Background. Extended-spectrum β-lactamase (ESBL)–producing Escherichia coli, particularly those producing CTX-M types of ESBL, are emerging pathogens. Bacteremia caused by these organisms represents a clinical challenge, because the organisms are frequently resistant to the antimicrobials recommended for treatment of patients with suspected E. coli sepsis.
Methods. A cohort study was performed that included all episodes of bloodstream infection due to ESBL-producing E. coli during the period from January 2001 through March 2005. Data on predisposing factors, clinical presentation, and outcome were collected. ESBLs were characterized using isoelectric focusing, polymerase chain reaction, and sequencing.
Results. Forty-three episodes (8.8% of cases of bacteremia due to E. coli) were included; 70% of the isolates produced a CTX-M type of ESBL. The most frequent origins of infection were the urinary (46%) and biliary tracts (21%). Acquisition was nosocomial in 21 cases (49%), health care associated in 14 cases (32%), and strictly community acquired in 8 cases (19%). Thirty-eight percent and 25% of patients had obstructive diseases of the urinary and biliary tracts, respectively, and 38% had recently received antimicrobials. Nine patients (21%) died. Compared with β-lactam/β-lactamase–inhibitor and carbapenem-based regimens, empirical therapy with cephalosporins or fluoroquinolones was associated with a higher mortality rate (9% vs. 35%; P = .05) and needed to be changed more frequently (24% vs. 78%; P = .001).
Conclusions. ESBL-producing E. coli is a significant cause of bloodstream infection in hospitalized and nonhospitalized patients in the context of the emergence of CTX-M enzymes. Empirical treatment of sepsis potentially caused by E. coli may need to be reconsidered in areas where such ESBL-producing isolates are present.
Extended-spectrum β-lactamases (ESBLs) are plasmid-encoded β-lactamases that confer resistance to penicillins, cephalosporins, and aztreonam. They are frequently found in Klebsiella species, Escherichia coli, and other enterobacteria . Antimicrobial therapy for infections caused by ESBL-producing organisms presents an additional challenge, because these organisms are also often resistant to other antimicrobials, such as trimethoprim-sulfamethoxazole, aminoglycosides, and fluoroquinolones. Until recently, the major problems posed by ESBL-producing organisms were related to nosocomial infections caused by Klebsiella pneumoniae, which produced mainly the TEM and SHV types of ESBL . However, in recent years and throughout most of the world, ESBL-producing E. coli—particularly strains producing CTX-M types of ESBL—have dramatically increased as a cause of infection [1, 2]. In contrast with infection caused by ESBL-producing K. pneumoniae, approximately one-half of infections caused by ESBL-producing E. coli affect nonhospitalized patients [3–7]. In addition, the epidemiology of infections caused by CTX-M–producing isolates has been shown to be different from infections due to TEM- or SHV-producing types of ESBLs: genes encoding CTX-M enzymes are associated with mobile genetic elements that favor the ease with which these enzymes spread through the community ; also, organisms producing CTX-M enzymes have been less frequently associated with clonal spread than those producing SHV or TEM ESBLs [8, 9].
Cephalosporins and fluoroquinolones are recommended as first-line empirical therapies for the treatment of community-acquired and nosocomially acquired sepsis originating in the urinary tract, polymicrobial soft-tissue infections, and intra-abdominal infections [10–12]. These recommendations would be challenged if a significant proportion of such infections turned out to be caused by ESBL-producing E. coli. In fact, recent data from our area indicate that 12%–16% of patients with infections due to ESBL-producing E. coli are bacteremic [8, 9]. However, little is known about the frequency of, risk factors for, and clinical features of bloodstream infection due to ESBL-producing E. coli in the context of the emergence of CTX-M enzymes. Therefore, we performed this study with the objective of analyzing the clinical and molecular epidemiology, predisposing factors, clinical features, and outcomes of bloodstream infections due to ESBL-producing E. coli.
The study was conducted in the Hospital Universitario Virgen Macarena, a 950-bed teaching hospital providing acute care for a population of 550,000 persons in Seville, Spain.
Study Design and Patients
A prospective cohort study was designed to include all episodes of bacteremia due to ESBL-producing E. coli that occurred during the period from January 2001 through March 2005. Cases were detected through the daily review of blood culture results. Demographic features and predisposing factors were obtained by reviewing the hospital and primary care clinical charts and interviewing, in all cases, the patient or their closest relatives, using a previously tested, structured questionnaire . The outcome was assessed prospectively. The following data were collected: age, sex, admission to any hospital during the previous year, nursing home residency, receipt of hemodialysis, home care, other types of health care contact, comorbidities, severity of underlying diseases [13, 14], surgery during the previous year, use of a vascular or urinary catheter, other invasive procedures performed during the previous 2 weeks, receipt of antimicrobial agents in the previous 2 months, source of bacteremia (determined using the Centers for Disease Control and Prevention [CDC] criteria ), acute severity of disease at presentation (according to the Pitt score ), presence of severe sepsis or septic shock , antimicrobial treatment received, and outcome.
The study was observational; however, all episodes of bacteremia in our hospital are notified and followed up by a clinical microbiologist or infectious diseases physician. Changes in antimicrobial treatment and general management were advised if considered necessary. The study was approved by the local ethics committee.
Bloodstream infections were primarily classified as nosocomially or community acquired, in accordance with the classic CDC criteria . Episodes of community-acquired bacteremia, in accordance with CDC criteria, were further classified as health care associated if any of the following criteria were present : >48-h hospital admission during the previous 90 days, receipt of hemodialysis, receipt of intravenous medication or home wound care in the previous 30 days, and residence in a nursing home or long-term care facility. Otherwise, cases were considered to be strictly community acquired. Underlying diseases were defined on the basis of standard criteria. Antimicrobial treatment was considered to be empirical if it was administered before susceptibility data were available. Empirical treatment was defined as inappropriate when an active antimicrobial agent (as determined by in vitro susceptibility testing) prescribed at the usual recommended dose had not been administered during the first 48 h. Oxyimino-β-lactams (cefuroxime, cefotaxime, ceftriaxone, ceftazidime, and aztreonam) were considered to be inappropriate regardless of the MIC. A change in empirical therapy was considered to be necessary and was recommended when the 2 following criteria were met: the patient had persistent signs of sepsis by the date when susceptibility data became available, and the organism demonstrated in vitro resistance to the administered antimicrobial. Crude mortality included all causes of death during admission; 14-day mortality included all deaths within 14 days after the date of the first positive blood culture result.
Bacterial strains and susceptibility assays. The first isolate obtained during each episode was studied. Identification was determined by the Vitek 2 system and API 20E strips (bioMérieux). ESBL production was screened and confirmed in accordance with CLSI standards . The in vitro activity of antimicrobial agents was determined using a microdilution assay . ESBL-producing organisms were considered to be multidrug resistant if they demonstrated in vitro resistance to ⩾3 of the following agents: amoxicillin-clavulanate, piperacillin-tazobactam, ciprofloxacin, gentamicin, amikacin, and trimethoprim-sulfamethoxazole.
Molecular typing. The clonal relationship between the isolates was determined by repetitive extragenic palindromic (REP)–PCR, as described elsewhere . Isolates were considered to be clonally related when band patterns differed by <3 bands. Isolates that were determined to be clonally related by REP-PCR were also studied by PFGE  using XbaI endonuclease (Roche Applied Sciences), and the findings were interpreted in the manner described by Tenover et al. .
β-Lactamase characterization. Isoelectric focusing [23–25] and PCR were used for the preliminary characterization of β-lactamases and β-lactamase genes, respectively. The presence of blaTEM, blaSHV, and blaCTX-M in each organism was studied by PCR, as described elsewhere . Oligonucleotide primers designed to amplify the genes that encode the most-common subgroups in the ESBL families were used [26–28]. E. coli J53 Rif-R was used as a negative control. Amplicons were sequenced in an external center (DNA Automatic Sequencing Service, Consejo Superior de Investigaciones Científicas; Madrid, Spain), which is equipped with an ABI PRISM 377 sequencer (Applied Biosystems). Sequences were analyzed using the Chromas application, the Basic Local Alignment Search Tool , and the Traduction Multiple program .
Continuous variables were compared using the Mann Whitney U test. Categorical variables were compared using the χ2 test or Fisher's exact test, as appropriate. Relative risks for mortality and 95% CIs were calculated. Adjusted ORs were calculated using logistic regression analysis. The data were analyzed using the SPSS software package (SPSS).
Epidemiology of bacteremia due to ESBL-producingE. coli. During the study period, there were 43 cases of bloodstream infection caused by ESBL-producing E. coli, and all cases were included in our study. One patient experienced 2 episodes separated by a 3-month interval. The distribution of the cases throughout the study period is shown in figure 1. The number of cases per year increased from 6 cases in 2001 to 16 cases in 2004. Twenty-two cases (51%) were community acquired, as defined using CDC criteria. During the study period, 8.8% of the episodes of bacteremia due to E. coli were caused by ESBL-producing isolates (12.9% and 6.5% of nosocomially and community-acquired episodes of E. coli bacteremia, respectively).
Of the 21 nosocomially acquired cases, 11 (52%) were in a medical service, 9 (43%) were in a surgical service, and 1 (5%) was in the intensive care unit (ICU). The median duration of the previous hospital stay was 26 days (range, 4–58 days). Of the 22 community-acquired cases, 14 were considered to be health care associated (32% of the total series), and 8 (19%) were considered to be strictly community acquired. Among the 14 patients with health care–associated episodes, 13 had been previously admitted to the hospital, 3 were nursing home residents, and 1 received hemodialysis.
Features of the case patients, according to acquisition category, are shown in table 1. In brief, patients with strictly community- and health care–associated episodes of infection had very similar characteristics, but some differences were found between patients with nosocomial episodes and those with strictly community-acquired episodes: severe underlying disease, neoplasia, and previous use of antimicrobials (particularly oxyimino-β-lactams) were more common among patients with nosocomial episodes of infection, whereas recurrent urinary tract infections were more common among patients with strictly community-acquired episodes.
Types of ESBL and molecular epidemiology. Susceptibility results and types of ESBL produced by the isolates are shown in table 2. ESBLs and their molecular relationships were studied in 37 isolates; 25 (70%) produced a CTX-M enzyme (24 of which were CTX-M-14 and 1 of which was CTX-M-9), and 12 (32%) produced an SHV enzyme (10 of which were SHV-12 and 2 of which were SHV-4). One isolate produced both CTX-14 and SHV-12. TEM-1 was present in 23 isolates, and TEM-2 was present in 1, but no TEM-type ESBLs were found. Although the differences were not statistically significant, isolates recovered from patients with strictly community-acquired episodes had the highest frequency of CTX-M enzymes and the lowest frequency of SHV enzymes (table 2). We found no other significant differences among the patients with regard to the type of ESBL produced. As determined using REP-PCR, all isolates were found to be genetically unrelated, except for the 2 strains isolated during different episodes in the same patient that had the same genetic profile, as well as for 3 other nosocomial isolates that were also found to be genetically related. Isolates that were considered to be clonally related by REP-PCR were also studied by PFGE, with identical results.
Clinical features and outcome. Clinical features and outcomes of the case patients are presented in table 2. The clinical presentation was somewhat less severe in patients with strictly community-acquired cases. The crude and 14-day mortality rates were 21% and 19%, respectively. Mortality occurred only in nosocomial and health care–associated cases. The mortality rate was 14% (4 of 29 patients) for episodes in which the origin was the urinary or biliary tract, and it was 36% (5 of 14 patients) for episodes with other origins.
Although, in all cases, empirical antimicrobial treatment was determined to be adequate on the basis of local guidelines, it was considered appropriate for only 51% of the patients according to in vitro susceptibility test results. Appropriate empirical therapy was administered more frequently to patients with strictly community-acquired episodes than for those with nosocomial episodes (table 2). Monotherapy was empirically administered to 31 patients, as follows: cephalosporins, 10 patients (4 of whom died); β-lactam/β-lactamase–inhibitors, 14 patients (2 of whom died); ciprofloxacin, 4 patients (2 of whom died); imipenem, 3 patients (all of whom survived); and combination therapy, 12 patients (1 of whom died). The outcomes of patients who received empirical treatment with cephalosporins or β-lactam/β-lactamase–inhibitors, according to the MIC of the specific drug for the isolate, are shown in table 3. The crude mortality rate was 35% (7 of 20 patients) among patients who received empirical treatment with a cephalosporin or fluoroquinolone-based regimen, and it was 9% (2 of 23) among those treated with a β-lactam/β-lactamase–inhibitor or carbapenem (relative risk, 4.0; 95% CI, 0.9–17.2; P = .05). After controlling for source and severity at presentation, the adjusted OR was 9.2 (95% CI, 1.2–70.2; P = .03).
If we consider patients who were alive when the susceptibility results were available, empirical antimicrobial treatment required changes more frequently among patients who had received a ciprofloxacin- or cephalosporin-based regimen than among those treated with β-lactam/β-lactamase–inhibitors or carbapenem (14 [78%] of 18 patients vs. 5 [24%] of 21 patients; relative risk, 3.6; 95% CI, 1.4–9.1; P = .001), even after adjusting for source of infection and severe sepsis or shock at presentation (adjusted OR, 10.7; 95% CI, 2.3–49.9; P = .002).
The clinical relevance of infections caused by ESBL-producing organisms has been outlined in several studies [1, 31, 32]. These studies exclusively or predominantly included infections caused by ESBL-producing K. pneumoniae. However, the recent emergence of ESBL-producing E. coli, related to the dissemination of genetic elements harboring CTX-M types of ESBL, is a new and distinct epidemiologic phenomenon [2, 7, 33].
Our study is, to our knowledge, the first to investigate the clinical and molecular epidemiology of bloodstream infections due to ESBL-producing E. coli and to characterize such organisms as a cause of community-acquired bacteremia in the present epidemiologic context. Association with health care, ESBL characterization, and clonality had not been investigated in the only 2 previous studies of bacteremia due to ESBL-producing E. coli. [34, 35]. Other studies have included both ESBL-producing E. coli and K. pneumoniae , but our data support the idea that the epidemiology of ESBL-producing E. coli is significantly different from that of K. pneumoniae . In fact, bacteremia due to ESBL-producing K. pneumoniae is typically nosocomially acquired [1, 16, 37] and caused by isolates producing SHV or TEM types of ESBL [37, 38], whereas only one-half of the episodes caused by ESBL-producing E. coli included in our study were nosocomially acquired, and 70% of the isolates produced CTX-M types of ESBL. These features, added to the fact that most isolates were clonally unrelated and that ciprofloxacin resistance was common, reflect the general epidemiologic situation of ESBL-producing E. coli in many areas [3, 6–9]. However, because clonal spread has been reported in other areas [4, 5], and because of the ease with which these enzymes are spreading , there is general concern that the situation may get worse in the future.
Our data indicate that, in this situation, ESBL-producing E. coli is a significant cause of nosocomial, health care–associated, and strictly community-acquired bloodstream infections. It is noteworthy that, contrary to most drug-resistant organisms, ESBL-producing E. coli is not yet frequently affecting ICU patients [9, 34, 35]. One of the main findings of our study is the fact that 19% of episodes were strictly community acquired, meaning that a multidrug-resistant pathogen may be the cause of serious community-acquired infections. Also, two-thirds of the so-called community-acquired cases, as defined according to classic CDC criteria, occurred in patients in association with health care, but these cases were similar to community-acquired cases with regard to other predisposing factors. It is possible that patients with health care–related cases acquired ESBL-producing E. coli during health care contact, but because CTX-M–producing E. coli is frequently a true community pathogen [2, 4, 5, 7, 8], the health care relation may have simply revealed patients who are at a higher risk of bacteremia among those who had previously been colonized with such strains [9, 39].
Previous antimicrobial use was common. Cefotaxime was the most commonly used oxyimino-β-lactam, a fact that may have favored the selection of CTX-M–producing isolates, because such types of ESBL are mainly cefotaximases [1, 2]. Fluoroquinolone use, which was also frequent, has been identified as a risk factor for other infections caused by ESBL-producing E. coli [8, 9], probably through the action of coselection .
The clinical importance of cephalosporin resistance posed by ESBL-producing organisms (even for apparently susceptible isolates) is now well established [1, 31]. Our data suggest this is also true for bacteremia caused by ESBL-producing E. coli in the setting of CTX-M enzymes: 4 of 10 patients who received empirical monotherapy with a cephalosporin died. Patients were treated with the recommended doses of cephalosporins, and although these data should be considered with caution because of the low number of patients, we could not find a trend for better survival among patients whose isolates had lower MICs. Also, ciprofloxacin therapy is associated with a worse prognosis for bacteremia due to ESBL-producing K. pneumoniae . We found that empirical therapy with cephalosporins or fluoroquinolones was associated both with a higher mortality rate and with a greater probability that a change in antimicrobial therapy would be required. Whether β-lactam/β-lactamase–inhibitor combinations are a good option is controversial, because these combinations are subject to the inoculum effect and may be affected by other resistance mechanisms . However, only 1 of 11 patients treated with intravenous amoxicillin–clavulanic acid died. We have previously reported that amoxicillin–clavulanic acid is much less affected by the inoculum effect than is piperacillin/tazobactam . However, the empirical use of this combination is hampered, because a significant proportion of ESBL-producing E. coli strains are resistant in vitro.
Inappropriate empirical therapy has been associated with increased mortality among patients with non–urinary tract infections caused by ESBL-producing E. coli and Klebsiella species , but appropriate empirical therapy may also be relevant for bacteremic urinary tract infections: 3 of 9 patients in our study who died had an infection of urinary tract origin, and empirical treatment was inappropriate in each case. Thus, carbapenems are probably the best options for treating bacteremia caused by ESBL-producing E. coli. As long as the emergence of carbapenem resistance is a concern , and bearing in mind that E. coli is the most common and fifth-most common cause of community- and nosocomially acquired bacteremia, respectively [18, 42], any recommendation regarding the empirical treatment of such infections must made with caution. We believe that the percentage of ESBL-producing strains among the E. coli strains found in our study (nearly 9%) is not yet enough to recommend the use of carbapenems for all serious infections caused by E. coli. However, until more data are available, we think that these drugs should be recommended in the empirical treatment of sepsis in patients with predisposing factors for bacteremia caused by ESBL-producing E. coli, as occurred in our study, in areas where the latter has been determined to be a relevant pathogen on the basis of local surveillance data. For some patients with urinary tract sepsis, addition of an aminoglycoside to a cephalosporin or fluoroquinolone, to broaden the empirical coverage until susceptibility data are available, may also be an option; however, more data are needed before this approach can be recommended.
The results of our study must be interpreted in light of its limitations. It is possible that some differences between subgroups may not have been detected as a result of the low number of patients involved. This is a not a randomized, controlled study; thus, comparisons of outcomes among the different antimicrobials might be biased. We did not compare the outcome of episodes caused by ESBL-producing E. coli with the outcome of episodes caused by non–ESBL-producing isolates, because this was beyond our objectives. Finally, our study was performed in a specific geographical location, and the results may not be applicable to areas with different epidemiologic situations.
In conclusion, ESBL-producing E. coli (particularly those isolates that produce CTX-M enzymes) is an emergent cause of nosocomial, health care–associated, and community-acquired bacteremia in areas where such organisms are present. In these areas, the protocols for empirical treatment of E. coli sepsis should be revised.
Financial support. REIPI (Spanish Network for Research in Infectious Diseases), Instituto de Salud Carlos III, Ministerio de Salud y Consumo (C03/14), FIS PI051019, and Junta de Andalucía (75/04). M.D.N. was the recipient of a fellowship from the Asociación Sanitaria Virgen Macarena.
Potential conflicts of interest. J.R.-B. has been a consultant for Chiron, Wyeth, Merck, and Pfizer and has served as speaker for Wyeth, Merck, Pfizer, and GlaxoSmithKline. A.P. has been a consultant for Merck and Pfizer; has served as speaker for Wyeth, Astra-Zeneca, Merck, and Pfizer; and has received research support from Merck and Pfizer. All other authors: no conflicts.