Hypereosinophilia (EO) is frequently observed in hemodialysis patients without apparent etiology. We prospectively studied hemodialysis (HD) patients during the period between May 1989 and December 1992, and divided the patients into 3 groups: non-eosinophilic control (EOC), intermittent EO (EOI), and persistent EO (EOP), according to the longterm measurements of their blood eosinophil counts. There were 91 patients enrolled in the study which were followed up for 24 months with 23 patients (25.3%) in the EOC group, 53 patients (58.2%) in EOI group, and 15 patients (16.5%) in the EOP group. There were no significant contribution of the type or reuse of the hemodialyzer to the development of EO, neither were age, serum albumin, KT/V, protein catabolic rate (PCR), or the duration of HD. The functional studies of eosinophils showed no significant difference in hypodense eosinophil distribution or eosinophilic cationic protein (ECP) levels between EOP group and EOC group. The distribution of hypodense eosinophils and serum ECP levels before and after dialysis procedure were not uniform, increasing in some but decreasing in the others. Thus the cause of EO is not merely due to the uremic state or dialysis procedures.