Bad Science

...with apologies to Ben Goldacre.
Today's hot news is the announcement by researchers in Thailand of a 'breakthough in the search for an anti HIV vaccine'.
But yet again, the meeja are getting all het up about scientific results they really know nothing about.
To explain, approx 8000 volunteers were given a trial vaccine and another 8000 were given a placebo. Revisiting them some years later they found about 70 of the placebo group were infected with HIV while only around 50 of the vaccinated group had become infected. And now they're claiming this as a 30% reduction in HIV infection rates.
Guff!
That is such a small sample of people (not the 16,000 volunteers, but the number of people who became infected) that it's hardly worth analysing! And to suggest that you can easily equate this to a 30% drop is just statistically daft. If you did the same trial again with another 2 groups of 8000 people but this time gave both groups a placebo and such a small percentage of people ultimately became infected, it's completely possible that could result in another 50/70 infection rate in the 2 groups again, purely by chance.
Aside from the rather dodgy sample size and their definition of 'statistically significant' being rather different than mine, personally I wouldn't like 30% odds, and secondly, there is more than one strain of HIV, this is not necessarily going to be a one-size-fits-all. Finally, it's a virus, it's out there in the population, it's mutating all the time, it's a moving target.
Sheesh. There is no magic bullet.

Lolcat of the day