A 24-hour food list for dietary assessment in large-scale epidemiological studies

Abstract

The validity of dietary assessment in large-scale cohort studies has been questioned. Combining different instruments for the assessment of consumption probability and amounts consumed might be feasible and improve the estimation of usual dietary intake in such studies. Thus, the objectives were (a) to develop a web-based 24-hour food list (24 h FL) for Germany to assess the consumption probability of foods during the previous 24 hours, (b) to evaluate the performance of the new questionnaire in a feasibility study, (c) to identify determinants of consumption-day amounts in order to derive person-specific standard consumption-day amounts and (d) to evaluate their relevance for the estimation of usual dietary intake distributions.<br /> Data from the German National Nutrition Survey II (NVS II) was used to develop a finite list of food items for the 24 h FL applying stepwise linear regression analysis. In addition, NVS II data was analyzed for determinants of consumption-day amounts across 22 food groups. A total of 508 individuals participating in the pilot study for the German National Cohort (GNC) were invited to fill in the 24 h FL via Internet up to three times during a three to six month period. In addition, GNC pilot study participants were asked to evaluate the questionnaire using a brief online evaluation form. Finally, usual intake distributions were compared that were either derived by using a parsimonious or a comprehensive prediction model.<br /> In total, 246 items were identified for the item list of the 24 h FL, reflecting more than 75% of variation in intake of 27 nutrients and four major food groups. Among individuals of the GNC pilot study invited, 64% participated in the feasibility study. Of these, 100%, 85%, and 68% of participants completed the 24 h FL one, two, or three times, respectively. The average time needed to complete the questionnaire was nine minutes and its acceptability in terms of understandability, usability, completeness and visual presentation was rated as high. Relevant determinants for consumption-day amounts of food groups were sex, age, body mass index (BMI), smoking status, years of education, household net income, living with a partner and employment status. However, the use of a comprehensive prediction model (sex, age, BMI, smoking status, years of education, household net income) compared to a parsimonious prediction model (sex, age, BMI) seemed to be important for less frequently consumed foods only.<br /> The 24 h FL represents a promising new dietary assessment tool, which can be employed as part of a blended approach combining multiple data sources for estimation of usual dietary intake in large-scale cohort studies.