The modified-relative-dose-response (MRDR) test and the relative-dose-response (RDR) test were compared in 49 mildly to moderately malnourished Bangladeshi children. The MRDR test had a significantly lower sensitivity, detecting only 71% of children with very low serum retinol (< or = 0.35 mumol/L) and 33% of children with low serum retinol (0.355-0.70 mumol/L) compared with 100% and 80% for the RDR test, respectively. The MRDR test showed a very strong dependency on retinol-binding protein (RBP) saturation (ie, percent saturation of RBP with retinol) compared with the RDR test. Only 3 (23%) of 13 children with RBP saturation > or = 55% but low vitamin A stores were diagnosed as abnormal by the MRDR test. This suggests that when apo-RBP concentration is limiting, as it is in malnourished children, didehydroretinol, the analog used in the MRDR test cannot effectively compete with retinol for binding to apo-RBP. Under these circumstances, the MRDR test is rendered ineffective. The possibility of increasing the sensitivity of the test by using a high dose of didehydroretinol needs to be investigated.