Dietary assessment in Whitehall II: the influence of reporting bias on apparent socioeconomic variation in nutrient intakes

OBJECTIVE: To assess socio-economic differences in nutrient intake, giving particular consideration to the influence of reporting bias. DESIGN: Cross-sectional study. Three methods of data analysis (inclusion of all subjects, exclusion of low energy reporters, and regression-based energy adjustment) were evaluated against biomarkers of fatty acid and antioxidant intakes. SETTING: London-based Civil Servants. SUBJECTS: Age and employment grade stratified random sub-sample of 459 men and 406 women aged 39-61 y who completed 7 d diet diaries at Phase 3 follow-up (1991-93) of the Whitehall II Study. DIETARY MEASURES: Mean daily intakes by employment grade (6 levels) of dietary energy, total fat, saturated fatty acids, polyunsaturated fatty acids (PUFA), linoleic acid, carbohydrate excluding fibre, dietary fibre, protein, alcohol, vitamin C, vitamin E, carotenes, potassium and calcium. Biomarkers: serum cholesterol ester fatty acids (CEFA), total cholesterol, plasma alpha-tocopherol and beta-carotene. RESULTS: Low energy reporting (LER), defined as a reported energy intake below 1.2 times calculated basal metabolic rate, was strongly associated with employment grade (top grade: men 17.3%, women 19.3%, bottom grade: men 45.7%, women 49.2% trend P < 0.0001 both sexes). This association is only to a small extent accounted for by the relative weight distribution across grades. The direct associations with grade--high status, high intakes--of total fat, saturated and polyunsaturated fatty acids seen in the results overall were greatly reduced or abolished when LER were excluded or when energy adjustment was used. Direct associations between grade and intakes of vitamin C and potassium in both sexes were evident regardless of data presentation method. Spearman correlations between biomarkers and reported intakes, for example CEFA and dietary linoleate (men 0.46, women 0.61), plasma beta-carotene/cholesterol ratio and dietary carotenes (men 0.16, women 0.21) together with theoretical considerations indicate that energy adjustment may be the preferred method for reducing the influence of reporting bias. CONCLUSIONS: Low energy reporting is a major source of bias in dietary surveys and its prevalence shows a marked inverse association with socio-economic status. The energy adjustment method provides an approach which reduces this bias without exclusion of low energy reporters. Intakes of micronutrients including vitamin C, rather than fatty acids, showed associations with socio-economic status consistent with a dietary explanation for social inequalities in cardiovascular disease.