The use and complications of totally implantable vascular access devices (TIVADs) were examined during multiple courses of antibiotics in cystic fibrosis (CF) patients. This retrospective study involved 36 CF centres. Risk factors for removal and septicaemia were sought by survival analysis of censored data. Multivariate Cox models were constructed with removal or septicaemia as the event and the characteristics of TIVADs as explanatory variables. TIVADs (n = 452) were implanted in 315 patients. The mean functional time per device was 32 +/- 25 months. Long-term complications occurred with 188 devices (42%); they consisted mainly of occlusion (21%, requiring removal in 77%), infection (9.3%, requiring removal in 851%; septicaemia in 7.3%; rate 0.3 per 1,000 days, Candida in 66%), and vascular thrombosis (4.7%, removal in 58%). Multivariate survival analysis showed that removal, whatever the reason, was associated with polyurethane (versus silicone) and routine use of the device for blood sampling (versus never). No risk factors, including heparin lock, were identified for septicaemia or for removal for obstruction. Totally implantable venous access devices appear to be safe and reliable for long-term intermittent venous access. Although retrospective, this study suggests that the characteristics of the material and blood sampling are risk factors for removal.