The relationship of anemia as a risk factor for child mortality was analyzed by using cross-sectional, longitudinal and case-control studies, and randomized trials. Five methods of estimation were adopted: 1) the proportion of child deaths attributable to anemia; 2) the proportion of anemic children who die in hospital studies; 3) the population-attributable risk of child mortality due to anemia; 4) survival analyses of mortality in anemic children; and 5) cause-specific anemia-related child mortality. Most of the data available were hospital based. For children aged 0-5 y the percentage of deaths due to anemia was comparable for reports from highly malarious areas in Africa (Sierra Leone 11.2%, Zaire 12.2%, Kenya 14.3%). Ten values available for hemoglobin values <50 g/L showed a variation in case fatality from 2 to 29.3%. The data suggested little if any dose-response relating increasing hemoglobin level (whether by mean value or selected cut-off values) with decreasing mortality. Although mortality was increased in anemic children with hemoglobin <50 g/L, the evidence for increased risk with less severe anemia was inconclusive. The wide variation for mortality with hemoglobin <50 g/L is related to methodological variation and places severe limits on causal inference; in view of this, it is premature to generate projections on population-attributable risk. A preliminary survival analysis of an infant cohort from Malawi indicated that if the hemoglobin decreases by 10 g/L at age 6 mo, the risk of dying becomes 1.72 times higher. Evidence from a number of studies suggests that mortality due to malarial severe anemia is greater than that due to iron-deficiency anemia. Data are scarce on anemia and child mortality from non-malarious regions. Primary prevention of iron-deficiency anemia and malaria in young children could have substantive effects on reducing child mortality from severe anemia in children living in malarious areas.