We assessed all episodes of methicillin-resistant Staphylococcus aureus (MRSA) bacteremia at our hospital during a 12-month period (n = 53) and compared those due to heterogeneous vancomycin-intermediate S. aureus (hVISA; n = 5, 9.4%) with those due to vancomycin-susceptible MRSA (n = 48). Patients with hVISA bacteremia were more likely to have high bacterial load infections (P = .001), vancomycin treatment failure (persistent fever and bacteremia for >7 days after the start of therapy; P < .001), and initially low serum vancomycin levels (P = .006). These clinical markers of hVISA bacteremia may help focus diagnostic efforts and treatment.
Infections due to methicillin-resistant Staphylococcus aureus (MRSA) constitute approximately one-half of all S. aureus infections in Australian hospitals, with treatment predominantly reliant upon the use of glycopeptides . Strains of MRSA with reduced vancomycin susceptibility were first reported in 1997 , and in Australia in 2001 , but their clinical significance has been debated [4–12]. S. aureus strains with reduced vancomycin susceptibility include vancomycin-intermediate strains (VISA; MIC, 8–16 µg/mL); vancomycin-resistant strains (VRSA; MIC, ⩾32 µg/mL); and heterogeneous VISA (hVISA) strains, which are defined by the presence of subpopulations of MRSA, typically at a rate of 1 organism per 105–106 organisms, with intermediate vancomycin resistance [3, 4].
The accurate laboratory detection of hVISA is time-consuming. Although susceptibility testing with the Etest (AB Biodisk) may be useful, the gold standard detection method appears to be population-analysis profile (PAP) testing .
Following the detection of the first Australian case of hVISA infection at our institution , we undertook a systematic assessment of the frequency of infections due to S. aureus with reduced vancomycin susceptibility to identify whether there are clinical features that may differentiate them from infections due to vancomycin-susceptible MRSA (VS-MRSA).
Methods. We identified all patients with MRSA bacteremia diagnosed at our institution during a 12-month period from July 2001 to June 2002 and retrieved at least 1 isolate from each episode from frozen storage. If possible, laboratory investigations were performed on the final blood culture isolate or on isolates obtained within the 48 h prior to the final blood sample for which culture results were positive. Isolates were assessed by PAP using the technique described by Wootton et al. . Isolates were defined as hVISA if they met NCCLS MIC criteria (i.e., MIC ⩽4 µg/mL) and they had a PAP area-under-the-curve (AUC) ratio of >0.9 when compared with Mu3 reference strains [4, 14]. Microbroth MIC analysis was performed according to NCCLS criteria , and the Etest analysis was performed according to the manufacturer's instructions. All laboratory testing was performed without knowledge about the clinical features of each case, and isolates were identified as being either VS-MRSA, hVISA, VISA, or VRSA. PFGE was performed on all isolates with methods and definitions described elsewhere [16, 17].
The medical records of all study patients were retrieved and assessed for the following information: age and sex, comorbidities (e.g., diabetes, renal or hepatic impairment, or immunosuppression), history of previous MRSA infection and/or colonization or vancomycin therapy, and number of hospital admissions during the previous 12 months. Clinical features and outcomes were assessed, including the following: the source of the current episode of MRSA bacteremia, the presence of high bacterial load infection (i.e., the presence of prosthetic device infection, an undrained abscess due to MRSA, or endocarditis confirmed by echocardiogram), evidence of vancomycin treatment failure (i.e., ongoing fever and bacteremia after >7 days of vancomycin therapy), overall response to therapy (i.e., time until afebrile for ⩾48 h, number of blood cultures positive for MRSA, time until clearance of bacteremia, length of hospital stay, and survival 1 month after the start of therapy), duration of vancomycin therapy, and presence of low trough serum vancomycin concentrations. These findings were compared for the 2 groups to assess for significant associations.
Continuous variables were assessed by the Kruskall-Wallis test, and categorical results were analyzed by the Fisher's exact test. Calculations were performed using Stata statistical software, version 6.0 (Stata), and a P value of <.05 was considered significant.
Results. During the study period, 53 bacteremic episodes were identified in 52 patients (for which 131 MRSA isolates were obtained). For all but 1 patient, laboratory investigations, including PAP analysis, were performed on the patient's final blood culture isolate (n = 44) or on an isolate from culture of a sample obtained 48 h before the final culture sample (n = 8). Five (9.4%) of 53 episodes were due to hVISA, and 48 episodes (in 47 patients) were due to VS-MRSA. No episode was due to VISA or VRSA. The 5 hVISA isolates had PAP AUCs of 0.91, 1.14, 1.17, 1.37, and 1.43, yet by broth microdilution the MICs of vancomycin were 2, 2, 2, 4, and 4 µg/mL, respectively, and by Etest the MIC of vancomycin was 4 µg/mL for all isolates. In comparison, all VS-MRSA isolates had a PAP AUC of ⩽0.9, and the MIC of vancomycin was 0.5–2 µg/mL (median, 1 µg/mL) by broth microdilution and 0.5–2 µg/mL (median, 1.5 µg/mL) by the Etest. The 5 hVISA isolates displayed 3 different PFGE patterns, suggesting they belonged to separate clones.
Patients with hVISA bacteremia were not significantly different from those with VS-MRSA bacteremia in terms of age (mean, 63.6 vs. 65.9 years), comorbidities, history of previous MRSA infection (1 of 5 vs. 18 of 48 patients), or number of previous hospital admissions or antibiotic courses. Of note, none of the patients with hVISA bacteremia had received vancomycin during the preceding 6 months, compared with 16 of 48 patients in the VS-MRSA group (P = .3).
The probable sources of bacteremia and treatment outcomes for patients with hVISA and those with VS-MRSA infection are summarized in table 1. The presence of prosthetic devices at the onset of bacteremia was a common feature; for 5 of 5 hVISA episodes and 32 of 48 VS-MRSA episodes, prosthetic devices were in situ at the time of obtainment of the first blood sample for which culture was positive (P = .3). A comparable number of prosthetic devices in each group (2 of 5 vs. 7 of 32) could not be removed promptly. Four patients with hVISA infection were treated with linezolid and responded promptly, and 1 patient with severe, inoperable prosthetic valve endocarditis had palliative treatment-withdrawal.
Episodes of bacteremia due to hVISA were significantly more likely than those due to VS-MRSA to be associated with high bacterial load infections (P = .001) and vancomycin treatment failure (P < .001). Patients with hVISA infection had a longer duration of fever (P < .001), a greater number of positive blood cultures (P < .001), a longer time until clearance of bacteremia (P = .002), and a greater length of stay in the hospital (P = .006). However, overall patient mortality at 1 month was similar. Of note, more patients with hVISA infection than patients with VS-MRSA infection had low (<10 µg/mL) trough serum vancomycin levels during the first week of therapy (P = .006), but, despite subsequent dose adjustments to ensure appropriate levels, vancomycin therapy still failed for patients with hVISA infection.
Restricting analysis to the 15 patients with high bacterial load infections (5 with hVISA and 10 with VS-MRSA infections) identified statistically significant differences in most of the associated features. Patients with hVISA infection had a higher rate of vancomycin treatment failure than did patients with VS-MRSA infection (5 of 5 patients vs. 1 of 10; P = .002), a longer time until afebrile (mean days ± SD, 35 ± 26 vs. 5.2 ± 5; P = .005), a greater number of positive blood cultures (mean ± SD, 8.2 ± 3.3 vs. 2.5 ± 2; P = .007), and a greater likelihood of low vancomycin concentrations during the first week of therapy (5 of 5 patients vs. 1 of 10; P = .002).
Since the duration of fever may have influenced the number of blood samples obtained for culture and therefore altered the rate of detectable bacteremia, we assessed those episodes in which fever persisted for >72 h after the commencement of vancomycin therapy and in which additional blood samples were obtained for culture. All 5 patients with hVISA infection and 7 patients with VS-MRSA infection had persistent fever; a total of 79 and 51 blood samples for culture were obtained from these patients, respectively. The mean number of positive blood cultures (± SD) was greater in the hVISA group (8.2 ± 3.3 vs. 2.7 ± 2.6; P = .01), as was the mean duration of bacteremia (39 ± 32 days vs. 6.7 ± 5.3; P = .01).
Discussion. We believe this study has a number of important implications. First, the rate of hVISA infection (5 [9.4%] of 53 episodes), was substantially higher at our institution than we expected on the basis of prior MIC data alone. Second, unlike some investigators [4–7], we found the presence of hVISA to be significantly associated with clinical evidence of both vancomycin treatment failure (defined as persistent fever and bacteremia for >7 days after commencement of vancomycin therapy) and high bacterial load infection. The latter, especially, highlights the importance of surgical debulking and removal of infected prosthetic material. The initially low trough vancomycin concentrations observed in patients with hVISA infection could result in either induction of any preexisting vancomycin resistance or the selection of new, resistant strains [12, 18]. However, since dosages adjusted to ensure appropriate vancomycin levels failed to improve outcomes among patients with hVISA infection, and most responded promptly to linezolid therapy, the presence of reduced vancomycin susceptibility among these strains appears important. The polyclonal nature of the hVISA strains in our study suggests either of these mechanisms is more likely than nosocomial transmission of a single clone. Finally, we found the currently recommended laboratory screening methods for hVISA  to be insensitive compared with PAP analysis results. As have other investigators , we found that a vancomycin MIC of ⩾4 µg/mL, using the Etest with a standard inoculum, would have differentiated between hVISA and VS-MRSA with reasonable accuracy.
Limitations of our study include its small size and retrospective design, so that observed associations may not absolutely confer clinical predictive value. To control for all potential confounding between the presence of infected prosthetic material, treatment failure, number of blood samples obtained for culture, and rate of bacteremia identified requires a prospective study—this we are planning.
In the meantime, we believe clinicians should consider the presence of high bacterial load infection and persistent fever and bacteremia for >7 days after commencement of vancomycin therapy to be potential clinical markers of hVISA infection. Given the current difficulties with laboratory screening for hVISA and the time-consuming nature of PAP analyses, these clinical markers may assist laboratories in targeting diagnostic efforts and assist clinicians in reassessing treatment options for such infections.