Reduced susceptibility to glycopeptides in methicillin-resistant Staphylococcus aureus (MRSA) clinical isolates is considered a risk factor for failure of glycopeptide therapy. We compared the prevalences of MRSA isolates with reduced glycopeptide susceptibility in patients with versus without persistent or recurrent MRSA bloodstream infections. A retrospective cohort study at the University Hospital of Geneva identified 27 patients with persistent or recurrent clonally related MRSA bacteremic episodes over an 8-year period, which included 208 consecutive nosocomial MRSA bacteremic episodes. Vancomycin and teicoplanin MICs were determined by a modified macrodilution assay allowing improved detection of glycopeptide-intermediate MRSA isolates (GISA), characterized by elevated teicoplanin or/and vancomycin MICs (≥ 4 μg/ml). For 16 patients (59%), their pretherapy and/or posttherapy MRSA isolates showed elevated teicoplanin MICs, among which 10 (37%) concomitantly displayed elevated vancomycin MICs. In contrast, 11 other patients (41%) were persistently or recurrently infected with non-GISA isolates. In comparison, only 39 (22%) of 181 single isolates from patients with no microbiological evidence of persistent or recurrent infections showed elevated teicoplanin MICs, among which 14 (8%) concomitantly displayed elevated vancomycin MICs. Clinical, microbiological, and pharmacokinetic variables for patients persistently or recurrently infected with GISA or non-GISA isolates were similar. Bacteremic patients with a poor response to glycopeptide therapy had a 2.8-fold- and 4.8-fold-higher rates of MRSA isolates displaying elevated teicoplanin and vancomycin MICs, respectively, than patients with single isolates (P < 0.0001). Detection of elevated teicoplanin MICs may help to predict a poor response to glycopeptide therapy in MRSA bacteremic patients.