BACKGROUND: Optimal iron levels in patients on hemodialysis are currently unknown, and a higher level than that for the healthy population is usually set for such patients considering the use of erythropoiesis-stimulating agents or the occurrence of chronic inflammation. However, excessive iron causes oxidative stress and impairment of its utilization by cells. Therefore we investigated the relationship between hemoglobin (Hb) level and iron status in hemodialysis patients to identify the optimal iron levels for patients undergoing hemodialysis. METHODS: A total of 208 outpatients on maintenance hemodialysis were followed up between July 2006 and June 2007. Men accounted for 64.9% cases [mean age, 59.3 ± 13.1 years and median dialysis history, 7.7 (3.6-13.2) years], and diabetic nephropathy accounted for 25.0% cases. Hemoglobin level was measured twice a month and serum ferritin, serum iron, and total iron-binding capacity were measured once a month. The doses of recombinant human erythropoietin and low-dose iron supplement were adjusted to maintain a hemoglobin level of 10-11 g/dL, according to the guidelines of the Japanese Society for Dialysis Therapy. Hepcidin was measured at baseline. Using the mean values for 1-year period, the relationships among hemoglobin, serum ferritin levels, and transferrin saturation levels were investigated based on a receiver operating characteristic curve and a logistic regression model. In addition, the correlations among serum ferritin, transferrin saturation, and hepcidin levels were analyzed by Pearson product-moment correlation coefficient and linear regression model. RESULTS: By receiver operating characteristic curve, the cutoff point of serum ferritin and transferrin saturation levels with a hemoglobin ≥10 g/dL showed <90 ng/mL (sensitivity: 69.1%, specificity: 72.1%, p < 0.001) and ≥20% (sensitivity: 77.6%, specificity: 48.8%, p = 0.302). Upon logistic regression model analysis with a hemoglobin ≥10 g/dL as the endpoint, the analysis of odds ratios relative to a group with serum ferritin ≥90 ng/mL and transferrin saturation <20% revealed that the group with serum ferritin <90 ng/mL and transferrin saturation ≥20% had the highest ratio: 46.75 (95% confidence interval: 10.89-200.70, p < 0.001). In Pearson product-moment correlation coefficient, hepcidin showed a strong positive correlation with serum ferritin [r = 0.78 (95% confidence interval: 0.72-0.83, p < 0.001)] and a weak positive correlation with transferrin saturation [r = 0.18 (95% confidence interval: 0.04-0.31, p = 0.010)]. In the multivariable analyses of the linear regression model, a positive relationship was shown between hepcidin and serum ferritin [β-coefficient of 0.30 (95% confidence interval: 0.27-0.34, p < 0.001)]; however, no relationship was shown with transferrin saturation [β-coefficient of 0.09 (95% confidence interval: -0.31-0.49, p = 0.660)]. CONCLUSIONS: In this study, the iron status of serum ferritin <90 ng/mL and transferrin saturation ≥20% was optimal in hemodialysis patients receiving recombinant human erythropoietin for anemia therapy. This result indicates that the threshold values for the optimal iron status may be lower than those currently recommended in iron-level management guideline.