The spread of extended-spectrum β-lactamase (ESBL)-producing Enterobacteriaceae (ESBL-E) is a growing public health threat worldwide.Reference Morrissey, Hackel, Badal, Bouchillon, Hawser and Biedenbach 1 A 2012 study from 72 hospitals in the United States reported that 16% of Klebsiella pneumoniae and 11.9% of Escherichia coli isolates were ESBL producers.Reference Morrissey, Hackel, Badal, Bouchillon, Hawser and Biedenbach 1 Much higher proportions were found in other populations,Reference Martelius, Jalava, Kärki, Möttönen, Ollgren and Lyytikäinen 2 with variation between countries and between hospitals. The limited antibiotic treatment options and adverse clinical outcomes of infections caused by ESBL-E have raised major concerns among clinicians and infection control services.Reference Castanheira, Farrell, Krause, Jones and Sader 3
In the 1980s and 1990s, K. pneumoniae was the predominant ESBL-E causing nosocomial outbreaks, and TEM and SHV were the main β-lactamases involved.Reference Ben-Ami, Schwaber and Nanon-Venezia 4 , Reference Burwen, Banerjee and Gaynes 5 In recent decades, E. coli has surpassed K. pneumoniae, and the most common β-lactamase is CTX-M.Reference Canton and Coque 6
Most previous studies that examined the epidemiology and outcome of infections caused by ESBL-producing organisms often combined different Enterobacteriaceae to one group assuming homogeneity; moreover, differences in outcomes between ESBL genotypes have not been considered.Reference Ben-Ami, Schwaber and Nanon-Venezia 4 , Reference Schwaber and Carmeli 7 However, this may not be the case. For example, unlike ESBL-producing K. pneumoniae (ESBL-KP), ESBL-producing E. coli (ESBL-EC) is common in community-acquired infections.Reference Canton and Coque 6 ESBL-EC has less predilection for hospital transmission than other ESBL-producing species, which has led the European Society of Clinical Microbiology and Infectious Diseases (ESCMID) to recommend contact precautions for patients colonized with all ESBL-E except ESBL-EC.Reference Tacconelli, Cataldo and Dancer 8
In this study, our aims were twofold: (1) to compare the epidemiology, clinical characteristics, and mortality of patients with bloodstream infections (BSIs) caused by ESBL-EC versus ESBL-KP; and (2) to examine the differences in clinical characteristics and outcome between BSIs caused by isolates with CTX-M versus other ESBL genotypes.
METHODS
Study Design and Patients
This analysis was part of the INCREMENT project, a retrospective cohort study designed to evaluate the epidemiology, clinical features, treatment efficacy, and prognosis of clinically significant BSI due to ESBL- or carbapenemase-producing Enterobacteriaceae.Reference Rodríguez-Baño, Navarro, Retamar, Picón and Pascual 9 In total, 33 tertiary-care hospitals in 12 countries participated in the project. For the present study, we included all adult patients (age>18 years) with clinically significant, monomicrobial BSIs caused by ESBL-EC or ESBL-KP from January 2004 through December 2013.
Variables and Definitions
Patient data were collected from hospital medical charts and included the following: demographic characteristics, ward type, site of acquisition (nosocomial, healthcare-associated or community), length of stay to bacteremia, source of infection (eg, urinary, biliary), comorbidities (individually and by Charlson score, dichotomized as ≤2 or >2),Reference Charlson, Pompei, Ales and MacKenzie 10 severity of underlying illness as measured by McCabe classification,Reference McCabe and Jackson 11 severity of BSI (presence of severe sepsis or septic shock and Pitt bacteremia score dichotomized as ≤4 or >4),Reference Hilf, Yu, Sharp, Zuravleff, Korvick and Muder 12 organism and resistance genotype, and 30-day mortality (calculated from the day of the first positive blood culture was taken). For patients who were no longer hospitalized at 30 days, we determined mortality by telephoning patients or their relatives and by consulting mortality registers.
Infections were defined as nosocomial if symptoms began >48 hours after hospital admission or within 48 hours of hospital discharge. Infections were defined as healthcare-associated if they were not nosocomial and if, in the previous 3 months, the patient was hospitalized in an acute-care hospital, long-term care facility or day hospital, had undergone dialysis, surgery or other invasive procedure, or received specialized home care. Patients who were neither nosocomial nor healthcare associated were considered strictly community acquired. The definitions of the Centers for Disease Control and Prevention (CDC) for nosocomial infections were used to classify the primary source of infection.Reference Garner, Jarvis, Emori, Horan and Hughes 13 Antimicrobial therapy was considered appropriate when including at least 1 drug active in vitro against the causative bacteria.
Laboratory Methods
Enterobacteriaceae strains were identified using standard microbiological techniques in each participating center. ESBL production and susceptibility to given antibiotics were screened and confirmed according to the guidelines of the Clinical and Laboratory Standards Institute (CLSI). 14 For some isolates, the ESBL genes had been characterized by polymerase chain reaction (PCR)Reference Tacconelli, Cataldo and Dancer 8 according to the needs of the local laboratories; no criteria were specified for the selection of the isolates. For secondary BSI, the source (eg, urine, wound) did not need to be microbiologically confirmed if enough clinical criteria were present.
Statistical Analysis
Risk factors for ESBL-EC versus ESBL-KP BSI and for death were initially examined by bivariate analysis. Categorical variables were compared using the χ2 test, and continuous variables were compered using the t test. Variables with P<0.10 were included in a multivariable logistic model. In addition, we decided a priori to include age and sex in the multivariable model. In the multivariate logistic model, P<0.05 was considered statistically significant.
SPSS version 22.0 software (SPSS, Chicago, IL) and SAS version 9.4 software (SAS Institute, Cary, NC) were used for this analysis.
Ethics
The Spanish Agency of Medicines and the Hospital Universitario Virgen Macarena Institutional Review Board approved the INCREMENT project. The need for written informed consent was waived. Approvals were also obtained from the institutional review boards of all participating centers.
RESULTS
Our sample included 909 patients: 687 (75%) with ESBL-EC and 222 (25%) with ESBL-KP BSI. The contributing institutes and the number of cases contributed by each center are summarized in Supplementary Table 1. The median age was 69 years (interquartile range [IQR], 56–79 years), and 55% of the patients were male. Only 149 patients (16%) had a strictly community-acquired infection. The median length of stay to bacteremia from admission was 1 day (IQR, 0–11 days). The urinary tract was the most common source of BSI (397 patients, 44%), followed by intra-abdominal infection (106, 12%) and the biliary system (104, 11%). Moreover, 44% of patients had a high Charlson comorbidity index (>2); 12.5% had a high Pitt bacteremia score (>4); and 35.5% developed severe sepsis or septic shock.
Risk Factors for BSI Caused by ESBL-EC Versus ESBL-KP
Table 1 compares epidemiological and clinical characteristics between patients with ESBL-EC BSI and ESBL-KP BSI. In bivariate analysis, patients with ESBL-KP were more likely to have cardiovascular, neurologic, or renal disease, and they were more likely to be in the ICU at the time of BSI diagnosis, to have a nosocomial BSI, and to have developed bacteremia >14 days after admission. A urinary source was more common in patients with ESBL-EC BSI. In the multivariable model, ESBL-KP BSI was independently associated with ICU admission, cardiovascular disease, neurological disease, later onset of bacteremia, and a nonurinary source.
TABLE 1 Demographic and Clinical Characteristics of Patients With ESBL-EC and ESBL-KP Bloodstream InfectionFootnote a
NOTE. ESBL, extended-spectrum β-lactamase; EC, Escherichia coli; KP, Klebsiella pneumoniae; IQR, interquartile range; OR, odds ratio; ICU, intensive care unit; HIV, human immunodeficiency virus; LOS, length of stay.
a Percentage of persons with known data.
b Odds ratio (OR) is for ESBL-KP infection.
c Death expected within 1 year or 5 years, as indicated.
In a subgroup analysis of 423 patients with nosocomial BSI, the same 5 risk factors were significant in a multivariable model (data not shown). No significant risk factors were found for patients with healthcare-associated community acquisition. Subgroup analysis of patients with strict community acquisition was not done due to small number of patients (21 patients with ESBL-KP).
Thirty-Day Mortality Following ESBL-EC Versus ESBL-KP BSI
The 30-day mortality rate was higher among patients with ESBL-KP BSI than with ESBL-EC BSI (33.7% vs 17.4%). Differences in mortality risk factors between ESBL-EC and ESBL-KP BSI are summarized in Table 2A and 2B. Septic shock/severe sepsis, bad prognosis according to McCabe classification, and inappropriate targeted therapy were risk factors for mortality in both ESBL-EC and ESBL-KP. Other risk factors for mortality, age and nonurinary source, were significant only in ESBL-EC. However, the odds ratios (ORs) exhibited a similar trend in ESBL-KP but did not reach statistical significance due to the smaller cohort. ESBL-KP remained an independent predictor of 30-day mortality (OR, 1.64; 95% confidence interval [CI], 1.1–2.5). In the healthcare-associated community-acquisition subgroup, mortality rates were also significantly higher in ESBL-KP versus ESBL-EC (27% vs 14%; P=.0125).
TABLE 2 Risk Factors for 30-Day Mortality in Patients With ESBL-EC (TABLE 2A) and ESBL-KP (TABLE 2B ) Bloodstream Infection by Univariate and Multivariate Analyses TABLE 2A
NOTE. ESBL, extended-spectrum β-lactamase; EC, Escherichia coli; KP, Klebsiella pneumoniae; OR, odds ratio; ICU, intensive care unit; HIV, human immunodeficiency virus.
a OR is for ESBL-KP infection.
Differences in Clinical Characteristics and Outcome by ESBL Genotype
A total of 286 bacterial isolates (218 EC and 68 KP) underwent PCR amplification of bla ESBL genes. No significant differences were found between patients whose isolates had and had not their ESBL genes characterized in terms of demographics, underlying conditions, bacterial species, sources of infection, severity of infection, or mortality (data not shown). CTX-M was the most prevalent ESBL subtype identified (218, 76%), both in E. coli (78%) and in K. pneumoniae (69%). The other ESBL genotypes present in our sample were SHV and TEM (22% of E. coli and 31% of K. pneumoniae). Table 3 compares clinical characteristics and outcomes in patients with BSI caused by bacteria producing CTX-M versus other ESBL enzymes. In bivariate analysis, non–CTX-M was significantly associated with pulmonary, cardiovascular, neurological, and connective tissue disease and with total Charlson score >2. We did not find significant differences in clinical severity of the BSI, site of acquisition, or mortality rate between CTX-M and non–CTX-M ESBLs.
TABLE 3 Epidemiological and Clinical Characteristics of Patients With ESBL-E Blood Stream Infection Characterized by PCR Amplification of bla ESBL Genes: CTX-M Versus Non–CTX-M
NOTE. ESBL-E, extended-spectrum β-lactamase-producing Enterobactericeae; KP, Klebsiella pneumoniae; OR, odds ratio; ICU, intensive care unit; HIV, human immunodeficiency virus; PCR, polymerase chain reaction.
Of the 68 patients, 13 (19%) with non–CTX-M were strictly community acquired compared with 30 of 218 patients with CTX-M (14%; P=.28). In addition, 10 patients (n=13) were from a single center, which might reflect local epidemiology and not a general trend.
DISCUSSION
In this large cohort of patients with ESBL-E BSI, we found differences between patient characteristics, clinical presentation, and outcome, depending on whether the causative organism was for ESBL-EC or ESBL-KP. Bacteremia due to ESBL-KP was more often of nonurinary origin, occurred later during hospitalization, was associated with comorbidities (cardiovascular and neurological) and had a more severe clinical presentation. Bacteremia due to ESBL-EC tended to occur more often upon admission (community onset) and to have a urinary source. Our results are in accordance with a previous study by Freeman et alReference Freeman, Rubin and McAuliffe 15 who demonstrated differences in risk profiles and clinical characteristics between patients with ESBL-EC BSI (eg, community-acquired infections) and patients with ESBL-KP BSI (eg, ICU admission).Reference Freeman, Rubin and McAuliffe 15
Gram-negative bacillary sepsis with shock has a mortality rate of 12%–38%.Reference Martelius, Jalava, Kärki, Möttönen, Ollgren and Lyytikäinen 2 , Reference Sakellariou, Gürntke and Steinmetz 16 In our cohort, the 30-day mortality rate was 21.4% and was significantly higher for ESBL-KP BSI (33.7%) than for ESBL-EC BSI (17.4%). We found that bacterial species is among the most important determinants of the risk for mortality. A study conducted in Finland likewise found higher 28-day mortality in patients with nosocomial BSI caused by ESBL-KP (28.0%) than by ESBL-EC (14.8%).Reference Martelius, Jalava, Kärki, Möttönen, Ollgren and Lyytikäinen 2 In contrast, Leistner et alReference Sakellariou, Gürntke and Steinmetz 16 did not find a difference in in-hospital mortality between BSI caused by ESBL-KP and ESBL-EC, although they did find significantly higher mortality with ESBL-negative KP than with ESBL-negative EC.
Here, CTX-M was the most prevalent type of ESBL in our sample, as in other studies.Reference Sakellariou, Gürntke and Steinmetz 16 There was no significant association between the type of ESBL (CTX-M vs non–CTX-M) and the organism or between the type of ESBL and whether the infection was nosocomial. These findings are consistent with previous data.Reference Xia, Fan and Huang 17
Epidemiologically, it has been suggested that CTM-X β-lactamase has higher rates of community transmission and an association with E. coli.Reference Xia, Fan and Huang 17 Our data support studies indicating that CTX-M is also prevalent in ESBL-KP.Reference Gürntke, Kohler and Steinmetz 18 We found that non–CTX-M was associated with pulmonary, cardiovascular, neurological, and connective-tissue diseases. It was also associated with total Charlson score >2, which may suggest that these patients’ infections were more healthcare associated. We conclude that, today, both ESBL-EC and ESBL-KP contain the CTX-M and non–CTX-M subtypes and that infections with all ESBL genotypes can occur in the community and in the hospital settings.
Our study has several strengths. The data were acquired from different hospitals in several countries, and the sample size was large. However, this study also has several limitations. The retrospective design of the study could be subject to information bias. However, we believe that the large number of patients included at least partly overcomes this problem. We analyzed the data from all countries together, but the epidemiology of ESBL genotypes may vary between countries. The ESBL genes were only characterized by PCR in a subgroup of isolates; however, the features of these patients were similar to those for which the ESBL gene were not studied.
In conclusion, the epidemiology, clinical characteristics, and risk of mortality differ significantly between ESBL-EC and ESBL-KP, and these factors do not differ by ESBL genotype. Therefore, we recommend that ESBL-E should not be considered a homogenous group. In addition to the different epidemiology of the bacteria, the clinical course differs; thus, different treatment options may be needed. Further studies should focus on individual bacteria and not on ESBL-E as a group.
ACKNOWLEDGMENT
We thank Hagit Mishali for her contribution to the study. We also thank the European Study Group of Bloodstream Infections and Sepsis (ESGBIS) from the European Society of Clinical Microbiology and Infectious Diseases (ESCMID) for endorsing the INCREMENT project.
Financial support: The study was funded by Plan Nacional de I+D+i 2013–2016 and Instituto de Salud Carlos III, Subdirección General de Redes y Centros de Investigación Cooperativa, Ministerio de Economía, Industria y Competitividad, Spanish Network for Research in Infectious Diseases (grant nos. REIPI DR12/0015/0010 and RD16/0016/0001) and was cofinanced by the European Development Regional Fund “A Way to Achieve Europe” Operative Program for Intelligent Growth, 2014–2020.
Potential conflicts of interest: All authors report no conflicts of interest relevant to this article.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/ice.2018.63.
