Thirty-four asymptomatic children, ages 5-35 mo, were studied to compare the conjunctival impression cytology technique with the relative dose response test in detection of subclinical vitamin A deficiency. Conjunctival smears were collected from the infero-temporal-bulbar conjunctiva of each eye with a strip of cellulose acetate filter paper and transferred onto a glass slide. Venous blood was drawn at 0 and 5 h after administration of an oral dose of 1000 micrograms of retinol palmitate (relative dose response test). An increase in serum retinol concentration (> or = 20%) in the 5-h value was considered indicative of an inadequate liver store of vitamin A and hence subclinical vitamin A deficiency. Of the 34 children, 26 (76.5%) had moderate to severe protein-energy malnutrition. Only three children (9%) had abnormal conjunctival impression cytology, whereas 23 (68%) had abnormal relative dose response. Even more striking was the finding that only two of the 23 children with abnormal relative dose response had abnormal conjunctival impression cytology. The results suggest that the conjunctival impression cytology test has poor agreement with the relative dose response test results in assessing vitamin A status in young children. If relative dose response is considered an acceptable reference method for assessing vitamin A status, then the conjunctival impression cytology test cannot be considered a valid measure of subclinical vitamin A deficiency in this population.