Why is it a problem? Well one of the reasons is the old “garbage in equals garbage out” when examining why a certain analysis didn’t work. When you put a lot of bad studies together and make it into one bad study, it’s still a bad study. The other issue is publication bias. People don’t get things published that are negative, they get things published that are positive in small numbers. At the time of CLASP, if you went to the Oxford database which tries to bring in all the information on trials that were going to start but never got published, there were as many patients in unpublished studies as there had been in the positive trials that had been published up until that time. So it looked like that was at least part of the explanation.