Epoetin (recombinant human erythropoietin; EPO) therapy adds a significant cost to the management of end-stage renal disease, the majority of the extra expense being attributable to its acquisition cost. In a Japanese multicentre, randomised, prospective study, a significant dose-dependent response was documented with epoetin given subcutaneously (SC) once a week or once every 2 weeks to patients receiving continuous ambulatory peritoneal dialysis. Three different dosages were studied over 5 months in patients with a haematocrit (Hct) of 0.28 or less, namely 6000U (107 U/kg), 9000U (167 U/kg) and 12 000U (211 U/kg). Epoetin was given weekly for the first 2 months until the target Hct value of 0.33 was reached. The rates of response were 81, 85 and 100% with the 6000U, 9000U and 12 000U regimens, respectively. Subsequently, responders were maintained at the target Hct for an additional 3 months, with the administration frequency eventually being reduced to fortnightly or 4-weekly. Patients in the epoetin 6000U and 9000U groups who did not respond after 2 months' treatment underwent induction and maintenance with the 12 000U regimen. During the maintenance phase, patients receiving the epoetin 6000U and 9000U dosages required weekly (54 and 64%, respectively) or fortnightly (46 and 36%, respectively) injections. Patients receiving the 12 000U regimen were found to require weekly (9%), fortnightly (73%) or 4-weekly (18%) injections. Using these data, we performed a decision analysis that quantitatively incorporated the probability of attaining and maintaining target Hct levels in all patients (i.e. the effectiveness of epoetin), and direct costs as a function of both cumulative doses and injections required in all 3 strategies over 5 months. Decision analysis indicated that the most cost-effective SC epoetin strategy in patients undergoing peritoneal dialysis is epoetin 6000U weekly for 2 months, followed by maintaining the target Hct with weekly or 2-weekly epoetin 6000U for the next 3 months. Nonresponders should restart epoetin therapy using the 12 000U strategy. The 9000U and 12 000U strategies were associated with similar costs, because the economic advantages associated with the lower administration frequency of the 9000U regimen compared with the 6000U regimen were offset by its higher cumulative acquisition cost. In other words, decision analysis indicated that the most cost-effective strategy was to use the lowest effective dose, reserving the highest dosage for patients who do not respond after 2 months. The superiority of this strategy was confirmed by a sensitivity analysis performed on the cost of drug administration, which was varied from zero to $US60 per dose. In addition, consistent results were obtained when the analysis was extended to cover a 1-year treatment period with all 3 strategies.