With the introduction of the on-line preparation of dialysis fluids, the hemofiltration technique, which has never had a widespread diffusion in its old version with the infusion bags, has gained a new interest. We planned a prospective, randomized, 3-year-long study comparing survival and morbidity in ultrapure bicarbonate dialysis (BD) with on-line predilution hemofiltration (HF). Since comorbidity is one of the main factors limiting survival, the study was addressed to patients with a severe degree of comorbidity. The paper presents the preliminary results of the trial. Sixty-four patients were enrolled and randomized to either BD (N = 32) or HF (N = 32). Mean age and dialysis vintage were comparable. Twenty patients died during the study, 12 in BD and 8 in HF. The relative risk of death was 11% higher in patients treated with BD compared to those in the HF group (p <0.005). The number of hospitalisation events per single patient was lower, even though not significantly, in HF compared to BD (1.94 ± 1.26 in HF vs 2.48 ± 1.98 in BD, p = NS). As concerns biochemistry, apart from beta-2-microglobulin, any other substantial difference was not found during the study, though the small solute concentration was generally a little more elevated in HF than in BD. Dialysis hypotension showed a trend to decrease in both the dialysis modalities up to near half of the trial, then, during the last year, it remained quite stable in HF, while, on the contrary, it increased in the BD group. By the end of the protocol, patients in HF showed a 2.5% incidence of acute dialysis hypotension, while patients in BD had 23%. In conclusion, this medium-long run study showed that, in spite of a comparable level of small solute removal, survival was longer in patients treated with on-line HF than with BD. Morbidity was lower too, but without a statistical relevance. Hemodynamic stability, in the long run, became clearly better with convection. We should reconsider what is dialysis adequacy and which are its best clinical markers.