Anemia protocols for hemodialysis patients usually titrate erythropoietin (ESA) according to hemoglobin and iron according to a threshold of ferritin, with variable response seen. A universally optimum threshold for ferritin may be incorrect, and another view is that ESA and iron are alternative anemia treatments, which should be selected based on the likely response to each. Hemodialysis patients developing moderate anemia were randomised to treatment with either an increase in ESA or a course of intravenous iron. Over 2423 patient-months in 197 patients, there were 133 anemia episodes with randomized treatment. Treatment failure was seen in 20/66 patients treated with ESA and 20/67 patients treated with iron (30.3 vs. 29.9%, p = 1.0). Successful ESA treatment was associated with lower C-reactive protein (13.5 vs. 28.6 mg/L, p = 0.038) and lower previous ESA dose (6621 vs. 9273 μg/week, p = 0.097). Successful iron treatment was associated with lower reticulocyte hemoglobin (33.8 vs. 35.5 pg, p = 0.047), lower hepcidin (91.4 vs. 131.0 μg/ml, p = 0.021), and higher C-reactive protein (29.5 vs. 12.6 mg/L, p = 0.085). A four-variable iron preference score was developed to indicate the more favorable treatment, which in a retrospective analysis reduced treatment failure to 17%. Increased ESA and iron are equally effective, though treatment failure occurs in almost 30%. Baseline variables including hepcidin can predict treatment response, and a four-variable score shows promise in allowing directed treatment with improved response rates.