Astrology, aspirin and your heart
One of many things to warm to from Richard Peto's appearance on The Life Scientific (the programme radio with the title odd)...
...was a re-telling of his argument with The Lancet when publishing findings about the beneficial effects of aspirin and other drugs in heart attack patients. Huge trial, uncontrovertible results, practical implications: great stuff:
During the first 5 weeks there were 804 (9.4%) vascular deaths in the aspirin-allocated group compared with 1016 (11.8%) in the placebo control group.
The benefit is modest, but clear and if followed everywhere, as it was, could save hundreds of thousands of lives worldwide.
The editors or reviewers responded with a request to analyse sub-samples of the data set, to see, for example, if the patients who benefitted were all or mostly male, or old, or middle-class, or...
That seems plausible, but to go back and re-interrogate the data for additional results is bound to show something, although this may be a mere consequence of doing lots of analysis (the problem of multiple testing), and Peto was opposed in principle to this sort of data dredging.
In the end, Peto and his colleagues agreed to perform and publish the requested analysis, but first, they performed and published an analysis by star sign:
subdivision of the patients in ISIS-2 with respect to their astrological birth signs appears to indicate that for patients born under Gemini or Libra there was a slightly adverse effect of aspirin on mortality (9% SD 13 increase; NS), while for patients born under all other astrological signs there was a strikingly beneficial effect (28% SD 5 reduction; 2p < 0-00001).
As they say, if you torture the data long enough, they'll confess to anything!
But if the results were significant for some biological or social sub-division that fitted our assumptions, would we be so quick to laugh?
Full story in readable format: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(15)61505-7/fulltext
Original article: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(88)92833-4/fulltext