Appropriate treatment for bloodstream infections due to carbapenem-resistant Klebsiella pneumoniae and Escherichia coli: A nationwide multicenter study in Taiwan

ABSTRACT
Background. We aimed to investigate the outcome of patients who received different antimicrobialtherapy in carbapenem-resistant Enterobacteriaceae bloodstream infections with a new definition oftigecycline use in a multicenter study from Taiwan.Methods. Patients receiving appropriate therapy for bloodstream infections due tocarbapenem-resistant Klebsiella pneumoniae and Escherichia coli from 16 hospitals in Taiwan wereenrolled between January, 2012 and June, 2015. A multivariate analysis using cox proportionalregression model was performed to identify independent risk factors of 14-day mortality. Tigecyclinewas defined as appropriate when the isolates had MIC ≤ 0.5 mg/L and we investigated whethertigecycline was associated with mortality among patients with monotherapy.Results. A total of 64 cases with carbapenem-resistant K. pneumoniae (n=50) and E. coli (n=14)bloodstream infections were analyzed. Of the 64 isolates, 17 (26.6%) had genes that encodedcarbapenemases. The 14-day mortality of these cases was 31.3 %. In the multivariate analysis,Charlson Comorbidity Index (HR: 1.21, 95% CI 1.03-1.42, P = .022) and colistin monotherapy (HR:5.57, 95% CI 2.13-14.61, P < .001) were independently associated with 14-day mortality. Among the55 patients with monotherapy, the 14-day mortality was 30.9% (n=17). Tigecycline use was notassociated with mortality in the multivariate analysis.Conclusion. Tigecycline monotherapy was a choice if the strains exhibited MIC ≤ 0.5 mg/L, andcolistin monotherapy was not suitable. Our findings can incite more clinical studies regarding theefficacy of tigecycline in carbapenem-resistant Enterobacteriaceae infections.

Introduction
The rapid spread of carbapenem-resistant (non-susceptible) Enterobacteriaceae has become a greatchallenge for physicians [1-3]. Clinical studies have demonstrated a high mortality rate among patientswith carbapenem-resistant Enterobacteriaceae infection [1-3]. An optimal antimicrobial regimen isimportant in the treatment of carbapenem-resistant Enterobacteriaceae infection. Tigecycline andcolistin are considered as the last resort for these infections [1-3].In previous studies of carbapenem-resistant Enterobacteriaceae bloodstream infections [4-14], anappropriate antimicrobial regimen was defined as at least one in vitro active agent according tobreakpoints established by the Clinical and Laboratory Standards Institute (CLSI) [15] or theEuropean Committee on Antimicrobial Susceptibility Testing (EUCAST) [16]. However, the CLSI didnot issue interpretative criteria for tigecycline susceptibility. The EUCAST recommends tigecyclinesusceptibility breakpoints in Enterobacteriaceae of susceptible MIC ≤ 1 mg/L and resistant MIC > 2mg/L [16]. In addition, most of the studies used the interpretative criteria from the US Food and DrugAdministration (FDA) for tigecycline (susceptible, MIC ≤ 2 mg/L; intermediate, MIC = 4 mg/L;resistant, MIC > 4 mg/L) [17].

Therefore, defining an appropriate therapy with tigecycline againstEnterobacteriaceae is challenging. Moreover, the steady-state maximum serum concentrations oftigecycline (0.6 mg/L) [18] were lower than the current breakpoints proposed by the EUCAST orFDA. However, the issue of low serum concentration had not been addressed in the abovementionedstudies [4-14], and the efficacy of tigecycline in the treatment of bacteremia is also debated in theliterature [19].In this multicenter study, we investigated the independent risk factors for mortality amongpatients with bloodstream infections caused by carbapenem-resistant Klebsiella pneumoniae andEscherichia coli. We proposed a new definition of appropriate antimicrobial therapy with tigecyclinein these infections. We aimed to investigate the impact of different regimens of antimicrobial therapy,especially tigecycline, on mortality in patients who received appropriate therapy in these infections.Patients with at least one positive blood culture were defined as bloodstream infections. Bloodstreaminfections caused by carbapenem-resistant K. pneumoniae and E. coli were identified from 16hospitals (12 medical centers and 4 regional hospitals; Supplementary Data) in Taiwan betweenJanuary 1, 2012 and June 30, 2015. Carbapenem resistance was defined as non-susceptibility toimipenem or meropenem (MIC ≥ 2 mg/L) based on the interpretative criteria from CLSI published in2012 [15].

Only the first episode of bloodstream infections was included for each patient. The clinicaldata were retrospectively collected, and patients aged < 20 years, polymicrobial infections, and thosewith incomplete medical records were excluded. Patients who did not receive at least 48 h of at leastone appropriate antibiotic were also excluded. The detailed information of appropriate therapy wasdescribed in the following section. The study protocol was approved by the institutional review boardof each participating hospital.Carbapenem-resistant K. pneumoniae and E. coli isolates were collected from blood culture in themicrobiological laboratories of each participating hospitals. The isolates were sent to the NationalHealth Research Institutes (Miaoli, Taiwan) and were stored at −70°C in 10% glycerol Luria–Bertanimedium prior to analysis. Bacterial identification was performed by a VITEK 2 automated system(bioMérieux, Marcy l’Etoile, France). MICs were determined by broth microdilution (Sensititre; TrekDiagnostic Systems, Cleveland, OH, USA) for all antibiotics except tigecycline. The MICs fortigecycline were determined using the Etest (bioMe´rieux) on Mueller–Hinton medium. The resultswere interpreted according to the breakpoints published by CLSI, except those for colistin andtigecycline. Colistin was interpreted according to breakpoints established by EUCAST, andtigecycline was interpreted according to breakpoints established by the US FDA.Carbapenem-resistant K. pneumoniae and E. coli isolates were screened for carbapenemase genes,plasmid-borne AmpC-like genes, and ESBL genes using polymerase chain reaction detection asdescribed previously [20-23].

Bacterial outer membrane proteins (OMPs) were isolated and the OMPprofiles (OmpK35 and OmpK36 for K. pneumoniae, and OmpC and OmpF for E. coli) were identifiedby sodium dodecyl sulphate-polyacrylamide gel electrophoresis (SDS-PAGE) followed by coomassieblue staining (Bio-Rad). K. pneumoniae ATCC 13883 and E. coli ATCC25922 were used as thecontrol strains [24, 25].The probable source of bloodstream infections, including pneumonia, urinary tract infection, surgicalsite infection, intra-abdominal infection, catheter-related infection, or primary bacteremia, wasdetermined on the basis of microbiological results and physicians’ findings. The definition ofhealthcare-associated infection was described previously [26]. Severity of illness at the time of onsetof infection was assessed by the Acute Physiology and Chronic Health Evaluation II (APACHE II)score. Appropriate antimicrobial therapy, including carbapenems, in patients enrolled in this study wasdefined as treatment with at least one agent to which the isolate was susceptible in vitro according toEUCAST breakpoints [16]. For tigecycline, target values of AUC:MIC ≥ 6.96 have been reported tobe predictive of clinical response [27]. The steady-state maximum serum concentration of tigecyclinewas 0.6 mg/L [28]. To achieve the target values with a standard dose (100-mg loading dose followedby 50 mg twice daily), > 90% probability of target attainment could be expected at tigecycline MIC ≤0.5 mg/L [28]. Therefore, we defined appropriate antimicrobial therapy with tigecycline when thestrains exhibited MIC ≤ 0.5 mg/L. Appropriate antimicrobial therapy with colistin was defined as anisolate being susceptible in vitro according to the EUCAST breakpoint (MIC ≤ 2 mg/L) [16].

Antimicrobial therapy in these patients usually varied, making it hard to classify them to a specificregimen; therefore, patients were assigned to a regimen only if it was initiated during the first 5 daysafter the sampling of blood culture and maintained for at least 48 h [29]. Appropriate combinationtherapy was defined as two or more appropriate antibiotics administrated simultaneously for > 48 h.Predictors of mortality and treatment regimensThe primary outcome was death within 14 days from the onset of bloodstream infection. Risk factorsfor mortality in patients with bloodstream infections due to carbapenem-resistant K. pneumoniae andE. coli were investigated by comparing clinical variables of survivor and non-survivor subgroups.The therapeutic regimens for patients with carbapenem-resistant K. pneumoniae and E. colibloodstream infections were selected at the discretion of the attending physicians. There was nostandard hospital guideline for antimicrobial therapy in carbapenem-resistant K. pneumoniae and E.coli infections.

The recommended total daily dose of colistin was usually 9 million IU given in two orthree divided dosages, and for tigecycline the total daily dose was 100 administered in two divideddosages. Usual doses of carbapenems were used: 500 mg for imipenem every 6 h, 500 mg fordoripenem and 1 g for meropenem every 8 h. Dosages were adjusted to creatinine clearance whenindicated.Categorical variables were evaluated with the χ2 or two-tailed Fisher’s exact test. Continuous variableswere compared with the Student t test (for normally distributed variables) or the Mann–Whitney U test(for non-normally distributed variables). We used cox proportional regression model to identifyindependent predictors of mortality. All biologically plausible variables with P< .20 in univariatetesting were incorporated into the model using a backward approach. Hazard ratio (HR) and 95%confidence interval (CI) were calculated. Two-tailed tests were used to determine statisticalsignificance and P < .05 was considered significant. Sensitivity analysis was performed as well amongpatients received monotherapy only. All statistical analyses were performed using SPSS version 17(SPSS Inc., Chicago, IL, USA).

Results
A total of 125 cases with bloodstream infections caused by carbapenem-resistant K. pneumoniae andE. coli were identified during the study period. Sixty-one cases were excluded because ofpolymicrobial infection (n=39), mortality within 48 h (n=14), inappropriate therapy (n=8). Finally, 64cases were analyzed in this study. K. pneumoniae accounted for the majority of infections (n=50,78.1%). The demographic and clinical characteristics of the patients are shown in Table 1. The ages ofthe patients ranged from 20 to 94 years, with a median age of 71 years, and 39 patients were male. The14-day mortality rate was 31.3% and the overall in-hospital mortality rate was 53.1%.Microbiological characteristics of carbapenem-resistant K. pneumoniae and E. coli isolatesOf the 64 isolates, 17 (26.6%) had genes that encoded carbapenemases, including KPC-2 (n = 11),IMP-8 (n = 1), VIM-1 (n = 2), OXA-48 (n = 1), NDM-1 (n = 1), and one with both KPC-2 and IMP-8.Almost all the carbapenemase-producing strains were K. pneumoniae, and only one E. coli withcarbapenemase (NDM-1) was identified. Other non-carbapenemase producing strains had genes thatencoded plasmid-borne AmpC/ESBL and the loss of outer membrane porins (supplementary data). Allthe strains were resistant to ceftriaxone or ceftazidime. The MIC of tigecycline was ≤ 0.5 mg/L for 46(71.9%), 1-2 mg/L for 16 (25%), > 2 mg/L for 2 (3.1%).

The MIC of imipenem or meropenem was ≥8 mg/L for 48 (75%) isolates, and 4 mg/L for nine isolates (6.3%).Table 2 showed the risk factors for 14-day mortality among patients with carbapenem-resistant K.pneumoniae and E. coli bloodstream infections. Chronic respiratory failure with mechanical ventilator,malignancy, Charlson Comorbidity Index, surgical drain, septic shock, APACHE II score, colistinmonotherapy, and monotherapy other than colistin or tigecycline were incorporated into multivariatecox regression. Combination therapy was not associated with 14-day survival in the univariateanalysis. In multivariate analysis, Charlson Comorbidity Index (HR: 1.21, 95% CI 1.03-1.42, P =.022) and colistin monotherapy (HR: 5.57, 95% CI 2.13-14.61, P < .001) were independentlyassociated with 14-day mortality.We further evaluated the risk factors for 28-day mortality among the 64 patients. The result wassimilar to that in the analysis for 14-day mortality. Charlson Comorbidity Index (HR: 1.21, 95% CI1.04-1.40, P = .011) and colistin monotherapy (HR: 5.31, 95% CI 2.24-12.6, P < .001) were stillindependently associated with 28-day mortality.Monotherapy with tigecycline in carbapenem-resistant K. pneumoniae and E. coli bloodstreaminfectionsTable 3 showed detailed antimicrobial regimens among the 64 patients. We compared the 14-daymortality among the 64 patients who received different antimicrobial regimen (Figure 1).

Most of thecases (n=55) in the current study received monotherapy. Colistin monotherapy was associated with ahigher mortality than that in tigecycline monotherapy (57.1% versus 18.2%, P = .035) andmonotherapy other than colistin or tigecycline (57.1% versus 13.0%, P = .002). No mortalitydifference was noted among other regimen comparison. We defined appropriate therapy withtigecycline when the strains exhibited MIC ≤ 0.5 mg/L in the current study. We further compared the11 patients with tigecycline monotherapy and those with other monotherapy (n=44). The clinicalcharacteristics and 14-day mortality were not different significantly between these two groups (Table4). Figure 2 showed no significant difference in mortality between the two groups in the survival curve(Log-Rank test, P=.392). Among the 55 patients with monotherapy, the 14-day mortality was 30.9%(n=17). Tigecycline was not associated with survival benefit independently in the multivariate coxregression model (data not shown).

Discussion
In this study, we investigated patients with bloodstream infections caused by carbapenem-resistant K.pnuemoniae and E. coli who received appropriate antimicrobial therapy. We proposed a newdefinition of appropriate antimicrobial therapy with tigecycline. We found that colistin monotherapyand Charlson Comorbidity Index were the independent predictors for 14-day mortality. Tigecycline monotherapy was not associated with a higher mortality among patients with monotherapy.In Taiwan, the major mechanism for carbapenem-resistant Enterobacteriaceae was not mediatedby carbapenemase till 2015 [20-23]. We included both carbapenemase-producing and noncarbapenemase-producing strains in this study, aiming to generalize our findings in the real-lifepractice. Many microbiology laboratories do not perform the molecular detection of carbapenemaseroutinely, and the physicians usually treat these infections according to the MICs interpreted by theCLSI or EUCAST. Therefore, the current study is able to help physicians manage bloodstreaminfections caused by carbapenem-resistant K. pnuemoniae and E. coli according to the MICs,regardless of the mechanisms of carbapenem resistance.Most studies regarding treatment for carbapenem-resistant Enterobacteriaceae bloodstreaminfections [4-10,12-14] defined appropriate use of tigecycline according to the US FDA criteria(susceptibile, MIC ≤ 2 mg/L). Tumbarello et al. used EUCAST criteria to define the appropriate use oftigecycline (susceptibile, MIC ≤ 1 mg/L) in KPC producing K. pneumoniae bacteremia [11]. Wedefined tigecycline use in bloodstream infection as appropriate only when the isolates exhibited MIC≤ 0.5 mg/L, based on previous pharmacodynamics study [27,28].

With tigecycline MIC ≤ 0.5 mg/L, a> 90% probability of target attainment could be expected, and the cumulative fraction of response was82.0%, based on previous pharmacokinetics and EUCAST wild-type MIC distributions of K.pneumoniae [28]. Tigecycline is considered as a potent therapeutic option for carbapenem-resistantEnterobacteriaceae infections [30], but the efficacy of tigecycline cannot be clearly defined becausethe lacking of suitable definition of appropriate therapy. One recent meta-analysis showed that theefficacy of tigecycline in treating carbapenem-resistant Enterobacteriaceae infections is similar to thatof other antibiotics [31], but the issue of suboptimal concentrations of tigecycline was still notdiscussed [31]. In the literature, cases treated with tigecycline monotherapy in bloodstream infectionsdue to carbapenem-resistant K. pneumoniae are limited. The current study firstly adopted a newdefinition of appropriate tigecycline in bloodstream infections caused by carbapenem-resistant K.pneumoniae and E. coli taking into account the low serum level of tigecycline. Our study found thattigecycline monotherapy was not associated with 14-day mortality if the strains exhibited MIC ≤ 0.5mg/L. Our result provided some insight on the issue of tigecycline treatment in carbapenem-resistantEnterobacteriaceae bacteraemia and might incite the further prospective studies to solidify thefindings.We found that colistin monotherapy was associated with poor outcome independently in thisstudy.

We cannot demonstrate whether colistin-based combination therapy is better because of thelimited cases with combination therapy in our analysis. One recent study conducted by de Oliveira etal. demonstrated that polymyxins was associated with a higher risk of mortality in KPC-producingEnterobacteriaceae infections and dosage was the major concern [32]. The dosing guidance of cilistinis necessary because an extensive interpatient variability in pharmacokinetics [33]. One recent studyproposed clinician-friendly dosing algorithms and suggested that monotherapy with intravenouscolistin may be suboptimal [33]. The difficulty to select an optimal dose in bloodstream infectionsmay be the reason for the low efficacy of colistin in our study. We also identified that CharlsonComorbidity Index influenced the outcome in this study, which emphasized the role of host factors incombating carbapenem-resistant bacteria.The major limitation of this study was that clinical data were obtained retrospectively frommedical records. Several missing variables, such as source control, might have potential effects onoutcome. The other limitation was the limited number of cases in the current study, especially patientswith combination therapy, which precluded further analysis. Finally, the limited cases treated withtigecycline monotherapy is the further limitation of this analysis. Nevertheless, our study firstlyprovided the new definition of appropriate tigecycline monotherapy in these serious infections.

In conclusion, we used a new definition of appropriate antimicrobial therapy with tigecycline in the treatment of bloodstream infections caused by carbapenem-resistant K. pneumoniae and E. coli. Our findings suggested that tigecycline monotherapy therapy was an appropriate choice if the strains exhibited MIC ≤ 0.5 mg/L, and colistin monotherapy is not suitable. In the era of limited new drugs, our findings can incite more clinical studies regarding the efficacy of tigecycline in carbapenem-resistant Enterobacteriaceae infections.