Carbapenem Therapy for Bacteremia Due to Extended-Spectrum-β-Lactamase-Producing Escherichia coli or Klebsiella pneumoniae: Implications of Ertapenem Susceptibility

Abstract

A retrospective study was conducted at two medical centers in Taiwan to evaluate the clinical characteristics, outcomes, and risk factors for mortality among patients treated with a carbapenem for bacteremia caused by extended-spectrum-beta-lactamase (ESBL)-producing organisms. A total of 251 patients with bacteremia caused by ESBL-producing Escherichia coli and Klebsiella pneumoniae isolates treated by a carbapenem were identified. Among these ESBL-producing isolates, rates of susceptibility to ertapenem (MICs ≤ 0.25 μg/ml) were 83.8% and 76.4%, respectively; those to meropenem were 100% and 99.3%, respectively; and those to imipenem were 100% and 97.9%, respectively. There were no significant differences in the critical illness rate (P = 0.1) or sepsis-related mortality rate (P = 0.2) for patients with bacteremia caused by ESBL-producing K. pneumoniae (140 isolates, 55.8%) and E. coli (111 isolates, 44.2%). Multivariate analysis of variables related to sepsis-related mortality revealed that the presence of severe sepsis (odds ratio [OR], 15.9; 95% confidence interval [CI], 5.84 to 43.34; P < 0.001), hospital-onset bacteremia (OR, 4.65; 95% CI, 1.42 to 15.24; P = 0.01), and ertapenem-nonsusceptible isolates (OR, 5.12; 95% CI, 2.04 to 12.88; P = 0.001) were independent risk factors. The patients receiving inappropriate therapy had a higher sepsis-related mortality than those with appropriate therapy (P = 0.002), irrespective of ertapenem, imipenem, or meropenem therapy. Infections due to the ertapenem-susceptible isolates (MICs ≤ 0.25 μg/ml) were associated with a more favorable outcome than those due to ertapenem-nonsusceptible isolates (MICs > 0.25 μg/ml), if treated by a carbapenem. However, the mortality for patients with bacteremic episodes due to isolates with MICs of ≤0.5 μg/ml was similar to the mortality for those whose isolates had MICs of >0.5 μg/ml (P = 0.8). Such a finding supports the rationale of the current CLSI 2011 criteria for carbapenems for Enterobacteriaceae.

INTRODUCTION

Production of β-lactamase is the most frequent β-lactam resistance mechanism in Gram-negative organisms. In the varied and complex world of β-lactamases, extended-spectrum β-lactamases (ESBLs) have played a leading role in the clinical field in recent decades (1, 15, 17, 19, 26). Infections caused by multidrug-resistant bacteria expressing ESBLs pose serious challenges to clinicians, since ESBL-producing bacteria are resistant to a broad range of β-lactams, including expanded-spectrum cephalosporins. Nosocomial infections caused by these organisms complicate therapy and limit treatment options (13, 19). The presence of ESBLs in various members of the Enterobacteriaceae family, particularly Klebsiella pneumoniae and Escherichia coli, is of great microbiological and clinical importance (14). It has been found that bloodstream infections by ESBL-producing Enterobacteriaceae isolates were associated with a delay in the institution of appropriate antimicrobial therapy compared with the time to appropriate therapy for infections by nonproducing isolates (21).

Carbapenems have been considered the treatment of choice against ESBL-producer infections (13, 18). This is mainly because they are not inactivated by these enzymes in vitro and have been demonstrated to have adequate effectiveness for the treatment of serious Gram-negative bacterial infections at various body sites (8, 18). The efficacy of carbapenem therapy was best illustrated in an observational study of bacteremia due to ESBL-producing K. pneumoniae (16), and carbapenems are considered the preferred agents for treatment of serious infections caused by ESBL-producing Enterobacteriaceae (13, 14, 18, 26). Carbapenem resistance is currently rare among Enterobacteriaceae, but some worrisome signs have appeared in recent years (14). There was increasing concern over the emergence of Gram-negative bacillary isolates with carbapenemases conferring resistance to carbapenems and β-lactam agents (14).

Because of the concerns over the carbapenemases, the Clinical and Laboratory Standards Institute (CLSI) revised the interpretive criteria of carbapenems for Enterobacteriaceae in document M100-S21 after their evaluation of pharmacokinetic-pharmacodynamic properties, MIC distributions, and limited clinical data (4). However, the clinical impact of such a revision in areas with a low prevalence of carbapenemases remained undefined. Therefore, the aim of this study was to assess the carbapenem susceptibility of blood isolates of ESBL-producing E. coli or K. pneumoniae and to evaluate the clinical outcome in adults with bacteremia due to ESBL producers with different carbapenem MICs when they were treated by a carbapenem.

MATERIALS AND METHODS

Study design and patients.

A retrospective study was undertaken among adults (ages, ≥18 years) with ESBL-producing E. coli and K. pneumoniae bacteremia at two main teaching hospitals, the National Cheng Kung University Hospital in southern Taiwan and the National Taiwan University Hospital in northern Taiwan, between May 2002 and August 2007. Carbapenem therapy was defined as the intravenous administration of a carbapenem for at least 48 h until the end of antimicrobial therapy or death. Empirical therapy was defined as the drug given within 5 days after bacteremia onset. Antimicrobial therapy was considered appropriate if the causative isolate was susceptible in vitro to the prescribed drug according to the current breakpoint criteria of CLSI (4). If the patients experienced more than one episode of bacteremia caused by an ESBL producer, only the first episode in each patient was included. The clinical choice of ertapenem (1 g every 24 h), imipenem (0.5 g every 6 h or 1 g every 8 h), or meropenem (1 g every 8 h) was at the discretion of the attending physicians, and their dosage could be adjusted in cases of renal insufficiency (4). In both hospitals, therapeutic use of carbapenems must be approved by infectious disease specialists and pharmacists for the indication and dosage.

In vitro susceptibility tests and ESBL detection.

In the clinical microbiology laboratory at both hospitals, ESBL production was detected by the phenotypic confirmatory test with the cephalosporin-clavulanate combination disks recommended by the CLSI (5). The MICs of each isolate for carbapenems were determined by the agar dilution method.

ESBL-producing E. coli and K. pneumoniae isolates were categorized into two groups, the carbapenem-susceptible and carbapenem-nonsusceptible groups. The former consisted of isolates with ertapenem MICs of ≤0.25 μg/ml or meropenem or imipenem MICs of ≤1 μg/ml, according to the M100-S21 document (4). The other isolates were the nonsusceptible group. The carbapenem susceptibility rates were compared with those defined by the MIC interpretive criteria recommended by the European Committee on Antimicrobial Susceptibility Testing (EUCAST) (7), in which the isolates with an ertapenem MIC of ≤0.5 μg/ml or an imipenem or meropenem MIC of ≤2 μg/ml were regarded as being susceptible to ertapenem or to imipenem or meropenem, respectively.

Clinical evaluation and outcomes.

Clinical information, including demographic information, underlying illness, serum laboratory and microbiological data, portals of entry, infectious foci, complications of bacteremia, antimicrobial therapy, and clinical outcome, was retrieved from medical charts and collected in a case record form. Patient outcomes were evaluated. In the present study, the primary outcome was sepsis-related mortality, which was the death of a patient with a clinical course suggestive of persistently active infection without an obvious explanation. The durations of hospital stay after bacteremia onset were evaluated only among the survivors.

Community-onset infections were defined as those for which the first positive blood sample for culture was collected within 48 h after admission, and the others were classified as hospital-onset infections. Immunosuppression was defined as the receipt of a corticosteroid (10 mg or an equivalent dosage daily) for more than 2 weeks or of antineoplastic chemotherapy or antirejection medication within 4 weeks before the onset of bacteremia. The severity of underlying medical illness was stratified according to the McCabe score as fatal, ultimately fatal, or nonfatal (12). The severity of bacteremia was graded on the day of bacteremia onset by the Pittsburgh bacteremia score (3).

Statistical analysis.

The data were analyzed by SPSS software for Windows, version 12.0. Continuous variables were expressed as mean values ± standard deviations (SDs) and compared by the Mann-Whitney U test or Student t test. Categorical variables were expressed as percentages of total numbers of patients analyzed and compared by the Fisher exact test or χ² test, as appropriate. Variables with a P value of 0.05 or less by the univariate analysis were included in a multiple conditional logistic regression analysis. A P value of less than 0.05 was considered statistically significant, and all tests were two-tailed.

In view of the differences in the baseline characteristics between the ertapenem-nonsusceptible and -susceptible groups, a propensity scoring matching analysis was used to adjust for risk factors of sepsis-related mortality. Each patient in the ertapenem-nonsusceptible group was matched to one in the ertapenem-susceptible group (1:1 match) by the closest propensity score. A maximal difference of 5% in the likelihood of sepsis-related mortality was allowed in the matching process. If there was more than one match with an identical propensity score, those with pneumonia (the first secondary matching variable) and similar McCabe scores (the backup secondary matching variable) would have a higher priority in the matching process.

RESULTS

During the study period, there were a total of 328 adult cases of ESBL-producing E. coli or K. pneumoniae bacteremia. Of them, 125 (49.8%) cases empirically and definitively received a carbapenem, 125 (49.8%) empirically received a cephalosporin and definitively received a carbapenem, and 1 empirically received amoxicillin-clavulanate (inactive in vitro against the bacteremic ESBL-producing isolate) and definitively received a carbapenem. Among 126 patients with empirical noncarbapenem therapy, the time lag in the initiation of carbapenem therapy was longer than that for patients with empirical carbapenem therapy (4.1 ± 1.4 versus 1.1 ± 1.0 days, P < 0.001). Therefore, a total of 251 (76.5%) cases treated by ertapenem, meropenem, or imipenem were included and analyzed for their clinical outcome during hospitalization (Fig. 1).

Fig 1.

Stratification of 251 patients with bacteremia due to ESBL-producing Escherichia coli and Klebsiella pneumoniae by carbapenem therapy and susceptibility.

E. coli accounted for 44.2% (n = 111) of 251 cases with ESBL-producer bacteremia, and K. pneumoniae accounted for 55.8% (n = 140). The MIC₉₀ of ertapenem for ESBL-producing E. coli and K. pneumoniae (0.5 μg/ml) was higher than that of meropenem (0.06 μg/ml), as shown in Table 1. With the modification of MIC susceptibility breakpoints for carbapenems, as recommended by the CLSI and EUCAST, the susceptibility rates among ESBL-producing E. coli and K. pneumoniae isolates varied. For ertapenem, susceptibility rates decreased from 100% (CLSI document M100-S20 method) to 83.8% (CLSI document M100-S21 method) and 91.9% (EUCAST 2011 method) in E. coli isolates and from 98.6% to 76.4% (M100-S21 method) and 90.7% (EUCAST 2011 method) in K. pneumoniae isolates. In contrast, susceptibility rates for meropenem or imipenem remained 100% in E. coli isolates, irrespective of the left shift of susceptibility breakpoints, and decreased to a lesser degree in K. pneumoniae isolates (meropenem, from 100% to 99.3% [M100-S21 method]; imipenem, from 100% to 97.9% [M100-S21 method] and 99.3% [EUCAST 2011 method]) (Table 1).

Table 1.

Carbapenem MICs and susceptibility rates for ESBL-producing Escherichia coli or Klebsiella pneumoniae isolates

Organism and method	MIC (μg/ml)						No. (%) of susceptible isolates
	Ertapenem		Meropenem		Imipenem		Ertapenem	Meropenem	Imipenem
	90%	Range	90%	Range	90%	Range
E. coli (n = 111)	0.5	0.03–2	0.06	0.03–0.12	0.25	0.03–2
CLSI M100-S20							111 (100)	111 (100)	111 (100)
CLSI M100-S21							93 (83.8)	111 (100)	111 (100)
EUCAST 2011							102 (91.9)	111 (100)	111 (100)
K. pneumoniae (n = 140)	0.5	0.03–16	0.06	0.03–2	0.5	0.06–4
CLSI M100-S20							138 (98.6)	140 (100)	140 (100)
CLSI M100-S21							107 (76.4)	139 (99.3)	137 (97.9)
EUCAST 2011							127 (90.7)	140 (100)	139 (99.3)

Patients with bacteremia caused by K. pneumoniae or E. coli were similar in terms of age, gender, and underlying medical diseases. Pneumonia (n = 51 [36.4%] versus n = 24 [21.6%]; P = 0.01) and skin and soft tissue infections (n = 13 [9.3%] versus n = 3 [2.7%]; P = 0.04) were more common in K. pneumoniae bacteremia. In contrast, urosepsis was more common in E. coli bacteremia (n = 40 [36.0%] versus n = 22 [15.3%]; P < 0.001). However, there were no significant differences in the critical illness (Pittsburg bacteremia score ≥ 4 points; n = 50 [35.7%] versus n = 28 [25.2%]; P = 0.1) or sepsis-related mortality rate (n = 22 [15.7%] versus n = 11 [9.9%]; P = 0.2) for cases with bacteremia caused by ESBL-producing K. pneumoniae and E. coli. Therefore, patients with E. coli and K. pneumoniae bacteremia were grouped together for further analysis.

Of the 251 patients, 75 (29.9%) patients were treated by ertapenem and 176 (70.1%) were treated by imipenem or meropenem. Considering the carbapenem MIC values of the causative isolates, 230 (91.6%) patients receiving a carbapenem treating bloodstream infections due to carbapenem-susceptible isolates, as defined by CLSI 2011 criteria (document M100-S21), were classified as the appropriate therapy group, and the remaining 21 (8.4%) were classified as the inappropriate therapy group. Patients infected by ertapenem-susceptible organisms (MICs ≤ 0.25 μg/ml) and treated by ertapenem had a lower sepsis-related mortality than those infected by ertapenem-nonsusceptible (i.e., ertapenem-intermediate and ertapenem-resistant) organisms (5.3% [3/57] versus 33.3% [6/18]; P = 0.002) (Table 2). However, if the breakpoint for ertapenem susceptibility is 0.5 μg/ml, the mortality rate for bacteremic episodes due to isolates with MICs of ≤0.5 μg/ml is similar to that for episodes due to isolates with MICs of >0.5 μg/ml (11.6% [8/69] versus 16.7% [1/6]; P = 0.8). With imipenem therapy, patients infected by imipenem-susceptible organisms had a better outcome than those infected by imipenem-nonsusceptible organisms (11.8% [16/136] versus 66.7% [2/3]; P = 0.04). In other words, the sepsis-related mortality rate was 38.1% (8/21) in the inappropriate therapy group, which is significantly higher than the 10.9% (25/230) in the appropriate therapy group (P = 0.002). Taking the concern of the adequacy of reduced doses in those with renal impairment into consideration, in a subgroup analysis of 153 cases with normal renal function receiving recommended carbapenem dosages, bacteremia due to ertapenem-nonsusceptible isolates still ensued with an outcome worse than that for cases with bacteremia due to ertapenem-susceptible isolates (37.0% [12/27] versus 7.1% [9/126]; P < 0.001).

Table 2.

With carbapenem treatment, therapeutic responses in patients with bacteremia due to ESBL producers, which were stratified by the susceptibility breakpoints of CLSI or EUCAST

Carbapenem therapy and response	CLSI M100-S21				EUCAST
	No. of fatal cases/total no. of cases (% mortality) with the following susceptibility breakpointsa:			P value	No. of fatal cases/total no. of cases (% mortality) with the following susceptibility breakpoints:			P value
	Susceptible	Intermediate	Resistant		Susceptible	Intermediate	Resistant
Ertapenem (n = 75)
Sepsis-related mortality	3/57 (5.3)	5/12 (41.7)	1/6 (16.7)	0.002	8/69 (11.6)	1/5 (20.0)	0/1 (0)	0.80
30-day mortality	6/57 (12.2)	5/12 (41.7)	1/6 (16.7)	0.028	11/69 (15.9)	1/5 (20.0)	0/1 (0)	0.88
Crude mortality	19/57 (33.3)	7/12 (58.3)	1/6 (16.7)	0.15	26/69 (37.7)	1/5 (20.0)	0/1 (0)	0.55
Meropenem (n = 37)
Sepsis-related mortality	5/37 (13.5)	0 (0)	0 (0)		6/37 (16.2)	0 (0)	0 (0)
30-day mortality	8/37 (21.6)	0 (0)	0 (0)		9/37 (24.3)	0 (0)	0 (0)
Crude mortality	16/37 (43.2)	0 (0)	0 (0)		17/37 (45.9)	0 (0)	0 (0)
Imipenem (n = 139)
Sepsis-related mortality	16/136 (11.8)	2/3 (66.7)	0 (0)	0.044	18/139 (12.9)	0 (0)	0 (0)
30-day mortality	23/136 (16.9)	2/3 (66.7)	0 (0)	0.08	25/139 (18.4)	0 (0)	0 (0)
Crude mortality	53/136 (39.0)	2/3 (66.7)	0 (0)	0.56	55/139 (39.6)	0 (0)	0 (0)
Meropenem or imipenem (n = 176)
Sepsis-related mortality	22/173 (12.7)	2/3 (66.7)	0 (0)	0.049	24/176 (13.6)	0 (0)	0 (0)
30-day mortality	34/173 (19.7)	2/3 (66.7)	0 (0)	0.1	34/176 (19.3)	0 (0)	0 (0)
Crude mortality	72/173 (41.6)	2/3 (66.7)	0 (0)	0.57	72/176 (40.9)	0 (0)	0 (0)

^aBreakpoints for susceptible, intermediate, and resistant for ertapenem were ≤0.25, 0.5, and ≥1 μg/ml, respectively, by the CLSI document M100-S21 method and ≤0.5, >0.5 to 1, and >1 μg/ml, respectively, by the EUCAST 2011 method. The corresponding breakpoints for meropenem, imipenem, and meropenem and imipenem were ≤1, 2, and ≥4 μg/ml, respectively, by the CLSI document M100-S21 method, and ≤2, >2 to 8, and >8 μg/ml, respectively, by the EUCAST method.

Of note, 51 (20.3%) patients acquiring bacteremia due to ertapenem-nonsusceptible isolates were more often associated with hospital onset (80.4% versus 65.0%; P = 0.04), rapidly fatal underlying diseases (29.4% versus 10.0%; P = 0.001), primary bacteremia (23.5% versus 9.0%; P = 0.007), inappropriate antimicrobial therapy (39.2% versus 0.5%; P < 0.001), and an unfavorable outcome (sepsis-related mortality, 33.3% [17/51] versus 8.0% [16/200]; P < 0.001) than patients with bacteremia due to ertapenem-susceptible isolates. However, in the multivariate analysis of prognostic factors among 251 patients with ESBL-producer bacteremia treated by a carbapenem, severe sepsis (odds ratio [OR], 9.70; 95% confidence interval [CI], 3.15 to 29.86; P < 0.001), hospital-onset bacteremia (OR, 5.61; 95% CI, 1.64 to 19.13; P = 0.006), and ertapenem-nonsusceptible isolates (OR, 5.17; 95% CI, 2.03 to 13.13; P = 0.001) were independent variables associated with sepsis-related mortality (Table 3).

Table 3.

Multivariate logistic regression analysis for risk factors of sepsis-related mortality due to extended-spectrum β-lactamase-producing Escherichia coli or Klebsiella pneumoniae bacteremia^a

Variable	Survivors (n = 218)	Nonsurvivors (n = 33)	Univariate analysis		Multivariate analysis
			OR (95% CI)	P value	OR (95% CI)	P value
Mean ± SD age (yr)	66.4 ± 16.6	70.5 ± 16.8	—	0.19	1.01 (0.98-1.04)	0.42
Male	118 (54.1)	19 (57.6)	1.15 (0.55–2.41)	0.85	1.13 (0.45–2.86)	0.8
Severe sepsis	55 (25.2)	27 (81.8)	13.34 (5.23–34.0)	<0.001	15.9 (5.84–43.34)	<0.001
Hospital-onset bacteremia	142 (65.1)	29 (87.9)	3.88 (1.32–11.44)	0.009	4.65 (1.42–15.24)	0.01
Rapidly fatal underlying disease	25 (11.5)	10 (30.3)	3.36 (1.43–7.86)	0.01	2.07 (0.7–6.17)	0.2
Pneumonia	58 (26.6)	17 (51.5)	2.93 (1.39–6.18)	0.007	1.56 (0.63–3.88)	0.34
Appropriate antimicrobial therapy	205 (94.0)	25 975.8)	0.2 (0.8–0.53)	0.002	0.44 (0.14–2.36)	0.57
Ertapenem-nonsusceptibleb isolates	34 (15.6)	17 (51.5)	5.75 (2.65–12.48)	<0.001	5.12 (2.04–12.88)	0.001

^aData are given as number (percent), unless otherwise specified. —, not available; OR, odds ratio; CI, confidence interval.

^bErtapenem nonsusceptible was an ertapenem MIC of >0.25 μg/ml, according to the breakpoint criteria of CLSI (document M100-S21).

To adjust the confounding factors of hospital-onset bacteremia and severe sepsis, propensity score matching was used for the outcome analysis. Fifty-one patients with bacteremia due to ertapenem-susceptible isolates could be fully matched with 51 patients with bacteremia due to ertapenem-nonsusceptible isolates. Their demographic and clinical characteristics are shown in Table 4. There were no differences in terms of age, gender, bacteremia due to E. coli, type and severity of underlying illness, bacteremic sources, and empirical carbapenem therapy between the two groups. However, infections due to ertapenem-nonsusceptible isolates were associated with a worse outcome (sepsis-related mortality rate, 35.3% versus 11.8%; P = 0.003).

Table 4.

Clinical characteristics of 51 matched pairs with bacteremia caused by ESBL-producing Escherichia coli and Klebsiella pneumoniae isolates which were susceptible or nonsusceptible to ertapenem^a

Characteristics	Ertapenem-susceptibleb group (n = 51)	Ertapenem-nonsusceptible group (n = 51)	P value
Hospital-onset bacteremia	42 (83.2)	41 (80.4)	1.0
Severe sepsis	22 (43.1)	22 (43.1)	1.0
Mean ± SD age (yr)	70.0 ± 9.6	70.0 ± 16.8	0.99
Gender, male	23 (45.1)	22 (43.1)	1.0
Bacteremia due to Escherichia coli	22 (43.1)	18 (35.3)	0.54
Polymicrobial bacteremia	17 (33.3)	18 (35.3)	0.36
Comorbidity
Diabetes mellitus	23 (45.1)	20 (39.2)	0.69
Chronic kidney disease	24 (47.1)	24 (47.1)	1.0
Malignancy	20 (39.2)	20 (39.2)	1.0
Immunosuppression	13 (25.5)	16 (31.4)	0.66
Liver cirrhosis	4 (7.85)	8 (15.7)	0.34
Organ transplant receipt	1 (2.0)	2 (3.9)	1.0
HIV infection/AIDS	1 (2.0)	0 (0)	1.0
None	2 (3.9)	5 (9.8)	0.44
Rapidly fatal underlying disease (McCabe classification)	12 (23.5)	15 (29.4)	0.65
Source of bacteremia
Pneumonia	20 (39.2)	20 (39.2)	1.0
Urosepsis	7 (13.7)	10 (19.6)	0.6
Vascular catheter-related infection	9 (18.0)	6 (11.8)	0.58
Intra-abdominal infection	4 (8.0)	3 (5.9)	1.0
Primary bacteremia	8 (15.7)	12 (23.5)	0.46
Skin and soft tissue infection	4 (7.8)	2 (3.9)	0.68
Empirical carbapenem therapy	48 (94.1)	42 (82.4)	0.12
Sepsis-related mortality	4 (8.1)	17 (33.3)	0.003
30-day mortality	6 (11.8)	18 (35.3)	0.009
Crude mortality	17 (33.3)	25 (49.0)	0.16

^aData are given as number (percent), unless otherwise specified.

^bErtapenem susceptible was an ertapenem MIC of ≤0.25 μg/ml, according to the breakpoint criteria of CLSI (document M100-S21).

In the subgroup analysis, the sepsis-related mortality rate for those with bacteremia due to ertapenem-nonsusceptible isolates was higher than that for those with bacteremia due to ertapenem-susceptible isolates, irrespective of empirical carbapenem therapy (9/22 [40.9%] versus 12/103 [11.7%]; P = 0.003) or noncarbapenem therapy (8/29 [27.6%] versus 4/97 [4.1%]; P = 0.001). Moreover, with imipenem or meropenem therapy, patients with bacteremia due to ertapenem-nonsusceptible isolates had a higher sepsis-related mortality rate than those with bacteremia due to ertapenem-susceptible isolates (33.3% [11/33] versus 9.1% [13/143]; P = 0.001) (Fig. 2). Similarly, of 75 patients with ertapenem therapy, infection by ertapenem-nonsusceptible isolates was associated with a worse outcome than infection by ertapenem-susceptible isolates (33.3% [6/18] versus 5.3% [3/57]; P = 0.005). However, among bacteremic episodes due to ertapenem-nonsusceptible or ertapenem-susceptible isolates, the sepsis-related mortality rates were similar, irrespective of ertapenem or of imipenem or meropenem therapy (Fig. 2).

Fig 2.

Outcome of sepsis-related mortality among patients with bacteremia caused by ESBL-producing Escherichia coli or Klebsiella pneumoniae isolates treated with a carbapenem, stratified by ertapenem MIC values.

DISCUSSION

Although many articles in the literature have reported the clinical features of ESBL-producing Enterobacteriaceae bacteremia, it is difficult to find and abstract clinical outcome data for individuals treated by a carbapenem. Moreover, there was heterogeneity between reports in terms of the definitions of carbapenem therapy and primary outcome, and thus, it is difficult to make comparisons. However, the majority of these published studies included a limited number of cases and reported low fatality rates for patients with carbapenem therapy. For example, in a study performed in Italy, only 1 (3.6%) of 28 patients with ESBL-producing Enterobacteriaceae bacteremia and initial carbapenem therapy died within 3 weeks (23). However, another study performed in Taiwan demonstrated that the discharge mortality rate for 54 adults with bacteremia due to ESBL-producing Enterobacteriaceae other than Klebsiella or E. coli was 27.3%, if they were treated by a carbapenem (9). The present work included the largest collection of about 250 patients with ESBL-producer bacteremia treated by a carbapenem. Their crude 30-day mortality rate was 18.3%, which was comparable to the 12.9% of 62 patients dying between 3 days and 30 days after the onset of bacteremia due to ESBL-producing E. coli and K. pneumoniae described in a report from South Korea (11). Thus, these studies indicated that carbapenem therapy for ESBL-producer bloodstream infections may fail in some patients. Identification of amendable prognostic factors for patients with ESBL-producer bacteremia is important when they are treated by one of the carbapenems, which are generally accepted as the drugs of choice in such a clinical setting.

By the multivariate analysis, several independent variables were found to be associated with sepsis-related fatality among adults with ESBL-producing E. coli or K. pneumoniae bacteremia and carbapenem therapy. These variables were mainly related to the underlying disease of the host or severity of the bloodstream infections, which were not remediable at the time of bacteremia onset. However, the factor of the appropriateness of antimicrobial therapy failed to be related to clinical outcome. One of the prognostic factors is the carbapenem susceptibility of causative isolates, which is a novel finding in the present study. Of patients with bacteremia due to isolates with decreased susceptibility to ertapenem (MICs > 0.25 μg/ml), clinical outcomes were worse than those among patients with bacteremia due to ertapenem-susceptible isolates (MICs ≤ 0.25 μg/ml).

The categorization of susceptibility or resistance is dependent on the breakpoints used, which in the case of carbapenem varies somewhat on the basis of the agency, the CLSI or EUCAST. However, such a small difference could have great implications. When the MIC breakpoint of ertapenem susceptibility followed the current EUCAST criterion (≤0.5 μg/ml), the survival disadvantage of carbapenem therapy for infections due to the isolates with an MIC of 0.5 μg/ml was not evident and these cases would be inappropriately treated by a carbapenem. Our results indicate that the preferred MIC breakpoint of ertapenem susceptibility for Enterobacteriaceae is 0.25 μg/ml and not 0.5 μg/ml.

The group 2 carbapenems, imipenem and meropenem, have been accepted to retain moderate activity against isolates with low-level ertapenem resistance (24) and have been used as salvage therapy to treat infections due to isolates in which ertapenem resistance emerges during therapy (6). This differential susceptibility to group 1 and 2 carbapenems among Enterobacteriaceae has generally been agreed on (20, 24). However, use of such a clinical practice deserves some precaution, since in the present study, bloodstream infections due to ertapenem-nonsusceptible ESBL producers treated by ertapenem or by imipenem or meropenem were associated with a fatality rate of 33%. Likewise, previous studies have proposed ertapenem nonsusceptibility to be a marker for carbapenem resistance (22). However, due to the limited number of cases included in the study, it is too early to preclude the use of imipenem or meropenem for severe infections due to ertapenem-nonsusceptible ESBL-producing isolates. Clinicians should be reminded to be alert about the therapeutic efficacy of carbapenem therapy for infections due to ertapenem-nonsusceptible pathogens.

Without exception, our retrospective study has several limitations. First, though there were some differences between the individuals in the ertapenem-nonsusceptible and ertapenem-susceptible groups, we did the subgroup analysis, multivariate analysis, and propensity score matching analysis to minimize the bias of the uneven distribution of clinical characteristics. All analytic results indicated the same finding that ertapenem susceptibility was significantly associated with sepsis-related mortality. Second, the outcome data for individuals with ESBL-producing E. coli or K. pneumoniae bacteremia were combined for analysis. It is generally assumed that E. coli and K. pneumoniae bacteremias behave similarly, since such a combination was commonly adopted in the literature (11, 23). Third, since only clinical data regarding the hospitalization period were available, we could analyze only the in-hospital outcome. It remains undefined if there is any difference in long-term outcome between different treatment groups. Finally, in Taiwan there are no indigenous isolates with KPC or NDM enzymes, and the mechanisms mediating carbapenem resistance in E. coli isolates were production of AmpC beta-lactamases in conjunction with porin deficiency (2, 25). The correlation between clinical outcome and carbapenem susceptibility may not be generalized to Western countries with a variable prevalence of KPC or NDM beta-lactamases among Enterobacteriaceae isolates. However, our study provides clinical evidence supporting the concept of the MIC-based therapeutic approach (10), in which knowledge of the precise resistance mechanisms in the causative strains would not be necessary.

In conclusion, reduced ertapenem susceptibility was observed, and the clinical outcomes of the patients treated by a carbapenem for bacteremia due to ESBL-producing E. coli or K. pneumoniae varied with the ertapenem susceptibility of the causative pathogens. Infections due to isolates with ertapenem MIC values of ≤0.25 μg/ml were associated with a favorable outcome if treated by any of the three carbapenems. Such a finding supports the rationale for the updated CLSI criteria of carbapenems for Enterobacteriaceae.