Complex correlation structure in consumption rates of major food groups: implications for the assessment of radiation exposure

Abstract

AbstractObjective:An extensive dataset on individual food consumption was analysed in order to study all pairwise correlations between the consumption rates of 11 major food groups. Additionally, the project aimed to examine and quantify the accuracy of a recently proposed estimator of total food consumption to be used for the estimation of radiation exposure by food. Such an inquiry seems justified, because the proposed estimator implicitly presumes an essentially positive correlation structure in food consumption rates.Design:Statistical analysis using representative data gathered in Germany in a nation-wide food consumption survey.Setting:Germany.Subjects:Individuals aged between 4 and 94 years namely 10901 males and 12308 females.Results:The consumption rates of 11 major food categories showed several significantly positive, but also a number of significantly negative, correlations. Negative associations between cereal and potato products persisted consistently over all age groups, independent of sex. Other significantly negative correlations were limited to certain age groups. Reflecting these negative correlations, a subsequent analysis of relative ranks of consumption revealed that no person in the sample had the highest consumption rates in all food groups simultaneously. Based on representative samples, overestimations of 34 to 53% were obtained if – as recently suggested in the context of radiation exposure prediction – the 95th percentiles of total food consumption were determined as sums of the corresponding percentiles of the food groups.Conclusions:The complex correlation structure of food group consumption rates, as identified in this study, bears important implications for various health-related issues. Ignoring them could lead to overly conservative estimations of radiation exposure due to food ingestion or to confounding effects in epidemiological studies on nutritional risk factors of diseases. The results also indicate that a distinction into different dietary patterns might be useful in characterising different consumption habits.