For matters of practical necessity, nutritional data in most epidemiologic studies is obtained from self-administered food frequency questionnaires (FFQ) requiring estimates of long-term average food consumption. Because of its central role epidemiologic research the ability of the FFQ to correctly characterize diet is essential to our ability to assess the role of diet in human health. We have hypothesized that some of the relatively large unexplained variability in FFQ observed in validation studies may be due to response set biases, especially social desirability (SD) and social approval(SA). Previous work showed a large downward bias in reported intake due to social desirability (SD) among women that is consistent both with theoretical constructs that we outline and with results from many validation tests of various FFQ. In the Womens Health Initiative (WHI), we have timely access to a major research study that will help to assess the nature and extent of this bias. A total of 1350 women from 6 clinical centers of the WHI, representing a wide range o geographical regions and ethnic/racial groups will be enrolled into this study. In addition to a baseline FFQ collected a art of the WHI, this study will assess SD collect four days of 24-hour diet recalls (24HR) in a 2-week period, and administer a second FFQ. A subset of 84 women will take part in a doubly labeled water (DLW) substudy to asses energy intake over the 2-week period.
The aims of this study are to: 1. test for bias in comparing fat scores derived from the WHI-FFQ and those from multiple 24HR overall and for Hispanics, Blacks, and Whites separately; 2. test for bias in the 24HR in estimating total energy intake using DLW, and 3. develop methods to correct for biases observed and to identify settings where improvements in precision can be expected. Statistical procedures will utilize primarily linear regression methods. Statistical power is adequate to detect overall differences about 20% as large as those observed earlier and therefore will allow for examination of subgroup differences and for developing and testing methods of adjustment. The overall implications of this work include: 1. reduced false negatives in dietary screening, 2. reduce sample sizes or increased statistical power in research studies, and 3. improved accuracy and precision in epidemiologic studies of diet and health utilizing the FFQ.
| Ma, Yunsheng; Olendzki, Barbara C; Pagoto, Sherry L et al. (2009) Number of 24-hour diet recalls needed to estimate energy intake. Ann Epidemiol 19:553-9 |
| Adams, Swann Arp; Matthews, Charles E; Ebbeling, Cara B et al. (2005) The effect of social desirability and social approval on self-reports of physical activity. Am J Epidemiol 161:389-98 |
