ABSTRACT
Methicillin-resistant Staphylococcus aureus (MRSA) is a leading cause of bacteremia and is associated with significant morbidity and mortality. Prior studies evaluating the association of vancomycin MICs with clinical outcomes in patients with MRSA bacteremia have been inconsistent. This study evaluated the association between vancomycin MICs and 30-day in-hospital mortality rates for patients with MRSA bacteremia. This was a retrospective cohort study of patients with MRSA bacteremia treated with vancomycin for ≥72 h from January 2013 to August 2016. Vancomycin MICs were determined by broth microdilution via automated susceptibility testing methods. Study groups consisted of patients with MRSA isolates that had vancomycin MICs of <2 μg/ml and those with vancomycin MICs of 2 μg/ml. Covariates included demographics, severity of illness, comorbidities, intensive-care unit (ICU) admission, infectious disease consultation, infectious sources, and hospital onset of bacteremia. The primary outcome was 30-day in-hospital mortality. Secondary outcomes included the duration of bacteremia, persistent bacteremia for ≥7 days, recurrence within 30 days, change to alternative antibiotic therapy, and length of hospital stay. Multivariate logistic regression models were analyzed to control for potential confounding variables. A total of 166 patients were included for analysis: 91 patients with vancomycin MICs of <2 μg/ml and 75 patients with vancomycin MICs of 2 μg/ml. In the multivariate logistic regression model, a vancomycin MIC of 2 μg/ml, compared to a MIC of <2 μg/ml, was not significantly associated with 30-day in-hospital mortality after adjustment for confounders. Additionally, all secondary outcomes were not statistically significantly different between study groups. In patients with MRSA bacteremia treated with vancomycin, the vancomycin MIC was not associated with differences in clinical outcomes.
INTRODUCTION
Staphylococcus aureus is one of the most common causes of bacteremia. Since the 1960s, there has been a rise in the level of drug resistance, especially methicillin resistance, which now accounts for close to half of all S. aureus infections, including bacteremia. In particular, methicillin-resistant S. aureus (MRSA) bacteremia is associated with higher morbidity, mortality, and health care costs than methicillin-susceptible S. aureus (MSSA) bacteremia. This is likely due to factors such as increased pathogen virulence, differences in patient populations affected, and the inferior efficacy of MRSA therapy (vancomycin) compared to MSSA therapy (beta-lactams) (1, 2). Vancomycin treatment is accepted as standard first-line therapy for MRSA bacteremia, as well as other MRSA infections (3). Antibiotic susceptibility breakpoints determined by the Clinical and Laboratory Standards Institute (CLSI) classify S. aureus isolates (including MRSA) as susceptible (S) when the MIC is ≤2 μg/ml, intermediate (I) when the MIC is 4 to 8 μg/ml, and resistant (R) when the MIC is >8 μg/ml (4).
There has been debate about whether S. aureus bacteremia caused by an isolate with a vancomycin MIC of 2 μg/ml is associated with poorer clinical outcomes than bacteremia caused by an isolate with a lower MIC. Results from primary studies and meta-analyses have been inconsistent; some have shown higher rates of mortality and treatment failure, whereas others have indicated no association between high vancomycin MICs and mortality (5–9). Interestingly, a paradoxical effect has also been observed in which a higher risk of mortality was determined by Etest for a MIC of <1 μg/ml than for a MIC of >1.5 μg/ml (10). Currently, guidelines from the Infectious Diseases Society of America (IDSA) for the treatment of MRSA infections recommend that for isolates susceptible to vancomycin (MIC, ≤2 μg/ml), “the patient's clinical response should determine the continued use of vancomycin, independent of the MIC” (3). Despite these recommendations, the literature describing the association between high vancomycin MICs and worsened outcomes is likely to influence changes in clinical practice for some practitioners solely on the basis of vancomycin MIC results, independently of the patient's clinical status. At the Robert Wood Johnson University Hospital (RWJUH), only CLSI categorical breakpoint interpretations (S, I, R) are reported (MIC values are suppressed), in order to prevent misinterpretation and bias from MIC values. Our objective for this study was to evaluate the association between vancomycin MICs and mortality for patients with MRSA bacteremia treated with vancomycin.
(This work was presented in part at IDWeek 2017, San Diego, CA, 4 to 7 October 2017.)
RESULTS
There were 247 adult patients with MRSA bacteremia at the RWJUH from January 2013 to August 2016. Of these, 81 patients were excluded for the following reasons: polymicrobial bacteremia (35 patients), death, discharge, or hospice within 24 h (22 patients), or a vancomycin therapy duration of <72 h (24 cases). After exclusion, a total of 166 patients with MRSA bacteremia were included for final analysis. Ninety-one patients were classified as having a vancomycin MIC of <2 μg/ml, and 75 patients were classified as having a vancomycin MIC of 2 μg/ml. A summary of baseline characteristics stratified by vancomycin MICs (study groups with a MIC of <2 μg/ml or 2 μg/ml) is shown in Table 1. Of particular note, the severity of illness (as measured by the Pitt bacteremia score) was slightly higher in the group with a MIC of 2 μg/ml (P < 0.001), as was the proportion of patients with corticosteroid usage (P = 0.028). The presumed sources of bacteremia (classified as low, intermediate, and high risk) were no different for the two study groups (P = 0.941). There were no other significant differences between the study groups.
Demographic and baseline characteristics
There were no statistically significant differences in clinical outcomes, either primary and secondary, between patients classified with a vancomycin MIC of <2 μg/ml and those with a MIC of 2 μg/ml (Table 2). Notably, there was no significant difference between patients with a MIC of <2 μg/ml and those with a MIC of 2 μg/ml in the proportion with 30-day in-hospital mortality (13.2% versus 24%, respectively [P = 0.072]), the median duration of bacteremia (3.5 versus 4 days, respectively [P = 0.679]), and the proportion of patients with persistent MRSA bacteremia (16.5% versus 17.3%, respectively [P = 0.884]).
Primary and secondary outcomes
The relatively low number of outcome events limited the number of covariates (maximum, two variables) that could be entered into the model to avoid overfitting. For the final logistic regression model, the Pitt bacteremia score and corticosteroid usage were identified as potential confounders (modestly associated with both the primary predictor and the outcome). In the final multivariate logistic regression analysis, there was no significant association between the vancomycin MIC (2 μg/ml versus <2 μg/ml) and 30-day in-hospital mortality after adjustment for both the Pitt bacteremia score (adjusted odds ratio [aOR], 1.56; 95% confidence interval [CI], 0.66 to 3.72) and corticosteroid use (aOR, 1.91; 95% CI, 0.84 to 4.33).
DISCUSSION
This study failed to demonstrate a statistically significant relationship between the vancomycin MIC and mortality. The lack of association between the vancomycin MIC and mortality was maintained in the multivariate logistic regression model after controlling for potential confounding variables. There was consistency among the secondary outcome measures as well, in particular the acute microbiological outcomes of duration of bacteremia and persistent bacteremia. The values for our long-term microbiological outcome measure, recurrence of MRSA bacteremia within 30 days, were also similar for the two study groups. The consistency of measures among all outcomes adds strength to the conclusion that there is no apparent association between the vancomycin MIC and clinical outcomes.
In contrast to most other institutions, the RWJUH has a distinct practice of suppressing the MIC values on antimicrobial susceptibility reports, instead giving categorical results only (susceptible, intermediate, or resistant). This is likely to reduce clinician bias that might be associated with the vancomycin MIC before a clinical response can be fully assessed, thus limiting changes in management or prescribing behavior in response to the reported MIC value. This result was most evident in the similar rates of change to alternative antibiotics (i.e., changing of vancomycin therapy) for patients with different vancomycin MICs. Additionally, the high rate of infectious disease (ID) consultation supports the notion that practice patterns are more likely consistent with IDSA recommendations of allowing the clinical response, not the MIC itself, to guide vancomycin therapy for MRSA bacteremia.
Our study results are consistent with previous publications on this topic (8, 9), and improvements in our design and data collection enhance the knowledge related to this controversy. We assessed direct microbiological outcomes, duration of bacteremia and persistent bacteremia, which are likely mediators in the exposure-outcome relationship between the vancomycin MIC and mortality. The lack of differences between study groups in these outcomes adds validity to the findings for the primary outcome of 30-day in-hospital mortality, particularly after adjustment for confounders. Our study adds confidence to the notion that vancomycin therapy is generally effective for MRSA bacteremia independently of the MIC value and strengthens the IDSA recommendation to let the clinical response guide therapy. This has the potential effect of reducing the use of more-expensive alternatives that may not provide additional benefit over vancomycin therapy on the sole basis of the vancomycin MIC.
There are multiple limitations to our study that should be noted when one is interpreting the results. First, given the retrospective nature of the study, assurance could not be made that all patients would have daily blood cultures that would allow for the accurate determination of the duration of bacteremia for research purposes. Despite this inherent limitation, the high rate of ID consultation likely provided the best practice of drawing repeat blood cultures when clinically indicated. Second, the use of automated susceptibility testing methods, particularly two different commercial systems, during the study period may have overestimated or underestimated the true MIC value. The change from the MicroScan to the BD Phoenix system in our microbiology laboratory was an operational decision that was independent of this study. The MicroScan system has been noted to overestimate vancomycin MICs (higher percentages of isolates with MICs of ≥2 μg/ml), whereas the BD Phoenix system has been shown to underestimate vancomycin MICs, both relative to the Etest (11). Either overestimation or underestimation could lead to nondifferential misclassification of the exposure, which can bias the effect measure toward the null. Although nonautomated susceptibility testing methods, such as agar dilution and the Etest, are argued to be more predictive of mortality (11), our study is more pragmatic and uses a susceptibility testing methodology that is valid per CLSI standards and reflects real-life clinical practice, thus improving the external validity of our results. Third, data on source control could not be accurately ascertained from retrospective chart reviews. Fourth, and finally, mortality and recurrence of infection were assessed only through inpatient hospitalization at our institution. There are no data from outpatient or non-RWJUH hospital systems regarding these long-term outcomes.
There are notable strengths to this study cohort and analysis. Although not mandatory at our institution, there was a high rate of ID consultation for both study groups (>85%). Consultation with an ID service for patients with S. aureus bacteremia has demonstrated significant benefits, including reductions in mortality and better adherence to standards of care (12, 13). Other institutions may prioritize ID consultations for patients with high vancomycin MICs due to the perception that these MICs are associated with a higher risk of mortality and infections that are more difficult to treat. This would potentially bias results, due to a differential protective effect from ID consultation, by exposure status. In other words, patients with an isolate with a vancomycin MIC of 2 μg/ml would be more “protected” from mortality, due to a higher rate of ID consultation, than patients with MICs of <2 μg/ml. Lastly, we were able to use regression modeling to adjust for various confounders given the lack of ability to randomize patients, which would otherwise have controlled for both measurable and unmeasurable differences between study groups.
In summary, our retrospective cohort study of patients with MRSA bacteremia treated with vancomycin, at an institution that suppresses MIC value reporting, demonstrated that the vancomycin MIC was not predictive of mortality or microbiological outcomes. Although randomized controlled trials addressing this research objective are unlikely to take place, further prospective studies with rigorous methodology are needed to continue to validate the conclusions described here.
MATERIALS AND METHODS
Study design and population.We performed a retrospective cohort study of patients ≥18 years old with MRSA bacteremia admitted to the RWJUH, New Brunswick, NJ, from January 2013 to August 2016. MRSA bacteremia was defined as the presence of MRSA growth in at least one blood culture and treatment with antibiotics accordingly. Patients were identified by consecutive sampling through the microbiology laboratory database and were further reviewed through the hospital electronic medical record (EMR). Only the first episode of MRSA bacteremia per patient at the RWJUH in the defined study period was included for analysis. All patients included were treated with vancomycin therapy for ≥72 h. We excluded patients with an index blood culture from an outside hospital, polymicrobial bacteremia, and death, discharge, or hospice status within 24 h of the index blood culture. This study was approved by the Rutgers University Institutional Review Board (IRB).
Measurements.Vancomycin MICs were determined by broth microdilution via automated susceptibility testing methods (MicroScan [Beckman Coulter, West Sacramento, CA] from January 2013 to July 2015 and BD Phoenix [BD, Sparks, MD] from August 2015 to August 2016) as per CLSI standards. At the RWJUH, all MIC values are suppressed on all standard antibiotic susceptibility reports for all culture types in the EMR. Therefore, in the final report following the Gram stain report in the EMR, only the organism identification and CLSI breakpoint interpretations (susceptible, intermediate, or resistant) are reported.
The primary predictor of interest was the vancomycin MIC; study groups were classified as having MRSA with a vancomycin MIC of <2 μg/ml or 2 μg/ml. Multiple covariates with a priori interest were recorded in order to assess for differences between study groups as well as to control for potential confounding factors by logistic regression modeling. Covariates included age, infectious disease (ID) consultation, comorbidities (measured by the Charlson comorbidity index), severity of illness (measured by the Pitt bacteremia score), intensive-care unit (ICU) admission on day 1 of the index blood culture, use of dialysis, use of corticosteroids (equivalent to 20 mg prednisone for ≥14 days), a history of transplantation (solid-organ or bone marrow), a history of infective endocarditis, current use of injection drugs or a history of such use, source of infection, and hospital onset of bacteremia (≥48 h after hospital admission).
Outcomes.The primary outcome was 30-day in-hospital mortality. This was a dichotomous outcome evaluating in-hospital survival up to 30 days; favorable outcomes were survival beyond 30 days in the hospital or live discharge, and the negative outcome was death in the hospital within 30 days. Secondary outcomes included the duration of bacteremia, persistent bacteremia, recurrence within 30 days, change to alternative antibiotic therapy, and hospital length of stay (LOS). The duration of bacteremia was defined as the time from the index blood culture to the first negative blood culture. Persistent bacteremia was defined as a duration of bacteremia of ≥7 days. Recurrence was defined as the presence of subsequent blood cultures positive for MRSA within 30 days of the index blood culture after initial clearance of bacteremia. Change to alternative antibiotic therapy was defined as switching from vancomycin after 72 h to another agent with activity against MRSA.
Statistical analysis.Descriptive statistics were performed for all variables. Continuous data were reported as means with standard deviations or medians with interquartile ranges (IQR). All categorical data are reported as percentages. The Student t test or Mann-Whitney U test (for nonparametric distributions) was used for the comparison of continuous variables. The χ2 test was used for the comparison of categorical variables. The significance level was set at a P value of <0.05 (two-sided).
Our a priori relationship of interest is the association of the vancomycin MIC and mortality. We performed a logistic regression analysis to estimate the crude odds ratio and 95% confidence intervals (CIs). A bivariate logistic regression analysis was performed to identify any potential confounders from a list based on a priori knowledge. Any variable in the bivariate analysis with a modest association (P < 0.2) with both the outcome (30-day in-hospital mortality) and the primary predictor (a vancomycin MIC of 2 μg/ml) and with theoretical clinical plausibility was entered into a multivariate logistic regression model to adjust for potential confounding effects. Through the multivariate logistic regression model, an adjusted odds ratio (aOR) and 95% CI were calculated in order to adjust for potential confounders and to determine the strength of association of the vancomycin MIC with 30-day in-hospital mortality. Data analysis was performed using Stata, version 15.0 (StataCorp).
ACKNOWLEDGMENT
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.
FOOTNOTES
- Received 8 December 2017.
- Returned for modification 3 January 2018.
- Accepted 19 January 2018.
- Accepted manuscript posted online 29 January 2018.
- Copyright © 2018 American Society for Microbiology.
