Serum retinol and hemoglobin levels were determined in 532 children aged 6-60 months living in urban slums of Karachi, Pakistan. Overall 67% (358 of 532) of children had hemoglobin levels of less than 11 g/dl, the World Health Organization definition of anemia for this age group. Estimations for red blood cell indices (hematocrit, mean corpuscular volume, mean corpuscular hemoglobin, mean corpuscular hemoglobin concentration, and red blood cell count) were done on a sample of 391 children to classify anemia morphologically. A similar percentage (69.8% [273 of 391 children]) of anemia was found in this group. The anemia was predominantly microcytic and hypochromic. Normocytic normochromic anemia was present in only 16.5%. Serum retinol levels were significantly correlated with hemoglobin (P < 0.002), hematocrit (P < 0.01), and red blood cell (P < 0.001) levels. However, anemia was found to be a poor predictor (positive predictive value [PPV] = 2.5%) for the presence of vitamin A deficiency (retinol < 10 μg/dl). The PPV increased to 54% if 20 μg/dl, which is an indicator of marginal vitamin A status, was used as the cutoff point. The sensitivity of the hemoglobin test was found to be 75% in correctly identifying vitamin A deficiency when retinol levels of 10 μg/dl and 20 μg/dl were both used as the cutoff points for deficiency. The specificity for the hemoglobin test varied from 33% to 40% when levels of 10 μg/dl and 20 μg/dl were used as the cutoff points for vitamin A deficiency. These results suggest that in planning community health programs, hemoglobin levels cannot be used as an accurate indicator for predicting vitamin A deficiency in Pakistani children.