Study Design and Setting

We conducted a prospective, observational, multicenter study during April 2011−June 2014 in 5 hospitals in Spain in which we included patients given a diagnosis of MSSA CRBSI according to standard a definition (14) that fulfilled 5 criteria: 1) age >18 years and a life expectancy >7 days; 2) no use of daptomycin within the 3 months preceding the incident MSSA CRBSI episode; 3) removal of the intravascular catheter suspected to be the source of CRBSI within a maximum of 72 hours from sampling of the first blood cultures yielding MSSA; 4) a maximum of 72 hours of treatment with glycopeptides; and 5) at least the first blood culture isolate available for microbiological analysis. All included patients signed a study-specific informed consent form. The study was approved by the ethics research boards at all participating centers.

Management of MSSA CRBSI

Although the nature of the study was not interventional, the adherence to evidence-based quality-of-care measures was strongly encouraged among participating investigators to optimize the therapeutic management of MSSA CRBSI. These measures were extraction of follow-up blood cultures, ordering echocardiography or venous Doppler ultrasound examination for patients with clinical indication, early use of intravenous cloxacillin or other highly effective antimicrobial drugs for treatment of for MSSA as definitive therapy, and adjustment of treatment duration according to the complexity of the infection.

Definitions

Recorded clinical variables were age; site of acquisition of infection; concurrent conditions, as assessed by the Comorbidity Index of Charlson et al. (15); prognosis of the underlying disease (classified according to the McCabe and Jackson modified criteria as rapidly fatal [when death was expected within 3 months], ultimately fatal [when death was expected within a period of >3 months but <5 years] and nonfatal [when life expectancy was >5 years]) (16); type of intravenous catheter; timing of catheter removal; treatment administered; duration of fever; duration of bacteremia; Pitt bacteremia score (17); development of septic shock; endovascular infection or complicated bacteremia (defined below); and all-cause and MSSA CRBSI−attributable mortality rates at 30 days. MSSA CRBSI−attributable deaths at 30 days were identified as per the clinical judgment of the investigator.

Antimicrobial drugs were classified within 1 of the following mutually excluding categories: glycopeptides; daptomycin; antistaphylococcal β-lactams (parenteral cloxacillin, cefazolin or other active β-lactams [including parenteral amoxicillin/clavulanate, piperacillin/tazobactam, imipenem, or meropenem]); non−β-lactam agents with in vitro activity against MSSA (antistaphylococcal quinolones, tigecycline, or linezolid); and noneffective agents. Timing to initiation of antimicrobial drug therapy and duration of treatment were also recorded. Empiric therapy was defined as therapy administered before identification and antimicrobial susceptibility testing of the isolate. Complicated bacteremia (study outcome) was defined by 1 of 5 events occurring after the incident episode of MSSA CRBSI: 1) development of endocarditis (defined according to modified Duke criteria) (18); 2) septic thrombophlebitis (defined by persistent MSSA bacteremia for >72 hours after the initiation of active therapy associated to the documentation of a thrombus at the site of catheter insertion); 3) septic metastatic infection, including arthritis, spondylitis, infections involving vascular or osteoarticular prostheses (excluding intravascular catheter) or end-organ hematogenous seeding to other locations remote from the primary focus; 4) persistent bacteremia for >72 hours; or 5) persistent fever for >4 days after extraction of the first blood culture yielding MSSA, provided that alternative causes had been reasonably excluded (19).

Microbiological Methods

Blood cultures were processed and isolates were identified according to standard techniques at each participating center. The first MSSA isolate from each patient was stored at −80°C until analysis and sent to the central reference laboratory (Department of Microbiology, University Hospital 12 de Octubre, Madrid, Spain). We subsequently subcultured strains twice and performed antimicrobial susceptibility testing by using the E-test method according to manufacturer’s recommendations (AB bioMérieux, Solna, Sweden). We tested vancomycin, daptomycin, oxacillin, and linezolid. In addition, the E-test macromethod was performed to screen for heteroresistant vancomycin-intermediate S. aureus (hVISA) (20). All the plates were visually read throughout the entire study period by 2 independent observers, who were blinded to each other and to the clinical outcomes. Interobserver discrepancies were resolved by reevaluation and a consensus decision.

Statistical Analysis

We used the Student unpaired t-test to compare normally distributed continuous variables, the Mann-Whitney U test to compare continuous variables with non-normal distribution, and the χ² and Fisher exact tests to compare proportions as appropriate. Areas under receiving operating characteristic curves for MICs for selected antistaphylococcal agents (on the basis of their discriminative value at the univariate comparisons) were plotted to select the optimal cutoff value in terms of sensitivity (S) and specificity (Sp) to distinguish between patients with or without complicated bacteremia by means of the Youden’s J statistic (J = S + Sp −1). We calculated the Pearson correlation coefficient (ρ) to assess the strength of the linear relationship among E-test MICs for daptomycin and vancomycin.

To test the effect of high MICs for antistaphylococcal agents and other clinical variables on the occurrence of complicated MSSA CRBSI, we plotted event-free Kaplan-Meier survival curves and performed comparisons between groups by using the log-rank test. Statistically relevant variables (p<0.10) at the univariate level were entered into multivariate Cox proportional hazard models with a backward stepwise selection, and hazard ratios (HRs) with 95% CIs were estimated. Some variables considered clinically relevant were also forced into the models regardless of their univariate p values. All statistical tests were 2-tailed, and the threshold of statistical significance was set at p<0.05. We used statistical software SPSS version 20.0 (SPSS, Inc., Chicago, IL, USA) to perform calculations of different analysis and generated graphics by using Prism version 6.0 (GraphPad Software Inc., La Jolla, CA, USA). Variables with an unacceptable proportion of missing values (>30%) were excluded from the analysis.

Study Population and Outcome

Of 108 patients initially recruited, 19 were excluded because they did not fulfill inclusion criteria after revision of the clinical charts (13 had received >72 hours of treatment with vancomycin and 6 had had a delay of >3 days in catheter removal). An additional 6 patients cases were excluded because of inability to recover the MSSA strain for microbiological analysis. Therefore, we included 83 patients in the final analysis; median follow-up was 479 days (range 5 days–1,014 days). No variables were excluded because of missing data. None of the included patients had received vancomycin, daptomycin, oxacillin, or linezolid within the 3 months preceding the incidence MSSA CRBSI episode.

We compared the main demographic and clinical characteristics of the 83 patients in the final cohort (Table 1). Catheter removal was performed for all patients within the first 72 hours from sampling of the first blood cultures and, in most (75 [90.6%]) patients, within the first 48 hours. Antistaphylococcal β-lactams were used as empiric therapy in 43 patients (51.8%), glycopeptides in 32 (38.5%), and daptomycin in 17 (20.4%). Combined therapy was used for 25 (31.1%) patients. Twelve (14.5%) patients received no empiric therapy or the regimen administered was based on antimicrobial drugs with no in vitro activity against MSSA. For all patients who received daptomycin, the dose used was >6 mg/kg/day (range 6−10 mg/kg/day). Venous Doppler ultrasound examination and echocardiogram were performed for 41 (49.5%) and 61 patients (73.5%), respectively.

Complicated MSSA CRBSI developed in 26 patients (31.3%) at a median of 4 days (interquartile range [IQR] 1–26 days) after sampling of blood cultures. No patients had late complications (i.e., >30 days). Septic thrombophlebitis was the most frequent complication (10 patients [12%]), followed by right-sided endocarditis (2 cases [2.4%]), hematogenous osteoarticular infection (2 cases [2.4%]), and pulmonary emboli (2 cases [2.4%]) (Technical Appendix Table 1). All-cause and MSSA CRBSI−attributable 30-day mortality rates were 12.0% and 2.4%, respectively.

Correlation between MICs and Outcome

All strains were sensitive to all antimicrobial drugs tested. None of the isolates had the hVISA phenotype. The 50% MIC and 90% MIC and distribution ranges values were 0.5 μg/mL and 0.75 μg/mL (range 0.5–0.75 μg/mL) for cloxacillin, 1.5 μg/mL and 2 μg/mL (range 0.5–2.5 μg/mL) for vancomycin, 0.38 μg/mL and 0.75 μg/mL (range 0.125–0.94 μg/mL) for daptomycin, and 1 μg/mL and 1.5 μg/mL (range 0.25–2 μg/mL) for linezolid.

We compiled a comparative description of geometric means of MICs determined by E-test for vancomycin, daptomycin, oxacillin, and linezolid among isolates from patients with or without complicated bacteremia (Table 2). We observed that the MICs for vancomycin and daptomycin were higher for isolates from patients in whom complicated bacteremia developed.

We subsequently performed an exploratory area under receiving operating characteristic curve analysis to establish the optimal cutoff point for the daptomycin MIC with the best discriminatory capacity to predict development of complicated bacteremia. An MIC for daptomycin of 0.5 μg/mL was selected (Technical Appendix Figure 1). This approach also confirmed that 1.5 μg/mL was the optimal cutoff point for the vancomycin MIC (Technical Appendix Figure 2). A total of 43 strains (41.8%) had a vancomycin MIC >1.5, μg/mL and 13 (15.7%) had a daptomycin MIC >0.5 μg/mL. On the basis of these thresholds, 11 (84.6%) of 13 isolates with high MICs for daptomycin also showed high MICs for vancomycin, and we found a moderate positive correlation between these variables (ρ = 0.21, p = 0.05) (Technical Appendix Figure 3). The percentage of MSSA strains with high MICs for vancomycin or daptomycin was higher in the group of patients with complicated bacteremia (Table 2). Strains isolated from all the patients in whom septic thrombophlebitis developed had high MICs for vancomycin or daptomycin.

We compared clinical characteristics of patients according to MICs for vancomycin or daptomycin (Table 3). Rates of complicated MSSA CRBSI were significantly higher among patients with episodes caused by strains with MICs for vancomycin (42% vs. 20%; p = 0.05) or daptomycin (69.2% vs. 24.3%; p = 0.004) than for those with episodes caused by strains with lower MICs. Strains with high MICs for vancomycin were not associated with higher rates of severe sepsis or septic shock at bacteremia onset, or with increased all-cause or attributable mortality rates. Conversely, we found that high MICs for daptomycin were more common among hemodynamically unstable episodes (i.e., severe sepsis or septic shock) than among remaining episodes (46.2% vs. 15.7%, p = 0.02). However, no effect on mortality rates was observed.

Risk Factors for Development of Complicated Bacteremia

Figure

Figure. Event-free Kaplan-Meier survival curves for complicated bacteremia according to E-test MICs for daptomycin (A) and vancomycin (B) for complicated catheter-related bloodstream infections with a methicillin-sensitive Staphylococcus aureus isolate.

We compared clinical characteristics of patients with or without complicated bacteremia (Table 4). Most (65.4%) patients with complicated cases were given a diagnosis at 1 recruiting center. Such an imbalance might be partially explained by the fact that such a center was the study promoter and the largest center at which the greatest diagnostic efforts were made. The probability of remaining free of complicated bacteremia at day 30 was significantly lower for patients with MSSA CRBSI caused by isolates with daptomycin MICs >0.5 μg/mL (33. 3% vs. 75%; log-rank test p = 0.002) and isolates with vancomycin MICs >1.5 μg/mL (59.2% vs. 79.6%; log-rank test p = 0.02) (Figure).

Because of low number of patients with complicated bacteremia (26 episodes), we could not assess the potential effect of MICs by performing a single logistic regression model adjusted for all the covariates found to be significant by univariate analysis. We alternatively attempted an exploratory approach based on different models. We used Cox regression models that incorporated a maximum of 3 variables at a time, always including the recruiting center as a potential confounder (Table 5; Technical Appendix Table 2). The presence of high MICs for vancomycin (minimum adjusted HR 2.4, 95% CI 1.2–5.5) and daptomycin (minimum adjusted HR 2.4, 95% CI 1.1–5.9) were kept as independent risk factors for development of complicated MSSA CRBSI in all models.

We performed a sensitivity analysis by restricting the study outcome to the occurrence of septic thrombophlebitis. These associations remained essentially unchanged, and we observed even higher HRs (minimum adjusted HR for vancomycin MIC 6.0 [95% CI 0.7–52.2]; minimum adjusted HR for daptomycin MIC 5.5 [95% CI 1.3–22.5]) (Technical Appendix Tables 2, 3).