Bacteria belonging to the genus Enterobacter spp. are Gram-negative Proteobacteria of the family Enterobacteriaceae and currently comprise 22 species [1]. They represent a diverse group which are widely distributed in nature [2] and possess multiple mechanisms to allow survival in a variety of environmental niches [3]. In humans, they may cause a wide variety of clinical infections and are a common cause of bacteraemia [2], especially within adult and neonatal intensive care units (ICUs) [4, 5].

The annual incidence of bacteraemia caused by Enterobacter spp. has been reportedly increasing in some parts of the world [6]. This bacterial genus presents particular challenges for the selection of optimal therapy due to the presence of chromosomally encoded AmpC beta-lactamase enzymes [7]. These enzymes are able to hydrolyse many beta-lactams, including third generation cephalosporins and may be induced by beta-lactam exposure. Furthermore, AmpC gene expression can become constitutively de-repressed by mutational loss of regulatory genes, leading to high-levels of AmpC production and a phenotype that demonstrates in vitro resistance to most beta-lactams and beta-lactam/beta-lactamase inhibitor (BLBLI) combination agents, except cefepime and carbapenems. [2] Such variants are usually present at low levels (e.g. between 10⁻⁵ to 10⁻⁷ of the total bacterial population) but may be rapidly selected during antibiotic therapy [2]. This phenomenon has been best described with the use of third generation cephalosporins (3GCs). In a prospective study of Enterobacter bacteraemia, emergence of resistance developed during treatment in 6 of 31 (19%) bacteraemia episodes treated with 3GCs [8]. Similar outcomes were replicated in a later study [9], although relapse rates have been reported as lower in other studies [10–12]. In a large cohort of patients from South Korea, emergent resistance during therapy with 3GCs was more likely to occur with a complex focus of infection (particularly malignant bile duct obstruction) but was never seen in urinary tract infections [10]. Development of resistance in Enterobacter infections has been associated with higher mortality and healthcare-related costs [13]. As a result, 3GCs are usually not recommended as first line therapy, even when susceptible in vitro.

Enterobacter may also acquire other major resistance determinants such as extended-spectrum beta-lactamase (ESBL) enzymes or carbapenemases [14, 15]. Carbapenem resistance may also develop from AmpC-hyperproduction in association with porin mutations [16] and multi-drug resistant (MDR) Enterobacter has been implicated in nosocomial outbreaks [17–21]. As such, Enterobacter spp. have become important ‘problem’ pathogens in the healthcare setting [22]. Most significant Enterobacter infections are treated with agents such as carbapenems, quinolones or aminoglycosides. Cefepime may be a useful option given its stability to AmpC enzymes [23]. Trimethoprim-sulphamethoxazole may also be effective [24], but is rarely used in contemporary practice for serious invasive Enterobacter infections. BLBLI agents, such as piperacillin-tazobactam, have an uncertain role in this context. [7] Although they are often avoided over concerns relating to the development of AmpC-mediated resistance, the risk of this occurring for this antibiotic class has rarely been examined, and may even be associated with improved outcome when used as empirical therapy in one study [25].

The aim of the study was to determine clinical factors associated with relapsed or persistent infection in patients with Enterobacter bacteraemia. A specific hypothesis to be tested was that the use of BLBLI agents as definitive therapy carried no additional risk of bacteraemia relapse when compared with alternative established therapies.

In this case–control study, 3.4% of patients with Enterobacter bacteraemia were found to have relapsed by 28 days. The most striking finding is the effect of bacteraemia source on the risk of this outcome; 74% of all relapsed cases had a line-associated source for the bacteraemia, and this variable had the greatest effect size in the multivariate logistic regression model (OR 3.87). Although data were not available to reliably determine the precise date of line-removal in the majority of cases, we would hypothesise delayed line removal as a likely reason for this association.

Given the well-described association with 3GC treatment and AmpC-mediated relapse when used for Enterobacter infections, this study also aimed to assess the effect of antibiotic choice on the risk of relapse. When compared with alternative effective therapies (such as carbapenems, cefepime, quinolones or aminoglycosides) BLBLIs were not significantly associated with bacteraemia relapse when used as definitive therapy. However, BLBLIs were used relatively infrequently for definitive therapy. Overall antibiotic use was heterogeneous and often included combination therapy, meaning that effects for individual agents were diluted. We defined BLBLI therapy as including ticarcillin-clavulanate and piperacillin-tazobactam. However, these agents have some important differences with respect to AmpC-enzymes. Clavulanate is a potent inducer of AmpC and a poor inhibitor, whereas piperacillin and tazobactam are only weak inducers of AmpC [33].

Many laboratories suppress susceptibility results for AmpC-producers (such as Enterobacter spp.) in favour of carbapenems or quinolones due to concerns of selecting for AmpC de-repressed variants. According to the findings of this study, this may not be justified as a routine policy, especially in cases where source control has been addressed or in the absence of significant immunosuppression. If BLBLIs are used against susceptible Enterobacter isolates there does not appear to be any clear association with microbiological failure; and failure with the emergence of AmpC-mediated resistance appears to be rare. However, given the small number of patients treated with BLBLIs in this study, the possibility of a type II error exists, and this needs further investigation in larger cohorts. The proportion of patients treated with BLBLIs was lower than expected, which compromised the power of the study to address the effect of antibiotic choice on outcome.

In this current era of emerging carbapenem-resistance, it is critical that alternatives to carbapenems are sought for common infections. Infectious disease practitioners in our region most frequently recommend carbapenems for AmpC beta-lactamase producing Enterobacteriaceae [34]. As such, it may be a significant driver of carbapenem use. Our findings would suggest that using a BLBLI when susceptibility is demonstrated, especially in immunocompetent patients in whom adequate source control has been achieved, is not clearly associated with microbiological failure. Further work would help to delineate patients for whom carbapenems may still be a superior choice.

Not only was the risk of relapsed Enterobacter bacteraemia low (3.4%) in this study, but in only 3 cases did resistance to 3GCs (a marker for AmpC de-repression) emerge during treatment. This figure is lower than described in some previous studies [8, 9]. There may be several reasons for this observed difference. Firstly, clinicians are more aware of this clinical phenomenon and 3GCs are currently rarely used as definitive treatment for Enterobacter bacteraemia. In this study only 3 of 159 (1.9%) bacteraemia events received either ceftriaxone or ceftazidime as definitive therapy, and in 2 of these 3GCs were given in combination with another agent. This may also reflect local laboratory practice for withholding susceptibility reporting for 3GCs against Enterobacter. Secondly, low rates of relapse may also result from better recognition of source control and the availability of other effective antibiotics, especially carbapenems, cefepime and fluoroquinolones. Other variables that may have influenced the low rate of relapse in this study might include the clinical characteristics of our patient population as well as long-standing infection control and antimicrobial stewardship programs.

The overall risk of Enterobacter bacteraemia with a de-repressed AmpC phenotype in our region was also relatively low over the study period, with only 38.4% of isolates being resistant to 3GCs (either ceftriaxone or ceftazidime). Very few isolates showed resistance to carbapenems, although MIC breakpoints were lowered during the study period. Both current and older clinical breakpoints may fail to detect the presence of carbapenemase in Enterobacteriaceae [35, 36]. Our findings may not be generalizable to populations encountering higher rates of baseline 3GC resistance, or frequent ESBL / carbapenemase acquisition in Enterobacter species.

Limitations of the study are acknowledged. Although the bacteraemia database was collected prospectively as part of routine surveillance, clinical and microbiological data extraction for Enterobacter cases was retrospective. Although a laboratory servicing several hospitals was used to identify cases, these occurred within a limited area in Australia and may not represent the diversity of infections and patient populations found in other states and countries. Whether all repeat bacteraemia episodes were true relapsed events or new infections is also not known, as isolates were not available for molecular typing. Equally, the presence of co-existing ESBLs was not sought systematically for isolates demonstrating resistance to 3GCs; inferring Amp-C de-repression based on the antibiogram alone may be unreliable. This study only examined relapsed bacteraemia, since such an endpoint is easily defined. However, the risk of relapsed infection and emergent resistance from other clinical sites infected with Enterobacter may be different and care should be taken when extrapolating from bacteraemia data alone.

Given the inherent limitations of retrospective studies, it is hoped that the efficacy of BLBLIs or other carbapenem-sparing options for Enterobacter infections could be addressed in large prospective observational studies or even randomised controlled trials (RCTs). A pilot RCT comparing piperacillin-tazobactam with meropenem for bloodstream infections caused AmpC-producers is currently recruiting (MERINO-2 Trial; ClinicalTrials.gov identifier: NCT02437045).

In a cohort of patients with bacteraemia caused by Enterobacter species, the overall risk of relapsed bacteraemia was low. Patient factors, such as a line-source for the bacteraemia or the presence of immunosuppression, were the strongest predictors of relapse. Use of a BLBLI agent as definitive therapy was not associated with treatment failure, but the total number of patients treated with a BLBLI was relatively low. Although this would suggest that susceptibility in vitro should translate into in vivo efficacy for BLBLI agents, especially if the patient has adequate source control, further studies are warranted to determine if BLBLIs are a safe and effective carbapenem-sparing option, especially for patients with severe disease, a complex focus of infection or immune compromise. Our data also reinforce the importance of line removal when suspected as the source for bacteraemia, especially for isolates with the genetic capacity to respond rapidly to antimicrobial exposure.