Objective: To assess the relative validity of two diet assessment methods, an extensive quantitative food frequency questionnaire (method A) and a novel shorter quantitative food frequency questionnaire with a 14 day food record (method B).
Design: A randomized prospective cohort study.
Setting: General community.
Subjects: 206 residents of the town of Malmö, aged between 50-69 years, 101 men and 105 women who completed the methods during one year.
Methods: Both diet methods were designed to cover the whole diet and portion sizes were estimated using a booklet with 120 photographs; method A comprised 250 items and method B combined a two-week food record measuring lunch and dinner meals and a shorter 130 item quantitative food frequency questionnaire for average consumption of foods, snacks and beverages during the past year. An 18 day dietary record comprising six 3-day weighed records evenly distributed over one year served as a reference method.
Results: Pearson's correlation coefficients varied from 0.25 for fat intake to 0.84 for milk products for method A and from 0.32 for fish to 0.88 for meat for method B. Correlations for most food groups ranged between 0.50-0.80, and were higher for method B. Only small changes were noted after adjustment for energy intake. On average for most food groups categorization of subjects into quartiles, 55% of subjects belonging to the lowest quartile, and 57-59% of those belonging to the highest quartile for method A and B were correctly classified.
Conclusion: A combined food record with a quantitative food frequency questionnaire is a better tool for food assessment than an extensive food frequency questionnaire.