There is a lot of general evidence about the effect of diet on our overall health, but when you start to look at the details, everything becomes more vague. Very few randomized prospective trials lasting for more than a year have been done, and breaking down the lessons they provide is not easy. What is more problematic, however, is combining evidence from different studies into general dietary guidelines. A quick review of past and present recommendations shows the difficulty; for example, recommendations about the amount and types of fats and carbohydrates for our diet have varied quite a bit. These are the problems faced by people using the mixtures of available evidence to make national recommendations about diet. An article released today in JAMA, the Journal of the American Medical Association stretches the evidence even further. They attempt to quantify the number of deaths that could be avoided by changing ten different dietary habits. Even with the use of “a comparative risk assessment modeling design” to estimate mortality, I don’t find these types of calculations informative or believable. Just because your are using the latest data and the best available techniques does not mean that the output will be reliable. I worry that articles like this presenting inaccurate predictions about saving hundreds of thousands of lives will undermine the impact of other reports that are more reasoned and less speculative.