A new analysis of TREAT shows that baseline characteristics and postrandomization factors do not explain the increased incidence of stroke in patients assigned to darbepoetin alfa. Baseline characteristics that predicted the risk of stroke in the overall TREAT population did not modify the risk of stroke associated with darbepoetin alfa in interaction analyses. Similarly, postrandomization factors, including blood pressure, hemoglobin levels, platelet count and treatment dose, did not explain the increased risk of darbepoetin alfa-related stroke.