Summary:In a prospective study, 218 cardiothoracic patients, in whom 'Abbocath-T' cannulae had been inserted preoperatively into the internal jugular vein, were randomized to receive skin preparation of the insertion site with tincture of iodine (108 controls) or tincture of iodine followed by application of sterile 2% calcium mupirocin ointment (110 test patients). Cannulae were usually removed within 48 h of the operation. Patients receiving mupirocin were less likely to develop significant colonization of one or more of their cannulae as judged by Maki's criterion of a yield of > 15 colony forming units (cfu) from a cannula segment rolled on an agar plate (17% of mupirocin treated patients compared with 54% of the controls, P < 0.001). Coagulase-negative staphylococci, micrococci, or both, were the commonest isolates and were cultured from 70% of the 186 control cannulae compared with 24% of 172 cannulae inserted through mupirocin-treated skin (P < 0.001). A count of more than 15 cfu was found on the tips of 25% control cannulae compared with 5% of the cannulae from mupirocin-treated patients, an effect which was independent of in-situ time (P < 0.001). For cannulae with colonized tips, the same species was isolated from the skin of the insertion site in 67%, from the exterior of the hub in 61% and from the lumen in only 15%. There were no side effects attributed to mupirocin or superinfection with resistant organisms. We conclude that in cardiothoracic patients the application of mupirocin after standard skin preparation with tincture of iodine significantly reduces the colonization of central venous cannulae by organisms derived from the skin insertion site.