The objectives of this study were to examine the predictive value of method-specific vancomycin (VAN) minimum inhibitory concentration (MIC) results on treatment outcomes of meticillin-resistant Staphylococcus aureus (MRSA) infections. VAN MIC values for MRSA strains were determined using Etest, VITEK-1, MicroScan (MScan) and broth microdilution (BMD), with additional screening for heterogeneous glycopeptide-intermediate S. aureus (hGISA) phenotype. Patients' charts were reviewed for outcome correlation. Performance characteristics of method-specific VAN MICs in predicting outcome were compared. Most (76%) of the 92 strains tested caused pneumonia or bacteraemia. The majority of strains tested (>70%) had a VAN MIC >1mg/L by Etest or MScan compared with 41% by Vitek and 7% by BMD. Agreement between test methods for high versus low MICs (>1mg/L vs. < or = 1mg/L) ranged from 36% to 71%. High versus low VAN MICs by Etest differentiated response of invasive strains to VAN. Performance characteristics (sensitivity/specificity/positive predictive value/negative predictive value) were: Etest, 55/81/89/38%; and Vitek, 56/62/81/32/%, respectively. Eight strains (9%) demonstrated a hGISA phenotype; more yielded high MICs by Etest, MScan and Vitek than BMD (87%, 87% and 75% vs. 50%). In conclusion, VAN MIC testing methods produce highly variable results. The Etest method appears to be relatively more reliable in predicting treatment response and yielded higher MICs for strains with a hGISA phenotype.