Eleven end-stage renal disease patients trained by stationary cycling during their hemodialysis treatments. After a 6-week control period, 12 weeks of training began and was increased to 30 to 60 minutes at > or = 70% of peak heart rate. Baseline, pretraining and, posttraining exercise tests were performed. Workload (WL), oxygen uptake (VO2peak), cardiac output (Q), heart rate (HR), and arterial oxygen content (CaO2) were measured. Stroke volume (SV), arteriovenous oxygen difference ((a-v)O2), and mixed-venous oxygen content (CvO2) were calculated. Rectus femoris biopsies were obtained pretraining and posttraining. At peak exercise, WL increased from 60 +/- 4 to 70 +/- 6 W (P < 0.05), VO2peak showed an upward trend from 14.8 +/- 0.9 to 16.8 +/- 1.3 mL/kg/min (P < 0.1), and Q, HR, SV, CaO2, CvO2, and (a-v)O2 were unchanged. Ten of the 11 patients increased WL, but only six increased VO2peak (five of 11 patients decreased VO2peak). The difference between groups (P < 0.02) was attributable to (a-v)O2, which increased in those who increased VO2peak (P < 0.02). There was an upward trend for succinate dehydrogenase activity (P < 0.06), and phosphofructokinase activity increased (P < 0.05). However, the rectus femoris capillary to fiber ratio, type I and II fiber areas, and fiber area variability were unchanged, and neither histomorphologic nor enzymatic activity changes were related to change in VO2peak. We conclude that not all dialysis patients increase VO2peak after training, but most can improve exercise capacity. Patients who improved VO2peak widened their (a-v)O2 difference, increasing oxygen extraction and showing that oxygen delivery is not always the limiting factor. Thus, the limitation of VO2peak in dialysis patients is a complex interaction of central and peripheral factors. Muscle therapies, such as exercise training, are needed in addition to increased oxygen delivery in rehabilitation of dialysis patients.