is not already NIH policy.