by Ben Goldacre
I am a doctor and a data geek. I worry that data geeks are too easily seduced by the glamour of laboratory science and forget about clinics. Randomised controlled trials are the best tool we have in medicine for finding out if a treatment works or not. Lots of trials are done. Unfortunately, the results of these trials can go missing in action after they are completed.
Missing data is always a challenge: but we also know that “negative results” are more likely to go missing. This means we have a biased sample, overestimating the benefits of treatments. To prevent all this happening, people have set up registers of trial protocols, to be completed before trials begin. These have not been correctly used, and they are not matched to published trials, which show up what data has been left unpublished.
I will describe a small project to fix this, illustrate how that can lead on to fixing other similar problems in medicine, and make a cry for help.
28th February to 1st March 2012