Patients having central venous catheters for three or more days were prospectively randomized to receive a transparent (n = 58) or gauze (n = 57) dressing to compare the incidence of insertion site colonization, local catheter-related infection, and catheter-related sepsis. Quantitative cultures of the catheter insertion site (25 cm2) revealed significantly greater colonization (P less than or equal to .009) after 48 h in the transparent versus the gauze dressing group. Local catheter-related infection occurred significantly more often (P = .002) in the transparent (62%) than in the gauze group (24%). Seven episodes of catheter-related bacteremia occurred in the transparent group (16.6%) and none in the gauze group (P = .015). Stepwise logistic regression analysis revealed that cutaneous colonization at the insertion site of greater than or equal to 10(3) cfu/mL (relative risk, 13.16) and difficulty of insertion (relative risk, 5.39) were significant factors for catheter-related infection. These data suggest that transparent dressings are associated with significantly increased rates of insertion site colonization, local catheter-related infection, and systemic catheter-related sepsis in patients with long-term central venous catheters.