Vancomycin-resistant enterococci (VRE) infections are a public health threat associated with increased patient mortality and healthcare costs. Antibiotic usage, particularly cephalosporins, has been associated with VRE colonization and VRE bloodstream infections (VRE BSI). We examined the relationship between antimicrobial usage and incident VRE colonization at the individual patient level. Prospective, weekly surveillance was undertaken for incident VRE colonization defined by negative admission but positive surveillance swab in a medical intensive care unit over a 17-month period. Antimicrobial exposure was quantified as days of therapy (DOT)/1000 patient-days. Multiple logistic regression was used to analyse incident VRE colonization and antibiotic DOT, controlling for demographic and clinical covariates. Ninety-six percent (1398/1454) of admissions were swabbed within 24 h of intensive care unit (ICU) arrival and of the 380 patients in the ICU long enough for weekly surveillance, 83 (22%) developed incident VRE colonization. Incident colonization was associated in bivariate analysis with male gender, more previous hospital admissions, longer previous hospital stay, and use of cefepime/ceftazidime, fluconazole, azithromycin, and metronidazole (P < 0·05). After controlling for demographic and clinical covariates, metronidazole was the only antibiotic independently associated with incident VRE colonization (odds ratio 2·0, 95% confidence interval 1·2-3·3, P < 0·009). Our findings suggest that risk of incident VRE colonization differs between individual antibiotic agents and support the possibility that antimicrobial stewardship may impact VRE colonization and infection.