Cancer patients are frequently anemic. Treatment of anemic patients with erythropoiesis-stimulating proteins (ESPs) such as epoetin and darbepoetin is associated with benefits that include a reduced transfusion risk and improved quality of life. The recent reports of two randomized trials in which ESP treatment was associated with a decreased survival raised valid concerns regarding the safety of these agents in oncology practice. Reports of erythropoietin receptors on non-hematologic human tumor cells have increased the level of concern and provided a relatively simple model for the effects of ESPs on tumor progression and resistance to treatment. This article reviews available data, which lead to a number of conclusions: 1) the two trials suggesting a negative impact on survival have serious methodologic issues that may compromise interpretation; 2) when used to treat rather than prevent anemia in cancer patients, ESPs show no significant negative impact on survival outcomes; 3) with the exception of erythroleukemia cell lines, the presence of functional erythropoietin receptors on human tumor cells has not been conclusively shown; and 4) a sound theoretical basis exists, supported by preclinical evidence, that any effect of ESP therapy on tumor outcomes may depend on baseline hemoglobin levels, with different effects when anemic and non-anemic individuals are treated. For the present, it is prudent to withhold ESP therapy unless hemoglobin concentrations fall below 12 g/dL and to titrate treatment to maintain a target of 12 g/dL, with adjustments in therapy to insure that levels do not exceed 13 g/dL.