Frequency of food use data and the multidimensionality of diet.

Cancer epidemiologists rely heavily on the frequency of food use technique to assess dietary risk factors. We found that the accepted procedures for arraying individuals along a continuum, based on their positions within distributions of intakes of specific nutrients, simultaneously array them on other dietary characteristics. A unidimensional approach to dietary assessment could confound cancer risk assessments, and the effects could differ for men and women. We found that men consumed more calories and energy-containing nutrients than women, who were more likely to consume larger amounts of vitamins A and C. Dietary variety was similar for men and women, despite compositional differences in diet. Diets of men contained proportionately more meats, grains and nuts, and alcohol. Diets of women contained more fruits, vegetables, and poultry and fish. Among men, we found stronger intercorrelations among energy intake and intakes of fat, fiber, and vitamin C and among fat intake and intakes of fiber and vitamins A and C. Compositional differences in diet across quartiles of nutrient intake were not consistent for men and women, suggesting that the constellation of dietary risk factors may differ for the two sexes. Men in the lowest quartiles of energy, fiber, and vitamin A intakes had the greatest percent contribution of alcohol to the diet. This pattern was not observed for fat intake levels. Measures of dietary patterns may be needed for more accurate descriptions of the associations of cancer risk and diet.