Clinical Outcomes of Extended-Spectrum Beta-Lactamase-Producing Enterobacteriaceae Infections with Susceptibilities among Levofloxacin, Cefepime, and Carbapenems

Purpose Highly resistant Gram-negative bacterial infections are associated with high mortality. Increasing resistance to standard therapy illustrates the need for alternatives when treating resistant organisms, especially extended-spectrum beta-lactamase- (ESBL-) producing Enterobacteriaceae. Methods A retrospective chart review at a community hospital was performed. Patients who developed ESBL-producing infections were included. Patients less than eighteen years old, who were pregnant, or who were incarcerated were excluded. The primary outcome was hospital mortality. The secondary outcomes included intensive care unit (ICU) mortality, ICU length of stay, and hospital length of stay. Results 113 patients with ESBL-producing infections met the criteria for review. Hospital mortality: carbapenem (16.6%), cefepime (0%), and levofloxacin (15.3%) (p=0.253). ICU mortality: carbapenem (4.5%), cefepime, (0%), and levofloxacin (3.7%) (p=0.616). Mean ICU and hospital length of stay: carbapenem (9.8 ± 16, 12.1 ± 1 days), cefepime (7.8 ± 6, 11.1 ± 10.5 days), and levofloxacin (5.4 ± 4.1, 11.1 ± 10.4 days) (p=0.805, 0.685). No predictors were clearly found between the source of infection and mortality. Conclusion Cefepime or levofloxacin can be a potential alternative agent for infections with ESBL-producing Enterobacteriaceae, and larger clinical trials investigating these outcomes are warranted.


Introduction
Extended-spectrum beta-lactamase-(ESBL-) producing Enterobacteriaceae infections are increasingly identi ed with subsequent hospitalization [1,2]. ESBL is a resistance mechanism in which the beta-lactam ring of antibiotics such as penicillins, cephalosporins, and aztreonam is hydrolyzed, inactivating the antibiotic. Increasing resistance was demonstrated in a 2013 Centers for Disease Control and Prevention (CDC) report, which included 26,000 ESBL-producing Enterobacteriaceae infections and 1700 deaths in the United States [3]. Traditionally, the treatment of choice for ESBLproducing Enterobacteriaceae infections has been the carbapenem class. However, trends in resistance demonstrate the need of conservative use of carbapenems and illustrate the need for alternative therapies against ESBL-producing organisms. Additionally, chromosomally mediated inducible resistance has manifested as porin loss, which decreases drug entry in strains already retaining high levels of AmpC [4]. Emerging incidence of community-acquired ESBL-identi ed infections has demonstrated a positive association with the increasing trend of carbapenem resistance in Enterobacteriaceae.
It is of interest to the patient for organisms (E. coli being the most common to exhibit ESBLs) to be detected and reported to healthcare providers in a timely fashion for both appropriate treatment to be initiated and resulting morbidity and mortality to be signi cantly reduced. However, reporting of ESBL can be delayed by a few days due to the screening and identi cation process. Many microbiology laboratories are challenged by increasingly changing and complicated recommendations [11]. Minimum inhibitory concentrations (MICs) of β-lactamase-producing Enterobacteriaceae might be reported as high but are still in the susceptible range (i.e., "hidden resistance") [7,8]. If diagnostic microbiology laboratories are unable to adequately test for ESBL production, it could be argued that suspect cases of hidden resistance will go unrecognized by the microbiologists and clinicians, increasing the likelihood for adverse consequences [8][9][10]. Other data showed that carbapenem use when MIC is ≥4 mg/liter had outcomes of increased mortality compared to patients whose microbiological isolate MICs were lower, demonstrating carbapenem therapy limitations [12].
Because Enterobacteriaceae organisms are becoming increasingly resistant to standard therapy, treatment selected by interpretation of MICs of ESBL producers and by clinical outcome may o er therapeutic alternatives in more challenging situations [7,[12][13][14]. Appealing to the cause of antimicrobial stewardship, the fourth-generation cephalosporin cefepime could be a considerable alternative against ESBL-producing organisms and could shift trends for carbapenem use. Cefepime has shown to have greater, stable activity against ESBL (i.e., AmpC enzymes) compared to the other extended-spectrum cephalosporins [9,10,14]. Cefepime does not rely on porins for entry into the bacterial cell, making it useful in the down-regulated porin resistance mechanism [15,16]. Appealingly, cefepime could be used in therapy for ESBL-EC and ESBL-K strains highly resistant to third-generation cephalosporins (including ceftazidime) and to aztreonam [14,[17][18][19][20][21]. Fluoroquinolones may also o er a place as alternative therapy in treating ESBL-producing Enterobacteriaceae. ough many bacteria are becoming increasingly resistant to uoroquinolones, treatments with this class when susceptibilities allow have demonstrated improved inpatient mortality when compared to carbapenems [22][23][24][25]. A small retrospective review with in vitro samples indicated no di erence in inpatient mortality with levo oxacin when comparing low, intermediate, and high MICs. However, they did demonstrate trends for a shorter improved mortality and shorter length of stay in the lower MIC group [22]. e objective of this study is to evaluate clinical outcomes among levo oxacin, cefepime, and carbapenem used for ESBL-producing Enterobacteriaceae infections. Additional objectives include identifying risk factors that can help tailor antibiotic therapy.

Methods
is is a retrospective chart review conducted at one 522 (including 22 ICU) bed tertiary community medical center. Patient charts were included from January 1, 2012, to September 30, 2015. Records and pro les of discharged patients who received cefepime, levo oxacin, or carbapenem for ESBL pathogens within the speci ed time span were identi ed by the microbiology lab department. is study received approval by the Texas Tech University Health Sciences Center Institutional Review Board. Individuals with ESBL-producing Enterobacteriaceae infections were identied, who also waived the need for informed consent. Eligible patients ful lled each of the following criteria: (1) aged 18 years or older, (2) clinically diagnosed with the ESBL-producing Enterobacteriaceae isolate demonstrated on culture, (3) received empirical treatment with cefepime, levo oxacin, or carbapenem for at least 48 hours after initial cultures had been drawn, and (4) admitted for inpatient treatment. Ninety days prior to culture, risk factors were identi ed. e researchers took no part in the conduct of the study, nor did they play a role in the analysis of the data.
An infection occurring in the ICU was de ned as the patient being present in the ICU when the culture was identi ed for the Gram-negative organism. Infectious and other diagnoses were identi ed by ICD-9 codes in the patient chart. Acute renal failure is identi ed when the serum creatinine is increased by 50% or more from the patient's baseline serum creatinine on the onset of infection. Included in this were patients with a past medical history of chronic kidney disease (CKD) noted in the chart. Antimicrobial therapy administered after diagnosis was regarded as empirical treatment, and therapy administered afterward was de ned as de nitive therapy.
is single medical center has a centralized clinical microbiology laboratory which processes 96,000 samples annually. At this laboratory, bacteria can be identi ed down to a genus species, and the susceptibilities are further outlined by prede ned antimicrobials. An automated broth microdilution system (MicroScan; Siemens AG, Germany) allows for susceptibility reporting, and an analysis is conducted in accordance with the Clinical and Laboratory Standards Institute (CLSI) criteria [26]. To identify ESBL-producing bacteria, isolates are processed using the MicroScan WalkAway system.
is system initially screens for ceftazidime and cefpodoxime resistance, which is subsequently con rmed on demonstration of synergy between these two antibiotics and additionally by clavulanic acid on disc synergy testing. e interpretation followed the current breakpoints recommended by the CLSI. e severity of underlying medical illness was calculated by the Elixhauser scoring system [27]. e primary study outcome was hospital mortality. e secondary outcomes were ICU mortality, ICU length of stay, and hospital length of stay. e following baseline characteristics were also collected: age, gender, type of infection, source of infection, culture species, and sensitivities.
All analyses were performed using Strata Version 13.1 (StataCorp, College Station, TX). Descriptive statistics were used to quantify patient characteristics, as well as for analysis of the primary and secondary outcomes. Categorical variables were expressed as percentages of total numbers of patients analyzed. Continuous variables were expressed as mean values ± SDs.
e Kruskal-Wallis test was used to compare di erences between groups. Multivariate regression analyses were used to identify risk factors that may have association with clinical outcomes. Data are summarized as n (%). Adjusted odds ratios (ORs) are presented with 95% con dence intervals and were calculated using multivariate logistic regression. OR p values correspond to adjusted ORs.

Baseline Characteristics.
In this review, 113 patients with ESBL-producing Enterobacteriaceae infections were successfully identi ed at the community hospital during the 3-year study period (Table 1). 66 patients were in the carbapenem group, 21 in the cefepime group, and 26 in the levo oxacin group. Table 1 shows the characteristics of the patients at baseline. e mean age of the patients was 69.8 years, and 60% were female. ere were 77.3% patients in the carbapenem group, 85.7% in the cefepime group, and 88.9% in the levo oxacin where the pathogen was identi ed as E. coli. e predominant infection treated was urinary tract infection (UTI) with 66.7% in the carbapenem group, 57.1% in the cefepime group, and 70.4% in the levo oxacin group. Additionally, 7.6% in the carbapenem group, 0% in the cefepime group, and 7.4% in the levo oxacin group were treated for intra-abdominal infection. erefore, no patients were treated with cefepime for intra-abdominal infection.
Furthermore, 15.2% in the carbapenem group, 28.6% in the cefepime group, and 7.4% in the levo oxacin group were treated for skin and soft tissue infection (SSTI). Of note, carbapenem and levo oxacin groups were treated for over 60% hospital-acquired or healthcare-associated infections, while the cefepime group was treated for 61.9% communityacquired infections.

Primary and Secondary
Outcomes. Hospital mortality and ICU mortality were not di erent in all three groups. No one died in the cefepime group even though there was no statistical di erence (Table 2). ICU length of stay was shorter in the cefepime and levo oxacin groups, by 2.0 and 4.4 days, respectively. Hospital length of stay was shorter by 1 day in the cefepime and levo oxacin groups compared to the carbapenem group, although this was not statistically di erent.

Multivariate Analysis.
ere were no predictors found between the source of infection and mortality (Table 3).

Discussion
ere is controversy in recent literature regarding alternative therapies against infections due to ESBL-producing Enterobacteriaceae. is study has outlined outcomes for a rural healthcare system where cefepime and levo oxacin were used as alternatives for treatment of infections.
Studies demonstrate improved inpatient mortality when susceptibilities have allowed treatment with uoroquinolones such as cipro oxacin and levo oxacin [22][23][24]. However, most of these studies only studied bacteremia as the primary type of infection, whereas our study allowed for all types of infection. Furthermore, these studies were following the earlier break points of MIC susceptibility established by the CLSI. Antimicrobial sensitivity to cefepime and levo oxacin among Enterobacteriaceae is de ned by the 2014 CLSI update as an MIC of less than or equal to 2 μg/mL [26]. A randomized control trial by Seo et al. demonstrated a high treatment failure comparing cefepime, piperacillin/tazobactam, and ertapenem in treatment of ESBL-producing E. coli infection [28]. However, the baseline characteristics of patients in the study were not well aligned. e sex di erences between cefepime and the other 2 groups varied, as well as the average age of patients in the cefepime group was approximately 10 years greater than the piperacillin/tazobactam and ertapenem groups. Our study, however, showed similar reporting between ages and sexes.
Previous literature showed an association between clinical outcomes and MICs. ese studies have suggested that a low MIC value is positively associated with better clinical outcomes versus a higher MIC [9,10,17,[19][20][21][22]. Recent literature has trended favorably from a sensitivitybased approach to an MIC-based approach in settings where ESBL producers are endemic [29][30][31]. Studies have shown that cefepime and levo oxacin can both favorably achieve pharmacodynamic targets against isolates of fully susceptible Enterobacteriaceae at lowered MIC susceptibilities [11,12,15,18,22]. ere have been underlying issues with studies regarding the lack of clinical experience using these medications, as well as the con icting reports of clinical outcomes in the literature when studying treatments for ESBLproducing E. coli infection. Additional research would help demonstrate that drug target attainment in ESBL-producing Enterobacteriaceae with cefepime and levo oxacin will achieve adequate drug target levels when using the recent updates for susceptible breakpoints for cefepime and levo oxacin.
Additionally, the doses of cefepime used in the study may have been subtherapeutic, whereas our doses were reviewed for therapeutic e cacy and we also took into account if renal dose adjustments had been made [28]. Lower doses of cefepime have been associated with treatment failure and higher mortality [28,32]. Literature suggests that higher maximal doses of cefepime, even when renally adjusted, may be more bene cial than standard cefepime dosing [32]. Because time above MIC is critical for clinical success of cefepime dosing, our study evaluated if the dose was appropriate based on the infection type and renal function [32,33].
Previous retrospective chart review had observed a trend of lower 30-day mortality in the uoroquinolone group compared to the carbapenem group, although not statistically signi cant (OR 4.53; 95% CI 0.98-21) [23]. Similarly, a higher 30-day mortality in the cefepime group was demonstrated when CLSI breakpoints were 8 or greater (OR 7.1; 95% CI 2.5-20.3) [21]. However, unlike our study, these two trials only looked at bacteremia infection and had used previously de ned CLSI breakpoints for cefepime. Although this study did not nd a statistically signi cant di erence either, when using a lower MIC breakpoint, there was a trend for lower mortality for both alternative therapy groups. ereby, it was suggested that cefepime therapy was limited for bacteremia caused by ESBL-producing Enterobacteriaceae organisms when cefepime MIC was less than or equal to 1 μg/mL [21,23]. Additional comparisons of length of hospital stay had shown that higher MIC resulted in increased length of stay of 5.67 days (95% CI 0.77-10.62 days; p � 0.02) with levo oxacin [22]. Similar to other studies, our study was limited by allowing other antimicrobial agents that could have altered the outcomes, although matched group analysis was applied to our study to o set this limitation.  Our study did not identify a statistically signi cant difference in hospital mortality between patients who were receiving treatment with carbapenem and those who received an alternative treatment. However, this study did nd a trend between carbapenem therapy and an increased risk of mortality and alternative therapy and a decreased risk of mortality. It follows that there might be greater in vitro activity in favor for cefepime, speci cally for not being induced by the AmpC gene. Interestingly, no one died in the cefepime group even though there was no statistical di erence. is could be explained by many community-acquired infections in the cefepime group, but this suggests that cefepime can be a good treatment option in ESBL community-acquired infections. Carbapenem therapy was associated with a trend for longer hospital stay than with cefepime therapy. is association is not entirely clear but is possibly related to increased level of severity of illness among patients receiving carbapenem therapy. More patients with septic shock and pneumonia were included in the carbapenem group, and more community-acquired infections were in the cefepime group. Although unable to demonstrate an association between mortality and source of infection for patients receiving cefepime or levo oxacin, it is possible that this study was underpowered to identify an association. Although unable to demonstrate an association between increased MIC of cefepime and mortality among patients receiving cefepime monotherapy, it is possible that the results might have been diluted due to the inability to detect de nitive MIC values.
Our study had a few strengths. First, this research was speci cally looking at clinical outcomes for the hospital and ICU. Additionally, this study allowed for identi cation of multiple Enterobacteriaceae organisms.
is study also allowed for several types of infection. Up until this point, previous literature had been limited to speci c types of Enterobacteriaceae (e.g., only E. coli) and had only studied outcomes in cases of bacteremia or UTI.
is study did have several limitations. First, there was a low sample size for the cefepime and levo oxacin groups compared to the carbapenem group. is is likely due to the physician's preference for the carbapenem group. Carbapenem therapy is used in this healthcare setting predominantly for ESBL-producing Enterobacteriaceae. Additionally, it was challenging to determine the exact MIC, as the MIC reported used automated log 2 dilution. In previous literature, an MIC of 1 or less proved to have a positive association between cefepime therapy and mortality. If this study had been able to detect an exact MIC of 1, there might have been a better detection for outcomes. Finally, there potentially was a physician selection bias for cefepime therapy in patients receiving treatment for community-acquired ESBLproducing infections.

Conclusion
Results from this study suggest that either cefepime or levo oxacin can be a potential alternative agent for infections with ESBL-producing Enterobacteriaceae for hospitalized patients when isolates are susceptible. Further studies with larger sample size and well-balanced type of infections between groups are warranted.

Conflicts of Interest
e authors declare that there are no con icts of interest regarding the publication of this article.