Research
Reproducibility and validity of food intake measurements from a semiquantitative food frequency questionnaire

https://doi.org/10.1016/0002-8223(93)91754-EGet rights and content

Abstract

Few food frequency questionnaires have been evaluated for their ability to assess intakes of individual foods that may be related to disease independently of their nutrient content. The reproducibility and validity of food intake measurements by a 131-item semiquantitative food frequency questionnaire were evaluated in a sample of 127 men from the Health Professionals Follow-up Study, a large longitudinal study of diet and disease. Each subject completed two questionnaires 1 year apart and two 1-week diet records 6 months apart during the intervening year. Pearson correlations assessing reproducibility between food intakes from the two questionnaires ranged from .31 for pie to .92 for coffee (mean = .59). Validity was measured by comparing food intakes from the second questionnaire with those from the diet records. Pearson correlations corrected for within-person weekly variation in diet record data ranged from .17 for other nuts to .95 for bananas (mean = .63). Large within-person variation precluded the calculation of accurate validity correlations for 29 foods. As we previously observed in women, the foods most often overreported were fruits and vegetables, and meats and dairy products were most often underreported. With few exceptions, reasonable levels of reproducibility and validity were observed for intake of individual foods in this extensive food frequency questionnaire.

Section snippets

MATERIALS AND METHODS

Men in this study were participants in the Health Professionals Follow-up Study, a prospective investigation of risk factors for heart disease and cancer among men aged 40 to 75 years at baseline in 1986. An initial FFQ (questionnaire 1) was mailed to all 51,529 members of the cohort. The questionnaire was mailed a second time (questionnaire 2) after a 1-year interval to the 157 men, from a random sample of 323 Boston-area cohort members, who agreed to participate in this validation study.

RESULTS

Mean daily intakes, expressed in serving-size units, for the 122 food items measured by FFQs 1 and 2 and by the two 1-week diet records for the 127 study participants are presented in Table 1. Food intakes were more likely to be overestimated than underestimated by the questionnaires, and the degree of overestimation was greater than the degree of underestimation. On questionnaire 2, 39 foods were estimated within 20% of the diet-record measurements; 47 foods were overestimated by more than 50%

DISCUSSION

In the small number of studies that have examined the reproducibility of intake measurements for specific food items on dietary questionnaires, correlation coefficients in the range of .4 to .7 are most common 9., 10., 11.. Other studies have used percent agreement to assess reproducibility 12., 13., but these are more difficult to compare because better agreement can be achieved by questionnaires with fewer response categories. Few validity studies of food intake measurements by dietary

APPLICATIONS

These analyses demonstrate that a comprehensive, semiquantitative FFQ can provide relative measures of food consumption that are reproducible and valid for most foods. As validated, the FFQ is most appropriate for use in studies that examine relationships between diet and risk of disease. Further evaluation is required before such questionnaires can be recommended for routine dietary assessment of individuals in clinical practice.

Though these findings apply only to the male health professionals

References (21)

There are more references available in the full text version of this article.

Cited by (928)

View all citing articles on Scopus

Supported by research grants HL 35464 and CA 55075 from the National Institutes of Health.

e

affiliated with all of the aforementioned sites.

View full text