Reforming Nutritional Epidemiologic Research

To the Editor The Viewpoint by Dr Ioannidis discussed the need for reform of nutritional epidemiologic research. We were surprised by the “nonexpert” calculations of gain or loss in life expectancy using data from our meta-analysis. We presented the findings in our meta-analysis as relative risks (RRs). Ioannidis calculated from those RRs the number of years of life gained or lost by consumption of a number of foods. Such calculations are not in line with statistical methodology and should not be used in a scientific debate about diet-disease associations and potential implications. An RR of 2 does not mean a shorter life expectancy of 40 years or an RR of 4 a shorter life expectancy of 60 years. An RR of 0.5 does not generate a life expectancy of 120 years or an RR of 0.85 a life expectancy of 92 years, a 12-year increase in life expectancy for 1 serving of nuts daily as calculated by Ioannidis. In the article, more misleading calculations of this type were done using other examples from our publication, but we would like to point out that heavy smoking, with an RR of 2.38 to 2.66, would result in a life expectancy of only about 35 years (assuming a life expectancy of 80 years), or a loss of life-years of 45 years according to the calculations of Ioannidis.