Methods. Comparisons between methods were made on 156 women, studied over 1 year at 3-monthly intervals at home. On each of four occasions, volunteers completed 4 days of weighed records and provided two 24-hour urine collections and a fasting blood sample.

Results. In comparison with the 16 days of weighed records, a food frequency questionnaire (FFQ) yielded higher values mainly due to greater reported consumption of milk and of vegetables, A 24-hour recall was as good as the FFQ in placing individuals in the distribution of habitual diet from weighed records. Results obtained from a 7-day estimated record were closest to those obtained from the weighed record. Correlations between 24-hour urine excretion and dietary N intake from weighed records were high (0.78-0.87) as were those with estimated food diaries (0.60-0.70). Correlations between urine N and the FFQ and 24-hour recall were lower (0.10 to 0.27), but improved by energy adjustment using residuals for N and K which are correlated with total energy intake. Comparisons between dietary estimates and urinary K and serum carotenoids and vitamin C showed broadly similar results. Limited biomarker information amongst 200 UK EPIC participants supported the findings of the validation study.

Conclusions. UK EPIC uses three methods (the 7-day diary, an improved FFQ, and the 24-hour recall) to assess diet. 93% of first food diaries are returned completed by participants. Repeated diaries are the main dietary assessment method for nested case-control analyses.