Tigecycline Therapy for Infections Caused by Extended-Spectrum β-Lactamase-Producing Enterobacteriaceae in Critically Ill Patients

We aimed to evaluate tigecycline on the clinical effectiveness in treating complicated skin and soft tissue infections (cSSTI), complicated intra-abdominal infections (cIAI), and pneumonia, caused by extended-spectrum β-lactamase (ESBL)-producing Enterobacteriaceae, as data are limited. From three medical centers in Taiwan, we retrospectively studied the cSSTI, cIAI, and/or pneumonia caused by ESBL-producing Enterobacteriaceae. Among the 71 patients, including 39 patients infected with Klebsiella pneumoniae, 30 infected with Escherichia coli and others, the clinical success rate of tigecycline-based therapy was 80–90% for pneumonia and cSSTI caused by E. coli and 50–60% for cIAI caused by K. pneumoniae and E. coli. Microbiological and clinical outcome of pneumonia caused by carbapenem-resistant K. pneumoniae was poor. Univariate Cox analysis showed that dyspnea, SOFA score, septic shock, thrombocytopenia, prolonged prothrombin time, and lesser microbiological eradication were significant factors associated with 30-day mortality after the end of therapy. Cox regression proportional hazards model revealed dyspnea and a SOFA score > 8 to be independently associated with time to death. For ESBL producers, tigecycline showed good effects for cSSTI and pneumonia by E. coli, ordinary for cIAI, but ineffective for pneumonia by K. pneumoniae. Dyspnea and a high SOFA score predict a poor outcome.


Introduction
Tigecycline is a glycylcycline antibiotic with a broad-spectrum antimicrobial activity against multidrug-resistant organisms (MDROs), such as methicillin-resistant Staphylococcus aureus, extended-spectrum β-lactamase (ESBL)-producing Enterobacteriaceae, Acinetobacter baumannii groups Cure was defined as the resolution of clinical signs and symptoms of infections compared with baseline, without a requirement for additional antibacterial treatment. Improvement was defined as the partial resolution of clinical signs and symptoms, with a scope of additional antibacterial treatment. Failure was defined as adding or switching to new antibiotics due to poor response, persistence, worsening signs and symptoms, unrelenting signs and symptoms, or persistent isolation of pathogens from the infection sites. Indeterminate was defined as the outcome that could not be assessed for any reason or adding/switching to new antibiotics due to emerging infections (other than initial infection sites) [5,6,15,27].

Secondary Clinical Outcomes by 30 Days after EOT
Mortality referred to overall death and was assessed on the 30th day after the end of therapy (EOT). Clinical success was stringently defined survival without readmission between EOT and the 30th day after EOT. Readmission could be a potential outcome confounder [28,29].

Microbiological Outcomes
Follow-up cultures of sampling from primary sites of infections were taken based on routine clinical practice. Eradication was defined as a documented disappearance of baseline infecting pathogens at primary sites of infections from the follow-up cultures. Persistence was defined as documented positive follow-up cultures from primary infection sites. If follow-up cultures were not available, it could be arbitrarily presumed to be eradication based on clinical success (cure or improvement) or Antibiotics 2020, 9,231 4 of 17 to be persistence through clinical failure [6]. Superinfection was defined as the emergence of a new isolate at the primary site of infection with worsening signs and symptoms of infection.

Data Collection
The collected data included demography, comorbidity, dates of admission and onset, source of infection, infection site, symptoms, severity status including the worst physiological variables of the baseline Acute Physiology and Chronic Health Evaluation (APACHE) II score within 48 h of intensive care unit (ICU) admission and real-time Sequential Organ Failure Assessment (SOFA) score during the first 48 h after infection onset, infectious complication, pathogen, empirical antimicrobial therapy, the date of end of therapy, outcome assessment, and the days of hospital stay after the onset of infection. Infections were classified as nosocomial vs. community-acquired depending on the onset of the infection after or before 48 h of hospitalization. Inadequate empirical antibiotic therapy was defined to be greater than 48 h between the time of culture and initiation of treatment using an antimicrobial for which the target pathogen was ultimately determined to be resistant in vitro [15].

Statistical Analyses
The baseline characteristics of selected patients from each hospital were analyzed using Pearson's Chi-square test or Fisher's exact test for categorical variables and analysis of variance (ANOVA) or Student's t-test for continuous variables. The Kaplan-Meier plot was employed to present the trend on 30-days mortality after infection onset, and the trend differences between the three hospitals were compared using the long-rank test.
The Cox proportional hazard regression analysis through three steps was used to estimate the relative 30-day mortality risk after EOT. Firstly, the above-analyzed significant baseline characteristics, including symptoms, comorbidity, disease severity, and complications, and other potentially confounding variables were further screened by the univariate Cox proportional hazards regression analysis. Secondly, all variables with p-values < 0.05 from the univariate analysis were included in the first multivariable prediction model. Thirdly, the second multivariable prediction model used Collett's model selection approach to find further predictors. This approach used a significance level of 0.10 for the univariate screening, after which the predictors with entry and stay criteria of 0.05 were selected in the final model. All statistical analyses were performed using SAS (version 9.4 for Windows, SAS Institute Inc., Cary, NC, USA). Significance was set at p-value < 0.05 (two-tailed).

Ethics Approval
The study was approved by local ethics committees, and was exempted from the requirement for informed consent by the Institutional Review Board (IRB) of the National Taiwan University Hospital (IRB no. 201710028RINB), Taipei; the IRB of Chi Mei Medical Center (IRB no. 10511-002) and the IRB of National Cheng Kung University Hospital (IRB no. A-ER-106-221), Tainan, Taiwan.

Results
A total of 71 cases that received tigecycline-based therapy for ESBL-producing Enterobacteriaceae with or without carbapenem resistance were enrolled in the study. Among them, 16 were from NTUH, 17 from CMMC, and 38 from NCKU ( Table 1). The infections were predominantly acquired at hospitals (90.1%). CMMC had more community-acquired cases (35.3%, p < 0.0001) and the lowest SOFA score (mean, 6.2, p = 0.033). The patients from NTUH had higher rates of post-operation status, source control, carbapenem resistance, septic shock, prolonged prothrombin time, and acute respiratory distress syndrome (ARDS) with statistical significance.

Patient Characteristics
The proportion of female patients in CMMC was significantly lower than others (p = 0.04). The patients from NTUH were significantly older in age (mean age, 72.1 years) than those from the other two hospitals (p = 0.02). The comorbidities of patients from NTUH involved significantly more respiratory disease, and solid cancer, with a higher Charlson score (mean, 9.5) than other hospitals. Other underlying diseases did not show significant differences among three institutes. No hematological malignancy or other immunocompromised hosts were found in the study. Most patients had multiple comorbidities with a mean Charlson score >5, most commonly cardiovascular disease (64.8%), metabolic diseases (57.8%), and solid cancer (42.3%). The patients from NTUH had a significantly higher rate of respiratory symptoms (dyspnea and cough) at the onset of infection. Most subjects were critically ill with a mean APACHE II score of more than 20, with 77.5% in the ICUs and 62.0% using mechanical ventilator (Table 1).
Nearly all patients received combination therapy, mostly with carbapenems (meropenem, imipenem, or doripenem), followed by cefepime, ceftazidime, piperacillin-tazobactam, cefpirome, piperacillin, and rarely, ciprofloxacin or levofloxacin. The rate of inappropriate empirical antimicrobial therapy was highly distributed among three hospitals without any significant difference (mean, 84.5%, Table 1). The remarkably high rate of inadequate empirical antibiotic therapy in three centers was due to a high proportion of carbapenem resistance and the unexpected presence of ESBL.
The outcome revealed that the patients from NTUH had significantly lower rates of clinical success (43.8%) and microbiological eradication (0%) at EOT, accompanied by the lowest survival rate by 30 days after EOT (37.5%). However, survival duration analysis for the 30-day mortality of the population among the three centers using the Kaplan-Meier technique showed no noteworthy difference between each of the paired comparisons ( Figure 1).

presence of ESBL.
The outcome revealed that the patients from NTUH had significantly lower rates of clinical success (43.8%) and microbiological eradication (0%) at EOT, accompanied by the lowest survival rate by 30 days after EOT (37.5%). However, survival duration analysis for the 30-day mortality of the population among the three centers using the Kaplan-Meier technique showed no noteworthy difference between each of the paired comparisons (Figure 1). Overall, the rates of clinical success and microbiological eradication at EOT were 63.4% and 50.7%, respectively. The overall mortality by 30 days after EOT was 36.6%. The mean length of stay at the hospital after the infection was 44.7 days, which was not significantly variable in any of the three hospitals (Table 1).
During the study period, the tigecycline susceptibility test was optional upon a physician request. We only can enroll tigecycline-susceptible cases (probably the physician would use tigecycline based on susceptibility), but persistent isolates at the end of therapy usually lacked tigecycline susceptibility (probably the physician switched to other drugs and did not request tigecycline testing). We could not say the development of tigecycline resistance explaining the cases with failure of microbiological eradication. From very limited cases with persistent isolates doing tigecycline susceptibility testing, tigecycline remained susceptible at the end of therapy. Overall, the rates of clinical success and microbiological eradication at EOT were 63.4% and 50.7%, respectively. The overall mortality by 30 days after EOT was 36.6%. The mean length of stay at the hospital after the infection was 44.7 days, which was not significantly variable in any of the three hospitals (Table 1).
During the study period, the tigecycline susceptibility test was optional upon a physician request. We only can enroll tigecycline-susceptible cases (probably the physician would use tigecycline based on susceptibility), but persistent isolates at the end of therapy usually lacked tigecycline susceptibility (probably the physician switched to other drugs and did not request tigecycline testing). We could not say the development of tigecycline resistance explaining the cases with failure of microbiological eradication. From very limited cases with persistent isolates doing tigecycline susceptibility testing, tigecycline remained susceptible at the end of therapy.
The clinical outcomes of tigecycline-based treatment per patient subgroup by infection sites are shown in Table 2. The cSSTI, cIAI, and pneumonia were successfully treated in 61.5%, 60.7%, and 52.2% patients at EOT respectively. We did not evaluate the detailed outcomes of the UTI, because tigecycline was not primarily used to treat UTI, and 11 of 14 UTIs (78.6%) were accompanied infections with either cSSTI, cIAI, and/or pneumonia. The UTI and multiple infections were successfully treated in 57.1% and 44.4% patients at EOT respectively.
For cSSTI, the clinical and microbiological outcomes at EOT and survival 30 days post EOT, the ESBL-CS group had apparently higher clinical success rates (about 60-70%) than ESBL-CR group (about 20-50%), which were statistically insignificant. Meanwhile, the E. coli group had a better clinical success than the K. pneumoniae group at EOT (90% vs. 50%, p = 0.08) and by 30 days after EOT (90% vs. 36%, p = 0.01).
For cIAI, the clinical and microbiological outcomes at EOT and 30 days after EOT, either by E. coli or K. pneumoniae (ESBL-CS or ESBL-CR), were about 40-60% successful, holding no significance.

Risk Factors for Poor Clinical Outcome
The characteristic of the subjects with clinical success (survival without readmission) or mortality by 30 days after EOT was evaluated using univariate analysis ( Table 3). The statistically significant predictors for mortality were found to be dyspnea, severity status (stay in ICU, ventilator use and SOFA score), and infectious complications (septic shock, acute renal failure, thrombocytopenia, prolonged prothrombin time, and neurologic complications). The clinical and microbiological outcomes at EOT were also correlated with clinical success by 30 days after EOT using both models (p < 0.0001). The mean length of stay after the infection was longer in the survival group than that in the mortality group (52.1 vs. 32.9, p = 0.03). The above-mentioned significant patient characteristics, microbiological eradication and other potential confounders, such as hospitals, Charlson score, carbapenem resistance, antibiotic therapy, and source control were further forwarded to the Cox proportional hazards regression analysis.   Table 4 shows independent factors associating with mortality by using Cox proportional hazards regression model. Univariate analysis revealed that dyspnea (p = 0.003), SOFA score (p = 0.0004), SOFA score >8 (p = 0.001), septic shock (p = 0.022), thrombocytopenia (p = 0.0004), prolonged prothrombin time (p = 0.018), and lack of microbiological eradication ability (p = 0.004) were considerably associated with mortality by 30 days after EOT.

Hazard Ratio for Outcome
Cox regression proportional hazard model revealed that only dyspnea (p = 0.008-0.009) and SOFA score > 8 (p = 0.041) were associated with the time to death. The prolonged prothrombin time and lack of microbiological eradication were found to be inconsistently associated with mortality when different approaching models were used. The Collett's model selection approach revealed that only dyspnea (p = 0.0002) and SOFA > 8 score (p = 0.004) significantly predicted mortality, with nearly 7.3-fold and 4.6-fold increased odds of dying respectively. Therefore, no specific factors of carbapenem resistance, empirical antimicrobial therapy, source control, ventilator use, ICU stay, or hospitals were associated with time to mortality (Table 4), even though NTUH had a relatively higher number of cases with source control (Table 1).
The effect of tigecycline-based therapy was lower than a previous report demonstrating a high overall clinical success rate (91.0%, 61/67) for ESBL-related infections [5]. The difference in the clinical success rate between our study and the previous report was statistically significant (63.4% vs. 91.0%, p = 0.0001). In another large series of cIAI patients, treatment was successful in 426 (70.3%) of 606 tigecycline-treated patients [28]. The difficulty of our study in attaining a high success rate could perhaps be justified by the data of K. pneumoniae (ESBL-CR) and a high proportion (80.5%) with inappropriate empirical antimicrobial therapy.
Carbapenem resistance is more common among infections caused by the Enterobacteriaceae and was associated with a significantly higher risk of mortality in comparison to carbapenem-susceptible Enterobacteriaceae [31].
In our previous study, we demonstrated that tigecycline activity was unaffected by ESBL production and had an MIC 90 ranging from 0.5 to 1.5 µg/mL for ESBL-producing K. pneumoniae, Serratia marcescens, and E. cloacae [32]. In a total of 2870 blood samples in India, all ESBL producers were sensitive to tigecycline [33]. In a small collection of serious infections by K. pneumoniae, the cure rate for tigecycline in patients with multidrug-resistant K. pneumoniae was 5/6 (83.3%). All six patients achieved microbiological eradication [6]. Nonetheless, due to the high frequency of carbapenem resistance in infections of ESBL-producing Enterobacteriaceae, the treatment of choice for these ESBL-related infections should not always consider a carbapenem as an empirical regimen. The tigecycline-based therapy should be efficiently developed ahead to treat these infections empirically rather than just reserving it as a specific therapy at a later stage.
Bloodstream infections caused by ESBL-producing Enterobacteriaceae have been associated with higher mortality than those by non-ESBL producers [23,34]. This was also seen in our patients having pneumonia or cSSTI. The success rates seemed to be lower in patients with pneumonia caused by K. pneumoniae (50% of ESBL-CS and 30% of ESBL-CR groups), due to difficulty in the microbiological eradication of K. pneumoniae (ESBL-CR) in cases of pneumonia (0/10). Our results add evidence that tigecycline is not recommended to treat pneumonia by K. pneumoniae (ESBL-CR) strains.
A higher concentration of tigecycline or combination of tigecycline with colistin rendering good in vitro synergistic activity against ESBL-producing K. pneumoniae or with carbapenem resistance could be further reassessed for clinical success [35][36][37]. The combination of tigecycline with colistin, aminoglycoside or doxycycline might be active against the ESBL-CR infections in several studies in Taiwan [37][38][39].
Most of our patients (78%) were critically ill and stayed in the ICUs. Although the univariate analysis revealed several risk factors for mortality, including dyspnea, SOFA score > 8, septic shock, thrombocytopenia, prolonged prothrombin time, and inability towards microbiological eradication, the Cox regression proportional model demonstrated that only dyspnea and SOFA score or SOFA > 8 were independently associated with the estimated time towards death. The SOFA score has been a useful predictor of clinical outcome and an initial score of 8-11 was associated with a 60% ICU mortality [40]. About 82% of the patients with dyspnea used the respiratory ventilator. A further study may be required to investigate the relationship of outcome and various ventilator modes. Our study proposes that late-onset mortality could be predicted by the useful warning sign of dyspnea and a SOFA score, which represents real-time disease severity during the infection onset.

Conclusions
The limitations of this study included retrospective observational design, a lack of information about tigecycline side effects and drug interactions, a lack of information about development of tigecycline resistance, and the low number and its heterogeneity in demographics. However, the infection sites of cohort were almost evenly distributed. The clinical experience of tigecycline in the ESBL-related infections was limited in current practice. This multicenter study enrolling a sizeable sample of patients provides an evident basis of future generalization in the design of tigecycline-based strategies against the critically ill infections by ESBL producers. The tigecycline-based therapy for ESBL-producers was highly active (80-90% success rates) against pneumonia and cSSTI by E. coli, but showed moderate response (50-60% success rates) for cIAI caused by E. coli or K. pneumoniae. Microbiological eradication of K. pneumoniae with carbapenem resistance was difficult to achieve in pneumonia, for which tigecycline-based therapy should not be recommended. The relative effects of tigecycline might be cautiously integrated into ASP recommendation as an alternative regimen in future. Some variables could be useful to predict outcomes of critically ill patients. High SOFA score and dyspnea were independently associated with time till death even when the patients were put on ventilator. Funding: The work was supported by Pfizer Inc. with Project no. WI223052 (IIR2016-03). The funding source did not involve in the study design; in the collection, analyses, and interpretation of data.