Evaluation of the two non-consecutive 24-h recall instrument for pan-European food consumption surveys

Background: The comparability of food consumption data originating from national nutritional surveys in Europe is currently hampered because of different methodologies used. Therefore, experts in the European Food Consumption Survey Method (EFCOSUM) consortium proposed to use two non-consecutive 24-h recalls for standardised dietary monitoring in European countries.
Aim: Within the European Food Consumption Validation (EFCOVAL) consortium (www.efcoval.eu), this thesis aimed to evaluate the data collected with two non-consecutive 24-h recalls using EPIC-Soft for comparisons of dietary intake in adults between countries in future pan-European surveys.
Methods: To evaluate the bias in protein and potassium intake as well as the ranking of individuals according to their fish and fruit & vegetable intake collected with two non-consecutive 24-h recalls, we developed a validation study within EFCOVAL. The study included biomarker data of 600 subjects from five European centres in Belgium, the Czech Republic, France, the Netherlands, and Norway. To gain further insight into the determinants of the accuracy of the method by using multilevel analysis, we combined EFCOVAL data from one day with similar data from twelve other centres participating in the European Prospective Investigation into Cancer and Nutrition (EPIC) calibration study. Then, we used the EFCOVAL data for assessing the impact of different modes of administration (telephone vs. face-to-face), recall days (1st vs. 2nd) and days of the week (weekdays vs. weekend) on the bias in protein and potassium intake. Finally, data from the Netherlands was used to explore the usefulness of collecting individual dietary data with the 24-h recalls for estimating dietary exposure to flavouring substances.
Results: On average, men and women underreported protein intake by 8% in the EFCOVAL study. Underreporting of potassium intake was 7% in men and 4% in women. The coefficient of variation of bias in observed protein and potassium intake between centres ranged from 4 to 7%. The prevalence of subjects with adequate protein and potassium intake according to the observed data at the lower and upper end of the usual intake distribution agreed fairly well (<10% difference) with the prevalence according to the excretion data. The results of the multilevel analysis indicated that the bias in observed protein intake for both genders and in potassium intake for women did not vary across centres and to a certain extent varied in potassium intake for men (coefficient of variation=9.5%). One of the factors mostly influencing the different performance of the method across European populations was BMI. Furthermore, two standardised 24-h recalls and a food propensity questionnaire appeared to be appropriate to rank individuals according to their fish and fruit & vegetable intake in a comparable manner between the five European centres. Moreover, we observed that in some centres protein intake reported by face-to-face interviews at the study site was less accurate than by telephone interviews, and that second 24-h recall assessments were less accurate than first recalls. In addition, in one out of five centres, protein intake estimated during weekends and potassium intake estimated during weekdays were less accurate than during other days of the week. Finally, the collection of detailed food consumption data at the individual level may be necessary to assess the dietary exposure to flavourings and adaptations of the databases used in EPIC-Soft software can provide more detailed information on the dietary exposure to the flavouring raspberry ketone than non-modified databases.
Conclusion: Two non-consecutive 24-h recalls using EPIC-Soft provides sufficiently valid and suitable data for comparing dietary intake across European populations.