Introduction

Enterococcus faecalis and Enterococcus faecium have become increasingly common causes of healthcare-associated infections (HAIs) in the United States (US) and Europe [1,2,3]. Six-year surveillance on Gram-positive organisms causing bacteraemia in US and European hospitals reported that 10.6% and 9.4% of bloodstream infections (BSI) were due to E. faecalis and E. faecium, which constitute, respectively, the second and the third pathogen in terms of frequency after Staphylococcus aureus [4]

The incidence of enterococci as the causative agent of BSI (E-BSI) continues to be also elevated during the COVID-19 pandemic, probably because of a more extensive empirical use of corticosteroids and cephalosporins which have variable in vivo activity against Enterococcus spp. [5]. In COVID-19 subjects, E. faecalis and E. faecium represented 18% and 9% of all BSI, according to a recent retrospective study by Giacobbe et al. [5].

Moreover, it is well established that E-BSI retained high in-hospital mortality estimated between 30 to 50% [6,7,8,9,10] and it has been identified as an independent predictor of increased 30-day mortality in elderly patients [11].

Furthermore, enterococci also play an increasingly important role in infective endocarditis, being the third leading causative agent in high-income countries and more frequently occurring in aged patients with high complexity of comorbidities [12].

Antibiotic combination is usually needed to achieve bactericidal activity in enterococcal-related endocarditis, although the possibility of combining antibiotics has become more challenging due to the increasing rate of acquired antimicrobial resistance among enterococci [13, 14]. Antibiotics’ timing and selection represent another critical issue, since two meta-analyses showed that inappropriate and delayed therapy in clinical practice is associated with more unfavourable outcomes and higher mortality [13, 14].

This retrospective study aimed to evaluate the clinical outcome and risk factors for mortality of patients with E-BSI admitted to internal medicine wards (IMW) in a large teaching hospital in northwest Italy over 2 years.

Methods

This retrospective study was conducted in IMWs, in a 1200-bed academic hospital with a primary and secondary referral (City of Health and Sciences, Molinette, Turin, Italy).

All adults with positive blood culture for E. faecalis or E. faecium between 1st January 2015 and 31st December 2016 were enrolled in the study, and the demographic, clinical and microbiological data were collected.

BSI were defined as community acquired (CA-BSI), hospital acquired (HA-BSI) or healthcare associated (HCA-BSI), according to the time and setting of the onset of infection. HA-BSI was defined as a first positive blood culture ≥ 48 h after hospital admission or within 48 h of discharge from the hospital. HCA-BSI was defined as a first positive blood culture obtained < 48 h of admission in patients with recent and/or frequent exposure to healthcare settings and procedures. Community-acquired infections are those with a first positive culture obtained < 48 h of admission that do not fulfil criteria for the healthcare-associated classification.

The Sequential Organ Failure Assessment (SOFA) score and quick-SOFA (q-SOFA) were used to assess the severity of the disease [15], and the Charlson Comorbidity Index [16], a weighted index of 17 coexisting medical conditions, was used as the measure of the risk of mortality of each comorbidity in this population.

According to Sepsis-2 guideline [17], sepsis was defined as the presence of at least two systemic inflammatory response syndrome (SIRS) criteria. Severe sepsis was defined accordingly as Sepsis-2, and sepsis was defined according to Sepsis-3 guideline [15].

Antibiotic therapy was defined as appropriate if all the following criteria were present: the administration of an active antibiotic against Enterococcus spp., the administration of an active combination therapy (at least two drugs simultaneously, except daptomycin) in infective endocarditis and a duration of at least 15 days of an active antibiotic treatment [18,19,20,21]. Antimicrobials considered active in monotherapy or in combination therapy against Enterococcus spp. were penicillin G, ampicillin, ampicillin/sulbactam, amoxicillin, amoxicillin/clavulanate, piperacillin/tazobactam, imipenem, vancomycin, teicoplanin, daptomycin, ceftriaxone, gentamicin, tigecycline and linezolid.

In case of death or discharge within 15 days from the beginning of antibiotic therapy, the therapy was considered appropriate if criteria 1 and 2 were satisfied. Moreover, the 21-day mortality was investigated.

The Vitek MS MALDI-TOF mass spectrometry (bioM erieux, Craponne, France) was routinely used for identifying Enterococcus spp. BSI, whereas the Vitek 2 automated system (bioMerieux, Craponne, France) was routinely used for susceptibility testing. The results of the susceptibility tests were interpreted according to the criteria of the European Committee on Antimicrobial Susceptibility Testing (EUCAST) (breakpoint tables for interpretation of minimum inhibitory concentrations [MIC] and zone diameters, version 10.0, 2020; http://www.eucast.org). For daptomycin, isolates were considered susceptible in the case of MIC 4 mg/L. Specifically, daptomycin was used at high dose, in combination therapy, in Enterococcus spp. strains reporting MIC < 4 mg/L.

The need for informed consent was waived due to the study’s retrospective nature, which the medical direction of the hospital approved (PROT.N.0115709). Data were collected according to the Italian laws on privacy.

At the time of blood cultures, demographic and clinical characteristics were summarized using absolute and percentage frequencies (qualitative variables) or using the mean and standard deviation or percentile (quantitative variables). Categorical variables were tested in the Chi-square test. Numeric variables were tested by Student’s t test or Wilcoxon test. Overall survival was defined as the time between the first blood sample culture and hospital death. The equivalence of survival curves was tested through the log-rank test. A Cox proportional-hazards model was used for multivariate analysis. Statistical analysis was made using STATA (StataCorp. 2017. Stata Statistical Software: Release 15. College Station, TX: StataCorp LP).

Results

Two hundred and eighteen episodes of E-BSI were included in the study. Due to incomplete medical records, the analysis was performed on 201 patients. The infection rate was 2.4/1000 days of hospital admission. Most E-BSIs were nosocomial related (78.1% HA-BSI, 12.4% HCA and 8.9% CA).

As shown in Table 1, E-BSI occurred mainly in female (61.7%), with a median age of 68.

Table 1 Descriptive analysis of demographic, clinical and microbiological data sorted according to the appropriateness of treatment in bloodstream infections due to Enterococcus spp.
Full size table

98.5% of all patients had comorbidities, and the median Charlson Comorbidity Index, adjusted for age, was five (range between four and six). Cancer was the most common comorbidity (N = 67, 33.3%), followed by diabetes (N = 52, 25.9%) and chronic obstructive pulmonary disease (COPD) (N = 51; 25.4%). Approximately 117 patients (58.5%) were hospitalized during the previous 3 months (67.8% in a medical ward, 12.9% in a surgical ward and 21.4% in an intensive care unit).

Patients with E-BSI frequently had invasive devices, mainly a central venous catheter (CVC) (73%) or a urinary catheter (61.7%) (Table 1).

21% of patients were on immunosuppressive therapy other than chronic corticosteroid treatment (12.4%) or had undergone solid organ or hematopoietic stem cell transplant (9.4% and 1.5%, respectively).

Among all E-BSI cases, in 47.9%, E. faecium was responsible, whereas E. faecalis was isolated in 52.1%. E. faecalis was resistant to ampicillin (N = 7; 3.1%) or vancomycin-resistant enterococci (VRE) N = 5; 2.2%. On the other hand, E. faecium were ampicillin resistant in 22.2% and VRE strains were 23.3% (N = 39).

Regarding the source of infection, abdominal infections were the most frequent (21.2%), followed by pulmonary infections (18.0%) and urinary tract infections (16.0%), according to clinical presentation and radiological features. Infective endocarditis was diagnosed in 5.5%, but overall only 58.6% of patients did not undergo echocardiography.

Following Sepsis-2 criteria, sepsis occurred in 41.7% of participants and septic shock in 8.0%. Data on severe sepsis have not been analysed for lack of complete data.

According to Sepsis-3 criteria, 15.9% of participants developed sepsis, and 5.4% developed septic shock.

Fifty-two patients (25.8%) received appropriate therapy. The most frequent antibiotic used was vancomycin (17.8%), followed by piperacillin/tazobactam (15.9%) alone or in combination (Table 2).

Table 2 Antibiotic molecules used in monotherapy or in combination during empirical or targeted therapy in bloodstream infections due to Enterococcus spp.
Full size table

Risk factors for mortality

According to the univariate analysis (Table 3), mortality was associated with resistance to vancomycin (p = 0.042, 95% CI [1.02, 3.31]), sepsis (p = 0.003, 95% CI [1.36, 4.58]) and septic shock (p = 0.007, 95% CI [1.39, 8.05]). SIRS presentation, instead, was linked to lower mortality (p = 0.047, 95% CI [1.39, 8.05]). The multivariate analysis (Table 4) confirmed resistance to vancomycin (OR 2.09, 95% CI [1.10, 3.99], p = 0.025), sepsis (OR 2.57, 95% CI [1.37, 4.82], p = 0.003) and septic shock (OR 3.82, 95% CI [1.53, 9.55], p = 0.004) as risk factors for death and sepsis, according to Sepsis-2 criteria as a protective factor (OR 0.50, 95% CI [0.26, 0.99], p = 0.047).

Table 3 Univariate analysis of risk factors for 21-day mortality associated with Enterococcus spp. bacteraemia
Full size table
Table 4 Multivariate analysis of risk factors for 21-day mortality-associated with Enterococcus spp. bacteraemia
Full size table

Survival

No difference was found between the group that received appropriate therapy and the group that did not receive appropriate therapy, except for endocarditis, which was more frequent in the first group (p = 0.003).

During hospitalization, 53 people (26%) died. Their median survival time was 65 days after the first blood draw. The other 75.4% lived at least 21 days, as shown by the Kaplan-Meier curve. After 28 days of follow-up, the survivorship curve showed no difference between the patients who received appropriate treatment and the patients who did not (Fig. 1).

Fig. 1
figure 1

Survival at 28 days from time from blood culture sample

Full size image

Discussion

Enterococci are Gram-positive pathogens frequently involved in BSI related to high in-hospital mortality, especially in high comorbid, frail patients [6, 7]. There is a high burden of widespread, difficult-to-treat multi-drug-resistant E. faecium, and E. faecalis is increasingly resistant to first-line antibiotic choices [7, 8]. While evidence suggests that combination therapy can decrease mortality in patients with enterococcal infective endocarditis, there is no strong evidence on the best therapy for patients with an E-BSI without endocarditis [13, 14].

We retrospectively reported epidemiological, microbiological and clinical features of the population with E-BSI in a single-centre, 2-year observational study. In our study, almost 80% of E-BSIs were hospital acquired, whereas, in similar studies, the incidence was about 40% [7, 22]. Nonetheless, the healthcare-associated incidence of E-BSI was much lower in our study than that in recent literature (about 10% vs 45%) [22]. Moreover, as previously described [7, 8, 22], we observed that BSI frequently occur in the elderly with several comorbidities. Among comorbidities, solid tumour, cardiovascular diseases and diabetes were the three most frequent underlying diseases. Interestingly in this series, COPD (25.3%) was reported at a higher rate than previous studies [7, 9]. This data might be explained, according to literature, by a frequent need for hospitalization, chronic corticosteroid treatment, and airways and gut bacterial colonization in people with COPD accounting for a higher risk of E-BSI. Inghammar et al. [23] described a 2.5-fold increase in bacteraemia in the COPD population in Sweden, and among pathogens involved, E-BSI was more frequent than in the control group [HR 1.9, 95% CI (1.3, 2.6)]. In previous works, COPD patients, notably in the ICU, reported frequent VRE colonization and higher related BSI rates [24, 25]. Moreover, a high comorbidity burden such as the one in our population with a median Charlson Comorbidity Index (CCI) of 5 must be taken into account in prognostic considerations given that in E-BSI a Charlson Comorbidity Index (CCI) higher than four has been independently associated with 30-day mortality [26].

Iatrogenic immunosuppression is a common risk for enterococcal-related infections, with up to 33% of these patients being admitted to receiving immunosuppressive therapy other than chronic corticosteroids, and 9.4% being admitted to receiving procedures resulting in more aggressive immune-suppression such as solid organ transplant. Besides, in COVID-19 subjects, the acute use of anti-inflammatory agents has been recently identified as an independent predictor of the development of E-BSI [4].

In our cohort, up to 54.7% of participants received enteral or parenteral nutrition during hospitalization. Enteral feeding has reported previously contrasting findings as a protective [23] or risk factor [7] for enterococcal infection, whereas parenteral nutrition is a well-known factor associated with catheter-related BSI in patients with central venous catheters (CVC) used to administer parenteral nutrition [28,29,30]. It is relevant to consider that the relative majority, almost one third, of our population received parenteral nutrition. Also, almost 80% of patients had a CVC. The primary source of infections was intraabdominal, including peritonitis, gastroenteritis and biliary tract infections, as stated in several studies [7,8,9, 24,25,26,27].

We also described a surprisingly high proportion of infection supposed to originate from the respiratory tract. Several authors have suggested a correlation between enterococcal pneumonia and enteral feeding, based on a higher risk of oropharynx colonization and the use of broad-spectrum antibiotics [28,29,30]. In this study, the number of BSI with pulmonary sources (18%) was the highest reported, to our knowledge, with more than 50% of the total population reporting an antibiotic treatment in the previous 6 months and more than 20% having been submitted previously to IMW admission to invasive mechanical ventilation.

Enterococcus spp. is commonly known as a healthcare-associated pathogen, according to the literature [22,23,24,25,26,27,28,29,30,31]. The in-hospital incidence combined with the advanced age of those infected, the high rate of medication and the high rate of comorbidities suggest that the infection typically occurs in frail patients.

E. faecalis was the most typical species isolated in this work. Furthermore, the prevalence of E. faecium was almost 50%, appearing higher among patients in our study than reported in other studies [8,9,10,11]. An estimated 3.1% of strains of E. faecalis and 22.2% of strains of E. faecium were resistant to ampicillin; moreover, resistance to gentamicin was found in 16.2% and 47.3% of strains, respectively.

The susceptibility reported for ampicillin is far from that reported in Asian literature [10], which reported susceptibility against 100% of E. faecalis isolates and from 3.8% up to 10.8% of E. faecium isolates. On the contrary, studies from Denmark and Canada reported susceptibility to ampicillin in 12.3% and 30% of E. faecium, respectively, and in 1.3% and 0.4% of E. faecalis, respectively [8, 9].

Nonetheless, ampicillin-resistant E. faecium was not reported frequently in previous studies and outside Europe [2, 4, 8,9,10,11]. This data did not likely represent Italy’s basic epidemiology [3, 5, 7, 11]; therefore, more recent homogeneous studies are warranted.

Our study reported VRE in 2.2% of E. faecalis strains and 23.3% of E. faecium strains. The National Healthcare Safety Network in the US reported a higher rate, with 6.9% and 80% of vancomycin-resistant faecalis and faecium, respectively [2]. Among Danish and Canadian studies [8, 9], VRE data in E. faecalis were 1.5% and 1%, respectively. These studies also reported lower rates of E. faecium VRE than us, at 1.3% and 7% in each group [8, 9]. Antibiotic resistance for ampicillin and vancomycin in our study was higher for both E. faecalis and E. faecium than that reported in northern European and Canadian literature; however, it was far from the high rates in the US, notably for VRE strains [2, 4, 8, 9].

Although 59% of patients in our study did not undergo echocardiography, the prevalence of endocarditis is in line with the literature where Enterococcus spp. is well known to be one of the most frequent causes of endocarditis [12]. Many infectious disease consultations discuss the importance of defining the source of infection, initiating appropriate antimicrobial therapy, managing treatment duration and de-escalation of the therapy once culture isolates are obtained [32, 33].

However, the management of BSI does not always occur after an infection specialist advises it [32, 33]. We demonstrated in enterococcal endocarditis a significant difference in survival when appropriate treatment is administered. Unfortunately, we did not adjust for inappropriate therapy as a confounding factor for survival. Appropriate case management driven by infectious diseases consults includes repeated cultures to assess clearance and ultrasonography to rule out endocarditis in E-BSI [32, 33]. These approaches might also have a role in mortality as previously shown [8,9,10,11].

The second aim of this study was to determine if combination therapy decreases mortality in patients with E-BSI. Only one out of four patients received appropriate therapy because clinicians stopped the therapy early; they decided only to use one drug or choose a drug not active against Enterococcus spp. To define the role of combination therapy, we identified differences between the number of patients who received appropriate therapy and those who received inappropriate therapy. Each variable was associated with choosing one of the two therapeutic regimens, except for diagnosed endocarditis which was more frequent in the group that received appropriate therapy. This finding can be explained considering that the length of treatment for endocarditis, antimicrobial choice and combinations could rely on many studies and guidelines [18,19,20,21].

At the univariate and multivariate analysis, we confirmed some risk factors for mortality previously described in the literature, such as resistance to vancomycin, sepsis and septic shock [13, 14]. On the contrary, appropriate antibiotic therapy was not associated with lower mortality in our casuistic analysis. This finding is likely since the enrolled patients who received appropriate therapy could still have a poor outcome due to their multiple comorbidities. Therefore, the high medicalization and frailty of these patients may be confounding factors. Furthermore, we found that sepsis seems to be protective for mortality: this result could be due to the fact that Sepsis-2 criteria excluded more critically ill patients. Unfortunately, we are not able to control the confounding factor of the use of combination therapy in patients with sepsis.

A survival of 28 days in 69.94% of patients seems to be higher than in other studies, in which a 21-day mortality was estimated in 21.4–24.8% of participants [13, 14]. However, it is necessary to consider that extensive literature has valued a 30-day survival rate showing a range of mortality from 21% up to 34% to be higher for E. faecium bacteraemia [13, 14].

There are several limitations to this study. First, this is a single-centre study that may not accurately reflect the general demographics of Italy. Second, the recurrence of BSI, surveillance blood culture and performance of surgical source control, especially in intraabdominal infections, were other missing data. Moreover, recently EUCAST has warned about the use of daptomycin in the treatment of Enterococcus spp. with daptomycin MIC of 4 mg/L, despite that the warning was published after the study design and after our data collection.

Our retrospective study demonstrates that many E-BSI sources in clinical practices within IMWs are not always adequately investigated, and the adequate rule-out of intracardiac vegetations lacks in E-BSI. Despite these elements, inappropriate therapy did not decrease survival in this population. We reported a higher rate of E. faecium than other studies. While the rate of ampicillin resistance was lower than expected, the VRE rate was higher than that of Asian and northern European countries [4, 6, 10]. This resistance pattern is consistent with the local epidemiology and the nosocomial acquisition of the infection for the large majority of our cohort. It reflects the microbiological counterpart of the features of the elderly comorbid population with a lengthy history of medicalization and recent previous antibiotic exposure that often populates the internist wards.