To improve estimates of vitamin A deficiency in children of pre-school age in the 2006 Uganda Demographic and Health Survey (UDHS 2006).

A cross-sectional study in which dried blood spot samples were analysed for C-reactive protein (CRP). Retinol-binding protein (RBP) had previously been analysed using a commercial enzyme immunoassay.

A population-based study in Uganda.

A systematically selected subset of the dried blood spot samples collected from children aged 6–59 months for UDHS 2006. Children were categorized into ‘normal CRP’ (Group A) and ‘raised CRP’ (Group B) using a CRP cut-off of 5 mg/l. A correction factor was calculated to adjust the Group B RBP values for the influence of the acute-phase response.

Geometric mean CRP was 6·2 (95 % CI 5·5, 7·0) mg/l, 1·6 (95 % CI 1·5, 1·8) mg/l and 17·9 (95 % CI 16·4, 19·6) mg/l in all children, in Group A and in Group B, respectively. Geometric mean RBP in all children, in Group A and in Group B was 1·18 (95 % CI 1·14, 1·22) μmol/l, 1·26 (95 % CI 1·20, 1·33) μmol/l and 1·12 (95 % CI 1·07, 1·17) μmol/l, respectively, before correction. Correction increased mean RBP in Group B to 1·26 (95 % CI 1·21, 1·31) μmol/l. The prevalence of vitamin A deficiency (RBP < 0·825 μmol/l) reduced from 18·4 % (95 % CI 17·2, 23·0 %) to 13·9 % (95 % CI 11·3, 16·5 %).

Correcting for the acute-phase response significantly reduced the prevalence of vitamin A deficiency; thus, the acute-phase response should be considered when vitamin A status is assessed using RBP in order to improve population-level estimates of vitamin A deficiency.