• E-mail
• Print
• Comment
• Font Size
• Digg
• del.icio.us
• Discuss article

The Norwegian Women and Cancer Study (NOWAC): Different Statistical Methods to Assess the Reproducibility of a Food Frequency Questionnaire

Posted on: Tuesday, 21 December 2004, 03:00 CST

The Norwegian Women and Cancer Study (NOWAC): Different Statistical Methods to Assess the Reproducibility of a Food Frequency Questionnaire. C. L. Parr,* M. B. Veierd,* P. Laake,* E. Lund,[dagger] and A. Hjartker.* * Institute of Basic Medical Sciences, Department of Statistics, University of Oslo, Norway, and [dagger] Institute of Community Medicine, University of Troms, Norway.

AIMS: Our aim was to assess the test-retest reproducibility of a food frequency questionnaire (FFQ) used in the Norwegian Women and Cancer Study (NOWAC) by applying statistical methods that are less established in traditional validation studies of dietary assessment methods

. In epidemiological studies of diet and cancer, FFQs are often used to measure dietary intake, but the data can be subject to substantial measurement error, giving biased risk estimates. This may explain some of the conflicting research results on diet-cancer associations. Measures of reproducibility can provide a useful first approximation of the quality of the dietary exposure measurements. However, different statistical methods may give different impressions of the degree of reproducibility. METHODS: NOWAC is a national population-based cohort study with 102,443 women enrolled at age 30-70 y from 1991 to 1997 to investigate risk factors for breast cancer, including diet. In this methodological substudy, the same FFQ was administered twice to a random sample of 2000 women from the cohort in 2002 at age 46-75 y. The FFQ was designed to assess habitual diet over the past year with special emphasis on fish consumption and covered 4 pages within a larger health and lifestyle questionnaire. RESULTS: The response rate was 79% for the test and 75% for the retest. Preliminary results for nutrients show that Pearson's correlation coefficient r range from 0.53 for cholesterol to 0.77 for alcohol. The analysis will continue during spring-summer 2004. Results for both nutrients and foods, including alternative reproducibility measures (e.g., intraclass correlation coefficient and graphical presentations), will be presented. Copyright American Institute of Nutrition Dec 2004

Source: Journal of Nutrition, The

More News in this Category