Observational studies have shown a strong positive correlation between the severity of anemia and the risk of poor outcomes in patients with chronic kidney disease (CKD). This observation was initially taken to imply that adverse outcomes in CKD are caused by anemia. However, the assumption of causality ignores the possibility that anemia and adverse outcomes might be unrelated and that both are caused by underlying inflammation, oxidative stress and comorbid conditions. Randomized clinical trials of anemia correction have revealed an increased risk of adverse cardiovascular outcomes in patients assigned to normal, rather than subnormal, hemoglobin targets. As a result, correction of anemia is now considered potentially hazardous in patients with CKD. Notably, individuals who did not reach the target hemoglobin level in the clinical trials, despite receiving high doses of erythropoietin and iron, experienced a disproportionately large share of the adverse outcomes. These observations point to overdose of erythropoietin and iron, rather than anemia correction per se, as the likely culprit. This Review explores the reasons for the apparent contradiction between the findings of observational studies and randomized clinical trials of anemia treatment in CKD. I have focused on data from basic and translational studies, which are often overlooked in the design and interpretation of clinical studies and in the formulation of clinical guidelines.