Background. Comparison of outcome of untunnelled catheters (UCs) and tunnelled cuffed catheters (TCCs) is difficult because they are usually used for different patients and conditions. The aim of the present study is to compare the outcome of TCCs with UCs limiting as much as possible the influence of confounding factors. The second purpose was to see whether our results support the time recommendations for maximum use of UCs outlined in the NKF-DOQI guidelines. Methods. Catheter and patient characteristics, catheter-related complications and all cultures taken from haemodialysis catheters inserted during a 3 year period were collected. Results. We analysed the outcome of 272 catheters (149 patients, 11612 catheter-days, 37 TCC and 235 UC). Patients with an UC suffered more often from acute renal failure (40 vs 8% for TCCs, P<0.001), their hospitalization rates were higher (54 vs 14%, P<0.001) and coumarins were used less (11 vs 27%, P<0.01). Rates of preliminary removal were 1.8 per 1000 catheter-days for TCCs, 35.3 for untunnelled femoral catheters (UFCs) and 17.1 for untunnelled jugular catheters (UJCs). Infection rates were 2.9 per 1000 catheter-days for TCCs, 15.6 for UJCs and 20.2 for UFCs. Hospitalization was an independent risk factor for an adverse outcome and more apparent in patients with an UC. After correction for patient differences, the strongest risk factor for preliminary removal (RR 9.69, P<0.001) and infection (RR 3.76, P<0.001) was having an UC inserted. Already, within 2 weeks actuarial and infection-free survival were better for TCCs (P<0.05 vs all separate groups). Conclusions. According to our results, a TCC should be used whenever it can be foreseen that a haemodialysis catheter is needed for more than 14 days.