If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
The prevalence and impact of ESBL Escherichia coli bacteraemia was investigated, with a focus on patients with haematological diseases.
A dramatic increase in the prevalence of ESBL-EC bacteraemia was identified.
On multivariate analysis, a risk factor for ESBL-EC bacteraemia was found to be previous ESBL-EC colonization.
ESBL-EC bacteraemia was not associated with a prolonged length of hospital stay.
Despite more inadequate initial antimicrobial treatment, mortality was not significantly increased.
The impact of extended-spectrum beta-lactamase (ESBL)-producing Escherichia coli (ESBL-EC) bacteraemia on outcome remains controversial.
A retrospective analysis of the prevalence, risk factors, clinical features, and outcomes of all ESBL-EC bacteraemia in one French hospital over a 5-year period was performed. A casecontrol study was undertaken: cases had at least one ESBL-EC bacteraemia and controls a positive non-ESBL-EC bacteraemia.
The prevalence of ESBL-EC bacteraemia increased from 5.2% of all positive E. coli blood cultures in 2005 to 13.5% in 2009 (p < 0.003). CTX-M represented 70% of ESBL-EC bacteraemia strains, and strains were not clonally related. On adjusted analysis, the only significant risk factor for ESBL-EC bacteraemia was a previous ESBL-EC colonization (odds ratio 11.3, 95% confidence interval 1.2107; p = 0.003). Initial antimicrobial therapy was less frequently adequate in the ESBL-EC group (48% vs. 85%; p = 0.003). The presence of ESBL-EC bacteraemia was not associated with a longer hospital stay (p = 0.088). Day 30 mortality was high, but not significantly different in the two groups (30% vs. 27%; p = 0. 82).
The prevalence of ESBL-EC bacteraemia has been increasing dramatically. Previous colonization with ESBL-EC was a strong risk factor for ESBL-EC bacteraemia. More inadequate initial antimicrobial therapy was noted in the ESBL-EC group, but mortality and length of hospital stay were not significantly different from those of patients with non-ESBL-EC bacteraemia.
Infections due to extended-spectrum beta-lactamase (ESBL)-producing Enterobacteriaceae have become a major health problem. ESBL-producing Escherichia coli (ESBL-EC), especially of the CTX-M type, have increased dramatically worldwide.
Risk factors for ESBL infection have been identified and include age, comorbidities, poor functional status, recurrent urinary tract infections, intensive care unit stay, prolonged hospital stay, previous use of antibiotics, and colonization with ESBL.
The impact of ESBL E. coli bacteraemia (ESBL-EC bacteraemia) on mortality is controversial. Some studies have shown that mortality is associated with inappropriate antimicrobial therapy, irrespective of ESBL production, whereas others have reported an increased mortality due to ESBL.
The aim of the current study was thus to assess the prevalence and risk factors for ESBL-EC bacteraemia and to analyse their impact on length of hospital stay and on day 30 mortality in hospitalized patients.
Prevalence was studied using data from the Microbiology Laboratory of Saint Louis Hospital, a 550-bed tertiary hospital with major clinical activity in the areas of HIV, haematology, and oncology. Data on all ESBL-EC and non-ESBL-EC bacteraemia from January 2005 to December 2009 were collected. A retrospective casecontrol study was then conducted, and all hospitalized adults from January 2005 to December 2008 with at least one episode of bacteraemia due to ESBL-EC were included for analysis. This study was performed in accordance with the ethical standards described in the 1964 Declaration of Helsinki. In this observational research, no additional medical procedure was performed and all data were retrieved from the medical charts of the treated patients. All information was, however, given to the patients and, in accordance with French legislation, patients could refuse the use of their medical data. Baseline characteristics at the date of the first positive blood culture drawn, clinical and microbiological data, antibiotic therapy, and outcomes were recorded carefully for each patient.
Cases were defined as adults with blood culture(s) yielding ESBL-EC during the study period. In the case of several positive ESBL-EC bacteraemia during the same infectious episode, only the first positive blood culture was considered for analysis. For each case, one control was selected among hospitalized patients with a positive non-ESBL-EC blood culture and matched to the closest date of an ESBL-EC bacteraemia positive defining-case.
The following data were retrieved from the medical charts: age, gender, underlying diseases and comorbidities (HIV, haematological malignancies, solid tumour, diabetes mellitus, solid organ transplant), severity as assessed by the APACHE II score (the Acute Physiology and Chronic Health Evaluation is a general prognostic model for mortality calculated at admission), severe neutropenia (absolute neutrophil count (ANC) <100/mm3), current chemotherapy, antibiotic use during the past 3 months, ESBL-EC colonization during the past 6 months, antibiotic susceptibility of the isolated strain, initial antibiotic therapy, length of hospital stay, and day 30 mortality (defined as the time from the first positive blood culture until discharge or death). Bacteraemia was classified as nosocomial, healthcare-related, or community-acquired, as described previously.
The source of bacteraemia was considered to be urinary, catheter, digestive, or respiratory when the strains recovered from the blood culture and from the source were phenotypically similar. Rare locations were classified as other, and when the portal of entry could not be identified clearly, the source of bacteraemia was classified as unidentified.
An adequate first antibiotic regimen was defined as a regimen containing at least one in vitro active drug, at the recommended dose, and initiated within 24 h after the blood sample was drawn. If one or more of these conditions was not fulfilled, then antimicrobial therapy was considered inadequate. Isolates were identified using the API System (bioMérieux, Marcy lEtoile, France). Strain susceptibility to antibiotics was determined by disk diffusion method and interpreted according to the French Society for Microbiology criteria with the 2008 susceptibility breakpoints (ESBL-EC strains were interpreted as resistant to amoxicillinclavulanate, piperacillintazobactam, and to broad-spectrum cephalosporins).
ESBL production was detected by double-disk synergy test between clavulanate and third-generation cephalosporins. ESBL type was identified by ESBL KPC Microarray (Check Points, Wageningen, Netherlands), according to the manufacturer's instructions. Previous colonization strains were also studied using the same procedure. ESBL-EC colonization strains were obtained by rectal swab during the 6 months prior to the onset of bacteraemia for all patients in the non-ESBL-EC group and for 31 (76%) of the patients in the ESBL-EC group. ESBL-EC infection was defined as the first positive ESBL-EC blood culture during the study period. Genotype comparison between ESBL-EC bacteraemia strains and colonization strains from the same patients was performed using DiversiLab system software, version 3.3.40 (bioMérieux).
Relationships between DiversiLab patterns were designated as recommended in the manufacture's guidelines: different strains (similarity <92%) with three or more band differences, or similar strains (similarity 92%) with two or fewer band differences.
2.1 Statistical analysis
The trend in ESBL-EC bacteraemia prevalence over time was compared with a Chi-square test for trend.
Baseline characteristics of the patients in the ESBL-EC and non-ESBL-EC groups were compared by Wilcoxon signed-rank tests and Liddell tests, accounting for the pair-matched design of the study. The cause-specific proportional hazards model with death as a competing risk was used for length of stay.
Conditional logistic regression was used for adjusted analysis. The variables introduced into the multivariate analysis included those with a marginal p-value of <0.15. APACHE II score was included in the model as a continuous variable.
Since day 30 mortality of matched patients had no reason to be correlated (matching only on the date of blood culture, with no major changes in outcome over the study period), the outcome analysis did not account for matching. It was nevertheless checked that intra-pair correlation was negligible for outcome (intra-class correlation coefficients <0.0001). Risk factors for day 30 mortality were analysed by logistic regression. Since there were only 23 deaths, only three variables were included in the model in order to limit over-fitting. Before including the variable inadequate antimicrobial therapy into the model, an interaction with ESBL-EC groups was tested. As a significant interaction was observed, results of the adjusted analysis on mortality are detailed with the effect of inadequate antimicrobial therapy in the ESBL-EC and non-ESBL-EC groups.
A sensitivity analysis was undertaken for patients in the ESBL-EC group treated with piperacillintazobactam, according to the current Comité de lantibiogramme de la Société Française de Microbiologie (CA-SFM) and European Committee on Antimicrobial Susceptibility Testing (EUCAST) 2015 guidelines (http://www.sfm-microbiologie.org/.../casfm/CASFM_EUCAST_V1_2015.pdf): ESBL strains were tested using the Etest method and considered susceptible when the minimum inhibitory concentration (MIC) for piperacillintazobactam was 8 mg/l. Patients were classified as having received an adequate treatment if the strain was susceptible and treatment initiated within 24 h after the first blood culture was drawn.
Analyses were performed using R statistical software, version 2.10.1 (R Development Core Team, Vienna, Austria). All statistical analyses were two-sided and p-values less than 0.05 were considered significant.
The prevalence of ESBL-EC bacteraemia increased significantly from 5.2% in 2005 to 13.5% in 2009 (p = 0.003; Figure 1). From 2005 to 2008, 631 episodes of E. coli bacteraemia were recorded and a first episode of ESBL-EC bacteraemia was noted for 45 patients (7.1% of ESBL-EC bacteraemia among E. coli bacteraemia). Medical charts could be reviewed for 41 of the 45 patients, and were matched with 41 controls with non-ESBL-EC bacteraemia.
Antibiotic resistance in the ESBL-EC and non-ESBL-EC groups was observed, respectively, in 23% and 8% of cases with gentamicin, in 5% and 0% with amikacin, and in 67% and 35% with ofloxacin.
Thirty-three isolates responsible for ESBL-EC bacteraemia could be characterized extensively: 21 strains produced CTX-M, seven strains produced TEM-ESBL, and three strains SHV-ESBL. Two E. coli produced two types of ESBL (CTX-M + TEM and CTX-M + SHV). Eighteen CTX-M strains belonged to the group CTX-M1, three to the group CTX-M9, one to the group CTX-M1 and TEM, and one to the group CTX-M1 and SHV. Genotypic comparison between these 33 isolates did not show any epidemiological relationship, suggesting a high variability among them (data not shown).
Thirteen patients had a positive ESBL-EC colonization strain. Previous colonization strains were available for 10 of the 13 patients, and a comparison between ESBL-EC bacteraemia strains and E. coli colonization strains was undertaken (Figure 2). DiversiLab patterns showed >92% similarity, with no band differences between the ESBL-EC bacteraemia strain and the colonization strain in eight out of 10 patients. For the two remaining patients, DiversiLab patterns between ESBL-EC bacteraemia strains and E. coli colonization strains (5/5c and 25/25c; Figure 2) showed 91% similarity (two different bands) and 59% similarity (six different bands), respectively. ESBL was typed for these two pairs of strains and only 25/25c had a different ESBL type. Overall, the same ESBL type was seen in nine of the 10 ESBL colonization strains studied.
Patient baseline characteristics, according to the study group, are shown in Table 1. Patients had severe underlying conditions. Indeed, haematological diseases were noted in 68% of cases (28 patients) in the ESBL-EC group vs. 49% (20 patients) in the non-ESBL-EC group (p = 0.046). Severe neutropenia (ANC <100/mm3) was noted in 27% of patients overall, without a significant difference between the two groups (p = 0.39). Fifty-two percent of patients overall had received chemotherapy during the past month. The mean APACHE score was high and similar in the two groups (p = 0.8). In the ESBL-EC and non-ESBL-EC groups, 83% and 71% episodes, respectively, were nosocomial or healthcare-related, with a non-significant difference between the groups (p = 0.27).
Table 1Baseline characteristics of patients with bacteraemia due to EBSL-EC or non-ESBL-EC and related risk factors for ESBL-EC bacteraemia
(n = 41)
(n = 41)
OR (95% CI)
OR (95% CI)
Age, years, median (IQR)
Female gender, n (%)
APACHE II score, median (IQR)
Solid organ transplant, n (%)
Haematological malignancy, n (%)
Solid tumour, n (%)
Severe neutropenia (ANC <100/mm3), n (%)
Chemotherapy in the past month, n (%)
HIV, n (%)
Diabetes mellitus, n (%)
Healthcare-related or nosocomial bacteraemia, n (%)
Risk factors for ESBL-EC bacteraemia are shown in Table 1. While more patients in the control group had a urinary portal of entry and more patients in the ESBL-EC group had haematological malignancies, a more favourable outcome in controls was not observed. On univariate analysis, having a haematological malignancy (odds ratio (OR) 3.67, 95% confidence interval (CI) 1.0213.1), antibiotic exposure during the past 3 months (OR 5.0, 95% CI 1.1022.8), and an ESBL-EC colonization (OR 12.0, 95% CI 1.5692.3) were significant risks factors for ESBL-EC bacteraemia. On multivariate analysis, the only independent risk factor for ESBL-EC bacteraemia was a known ESBL-EC colonization in the last 6 months (OR 11.3, 95% CI 1.2107; p = 0.035).
The initial antibiotic treatment was based on the association of a broad-spectrum beta-lactam and an aminoglycoside or fluoroquinolone in 72% of cases and 68% of controls (p = 0.81, Table 2). Patients with ESBL-EC bacteraemia were less frequently treated with an adequate antibiotic therapy compared to controls (48% adequacy vs. 85%, respectively; p = 0.003) despite more carbapenem use in the ESBL-EC group (35% vs. 10%; p = 0.01).
Table 2Initial regimen administered for patients with bacteraemia in the ESBL-EC and non-ESBL-EC groups
Initial antimicrobial treatment
ESBL-EC (n = 40) n (%)
Non-ESBL-EC (n = 40) n (%)
Univariate analysis p-value
Association with aminoglycosides or fluoroquinolones
The median length of hospital stay was not significantly different between groups: 15 days (interquartile range (IQR) 1021 days) in the ESBL-EC group and 11 days (IQR 717 days) in controls (p = 0.88) (data not shown).
The 30-day mortality was high: 30% for ESBL-EC and 27% for controls (p = 0.82). Risk factors for mortality according to the groups studied are presented in Table 3. The first model tested the effect of ESBL-EC adjusted only for the APACHE score and no significant effect of ESBL-EC on mortality was noted (OR 1.23, 95% CI 0.364.23). The second model was adjusted for APACHE score and initial treatment adequacy. Since an interaction was observed between treatment adequacy and ESBL-EC, results took into account having an adequate or inadequate initial treatment separately. Mortality was studied in the group with an initial adequate treatment (OR 2.99, 95% CI 0.5915.2) and in the group with an initial inadequate treatment (OR 0.074, 95% CI 0.0051.19), both adjusted for the APACHE score. In both groups, the effect of ESBL-EC adjusted for the APACHE score and initial treatment was not significant. On adjusted analysis, a higher APACHE score was significantly associated with a higher mortality rate in both models.
Table 3Adjusted analysis of the impact of ESBL-EC on Day 21 mortality
OR (95% CI)
First model: model without the variable inadequate antimicrobial therapy
ESBL-EC non ESBL-EC
1.23 (0.36;4.23) 1
Second model: model with the variable inadequate antimicrobial therapy (Interaction with ESBL: p = 0.025)
Adequate antimicrobial therapy: - and non ESBL-EC - and ESBL-EC
1 2.99 (0.59;15.2)
Inadequate antimicrobial therapy: - and non ESBL-EC - and ESBL-EC
30.3 (2.30-399.2) 2.25 (0.29;17.7)
Apache score (per unit)
ESBL, extended-spectrum beta-lactamase; EC, Escherichia coli; OR, odds ratio; CI, confidence interval; APACHE, Acute Physiology and Chronic Health Evaluation.
In the sensitivity analysis, five patients treated with piperacillintazobactam in the ESBL-EC group were classified as having had an adequate treatment after taking into account the current breakpoints (MIC value for piperacillintazobactam 8 mg/l for susceptible strains) and the delay between the first blood culture and first antibiotic regimen. Two of the five patients died before day 30 and three were still alive at day 30. In this sensitivity analysis, patients in the ESBL-EC group still had less adequate treatment (p = 0.049), with no difference in day 30 mortality in the adjusted analysis (Supplementary Material, Tables S1 and S2).
In this study, a dramatic increase in the prevalence of ESBL-EC bacteraemia was observed, with a prevalence that doubled in 4 years. This increase in resistance has also been reported in recent studies involving either immunocompetent or immunocompromised patients.
In Europe, the Antimicrobial Resistance Surveillance Network (EARS-Net) showed that the proportion of reported E. coli isolates resistant to third-generation cephalosporins (proxy for ESBL-EC) increased significantly during the 20062010 period, with a resistance rate for invasive E. coli infections in France and the UK in 2010 of 7.2% and 8.3%, respectively.
More than 70% of patients (cases and controls) had nosocomial or healthcare-related E. coli bacteraemia, and, as described in previous studies, ESBL-EC strains were predominantly CTX-M strains (70%).
In contrast to other studies on bacteraemia due to ESBL-EC reporting urinary tract as the primary infection site, the source of bacteraemia was unknown in about 25% of cases and was mainly of digestive origin in the remaining cases studied here. The baseline characteristics of the population involved in the present work (36 neutropenic patients (44%), including 22 cases (27%) with severe neutropenia) may explain this discrepancy.
Regarding risk factors for ESBL-EC bacteraemia in a matched control study, having a haematological malignancy, prior exposure to antibiotics in the past 3 months, and previous colonization with ESBL-EC strains were independent risk factors on univariate analysis, while only colonization remained a risk factor on multivariate analysis. Of note, all patients previously colonized had received antibiotics in the past 3 months.
The similarity between colonizing strains and strains responsible for bacteraemia was studied and it was found that nine out of the 10 pairs of strains that could be analysed were similar. Nevertheless, no definitive conclusions can be drawn due to the small number of strains that were available for analysis, although this could be an argument for screening of ESBL colonization in immunosuppressed patients, particularly in neutropenic patients, to help define the empirical therapy in the case of fever. This differs from a single study on ESBL-EC colonization and bacteraemia in neutropenic patients, which found that colonization did not appear to have any significant clinical relevance.
In the present study, despite significantly more inadequate probabilist treatment, there was no significant increase in mortality or length of hospital stay in patients with bacteraemia due to ESBL-EC. There were slightly more patients with a urinary portal of entry in the control group and more patients with haematological malignancies in the ESBL-EC group. Despite these favourable characteristics in the control group, a more favourable outcome was not observed for controls. Unsurprisingly, in the adjusted analysis, a higher APACHE score was significantly associated with a higher mortality rate.
In the sensitivity analysis, five patients in the ESBL-EC group were reclassified as treated with an adequate antibiotic therapy, according to the current susceptibility breakpoints for ESBL strains (EUCAST/CA-SFM, 2015). In this sensitivity analysis, results with regards to inadequate treatment and day 30 mortality did not change when considering the recent guidelines.
In a meta-analysis including 15 studies, ESBL-EC bacteraemia was associated with higher mortality after adjusting for confounding factors (OR 1.51), and treatment inadequacy was the explanation for the higher death rate.
Patients in both groups had severe underlying diseases and a critical clinical condition at baseline, as attested by the APACHE score. The probability of death in patients with an initial APACHE II score of 18 is 29%, which corresponds to the mortality rate observed in this study. The same results were found in the study published by García Hernández et al. on ESBL-EC bacteraemia with more inadequate initial treatment, but with a mortality statistically associated with the severity of underlying diseases and critical severity of illness at onset, but not with the presence of ESBL-producing strains.
In seriously affected patients, as described in the present study, the real impact of ESBL-EC bacteraemia on the overall mortality is difficult to assess.
The length of stay, rarely studied elsewhere, was not affected by the antibiotic resistance of E. coli responsible for bacteraemia. However the high mortality and the small number of patients included may explain the absence of impact of resistance on length of stay.
Some limitations of this study should be acknowledged. First it was a retrospective study performed at a single institution. Second, it is underpowered due to the small size of the effective.
In conclusion, it was found that the prevalence of ESBL-EC bacteraemia increased dramatically throughout the study period, with a large predominance of CTX-M strains that were not clonally related. Previous colonization with ESBL-EC was the only significant risk factor for ESBL-EC bacteraemia on multivariate analysis. ESBL-EC bacteraemia was not associated with a prolonged length of hospital stay. Overall mortality was high in the immunosuppressed population studied. Despite more inadequate initial antimicrobial therapy in patients with ESBL-EC bacteraemia in this study, there was no significant increase in mortality rate.
The following are acknowledged: microbiology laboratory technicians; departments of intensive care, internal medicine, haematology, gastroenterology, and oncology.
Ethical approval: This study was performed in accordance with the ethical standards described in the 1964 Declaration of Helsinki. In this observational research, no additional medical procedure was performed and all data were retrieved from the medical charts of the treated patients. All information was, however, given to the patients and, in accordance with the French legislation, patients could refuse the use of their medical data.