C-reactive protein mediates the association between fruit and vegetable intake, and mortality among US adults

Abstract

Background Fruit and vegetable (FV) are part of a healthy diet as their consumption has been consistently associated with a reduction in the risk of premature mortality and various biomarkers of inflammation have been hypothesized to partially explain such relationships. However, empirical evidence quantifying the mediating role of such biomarkers in the relationship between FV intake and mortality is not available at the population level. The present study examines the association between FV daily servings or objectively measured biomarkers of FV intake and mortality and the extent to which a biomarker of inflammation mediates these associations. Methods We used data on 13,892 adults aged ≥ 20 years enrolled in the NHANES-III between 1988 and 1994 and followed up through Dec 31, 2015. Daily serving of FV was assessed using 24-hour dietary recall and mortality was determined from the National Death Index records. Serum concentrations of 6 biomarkers of FV intake and C-reactive protein (CRP) were measured in blood samples. We used a causal mediation analysis with Aalen additive hazards models to decompose the total effect of FV intake on mortality into natural direct and indirect (through CRP level) effects while controlling for a priori-defined confounders. We also calculated the proportion mediated. Results Over the follow-up period, 5023 participants died (1503 from cardiovascular diseases, and 1110 from cancer). High daily FV serving (at least 5 servings) and high level of FV biomarker (> median) were associated with a lower risk of death compared to low daily servings and low level of FV intake biomarker, respectively. Comparing participants with high scores to low scores of the FV intake biomarker, we observed reduced hazard rate difference for all-cause (-0.54 per 10000 person-year [95% CI -0.72 to -0.33]), cardiovascular disease (-0.19 per 10000 person-year [95% CI -0.26 to -0.12]), and cancer (-0.10 per 10000 person-year [95% CI -0.14 to -0.06]) mortality that can be attributed to subsequent decreases in CRP. The corresponding proportions mediated were 16%, 17%, and 10% for all-cause, cardiovascular, and cancer mortality, respectively. Conclusion Our findings confirmed the beneficial effects of FV intake and that the effects were partly mediated by a reduction in circulating CRP.