Due to the weak immune system and the accumulation of various factors, such as repeated chemotherapy and long hospital stays, patients with haematologic malignancies suffer from greater morbidity and mortality from BSI than patients with other cancers [5, 23]. In recent years, the expansion of GNB and antibiotic-resistant strains associated with significant mortality has elicited considerable concern because there are few possible therapeutic alternatives. It is essential to elucidate the pathogen distribution and identify the key factors that could help in assessing the risk of infection by antibiotic-resistant bacteria and mortality in haematologic malignancy patients, which can then be used to provide appropriate and immediate interventions (such as antibiotic administration) to improve patient survival.
In this large-scale, single-centre study, the microbiological and clinical characteristics and outcomes of haematologic malignancy patients diagnosed with GNB BSI were recorded and analysed. The three most common GNB species were E. coli, Klebsiella pneumoniae and Pseudomonas aeruginosa, which together accounted for 78.6% of the GNBs causing BSI. This finding is consistent with the results of recent studies from other cities in China and other countries, such as Italy and Lebanon [3, 4, 24]; however, the proportion of E. coli (28.8%) in our study was much lower than that in Italy (52.7%) and Lebanon (45.6%), corresponding to a greater proportion of Klebsiella pneumoniae and Pseudomonas aeruginosa, thus suggesting that the prevalence of these bacterial species differs among global regions. It is necessary for the implementation of regular local surveillance of pathogen epidemiologic status and antibiotic susceptibility to evaluate antimicrobial strategies and adapt them to mitigate the effects of emerging pathogens.
Carbapenems have been used as last-line antibiotics to treat infections caused by GNB, which are typically resistant to many first- and second-line antibiotics, such as cephalosporins and monobactam. Unfortunately, GNB rapidly developed resistance to carbapenems due to the selective pressures exerted by using this antibiotic family, and has now spread throughout the world. CRGNB strains usually include lactose nonfermenters such as CREc, CRKP and other Enterobacteriaceae, as well as nonfermenters such as CRAB and CRPA [25]. The impact of resistance may depend on the type of gram-negative bacteria, with resistance in Klebsiella pneumoniae and Pseudomonas aeruginosa having a greater impact than in Acinetobacter baumannii [7]. In this study, 25.9% of GNB were carbapenem resistant, which was similar to the findings of studies from Italy and Israel [4, 7]. Not surprisingly, CRKP was the most frequently isolated CRGNB in this study, as it has been one of the most common pathogens causing BSIs that increasingly cause carbapenem resistance worldwide [26]. CRKP has been previously reported as being the bacterial species most frequently responsible for infections (mainly BSIs) in haematologic malignancies patients [24, 27, 28]. The number of CRPA-infected strains and the proportion of CRPA-infected strains among all Pseudomonas aeruginosa strains were lower than those of CRKP-infected strains in this study (but not much lower). In addition, CREc and CRAB were also observed in this study and have been reported in BSIs of haematologic malignancies patients in other studies [7, 24].
Timely recognition of the most at-risk patients with CRGNB infection is critical for early appropriate empirical regimen selection within the first 24–48 h of infection and for source control of nosocomial dissemination, due to the fact that the standard empiric antibiotic therapies recommended for haematologic malignancy patients do not correspond to first-line CR-bacteria-targeted treatments, as well as the fact that CR-bacteria identification takes at least 48 h [15]. In the analysis of CRGNB BSI, we found that patients with underlying chronic liver disease were at an independent risk for infection BSI caused by CRGNB when GNB BSI was performed. This is due to the fact that the liver is essential for clearing bacteria and related toxins (such as endotoxin) from the bloodstream [29]. Liver dysfunction caused by chronic liver disease, such as alcoholic liver disease, is associated with a higher incidence of infections and a higher mortality rate due to sepsis [30]. Consistent with other studies [7, 31, 32], we also found that recent carbapenem treatment was significantly related to the occurrence of CRGNB bacteraemia. China is one of the main regions of the world with a high prevalence of antibiotic-resistant bacteria, and it is unavoidable that carbapenems are frequently used for empirical and targeted treatment, especially for patients in intensive care units and malignancy wards. Carbapenem treatment could eliminate the carbapenem-susceptible strains which resulted in carbapenem-resistant strains become the dominant population for infection.
Albumin is synthesized by the liver and reflects liver function and nutritional status. Patients with low albumin concentrations are more likely to have lower immunity to resist bacterial infection. A previous study demonstrated that a low albumin concentration was a risk factor for BSI and CRGNB BSI among patients with haematological malignancies [7]. Consistently, our study demonstrated that an albumin concentration < 30 g/l was closely associated with a CRGNB BSI. Furthermore, a platelet count < 30×109/l was also identified as being a risk factor significantly related to CRGNB BSI in our study, and which is rarely included in risk factor analysis of BSI in patients with haematological malignancies. Platelets are considered to be the sentinels of the bloodstream for rapid identification of microbial invasion by providing a comprehensive armamentarium of pathogen detection systems [31]. It is common knowledge that patients with haematological malignancies are immunocompromised because they receive chemotherapy. It could be hypothesized that a low concentration of albumin and a low platelet count combined with an immunocompromised state associated with a haematological malignancy may increase the risk for GNB BSI.
The 30-day mortality of GNB BSIs was 24.5% in our study, whereas other studies have reported mortality rates ranging from 16.3–32.1% [24, 32, 33]. Similar to other studies [4, 7], the 30-day mortality rate was much greater in CRGNB patients than in CSGNB patients. Several studies have reported of the risk factors for mortality in haematological malignancy patients with BSIs and GNBs; however, to our knowledge, there is no published clinical predictive model that allows us to accurately identify haematological malignancy patients at high risk of mortality within 30 days due to GNB BSIs. Herein, we established a multivariate regression model for predicting the 30-day mortality of haematological malignancy patients with GNB BSIs. Risk factors retained in the final model that were significantly and independently associated with 30-day mortality included septic shock, mechanical ventilation after BSI, neutropenia before BSI and an albumin concentration < 30 g/l before BSI. The model had satisfactory discrimination, consistency and clinical benefit based on the assessments of the C-indices, calibration plots and decision curves.
Septic shock is defined as life-threatening organ dysfunction caused by a dysregulated host response to infection and is associated with unacceptably high mortality [34]. According to numerous studies on bacteremia in patients with haematological malignancies [3, 7, 35], septic shock is an independent risk factor for mortality. Early screening of patients who exhibit low oxygen saturation, hypotension, or disturbance of consciousness, as well as timely intervention, including fluid resuscitation, vasopressors and appropriate antimicrobial therapy, are critical for removing the source of infection. In addition to septic shock, we also found that mechanical ventilation after BSI was strongly related to mortality. Few studies have included mechanical ventilation for potential risk factor analysis in haematological malignancy patients with BSI, although it has been widely identified as being a risk factor for mortality after BSI in other studies. Patients with bacteraemia requiring mechanical ventilation tend to have high APACHE and SOFA scores, longer lengths of hospital stay, multiple vascular lines, and greater risks of colonization and multiple infections. This emphasizes how crucial it is to account for serious underlying risk factors when evaluating prognosis in these patients.
Neutropenia is a common adverse effect of chemotherapy, haematopoietic stem cell transplantation, and disease course in patients with haematological malignancies. Previous studies have suggested that neutropenia is an independent risk factor for BSI and has a significant impact on survival rates [5, 27, 36]. Neutrophils are essential for the acute inflammatory response and bacterial elimination. Neutropenia reduces the inflammatory response to developing infections, which promotes bacterial growth and invasion and renders patients vulnerable to recurrent infections. As life-threatening infections are more likely to occur in haematological malignancy patients with neutropenia, effective management of neutropenia is indispensable to maximize patient outcomes. Studies have demonstrated that an ALB concentration < 30 g/l is significantly related to 30-day mortality in haematological malignancy patients with BSIs [3, 36]. In our study, an ALB concentration < 30 g/l was not only a dependent factor for CRGNB infection but also for GNB mortality. The monitoring of the albumin level and intervention by infusing albumin are important for preventing infection and improving the prognosis of haematological malignancy patients with GNB BSIs.
Our study was limited by the single-centre retrospective nature of the study. There are likely clinical aspects that were not quantified or measured among haematological malignancy patients, and residual confounding factors cannot be excluded. Regional differences in the diagnosis, laboratory testing, therapeutic regimens and epidemiology of antibiotic-resistant isolates may result in different findings. Our predictive model should be optimized through further external validation and prospective studies.
In conclusion, our study investigated the occurrence of GNB in haematological malignancy patients with BSIs and evaluated the demographic and clinical factors related to CRGNB BSI and GNB BSI-related mortality. CRGNB BSIs are frequently observed and are associated with poor prognosis. The risk of CRGNB BSI can be estimated when GNB BSI is confirmed by evaluating factors such as chronic liver disease, previous exposure to carbapenem therapy in one month, a platelet count < 30×109/l and an albumin concentration < 30 g/l before BSI. Furthermore, neutropenia before BSI, an albumin concentration < 30 g/l before BSI, septic shock and mechanical ventilation after BSI were found to be independent risk factors for 30-day mortality in haematological malignancies patients with GNB BSIs. A nomogram with satisfactory predictive ability based on these factors was developed. It allows for patients to be stratified according to their risk of poor prognosis, and early effective management can be used to improve patient survival.