An increasing number of reports documenting resistance to human recombinant erythropoietin (rHuEPO) therapy are challenging the concept that erythropoietin deficiency is the main cause of the anaemia of chronic kidney disease (CKD). In an attempt to establish whether other factors play a more predominant role in the anaemia of CKD, 988 patients receiving dialysis were assessed for a wide range of variables. Data were collected on haematocrit (Hct) levels, rHuEPO dose, dry weight, serum ferritin, transferrin saturation, serum albumin, serum aluminium, serum parathyroid hormone intact, eKt/V for urea, gender, dose of i.v. iron administered, time in hospital, and use of i.v. vancomycin. Hyporesponsiveness to rHuEPO was defined as patients requiring >500 IU/kg/week or failing to achieve Hct levels of >30%. Ninety-two (9.2%) of the 988 patients met the above criteria for hyporesponsiveness to rHuEPO. In 21 of these patients, Hct concentrations remained <30% at 6-month follow-up. There were known haematological causes of refractoriness to rHuEPO in nine of these patients. During extended follow-up, probable causes of hyporesponsiveness were discovered in all but two of the remaining 13 patients. Of 62 dialysis patients who received rHuEPO at doses >500 IU/kg/week, 45 (73%) had Hct concentrations of 33-42%. These patients were responding to the higher doses of rHuEPO with no obvious adverse effects. Lower values of serum ferritin, transferrin saturation, and eKt/V, or higher levels of parathyroid hormone or serum aluminium were not associated with higher rHuEPO dose requirements. These results suggest that erythropoietin deficiency is still the main cause of the anaemia of CKD. Erythropoietin replacement therapy can correct the anaemia in almost all iron replete patients providing enough hormone is given, functional iron deficiency is avoided, aluminium levels and parathyroid toxicities are controlled and that no de novo haematological condition that affects erythropoiesis or red blood cell survival develops. Consideration should be given to modifying the definition of rHuEPO hyporesponsiveness. The US Hct target of 33-36% for haemodialysis patients is narrow and the European target of Hct >33% may be significantly more practical and physiologically relevant.