Abstract 4368203: A validated metabolite-based biomarker score for fruit and vegetable intake and associations with all-cause mortality and incident cardiometabolic diseases

Background: Epidemiological evidence underlying beneficial associations of fruit and vegetable intake with all-cause mortality and cardiovascular diseases largely relies on subjective dietary assessment. Research question: To identify a circulating biomarker score predictive of fruit and vegetable intake, perform validation in a feeding trial, and assess associations with all-cause mortality and incident cardiometabolic diseases. Methods and results: We included 3,979 EPIC-Norfolk participants, aged 40-79y, with plasma carotenoids and metabolites, measured by reversed-phase high-performance liquid chromatography and ultra-high performance liquid chromatography-tandem mass spectrometry respectively, and 7-day diet diaries collected at baseline (1993-1998). Using elastic net regression with bootstrapping followed by ridge regression, we identified a metabolite score comprising 81 metabolites and 5 carotenoids, robustly correlated with self-reported fruit and vegetable intake ( r =0.51), explaining 44% of the variance, with discriminative ability across quintiles. We selected 17 metabolites with available chemical standards plus the 5 carotenoids, explaining 33% of the variance, for validation in a randomised controlled cross-over trial of 39 pre-hypertensive participants (NCT03410342). Participants consumed standardized diets low or high in fruit and vegetable types for two weeks, separated by a washout week. Mixed-effect regression analysis showed that the metabolite score significantly discriminated between low and high fruit and vegetable diets ( P <6x10 -5 ). Top 5 metabolites positively contributing to the metabolite score were erythritol, methyl glucopyranoside (alpha+beta), threonate, s-methylcysteine, and dopamine 3-O-sulfate. After a median follow-up ranging 16-20y and adjustment for confounders, each 1SD higher metabolite score was inversely associated with type 2 diabetes (HR: 0.80; 95% CI: 0.69-0.93) and renal disease (HR: 0.85; 0.75-0.96), with borderline associations for all-cause mortality (HR: 0.95; 0.90-1.00), and cardiovascular diseases (HR: 0.94; 0.98-1.00. Associations of self-reported fruit and vegetable intake with outcomes were comparable to those using the metabolite score, with similar or wider 95% confidence intervals. Conclusion: These findings indicate the potential to use objective indicators of the physiological response to fruit and vegetable intake, and to elucidate mechanisms of association with disease.