Abstract
Exposures to antioxidants (AO) are associated with levels of C-reactive protein (CRP), but the pattern of evidence is mixed, due in part to studying each potential AO, one at a time, when multiple AO exposures might affect CRP levels. By studying multiple AO via a composite indicator approach, we estimate the degree to which serum CRP level is associated with serum AO level. Standardised field survey protocols for the US National Health and Nutrition Examination Survey (NHANES) 2003-2006 yielded nationally representative cross-sectional samples of adults aged 20 years and older (n 8841). NHANES latex-enhanced nephelometry quantified serum CRP levels. Liquid chromatography quantified serum concentrations of vitamins A, E and C and carotenoids. Using structural equations, we regressed CRP level on AO levels, and derived a summary estimate for a composite of these potential antioxidants (CPA), with covariates held constant. The association linking CPA with CRP was inverse, stronger for slightly elevated CRP (1·8≤CRP<10 mg/l; slope= -1·08; 95 % CI -1·39, -0·77) and weaker for highly elevated CRP (≥10 mg/l; slope= -0·52; 95 % CI -0·68, -0·35), with little change when covariates were added. Vitamins A and C, as well as lutein+zeaxanthin, were prominent contributors to the composite. In these cross-sectional data studied via a composite indicator approach, the CPA level and the CRP level were inversely related. The stage is set for more confirmatory longitudinal or intervention research on multiple vitamins. The composite indicator approach might be most useful in epidemiology when several exposure constructs are too weakly inter-correlated to be studied via formal measurement models for underlying latent dimensions.