National Cancer Institute Home at the National Institutes of Health |
Please wait while this form is being loaded....
The Applied Research Program Web site is no longer maintained. ARP's former staff have moved to the new Healthcare Delivery Research Program, the Behavioral Research Program, or the Epidemiology & Genetics Research Program, and the content from this Web site is being moved to one of those sites as appropriate. Please update your links and bookmarks!

Comparative Validation of the Block, Willett, & NCI Food Frequency Questionnaires

Summary: The 1997-1998 Eating at America's Table Study (EATS) validated a new NCI-developed food frequency questionnaire (FFQ), the Diet History Questionnaire (DHQ), and compared its performance to that of two other FFQs widely used at the time, the 1995 Block FFQ and the purple version of the Willett FFQ. The DHQ performed as well as or better than the other two FFQs. Results from this study also point the way to future research in exploring the structure of measurement error in current methods of assessing dietary intake.

Reference: Subar AF, Thompson FE, Kipnis V, Midthune D, Hurwitz P, McNutt S, McIntosh A, Rosenfeld S. Comparative Validation of the Block, Willett, and National Cancer Institute Food Frequency Questionnaires. Am J Epidemiol 2001;154(12):1089-99.

The most practical and economical way to collect comprehensive dietary data in large epidemiologic studies is the food frequency questionnaire (FFQ). FFQs are designed to capture individuals' usual dietary intake by asking respondents to report their frequency of consumption of a list of foods over a specific period of time. Efforts to improve the accuracy of FFQs are essential to improving our ability to measure dietary intake patterns and thereby to better understand of the role of diet in the etiology and prevention of chronic diseases such as cancer.

The Eating at America's Table Study (EATS) was designed to validate the Diet History Questionnaire (DHQ), a new and improved FFQ developed by the National Cancer Institute, and to compare it against two other widely used FFQs -- the 1995 NCI-Block Health Habits and History Questionnaire and the Willett (purple version) FFQ. Of the 1,640 eligible individuals recruited for the study, 1,301 participants first completed four 24-hour dietary recalls over the course of a year. Following this phase, participants were randomized into two groups, one of which completed the DHQ and the Block FFQ. The other group completed the DHQ and the Willett FFQ. The DHQ was validated by assessing the correlations and attenuation coefficients between nutrient intakes estimated on the DHQ and "true" intakes estimated from the four 24-hour recalls using a measurement error model. The investigators also determined correlations between the Block and Willet FFQs and these "true" intakes.

Results showed that absolute nutrient intakes for women estimated by the DHQ and Block FFQs were fairly comparable to the 24-hour recalls; the Willett instrument tended to overestimate nutrient intake for women. All three instruments underestimated absolute nutrient intake for men, compared with the recalls, though the Willett instrument underestimated to a greater degree than did the other two FFQs. Correlation and attenuation coefficients were somewhat higher for the DHQ than for the Block FFQ, and both were better than the Willett FFQ in models unadjusted for energy. Energy adjustment increased correlation and attenuation coefficients dramatically for the Willett FFQ and modestly for the other two FFQs, suggesting that energy adjustment, in general, reduces measurement error in all FFQs. The lack of portion size information in the Willett FFQ may partially explain the differences in absolute intakes among the three FFQs.

To better understand the structure of measurement error in FFQs, the authors urge further research in varied populations with biomarker reference instruments that are not based on self-report. At the same time, EATS has shown that this improved, cognitively-based FFQ is as good as or better than two FFQs that were in wide use at the time and is a reasonable alternative for investigators to use in a data collection domain that is still primarily dependent on self-reported information.

<< Previous Summary | Next Summary >>

Last Modified: 11 Apr 2014