Errors in the interpretation of dietary assessments

Two years ago, I reviewed the analytic effect of error in the estimation of dietary intake, describing the emphasis on the "random" day-to-day variation in reported intake. Interest in this area is increasing and there are signs of progress in analytic strategies. This paper focuses on two concerns about the use of dietary data in analyses. The effect of different methods of adjusting analyses of fat and a health outcome for energy is illustrated through an exploration of the association between fat intake and high body mass index in data sets from the US Department of Agriculture and a Dutch national survey. Both a shift in the analytic question and a change in the error structure occur as analysis strategies are changed, leading to confusion in interpretation. The paper also addresses the growing concern about bias in the estimation of intake and the possibility that differential bias moves with stratification variables of analytic interest. The increasing use of doubly labeled water estimates of energy expenditure as a gold standard for checking on overall bias in reporting is commendable. There will always be error in dietary assessments. The challenge is to understand, estimate, and make use of the error structure during analysis.