Risk Factors and Clinical Impact of Klebsiella pneumoniae Carbapenemase–Producing K. pneumoniae

Abstract

BACKGROUND

Klebsiella pneumoniae carbapenemase (KPC)–producing K. pneumoniae is an emerging pathogen with serious clinical and infection control implications. To our knowledge, no study has specifically examined risk factors for KPC-producing K. pneumoniae or its impact on mortality.

METHODS

To identify risk factors for infection or colonization with KPC-producing K. pneumoniae, a case-control study was performed. Case patients with KPC-producing K. pneumoniae were compared with control subjects with carbapenem-susceptible K. pneumoniae. A cohort study evaluated the association between KPC-producing K. pneumoniae and in-hospital mortality.

RESULTS

Fifty-six case patients and 863 control subjects were identified. In multivariable analysis, independent risk factors for KPC-producing K. pneumoniae were (1) severe illness (adjusted odds ratio [AOR], 4.31; 95% confidence interval [CI], 2.25–8.25), (2) prior fluoroquinolone use (AOR, 3.39; 95% CI, 1.50, 7.66), and (3) prior extended-spectrum cephalosporin use (AOR, 2.55; 95% CI, 1.18, 5.52). Compared with samples from other anatomic locations, K. pneumoniae isolates from blood samples were less likely to harbor KPC (AOR, 0.33; 95% CI, 0.12, 0.86). KPC-producing K. pneumoniae was independently associated with in-hospital mortality (AOR, 3.60; 95% CI, 1.87–6.91).

CONCLUSIONS

KPC-producing K. pneumoniae is an emerging pathogen associated with significant mortality. Our findings highlight the urgent need to develop strategies for prevention and infection control. Limiting use of certain antimicrobials, specifically fluoroquinolones and cephalosporins, use may be effective strategies.

In North America, Klebsiella species are among the most common pathogens recovered in intensive care units.^1,2 Klebsiella pneumoniae is also among the bacteria that most readily develop resistance mechanisms to multiple classes of antibiotics, and the prevalence of drug resistance is increasing at an alarming rate.^3,4 Because of the emergence of extended-spectrum β-lactamases, carbapenems have been the agents of choice for the management of multidrug-resistant K. pneumoniae infection.

Carbapenem resistance among K. pneumoniae emerged a decade ago.⁵ Reports of carbapenem resistance were initially sporadic, and such resistance was attributed to several mechanisms, including high-level production of an AmpC β-lactamase combined with loss of outer membrane proteins or, rarely, efficient carbapenem-hydrolyzing β-lactamases (eg, the class B metallo-β-lactamases).⁶ However, in 2001, a plasmid-mediated class A β-lactamase enzyme, K. pneumoniae carbapenemase type 1 (KPC-1), was identified in K. pneumoniae. Initially limited to the northeastern United States, where the prevalence of the KPC enzyme among K. pneumoniae is as high as 24% in some areas, KPC-type enzymes (ie, KPC-2 and KPC-3) have spread globally at an alarming rate.^7–11 KPC production may be an even greater problem than currently recognized, because carbapenem resistance in KPC-producing K. pneumoniae isolates is frequently not detected by automated microdilution susceptibility testing routinely used in clinical microbiology laboratories.^8,12 In most cases, because of the existence of multiple resistance mechanisms associated with KPC-containing plasmids, few therapeutic options exist for treating infections due to KPC-producing K. pneumoniae, and high mortality has been reported.^8,13,14

KPC-producing K. pneumoniae appears to be the next major challenge in antimicrobial resistance. Closer investigation of the risk factors and impact of this organism is warranted to develop interventions aimed at curbing its rapid emergence. Although risk factors for carbapenem-resistant K. pneumoniae have been reported by several investigators,^14–17 we are unaware of any study that has investigated risk factors for carbapenem resistance specifically caused by the presence of KPC. As such, this study aims to identify risk factors for isolation of KPC-producing K. pneumoniae in clinical cultures among a large cohort of patients. We also assessed the independent impact of KPC-producing K. pneumoniae on mortality.

METHODS

This study was conducted at the Hospital of the University of Pennsylvania (Philadelphia), a 725-bed academic tertiary care medical center, and Penn Presbyterian Medical Center (Philadelphia), a 344-bed urban hospital. The study was approved by the Committee on Studies Involving Human Beings of the University of Pennsylvania.

A case-control study was performed to identify risk factors for KPC-producing K. pneumoniae. All patients older than 18 years who had an inpatient clinical culture positive for K. pneumoniae from October 1, 2006 through April 30, 2008, were eligible for inclusion. Individuals became eligible on the first day that a K. pneumoniae isolate was identified, and each individual was eligible only once, regardless of whether K. pneumoniae was isolated on subsequent days. Patients with KPC-producing K. pneumoniae were designated as case patients. To evaluate for the presence of KPC, K. pneumoniae isolates were screened for carbapenem nonsusceptibility with use of ertapenem by disk diffusion testing or by Vitek2 panel GN-20 (bioMérieux).¹⁸ Ertapenem has been shown to be more sensitive in identifying carbapenem resistance, compared with imipenem and meropenem.^8,19 KPC-producing K. pneumoniae was then confirmed using either polymerase chain reaction (PCR) to detect bla_KPC or the Modified Hodge test phenotypic assay using meropenem as the indicator.²⁰ In the United States, where carbapenemases in K. pneumoniae are almost exclusively attributable to KPC, the Modified Hodge test has shown 100% specificity and sensitivity in published reports²⁰ and in testing at our institution. All carbapenem-nonsusceptible isolates with negative PCR and/or phenotypic testing results were excluded. The mechanism of carbapenem nonsusceptibility in these isolates was presumed to be a porin mutation plus an extended-spectrum β-lactamase and/or AmpC, rather than KPC production.

All patients with carbapenem-susceptible K. pneumoniae yielded by an inpatient clinical culture during the study period were designated as control subjects. Although there has been considerable debate regarding the choice of control groups in studies of antimicrobial resistance, we chose persons with carbapenem-susceptible K. pneumoniae to be our control subjects, because we sought specifically to evaluate risk factors for KPC production among patients with K. pneumoniae infection or colonization.²¹

Potential risk factors were assessed through the use of a comprehensive health system computer database, which has been used effectively for similar studies of antimicrobial use in the past.²² Data obtained included age, sex, race, origin of patient at the time of hospital admission (eg, home or transferred from another institution), hospital location at the time that the clinical isolate was identified, and duration of hospital stay prior to identification of the isolate. The anatomic source of the K. pneumoniae isolate and the antibiotic susceptibility profile of the isolate were noted. If K. pneumoniae was identified in more than 1 anatomic site on the same day, other sites were noted. Nosocomial acquisition was defined as an isolate identified more than 48 hours after admission to the hospital or an isolate identified no later than 48 hours after admission in a patient transferred from another medical institution (eg, other hospital or long-term care facility). Prior antibiotic exposure was defined as at least 2 days of therapy administered during the 30 days prior to the culture. The presence of comorbid conditions was documented; such conditions included hepatic dysfunction and/or cirrhosis, anemia (indicated by hemoglobin level less than 10.0 g/dL), malignancy, diabetes mellitus, renal insufficiency (indicated by creatinine level >2.0 mg/dL or the requirement of dialysis), human immunodeficiency virus (HIV) infection, cardiopulmonary disease, and a history of solid organ transplantation. To adjust for severity of illness, we used the All Patient Refined Diagnosis Related Group (APR-DRG) method to classify patients into 4 severity of illness categories: minor, moderate, major, and extreme.²³

Bivariable analyses were conducted to determine the unadjusted association between potential risk factors and the presence of KPC-producing K. pneumoniae. Categorical variables were compared using the Fisher exact test. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated to evaluate the strength of any association. Continuous variables were compared using the Student t test or the Wilcoxon rank-sum test, depending on the validity of the normality assumption. Multivariable analysis was performed using forward stepwise multiple logistic regression. Variables with a P value <.20 in bivariable analyses were considered for inclusion in a multivariable model. Backward stepwise multiple logistic regression was also performed to determine whether identification of risk factors varied with the approach to multivariable analysis.

A cohort study was performed to evaluate the association between isolation of a KPC-producing K. pneumoniae isolate and in-hospital mortality. Individuals previously identified as case patients and control subjects comprised the exposed (patients with KPC-producing K. pneumoniae infection) and unexposed (patients with carbapenem-susceptible K. pneumoniae infection) participants in the cohort study, respectively. Bivariable analyses were performed, and variables with P < .2 in bivariable analysis were considered for inclusion in a multivariable model. Variables remained in the final model if their inclusion resulted in a >15% change in the effect size for the association between isolation of KPC-producing K. pneumoniae and mortality.

For all calculations, a 2-tailed P value <.05 was considered to be statistically significant. All statistical calculations were performed using standard programs in Stata, version 10.0 (Stata).

RESULTS

During the study period, 928 inpatients had a clinical culture positive for K. pneumoniae. Carbapenem-resistant K. pneumoniae was identified in 65 individuals, and KPC-producing K. pneumoniae was confirmed in 56. Carbapenem resistance was attributable to alternative mechanisms in 6 isolates, and 3 isolates were not available for PCR and/or Modified Hodge testing; individuals with these isolates were excluded. Anatomic sites of cultures were urine in 550 individuals (59.9%), blood in 162 (17.3%), respiratory tract in 111 (12.1%), abdomen in 81 (8.1%), and other or unknown in 61 (6.6%). Of note, K. pneumoniae was simultaneously identified in multiple sites in 41 individuals, including blood and urine in 30 and blood and abdomen in 5.

Case patients were more often critically ill, as indicated by a higher rate of intensive care unit admission, the use of mechanical ventilation, and categorization of illness as extreme by APR-DRG (Table 1). Prior use of numerous antibiotics was also significantly more common among case patients. There were no differences in age or race, aminoglycoside use, and the presence of comorbidities (ie, malignancy, liver disease, renal insufficiency, cardiopulmonary disease, HIV infection, diabetes, and history of solid organ transplantation). In multivariable analysis (Table 2), APR-DRG extreme illness category, prior fluoroquinolone use, and prior extended-spectrum cephalosporin use were significantly associated with isolation of KPC-producing K. pneumoniae. Isolates obtained from the blood were less likely to harbor KPC, compared with those obtained from other anatomic sites. Results of backward and forward stepwise regression were not substantively different.

TABLE 1.

Factors Associated with Klebsiella pneumoniae Carbapenemase–Producing K. pneumoniae in Bivariable Analysis

Variable	Case patients (n = 56)	Control subjects (n = 863)	OR (95% CI)	Pa
Time from hospital admission to K. pneumonia isolation, median days (IQR)	4 (1–15)	1 (0–6)	…	.004
Transferb	20 (35.7)	153 (17.7)	2.58 (1.37–4.72)	<.001
Female sex	28 (50.0)	354 (41.0)	1.44 (0.80–2.56)	.19
Used mechanical ventilation at time of isolate identification	17 (30.4)	126 (14.6)	2.55 (1.31–4.78)	.002
Surgical service	31 (55.4)	336 (38.9)	1.94 (1.09–3.50)	.02
Stayed in ICU at the time of culture	19 (33.9)	192 (22.3)	1.79 (0.95–3.29)	.04
APR-DRG severity of illness extreme	42 (75.0)	317 (36.7)	5.17 (2.71–10.40)	<.001
Prior antimicrobial use
Fluoroquinolones	10 (17.9)	30 (3.5)	6.03 (2.47–13.62)	<.001
Carbapenems	2 (3.6)	8 (0.9)	3.96 (0.40–20.46)	.06
Extended-spectrum cephalosporins	11 (19.6)	46 (5.3)	4.34 (1.89–9.22)	<.001
First- or second-generation cephalosporins	5 (8.9)	35 (4.1)	2.32 (0.68–6.30)	.08
Piperacillin-tazobactam	6 (10.7)	31 (3.6)	3.22 (1.05–8.32)	.009
Anti-anaerobic drugc	20 (35.7)	137 (15.9)	2.94 (1.56–5.40)	<.001
Vancomycin	18 (32.1)	95 (11.0)	3.83 (1.97–7.19)	<.001
Microbiologic characteristic
Healthcare acquired	38 (67.9)	408 (47.3)	2.35 (1.29–4.45)	.003
Urine isolate	29 (51.8)	521 (60.4)	0.71 (0.40–1.26)	.20
Blood isolate	5 (8.9)	157 (18.2)	0.44 (0.14–1.12)	.08
Respiratory isolate	13 (23.2)	98 (11.4)	2.36 (1.12–4.66)	.008

NOTE. Data are no. (%) of participants, unless otherwise indicated. Only factors with P ≤ .20 are shown. These variables were considered for inclusion in multivariable analysis. APR-DRG, All Patient Refined Diagnosis Related Group; CI, confidence interval; ICU, intensive care unit; IQR, interquartile range; OR, odds ratio.

^aThe Fisher exact test was used for categorical variables, and the Wilcoxon rank-sum test was used for continuous variables.

^bTransferred from another medical facility, including long-term care facilities.

^cAmoxicillin-clavulanate, ampicillin-sulbactam, piperacillin-tazobactam, meropenem, metronidazole, and clindamycin.

TABLE 2.

Multivariable Analysis (Case-Control Study)

Variable	AOR (95% CI)	P
APR-DRG severity of illness extreme	4.31 (2.25–8.25)	<.001
Prior fluoroquinolone use	3.39 (1.50–7.66)	.003
Prior extended-spectrum cephalosporin use	2.55 (1.18–5.52)	.02
Blood isolate	0.33 (0.12–0.86)	.02

NOTE. AOR, adjusted odds ratio; APR-DRG, All Patient Refined Diagnosis Related Group; CI, confidence interval.

Among the 56 persons who had KPC-producing K. pneumoniae isolated, 18 (32.1%) died during hospitalization, compared with 85 (9.9%) of the 863 patients who had carbapenem-susceptible K. pneumoniae isolated (P < .001). Among patients who died, time to death after the date of the first positive culture result was significantly longer for patients with KPC-producing K. pneumoniae than for patients with carbapenem-susceptible K. pneumoniae. The median period from the first positive K. pneumoniae culture result to death was 18 days (interquartile range, 5–23 days) among patients with KPC-producing K. pneumoniae infection, compared with 8 days (interquartile range, 3–16 days) among patients with carbapenem-susceptible K. pneumoniae infection (P = .02). In a multivariable model, APR-DRG extreme severity of illness and use of mechanical ventilation on the day of culture confounded the relationship with isolation of KPC-producing K. pneumoniae; however, infection or colonization with KPC-producing K. pneumoniae remained significantly associated with in-hospital mortality (adjusted OR, 2.28; 95% CI, 1.18–4.40; P < .02).

DISCUSSION

Our study demonstrates that severity of illness, prior fluoroquinolone use, and prior extended-spectrum cephalosporin use are risk factors for isolation of KPC-producing K. pneumoniae. We also found that isolation of KPC-producing K. pneumoniae is independently associated with in-hospital mortality.

There is ample laboratory evidence to support an association between KPC and prior fluoroquinolone use. Plasmid-encoded qnr genes, which convey low-level fluoroquinolone resistance, have been identified in the same conjugative K. pneumoniae plasmid as KPC genes (specifically bla_kpc-2 and qnrB4).²⁴ Similarly, Endimiani et al identified both bla_kpc-3 and qnrB19 on the transferable K. pneumoniae plasmid pLRM24.²⁵ Those authors subsequently identified a complex genetic region, the KQ element (KPC and qnr), which not only contains bla_kpc-3 and qnrB19 but also is associated with a Tn4401-like element.²⁶ Previous work suggested that the putative KPC-2 encoding transposon Tn4401 contributes to the dissemination of these enzymes.²⁷ Indeed, fluoroquinolone use has been identified as a risk factor for carbapenem-resistant K. pneumoniae infection in past epidemiologic studies, although those studies did not focus specifically on carbapenem resistance due to KPC production.^14,15

That yet another resistance pattern appears to be associated with fluoroquinolone use is concerning, especially in light of widespread prescribing of these drugs in recent years.²⁸ Fluoroquinolone use has been associated with the emergence of resistance to other nonfluoroquinolone agents and multidrug resistance in a variety of bacteria.^22,29–32 Not surprisingly, plasmids carrying qnr usually also encode other non-KPC resistance genes, particularly extended-spectrum β-lactamases.³³

Prior cephalosporin use^16,17 and prior carbapenem use^14,16,17 have been associated with carbapenem-resistant K. pneumoniae in previous studies. In our study, very few patients had a history of carbapenem use (1.1%). As such, prior use of carbapenems as a risk factor for KPC production in K. pneumoniae isolates cannot be excluded. However, KPC enzymes have been shown to confer greater resistance to cephalosporins than to carbapenems, and therefore, a significant selective pressure for KPC-harboring bacteria induced by cephalosporin use is certainly plausible.³⁴ Because existing studies included all carbapenem-resistant K. pneumoniae isolates, regardless of mechanism of resistance, future studies should better clarify the role of carbapenem use in the acquisition of KPC-producing K. pneumoniae.

In the present cohort study, we confirmed that identification of a KPC-producing K. pneumoniae isolate is associated with in-hospital mortality. We also found that death was delayed among case patients, compared with control subjects. Although these findings are interesting, several limitations of our study warrant future studies to more closely examine factors associated with mortality to better elucidate the clinical course of infection due to KPC-producing K. pneumoniae. Although we controlled for severity of illness, we were not able to differentiate between infection and colonization, and the high in-hospital mortality may reflect the severity of underlying disease rather than infection with KPC-producing K. pneumoniae. Of note, because we did not follow up with patients after the hospital admission during which their first K. pneumoniae isolate was identified, the impact on mortality may also be underestimated. We did not evaluate choice of therapy or time to appropriate therapy. In a prior study, only failure to remove the source of infection, not time to therapy, was associated with mortality among patients with infection due to carbapenem-resistant K. pneumoniae.¹⁷ In this study, however, no adjustment was made for severity of illness, and these findings may not hold true for carbapenem resistance due to KPC-producing K. pneumoniae.

Our study is unique in several ways. There have been few studies examining carbapenem resistance among K. pneumoniae, but none specifically examined risk factors for KPC production as confirmed by PCR and/or Modified Hodge testing.^14–17,35 Several of these studies were performed before the documentation of KPC in the geographic area of the study setting.^15,16,35 Furthermore, many used automated carbapenem testing, which does not detect carbapenem resistance in 6%–87% of known resistant isolates.¹² Finally, no study used ertapenem to screen for carbapanem resistance; ertapenem has been shown to be a more sensitive indicator of KPC than meropenem and imipenem.^12,20 Although carbapenem-resistant isolates may be phenotypically similar, risk factors and outcomes may differ according to the mechanism of resistance.

Several additional potential limitations exist. We did not include details regarding prior antibiotic exposure in an out-patient context, because these data were not consistently available. Our study was conducted in a large tertiary care medical center, and the results may not be generalizable to other institutions. Of interest, time to identification of KPC-producing K. pneumoniae was relatively short in our study, compared with others (4 days). This may reflect differences in patient population and/or transmission dynamics at our institution. Although we attempted to control for potential confounders, residual confounding by unknown or unmeasured confounders is possible in any nonrandomized study. As mentioned, we did not determine whether a positive clinical culture represented colonization or infection and used only clinical cultures, rather than performing surveillance, to detect colonization with KPC-producing bacteria. Of note, the optimal method for detecting asymptomatic patients colonized with KPC-producing K. pneumoniae is still unknown.

KPC-producing K. pneumoniae is an emerging pathogen associated with significant mortality. Our current antimicrobial armamentarium is ill equipped to battle infections caused by this organism. Our findings highlight the urgent need to develop strategies for prevention and infection control and underscore the importance of using antimicrobial stewardship to minimize unnecessary and inappropriate antimicrobial use, particularly of fluoroquinolones and extended-spectrum cephalosporins. Restriction of fluoroquinolone use appears to be particularly important in the fight against the emergence of antimicrobial resistance.