Prolonged venous access is desirable in very-low-birth-weight infants and infants for whom feedings are contraindicated. We prospectively evaluated 481 small-diameter venous catheters placed percutaneously in 317 patients over 3 years. Of 478 catheters, 241 (50%) were placed in infants weighing 1 kg or less. Mean catheter stay was 13 days (range, less than 1 to 77 days). Almost half (49%) of the central and thoracic catheters (91% of placements) were removed nonelectively: 43% due to problems such as leaking or clotting and 6% to suspicion of sepsis or venous occlusion. Of the 23 episodes of possible sepsis in the 478 catheter stays, six (1.3%) were confirmed catheter-related sepsis; 12 (2.5%) were confirmed alternate locus sepsis. Three factors specific to percutaneous central venous catheter-related sepsis were prolonged catheter stay (3 to 5 weeks), Staphylococcus epidermidis, and weight less than or equal to 1 kg. Four factors specific to alternate locus sepsis were presence of an alternate infection site, earlier infection (1 to 2 weeks), extremely low birth weight, and prolonged clinical instability. Percutaneous central venous catheterizations reduced the need for the stress of repeated venipuncture, resulting in lower complication rates than those reported with surgically placed central venous catheters, and leading to identification of risk factors specific to catheter sepsis and alternate locus sepsis.