Accurate estimation of the prevalence of vitamin A deficiency (VAD) is challenging because serum retinol concentrations (SR) decrease during an acute phase response (APR). The objective was to investigate different adjustment algorithms to correct SR during inflammation among preschool children (PSC).
As part of the Biomarkers Reflecting Inflammation and Nutrition Determinants of Anemia (BRINDA) initiative, cross-sectional data from PSC living in Colombia (n=3794), Mexico (n=2512), Nicaragua (n=1420) and Pakistan (n=7318) were adjusted for inflammation. As markers of the APR, C-reactive protein (CRP) concentrations were available from Mexico and Columbia, and a-1-acid glycoprotein (AGP) concentrations from Nicaragua and Pakistan.
In general, making no adjustments to SR in the presence of inflammation over-estimated the prevalence of VAD (retinol<0.7µmol/L). In Colombia, Mexico, Nicaragua and Pakistan: using correction factors (CFs) derived by Thurnham et al decreased the prevalence of VAD by 2.8, 3.3, 1.0 and 3.5 percentage points respectively compared to unadjusted prevalences; using BRINDA derived CFs changed the prevalence of VAD by 2.5, 3.0, -0.4 and 0.1 percentage points; finally using BRINDA derived regression co-efficients (RC) decreased VAD by 3.3, 3.0, 0.2 and 0.8 percentage points. VAD prevalence was comparable in Mexico (16%) and Colombia (24%), where the adjusted change in prevalence of VAD using CFs or RC was similar. VAD prevalence was lowest in Nicaragua (2%), and highest in Pakistan (52%), and the change in adjusted VAD prevalence was variable by algorithm.
In all countries, only one acute phase protein was measured which may not capture the full effect of the APR.