Magnitude, determinants and impact of under-reporting of energy intake in a cohort study in Greece.

Abstract

To investigate the frequency and the determinants of under-reporting in a semi-quantitative food frequency questionnaire used in the Greek segment of the European Prospective Study on Nutrition, Cancer and Health (EPIC study). A food frequency questionnaire was completed by 9262 adult men and women. The questions included in this questionnaire covered the average intake of approximately 150 food items and beverages over 1 year. Evaluation of under-reporting was conducted on an individual basis taking into account the expected daily variation of nutritional intakes during the time period of recording. Individuals whose energy intake was lower than 1.14*BMR (basal metabolic rate) were defined as under-reporters. Urban and rural population of Greece. The data indicated underestimation of energy intake by 11.8% of individuals enrolled. Results from a logistic regression model indicated that body mass index (BMI), gender, age and educational level were significant predictors of under-reporting. The proportion of overweight participants (BMI >30) who tend to under-report energy intake was more than twice that of normal-weight individuals. Men were significantly more prone to under-reporting compared to women, while low education individuals under-report more often than others. Exclusion of under-reporters generated, as expected, mean nutrient values that were significantly higher (by about 6%) than those derived from the total number of participants. When the nutrient values were energy-adjusted, however, or were expressed as percentages of energy intake for macronutrients or as nutrient densities for micronutrients, the emerging differences were minimal and generally statistically not significant. Under-reporting does exist and it is more extensive among men, those with low education levels and the overweight participants. Adjustment for energy intake minimizes the bias generated by under-reporting with respect to particular nutrients and their association with various disease outcomes in the cohort.