| Summary: | Background: Nutrient profiling (NP) aims to identify healthier food options according to the content of selected ‘positive’ nutrients e.g. fibre, protein, and ‘negative’ nutrients e.g. sodium, saturated fat. The British and French food safety agencies developed the WXYfm and SAIN,LIM models, respectively. Their predictive validity in relation to chronic disease has yet to be demonstrated. Aim: To test the hypothesis that ‘healthy’ diets as defined by NP have predictive validity. Methods: Between 1991-93, 7,251 participants of the Whitehall II study completed a 127-item food frequency questionnaire (FFQ). WXYfm and SAIN,LIM scores for each FFQ-item were used to derive energy-weighted aggregate diet scores (AS) for each participant and NP model. Validity was assessed against baseline factors including dietary quality indices. Prospective associations were examined with incident CHD, diabetes and cancer, and all-cause mortality (318, 754, 251, and 524 events, respectively—median follow-up time was approximately 17 years). Results: AS were weakly associated with dietary quality indices. Cox modelling identified U-shaped associations (p quadratic trend <.05) between both AS and all outcomes except diabetes. Participants with middle AS had slightly reduced risk; SAIN,LIM estimates were significant for CHD and all-cause mortality. Dietary misreporting, particularly of energy-dense foods with high ‘negative’ nutrient content, was associated with BMI, hypertension and other risk factors, and explained much of the unexpected U-shaped AS-outcome associations. Alternative AS less sensitive to dietary misreporting confirmed the potential of NP as a public health tool. In particular, the WXYfm ‘positive’ nutrients predicted risk reduction for all outcomes. Conclusions: Predictive validity of the NP approach was partly established. The prospective effects of AS on chronic disease outcomes were confounded by the association between vascular risk and energy misreporting. Further predictive validity studies of NP methods ideally require food-based dietary assessment (e.g. diet diaries, 24h recalls) with less reporting bias.
|
|---|
