The purposes of this study were to analyze the causes and routes of infection of indwelling central venous catheters and to improve the diagnosis of catheter sepsis before the removal of the cannula. One hundred forty catheter tips were prospectively studies; also, cultures of 52 proximal segments of catheters, 44 swabs of the subcutaneous segment, 195 skin entry sites, 181 infusional fluids, 208 blood samples, and 106 infected distant sites were examined. The catheter sepsis rate was 7.6%, but this sepsis was primary in only 3.4%, because in 4.2% prior isolation of organisms from the wound, urine, throat, or sputum was possible, indicating that the catheter was not primarily responsible for the infection. Primary infection always disappears with removal of the cannula (with or without antibiotics), whereas the course of the secondary infection is related to the gravity of the infected foci and the involved microorganisms. Contamination of the infusional fluid, the skin entry site, and some distant foci carry a real risk of seeding the catheter (from 5.8% to 19.5%). The cultures of the skin entry sites, infusional fluids, distant foci, and the subcutaneous segment of the catheter did not prove useful in predicting the infection. Only the blood cultures were a reliable diagnostic tool: a positive blood culture meant colonization of the catheter tip in 44% of cases and sepsis in 36%. Although the potential colonization varied greatly for different microorganisms, the growth of microorganisms in the blood was a strong indication for removing the cannula.