Retrospective studies have shown hospitalization and mortality rates during haemodialysis (HD) to be associated with anaemia. The prospective, multicentre Morbidity-and-mortality Anaemia Renal (MAR) study was designed to establish the burden of anaemia by controlling for other risk factors. Charlson index was used for comorbid adjustment. Finally, 1428 patients from 119 centres (60% men, aged 64.4 years, time on HD 15.3 months, Charlson comorbidity index 6.5 +/- 2.3) completed follow-up. They had hypertension (75.8%), diabetes mellitus (25.9%), heart failure (13.9%) and coronary disease (16.7%). Of the total patients, 94.8% were receiving erythropoietin (111.6 +/- 70.6 U/kg/week) and 76.7% i.v. iron, and haemoglobin (Hb) at inclusion was 11.7 +/- 1.5 g/dl. Hospitalization rate was 1.1 admissions/patient/year. Yearly mortality was 12% [35% cardiovascular (CV)]. The relative risk and confidence interval (CI) for hospitalization and death were 0.86 (0.81-0.91) and 0.82 (0.73-0.91), respectively, per 1 g/dl increase in initial Hb after adjustment for comorbidity, vintage, aetiology, access type, albumin and Kt/V. The probability of remaining free from hospitalization (CI) was 0.34 (0.27-0.41) for initial Hb 12 g/dl. Same analysis for patient survival was 0.77 (0.71-0.83) for Hb 12 g/dl, P < 0.001. The Cox regression model for hospitalization-free survival included the risk factors initial Hb (relative risk 0.86 per 1 g/dl increase, P < 0.001) Charlson, albumin and prior CV event. Hb level predicted 1-year-survival and hospitalization. This effect persisted after adjustment for comorbidity and other prognostic factors.