Prior to beginning to administration erythropoietin (EPO) in 1989, we examined the relationships of hematocrit (HCT), mean corpuscular volume (MCV), and serum ferritin (FER) in one group of hemodialysis patients (group A, n = 117) and replicated the findings in a second group (group B, n = 73). The groups had similar mean ( ± SD) HCT (A: 25.7 ± 5.3%; B: 25.2 ± 5.1%), MCV (A: 83.3 ± 6.5 fl; B: 83.5 ± 7.5 fl) and FER (A: 607 ± 1446 µg/l; B: 374 + 601 µg/l). For group A, iron stores [log (FER)] correlated inversely with HCT (r = -0.44, p < 10-4) and directly with MCV (r = 0.32, p < 10-3). After dividing group A into octiles by the FER level, the lowest octile (mean FER 17.8 ± 6.2 µg/l) had the highest mean HCT (29.5 ± 6.4%) and lowest mean MCV (80.8 ± 7.1 fl), while the highest octile (mean FER 3,312 ± 3,005 µg/l) had the lowest mean HCT (21.9 ± 2.8%) and the second-highest mean MCV (86.4 ± 4.9 fl). The trends were similar in group B. We conclude that increased erythropoiesis appeared to cause or, at least, unmask iron deficiency in HD patients even prior to the advent of EPO therapy. Variations in the level of erythropoiesis among these patients (presumably due to variation in EPO levels, chronic inflammation) strongly influenced the determinants of iron stores (i.e., marrow utilization of iron, transfusion need); iron stores, in turn,