Harveian Oration 2008

Editor – Sir Michael Rawlins' fascinating review of decision making in the use or approval of therapeutic interventions (Clin Med December 2008 pp 579–88) mentions the belief that the results of the GREAT trial of GP home thrombolysis were ‘too good to be true’. Oddly, suspicions of biological implausibity are employed most often by statisticians, rather than doctors, to discount statistical hypotheses. Here, a delay of one or two hours immediately after an infarct could easily halve the benefit of thrombolysis, as was in fact observed in the study. To explain this away, an entirely imaginary prior scenario was introduced to ‘pull back’ the results to a more acceptable range. This example discredits rather than supports the use of Bayesian analysis. It reminds one of the bad old days when ‘We set out to prove (or disprove) …’ was tolerated as a preamble to a paper.

• © 2009 Royal College of Physicians