Skip Navigation

Molecular Correlation for the Treatment Outcomes in Bloodstream Infections Caused by Escherichia coli and Klebsiella pneumoniae with Reduced Susceptibility to Ceftazidime

  1. Annie Wong-Beringer1,
  2. Janet Hindler2,
  3. Michael Loeloff5,
  4. Anne Marie Queenan5,
  5. Nancy Lee3,
  6. David A. Pegues4,
  7. John P. Quinn6, and
  8. Karen Bush5
  1. 1Western University of Health Sciences, College of Pharmacy, Pomona, California
  2. 2Departments of Clinical Microbiology, University of California Los Angeles Medical Center, Los Angeles, California
  3. 3Pharmaceutical Services, University of California Los Angeles Medical Center, Los Angeles, California
  4. 4Medicine, University of California Los Angeles Medical Center, Los Angeles, California
  5. 5The R. W. Johnson Pharmaceutical Research Institute, Raritan, New Jersey
  6. 6University of Illinois, Chicago
  1. Reprints or correspondence: Dr. Annie Wong-Beringer, Western University of Health Sciences, College of Pharmacy, 309 East Second Street, Pomona, CA 91766 (anniewb{at}westernu.edu).
 
Next Section

Abstract

Data are limited on outcomes of treatment with extended-spectrum cephalosporins (ESCs) for infections caused by Enterobacteriaceae that produce extended-spectrum β-lactamases (ESBLs). This study describes the largest treatment experience of a nonoutbreak series of bloodstream infections caused by strains of Escherichia coli (23 episodes) and Klebsiella pneumoniae (13 episodes) with a ceftazidime minimal inhibitory concentration of ⩾2 µg/mL. E. coli isolates produced a greater variety of β-lactamase types than did K. pneumoniae isolates, among which ESBL production was predominant. Five ESBL types were identified: TEM-12, TEM-71, TEM-6, SHV-12, and SHV-5. Most patients were treated empirically with an ESC-based regimen. A favorable response to treatment with a nonceftazidime ESC was observed when the causative pathogen produced either TEM-6 or TEM-12; ceftazidime treatment was associated with failure of therapy in all patients. Despite the limited clinical success, ESCs are currently not recommended for the treatment of serious infections caused by ESBL-producing strains.

During the past 2 decades, antibiotic-resistant mutant strains that produce extended-spectrum β-lactamases (ESBLs) have emerged among the Enterobacteriaceae, predominantly Escherichia coli and Klebsiella pneumoniae. The emergence of such strains has important clinical and therapeutic implications. First, ESBLs are often derived from TEM and SHV enzymes, which are present in 75% of Enterobacteriaceae [1]. The resistance determinants for ESBLs are often found on transmissible plasmids, which facilitate the spread of the determinants to other organisms [1]. Second, because of the lack of an obvious marker to indicate the presence of such enzymes, routine susceptibility testing may not detect the presence of ESBLs. To improve laboratory detection, the National Committee on Clinical Laboratory Standards (NCCLS) currently recommends that laboratories perform special screening to identify possible ESBL producers, using a break point MIC of ⩾2 µg/mL for ⩾1 of the following antimicrobial agents: cefpodoxime, cefotaxime, ceftriaxone, ceftazidime, or aztreonam. This testing should be followed by a confirmatory test that depends on the observation of increased susceptibility to cefotaxime or ceftazidime in the presence of clavulanic acid [2]. However, according to a recent survey, many microbiology laboratories currently do not use an adequate screening method for detection of ESBL-producing organisms [3].

Adding to the challenge is the problem of selecting appropriate antibiotics for the treatment of serious infections due to these organisms. Depending on the ESBL type or types produced, these organisms can exhibit various levels of in vitro resistance to cephalosporins [4, 5]. In addition, strains may be resistant to multiple other classes of antibiotics [5]. Reports of treatment failure despite apparent in vitro susceptibility have called into question the usefulness of the extended-spectrum cephalosporins (ESCs) for the treatment of serious infections due to these organisms [6]. Numerous studies have been published describing the epidemiology, laboratory detection methods, and molecular characterization of the ESBLs; however, few have presented sufficient clinical data establishing the relationship between the ESBL type and treatment outcomes, particularly with the ESCs.

In this study, we evaluated 36 cases of bloodstream infection due to presumptive ESBL-producing E. coli and K. pneumoniae that occurred during a 12-year period and were not associated with an outbreak of such infections. The primary objectives of this study were to characterize the phenotypes of the study organisms, to evaluate the performance characteristics of ESBL screening and confirmatory tests, and to examine the molecular basis for treatment outcomes, with a focus on ESCs.

Previous SectionNext Section

Patients, Materials, and Methods

Patients and bacterial strains. We searched the computerized microbiology database at the University of California Los Angeles (UCLA) Medical Center, a 750-bed tertiary-care facility, for the 12-year period 1986–1997 to identify E. coli and K. pneumoniae blood isolates with ceftazidime MICs of ⩾2 µg/mL. During this period, samples of E. coli and K. pneumoniae strains (isolated from various body sites) had been saved if they had elevated ceftazidime MICs and were suspected to have unusual resistance. Saved isolates had been stored in Brucella broth supplemented with 15% glycerol at -70°C until retrieved for the present study. Blood isolates from study patients were subsequently subjected to phenotypic and molecular analysis. Results of antimicrobial susceptibility tests and ESBL screening and confirmatory tests were used to characterize the phenotypes of the saved isolates. Molecular analysis included performance of isoelectric focusing studies, outer membrane protein (OMP) characterization, and PCR amplification and sequencing of PCR products for blaTEM and blaSHV.

Treatment outcomes were assessed by reviewing the medical records of all patients for whom a blood culture yielded a strain of E. coli and K. pneumoniae with a ceftazidime MIC of ⩾2 µg/mL. Hospitalized patients who had bloodstream infections and had received ⩾72 h of antimicrobial therapy that was targeted for gram-negative organisms were considered assessable for outcome analysis. Information extracted from medical charts included the following: demographic data, presence of comorbid illnesses, findings of physical examination, presenting symptoms, vital signs, laboratory test results, and radiographic findings. In addition, any history of recent surgery, presence of invasive devices (e.g., a central venous catheter), and ceftazidime exposure ⩽30 days before the first positive blood culture result were noted. The antimicrobial treatment regimen was recorded, including the agent or agents administered, the duration of treatment, and clinical response.

Antimicrobial susceptibility testing. Antimicrobial susceptibility was determined by broth microdilution MIC tests according to the procedures described by NCCLS in M7-A5 [2]. The organisms were tested against a wide variety of antimicrobial agents to determine in vitro susceptibility: gentamicin, tobramycin, amikacin, netilmicin; ciprofloxacin, ofloxacin, levofloxacin; doxycycline; trimethoprim-sulfamethoxazole; imipenem, meropenem; ampicillin, ampicillin-sulbactam, piperacillin, piperacillin-tazobactam, ticarcillin, ticarcillin–clavulanic acid; cefazolin; cefoxitin, cefotetan; cefotaxime, ceftizoxime, ceftriaxone, ceftazidime; aztreonam; and cefpodoxime. Tests were also performed with combinations of clavulanic acid (concentration, 4 µg/mL) and cefotaxime, ceftriaxone, aztreonam, ceftazidime, or cefpodoxime. Quality control was performed by testing E. coli ATCC 25922 and K. pneumoniae K6, as controls. Inoculated plates were incubated in ambient air at 35°C for 16–20 h. The MIC of each antimicrobial agent or combination of agents was defined as the lowest concentration that inhibited visible growth of the organism.

Screening and confirmatory tests for ESBL-producing strains. To screen for ESBL production, we used a break point of ⩾2 µg/mL for aztreonam, ceftazidime, cefpodoxime, cefotaxime, or ceftriaxone. The original E-test ESBL test strips (i.e., the first version of the strips) containing ceftazidime with and without clavulanic acid (AB Biodisk) were used according to the manufacturer's instructions. In addition, ESBL production was confirmed with use of broth microdilution tests with ceftazidime or cefotaxime, each alone and in combination with clavulanic acid, as stated above. These tests were considered positive for ESBL production if the MIC for ceftazidime or cefotaxime in combination with clavulanic acid was ⩾3 2-fold dilutions lower than the MIC for either drug alone. The performance characteristics of the ESBL screen, ESBL E-test, and broth microdilution tests were determined on the basis of genotype analysis, and the sensitivity, specificity, positive predictive values, and negative predictive values were calculated.

β-Lactamase extraction and isoelectric focusing studies. Crude extracts were prepared from log phase cultures grown in trypticase soy broth. A cell suspension in 0.2 M sodium acetate (pH 5.5); extracts were subjected to 4 freeze-thaw cycles, then centrifuged. We determined isoelectric points for β-lactamases in the supernatants by means of isoelectric focusing, using Multiphor apparatus and Ampholine PAGplates (Amersham-Pharmacia) with a pH of 3.5–9.5. As standards, we used TEM-1, TEM-26, SHV-1, and P99 β-lactamases. β-Lactamase activity was detected by use of a nitrocefin overlay. AmpC-type cephalosporinases were identified by observing β-lactamase inhibition when an overlay of aztreonam (concentration, 1 µg/mL) was applied 10 min before the application of nitrocefin.

OMP characterization and transfer of resistance. Selected E. coli isolates underwent characterization of OMP and transfer of resistance. Bacterial cell membranes were isolated from strains grown overnight in trypticase soy broth by means of sonication (3 pulses of 30 s alternated with 15 s of cooling on ice). The suspension was then centrifuged at 100,000 g for 40 min at 4°C. OMPs were separated from total membrane preparations by solubilizing the inner membrane with 20% Sarkosyl, followed by centrifugation at 40,000 g for 30 min at 4°C. These pellets were then resuspended in 1% Sarkosyl and centrifuged a second time. The inner membrane was dissolved facilitating the extraction of the insoluble outer membrane. OMP was analyzed by use of electrophoresis on a 10% sodium dodecyl sulfate–polyacrylamide gel and visualized by staining with 0.125% Coomassie blue stain.

Mating experiments were performed with E. coli C600RifR/NAR as the recipient. Ceftazidime (concentration, 10 µg/mL), nalidixic acid (concentration, 50 µg/mL), and rifampin (concentration, 60 µg/mL) were used as selecting agents.

PCR amplification and sequencing of PCR products for blaTEM and blaSHV. Genomic DNA from KP3160 (isolated from case 15) and EC2859 (isolated from case 12) was digested with HindIII and ligated into HindIII digested pACYC184 (New England Biolabs) and transformed into E. coli DH5α and selected on trypticase soy agar plus 50 µg/mL of ampicillin. Clones were tested for the presence of the desired SHV enzyme by means of isoelectric focusing. The full-length SHV PCR product was generated with use of Pfu polymerase (Stratagene) and the primers SHV162F (5′-GCC TTT ATC GGC CCT CAC TCA-3′) and SHV1135B (5′-ATG CCG CCG CCA GTC ATA TC-3′). The purified PCR products were sequenced (2 PCR reactions per sample) with the primers SHVF, SHVB, SHV671F (5′-GCT GGT TTA TCG CCG ATA AG-3′), and SHV671B (5′-TCT TAT CGG CGA TAA ACC AGC-3′). Isolates EC1924 (from case 10), EC3022 (from case 15), and KP2679 (from case 17) had their TEM-containing plasmid isolated and transformed into E. coli DH5α or HB101. E-test ESBL strips were used to confirm that the resulting clones were ESBL producers. The TEM gene was amplified by PCR with use of Pfu polymerase and the primers BLAOT3 (5′-ATG AGT ATT CAA CAT TTC CG-3′) and BLAOT4 (5′-CCA ATG CTT AAT CAG TGA GG-3′) [7]; it was then placed into PCR blunt vector (Invitrogen). Plasmid DNA (2 clones for each enzyme) was sequenced with the primers BLAOT3, BLAOT4, BLA189F (5′-ACG TTT TCC AAT GAT GAG CAC T-3′), and BLA608B (5′-CGG GAA GCT AGA GTA AGT AGT-3′). All sequencing was performed by ACTG, Inc. (Northbrook, IL), and analyses were performed in-house by use of Vector NTI 5.2 (InforMax)

Assessment of outcomes. An episode of bloodstream infection was defined by the isolation of E. coli and K. pneumoniae strains that had a ceftazidime MIC of ⩾2 µg/mL from at least 1 blood culture and the presence of at least 1 of the following clinical signs or symptoms of infection: temperature of >38°C, chills, or hypotension (for patients aged ⩽12 months: temperature of >38°C, hypothermia [temperature <37°C], apnea, or bradycardia). Bloodstream infection was classified as primary or secondary, and acquisition was classified as nosocomial or community by use of a standard surveillance definition [8].

Clinical outcomes were classified as one of the following: complete response, partial response, failure, relapse, or unassessable. Complete response was defined as resolution of fever, leukocytosis, and local signs of infection. Partial response was defined as improvement of fever, leukocytosis, and local signs of infection without complete resolution. Treatment failure was defined as absence of resolution or worsening of signs and symptoms of infection. Relapse was defined as recurrence of infection with same organism at any body site in ⩽1 month after discontinuation of therapy. An outcome was defined as unassessable if medical records were incomplete.

Clinical outcome was assessed for all assessable episodes of bloodstream infection according to strain type, initial antimicrobial regimen, and type of treatment. A patient who had complete or partial response to treatment was considered a responder, whereas those who experienced a relapse or those whose infections failed to respond to treatment were considered nonresponders.

Statistical analysis. Treatment outcomes were analyzed according to strain type (E. coli or K. pneumoniae), antimicrobial regimen (including cephalosporin or not including cephalosporin) initially and at the end of treatment, and type of treatment (monotherapy or combination therapy). To compare treatment outcomes for each of the subgroups, χ2 or Fisher's exact test was used, as appropriate. P < .05 indicated statistical significance. All analyses were performed with GraphPad Prism software, version 3.00 for Windows (GraphPad Software).

Previous SectionNext Section

Results

Patients and bacterial strains. A total of 44 episodes of bacteremia for which tests were positive for E. coli or K. pneumoniae with ceftazidime MICs of ⩾2 µg/mL were identified from the computerized microbiology database at UCLA Medical Center for the years 1986–1997. Of those 44 episodes, 36 (in 34 patients) were treated with ⩾3 days of antimicrobial therapy and were included in the analysis of treatment outcome. Two patients each had 2 separate episodes of bloodstream infections due to different organisms. E. coli was more frequently isolated from study patients than was K. pneumoniae (23 vs. 13 patients, respectively). At least 1 blood isolate was saved for 21 episodes of bloodstream infection; 14 E. coli isolates and 7 K. pneumoniae isolates were available for phenotypic and molecular analysis.

Patient characteristics are shown in table 1. Bacteremia was primary in 14 (39%) of 36 episodes and secondary in 22 episodes (61%). For patients with secondary bacteremia, urine was the most frequent source of primary infection. Half of the episodes occurred in solid-organ transplant recipients. End-stage hepatic or renal disease and malignancy were the next most common underlying conditions. Most episodes (24 [67%] of 36) were classified as nosocomial. Approximately one-third of the patients had a recent history of surgery or ceftazidime exposure (⩽30 days before the first positive blood culture result). At the time that the positive blood culture result was obtained, most patients with episodes were hospitalized (in 24 [67%] of 36 episodes) in an acute-care unit. The crude mortality rate was 22% (8 of 36 episodes), and attributable mortality due to sepsis was 8% (3 of 36 episodes).

Figure 1
View larger version:
Figure 1

Response to the initial empirical regimen for patients with bacteremia due to ceftazidime-resistant Escherichia coli (EC) and Klebsiella pneumoniae (KP) by strain type. A, Regimens that included an extended-spectrum cephalosporin (ESC) compared with other regimens (n = 36 patients). B, Monotherapy compared with combination therapy (all regimens contained ⩾1 agent to which the strain was susceptible; n = 25 patients). Ceph, ESC-containing regimen; non-ceph, non–ESC-containing regimen; responders, patients who responded to treatment.

Figure 2
View larger version:
Figure 2

Response to the initial empirical regimen (empiric) versus response at the end of treatment (EOT), according to the isolates' extended-spectrum β-lactamase (ESBL) production status, for patients with bacteremia due to ceftazidime-resistant Escherichia coli (EC) and Klebsiella pneumoniae (KP). A, Regimens with ESC compared with other regimens (n = 21 cases). B, Monotherapy versus combination therapy (all regimens contained ⩾1 agent to which the strain was susceptible; n = 13 cases with initial response to empirical therapy, n = 19 cases with response at EOT. Ceph, extended-spectrum cephalosporin (ESC)–containing regimen; comb, combination therapy; mono, monotherapy; non-ceph, non–ESC-containing regimen; responders, patients who responded to treatment.

Table 1
View larger version:
Table 1

Demographic characteristics of 34 patients with bacteremia due to ceftazidime-resistant Escherichia coli (EC) and Klebsiella pneumoniae.

Phenotypic characterization: antimicrobial susceptibility testing. Table 2 summarizes the results of in vitro antimicrobial susceptibility tests of selected antimicrobial agents. None of the 21 saved isolates were fully susceptible to ticarcillin and piperacillin, except 1 strain that had a piperacillin MIC of 8 µg/mL; this strain produced an AmpC enzyme at a low level. Combination with tazobactam restored the activity of piperacillin (MIC, ⩽8 µg/mL) against 11 (79%) of 14 E. coli strains, regardless of whether the strains produced ESBL, but restored its activity against only 2 (29%) of 7 K. pneumoniae strains, each of which produced 2 β-lactamases (TEM-6 and SHV-1; table 3). On the other hand, combination with clavulanic acid restored the activity of ticarcillin against only 2 E. coli strains, neither of which produced ESBL. All isolates were susceptible to imipenem and meropenem at an MIC of ⩽0.5 µg/mL. Most isolates were susceptible to gentamicin (14 of 21 isolates) and to ciprofloxacin (17 of 21 isolates). All strains but 2 were resistant to trimethoprim-sulfamethoxazole (data not shown).

Table 2
View larger version:
Table 2

In vitro susceptibilities of Escherichia coli (EC) and Klebsiella pneumoniae (KP) blood isolates from episodes (cases) of bacteremia.

Table 3
View larger version:
Table 3

Profiles of putative extended-spectrum β-lactamase–producing Escherichia coli (EC) and Klebsiella pneumoniae (KP) isolates from epsiodes (cases) of bacteremia.

Phenotypic characterization: ESBL screening and confirmatory tests. Among the antimicrobial agents recommended by NCCLS for identifying strains that produce ESBL, ceftazidime, aztreonam, and cefpodoxime at an MIC of ⩾2 µg/mL had sensitivities of 100%, 90%, and 70% and positive predictive value of 52%, 55%, and 55%, respectively, for identifying ESBL-producing strains. Strains that have either TEM-12 or TEM-6 are less efficient in hydrolyzing ceftriaxone and cefotaxime [8]; for such strains, both agents have MICs in the range of 0.5–2 µg/mL. Cefoxitin MICs were generally higher for strains that produced AmpC enzyme (MIC50, 64 µg/mL) than they were for strains that produced ESBLs (MIC50, 16 µg/mL; tables 3 and 4). It is of note that cefoxitin susceptibility (MIC, ⩽8 µg/mL) was observed in only 5 (45%) of 11 ESBL-producing strains.

Table 4
View larger version:
Table 4

Profiles of AmpC-producing strains of Escherichia coli (EC) isolated from epsiodes (cases) of bacteremia.

Using either the original ESBL E-test strip or broth microdilution tests, we performed an ESBL confirmatory test of ceftazidime alone and in combination with clavulanic acid and found that it discriminated between ESBL-producing and non–ESBL-producing strains (sensitivity, 100%; positive predictive value, 78%; specificity, 70%; negative predictive value, 100%; tables 3 and 4). Broth microdilution tests with cefotaxime alone or in combination with clavulanic acid were less sensitive but more specific in confirming ESBL production (sensitivity, 91%; positive predictive value, 83%; specificity, 80%; negative predictive value, 89%). Two E. coli strains had a positive ESBL confirmatory test result for all 3 of these tests but did not produce an ESBL; both strains produced an atypical cephalosporinase that was significantly inhibited by clavulanic acid. For 1 K. pneumoniae strain that produced SHV-1, there were false positive ESBL confirmatory test results for the ESBL E-test and broth microdilution test of clavulanic acid alone and in combination with ceftazidime, but not clavulanic acid in combination with cefotaxime. On the other hand, for 1 strain of E. coli that produced TEM-12 (isolated from patient 10), there was a false negative results for the ESBL confirmatory broth microdilution test of cefotaxime alone and in combination with clavulanic acid.

The clinical relevance of performance of ESBL confirmatory test for isolates with elevated MICs for the ESCs was demonstrated by cases 8 and 12 (tables 2 and 5). These 2 cases of bloodstream infection were caused by an E. coli isolate that had cefotaxime and ceftriaxone MICs of 1–4 µg/mL and ceftazidime MICs of 8 and 16 µg/mL. Both patients were treated with ceftizoxime; the patient with case 8 had a complete response (ceftizoxime MIC, 4 µg/mL), but the patient with case 12 died as a result of urosepsis after 6 days of treatment (ceftizoxime MIC, 2 µg/mL). The first case was associated with a non–ESBL-producing strain that produced both a TEM-1 and AmpC β-lactamase, which indicates that a combination of non-ESBL enzymes can result in an ESBL screening phenotype that has negative ESBL confirmatory test results. The second case was caused by a strain that produced the enzymes SHV-12 and TEM-1; the ESBL phenotype was confirmed by both the ESBL E-test and the broth microdilution test of ceftazidime alone and in combination with clavulanic acid. In this case, the failure to detect ESBL-mediated ESC resistance and modify therapy from ceftizoxime likely contributed to the fatal outcome. The difference in responses to treatment with an ESC in these 2 cases may be attributable to ESBL production in one strain but not in the other. Undoubtedly, if there had been a positive ESBL confirmatory test result at the time the bacterial strain was isolated and its identification and antibiotic suscetibility profile were reported to the clinician, as well as a lack of clinical response, these findings would have guided the clinician to promptly change therapy and, possibly, would have prevented the patient's death from sepsis.

Table 5
View larger version:
Table 5

Antimicrobial regimens and outcomes for patients with episodes (cases) of bacteremia due to ceftazidime-resistant strains of Escherichia coli (EC) and Klebsiella pneumoniae (KP).

Molecular analysis. Tables 3 and 4 show the isoelectric points for each isolate and the β-lactamases identified for selected isolates. The isolates were grouped into 11 types (designated A–K) on the basis of the isoelectric points of the β-lactamases detected. The isolates were further grouped into AmpC-producing strains (n = 9) and putative ESBL-producing strains (n = 12) on the basis of the results of β-lactamase extraction and isoelectric focusing studies. PCR amplification and sequencing of the PCR products for blaTEM and blaSHV were performed for each of the non-AmpC enzymes that had a unique isoelectric point.

A majority (9 [64%] of 14) of the E. coli strains (types A, B, C, and D) produced an AmpC enzyme, compared with 0 of the 7 K. pneumoniae strains. Among the AmpC-producing E. coli strains, type B strains were unique because the β-lactamase had a high isoelectric point (9.0), a finding consistent with the presence of AmpC, but the enzyme was inhibited by clavulanic acid and tazobactam, whereas type D strains produced concomitant TEM-1 enzyme. In addition, OMP characterization was performed for 3 E. coli isolates. The cefoxitin-susceptible isolate from patient 5 showed 3 proteins in the 31,000–42,700-kDa range. The cefoxitin-resistant isolates from cases 2 and 8 exhibited only 2 OMP bands and showed the loss of a 38-kDa protein. In addition, cefoxitin resistance could not be transferred to the E. coli C600RifR/NAR recipient by conjugation with either of the 2 strains examined. These data suggest that decreased permeability contributed to cefoxitin resistance in these strains.

On the basis of PCR amplification and sequencing of the PCR products for blaTEM and blaSHV, 5 of 14 E. coli strains and 6 of 7 K. pneumoniae strains produced ESBLs, as follows: TEM-12 (n = 1), SHV-12 (n = 4), TEM-71 (n = 1), TEM-6 (n = 2), and SHV-5 (n = 3). Nine of 11 strains produced 2 β-lactamases. No strain produced multiple ESBLs. For 1 K. pneumoniae strain (isolated from patient 16), the ceftazidime MIC was elevated (2 µg/mL), the ESBL E-test result was positive, and the strain produced SHV-1 enzyme. One K. pneumoniae strain (isolated from patient 14) had an isoelectric point profile identical to that of E. coli type F isolates, which suggests that the gene encoding for this ESBL (SHV-12) may have been transferred between E. coli and K. pneumoniae strains.

Treatment. Initial treatment regimens for the 36 clinically assessable episodes of bloodstream infection included administration of an ESC agent (22 episodes [61%]), most commonly ceftazidime (7 episodes [31%]) or ceftizoxime (7 episodes [19%]). Therapy with an agent other than a cephalosporin was primarily with ampicillin-sulbactam (1 episode [3%]), ciprofloxacin (5 episodes [14%]), and imipenem (4 episodes [11%]) alone or in combination with an aminoglycoside. Initial treatment regimens included a single agent or combination of agents (18 episodes each). Most patients received multiple courses of antimicrobial therapy for the episode of bloodstream infection, for a total of 49 courses of treatment (range, 1–4 regimens per episode). Table 5 summarizes the treatment regimens, duration, and associated outcomes for 21 episodes from which an isolate was available for phenotypic and molecular analysis.

Outcomes. Overall, patients responded to treatment in 20 (56%) of 36 assessable episodes. Patients with bloodstream infections caused by E. coli strains with elevated ceftazidime MICs were almost twice as likely to respond to treatment as patients with infections caused by K. pneumoniae, although the difference was not statistically significant (15 [65%] of 23 vs. 5 [38%] of 13 episodes; P = .12). A trend toward better treatment response was observed for noncephalosporin regimens compared with regimens containing an ESC, both for E. coli infections (6 [86%] of 7 vs. 9 [56%] of 16 episodes; P = .34) and K. pneumoniae infections (3 [50%] of 6 vs. 2 [28%] of 7 episodes; P = .59; figure 1). When treatment response was analyzed according to regimens that contained at least 1 agent to which the strain was susceptible in vitro, monotherapy and combination therapy were equally effective for the treatment of E. coli bacteremia (4 [44%] of 9 vs. 4 [57%] of 7 episodes; P = 1.0). In contrast, none of the episodes of K. pneumoniae bacteremia responded to monotherapy, whereas 4 (57%) of 7 episodes responded to combination therapy. However, the small number of patients in each subgroup precluded meaningful statistical comparison.

Clinical and microbiologic correlation. The relationship between β-lactamase type and treatment outcomes, particularly with the ESCs, was examined in detail for the 21 episodes of bloodstream infection for which isolates were saved (table 5). Overall, 35 courses of antimicrobial therapy were prescribed for these 21 episodes of bloodstream infection; 19 courses for 11 ESBL-associated episodes and 16 courses for non–ESBL-associated episodes. The clinical response to the initial treatment regimen was significantly lower for episodes of bloodstream infection caused by ESBL-producing strains than for episodes of infection with non–ESBL-producing strains (2 [18%] of 11 vs. 8 [80%] of 10 episodes; P = .009). However, when clinical response was determined at the end of treatment, a similar result was observed (ESBL vs. non–ESBL-associated infection: 7 [64%] of 11 vs. 9 [90%] of 10 episodes; P = .31). Among the 11 episodes associated with ESBL-producing strains, the initial and end-of-treatment response rates did not differ according to whether the strains had TEM or SHV-type enzymes (data not shown).

When clinical response was assessed by treatment type, we observed that the response to empiric therapy was similar for ESC-containing courses of antimicrobials and for non–ESC-based courses (5 [38%] of 13 vs. 5 [63%] of 8 episodes; P = .39; figure 2); a similar response was noted at the end of treatment (7 [70%] of 10 vs. 9 [82%] of 11 episodes; P = .64). Among the 16 ESC-based courses, the response rate for nonceftazidime ESCs was more than double that for ceftazidime (7 [70%] of 10 vs. 2 [33%] of 6 episodes; P = .3), although this difference did not reach statistical significance. Partial or complete response to nonceftazidime ESCs was observed for 2 patients infected with strains that produced TEM-6 or TEM-12 (1 of each). Two patients died of sepsis as a result of failure to respond to treatment with ceftazidime therapy alone or in combination with ciprofloxacin. Both of these patients had bloodstream infections with K. pneumoniae strains that harbored either a TEM-6 or SHV-5 ESBL. In addition, imipenem was prescribed for 4 courses of therapy, 2 of which resulted in relapse. One case of relapse was caused by infection with a strain of E. coli that produced the TEM-71 enzyme; K. pneumoniae that produced both SHV-1 and SHV-5 enzymes was associated with the other case of relapse. When treatment response was analyzed according to regimens that contain at least 1 agent to which the strain is susceptible in vitro, response rates for monotherapy and combination empiric therapy were similar (3 [38%] of 8 episodes, vs. 2 [40%] of 5 episodes; P = 1.0). Response rates noted at the end of treatment for the same comparison (10 [83%] of 12 vs. 4 [57%] of 7 episodes; P = .3) favor monotherapy, but the difference in rates did not reach statistical significance (figure 2). Of the 5 non–ESC-containing treatment courses that resulted in a positive response, 4 (80%) were combination therapy.

Previous SectionNext Section

Discussion

We identified 36 episodes of bloodstream infection that occurred during a 12-year period among patients at a large tertiary-referral medical center and were associated with strains of E. coli and K. pneumoniae that had reduced susceptibility to ceftazidime. Infections occurred sporadically throughout the study period and included both nosocomial and community-acquired cases of bacteremia. Blood isolates from 21 of the 36 episodes were available for phenotypic and molecular analysis, which facilitated the correlation of phenotypic and genotypic characteristics of the strains with treatment outcome for most of the episodes. Previous studies have focused on the epidemiology, laboratory detection methods, and molecular characterization of ESBL-producing organisms. Data on treatment outcomes are limited to case reports or descriptions of nosocomial outbreaks [6, 9–25]. Thus, sufficient clinical data to establish the relationship between the type of ESBLs produced by the infecting strains and treatment outcomes is lacking. To our knowledge, this is the largest reported series (not associated with an outbreak investigation) to examine treatment of bloodstream infections caused by strains of E. coli and K. pneumoniae with reduced susceptibility to ceftazidime.

On the basis of in vitro susceptibility results and limited reports of clinical experience, imipenem has emerged as the agent of choice for the treatment of serious infections associated with ceftazidime-resistant strains of E. coli and K. pneumoniae. This has occurred despite infrequent reports of the failure of ESCs treatment of infections caused by strains that appear to be susceptible in vitro [6]. Lacking sufficient outcome data on other agents, many investigators recommend the use of imipenem as the treatment of choice for serious infections due to ESBL-producing isolates [12, 23]. Several investigators have concluded that initial treatment of bloodstream infections caused by ESBL-producing strains with noncarbapenem agents may be associated with higher mortality than is treatment with a carbapenem agent; however, details on treatment information for most patients were not available in those studies [26, 27]. Currently, the NCCLS recommends that ESBL-producing strains should be reported by microbiology laboratories as resistant to all penicillins, cephalosporins, and aztreonam [2]. These criteria and other expert recommendations may promote the exclusive use of imipenem for the treatment of serious infections caused by ceftazidime-resistant E. coli and K. pneumoniae strains, limiting clinical experience with the use of other agents, particularly the ESCs. Increased use of imipenem also may promote antimicrobial resistance. Rahal and associates [28] found that reduction in the use of ESC resulted in a dramatic decrease in the incidence of ESBL producing–strains, but reduction in use was also associated with a 140% increase in the use of imipenem and a 70% increase in the incidence of infection with imipenem-resistant strains of Pseudomonas aeruginosa. Therefore, therapeutic options other than carbapenems would be an attractive alternative for ceftazidime-resistant ESBL-producing organisms.

In this study, we retrospectively assessed treatment outcomes and examined clinical response after both initial empirical therapy and at end of treatment. Of 36 episodes, more than one-half were treated with an ESC, and only 4 courses of imipenem were administered. In addition, although other studies have reported treatment experience with outbreak strains of K. pneumoniae that produced only 1 or 2 different types of ESBLs, we evaluated treatment outcomes for bloodstream infections associated with 5 different types of ESBL-producing E. coli and K. pneumoniae strains in a situation in which these strains were endemic.

Our study confirms that the current NCCLS-recommended ESBL screening and confirmatory tests (which use ceftazidime or cefotaxime, each alone and in combination with clavulanic acid) for E. coli and K. pneumoniae are highly sensitive in detecting a variety of ESBL phenotypes. Although strains producing atypical AmpC or SHV-1, or a combination of a TEM-1 with an AmpC enzyme, yielded false positive screening results for ESBL production, the ESBL confirmatory tests were associated with high positive predictive value (78%–83%). It is of note that, because cefotaxime is less susceptible to hydrolysis by most North American ESBLs than is ceftazidime, the cefotaxime-based ESBL confirmatory test is likely to be less sensitive than are ceftazidime-based tests. Promptly informing clinicians of an E. coli or K. pneumoniae strain that is suspected to produce ESBL may assist in directing or altering antimicrobial therapy.

We found that the overall treatment response for bloodstream infections due to ceftazidime-resistant E. coli strains was higher than for K. pneumoniae strains. The difference in treatment response by strain could be explained largely by the variation in the mechanisms of resistance observed among study organisms. β-Lactamase extraction and OMP characterization of available isolates revealed that the underlying mechanism of ceftazidime resistance in the majority of E. coli isolates was the activity of an AmpC enzyme with or without a concomitant TEM-1 enzyme and loss of OMPs. On the other hand, 86% of the K. pneumoniae strains were found to produce single or multiple TEM- or SHV-type ESBLs. The difference in responses may be attributable in part to the enhanced ability of TEM and SHV ESBLs to hydrolyze ESCs, especially ceftazidime, compared with the ability of AmpC enzymes [29]. Cases 5, 8, and 16 were due to AmpC-producing organisms, which had ceftazidime MICs of 16, 8, and 2 µg/mL, respectively, and all had complete response to nonceftazidime ESC monotherapy. In contrast, cases 12, 18, and 21 were due to organisms that had 1 of the following combinations of enzymes: SHV-12 and TEM-1, TEM-6 and SHV-1, or SHV-5 and SHV-1. All 3 patients died as a result of sepsis after treatment with an ESC-based regimen. Of note, we did not study porin changes for the K. pneumoniae isolates. Thus it is possible that a concomitant loss of OMPs may have contributed to the resistance profile as well as the lower response to treatment.

When outcomes of episodes of bacteremia caused by ESBL-producing strains were analyzed with respect to treatment regimen, no significant difference was found between regimens containing an ESC and other regimens. It is noteworthy that positive response to treatment with an ESC-containing regimen was observed for bacteremia caused by pathogens that produced either TEM-6 or TEM-12. Response to treatment with a nonceftazidime-ESC may be explained by the relatively less efficient ceftazidime-hydrolyzing capability of TEM-6 and TEM-12 ESBLs compared with TEM-71. Ceftazidime treatment was associated with treatment failure in all patients, as expected, because the isolates involved were all resistant to this agent (MIC ⩾32 µg/mL). In contrast to other reports, only a few of our patients received imipenem therapy; treatment success was not observed in all patients despite an MIC of ⩽0.25 µg/mL.

We recognize that we could not evaluate the impact of host and other nondrug factors on the overall clinical response because of the retrospective design of our study. Nonetheless, several important insights can be gained. First, an NCCLS confirmatory test or equivalent should be performed for all E. coli and K. pneumoniae isolates that test positive for an ESBL phenotype, and the results of these confirmatory tests should be used to guide decisions about therapy in the context of clinical response. Second, we found that ceftazidime-resistant E. coli strains produced a greater variety of β-lactamase types than did K. pneumoniae strains. Finally, we observed treatment success with the nonceftazidime ESCs for bloodstream infections caused by pathogens that produce ESBLs, which are known to be relatively weak hydrolyzers of the ESCs. Other investigators have also shown that serious infections, including meningitis, caused by ESBL-producing isolates can be successfully treated with an ESC [9, 10].

On the basis of the results of our study, we strongly recommend that the use of ceftazidime for empirical therapy should be avoided whenever possible, both because previous drug exposure may promote acquisition of ESBL-producing organisms and because ceftazidime is more susceptible to hydrolysis by most North American ESBLs than are ceftriaxone or cefotaxime. Our findings support the current NCCLS recommendation that ESBL-producing strains should be reported by microbiology laboratories as resistant to all penicillins, cephalosporins, and aztreonam. [2]. Despite the fact that a minority of strains that produced an atypical cephalosporinase were falsely confirmed to produce ESBL and that limited cases of clinical success were observed for the treatment of bloodstream infections with an ESC in the patients we reviewed, we currently recommend that the use of ESCs be avoided for the treatment of serious infections caused by a strain with a confirmed ESBL phenotype.

Previous SectionNext Section

Acknowledgments

We thank Lisa Licata for performing initial isoelectric focusing studies.

Previous SectionNext Section

Footnotes

  • Presented in part: 37th Interscience Conference on Antimicrobial Agents and Chemotherapy, Toronto, Canada, September 1997 (abstract J-8); 98th American Society of Microbiology General Meeting, Atlanta, GA, May 1998 (abstract C-24); 39th Interscience Conference on Antimicrobial Agents and Chemotherapy, San Francisco, CA, September 1999 (abstract 1478).

  • Received March 23, 2001.
  • Revision received July 24, 2001.
Previous Section
 

References

  1. ↵
    1. Livermore DM
    . β-Lactamases in laboratory and clinical resistance. Clin Microbiol Rev 1995;8:557-84.
    Abstract/FREE Full Text
  2. ↵
    1. National Committee for Clinical Laboratory Standards (NCCLS)
    . Approved standard M7-A5. Wayne, PA: NCCLS; 2000. Methods for dilution antimicrobial susceptibility tests for bacteria that grow aerobically.
  3. ↵
    1. Centers for Disease Control and Prevention
    . Laboratory capacity to detect antimicrobial resistance, 1998. MMWR Morb Mortal Wkly Rep 2000;48:1167-71.
    Medline
  4. ↵
    1. Bush K
    . Is it important to identify extended-spectrum β-lactamase-producing isolates? Eur J Clin Microbiol Infect Dis 1996;15:361-4.
    CrossRefMedlineWeb of Science
  5. ↵
    1. Jacoby GA,
    2. Sutton L
    . Properties of plasmids responsible for production of extended-spectrum βlactamases. Antimicrob Agents Chemother 1991;35:164-9.
    Abstract/FREE Full Text
  6. ↵
    1. Karas JA,
    2. Pillay DG,
    3. Muckart D,
    4. et al
    . Treatment failure due to extended spectrum β-lactamase. J Antimicrob Chemother 1996;37:203-4.
    FREE Full Text
  7. ↵
    1. Arlet G,
    2. Brami G,
    3. Decr'e D,
    4. et al
    . Molecular characterization by PCR-restriction length polymorphism of TEM β-lactamases. FEMS Microbiol Lett 1995;134:203-8.
    MedlineWeb of Science
  8. ↵
    1. Garner JS,
    2. Jarvis WR,
    3. Emori TG,
    4. et al
    . CDC definitions for nosocomial infections, 1988. Am J Infect Control 1988;16:128-40.
    CrossRefMedlineWeb of Science
  9. ↵
    1. Smith CE,
    2. Tillman S,
    3. Howell AW,
    4. et al
    . Failure of ceftazidime-amikacin therapy for bacteremia and meningitis due to Klebsiella pneumoniae producing an extended-spectrum β-lactamase. Antimicrob Agents Chemother 1990;34:1290-3.
    Abstract/FREE Full Text
  10. ↵
    1. Quinn JP,
    2. Miyashiro D,
    3. Sahm D,
    4. et al
    . Novel plasmid-mediated β-lactamase (TEM-10) conferring selective resistance to ceftazidime and aztreonam in clinical isolates of Klebsiella pneumoniae. Antimicrob Agents Chemother 1989;33:1451-6.
    Abstract/FREE Full Text
    1. de Champs C,
    2. Guelon D,
    3. Joyon D,
    4. et al
    . Treatment of a meningitis due to an Enterobacter aerogenes producing a derepressed cephalosporinase and a Klebsiella pneumoniae producing an extended-spectrum β-lactamase. Infection 1991;19:181-3.
    CrossRefMedlineWeb of Science
  12. ↵
    1. Paterson DL,
    2. Singh N,
    3. Gayowski T,
    4. et al
    . Fatal infection due to extended-spectrum β-lactamase-producing Escherichia coli: implications for antibiotic choice for spontaneous bacterial peritonitis. Clin Infect Dis 1999;28:683-4.
    FREE Full Text
    1. Spencer RC,
    2. Wheat PF,
    3. Winstanley TG,
    4. et al
    . Novel β-lactamase in a clinical isolate of Klebsiella pneumoniae conferring unusual resistance to β-lactam antibiotics. J Antimicrob Chemother 1987;20:919-27.
    FREE Full Text
    1. Segal-Maurer S,
    2. Mariano N,
    3. Qavi A,
    4. et al
    . Successful treatment of ceftazidime-resistant Klebsiella pneumoniae ventriculitis with intravenous meropenem and intraventricular polymixin B: case report and review. Clin Infect Dis 1999;28:1134-8.
    Abstract/FREE Full Text
    1. Naumovski L,
    2. Quinn JP,
    3. Miyashiro D,
    4. et al
    . Outbreak of ceftazidime resistance due to a novel extended-spectrum β-lactamase in isolates from cancer patients. Antimicrob Agents Chemother 1992;36:1991-6.
    Abstract/FREE Full Text
    1. Brun-Buisson C,
    2. Legrand P,
    3. Philippon A,
    4. et al
    . Transferable enzymatic resistance to third-generation cephalosporins during nosocomial outbreak of multiresistant Klebsiella pneumoniae. Lancet 1987;2:302-6.
    CrossRefMedline
    1. Prodinger WM,
    2. Fille M,
    3. Bauernfeind A,
    4. et al
    . Molecular epidemiology of Klebsiella pneumoniae producing SHV-5 β-lactamase: parallel outbreaks due to multiple plasmid transfer. J Clin Microbiol 1996;34:564-8.
    Abstract/FREE Full Text
    1. French GL,
    2. Shannon KP,
    3. Simmons N
    . Hospital outbreak of Klebsiella pneumoniae resistant to broad-spectrum cephalosporins and β-lactam-β-lactamase inhibitor combinations by hyperproduction of SHV-5 β-lactamase. J Clin Microbiol 1996;34:358-63.
    Abstract/FREE Full Text
    1. Bingen EH,
    2. Desjardins P,
    3. Arlet G,
    4. et al
    . Molecular epidemiology of plasmid spread among extended broad-spectrum β-lactamase-producing Klebsiella pneumoniae isolates in a pediatric hospital. J Clin Microbiol 1993;31:179-84.
    Abstract/FREE Full Text
    1. Pena C,
    2. Pujol M,
    3. Ardanuy C,
    4. et al
    . Epidemiology and successful control of a large outbreak due to Klebsiella pneumoniae producing extended-spectrum β-lactamases. Antimicrob Agents Chemother 1998;42:53-8.
    Abstract/FREE Full Text
    1. Rice LB,
    2. Willey SH,
    3. Papanicolaou GA,
    4. et al
    . Outbreak of ceftazidime resistance caused by extended-spectrum spectrum β-lactamases at a Massachusetts chronic-care facility. Antimicrob Agents Chemother 1990;34:2193-99.
    Abstract/FREE Full Text
    1. Royle J,
    2. Halasz S,
    3. Eagles G,
    4. et al
    . Outbreak of extended spectrum beta lactamase producing Klebsiella pneumoniae in a neonatal unit. Arch Dis Child Fetal Neonatal Ed 1999;80:64-8.
    CrossRef
  23. ↵
    1. Meyer KS,
    2. Urban C,
    3. Eagan JA,
    4. et al
    . Nosocomial outbreak of Klebsiella infection resistant to third-generation cephalosporins. Ann Intern Med 1993;119:353.
    Abstract/FREE Full Text
    1. Emery CL,
    2. Weymouth LA
    . Detection and clinical significance of extended-spectrum β-lactamases in a tertiary-care medical center. J Clin Microbiol 1997;35:2061-7.
    Abstract/FREE Full Text
  25. ↵
    1. Siu LK,
    2. Lu PL,
    3. Hsueh PR,
    4. et al
    . Bacteremia due to extended-spectrum β-lactamase-producing Escherichia coli and Klebsiella pneumoniae in a pediatric oncology ward: clinical features and identification of different plasmids carrying both SHV-5 and TEM-1 genes. J Clin Microbiol 1999;37:4020-7.
    Abstract/FREE Full Text
  26. ↵
    1. Schiappa DA,
    2. Hyden MK,
    3. Matushek MG,
    4. et al
    . Ceftazidime-resistant Klebsiella pneumoniae and Escherichia coli bloodstream infection: a case-control and molecular epidemiologic investigation. J Infect Dis 1996;174:529-36.
    Abstract/FREE Full Text
  27. ↵
    1. Paterson DL,
    2. Ko W,
    3. von Gottberg A,
    4. et al
    . Outcome of cephalosporin treatment for serious infections due to apparently susceptibe organisms producing extended-spectrum β-lactamases: implications for clinical microbiology. J Clin Microbiol 2001;39:2206-12.
    Abstract/FREE Full Text
  28. ↵
    1. Rahal JJ,
    2. Urban C,
    3. Horn D,
    4. et al
    . Class restriction of cephalosporin use to control total cephalosporin resistance in nosocomial Klebsiella. JAMA 1998;280:1233-7.
    Abstract/FREE Full Text
  29. ↵
    1. Bush K,
    2. Jacoby GA,
    3. Medieros AA
    . A functional classification scheme for β-lactamases and its correlation with molecular structure. Antimicrob Agents Chemother 1995;39:1211-33.
    FREE Full Text
| Table of Contents

This Article

  1. Clin Infect Dis. (2002) 34 (2): 135-146. doi: 10.1086/324742
  1. AbstractFree
  2. » Full Text (HTML)Free
  3. Full Text (PDF)Free

Classifications

Share

  1. Email this article

Navigate This Article

Current Issue

  1. March 1, 2012 54 (5)
  1. Clinical Infectious Diseases
  1. Alert me to new issues

Most