Ben Goldacre
Thursday December 2, 2004
The Guardian

· I love Christmas. All those festive news stories planted by industry about how chocolate and booze are good for you, and then the excitement of a season of new year bad science detox stories to look forward to. “According to the British Nutrition Foundation, most adults tend to put on at least 5-7Ib over the four-week Christmas period.” So says the PR department of Phase 2, a white kidney bean extract that’s “the low-carb diet in a pill”. It’s a great idea: an amylase inhibitor that stops your body breaking down carbohydrates, so you can’t absorb them and you presumably poo them out. Whole chips perhaps.

· It’s not a particularly new idea. In fact, in 1983, Professor John Garrow and some colleagues read about a similar product, and came up with an ingenious way of testing it. We eat plants and then break down their carbohydrates, release their energy for our own dubious purposes and breathe out carbon dioxide. There are two isotopes of carbon: carbon-13 and carbon-12. Some plants have a higher amount of carbon-13 in their sugars than others, so if you eat the carbohydrates from them, you soon breathe out carbon dioxide that has more carbon-13 than usual. To measure carbon-13 against carbon-12 you need a mass spectrometer: the joy of being a scientist is that you might have a friend who has one lying around on the day you feel like causing some trouble. They gave five obese women starch meals containing lots of carbon-13, along with a placebo tablet, or Starchex, the amylase inhibitor tablet of the day. The result? Placebo or amylase inhibitor, it made no difference, they still breathed out the same amount of carbon-13; they also had the same blood sugar changes and insulin levels after their meals. As far as you can tell, the pill made no difference to starch absorption.

· But wait: Phase 2 boasts “eight separate clinical trials”. I could find one properly published, in the Alternative Medicine Review. The Phase 2 group lost 3.79lb, and the placebo group lost 1.65lb, on average: but the difference was only statistically significant at p=0.35. That’s not even worth printing, because it means (and bear with me professional statisticians, I only have 400 words to make my point here) that there was a one in three chance that the finding was completely by chance. One in 20 is the traditional cut-off for people to take a finding seriously. And regardless of statistical significance, I leave it to you to decide if a difference of 2lb has any practical significance. Keep eating those chips.