Comparison of the Clinical Characteristics and Outcomes Associated with Vancomycin-Resistant Enterococcus faecalis and Vancomycin-Resistant E. faecium Bacteremia

Abstract

In published studies, cohorts of patients with bacteremia due to vancomycin-resistant Enterococcus (VRE) have predominantly been infected with Enterococcus faecium. Little is known about the epidemiology and outcomes associated with bacteremia due to VR Enterococcus faecalis. A retrospective study of isolates obtained from January 2008 to October 2010 was conducted at Detroit Medical Center (DMC). Unique patients with blood cultures positive for VRE were reviewed. Outcomes were analyzed using logistic regression. During the study period, 105 cases of bacteremia due to VR E. faecalis and 197 cases of bacteremia due to VR E. faecium were identified. The mean age in the study cohort was 61.5 ± 15 years; 162 subjects (53.6%) were male. After controlling for a propensity score, bacteremia due to VR E. faecalis was associated with >2-fold-lower in-hospital mortality than bacteremia due to VR E. faecium. Interestingly, bacteremia due to VR E. faecalis was associated with longer hospital stay after VRE isolation, although total length of stay was similar for groups with VR E. faecalis and VR E. faecium. Bacteremia due to VR E. faecalis was associated with a >2-fold-lower risk for mortality than bacteremia due to VR E. faecium, possibly due to the availability of β-lactam therapeutics for treatment of VR E. faecalis.

INTRODUCTION

Enterococci are the third most common type of health care-associated pathogen in the United States (17). The two most common species responsible for enterococcal infections in humans are Enterococcus faecalis and Enterococcus faecium. Enterococci are inherently resistant to several antimicrobials and also have an ability to acquire mobile genetic resistance determinants (22). The emergence of resistance to various antimicrobial agents, specifically to vancomycin, has become a major clinical and epidemiological threat. Vancomycin-resistant Enterococcus (VRE) infections are associated with increased morbidity, mortality, and costs compared to vancomycin-susceptible enterococcal infections (3).

VRE infections impose a huge burden on facilities in metropolitan Detroit. At Detroit Medical Center (DMC) in 2009, 530 of 4,377 (12.1%) isolates of E. faecalis and 846 of 1,150 (73.6%) isolates of E. faecium were VRE (16). Both the proportion of total VREs accounted for by VR E. faecalis and the total absolute number of unique clinical isolates of VR E. faecalis increased significantly from 2003 to 2009 (16). The prevalence of VR E. faecalis at DMC is much higher than in other regions in North America and worldwide. According to National Healthcare Safety Network (NHSN) data, 6.9% of E. faecalis isolates reported in 2006 and 2007 were VRE, compared to over 12% of E. faecalis isolates reported at DMC (17).

Past studies of the epidemiology and outcomes associated with VRE infections were conducted with cohorts consisting predominantly of E. faecium, and relatively little is known about the epidemiology of vancomycin-resistant E. faecalis (5). Only a few studies with small sample sizes have compared the epidemiology of infections due to VR E. faecalis and VR E. faecium (7, 15).

The growing prevalence of VR E. faecalis is worrying, given the fact that there are limited epidemiological data available pertaining specifically to infections due to VR E. faecalis. Better understanding the epidemiology of and outcomes associated with infection due to VR E. faecalis is an essential first step in limiting the continued proliferation and spread of these organisms. The aims of this study were to investigate the epidemiology of bacteremia due to VR E. faecalis and to compare the epidemiology of bacteremia due to VR E. faecalis to that of VR E. faecium.

MATERIALS AND METHODS

Study settings and design.

A retrospective analysis of risk factors and outcomes of bacteremia due to VR E. faecalis was conducted at DMC. The DMC health care system consists of 8 hospitals with more than 2,200 inpatient beds and serves as a tertiary referral hospital for metropolitan Detroit and southeast Michigan. All VR E. faecalis and VR E. faecium clinical isolates from 1 January 2008 to 31 October 2010 were analyzed and reviewed. Institutional review boards at Wayne State University and DMC approved the study before its initiation.

Patients and variables.

The study cohort consisted of all patients who had a positive blood culture for VR E. faecalis or VR E. faecium as identified by the DMC clinical microbiology laboratory. For patients who had >1 blood culture positive for VRE during the study period, only the first episode of VRE bacteremia was analyzed (i.e., unique patient episodes).

Parameters retrieved from patient records included demographics, background conditions and comorbid conditions (including Charlson's scores [6]), recent health care-associated exposures, such as a stay in a health care facility, invasive procedures, or the presence of indwelling devices (13, 20), acute illness indices, including McCabe score (2), and the severity of sepsis levels per systemic inflammatory response syndrome (SIRS) criteria (10). Information regarding infectious clinical syndromes causing bacteremia (sources of infections) was determined according to CDC/NHSN definitions and the Infectious Diseases Society of America (IDSA) guideline (4, 18, 21) and, when present, according to consult notes from the infectious diseases service. For cases categorized as catheter-related bloodstream infections, the information pertaining to catheter removal was extracted from the patient chart. Other information retrieved from patient records included presence of VRE on admission, defined as a hospital stay before VRE isolation of ≤2 days, cocolonization with methicillin-resistant Staphylococcus aureus (MRSA), defined as isolation of MRSA from any body site within 7 days of VRE isolation (before and after), and factors related to antimicrobial therapy, including recent (within 3 months) exposures to antimicrobials, empirical therapy (defined as therapy administered from 48 h before to 72 h following the VRE culture date), the main “consolidative” regimen (i.e., antibiotics provided from 72 h to 14 days following VRE isolation), and the time (in hours) to initiation of effective therapy. Effective therapy was defined based on in vitro activity as reported by the DMC clinical microbiology laboratory, in accordance with the Clinical and Laboratory Standard Institutions (CLSI) criteria (8). In addition, hours to effective therapy (defined by hours between the time of VRE isolation from blood and the initiation of effective therapy) and hours of effective therapy (defined by the duration of effective antimicrobial therapy administered from 48 h before to 2 weeks following the VRE culture date) were analyzed. Last, outcomes, including in-hospital and 90-day mortality, length of hospital stay (LOS), functional status deterioration (defined as deterioration from admission to discharge in one or more activities of daily living [ADL] according to Katz criteria [19]), and discharge to a long-term facility after being admitted from home, were also noted.

Microbiology.

DMC has a single centralized Clinical Microbiology Laboratory, which processes ∼500,000 samples annually. Bacteria are identified to the species level, and susceptibilities to predefined antimicrobials are determined based on an automated broth microdilution system (MicroScan; Siemens AG, Germany) and in accordance with CLSI criteria (8). All MicroScan results are assessed by a medical technologist in a microbiology laboratory, and supplemental tests are utilized where indicated, such as atypical susceptibility patterns of enterococci (e.g., E. faecium sensitive to ampicillin or vancomycin). Surveillance cultures and routine screening for VRE or MRSA were not conducted routinely at DMC during the study period and were excluded from the analysis.

Statistical analysis.

All analyses were performed by using SPSS 18 (SSPS, Inc., Chicago, IL). Bivariate analyses were performed using Fisher's exact test or the chi-square test for categorical variables and the independent samples t test or the Mann-Whitney U test for continuous variables. Multivariable models for outcomes were constructed using logistic regression. All variables with a P value of <0.1 in the bivariate analyses were considered for inclusion in the multivariate analysis. A stepwise selection procedure was used to select variables for inclusion in the final model. The final selected model was tested for confounding variables to assess any additional impact on the model. If a covariate affected the β coefficient of a variable in the model by >10%, then the confounding variable was maintained in the multivariable model. All P values were two-sided. In addition to the examination of statistical significance and confounding variables, the effect modification between variables was evaluated by testing appropriate interaction terms for statistical significance. Here, the percentages displayed are the “valid percent,” which indicates the percent excluding the missing data from the denominator, unless otherwise stated.

For each outcome, logistic regression analysis was conducted adjusting for a propensity score predicting the likelihood of bacteremia due to VR E. faecalis versus VR E. faecium. The propensity score was assigned based on the β coefficients of the variables in the logistic regression model.

RESULTS

During the study period, 302 cases of VRE bacteremia were identified, including 105 (34.8%) cases of bacteremia due to VR E. faecalis and 197 (65.2%) cases of bacteremia due to VR E. faecium. The overall mean age of the study cohort was 61.5 ± 15 years; 162 (53.6%) were male, 218 (72.2%) were African-American, and 179 (59.3%) were admitted from long-term care facilities or transferred from other hospitals. In 140 (46.4%) cases, bacteremia due to VRE was present on admission, i.e., occurring ≤48 h following hospitalization. The most common sources of bacteremia included central venous catheter (i.e., catheter-related bloodstream infection) (n = 109; 36.3%); skin and soft tissue infection (n = 65; 21.5%); intra-abdominal infection (n = 34; 11.3%); urinary tract infection (n = 33; 11.0%); unknown focus (n = 27; 9.0%); contaminant or line colonization (n = 24; 8.0%); infective endocarditis (n = 5; 1.7%); and pneumonia (n = 2; 1.0%) (the two patients with pneumonia had VRE coisolated with Pseudomonas aeruginosa).

The results of a bivariate analysis comparing bacteremia due to VR E. faecalis and VR E. faecium are displayed in Table 1. Variables associated with bacteremia due to VR E. faecalis included African-American race, residence in an institutional setting, dependent functional status at the time of admission, and rapidly fatal status according to McCabe score at the time of admission. Patients with bacteremia due to VR E. faecalis more frequently had comorbid conditions, including diabetes mellitus, heart valve diseases, chronic skin ulcers, and peripheral vascular diseases, and were more likely to have permanent devices and to be on chronic hemodialysis.

Table 1.

Bivariate analysis of vancomycin-resistant Enterococcus (VRE) bacteremia: VR E. faecalis versus VR E. faecium^a

Category	Parameterb	VR E. faecalis (n = 105)	VR E. faecium (n = 197)	Odds ratio (95% confidence interval)	P value
Demographics	Mean age (years) ± SD	62.9 ± 14.9	60.7 ± 15.1		0.210
	No. (%) ≥65 years of age	49 (46.7)	81 (41.1)	1.25 (0.78–2.02)	0.394
	No. (%) of males	56 (53.3)	106 (53.8)	1.02 (0.63–1.64)	1.000
	No. (%) of African-Americans	87 (82.9)	131 (66.5)	2.44 (1.35–4.38)	0.003
	No. (%) in non-private-home residence	57 (54.3)	66 (33.5)	2.36 (1.45–3.83)	0.001
Acute and chronic conditions on admission	No. (%) with dependent functional status	82 (78.1)	125 (63.8)	2.02 (1.17–3.50)	0.013
	No. (%) with cognition impairment	44 (41.9)	77 (39.3)	1.12 (0.69–1.81)	0.659
	No. (%) with rapidly fatal McCabe score (2)	37 (35.6)	46 (23.6)	1.79 (1.06–3.01)	0.031
	No. (%) with diabetes mellitus	60 (57.1)	86 (43.7)	1.72 (1.07–2.78)	0.030
	No. (%) with hemiplegia	17 (16.2)	16 (8.1)	2.19 (1.06–4.53)	0.051
	No. (%) with dementia	27 (25.7)	29 (14.7)	2.01 (1.11–3.61)	0.029
	Charlson's weighted index comorbidity score (6), mean ± SD	5.5 ± 3.1	4.3 ± 3.0		0.000
	No. (%) with Charlson's weighted index co-morbidity score(6) ≥ 4	79 (75.2)	108 (54.8)	2.50 (1.48–4.23)	0.000
	No. (%) with heart valve diseases	21 (20)	16 (8.1)	2.83 (1.40–5.70)	0.013
	No. (%) with chronic skin ulcers	44 (41.9)	57 (28.9)	1.77 (1.08–2.91)	0.029
	No. (%) with peripheral vascular diseases	42 (40)	37 (18.8)	2.88 (1.70–4.90)	0.000
	No. (%) in an immunosuppressive statec	29 (27.6)	61 (31.0)	0.85 (0.50–1.44)	0.598
Exposure to health care settings and environments before VRE isolation	No. (%) hospitalized in the past 3 months	80 (76.2)	142 (72.1)	1.24 (0.72–2.14)	0.495
	No. (%) undergoing chronic hemodialysis	45 (42.9)	38 (19.3)	3.14 (1.86–5.30)	0.000
	No. (%) with surgery or invasive procedure in the past 6 months	72 (68.6)	150 (76.1)	0.68 (0.40–1.16)	0.172
	No. (%) with permanent devicesd	81 (77.1)	128 (65)	1.82 (1.06–3.13)	0.036
	Median LOS (in days) before isolation, days (IQR)	1.0 (0.0–16.5)	5.2 (0.7–17.6)		0.013
	No. (%) with ICU stay in current hospitalization prior to VRE isolation	39 (37.5)	98 (50.3)	0.59 (0.37–0.97)	0.039
Microbiology	No. (%) with VRE present on admissione	62 (59.0)	78 (39.6)	2.20 (1.36–3.56)	0.002
	No. (%) with cocolonization with MRSAf	22 (21.0)	22 (11.2)	2.11 (1.11–4.02)	0.026
	No. (%) with polymicrobial bacteremia	50 (48.1)	65 (33.0)	1.88 (1.16–3.06)	0.013
	No. (%) with ampicillin-resistant isolates	3 (2.9)	195 (99.0)	0.00 (0.00–0.00)	0.000
	Median ampicillin MIC (μg/ml) (IQR)	1 (0.5–1)	16 (16–16)		0.000
	No. (%) with high-level-gentamicin-resistant isolatesg,h	99 (94.3)	42 (21.5)	60.11 (24.64–146.66)	0.000
	No. (%) with high-level-streptomycin resistant isolatesh	76 (72.4)	145 (73.6)	0.94 (0.55–1.60)	0.892
	No. (%) with daptomycin-resistant isolatesg	0 (0.0)	8 (4.3)	0.96 (0.93–0.98)	0.054
	Median daptomycin MIC (μg/ml) (IQR)	1 (1–1)	4 (4–4)		0.000
	No. (%) with linezolid-resistant isolatesg	0 (0.0)	3 (1.5)	0.99 (0.97–1.00)	0.56
	Linezolid MIC (μg/ml) (IQR)	2 (0.5–2)	2 (0.25–2)		0.325
Antimicrobial exposure within 3 months prior to VRE isolation	No. (%) receiving any antibiotics	86 (81.9)	137 (69.5)	1.98 (1.11–3.55)	0.020
	Penicillinsi	32 (30.5)	52 (26.4)	1.22 (0.73–2.06)	0.501
	Cephalosporins	64 (61.0)	94 (47.7)	1.71 (1.06–2.77)	0.030
	Carbapenems	17 (16.2)	25 (12.7)	1.33 (0.68–2.59)	0.485
	Fluoroquinolones	23 (21.9)	27 (13.7)	1.77 (0.95–3.27)	0.075
	Vancomycin	74 (70.5)	76 (38.6)	3.80 (2.29–6.32)	0.000
	Aminoglycosides	14 (13.3)	12 (6.1)	2.37 (1.05–5.34)	0.050
	Trimethoprim-sulfamethoxazole	12 (11.4)	5 (2.5)	4.96 (1.70–14.48)	0.003
	Daptomycin	6 (5.7)	6 (3.0)	1.93 (0.61–6.14)	0.353
	Linezolid	6 (5.7)	7 (3.6)	1.65 (0.54–5.03)	0.386
	Clindamycin	9 (8.6)	19 (9.6)	0.88 (0.38–2.02)	0.837
	Metronidazole	16 (15.2)	24 (12.2)	1.30 (0.66–2.56)	0.479
Severity of illness indices at the time of VRE isolation	No. (%) with reduced consciousness	56 (55.4)	96 (49.2)	1.28 (0.79–2.08)	0.329
	No. (%) with severe sepsis/septic shock/multiorgan failure per SIRS criteria (10)	19 (18.8)	32 (17.8)	1.07 (0.57–2.01)	0.872
	No. (%) transferred to ICU	25 (24.3)	55 (28.2)	0.82 (0.47–1.41)	0.495
	No. (%) with necessitated acute mechanical intubation/ventilation	28 (27.2)	53 (27.2)	1.00 (0.59–1.71)	1.000
	No. (%) who developed acute renal failurej	44 (42.3)	67 (34.4)	1.40 (0.86–2.28)	0.209
	No. (%) who developed acute liver injuryk	2 (2.0)	24 (12.3)	0.15 (0.03–0.63)	0.002
Infectious clinical syndromes causing bacteremia	No. (%) with catheter-related bloodstream infection	47 (44.8)	62 (31.8)	1.76 (1.05–2.96)	0.022
	No. (%) with skin and soft tissue infection	25 (23.8)	40 (20.5)	1.23 (0.67–2.24)	0.480
	No. (%) with intra-abdominal infection	9 (8.6)	25 (12.8)	0.65 (0.27–1.52)	0.280
	No. (%) with urinary tract infection	8 (7.6)	25 (12.8)	0.57 (0.23–1.38)	0.179
	No. (%) with focus unknown	7 (6.7)	21 (10.8)	0.60 (0.22–1.55)	0.250
	No. (%) with contamination or catheter colonization	7 (6.7)	17 (8.7)	0.76 (0.27–2.02)	0.548
	No. (%) with infective endocarditis	1 (1.0)	4 (2.1)	0.46 (0.02–4.47)	0.662
	No. (%) with pneumonial	1 (1.0)	1 (0.5)	1.88(0.12–30.44)	1.000
Antimicrobial therapy	No. (%) with effective therapy administeredm	89 (84.8)	131 (70.1)	2.38 (1.28–4.41)	0.005
	Hours to effective therapy (mean ± SD)	59.7 ± 70.3	67.2 ± 63.2		0.134
	Hours of effective therapy (mean ± SD)	255.8 ± 147.4	210.4 ± 126.7		0.017
Outcomes	No. (%) with in-hospital mortality	22 (21)	70 (35.5)	0.48 (0.28–0.84)	0.009
	No. (%) with 3-month mortality	7 (8.5)	18 (14.5)	0.55 (0.22–1.38)	0.276
	No. (%) with functional status deterioration	23 (27.7)	58 (45.7)	0.46 (0.25–0.83)	0.009
	No. (%) discharged to LTCF after being admitted from home	7 (8.5)	18 (14.5)	0.55 (0.22–1.38)	0.351
	No. (%) with additional hospitalizations in 6 months after VRE isolation	63 (75.9)	95 (74.8)	0.94 (0.50–1.79)	1.000
	No. (%) with surgeries/invasive procedures within 3 months after VRE isolation	54 (56.8)	59 (33.1)	2.66 (1.59–4.43)	0.000
	Median total LOS (days) (IQR)	17.0 (8.3–39.8)	17.8 (8.2–32.9)		0.686
	Median LOS after isolation (days) (IQR)	11.5 (7.0–21.6)	8.5 (4.2–18.4)		0.007
	Median LOS after isolation excluding those who died (days) (IQR)	10.9 (7.2–21.8)	9.1 (5.2–20.1)		0.099
	No. (%) with LOS after isolation of ≥7 days (excluding those who died), n (%)	66 (79.5)	83 (65.4)	2.06 (1.08–3.93)	0.030
	Median ICU LOS after isolation (days) (IQR)	0.8 (0.0–11.8)	0.9 (0.0–6.1)		0.845

^aAbbreviations: ICU, intensive care unit; IQR, interquartile range; LOS, length of hospital stay; LTCF, long-term-care facility; MRSA, methicillin-resistant Staphylococcus aureus; SIRS, systemic inflammatory response syndrome.

^bFor each parameter, “No. (%)” refers to the number and the percentage of patients for whom data were available, i.e., excluding the missing cases.

^cIncludes one ore more of the following: neutropenia (<500 neutrophils) at time of culture, glucocorticoid/steroid use in the past month, chemotherapy in the past 3 months, radiotherapy in the past 3 months, HIV infection, transplantation, or anti-tumor necrosis factor alpha therapy in the past 3 months.

^dChronic or permanent devices (e.g., tracheotomies, central lines, urinary catheters, or orthopedic external fixators) that were in place at least 48 h prior to VRE isolation.

^eDefined as a hospital stay of ≤2 days before VRE isolation.

^fDefined as isolation of MRSA from any body site within 7 days of VRE isolation (before or after).

^gIn vitro susceptibility test results were not available for the following numbers of isolates: daptomycin, 12 (6.1%) VR E faecium and 9 (8.6%) VR E. faecalis isolates; linezolid, 1 (0.5%) VR E. faecium and 26 (24.8%) VR E. faecalis isolates; high-level gentamicin, 2 (1%) VR E. faecium isolates.

^hHigh-level resistance is defined as a MIC of ≥500 μg/ml for gentamicin or ≥2,000 μg/ml for streptomycin.

ⁱPenicillins include β-lactam–β-lactamase inhibitor combinations.

^jDefined as elevation of creatinine to 50% or more above baseline.

^kDefined as the presence of at least one of the following conditions: transaminases doubled compared to baseline value, INR above 1.5 without exposure to anticoagulants, and total bilirubin tripled compared to baseline.

^lThese patients had P. aeruginosa coisolated.

^mDefined as an antibiotic administered in the 14 days following the VRE culture that demonstrated in vitro efficacy against the VRE pathogen.

Additional variables associated with bacteremia due to VR E. faecalis compared to VR E. faecium were cocolonization with MRSA, polymicrobial bacteremia, a significantly shorter duration of hospitalization stay from the time of admission to occurrence of VRE bloodstream infection, and a higher frequency of bloodstream infections that were present on admission (i.e., isolated within the first 48 h of hospitalization). Regarding polymicrobial bacteremia, coagulase-negative staphylococcus (CNS) was the most common pathogen isolated from the same blood culture as VRE (20 [40%] of VR E. faecalis cultures and 22 [33.8%] of VR E. faecium cultures; P = 0.46), and the next most common pathogen isolated was MRSA. MRSA isolation from the same blood culture was more common in bacteremia due to VR E. faecalis than VR E. faecium (10 [20%] and 5 [7.7%], respectively; P = 0.05). Subcohort analysis of cases of bacteremia due to VRE which were present on admission revealed that patients with bacteremia due to VR E. faecalis were more likely to be on chronic hemodialysis than patients with VR E. faecium (34 [54.8%] and 21 [26.9%], respectively; odds ratio [OR], 3.30 [1.63 to 6.69], P = 0.001). Among the clinical infectious syndromes, only catheter-related bloodstream infection was significantly more common among patients with bacteremia due to VR E. faecalis. The catheter was removed after the bacteremic episode during hospitalization in 30 (88.2%) patients with bacteremia due to VR E. faecalis and in 45 (72.6%) of patients with bacteremia due to VR E. faecium (P = 0.12). There was no statistical difference observed between VR E. faecalis and VR E. faecium regarding multiple isolations from the same patient of the same species of VRE from blood during the study period (21 [20%] of VR E. faecalis infections and 42 [21.3%] of VR E. faecium infections; P = 0.83).

As expected, VR E. faecalis isolates were more susceptible to ampicillin than were isolates of VR E. faecium (102 [97.1%] of VR E. faecalis, 2 [1%] of VR E. faecium). However, these isolates displayed high-level resistance to gentamicin more frequently than VR E. faecium. No vancomycin-resistant S. aureus (VRSA) strains were isolated from the study cohort. Recent antimicrobial exposures were more common among patients with bacteremia due to VR E. faecalis, and these patients were more frequently exposed to cephalosporins, trimethoprim-sulfamethoxazole, and vancomycin than were patients with bacteremia due to VR E. faecium. The patients with bacteremia due to VR E. faecalis were more likely to receive effective antibiotics and to receive longer durations of effective therapy. However, among patients who received effective therapy, the time to effective therapy did not differ between the groups.

In multivariate analysis, independent predictors for bacteremia due to VR E. faecalis compared to VR E. faecium were previous antimicrobial exposures, chronic hemodialysis, cocolonization with MRSA, and polymicrobial bacteremia. ICU stay during the current hospitalization prior to VRE isolation was independently associated with bacteremia due to VR E. faecium (Table 2). Based on this model, a propensity score predicting the likelihood of bacteremia due to VR E. faecalis as opposed to VR E. faecium was developed for each subject based on the β coefficients of variables in the multivariate model.

Table 2.

Multivariate analysis and propensity score predicting the likelihood of bacteremia due to vancomycin-resistant E. faecalis versus vancomycin-resistant E. faecium

Variable	Odds ratio (95% confidence interval)	P value	Propensity score (β coefficient)
Rapidly fatal McCabe score (2)	1.80 (0.99–3.28)	0.053	0.59
Charlson's weighted index comorbidity score (6) ≥ 4	1.88 (0.85–4.20)	0.122	0.63
African-American race	1.90 (0.99–3.61)	0.052	0.64
Heart valve diseases	1.91 (0.87–4.23)	0.109	0.65
Polymicrobial bacteremia	1.92 (1.10–3.34)	0.022	0.65
Non-ICU stay in current hospitalization prior to VRE isolation	1.94 (1.09–3.43)	0.023	0.66
Antimicrobial exposure within 3 months before VRE isolation	1.97 (1.02–3.79)	0.043	0.68
Cocolonization with MRSA	2.41 (1.11–5.27)	0.027	0.88
Chronic hemodialysis	2.72 (1.50–4.93)	0.001	1.00

In the bivariate outcomes analysis (Table 1), patients with bacteremia due to VR E. faecium were more likely to have acute liver injury at the time of infection. Patients with bacteremia due to VR E. faecium were more likely to die in the hospital and to experience a deterioration in functional status after the onset of bacteremia. Bacteremia due to VR E. faecalis was associated with longer duration of hospitalization after VRE isolation, and a higher frequency of surgeries and invasive procedures in the 3 months following VRE isolation than was bacteremia due to VR E. faecium.

In multivariate analysis of clinical outcomes, after controlling for the propensity score (Table 3), bacteremia due to VR E. faecalis was associated with a >2-fold-lower in-hospital mortality. However, bacteremia due to VR E. faecalis was associated with a >2-fold increase in likelihood of prolonged duration (≥7 days) of hospitalization after culture, although total duration of hospitalization was similar compared to patients with VR E. faecium. Controlling for both administration of effective therapy and propensity score in multivariate analysis did not affect the significance of the association between bacteremia due to VR E. faecalis and decreased mortality (P = 0.02; OR = 0.49; 95% CI, 0.26 to 0.89).

Table 3.

Multivariate analysis of clinical outcomes associated with bacteremia due to vancomycin-resistant E. faecalis compared to vancomycin-resistant E. faecium^a

Variable	No. (%) of patients with:		Bivariate analysisb		Multivariate analysisc
	VR E. faecalis (n = 105)	VR E. faecium (n = 197)	OR (95% CI)	P value	OR (95% CI)	P value
In-hospital death	22 (21.0)	70 (35.5)	0.48 (0.28–0.84)	0.009	0.47 (0.26–0.85)	0.013
Functional status deterioration	23 (27.7)	58 (45.7)	0.46 (0.25–0.83)	0.009	0.68 (0.36–1.32)	0.257
Duration of hospitalization after VRE bacteremia (excluding those who died) ≥ 7 days	66 (79.5)	83 (65.4)	2.06 (1.08–3.93)	0.030	2.03 (1.00–4.12)	0.050

^aControlled for propensity score for the likelihood of having VR E. faecalis.

^bAssociation between outcome and VR E. faecalis.

^cControlled for propensity score: independent association between outcome and VR E. faecalis.

DISCUSSION

Bacteremia due to VR E. faecalis was independently associated with a >2-fold-lower in-hospital mortality than bacteremia due to VR E. faecium, although patients with bacteremia due to VR E. faecalis had a greater degree of acute severity of illness (as determined by the McCabe score) than patients with VR E. faecium. The findings of our study that VR E. faecium bacteremia is associated with worse prognosis are in accordance with previous reports (7, 15). It was discussed previously that the higher mortality might be due to underlying disease or patient characteristics (23). However, in multivariate analysis adjusted by propensity score and receipt of effective therapy, bacteremia due to VR E. faecium remained significantly associated with mortality. This association might be related to virulence differences between species. Such virulence differences have been reported (12, 27). Another possible explanation for the lower mortality risk among patients with VR E. faecalis is the efficacy of β-lactam therapeutics for VR E. faecalis infections but not for infections due to VR E. faecium. In this study, penicillins were most frequently used as main therapy for bacteremia due to VR E. faecalis (57.1% of patients were treated with β-lactam therapy), whereas linezolid, a bacteriostatic agent, was most frequently used as main therapy for bacteremia due to VR E. faecium (48.7% of patients were treated with linezolid). Many uncertainties exist regarding the optimal treatment for VRE infections (1), and further studies are needed.

In a bivariate analysis, bacteremia due to VR E. faecalis, in addition to being associated with decreased mortality, was associated with decreased risk for functional deterioration and prolonged duration of hospitalization after VRE isolation. Thus, patients with bacteremia due to VR E. faecalis were more likely to survive and experience prolonged durations of hospital stay after VRE bacteremia. In addition, patients who developed VR E. faecalis bacteremia had more frequent comorbid conditions and a greater severity of acute illness than patients with bacteremia due to VR E. faecium. Thus, patients with bacteremia with VR E. faecalis were at increased risk for MRSA cocolonization and prolonged duration of hospitalization and were more likely to have a greater severity of chronic and acute severity of illness.

Of note, VR E. faecalis blood isolates from study subjects displayed high-level resistance to gentamicin more frequently than VR E. faecium isolates in our study. Data pertaining to high-level resistance to gentamicin among VR E. faecalis are limited; previous reports on bacteremia due to VRE did not find a significant difference in the frequency of high-level resistance to gentamicin between VR E. faecalis and VR E. faecium (7, 15). The number of VR E. faecalis isolates in the current study displaying high-level resistance to gentamicin was much greater than that in previous studies (11, 14).

Another key finding of this study was that patients with bacteremia due to VR E. faecalis were more frequently exposed to cephalosporins, trimethoprim-sulfamethoxazole, and vancomycin than were patients with VR E. faecium bacteremia. These findings identify agents which impart a selective pressure for the selection of VR E. faecalis. The finding in this study that exposure to cephalosporins, but not other β-lactam antibiotics, was more common among patients with bacteremia due to VR E. faecalis suggests that exposure to broad-spectrum cephalosporins might provide a selective advantage for isolation of VR E. faecalis. Further studies are needed to determine the association between cephalosporins, as well as other antimicrobials, and VR E. faecalis.

VRE strains can horizontally transfer the vanA gene complex to S. aureus, resulting in VRSA in patients cocolonized with VRE and S. aureus (24). In the majority of VRSA cases studied, VR E. faecalis served as the vanA donor to S. aureus (24) and cocolonization with S. aureus and VR E. faecalis, rather than VR E. faecium, was associated with VRSA isolation (25). Since 2002, 12 cases of VRSA have been reported in the United States (9, 26), and 8 have occurred in Southeast Michigan. It is unclear why VRSA has a predilection for this region, although clues have recently emerged (28). VRE isolates with an Inc18-like vanA plasmid, which has been reported to facilitate the transfer of the vanA gene to S. aureus, have been reported to be more prevalent in Michigan (3.9%) than other U.S. states (0.6%) and also are more commonly identified among VR E. faecalis (identified in 12.5% of isolates in Southeast Michigan) than VR E. faecium (identified in 1.0% of isolates) (28). Thus, the epidemiology of VR E. faecalis revealed in this study, in the location in the world where VRSA isolation is most prevalent, might add important implications for the endemicity of VRSA in this region.

To the best of our knowledge, this study is the largest epidemiologic investigation of bacteremia due to VR E. faecalis. This study utilized advanced epidemiological tools (including a propensity score) to control for confounding and overcome inherent weaknesses associated with retrospective studies.

A limitation of this study, due to its retrospective nature, was that isolates were unavailable for further molecular analysis. This precluded the conduction of population analysis. Thus, we were unable to investigate the presence of potential clonal relationships among the isolates or the presence of high-risk clonal complexes. In addition, we were unable to analyze mechanisms of vancomycin resistance.

An ominous finding in this study was that bacteremia due to VR E. faecalis was detected within the initial 48 h of hospitalization in more than 50% of patients with VR E. faecalis and was significantly more likely to be “present on admission” than was bacteremia due to VR E. faecium. This suggests the presence of a “nonhospital” pool of VR E. faecalis, which is concerning. As attempts are made to limit the spread of vancomycin resistance, more attention needs to be focused on outpatient, health care-associated settings. Patients with VR E. faecalis bacteremia were more likely to be on chronic dialysis and permanently reside in “non-private-home” institutions, and these populations and nonhospital settings represent opportune targets for infection control and antimicrobial stewardship and resistance prevention strategies.