Assessment of Mortality-Related Risk Factors and Effective Antimicrobial Regimens for Treatment of Bloodstream Infections Caused by Carbapenem-Resistant Enterobacterales

Bloodstream infections (BSIs) attributable to carbapenem-resistant Enterobacterales (CRE-BSIs) are dangerous and a major cause of mortality in clinical settings. This study was therefore designed to define risk factors linked to 30-day mortality in CRE-BSI patients and to examine the relative efficacies of different antimicrobial treatment regimens in affected individuals. ABSTRACT Bloodstream infections (BSIs) attributable to carbapenem-resistant Enterobacterales (CRE-BSIs) are dangerous and a major cause of mortality in clinical settings. This study was therefore designed to define risk factors linked to 30-day mortality in CRE-BSI patients and to examine the relative efficacies of different antimicrobial treatment regimens in affected individuals. Data pertaining to 187 CRE-BSI cases from four teaching hospitals in China collected between January 2018 and December 2020 were retrospectively analyzed. For the 187 patients analyzed in this study, the 30-day mortality of CRE-BSI was 41.7% (78/187). Multivariate logistic regression analyses revealed that Pitt bacteremia score, immunocompromised status, meropenem MIC of ≥8 mg/liter,absence of source control of infection, and appropriate empirical therapy were independent predictors of CRE-BSI patient 30-day mortality. After controlling for potential confounding factors relative to ceftazidime-avibactam (CAZ-AVI) treatment, combination therapies including CAZ-AVI (odds ratio [OR], 1.287; 95% confidence interval [CI], 0.124 to 13.403; P = 0.833) were not related to any significant change in patient mortality risk, whereas the 30-day mortality risk was higher for patients administered other antimicrobial regimens (OR, 12.407; 95% CI, 1.684 to 31.430; P = 0.011). When patients were treated with antimicrobial regimens not containing CAZ-AVI, combination therapy (OR, 0.239; 95% CI, 0.077 to 0.741; P = 0.013) was related to a decreased 30-day mortality risk relative to monotherapy treatment. The mortality-related risk factors and relative antimicrobial regimen efficacy data demonstrated in this study may guide the management of CRE-BSI patients.


RESULTS
Patient overview. In total, 226 hospitalized patients with blood cultures positive for CRE were screened for this study, with 187 nonduplicate patients ultimately being enrolled in our analysis ( Fig. 1), including 164, 21, 1, and 1 CRE-BSI cases caused by K. pneumoniae, E. coli, Enterobacter cloacae, and Citrobacter freundii, respectively.
Risk factors associated with CRE-BSI patient 30-day mortality. Relative to the surviving CRE-BSI patients, those that were deceased at the end of the 30-day period exhibited higher acute physiology and chronic health evaluation (APACHE II) scores (median scores, 13.0 versus 10.0; P , 0.001) and Pitt scores (median scores, 3.0 versus 1.5; P , 0.001) at the time of BSI. Deceased patients also exhibited higher incidences of immunocompromising conditions (42.3% versus 17.4%; P , 0.001), primary BSI (11.5% versus 2.8%; P = 0.016), and ICU hospitalization at time of BSI onset (42.3% versus 26.6%; P = 0.025). The number of days of appropriate antimicrobial treatment was also decreased for deceased patients relative to survivors (median numbers of days, 10.0 versus 12.0; P = 0.049) ( Table 1).
The impact of definitive antimicrobial treatment on CRE-BSI patient mortality. The dosages of definitive antimicrobials are shown in Table S4, and dosages were adjusted to creatinine clearance when indicated.
After controlling for Pitt score, meropenem MIC of $8 mg/liter, immunocompromised status, source control of infection, and the administration of appropriate empirical therapy, an additional multivariate backward stepwise logistic regression analysis revealed that definitive therapy with CAZ-AVI alone, CAZ-AVI plus tigecycline (OR, 1.645; 95% CI, 0.106 to 25 Fig. 2).
After controlling for potential confounding variables, definitive CAZ-AVI therapy (OR, 0.088; 95% CI, 0.020 to 0.379; P = 0.001) was associated with a lower risk of 30-day  34.151; P = 0.005) were associated with higher risks of clinical failure even after adjustment for Pitt score, meropenem MIC of $8 mg/liter, immunocompromised status, source control of infection, appropriate empirical therapy, possible source of BSI, and days of appropriate antimicrobial therapy (Table S6).
The impact of monotherapy and combination therapy regimens on CRE-BSI patient mortality. After adjusting for Pitt score, meropenem MIC of $8 mg/liter, immunocompromised status, source control of infection, and appropriate empirical therapy administration, an additional multivariate backward stepwise logistic regression analysis indicated that, compared with definitive CAZ-AVI monotherapy, treatment with combination therapies containing CAZ-AVI (OR, 1.287; 95% CI, 0.124 to 13.403; P = 0.833) was linked to comparable 30-day mortality risks among CRE-BSI patients. In contrast, when patients were administered definitive antimicrobial regimens not containing CAZ-AVI, combination therapy (OR, 0.240; 95% CI, 0.077 to 0.745; P = 0.014) was related to lower 30-day mortality risks than was monotherapy (Table 4).
Consistent with the above-described data, Cox regression survival curves indicated that CAZ-AVI monotherapy-treated-patient 30-day mortality rates were similar to those of patients treated with a combination of antimicrobial agents including CAZ-AVI, whereas the 30-day mortality for patients that underwent monotherapy treatment was significantly higher than that of patients administered combination therapies not containing CAZ-AVI (Fig. 3).

DISCUSSION
The present multicenter real-world study enabled us to successfully identify certain predictors of CRE-BSI patient mortality and to compare the relative efficacies of different antimicrobial regimens used to treat CRE-BSI patients in China. Together, our data highlight the most promising therapeutic options available at present to treat BSIs caused by CRE.
We found that CRE-BSI patients in the present study cohort exhibited a 30-day mortality rate of 41.7%, which was consistent with prior reported rates ranging from 30 to 80% (6)(7)(8). Much as with other infections, CRE-BSI infection outcomes are influenced by pathogen type, host factors, and medical interventions (14). In previous studies, patients infected with CRE isolates exhibiting a meropenem MIC of #8 mg/liter were found to have better survival when administered antimicrobial regimens containing a high-dose carbapenem, particularly when they underwent combination or prolonged- In addition to being a previously reported risk factor for CRE infection (18), immunocompromised status was identified as a risk factor associated with increased CRE-BSI patient mortality in our study, in line with the work of Gomez-Simmonds et al. (19). Even when effectively treated with antibiotics, immunocompromised patients suffer from infections more and exhibit higher death rates than immunocompetent individuals. As 95% of the cases in our study were of secondary BSIs, the source control of infections via abscess drainage, urinary catheterization, CVC removal or replacement, and related techniques was critical. Appropriate empirical therapy has previously been reported to be associated with better CRE-BSI patient prognosis, and prompt treatment within 48 h of BSI onset with such empirical antibiotics is agreed to be associated with better severe infection outcomes (8,20,21). We did not confirm whether BSI patients can benefit from treatment within a shorter duration, such as the 24-h period proposed by Falcone et al. (22), potentially due to the limited sample size in the present report. Future large-scale studies will be required to further examine the association between the timing of treatment and patient outcomes.
Due to its relatively recent introduction to the Chinese market, high price, and restricted clinical indications, only 35 patients in the present study were treated with CAZ-AVI. We found that patients treated with CAZ-AVI-containing regimens exhibited reduced 30-day mortality relative to patients treated without CAZ-AVI (17.1% versus 47.4%; P , 0.001), with comparable differences in rates of clinical failure (28.6% versus 68.4%; P , 0.001). Even after controlling for confounding variables, regimens containing CAZ-AVI were associated with increased survival rates and lower clinical failure rates, consistent with prior reports (23-25). van Duin et al. (24) previously examined the relative efficacies of CAZ-AVI and colistin for the treatment of CRE infections. After adjusting for the inverse probability of treatment weighting (IPTW), these authors found that the all-cause mortality in patients treated with CAZ-AVI (n = 38) was just 9%, as opposed to 32% for patients treated with colistin (n = 99), with CAZ-AVI-treated patients having IPTW-adjusted odds of good clinical outcomes at 30 days of 64% (95% CI, 57% to 71%). In a retrospective analysis conducted by Shields et al. (25), clinical success was observed to be more common among patients treated with CAZ-AVI-containing regimens (85% [11/13]) than among patients treated with other regimens (P = 0.006), including those composed of $2 agents exhibiting in vitro activity (44% [12/27]; P = 0.02), with CAZ-AVI administration being an independent predictor of successful clinical outcomes in their multivariate logistic regression analysis (OR, 8.64; 95% CI, 1.61 to 43.39; P , 0.01). In our study, all 35 isolates from cases treated with CAZ-AVI were susceptible to CAZ-AVI. The development of CAZ-AVI resistance among CRE during treatment has been reported previously (26,27). Potential impacts of this on clinical outcomes need to be further studied.
Prior reports indicate that combination therapies are associated with lower mortality rates than monotherapies when used to treat CRE infections (15,16,19,28). Indeed, one prior meta-analysis of 44 observational studies incorporating 3,195 patients with CRKP infections found monotherapy to be linked to a higher mortality risk than combination therapy (OR, 1.45; 95% CI, 1.18 to 1.78%) (29). Here, we found that combination therapy benefits were only evident for therapeutic regimens not containing CAZ-AVI, whereas no differences were observed in patient survival when comparing CAZ-AVI monotherapy and combination regimens incorporating CAZ-AVI. Caston et al. (23) and Tumbarello et al. (30) similarly found that CAZ-AVI-containing combination therapies were not linked to any improvements in the success of clinical treatment compared with CAZ-AVI monotherapy. As such, we posit that CAZ-AVI is a more promising therapeutic choice for treating patients infected with susceptible CRE isolates, even when used as a monotherapy.
We found that there were no significant differences among different antimicrobial regimens that did or did not contain carbapenems, polymyxins, tigecycline, or aminoglycosides with respect to CRE-BSI patient 30-day mortality. Using carbapenems to treat CRE infections remains a matter of controversy, with the relative benefits of including these antibiotics being dependent upon the carbapenem-specific MIC of the infecting pathogen (15,19,29,31). Most isolates in this study exhibited a high carbapenem MIC (.8 mg/liter), indicating that carbapenem-containing regimens were unlikely to be effective. While isolates herein exhibited high rates of susceptibility to tigecycline, polymyxins, and aminoglycosides in vitro, the treatment of patients with these agents largely failed to improve survival outcomes, likely due to their low concentrations in serum and sites of infection, such as the LRT, following administration (32,33). Tigecycline monotherapy has even been implicated in increased CRE-BSI patient mortality in prior studies (34). As polymyxins exhibit  a narrow therapeutic window and highly variable pharmacokinetics, their optimal dosage range also remains poorly understood (20). There are several potential limitations to this analysis. First, our study had a retrospective design and was thus susceptible to selection bias and recall bias. Second, other factors beyond the categories of antimicrobial agents utilized, such as their doses and the duration of effusion, can influence treatment efficacy. As our study was retrospective and sample subgroups were limited, we were not able to analyze these data. The punctual meropenem MIC value may be an important tool for understanding the response or absence of response to therapy. However, it was possible to measure the punctual MIC value. Additional analyses of drug resistance genes in the bacteria isolated from these patients would also be of value as a means of further exploring drug resistance-related characteristics and related treatments.
In summary, the analysis of mortality-related risk factors and antimicrobial agent efficacies in CRE-BSI patients reported herein has the potential to guide the management of these critically ill patients. However, future prospective cohort studies or randomized trials are urgently needed to validate and expand upon the findings of the present study.

MATERIALS AND METHODS
Study design and patient selection. Medical records for hospitalized patients at four teaching hospitals in China (Table S1) with positive CRE blood cultures between 1 January 2018 and 31 December 2020 were reviewed retrospectively. Patients were excluded if they (i) were ,14 years old, (ii) exhibited polymicrobial bacteremia, (iii) did not have medical records that were fully available, (iv) did not exhibit clinical manifestations consistent with bacteremia, or (v) experienced more than one CRE-BSI, in which case only the first positive blood culture report was included.
The primary study outcome was 30-day mortality following BSI onset, while the secondary outcome was clinical failure, including (i) death, (ii) symptom persistence or evidence of infection at day 30, and (iii) the recurrence of symptoms or evidence of infection after treatment ended.
The study design was approved by the Ethics Committee of Beijing Jishuitan Hospital (no. 201911-15). Given the retrospective nature of the study, the Ethics Committee determined that informed consent was not necessary.
Microbiology. The Vitek 2 system (bioMérieux, Marcy l'Etoile, France) or matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry (MALDI Biotyper [Bruker Daltonics GmbH, Leipzig, Germany] or Vitek-MS [bioMérieux]) was employed for isolate identification. Testing for antibiotic susceptibility was conducted according to the standard protocols of each hospital, with most utilizing the Vitek 2 system or a broth microdilution (BMD) method. MICs for tigecycline, colistin, and CAZ-AVI were determined via standard BMD and were interpreted according to Clinical and Laboratory Standards Institute (CLSI) breakpoints. In accordance with the 2018 CLSI guidelines, carbapenem resistance was defined by a MIC of $2 mg/ml for ertapenem or $4 mg/ml for meropenem or imipenem (35). The tigecycline ($8 mg/ml) and colistin (.2 mg/ml) breakpoints were defined according to the U.S. FDA (36) and European Committee on Antibiotic Susceptibility Testing guidelines (37), respectively.
Study definitions. Bacteremia was defined by the detection of a minimum of one blood culture positive for a known pathogen that coincided with consistent clinical features. Immunocompromised status included primary immune deficiency diseases, active malignancy, human immunodeficiency virus infection with a CD4 T-lymphocyte count of ,200 cells/ml or percentage of ,14%, immunosuppressive therapy, solid organ transplantation, hematopoietic stem cell transplantation, and splenectomy. Antimicrobial drug exposure was defined as utilizing any antibiotics for .72 h within 30 days prior to CRE-BSI diagnosis. Empirical therapy was defined by the administration of antimicrobial agents before blood culture reports were available, while definitive therapy was defined by antimicrobial therapy administration following susceptibility testing result availability. Regimens were considered to be "appropriate" when they consisted of a minimum of one drug to which the causative bacterium was noted to be susceptible upon in vitro susceptibility testing, whereas they were otherwise deemed "inappropriate." Combination therapy was the administration of more than one antimicrobial treatment exhibiting in vitro activity, whereas monotherapy was the administration of just one antimicrobial agent exhibiting in vitro activity (28).
Data collection. Data pertaining to patient demographics (age and sex), comorbidities (see Text S1 for definitions of underlying conditions), hospital wards, prior antibiotic exposure, invasive procedures (mechanical ventilation, urinary catheterization, gastric catheterization, and central venous catheterization), APACHE II scores at BSI onset, Pitt bacteremia scores at BSI onset, empirical and definitive antimicrobial treatments administered, and 30-day all-cause mortality rates were obtained from patient medical records and analyzed retrospectively.
Statistical analysis. Kolmogorov-Smirnov tests were used to assess data normality. Normally distributed data are given as mean values 6 standard deviations (SD), while other data are given as median values and interquartile ranges. Categorical variables were analyzed with the chi-square test or Fisher's exact test, whereas continuous data were evaluated via Student's t test or the Mann-Whitney U test. A two-tailed P value of ,0.05 was the significance threshold for all studies, and all analyses were conducted using SPSS 22.0 (IBM, Armonk, NY, USA).
Baseline features were compared between patients that were and were not alive at the 30-day time point, with those variables yielding a P value of ,0.1 in univariate analyses being incorporated into a multivariate backward stepwise logistic regression model to establish independent predictors of CRE-BSI patient 30-day mortality. The relative efficacies of different antimicrobial regimens were assessed by treating these risk factors as confounding variables in a multivariate backward stepwise logistic regression model analysis.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.4 MB.