There have been few clinical studies on the association between the 24-h area under the concentration-time curve (AUC24) to minimum inhibitory concentration (MIC) ratio and vancomycin treatment outcomes in methicillin-resistant Staphylococcus aureus (MRSA) infections. Patients with MRSA bacteraemia between July 2009 and January 2012 were analysed retrospectively. All adult patients treated with vancomycin for ≥72 h without dialysis were included. The MIC was determined by Etest and broth microdilution (BMD). Initial steady-state AUC24 was estimated using a Bayesian model, and the AUC24/MIC cut-off value for differentiating treatment success and failure was calculated by classification and regression tree (CART) analysis. In total, 76 patients were enrolled; vancomycin treatment failure occurred in 20 patients (26.3%). Catheter-related infection was the most frequent (35.5%), followed by surgical site infection (26.3%), whilst 25 (32.9%) had complicated infections. In univariate analysis, decreased MRSA vancomycin susceptibility (MIC≥1.5 mg/L) and vancomycin trough levels (15-20 mg/L) were not associated with treatment outcomes. In the CART analysis, low initial vancomycin AUC24/MIC (<430 by Etest; <398.5 by BMD) was associated with a higher treatment failure rate (50.0% vs. 25.0%, P=0.039 by Etest; 45.0% vs. 23.2%; P=0.065 by BMD). In multivariate analysis, low initial vancomycin AUC24/MIC was a significant risk factor for treatment failure [adjusted odds ratio (aOR)=4.39, 95% confidence interval (CI), 1.26-15.35 by Etest; aOR=3.73, 95% CI 1.10-12.61 by BMD]. In MRSA bacteraemia, a low initial vancomycin AUC24/MIC is an independent risk factor for vancomycin treatment failure.