REPRODUCIBILITY AND RELATIVE VALIDITY OF AN EXTENSIVE SEMIQUANTITATIVE FOOD FREQUENCY QUESTIONNAIRE USING DIETARY RECORDS AND BIOCHEMICAL MARKERS AMONG GREEK SCHOOLTEACHERS

Background. We evaluated the reproducibility and relative validity of a 190-item semi-quantitative food frequency questionnaire (FFQ) to be used in a large prospective study in the Athens area of Greece. Method s. In all, 42 men and 38 women, aged 25-67 years, completed two self-a dministered semi-quantitative FFQ spaced approximately 1 year apart. W ithin this 1-year interval, participants visited the study centre mont hly and completed an interviewer-administered 24-hour diet recall ques tionnaire, We also collected two venous blood and three 24-hour urine samples from participants at randomly selected periods during the year between the two administrations of the dietary questionnaire. Results . Mean values for intake of most nutrients assessed by the two FFQ wer e in good agreement while those assessed by the two different dietary methods were reasonably similar for most nutrients. Intraclass correla tion coefficients for energy-adjusted nutrient intakes assessed by que stionnaires one year apart averaged 0.57 (range 0.24-0.75) and were no t substantially different between genders. Correlation coefficients be tween the energy-adjusted nutrients measured by repeated 24-hour recal ls and the semi-quantitative FFQ ranged from 0.25 for beta-carotene an d polyunsaturated fats to >0.50 for saturated fats, cis-linoleic acid, calcium and phosphorus (average: 0.46 for men and 0.39 for women). Re liability and relative validity were very high for alcohol intake. Die tary intakes from the FFQ were significantly correlated to plasma leve ls of vitamin C, but not beta-carotene or cholesterol. Dietary protein intake assessed by both FFQ was correlated (0.36, 0.30) with average urinary nitrogen excretion levels from three 24-hour urine samples. Co nclusions. These data indicate that the semi-quantitative FFQ is repro ducible and provides a reasonably reliable measure of intake. Signific ant correlations between diet and an independent biochemical marker fu rther corroborate the relative validity of our questionnaire in this G reek population.