Acquired bone marrow failure (BMF) syndromes include aplastic anemia (AA), paroxysmal nocturnal hemoglobinuria (PNH), and myelodysplastic syndromes (MDS). In this study, to clarify what the frequencies of borderline PNH patients with lower proportions of glycosylphosphatidylinositol (GPI)-protein-deficient erythrocytes and no apparent clinical symptoms of PNH, such as visible hemoglobinuria and thrombotic events, are and what the significance of serum haptoglobin (Hpl) in patients with BMF is, we examined flow cytometry of erythrocytes with expression of CD59 and quantified the concentrations of serum Hpl in 142 Japanese patients with BMF, including 54 AA, 36 PNH, and 52 MDS patients. The diagnosis and grading of the severity of AA were based on the criteria of the International Agranulocytosis and Aplastic Anemia Study Group (Blood, 1987) and that of Frickhoten et al (N Engl J Med, 1991), respectively. A patient with over 1% of CD59− erythrocytes, determined by flow cytometry, was judged to have PNH erythrocytes. The diagnosis and phenotypes of MDS were determined according to the French-American-British criteria (Br J Haematol, 1982). The concentrations of serum Hpl were measured by the nephelometric procedure, developed by Van Lente et al (Clin Chem, 1979), with some modifications. The IgG bound to erythrocytes on the cell surfaces was measured by immunoradiometoric assay, developed by Jeje et al (Transfusion, 1984), with some modifications. The proportions of CD59− erythrocytes from 32 healthy individuals, AA patients, MDS patients, and PNH patients were 0.047 ± 0.052% (mean ± standard deviation), 0.143 ± 0.183%, 0.103 ± 0.126%, and 37.529 ± 33.442%, respectively. Thirteen (36.1%) of 36 PNH patients had lower proportions (range, 1–10%) of CD59− erythrocytes and no apparent clinical symptoms. Subsequently, the concentrations of serum Hpl in 436 Japanese healthy individuals, AA patients, MDS patients, and PNH patients were 152.4 ± 72.7 mg/dl (range, 42.6–309 mg/dl), 127.6 ± 130.4 mg/dl (range, 7–551 mg/dl), 73.2 ± 74.3 mg/dl (range, 3–430 mg/dl), and 13.2 ± 24.6 mg/dl (range, 2–130 mg/dl), respectively. There were significant differences in the values between AA and MDS (p<0.01), MDS and PNH (p<0.01), and AA and PNH patients (p<0.001). The frequencies of AA, MDS, and PNH patients with below 40 mg/dl of serum Hpl were 27.8%, 38.5%, and 94.4%, respectively. The frequency of PNH patients was significantly higher than that of AA or MDS patients (p<0.001 or p<0.001, respectively). Surprisingly, the white blood cell counts, absolute neutrophil counts, and platelet counts in the group (n=15) with low concentrations of serum Hpl in AA patients or the concentrations of hemoglobin and serum iron values in the group (n=20) with low concentrations of serum Hpl in MDS patients significantly decreased compared with those in the group with normal concentrations of serum Hpl in AA patients (n=39) or in MDS patients (n=32), respectively (p<0.005, p<0.01, and p<0.02; or p<0.05 and p<0.02, respectively). These findings suggest the possibility that decrease of serum Hpl might be due to the destruction of hematopoietic precursor cells via immunological mechanisms in AA or due to ineffective erythropoiesis in MDS, although it is unclear how the concentrations of serum Hpl decrease in AA patients. Recently, Theilgaard-Mönch K et al (Blood, 2006) reported that the production of Hpl by immature granulocytes in addition to hepatocytes may partially contribute to concentrations of serum Hpl. Then, we also considered another reason of decrease of concentrations of serum Hpl in patients with AA and MDS. The mean erythrocyte-associated IgG in 100 healthy individuals or 15 AA and 17 MDS patients with low concentrations of serum Hpl was 33 ± 13 molecules/one erythrocyte or 76.5 ± 64.5 molecules/one erythrocyte and 93.5 ± 60.8 molecules/one erythrocyte, respectively. The values of erythrocyte-associated IgG of 3 (20%) of 15 AA patients and 5 (29.4%) of 17 MDS patients were clinically significant, indicating that autoimmune hemolysis was associated with the decrease of concentrations of serum Hpl in these patients. In conclusion, our findings suggest that the concentrations of serum Hpl can be used as a hallmark knowing existence of over 1% of GPI-protein-negative erythrocytes in PNH, and that some mechanisms, including autoimmune hemolysis, contribute to decrease of concentrations of serum Hpl in AA and MDS.