Relative validity of a food frequency questionnaire used in the Inter99 study

To evaluate the validity of the Inter99 food frequency questionnaire (FFQ) compared with a 28-days' diet history and biomarkers.

A random sample of 13 016 individuals were drawn from a general population and invited for a health screening programme. Participation rate was 52.5%. All high-risk individuals were re-invited for assessment after 1 and 3 years and completed a 198-item FFQ at all three occasions. Participants attending for 3 years follow-up were invited to participate in the validation study, including a 28-days' diet history, a 24-h urine collection and a fasting blood sample. Overall, 264 subjects participated.

Spearman's rank correlation coefficients between the two dietary methods ranged from 0.31(beta-carotene) to 0.64 (fruits) in men and from 0.31 (polyunsaturated fat and sodium) to 0.64 (fruits) for women. The proportion of individuals classified in the same or adjacent quintiles were, on average, 72% for men and 69% for women. Gross misclassification was found on average in 2%. The correlation coefficients of the residuals ranged from 0.27 (sodium) to 0.61 (fruits) for men and from 0.21 (sodium) to 0.62 (B12-vitamin) for women. Correlation coefficients between fruit and vegetable intake and carotenoids ranged from -0.08 (lycopene) to 0.44 (alpha-carotene). For the residuals the correlation coefficients ranged from -0.004 (lycopene) to 0.47 (alpha-carotene).

The Inter99 FFQ and the residuals of the intake provide acceptable classification of individuals according to their dietary intakes and the FFQ gives a good quantitative measurement of key dietary components.