Anemia treatment with epoetin has led to dramatic increases in hematocrit levels since 1989. Studies have demonstrated that morbidity and mortality rates are lower when hematocrit values are within the Disease Outcomes Quality Initiative (DOQI) target range (33 to 36%). Recently, clinical studies demonstrated that patients without cardiovascular disease exhibited lower morbidity rates and improved cognitive function with hematocrit values of >36%. One prospective trial, in contrast, demonstrated that normal hematocrit values among patients with cardiac disease were associated with higher mortality rates. These conflicting results have led to concerns regarding the risks and benefits associated with hematocrit values between 36 and 42%. To address these concerns, a recent cohort of 1996 to 1998 incident hemodialysis patients was studied, with assessments of the risks of death and hospitalization and the medical costs associated with hematocrit values of >36%. Patients survived at least 9 mo after dialysis initiation, and comorbidity, disease severity, and hematocrit levels were determined for months 4 to 9. Patients were grouped on the basis of hematocrit values, i.e., <30, 30 to <33, 33 to <36, 36 to <39, or > or =39%, with 1 yr of follow-up monitoring. A Cox regression model was used to evaluate all-cause and cause-specific mortality and hospitalization rates. The economic evaluations included analyses with Medicare Parts A and B allowable expenditures as the dependent variable and the same clinical characteristics as independent variables. For patients with hematocrit values of > or =36%, mortality rates were not different, hospitalization rates were 16 to 22% lower, and expenditures were 8.3 to 8.5% less, compared with patients with hematocrit values of 33 to <36%. These observations do not demonstrate causality. Additional long-term studies are needed to assess the risks of higher hematocrit values among all patients and patients with cardiovascular disease.