Abstract
The accurate estimation of vitamin A deficiency (VAD) is critical to informing programmatic and policy decisions that could have important public health implications. However, serum retinol and retinol binding protein (RBP) concentrations, two biomarkers often used to estimate VAD, are temporarily altered during the acute phase response, potentially overestimating the prevalence of VAD in populations with high levels of inflammation. In 22 nationally-representative surveys, we examined (1) the association between C-reactive protein (CRP) or α1-acid glycoprotein (AGP) and retinol or RBP, and (2) how different adjustment approaches for correcting for inflammation compare with one another. In preschool age children (PSC) and school age children (SAC), the association between inflammation and retinol and RBP was largely statistically significant; using the regression approach, adjustments for inflammation decreased the estimated prevalence of VAD compared to unadjusted VAD (range: -22.1 to -6.0 percentage points). In non-pregnant women of reproductive age (WRA), the association between inflammation and vitamin A biomarkers was inconsistent, precluding adjustments for inflammation. The burden of VAD can be overestimated if inflammation is not accounted for, and the regression approach provides a method for adjusting retinol and RBP for inflammation across the full range of concentrations in PSC and SAC.