Using a semiquantitative technique for culturing material from vascular catheters, we studied by random allocation the efficacy of three regimens for site care of 827 catheters used in adult patients: an iodophor ointment (PI2), ointment containing polymyxin, neomycin and bacitracin (PNB), and use of no topical agent whatsoever (control). Even though this is the largest study of this subject, there was not a sufficient number of catheter-related septicemias to permit valid comparisons (two in each group, 0.7 percent). However, the rate of local catheter-related infection (greater than or equal to 15 CFU on semiquantitative culture), the prelude to related septicemia, was significantly lower in the PNB group (2.2 percent, P = 0.02) as compared with controls (6.5 percent). Use of PI2-treated catheters resulted in one-half fewer infections (3.6 percent) than use of control catheters (P = NS). Staphylococcal infections occurred with 15 control catheters, eight treated with PI2 and two with PNB (P = 0.002). Infections by gram-negative bacilli occurred less frequently in both treatment groups than in controls, but three of four Candida infections, including one septicemia, occurred in the PNB group. Topical antimicrobial agents confer modest benefit in protection against catheter-related infection, primarily for peripheral venous catheters that must remain in place for more than four days. If an ointment is to be used, topical PNB may be preferable for peripheral venous catheters and PI2 ointment for central venous catheters used for parenteral nutrition and for arterial catheters.