The ratio plasma retinol-binding protein (RBP) : transthyretin (TTR) has been proposed as a means to improve the assessment of vitamin A status of individuals with concurrent infection or inflammation. We have measured RBP and TTR in stored sera from South African children who had accidentally ingested kerosene. Samples were collected from these children in hospital when suffering acute inflammation and respiratory distress, and from them and neighbourhood control children 3 months later. Vitamin A status was defined by modified relative dose response (MRDR) tests of liver retinol stores at 3 months and by serum retinol concentration both when children were ill and when they were well. Illness was defined as either being in hospital or, at follow-up, as having a raised plasma α1-acid glycoprotein (AGP) level. The RBP : TTR value was significantly decreased by both illness and low liver retinol stores. When the effects on RBP : TTR of illness and vitamin A stores were considered together for the 3-month follow-up samples, only vitamin A status significantly decreased the value. We calculated sensitivity and specificity of the RBP : TTR ratio against established measures of vitamin A status using a cut-off value of 0·3 for RBP : TTR and standard cut-off values for MRDR (0·06) and plasma retinol (0·7 μmol/l). Compared with MRDR, RBP : TTR had sensitivities of 76 % and 43 % and specificities of 22 % and 81 % to detect vitamin A deficiency in hospitalized and well children respectively. Compared with plasma retinol, sensitivities were 88 % and 44 % and specificities were 55 % and 64 % in hospitalized and well children respectively. Only for the case of clinically well children with biochemical evidence of subclinical inflammation did sensitivity (62 % and 100 % against MRDR and plasma retinol respectively) and specificity (100 % and 60 % against MRDR and retinol) approach useful levels for an assessment tool. Overall, although a trend supporting the theory behind the use of the RBP : TTR for assessment of vitamin A status in infection was observed in the current study, the ratio did not provide adequate sensitivity and specificity to be a useful assessment tool.