Results of studies in which biological markers have been used as the reference method for assessment of dietary intake for selected nutrients suggest that the degree of error associated with food-frequency questionnaires (FFQs) is considerably larger than previously estimated. This could explain the lack of an association between increased fat intake and breast cancer in population studies.
Sheila Bingham from the MRC Dunn Human Nutrition Unit, Cambridge, UK, and Cambridge University colleagues assessed the relation between breast-cancer risk and fat intake with an FFQ similar to those used in previous population (cohort) studies and a 7-day food diary completed by women in the European Prospective Investigation of Cancer and Nutrition (EPIC) Norfolk study. Around 13,000 women were studied using a FFQ and seven-day food diary between 1993 and 1997. 168 women had developed breast cancer by the year 2000, at which time analysis of fat intake was assessed for each woman with breast cancer compared with four healthy controls matched for age and other factors to take account of possible bias.
Women in the upper quintile (top 20%) for saturated-fat consumption were at twice the risk of breast cancer than women in the lowest quintile when fat intake was assessed by the food diary; however no association was evident between increased saturated fat intake and breast-cancer risk with use of the FFQ
Sheila Bingham comments: "Inconsistency between experimental and epidemiological data on fat and breast-cancer risk could thus be accounted for by problems with methods used in cohort studies to measure diet. The food diary is more expensive to code for conversion into nutrients than the FFQ, but we have shown that its use is acceptable and feasible in large cohort studies."
She concludes: "No biomarkers exist for fat intake, so none of the associations shown can be said to be free of measurement-error effects. However, our preliminary findings suggest that use of the food diary can detect relations between diet and cancer risk within a relatively homogeneous population."
In an accompanying Commentary (p 182), Ross Prentice from the Fred Hutchinson Cancer Research Center, Seattle, concludes: "Bingham and colleagues' report highlights the importance of the methodological issues in nutrition and chronic disease research by providing an example in which two valuable dietary self-reporting instruments give qualitatively different results.
Their report will help to move the debate about measurement error in dietary assessment, which has been ongoing for several years, into a more practical arena. However, it will be useful to compare disease-risk associations with food diary and food frequency assessments of fat intake in additional cohort studies, as the Norfolk-EPIC study is based on a modest number of cases of breast cancer-and especially because the estimated relative risks across fat intake quintiles (whether based on food diaries or food frequency questionnaires) seem larger than would be expected from the international correlational analyses that provided much of the stimulus for the fat and breast cancer hypothesis. More generally, the reliability and interpretation of cohort-study data on the controversial topic of an association between dietary fat and breast cancer will unfortunately remain unclear until further objective information becomes available about the measurement properties of the dietary assessment methods used."
Issue 19 July 2003 (pp 182, 212)
Embargoed 0001 h (London time) 18 July 2003. In North America, the embargo time for Lancet press material is 6:30pm ET Thursday 17 July 2003.
Contact: MRC Press Office, Park Crescent, London W1, UK; T) +44 (0)20 7637 6011; E) press.office@headoffice.mrc.ac.uk
Steve Palmer, Cancer Research UK Press Office, 42 Lincolns Inn Fields, London EC1, UK; T) +44 (0)20 7061 8312; E)
Stephen.palmer@cancer.org.uk
Dr Ross L Prentice, Division of Public Health Sciences, Fred Hutchinson Cancer Research Center, Seattle, WA 98109-1024, USA; T) +1 206 667 6756; F) +1 206 667 4142; E) rprentic@WHI.org
Journal
The Lancet