Background

Escherichia coli is one of the most common strains of extended-spectrum β-lactamase (ESBL)-producing bacteria [1]. Since the first ESBL-producing bacterial strains were discovered in the 1980s, new genotypes have frequently emerged [2]. In the past 10 years, the detection rate of ESBL-producing E. coli has continued to increase dramatically worldwide. In Asia, and particularly in China, the prevalence rate is much higher than that in Western developed countries [3, 4]. The production of β-lactamase enzymes is the principal mechanism by which Gram-negative bacteria resist the action of β-lactam antibiotics. According to relevant domestic monitoring data, E. coli accounted for the largest proportion of resistant bacteria detected in China, followed by K. pneumoniae [5].

Previous studies have shown that prior use of broad-spectrum antibiotics, poor general condition, advanced age, and some invasive procedures (such as central venous catheterization) are risk factors for ESBL-producing E. coli infection [6, 7]. For patients with ESBL-producing E. coli bloodstream infections (BSIs), improved prognoses depend on a detailed understanding of the factors that increase mortality. Increased mortality due to inappropriate antibiotic treatment (whether empirical or definite) has been widely reported in previous studies. Furthermore, other risk factors (e.g., poor general health condition, liver disease, septic shock, and antibiotic exposure) prior to infection have also been reported in the literature [8, 9].

ESBL-producing E. coli represents a serious public health issue and infection control challenge, and the selection of appropriate treatments for ESBL-producing E. coli BSIs remains controversial as there are insufficient large-scale randomized controlled trials to support various therapeutic approaches. Carbapenem antibiotics are still considered to be first-line antibiotics for treating ESBL-producing E. coli BSIs [10]. However, in recent years, widespread use of these antibiotics has driven the development of carbapenem-resistant Enterobacteriaceae (CRE). Available monitoring data showed that the resistance rate of Enterobacteriaceae to β-lactam antibiotics such as cefotaxime, ceftazidime, cefepime, and piperacillin-tazobactam decreased from 2005 to 2014, while the meropenem (2005, 2.8%; 2014, 4.5%) and ertapenem (2005, 5.5%; 2014, 8.9%) resistance rates showed an increasing trend [11]. Currently, a growing number of experts believe that there is a wealth of data showing that the effectiveness of treatment with β-lactam-β-lactamase inhibitor (BLBLI) combination antibiotics is similar to that of carbapenems in patients infected with ESBL-producing strains. Antibiotic drugs such as cephalosporins and BLBLI combination antibiotics (e.g., piperacillin-tazobactam) have once again gradually attracted attention.

It is worth noting that ESBL-producing E. coli bacteremia occurs often in solid organ transplant recipients, which may be due to their frequent hospitalization, longer hospital stays, and long-term exposure to immunosuppressive agents. Moreover, some studies have shown that these patients have higher ESBL-producing E. coli infection and mortality rates [12,13,14]. To the best of our knowledge, no studies have analyzed the risk and prognosis of non-transplanted ESBL-producing E. coli BSIs; therefore, in this study, patients with solid organ and hematopoietic stem cell transplants were excluded.

This retrospective analysis was designed to investigate the clinical characteristics of non-transplant patients with ESBL-producing E. coli infections as well as to analyze the risk and prognostic factors and the therapeutic effects of different antibiotic regimens with the goals of strengthening our understanding of ESBL-producing E. coli infections and providing new guidance for clinical practice.

Methods

Population

This study retrospectively analyzed the clinical and microbiological data of patients with BSIs caused by Escherichia coli in the First Affiliated Hospital of Zhejiang University from January 2013 to December 2016 after receiving approval from the research ethics committee. Data from patients with the following characteristics were included in the analysis: a) a blood culture positive for E. coli, b) clinical manifestations of infection, and c) hospitalization for more than 48 h with a complete clinical data set. Patients aged < 16 years and those without complete medical records were excluded. Patients with histories of organ or hematopoietic stem cell transplantation before the BSI and those with carbapenem-resistant E. coli infections were excluded. If the same patient had more than one BSI within 6 months, only data from the first BSI were included. Ultimately, based on the inclusion and exclusion criteria, 491 patients were included in the study.

Bacterial identification and drug sensitivity testing

The VITEK 2 COMPACT automatic microbial identification system was used for bacterial identification and drug susceptibility testing. According to the guidelines of the Clinical and Laboratory Standards Institute (CLSI) standards (2015) [15], ESBL production was determined using a double-disk potentiation test with amoxicillin-clavulanic acid and cefotaxime, ceftazidime, or cefepime or ESBL-positive results from the VITEK-2 N131 analysis. Carbapenem resistance was defined as a minimum inhibitory concentration (MIC) of ≥2 mg/L for ertapenem or a MIC of ≥4 mg/L for imipenem or meropenem.

Data collection

Demographic and clinical data were collected from the electronic case system, including age, gender, underlying disease, comorbidities, length of hospital stay, admission to the ICU, invasive procedures before and after infection, and antibiotic treatments. The severity of the disease was assessed via the Acute Physiology and Chronic Health Evaluation (APACHE) II scores and Pitt scores [16]. The Charlson index was used to assess the burden of any comorbidities [17]. A three-part analysis was conducted. First, risk factors associated with ESBL-producing E. coli infection were evaluated by comparing the ESBL-producing and non-ESBL-producing patient groups. Second, to investigate the risk factor of mortality, the 491 E. coli BSI patients were divided into survival and death groups according to their survival status after 28 days of infection. Finally, 283 patients with ESBL-producing E. coli BSIs were analyzed to assess the risk factors associated with 28 day mortality and various antibiotic treatments.

Definitions

E. coli BSI was defined as an infection manifested by the presence in at least one blood culture that grew a E. coli strain. BSI refers to the systemic inflammatory response syndrome caused by toxins and metabolites produced by pathogens. Possible sources of E. coli BSIs are based on the Centers for Disease Control (CDC) and Prevention/National Healthcare Safety Network (NHSN) surveillance definitions [18]. Glucocorticoid therapy was defined as prednisolone > 20 mg/day for more than 7 days. Antimicrobial drug exposure referred to the use of antibiotics for > 72 h 30 days prior to the BSI diagnosis. Empirical therapy was defined as a therapeutic drug administered at the time of the blood culture test or before the blood culture report based on clinical experience. Definitive therapy referred to antimicrobial therapy administered based on drug susceptibility results. Treatments were classified as “appropriate” if the regimen contained at least one drug effective against E. coli; otherwise, they were classified as “inappropriate”. All-cause mortality was defined as death from an E. coli BSI within 7, 14 and 28 days of the onset of bacteremia.

Data analysis

In the univariate analysis of ESBL-producing E. coli BSIs, the χ2 or Fisher’s two-tailed test was used for categorical variables, and Student’s t test (for variables with normal distributions) and the Mann-Whitney U test (for variables with non-normal distributions) were used for continuous variables. For continuous variables, the results are expressed as mean ± standard deviation (SD) or median (interquartile range [IQR]). For categorical variables, the percentages for each group are reported. Variables with P values ≤0.05 in the univariate analysis were included in the multivariate analysis, and binary logistic regression (backward: condition) was used to identify independent predictors. The survival distribution functions in different groups were compared via the Kaplan-Meier product limit method. All data were statistically analyzed using SPSS version 23.0, and P values ≤0.05 were considered statistically significant.

Results

Clinical characteristics

From January 2013 to December 2016, 491 patients with E. coli BSIs were included in this study, except for 67 patients who received solid organ or hematopoietic stem cell transplants and 22 patients with CRE BSIs. Among the included patients, 283 (57.6%) had ESBL-producing E. coli BSIs and 208 (42.4%) had non-ESBL-producing E. coli BSIs. There were no significant differences in age and gender between the two groups (60.8 ± 16.7 vs. 60.5 ± 16.3, P = 0.836; male, 112 vs. 150, P = 0.853). Of the primary infection sites of the 491 patients with E. coli BSIs, abdominal infections were the highest (n = 249), followed by respiratory infections (n = 108) and urinary tract infections (n = 84). The annual incidence rates of ESBL-producing E. coli BSIs from 2013 to 2016 were 66.2% (86/130), 59.1% (88/149), 49.1% (53/108), and 53.8% (48/104), respectively, and while the incidence rate fluctuated, it did not increase significantly (P = 0.638).

Risk factors for ESBL-producing E. coli BSIs

A univariate analysis (Table 1) showed that risk factors for ESBL-producing E. coli BSIs included urinary system infection, prior surgery, prior invasive procedures (central venous catheterization, gastric catheterization, and percutaneous catheterization), and antibiotic use within 30 days before infection (mainly cephalosporins and BLBLI combination regimens). In a multivariate analysis, logistic regression analysis showed that urinary tract infections were the primary site of E. coli BSIs (OR = 1.897, P = 0.014) and that cephalosporin exposure (OR = 2.767, P = 0.007) and treatment with BLBLI combination regimens in the 30 days prior to the BSI (OR = 1.950, P = 0.010) were independent risk factors for ESBL-producing E. coli BSIs.

Table 1 Clinical and Demographic Characteristics of Patients with BSI Caused by E. coli
Full size table

Risk factors affecting the mortality of patients with E. coli infection

The non-transplanted patients with E. coli BSIs were classified into survivor and death groups based on the outcome at 28 days. The overall all-cause 28 day mortality rate of the 491 patients was 14.46% (71/491). A survival curve analysis (Fig. 1) showed that the mortality of the patients in the ESBL-producing E. coli group was higher than that of the patients in the non-ESBL-producing E. coli group (14.9% vs. 14.2%, χ2 = 0.003, P = 0.953), and the difference between the mortalities was not statistically significant. Furthermore, the patients were divided into community infection and nosocomial infection groups (Additional file 1: Figure S1). In 187 community-infected BSI patients (8.4%), there was no statistically significant difference in the mortality between those with non-ESBL-producing E. coli infections and those with ESBL-producing E. coli infections (8.3% vs. 11.7%; χ2 = 0.472, P = 0.492). There was also no significant difference in the mortalities between these two groups in the patients with nosocomial infections (17.7% vs. 16.7%; χ2 = 0.240, P = 0.624).

Fig. 1
figure 1

Kaplan-Meier survival estimates among patients with BSIs caused by non-ESBL-producing E. coli and ESBL-producing E. coli

Full size image

The main characteristics of the E. coli BSI survivor and non-survivor subgroups are shown in Table 2. Based on a multivariate regression analysis, the following factors were independently associated with a higher mortality risk in patients with E. coli BSIs: hepatobiliary disease (OR = 1.890, P = 0.034), prior use of carbapenem antibiotics (OR = 2.839, P = 0.016), and higher APACHE II scores (OR = 1.313, P < 0.001). Normal serum total protein (OR = 0.959, P = 0.006) and appropriate definite antibiotic treatment were independent protective factors (OR = 2.777, P = 0.017) for the outcomes of patients with E. coli BSIs.

Table 2 Analysis of risk factors for 28-day Mortality in 491 patients with E. coli bloodstream infections
Full size table

Risk factors for mortality in ESBL-producing E. coli BSIs and treatment regimens

A total of 283 ESBL-producing E. coli BSI patients were included in this analysis, and the 28 day mortality was 14.8%. In the univariate analysis (Additional file 2: Table S1), pulmonary infection, invasive procedures and/or device implementation before and after development of a BSI, antibiotic use within 30 days prior to infection, hypoproteinemia, and higher APACHE II scores were risk factors for increased mortality in patients with ESBL-producing E. coli BSIs. A multivariate analysis revealed that hypoproteinemia (OR = 0.941, P = 0.045) and higher APACHE II scores (OR = 1.103, P = 0.003) were independent risk factors for poor outcomes.

As shown in Additional file 2: Table S1, of the patients infected with ESBL-producing E. coli, 87.99% (249/283) received appropriate empirical antibiotic therapy and 94.35% (267/283) received appropriate definite antibiotic therapy. Appropriate empirical antibiotic treatment was a protective factor in the univariate analysis, and the rate of patients receiving appropriate empirical antibiotic treatment in the death group was significantly lower than that in the survival group (78.2% vs. 89.6%, P = 0.042). Based on our multivariate analysis, appropriate empirical antibiotic treatment also appeared to be a protective factor that can reduce the mortality rate, although this effect was marginally significant (OR = 2.526, P = 0.058).

In terms of empirical treatment, 117 patients with ESBL-producing E. coli BSIs were treated with carbapenem monotherapy, and 95 patients were treated with BLBLI combination monotherapy regimens. Figure 2 shows that a single-antibiotic regimen with carbapenem (n = 117) or a BLBLI regimen (n = 95) resulted in no differences in 28 day mortality between the two groups (12.8% vs. 17.9%; χ2 = 0.759, P = 0.384), and their APACHE II scores were 10 and 9, respectively. Considering the potential confounding effect of severity, the patients were divided into two groups according to their APACHE II scores (< 9 and ≥ 9), and the difference in the prognoses of the two groups was compared (Additional file 1: Figure S2). For patients with APACHE II scores ≥9 at the onset of bacteremia, there was no statistically significant difference in the 28 day mortality rate between the patients who received carbapenem and those who received BLBLI monotherapy (16.4% vs. 22.4%; χ2 = 0.249, P = 0.617); similarly, no significant difference was observed between the two groups in patients with APACHE II scores < 9 (6.8% vs. 10.8%; χ2 = 0.694, P = 0.405).

Fig. 2
figure 2

Kaplan-Meier 28 day survival estimates among ESBL-producing E. coli BSI patients depending on treatment with carbapenem or β-lactam-β-lactamase inhibitor (BLBLI) combination antibiotics

Full size image

Discussion

Escherichia coli is one of the most common pathogens in clinical infections. The primary mechanism of drug resistance in these bacteria is the production of ESBLs, and this mechanism underlies a high proportion of the antibiotic resistance cases in China [5]. This study focused on studying the risk factors associated with ESBL-producing E. coli BSIs in non-transplanted patients.

Antimicrobial use prior to BSI is believed to be an important factor in drug-resistant infections [19, 20], although some studies have shown no association between ESBL-producing E. coli infections and prior antibiotic therapy; for example, Denis et al. reported that there was no significant difference in the prevalence of antibiotic exposure between the ESBL-producing E. coli and non-ESBL-producing E. coli groups, which included 82 patients (P = 0.32) [21]. Our results also demonstrated that cephalosporins and BLBLI combination regimens within 30 days prior to infection were independent risk factors for ESBL-producing E. coli BSIs, which is consistent with the findings of many previous studies. Regarding the likely infectious source of E. coli BSIs, intra-abdominal infections were most common, followed by respiratory tract infections and urinary tract infections. However, most studies showed that ESBL-producing E. coli strains are primarily derived from urinary tract infections [7, 9, 22], followed by respiratory and bloodstream infections. Furthermore, we found that patients with nosocomial infections accounted for 50% of the patients with E. coli urinary tract infections (n = 29) and 66% of the patients with intra-abdominal infections (n = 93); therefore, this significant difference between the frequencies of patients with infections in these two sites is likely due to a higher likelihood of acquiring an abdominal infection in the hospital. Based on this observation, we suspect that the reason for the high proportion of intra-abdominal infections in this study may be related to the high number of patients undergoing hepatology-related procedures and hepatobiliary surgery in our hospital. It is well known that E. coli is an opportunistic bacterial pathogen that can invade the body when invasive procedures disrupt the mucosa. The multivariate analysis of the risk factors associated with ESBL-producing E. coli in this study showed that urinary tract infection was an independent risk factor (P = 0.014), consistent with previous studies [23,24,25]. Currently, ESBL-producing E. coli BSIs originating from urinary tract infections are more commonly studied, while articles focusing on abdominal infections are rare, suggesting a further research direction.

To explore the high mortality rate associated with E. coli BSIs, we evaluated the patient characteristics and the treatments they received. In this study, there was no significant difference in the 28 day mortality between patients infected with ESBL-producing E. coli and those infected with non-ESBL-producing E. coli (P = 0.953), similar to the findings of some previous studies [21, 26]. Conversely, other studies demonstrated significantly higher mortality in patients infected with ESBL-producing E. coli than in those infected with non-ESBL-producing E. coli [6]. A possible reason for the similar mortality rates in the two groups in the present study is the use of many broad-spectrum antibiotics due to the current high prevalence of ESBL-producing E. coli. For patients with end-stage liver disease, bacterial infection is one of the most common causes of death [27], which explains the observation that hepatobiliary system disease was an independent risk factor for mortality in the present study. Furthermore, an analysis of patients with potential liver disease and E. coli infection by Kang et al. yielded similar results [28]. Our analysis showed that disease severity is also an independent risk factor for mortality in patients with ESBL-producing E. coli BSIs, which is also similar to the results of previous studies [29]. We also found that low serum albumin is an independent risk factor for death in patients with E. coli and ESBL-producing E. coli BSIs. Serum albumin level is a common indicator for assessing a patient’s nutritional status, organ function, and comorbidity. The inflammatory state resulting from bacterial infection, which leads to the production of IL-1, TNF, and other cell mediators, can interfere with liver albumin synthesis, resulting in hypoalbuminemia [30]. There is currently a lack of literature on the relationship between mortality and serum albumin level in patients with E. coli infections. Akirov et al. [31] studied the relationship between serum albumin levels and prognosis in hospitalized patients (without any detailed division) and found that low albumin levels were positively correlated with short- and long-term mortality in the entire hospitalized patient population. Albumin levels are often lower in patients with hepatic insufficiency, which was also associated with higher mortality in the patients with hepatobiliary disease included in this study. Du et al. assessed the outcomes of 85 patients with E. coli infections and found that prior use of BLBLI combination antibiotics increased the mortality rate [32]. In the present study, we also found that use of carbapenems within 30 days prior to infection was an important independent risk factor for the patients who died, which has been rarely reported previously. Previous studies [33] have shown that short-term use of such antibiotics before infection may also lead to the production of and infection by carbapenem-resistant bacteria. These outcomes increase the complexity of treating critical infections, which might explain the increased mortality observed in this study.

In terms of the antibiotic options for treating E. coli BSIs, we believe that appropriate, defined antibiotic treatments are a protective factor for reducing death due to E. coli BSIs, as reported in this study (OR = 2.777, P = 0.017), and this observation is consistent with those of some other studies [7, 34]. The findings of attempts to determine the best treatment options for ESBL-producing E. coli have been inconclusive, although carbapenems have long been considered the best antibiotic for treating infections with ESBL-producing E. coli [35]. However, the wide use of carbapenems has been associated with the emergence of CRE. In recent years, the efficacy advantages of carbapenems over other antibiotics have been increasingly questioned. The latter antibiotics include BLBLI combination antibiotics, cefepime, and quinolones (among others), and the feasibility of using the above-mentioned older antibiotics to treat ESBL-producing E. coli infections is becoming more and more accepted by researchers [36,37,38]. We analyzed the outcomes of 283 patients with ESBL-producing E. coli BSIs, including 95 patients with BLBLI combination monotherapy regimens and 117 with carbapenem monotherapy (median APACHE II 9 and 10). Their 28 day mortality rates were 12.8 and 17.9%, respectively, and the difference was not statistically significant (χ2 = 0.759, P = 0.384). Because disease severity may affect prognoses, we divided the ESBL-producing E. coli BSI patients into two groups according to their APACHE II scores (< 9, ≥ 9) and assessed their 28 day mortality rates. We found no statistically significant difference in 28 day mortality between the two groups (Additional file 1: Figure S2 P = 0.405; P = 0.617), which confirms that carbapenem antibiotics are no longer more effective than BLBLI combination antibiotics against ESBL-producing E. coli BSIs. These findings are consistent with previous reports that proposed the use of BL/BLIs to treat ESBL-producing E. coli BSIs [22, 39]. For example, Muhammed et al. [39] conducted a meta-analysis of more than 1000 patients in 14 relevant articles and found that there was no significant difference in prognosis between patients treated with carbapenems and those treated with BLBLI combination antibiotics for either empirical or definite antibiotic treatment. However, some reports support the opposite opinion that carbapenem antibiotics can significantly reduce mortality compared with other antibiotics [40]. Currently, as stated above, it remains unclear in all of the relevant studies whether carbapenems were primarily administered to patients with more comorbidities or to those with more severe clinical conditions.

Regarding the limitations of this study, our findings are based mainly on retrospective data and studies, which makes it impossible to eliminate certain types of bias. Furthermore, while this study excluded patients with solid organ and hematopoietic stem cell transplantation, some patients with underlying diseases may have been treated with immunosuppressive agents. Finally, we were unable to perform further stratification of the outcomes based on patient genotyping due to a lack of phenotypic testing. In this context, we should perform more detailed drug resistance gene detection and drug susceptibility testing in ESBL-producing E. coli BSI patients to better understand the drug resistance profiles involved as well as the relevant treatment options.

Conclusion

In summary, prior exposure to cephalosporin and BLBLI combination antibiotics increases the risk of acquiring ESBL-producing E. coli infections. Prior carbapenem use and a poor systemic condition were risk factors for increased mortality in patients with ESBL-producing E. coli and non-ESBL-producing E. coli BSIs; however, no significant difference was found in the prognoses of patients with ESBL-producing E. coli and non-ESBL-producing E. coli BSIs. Hypoproteinemia and severe disease lead to worse outcomes for patients with ESBL-producing E. coli BSIs. Carbapenem offered no significant advantage over BLBLI combination antibiotics for improving the 28 day mortality of patients with ESBL-producing E. coli BSIs.