Adjusting iron and vitamin A status in settings of inflammation: a sensitivity analysis of the Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia (BRINDA) approach

Abstract

ABSTRACTBackgroundAccurate assessment of iron and vitamin A status is needed to inform public health decisions, but most population-level iron and vitamin A biomarkers are independently influenced by inflammation.ObjectivesWe aimed to assess the reproducibility of the Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia (BRINDA) regression approach to adjust iron [ferritin, soluble transferrin receptor (sTfR)] and vitamin A [retinol-binding protein (RBP), retinol] biomarkers for inflammation (α-1-acid glycoprotein and C-reactive protein).MethodsWe conducted a sensitivity analysis comparing unadjusted and adjusted estimates of iron and vitamin A deficiency using the internal-survey regression approach from BRINDA phase 1 (16 surveys in children, 10 surveys in women) and 13 additional surveys for children and women (BRINDA phase 2).ResultsThe relations between inflammation and iron or vitamin A biomarkers were statistically significant except for vitamin A biomarkers in women. Heterogeneity of the regression coefficients across surveys was high. Among children, internal-survey adjustments increased the estimated prevalence of depleted iron stores (ferritin <12 µg/L) by a median of 11 percentage points (pp) (24 pp and 9 pp in BRINDA phase 1 and phase 2, respectively), whereas estimates of iron-deficient erythropoiesis (sTfR >8.3 mg/L) decreased by a median of 15 pp (15 pp and 20 pp in BRINDA phase 1 and phase 2, respectively). Vitamin A deficiency (RBP <0.7 µmol/L or retinol <0.7 µmol/L) decreased by a median of 14 pp (18 pp and 8 pp in BRINDA phase 1 and phase 2, respectively) in children. Adjustment for inflammation in women resulted in smaller differences in estimated iron deficiency than in children.ConclusionsOur findings are consistent with previous BRINDA conclusions that not accounting for inflammation may result in an underestimation of iron deficiency and overestimation of vitamin A deficiency. Research is needed to understand the etiology of the heterogeneity in the regression coefficients before a meta-analyzed regression correction can be considered.