The complications associated with the placement and use of Hickman catheters (n = 120), Broviac catheters (n = 146), and implantable ports (n = 93) in children with cancer were analyzed. Percutaneously placed central venous access devices (CVADs) tended to fail less often (P = .86) and to develop infections less often (P = .056) than surgically placed CVADs. The difference in complications with percutaneous versus surgically placed CVADs requires confirmation in a randomized trial to assure they are not a result of differences in patient characteristics. When all catheter failures (removal due to infection, obstruction, or dislodgement) were considered, ports had a significantly longer failure-free duration of use than externalized Hickman and Broviac catheters (P = .0009). Ports also remained infection-free longer than externalized catheters (P = .0014). The greatest risk of infection occurs in the first 100 days of use, particularly for ports. This study demonstrates that for long-term use (greater than 100 days) ports are superior to externalized catheters in children with cancer.