Adjusting retinol-binding protein concentrations for inflammation: Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia (BRINDA) project

Abstract

Background: The accurate estimation of the prevalence of vitamin A deficiency (VAD) is important in planning and implementing interventions. Retinol-binding protein (RBP) is often used in population surveys to measure vitamin A status, but its interpretation is challenging in settings where inflammation is common because RBP concentrations decrease during the acute-phase response.Objectives: We aimed to assess the relation between RBP concentrations and inflammation and malaria in preschool children (PSC) (age range: 6-59 mo) and women of reproductive age (WRA) (age range: 15-49 y) and to investigate adjustment algorithms to account for these effects.Design: Cross-sectional data from 8 surveys for PSC (n = 8803) and 4 surveys for WRA (n = 4191) from the Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia (BRINDA) project were analyzed individually and combined with the use of a meta-analysis. Several approaches were explored to adjust RBP concentrations in PSC in inflammation and malaria settings as follows: 1) the exclusion of subjects with C-reactive protein (CRP) concentrations >5 mg/L or α-1-acid glycoprotein (AGP) concentrations >1 g/L, 2) the application of arithmetic correction factors, and 3) the use of a regression correction approach. The impact of adjustment on the estimated prevalence of VAD (defined as <0.7 μmol/L) was examined in PSC.Results: The relation between estimated VAD and CRP and AGP deciles followed a linear pattern in PSC. In women, the correlations between RBP and CRP and AGP were too weak to justify adjustments for inflammation. Depending on the approach used to adjust for inflammation (CRP+AGP), the estimated prevalence of VAD decreased by a median of 11-18 percentage points in PSC compared with unadjusted values. There was no added effect of adjusting for malaria on the estimated VAD after adjusting for CRP and AGP.Conclusions: The use of regression correction (derived from internal data), which accounts for the severity of inflammation, to estimate the prevalence of VAD in PSC in regions with inflammation and malaria is supported by the analysis of the BRINDA data. These findings contribute to the evidence on adjusting for inflammation when estimating VAD with the use of RBP.

Keywords: anemia; inflammation; meta-analysis; nutritional assessment; retinol-binding protein; vitamin A deficiency.

FIGURE 1

Estimated prevalence [percentage (95% CI)] of vitamin A deficiency in preschool children by CRP (A) and AGP (B) deciles: the Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia (BRINDA) project. The analysis was restricted to surveys (Bangladesh, Cameroon, Côte d’Ivoire, Kenya 2007, Kenya 2010, Liberia, Philippines, and Papua New Guinea) that measured both CRP and AGP for comparability between CRP and AGP relations with biomarkers (n = 8803). Vitamin A deficiency was defined as an RBP concentration <0.70 μmol/L. Bold vertical lines indicate commonly used cutoffs for CRP and AGP. AGP, α-1-acid glycoprotein; CRP, C-reactive protein; RBP, retinol-binding protein.

FIGURE 2

Estimated prevalence [percentage (95% CI)] of vitamin A insufficiency in women of reproductive age by CRP (A) and AGP (B) deciles: the Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia (BRINDA) project. The analysis was restricted to countries (Cameroon, Côte d’Ivoire, Liberia, and Papua New Guinea) (n = 4191).Vitamin A insufficiency was defined as an RBP concentration <1.05 μmol/L. Bold vertical lines indicate commonly used cutoffs for CRP and AGP. AGP, α-1-acid glycoprotein; CRP, C-reactive protein; RBP, retinol-binding protein.

FIGURE 3

Estimated prevalence [percentage (95% CI)] of vitamin A deficiency on the basis of IRC-CRP, IRC-AGP, or IRC-CRP+AGP in preschool children: the Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia (BRINDA) project. Vitamin A deficiency prevalence was defined as retinol binding protein <0.7 mmol/L. Bars without a common lowercase letter within a given survey differ, P < 0.05 (adjusted by using Bonferroni correction). Internal regression correction reference values were as follows: C-reactive protein: −2.26 ln(mg/L) [QE (df = 10) = 439.90, P < 0.0001]; α-1-acid glycoprotein: −0.52 ln(g/L) [QE (df = 10) = 584.5546, P < 0.0001]. IRC-AGP, internal regression correction adjusting for α-1-acid glycoprotein; IRC-CRP, internal regression correction adjusting for C-reactive protein; IRC-CRP+AGP, internal regression correction adjusting for C-reactive protein and α-1-acid glycoprotein; QE, QE test of residual heterogeneity.

FIGURE 4

Estimated prevalence [percentage (95% CI)] of vitamin A deficiency with the use of different inflammation-adjustment approaches adjusting for CRP and AGP in preschool children: the Biomarkers Reflecting Inflammation and Nutrition Determinants of Anemia (BRINDA) project. Vitamin A deficiency was defined as retinol binding protein <0.7 μmol/L. Bars without a common lowercase letter within a given survey differ, P < 0.05 (adjusted by using Bonferroni correction). BCFs were as follows: incubation phase: 1.22 (95% CI: 1.16, 1.28); early convalescence phase: 1.38 (95% CI: 1.31, 1.44); and late convalescence phase: 1.09 (95% CI: 1.06, 1.11) [QE (df = 28) = 261.6964, P < 0.0001). IRC and BRC reference values were as follows: ln CRP = −2.26 ln(mg/L) [QE (df = 10) = 439.90, P < 0.0001]; ln AGP = −0.52 ln(g/L) [QE (df = 10) = 584.5546, P < 0.0001]. BRC coefficients were as follows: ln CRP = −0.06; and ln AGP = −0.09 [QE (df = 21) = 240.4563, P < 0.0001]. AGP, α-1-acid glycoprotein; BCF, Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia correction factor; BRC, Biomarkers Reflecting Inflammation and Nutritional Determinants of Anemia regression correction; CRP, C-reactive protein; ICF, internal correction factor; IRC, internal regression correction; QE, QE test of residual heterogeneity.

FIGURE 5

Estimated prevalence [percentage (95% CI)] of vitamin A deficiency in preschool children with the use of different malaria-adjustment approaches. Vitamin A deficiency was defined as retinol-binding protein <0.7 μmol/L. Bars without a common lowercase letter within a given survey differ, P < 0.05 (adjusted by using Bonferroni correction). Internal regression correction reference values were as follows: ln C-reactive protein = −2.26 ln(mg/L) [QE (df = 10) = 439.90, P < 0.0001]; and ln α-1-acid glycoprotein = −0.52 ln(g/L) [QE (df = 10) = 584.5546, P < 0.0001]. IRC-CRP+AGP, internal regression correction adjusting for C-reactive protein and α-1-acid glycoprotein; IRC-CRP+AGP+malaria, internal regression correction adjusting for C-reactive protein, α-1-acid glycoprotein, and malaria; IRC-malaria, internal regression correction adjusting for malaria; QE, QE test of residual heterogeneity.