Monday, 26 November 2012

Risk Reduction in the Jungle

I was recently watching 'I'm a Celebrity Get me out of here' (under duress from Mrs B, he lamely claims!!)

During the interval there was a  phone-in competition with the usual asinine question and an amazing prize of a fully paid trip to Queensland. The SMS cost is daylight robbery-like. However, seduced by the amazing prize (and a few glasses of vino rosso), I entered the competition. I immediately got a text back from ITV inviting me to resubmit the answer for a second time at a reduced fee and a 100% increase in the chances of winning the competition. This little episode reminded me of the whole debate about how treatment effects should be expressed when the results of randomised control trials are published - i.e as a relative risk reduction or as absolute risk reduction. I have always felt that using relative risk reduction/change, i.e. the method used in the text from ITV stating that a second text increases my chance of winning by 100%, is, although factually correct,  very misleading to the uninitiated. In absolute risk reduction terms the same change could be expressed as an increase in winning from 1 in 3 million to 2 in 3 million or as a decrease/fall in the chances of NOT winning from 2,999,999 out of 3 million to 2,999,998 out of 3 million i.e. an infinitesimally small decrease. This surely should be the way results are expressed when trials are published. The fact that science is increasingly being reported in the popular press is good news but exaggerating findings by using relative risk reductions as often happens does a great disservice to science journalism.
When it comes to winning phone-in competitions or the lottery, just remember that however many tickets you have, you chances of winning will always remain infinitesimally small!