A prospective randomized trial was organized at the Institut Gustave-Roussy to assess the reliability of classical external catheters (CE) versus totally implanted access systems (TI) for delivering intravenous chemotherapy for a duration of at least 6 months. The analysis was performed on the 96 patients whose implantation succeeded (CE 46, TI 50). Failure was defined as loss of ability to function (followed by removal) within the 6-month period of the survey. Patients dying with functional catheters were considered as censored (15 cases) at the time of death. Twenty-four access systems were removed. The removal-free curves differ significantly (P less than 0.001), favoring the TI access systems. The main reasons for removal were: catheter fall (CE6, (TI 0), migration (CE 1, TI 1), infection (CE 5, TI 1), thrombotic occlusion (CE 1, TI 0) and venous complications (CE 1 thrombosis plus 1 pulmonary embolism, TI 1 thrombosis). In addition, a survey by questionnaire demonstrated a significantly better patient activity rate (P = 0.02) and hygiene (P less than 0.001) in the TI group. This prospective randomized study demonstrates that totally implanted access systems are more reliable, safer and better tolerated than classical external catheters for solid tumor patients undergoing intravenous chemotherapy for longer than 6 months.