March 1998, Volume 52, Issue 3, Pages 186 – 192

Original Communication

Effect of additional questions about fat on the validity of fat estimates from a food frequency questionnaire A Wolk1,,7, H Ljung1, B Vessby2, D Hunter3,4, WC Willett3,4,5 & the Study Group of MRS SWEA6 1Department of Medical Epidemiology, Karolinska Institute, Stockholm, Sweden     2Department of Geriatrics, University Hospital, Uppsala, Sweden     3Department of Epidemiology, Harvard School of Public Health, Boston, MA, USA     4Channing Laboratory, Harvard Medical School, Boston, MA, USA     5Department of Nutrition, Harvard School of Public Health, Boston, MA, USA     6The Study Group: H-O Adami, L Bergkvist, Å Bruce, L Holmberg, J Leppert, S Ljunghall, G Nilsson, I Ringqvist and S Sörensen     7Correspondence: Dr A Wolk, Department of Medical Epidemiology, Karolinska Institute, S-171 77 Stockholm, Sweden.

Keywords

fats;   questionnaire;   adipose tissue;   validity

Abstract

Objective: We studied whether the validity of fat estimates from food frequency questionnaires (FFQ) can be increased by using in nutrient calculation an additional qualitative information about the type of fat and reduced consumption of visible fat and skin. Design: A random sample of women answered an 88-item self-administered FFQ and performed 4×1-week weighed dietary records (DR). Setting: Uppsala County in central Sweden. Subjects: One hundred and eighty-four women aged 30-77 y, with FFQ and complete DR; 73 women with subcutaneous adipose tissue (AT) samples. Methods: Fat intake from the FFQ was calculated with/without use of qualitative information and compared to DR and fat composition of AT. Main outcome measures: Estimates of long-time intake of total fat, saturated, monounsaturated, polyunsaturated fat and ten specific fatty acids based on FFQ, DR and composition of AT. Results: Mean absolute fat intake estimates based on FFQ (without vs with use of additional fat information) were 21.2 vs 20.2 g/d for saturated, 17.1 vs 16.0 g/d for monounsaturated and 7.3 vs 7.3 g/d for polyunsaturated fat. The Pearson correlation coefficient between the FFQ and AT for polyunsaturated fat was 0.65 vs 0.67. Corresponding correlation between the FFQ and DR was 0.40 vs 0.41; adjustment for energy intake increased this correlation from 0.40 to 0.52. Conclusions: The increase in the validity of fat estimates due to use of qualitative information about fat was negligible; energy adjustment had greater impact than asking additional questions. Sponsorship: The study was supported by grants from the Swedish Cancer Society.

Received 13 May 1997; Revised 15 October 1997; Accepted 24 October 1997

^© Macmillan Publishers Ltd 1997