Development and Validation of a Food Frequency Questionnaire in Spain

Abstract

To create a suitable instrument to estimate intakes of total calories, protein, carbohydrate, fats (saturated, mono and polyunsaturated), alcohol, cholesterol, fibre, vitamin A and vitamin C in epidemiological studies conducted in Spain, a food frequency questionnaire was developed and tested. In particular, the questionnaire was designed to be used in a large population-based case-control study of dietary factors in relation to breast and colorectal cancer among women from different Spanish regions. After identifying the most important food sources of the relevant nutrients in the study population, the final version of the questionnaire asked about consumption of 118 food items. Its reproducibility and validity were tested among 147 Spanish women aged 18–74 years. These subjects were asked to complete the questionnaire before and after completing four 4-day food records. The records were obtained at 3-month intervals designed to represent daily and seasonal changes (between 1990 and 1991). Using the information available from standard Spanish food composition tables, an ad hoc computer program was created to translate food consumption into nutrient intake. The reproducibility of the questionnaire was assessed by means of estimating correlations between nutrient scores measured with the same instrument twice, with a period of 1 year between estimates. Pearson correlation coefficients ranged from 0.51 for saturated fat to 0.88 for alcohol. In the validity study, correlation coefficients between diet records and the first and second questionnaires ranged between r=0.20 for vitamin A and r=0.88 for alcohol. Simple adjustment for total energy intake did not increase these estimates, but the energy-adjusted correlations improved after considering the attenuation derived from within-person variation. The de-attenuated correlation coefficients between the second questionnaire and diet records ranged from 0.45 for vitamin A to 0.91 for alcohol. With regard to gross misclassification, on the average 3% of subjects classified in the highest or lowest quintile by food record were assigned to the lowest or highest quintile by the food frequency questionnaire. These results indicate that our Spanish questionnaire is fairly reproducible and provides a potentially reliable scale for categorizing individuals by level of past nutrient intake.