The objective of this study was to quantify the incremental medical costs that are associated with untreated anemia among elderly patients with predialysis chronic kidney disease (CKD). An analysis of claims and laboratory data between January 1999 and February 2005 was conducted. Inclusion criteria were age >/=65 yr, two or more hemoglobin readings, one or more claims for CKD, and two or more GFR values of <60 ml/min per 1.73 m(2) (stages 3 to 5 CKD). Patients were excluded when they had cancer or lupus, had received organ transplantation, or were treated for anemia. An open-cohort design was used to classify patients' observation periods into anemia and nonanemia. Both univariate and multivariate analyses were conducted to compare periods of anemia and nonanemia for average monthly medical costs; the latter was adjusted for age, gender, GFR, diabetes, hypertension, liver cirrhosis, coronary artery disease, myocardial infarction, and left ventricular hypertrophy. A subset analysis of patients with moderate CKD (stage 3) was conducted. A total of 2001 patients were identified. Untreated anemia was associated with a significant increase in medical costs, with an unadjusted incremental monthly cost of $1089 (P < 0.0001) and a cost ratio of 1.8:1 relative to nonanemia. After controlling for covariates, untreated anemia remained significantly associated with a cost increase (adjusted incremental monthly cost $503; cost ratio 1.4:1; P < 0.0001). Similar significant cost burden was observed in the subset of patients with moderate CKD. The retrospective observational design may be more susceptible to bias than a randomized, controlled trial. This large study, which was based on real-life practice data, demonstrated that untreated anemia in elderly patients with predialysis CKD was associated with a significant increase in medical costs.