The study was carried out in order to evaluate in maintenance hemodialysis (MH) patients: (1) the reliability of serum ferritin (SF) measurement in iron deficiency diagnosis and therapy; (2) the possibility to improve iron stores assessment through laboratory indexes routinely used in clinical practice; (3) the most effective iron deficiency treatment. After a preliminary assessment of SF reference values in 250 healthy volunteers, we studied 72 MH patients divided into three groups according to their SF baseline values: high (group A), normal (group B), low (group C) (normal range 19–191 ng/ml). Each group was further divided into three subgroups receiving three different iron treatments for 6 months: (1) oral administration of 67.5 mg/day of Fe3+ as Fe-ferritin (subgroups A1, B1, C1); (2) oral administration of 60 mg/day of Fe3+ as Fe-condroitin sulfate (subgroups A2, B2, C2); (3) i.v. administration at the end of each dialytic session of 31 mg of Fe3+ as Fe-gluconate-Na (subgroups A3, B3, C3). The response to the iron therapy was considered positive when the hemoglobin (Hb) and the hematocrit (Ht) increased to > 15% of the baseline values. The rate of positive responses in each subgroup was as follows: Al 0/5, A2 0/5, A3 0/7, B12/10, B2 1/6, B3 5/11, C11/7, C2 3/7, C3 10/16. We concluded that SF values above 191 ng/ml allow to exclude iron deficiency whereas SF values ≤ the normal range are inadequate. In an attempt to improve diagnostic sensitivity we divided patients of subgroup B3 and C3 into responders (R) and nonresponders (NR). In these patients the hematological indexes assessed prior to iron therapy and significantly different between the two groups were used to calculate a discriminant function (DF). Compared to NR, R showed significantly lower values of SF, blood iron, transferrin saturation, mean corpuscular volume, mean corpuscular hemoglobin, and significantly higher vlaues of total iron-binding capacity and red cell distribution width. Nevertheless, none of these parameters alone allows a reliable discrimination between the patients of the two groups. On the contrary through DF, diagnostic sensitivity rises to 93.3% with a specificity of 91.7%. Discriminant scores resulted as 796 for NR, 41 for R, and 450 resulted as the best discriminant value. In order to assess the validity of the iron deficiency diagnosis and treatment criteria developed in the previous parts of the study 20 mg of Fe3+ as Fe-gluconate-Na were given i.v. at the end of each dialytic session to 11 new patients diagnosed as iron deficient by our criteria. All patients showed a positive response to the therapy. SF values increased slowly during the phase of Hb increment and rapidly when Hb values reached stabler values. Finally, we can summarize the following: Diagnostic criteria – SF > 191 ng/ml: no iron deficiency; SF < 191 ng/ml use DF: score > 796 no iron deficiency, 450≤ score < 796 doubtful iron deficiency, score≤450: iron deficiency; and treatment modalities – i.v. administration of 20 mg of Fe3+ as Fe-gluconate-Na at end of each dialytic session, the therapy should be ended when stable Hb values are reached together with a rapid SF level increment or when SF > 191 ng/ml.