Development and testing of an algorithm to predict fruit and vegetable consumption

A high intake of fruit and vegetables (F&V) may reduce the risk of disease, particularly heart disease. However, it is important that F&V intake is measured accurately when trying to work out whether increasing consumption of these foods will reduce the risk of disease. It is currently very difficult to measure F&V intake accurately, as standard methods used are imprecise. The introduction of a blood test to measure F&V consumption would be a significant development in nutrition research as it would allow researchers to make better estimates of F&V intake in the general population, and also lead to a better understanding of whether and how F&V reduce disease risk. We propose that, because F&V are diverse and complex foods with many biologically active compounds, that a single blood test is never likely to accurately reflect F&V consumption. As an alternative, we propose that the results of measuring a panel of compounds from a single blood test could be combined using statistical methods, and that this approach would lead to a more accurate estimate of F&V consumption. We will carry out a study where we feed healthy volunteers a known number of portions of F&V per day and will carry out these blood tests, to allow the testing of the statistical methods. We will then test the results of these statistical analyses in a series of F&V studies that have already been carried out where we know how many portions of F&V the participants were consuming in order to compare reported intake with the more objective blood results.

Single biomarkers of fruit and vegetable (F&V) consumption, such as vitamin C and certain carotenoids and flavonoids, have been proposed but associations between these compounds and F&V intake are relatively weak. We propose that, given the complexity of F&V, production of an algorithm based on a wide variety of bioactive compounds found in F&V (including those outlined above), might better predict F&V consumption. Predominant flavonoids and polyphenols will be identified through food analysis of a wide variety of F&V. A controlled feeding study (30 volunteers randomised to 2, 5 or 8 portions of F&V/day for 4 weeks) will allow development of the algorithm, and this will be tested in a number of F&V intervention studies (approximately 1300 blood/intake comparisons) and a cross-sectional population study (n=300). This project will therefore generate a biomarker of F&V consumption using blood-based analyses of a panel of bioactive compounds.