Are the findings reliable?

There is little to suggest they are. Tom Sanders, head of nutritional research at King's College London, says that the strain of rat the French team used gets breast tumours easily, especially when given unlimited food, or maize contaminated by a common fungus that causes hormone imbalance, or just allowed to age. There were no data on food intake or tests for fungus in the maize, so we don't know whether this was a factor.

But didn't the treated rats get sicker than the untreated rats?

Some did, but that's not the full story. It wasn't that rats fed GM maize or herbicide got tumours, and the control rats did not. Five of the 20 control rats – 25 per cent – got tumours and died, while 60 per cent in "some test groups" that ate GM maize died. Some other test groups, however, were healthier than the controls.

Toxicologists do a standard mathematical test, called the standard deviation, on such data to see whether the difference is what you might expect from random variation, or can be considered significant. The French team did not present these tests in their paper. They used a complicated and unconventional analysis that Sanders calls "a statistical fishing trip".

Anthony Trewavas of the University of Edinburgh, UK, adds that in any case, there should be at least as many controls as test rats – there were only 20 of the former and 80 of the latter – to show how variably tumours appear. Without those additional controls, "these results are of no value", he says.