Standardization of the 24-hour diet recall calibration method used in the European prospective investigation into cancer and nutrition (EPIC): General concepts and preliminary results

Objectives: Despite increasing interest in the concept of calibration in dietary surveys, there is still little experience in the use and standardization of a common reference dietary method, especially in international studies. In this paper, we present the general theoretical framework and the approaches developed to standardize the computer-assisted 24 h diet recall method (EPIC-SOFT) used to collect about 37 000 24-h dietary recall measurements (24-HDR) from the 10 countries participating in the European Prospective Investigation into Cancer and Nutrition (EPIC). In addition, an analysis of variance was performed to examine the level of standardization of EPIC-SOFT across the 90 interviewers involved in the study. Methods: The analysis of variance used a random effects model in which mean energy intake per interviewer was used as the dependent variable, while age, body mass index (BMI), energy requirement, week day, season, special diet, special day, physical activity and the EPIC-SOFT version were used as independent variables. The analysis was performed separately for men and women. Results: The results show no statistical difference between interviewers in all countries for men and five out of eight countries for women, after adjustment for physical activity and the EPIC-SOFT program version used, and the exclusion of one interviewer in Germany (for men), and one in Denmark (for women). These results showed an interviewer effect in certain countries and a significant difference between gender, suggesting an underlying respondent's effect due to the higher under- reporting among women that was consistently observed in EPIC. However, the actual difference between interviewer and country mean energy intakes is about 10%. Furthermore, no statistical differences in mean energy intakes were observed across centres from the same country, except in Italy and Germany for men, and France and Spain for women, where the populations were recruited from areas scattered throughout the countries. Conclusion: Despite these encouraging results and the efforts to standardize the 24-HDR interview method, conscious or unconscious behaviour of respondents and/or interviewer bias cannot be prevented entirely. Further evaluation of the reliability of EPIC-SOFT measurements will be conducted through validation against independent biological markers (nitrogen, potassium).