Disease Severity Is an Independent Risk Factor of Mortality and Outcomes in Critically Ill Patients With Carbapenem-resistant Klebsiella Pneumoniae Bloodstream Infections: a 5-year Retrospective Analysis


 Background: Carbapenem-resistant Klebsiella pneumoniae bloodstream infections (CRKP-BSIs) are associated with high morbidity and mortality rates, especially in critically ill patients. Comprehensive mortality risk analyses and therapeutic assessment in real-world practice are beneficial to guide individual treatment.Methods: We retrospectively analyzed 87 patients with CRKP-BSIs (between July 2016 and June 2020) to identify the independent risk factors for 28-day all-cause mortality. The therapeutic efficacies of tigecycline-and polymyxin B-based therapies were analyzed.Results: The 28-day all-cause mortality and in-hospital mortality rates were 52.87% and 67.82%, respectively, arising predominantly from intra-abdominal (56.32%) and respiratory tract infections (21.84%). A multivariate analysis showed that 28-day all-cause mortality was independently associated with the patient’s APACHE II score (p = 0.002) and presence of septic shock at BSI onset (p = 0.006). All-cause mortality was not significantly different between patients receiving tigecycline- or polymyxin B-based therapy (55.81% vs. 53.85%, p = 0.873), and between subgroups mortality rates were also similar. Conclusions: Critical illness indicators (APACHE II scores and presence of septic shock at BSI onset) were independent risk factors for 28-day all-cause mortality. There was no significant difference between tigecycline- and polymyxin B-based therapy outcomes. Prompt and appropriate infection control should be implemented to prevent CRKP infections.


Introduction
In recent years, bacterial resistance has become a signi cant problem, worldwide. As one of the most threatening and critically resistant carbapenem-resistant Enterobacteriaceae (CRE) classi ed by the World Health Organization, carbapenem-resistant Klebsiella pneumoniae (CRKP) are gradually increasing (1). According to a resistance report from the China Antimicrobial Surveillance Network, the imipenem and meropenem resistance rates for K. pneumoniae had increased to 25% and 26.3% in 2018, respectively (2). CRKP infections often occur in intensive care units (ICUs), leading to severe nosocomial infections, such as pneumonia, bloodstream infections (BSIs), complicated intra-abdominal infections, and urinary tract infections. These infections frequently cause multiple organ dysfunctions and septic shock, leading to an increased risk of in-hospital mortality, prolonged hospitalization, and high economic costs (3,4). It is worth noting that the mortality rate (33-70%) associated with CRKP-BSIs is three-fold higher than that of infections occurring at other bodily locations, and has become an urgent and widely acknowledged concern in clinical practice (5). Previous studies have shown that, compared with carbapenemsusceptible K. pneumoniae BSIs, non-transplant patients with histories of previous gastric catheterization, carbapenem use, hypoproteinemia, and high acute physiology and chronic health evaluation (APACHE) II scores have an increased risk of acquiring CRKP-BSIs (6). Likewise, septic shock and high APACHE II, Pitt bacteremia, and Charlson comorbidity index scores are independent risk factors for CRKP-BSI mortality (6, 7). However, the prognoses of ICU patients with CRKP-BSIs are related to their underlying diseases and demonstrates regional differences (8). Comprehensive risk assessments, based on baseline clinical features and infection severity, are of paramount importance to the early, individualized treatment of patients with CRKP-BSIs.
Although the mortality of patients with CRKP-BSIs is high, the e cacies of different antibiotic therapies vary according to the bacterial resistance mechanism(s) possessed by the infecting organisms (9).
Hence, the optimal treatment remains uncertain and the clinical antibiotic therapy options are limited, despite several antibiotics having been reported as effective against CRKP infections (10). In recent years, tigecycline-and polymyxin B-based antimicrobial therapies have been recommended for and widely used against drug-resistant Gram-negative bacilli according to the Chinese consensus statement (11). In May 2019, the new β-lactamase inhibitor complex ceftazidime/avibactam was approved and available for use, in China. Nevertheless, the exact therapeutic effects of these two CRKP-BSI treatment options remains uncertain, especially in critically ill patients. Therefore, we retrospectively reviewed the clinical data for patients with CRKP-BSIs treated, between June 2016 and June 2020, in our hospital's ICUs. Our aim was to identify the predictors of mortality associated with CRKP-BSIs and explore the therapeutic e cacies of tigecycline-and polymyxin B-based therapies in critically ill patients.

Study design and population
This retrospective study was conducted in Ruijin Hospital, a liated to Shanghai Jiao Tong University School of Medicine, a teaching hospital with 3300 beds in Shanghai, China. The study protocol was approved by the hospital's Medical Ethics Committee (approval, 2019-1-3). Between July 2016 and June use of mechanical ventilation) at BSI onset were assessed, and further calculated by APACHE II score, pitt bacteraemia score (PBS) and sequential organ failure assessment score (SOFA score). We extracted the antimicrobial treatment administered to each patient from their medical records. Additionally, we evaluated the attributable and 28-day all-cause and in-hospital mortality rates following each patient's initial positive blood culture, along with the total lengths of hospital and ICU stays.

Microbiological methods
All blood culture isolates were processed in the hospital's clinical microbiology laboratory. Bacterial species identi cations were con rmed using matrix-assisted laser desorption/ionization time-of-ight mass spectrometry (bioMérieux, Marcy-l'Etoile, France) and susceptibility testing was performed (VITEK 2, bioMérieux) (12). Minimum inhibitory concentrations (MICs) were classi ed according to breakpoints established by the Clinical and Laboratory Standards Institute (CLSI; Annapolis Junction, MD, USA). The CRE de nition was based on CLSI guidelines, with carbapenem resistance de ned as in vitro resistance to at least one of the following: ertapenem (MIC ≥ 2 mg/L), imipenem (MIC ≥ 4 mg/L), or meropenem (MIC ≥ 4 mg/L, disk diffusion zone ≤ 19 mm) (13).

De nitions
Patients with CRKP-BSIs were de ned as those with at least one CRKP-positive blood culture and clinical infection symptoms consistent with bacteremia (14). BSI onset was de ned as the date on which the rst positive blood culture was collected. The probable infection sources were classi ed into the following categories by the attending physicians: BSI from central lines, such as central venous or hemo ltration catheters; intra-abdominal infection (IAI); respiratory tract infection; skin and soft tissue infection (SSTI); and cardiovascular system infections such as endocarditis; or primary BSI, when a source was not identi ed (14). Septic shock was de ned according to the de nition of sepsis-3 (15). AKI was classi ed according to the Kidney Disease: Improving Global Outcomes clinical guidelines. Empiric antibiotic therapy was selected, according to clinical judgment, prior to receipt of the blood culture report (16).
Treatment with tigecycline and/or any other antibiotics, except polymyxin B, was considered to be tigecycline-based therapy; any treatment involving polymyxin B was considered to be polymyxin B-based therapy (17,18). Combination therapy was de ned as a regimen including more than one in-vitro active antimicrobial (19,20).

Procedures and assessments
The primary outcome of this study was the 28-day all-cause mortality rate; survivor and non-survivor subgroups were compared to identify mortality predictors. We analyzed the therapeutic e cacies of tigecycline-and polymyxin B-based therapies because the antibiotic treatment options were varied. The e cacy analyses also involved subgroup analyses based on age (≤ 65 years vs. > 65 years), men versus women, BMI (≤ 25 vs. > 25), APACHE II score (≤ 20 vs. > 20), with or without CRRT, and with or without mechanical ventilation.

Statistical analysis
Categorical variables were reported as frequencies and percentages; continuous variables were reported as means and standard deviations (SDs), if they were normally distributed, or as medians and interquartile ranges (IQRs), if they were non-normally distributed. Categorical variables were compared using chi-square or Fisher's exact tests; continuous variables were compared using Student's t-test or the Mann-Whitney U-test, according to their distribution. Variables that were signi cant in the univariate analyses (p < 0.1) were added to a stepwise, multiple logistic regression model to identify the independent risk factors for 28-day mortality. The survival analysis was performed using the Kaplan-Meier method, and a p-value < 0.05 was considered statistically signi cant. The Cochran-Mantel-Haenszel test was used for the various subgroup analyses (age, sex, BMI, APACHE II score, CRRT, and mechanical ventilation). All statistical tests were reported as two-tailed tests, and p-values < 0.05 were considered statistically signi cant. All statistical analyses were conducted using SPSS (version 23.0; IBM, Armonk, NY, USA).

Characteristics of critically ill patients with CRKP-BSIs
A total of 87 patients (67 males, 77.0%) with CRKP-BSI were included in the study, the baseline characteristics and univariate analyses of the study population are shown in Table 1. The most frequent sources of the bacteremia were intra-abdominal (56.32%), respiratory tract (21.84%), central lines (9.20%), primary (5.75%), skin and soft tissue infections (3.45%), urinary tract infections (2.29%), and cardiovascular system infections (1.15%). AKI and septic shock occurred in 51.72% and 73.56% of patients, 33.33% of the patients received CRRT and 78.16% required mechanical ventilation support at the BSI onset. The mean APACHE II score at CRKP-BSI onset was 23.3 ± 7.58 and SOFA was 8 (range, 5.5-11). The 28-day all-cause and in-hospital mortality rates were 52.87% and 67.82%, respectively. The 28day mortality and in-hospital mortality rates attributable to infection were 47.13% and 55.17%, respectively. Furthermore, the 28-day all-cause mortality rates, by major source of infection, were 53.06% (26/49) in those with IAIs, 63.16% (12/19) in those with respiratory infections, and 50.00% (4/8) in those with central line infections; there was no signi cant difference in the CRKP-BSI mortality rates strati ed by site of infection.

Therapeutic e cacy of tigecycline-and polymyxin B-based antimicrobial regimens
To compare the e cacy of tigecycline-and polymyxin B-based therapies, we compared the demographics and clinical characteristics between the tigecycline and polymyxin B groups ( 88.37%, p = 0.009). There were no other signi cant between-group differences in clinical characteristics or illness severities at BSI onset. The 28-day all-cause mortality rates were not signi cantly different (55.81% vs. 53.85%, p = 0.873) for patients receiving either tigecycline-or polymyxin B-based therapies. Similarly, there were no between-group differences in all-cause mortality, attributable mortality, or attributable inhospital mortality rates. The Kaplan-Meier curve for 28-day survival distributions is shown in Fig. 1 (logrank, p = 0761). The lengths of hospital and ICU stays were comparable between the two groups. Moreover, we compared the tigecycline-based regimens with polymyxin B-based regimens not including ceftazidime-avibactam treatment and did not nd any signi cant between-group difference in 28-day mortality rates (55.81% vs 54.17%, p = 0.897). Additionally, there was no signi cant difference in 28-day mortality rates between the tigecycline-and polymyxin B-based combination therapies (51.28% vs. 56.0%, p = 0.712). clarify the impact of these variables and indicators of disease severity on tigecycline-and polymyxin Bbased therapy outcomes. However, no signi cant differences in treatment e cacies were evident in the subgroup analyses (Fig. 2).

Discussion
CRKP infections are life-threatening health problems, worldwide. Critically ill patients in ICUs have a higher risk of contracting CRKP-BSIs because of their illness severity, immunosuppression, and frequent use of invasive procedures; ICU-acquired infections were previously shown to be a powerful causative factor leading to the development of CRKP-BSIs (3). Notably, patient population heterogeneity, such as infection source, underlying disease severity, and response syndrome severity at presentation, are important for determining CRKP-BSI prognoses. This is because these BSIs are inherently di cult to eradicate, eventually spreading to different sites throughout the body, and are associated with high morbidity and mortality rates. Therefore, evaluations of the pathogenic characteristics, in-hospital death risk factors, and treatment outcomes of CRKP-BSIs are important.
Here, we analyzed the risk factors for CRKP-BSI mortality rates and treatment outcomes in 87 critically ill patients. The average APACHE II and SOFA scores and the 28-day all-cause mortality rates were higher than those previously reported. The higher mortality rate may be attributed to differences in the bacteremia sources and patient disease severity heterogeneity. The infection source for CRKP-BSIs frequently affects patient prognoses, as evidenced in PANORAMA study, primary bloodstream infections were the most predominant source of BSIs (48%), followed by gastrointestinal system infections (21%) (8). In the EUROBACT study, primary bloodstream (23.7%) and catheter-related (21.4%) infections were the most common sources, followed by respiratory tract (21.1%) and intra-abdominal (11.6%) infections (21). Most patients admitted to our ICU had complicating abdominal and respiratory tract infections, and 78.16% of the CRKP-BSIs originated from these two sources, leading to 28-day mortality rates of 53.1% and 63.2%, respectively. Abdominal infections often require more than one source control interventions, and the effect of drainage is di cult to evaluate (22). Respiratory tract infections are considered to less "drainable" and such uncontrolled infection sources seem to be independently associated with poor outcomes (23). Moreover, the patient's underlying condition and disease severity directly affect the prognosis. In our study, septic shock developed in 73.56% of patients and was 93.48% among the nonsurvivors; the 28-day CRKP-BSI mortality rate in patients with septic shock was 67.19%. Similarly, the risk factors for CRKP-BSI mortality identi ed in our analysis are consistent with those reported previously, including older age; biliary tract disease; high procalcitonin levels; high APACHE II, SOFA, and Pitt bacteremia scores; and development of AKI and septic shock upon BSI onset (in our univariate analysis). Furthermore, our multivariate analysis revealed that the APACHE II score and septic shock were independent risk factors for crude 28-day mortality (7,24). Therefore, we should be aware that the mortality risk is based on the illness severity at BSI onset; thus, early and adequate source control and appropriate antimicrobial treatment should be implemented as soon as possible (25,26).
The selection of antimicrobial therapies for CRKP is very limited; tigecycline and colistin/polymyxin B have been recommended by a Chinese consensus panel on the treatment of multidrug-resistant Gramnegative bacilli, and are widely used in China until now (11). The resurgence of colistin has resulted from its unique mechanism of action: disruption of the Gram-negative bacterial outer membrane and provision of rapid bactericidal activity in conjunction with other antibiotics. Polymyxin B is the preferred agent for systemic use in invasive infections because of its superior pharmacokinetics and reduced nephrotoxicity potential (27). A recent retrospective cohort study found that combination therapies, mostly polymyxin B plus amikacin, had signi cantly lower 30-day mortality rates than monotherapies (37.5% vs. 64.75%, p = 0.01) in 82 patients with CRKP-BSIs (28). However, reports still emerge of inadequate dosages, treatment failures, adverse effects, and heterogeneous resistance during polymyxin B treatment (29).
Tigecycline is a broad-spectrum glycyl cyclic peptide antibiotic that has shown in vitro activity against CRE; however, the e cacy of tigecycline against CRE infections, especially BSIs, remains debatable because of insu cient dose and low serum concentration (30). Multivariate analyses have shown that the use of tigecycline is not only an independent risk factor for CRKP-BSI development (OR, 3.915; p = 0.005), but its use was also associated with an increased risk of 28-day mortality (31). However, a systematic review and meta-analysis that included 21 controlled studies (1595 patients) and 5 single-arm studies (113 patients) revealed that the tigecycline groups had similar overall mortality rates as the control groups (OR, 0.96; 95% CI, 0.72-1.22; p = 0.73); but tigecycline combination therapies and high-dose regimens have been reported to be associated with better outcomes, without signi cant adverse effects in the subgroup analysis (32). These data highlight the utility of combination antimicrobial therapy for CRKP-BSIs, particularly in high-risk, severely ill patients (20,33). Our study re ects the broad acceptance of combination therapy for the treatment of CRKP-BSIs in that 90.69% and 96.15% of cases were prescribed either tigecycline-or polymyxin B-based combination therapy, respectively. Despite this high rate of combination therapy use, the 28-day mortality rate remained > 50%, suggesting that we need to rethink and assess the e cacy of these two drugs.

Conclusions
In conclusion, critically ill patients with CRKP-BSIs showed poor outcomes and high mortality rates. Disease severity indicators, speci cally the APACHE II score and evidence of septic shock at BSI onset, were independent risk factors for 28-day all-cause mortality. Although combination therapy was commonly used, the 28-day mortality rate was high and there was no signi cant difference in e cacy between the tigecycline-and polymyxin B-based regimens. In this era of limited new drugs for CRKP infections, prompt and appropriate infection control should be implemented to prevent the dissemination of resistant microorganisms among high-risk patients, especially those in the ICU. The study protocol was approved by the Institutional Ethic Committees of the Shanghai Jiao Tong University School of Medicine and Ruijin Hospital (approval, 2019-1-3).

Consent for publication
Not applicable.
Availability of data and materials The raw data can be made available to the interested researchers by the authors of this article if requested.

Competing interests
The authors declare that they have no competing interests.