Central venous catheters (CVC) are an important source of nosocomial infection in intensive care units. The unnecessary removal of CVC suspected to be infected can probably be minimized. In order to test the accuracy of non-invasive methods for predicting catheter colonization, we prospectively compared the results of 50 consecutive CVC tip cultures, with cultures of the CVC hub and the skin at the insertion site. The CVC were separated into two groups based upon the underlying reason for CVC removal: group I (n = 20), suspicion of infection; group II (n = 30), no suspicion of infection. The skin culture (with a threshold of 15 CFU) was useful in both groups for assessing catheter colonization since it was always positive in cases of catheter colonization and always negative in the absence of catheter colonization. The contribution of the CVC hub cultures alone was minimal since there was no case of catheter colonization with negative skin cultures and positive hub cultures suggesting that the main route of catheter colonization was via the skin. Catheter-related bacteremia was identified in seven patients (six in group I and one in group II). In these patients, the ratio of bacterial colony counts (central/peripheral) was greater than 10:1 in only two cases.