Phenotype, molecular characterisation and risk factors for postoperative meningitis caused by ESBL-producing-Enterobacteriaceae: a six years multi-Centre comparative cohort study

Background To determine the phenotype, molecular characterisation and risk factors of postoperative meningitis induced by Extended-spectrum β-lactamase (ESBL)-producing Enterobacteriaceae (EPE) in China. Methods We performed a multi-centre comparative cohort study of postoperative meningitis patients infected with Enterobacteriaceae in 4 neurosurgical centres in China from January 2014 to December 2019. Phenotype and molecular characteristics of the isolates were reviewed and tested, and independent risk factors of the EPE meningitis were evaluated by binary logistic regression. Results In total, 220 Enterobacteriaceae include 78 EPE were available in this study. 85.6% (67/78) ESBL-related genes were tested, and blaSHV (14.9%) and blaSHV + blaTEM + blaCTX-M-9 (20.9%) were found to be the most frequent mono and combined ESBL-related genes harboured by Enterobacteriaceae. On binary logistic analysis, craniotomy (OR. 2.583, 95% C.I. 1.274–5.235, P = 0.008) and malignancy (OR. 2.406, 95% C.I. 1.299–4.456, P = 0.005) were the associated independent risk factors to meningitis induced by EPE. Conclusions To the best of our knowledge, this is the largest series focusing on risk factors of EPE meningitis which has been conducted in China. Craniotomy and malignancy were independent risk factors for EPE meningitis. The risk factors identified may be further utilized in clinical practice and research to avoid and reduce the mortality in future.


Background
Postoperative meningitis, secondary to neurosurgical procedures, trauma, or shunt devices, is one of common healthcare-associated infections, which are a significant cause of perioperative morbidity, mortality and healthcare cost [1,2]. Incidence of postoperative meningitis in recent prospective studies varied between 0.7 and 25% [3][4][5][6]. In recent years, postoperative meningitis has been found to be closely related to the success of neurosurgical operation and mortality of neurosurgery patients; therefore, it has received increasing attention in clinical settings. Patients who underwent neurosurgery are known to have lower immunity, experience complicated longer duration of operations, and are more difficult to treat [7]. Therefore, reducing the incidence of postoperative meningitis is an important task for neurosurgical physicians.
Various pathogens can cause postoperative meningitis, of which, and Enterobacteriaceae is a critical branch. Enterobacteriaceae has been reported to account for more than 20% of nosocomial infections [8]. Multidrugresistant Enterobacteriaceae are more harmful due to their extensive drug resistance. Extended-spectrum βlactamase (ESBLs) production is one the most common antibiotic resistance mechanism of Enterobacteriaceae, and ESBL-producing Enterobacteriaceae (EPE) has been an increasingly implicated as a cause of infection. A previously study reported that prevalence of EPE rectal colonisation in healthy human beings was 14% globally, and the distribution varied extremely in different locations [9]. Also, EPE, as a nosocomial pathogen, is a health threat among medical institutions, especially in surgery patients with immune-compromised or other comorbidities. In 2017, the WHO released a list of antibioticresistant bacteria that pose the greatest threat to human health, and for which new antibiotics are desperately needed. From that, EPE belongs to the first-grade priority (critical) and is a clinically ultra-threatening pathogen [10]. The risk factors for EPE include prolonged mechanical ventilation [11], ICU admission [12], catheter usage [13], severe illness [14], and frequent antibiotic usage [15]. However, few reports are target on risk assessment of postoperative meningitis. Existence of the blood-brain barrier may block antibiotics such as polymyxin B entering the central nervous system, and lead to EPE postoperative meningitis more harmful [16], therefore, risk factor assessment possess great clinical significance.
To better prevent and determine the treatment strategy of postoperative meningitis caused by EPE, we conducted a multi-centre comparative cohort study to explore the independent risk factors and assess clinical molecular characteristics of postoperative meningitis caused by EPE. To our knowledge, this is the first study globally to assess the molecular characteristics of EPE and risk factors of EPE postoperative meningitis.

Patients
In the four centres, adult neurosurgical patients(> 18 years old) were included if they were survived at least 7 days with at least one cerebrospinal fluid (CSF) culture positive for Enterobacteriaceae. Patients who underwent only external ventricular drain (EVD) or CSF shunt or stereotactic surgery, without antimicrobial therapy not done in hospital and incomplete clinical medical records were excluded. All of the patients were followed-up for the diagnosis of postoperative meningitis during the first 30 days after neurosurgery. A flow chart of this study is shown in Fig. 1.
Patients with Enterobacteriaceae postoperative meningitis were diagnosed by the diagnostic criteria of the Centres for Disease Control [17,18]. The criteria of Enterobacteriaceae meningitis was as follows: 1) Enterobacteriaceae cultured from neurosurgical patients' CSF; 2) one or more of the clinical symptoms or signs, including fever(> 38°C), headache, meningeal signs (stiff neck, cranial nerve signs or irritability), with anti-infective treatment; and one or more of the clinical laboratory tests as follows: 1) elevated CSF protein level, increased CSF leucocyte count, and/or decreased CSF glucose level; 2) appearance of gram-negative bacillus on gram stain of CSF; 3) growth of Enterobacteriaceae in patients' blood (by bacterial culture).

Microbiology
All bacteria with the same growth characteristics of Enterobacteriaceae were classified to standard microbial identification procedure, and the identification system is VITEK-2 Compact system ((bioMerieux, Marcy l 'etoile, France, based on biochemical reaction) and VITEK MS (bioMerieux, Marcy l 'etoile, France, based on matrixassisted laser desorption/ionization time-of-flight mass spectrometry) system. Antimicrobial testing of susceptibility (AST) of Enterobacteriaceae was performed by the disc diffusion tests (Kirby-Bauer method) and broth microdilution method (MIC) and classified as sensitive, intermediate and resistant according to the Clinical and Laboratory Standards Institute (CLSI) 2019. Isolates tested with aztreonam or ceftazidime MICs > 1 mg/L were screened for the presence of ESBL. Judgment of an ESBL phenotype of four Enterobacteriaceae (Klebsiella pneumoniae, Klebsiella oxytoca, Escherichia coli and Proteus mirabilis) was performed by Kirby-Bauer method according to the CLSI 2019. The test procedure is as follows: ceftazidime/clavulanate and cefotaxime/clavulanate discs, in comparison with ceftazidime and cefotaxime discs alone (Oxoid Ltd., Basingstoke, Unite Kingdom). According to the CLSI 2019 antimicrobial susceptibility testing standards [19], EPE was defined when an increase in a zone diameter ≥ 5 mm for either antibiotic together with clavulanate vs the zone diameter of the agent, and it is reported that the ESBL phenotype of other Enterobacteriaceae can also be determined by the same method as the above four Enterobacteriaceae [20]. The left were defined as non-ESBL-producing Enterobacteriaceae (NEPE).
Molecular detection of bla CTX-M-1 , bla CTX-M-9 , bla SHV , bla TEM , bla OXA-23 and bla OXA-66 were done by micro/ nanofluidic chip platform (MNCP) based on the loopmediated isothermal amplification (LAMP) method [21]. The extraction of the nucleic acid was done according to MNCP manufacturer's instructions. In addition, preparation of the isothermal amplification reaction solution and detection using the MNCP were the same as previously described [22].

Therapy and clinical variables for evaluation
In this study, all of the patients' qualified daily progress records were established by analyse of the clinical database of the neurosurgery, infectious diseases and microbiology departments in the four centres described above. From that, we summarised and measured the status of antibiotic use and clinical outcomes in patients with postoperative meningitis, including antibiotic prophylaxis, empirical, definitive therapy and mortality. Antibiotics were employed by the neurosurgical doctors according to local or international common standards and antimicrobial susceptibility testing. Twenty-one characteristics of the postoperative meningitis patients were extracted from the clinical database for risk factor evaluation, including patients' routine information (age, male%), fever(> 38°C), assist mechanical ventilation (AMV), bacteraemia, craniotomy, CSF leakage, diabetes mellitus, EVD, Glasgow Coma Scale (GCS), hospitalacquired pneumonia, hypertension, intensive care unit (ICU) admission, lumbar drainage (LD), long-time surgery duration(> 180 min), length of hospital stay (LOS), malignancy, postoperative infection time, reoperation, surgical wound classification, and time of cure of infection.

Statistical analysis
All of the variables differences were evaluated by univariate analysis. Of them, the categorical variables were assessed by Pearson's chi-squared test. Quantitative data were assessed using Student's t-test or one-way variance analysis test, and abnormally distributed quantitative variables were processed using Kruskal-Wallis or Mann-Whitney U test. Binary logistic algorithm was built to evaluate differences between the EPE and NEPE postoperative meningitis patients. Any variables with P < 0.1 in the univariate analysis were carried forward in binary logistic regression algorithm to analyse the independent risk factor for EPE postoperative meningitis. Significance was defined as a P < 0.05, and calibration was analysed by Hosmere Lemeshow (H-L) test for goodness-of-fit. Statistical analyses were carried out by SPSS 22.0 (IBM, New York, USA). The graph was performed using Prism 7.0 (Graphpad, San Diego, USA).

Patients
Over the 6 years of the study, a total of 51,227 neurosurgery patients and 4198 postoperative meningitis patients were included in the four centres described above. The infection rate was 8.2% (4198/51,227). Two hundred seventy-two cases of Enterobacteriaceae postoperative meningitis were recorded. Of them, 52 were excluded based on criteria (Fig. 1).

Microbiology
The distribution of Enterobacteriaceae is shown in Table 1. From that, Klebsiella pneumoniae has the highest proportion (40.9%, 90/220), followed by Escherichia coli (17.3%, 38/220), Enterobacter aerogenes (10.0%, 22/220), and Enterobacter cloacae (9.1%, 20/220). The antimicrobial susceptibility test of the Enterobacteriaceae is shown in Fig. 2. The proportion of EPE was 35.5%. In carbapenem, sensitivity to meropenem and imipenem were both 85.0%. Among the 78 EPE, 67 isolates of ESBLrelated genes were detected by MNCP, and the bla SHV gene is the most frequent ESBL-related gene. Majority (58.2%, 39/67) of the gene-harbouring EPE contained multiple genes, and the most frequent ESBL-related gene combination of EPE is bla SHV + bla TEM + bla CTX-M-9 . The whole distribution of the genes is shown in Fig. 3-A. Also, we classified all of the 67 EPE into 3 groups, including K. pneumonia, E. coli and others. The numbers of each groups' genes are shown in Fig. 3

Binary logistic analysis of risk factors for EPE postoperative meningitis
We conducted a binary logistic analysis to evaluate the independent risk factor of EPE postoperative meningitis (Table 4). Firstly, we included all the factors with P < 0.1 in the univariate analysis into the binary logistic analysis. Then, the H-L test of the model was calculated for calibration. All the variables embedded in the algorithm by binary logistic analysis were listed in Table 4. From that, craniotomy (odds ratio (OR) 2.583, 95% confidence interval(C.I.) 1.274-5.235, P = 0.008) and malignancy

Discussion
In this multi-centre study, we conduct the largest neurosurgical individuals with Enterobacteriaceae postoperative meningitis up to now, including 220 adult patients during 2014-2019. Characteristics of the molecular epidemiology, and risk factors in EPE postoperative meningitis were screened. Of them, craniotomy (OR 2.583, 95% C.I. 1.274-5.235, P = 0.008) and malignancy (OR 2.406, 95% C.I. 1.299-4.456, P = 0.005) were independent risk factors for EPE postoperative meningitis, and the mortality attributed to infection between the two groups is significantly different. It was also confirmed that the majority of EPE contained more than one ESBL related gene, while bla SHV + bla TEM + bla CTX-M-9 is the most frequent gene combination.
Enterobacteriaceae is one of the critical pathogenic bacteria that cause nosocomial infections, occupying a large proportion of all pathogenic bacteria. As previously reported, EPE has played a vital role in nosocomial infection [23]. It has the properties of easy to acquire resistance genes (e.g., horizontal gene transfer) [24], and high pathogenicity [25]. It has strong drug resistance characteristics, and the incidence of ESBL production is 35.5%, but the proportion of ESBL production by K. pneumoniae and E. coli is as high as 46.7 and 52.6%. The WHO ranks EPE as the first priority of grade (critical), it has extensive drug resistance and is resistant to most β-lactam antibiotics  This study confirmed that the majority of EPEs does not pose single ESBL-resistance gene. The proportion of EPE with a single gene is 41.8%. Among them, the bla SHV gene is the most frequent ESBL-related gene where bla SHV were more common [28]. In the Netherlands, it's reported that bla CTX-M-1 was the predominant gene [29]. Previously studies above certified   [30]. The most frequent ESBL-related gene combination of EPE is bla SHV + bla TEM + bla CTX-M-9 (20.9%), followed by bla SHV + bla CTX-M-9 (13.4%), and the possible reason as follows: Firstly, plasmids possessing bla CTX-M genotypes are admitted to possessing other ESBL-related genes transmitting resistance to a series of antibiotics. Secondly, single replicon in distinct resistance gene location may lead to co-selection and may contribute to the dissemination [31].
The second-or third-generation cephalosporins were employed as antibiotic prophylaxis for postoperative meningitis patients ahead of the neurosurgery, which can prevent the invasion of pathogenic bacteria during the operation. In the empirical treatment, there are more antibiotics for selection, but vancomycin+meropenem is the main choice. This is because neurosurgery is generally complicated, and the use of the most effective antibiotics, such as carbarpenem, can most likely reduce the infection rate. In the definitive therapy, small percentage of patient population received monotherapy with meropenem. Carbarpenem seems to be the most frequently used antibiotic to cure EPE infections. Practice guidelines for management of ventriculitis/meningitis have been announced previously [32], a third-generation generation cephalosporin, such as ceftriaxone is endorsed for NEPE meningitis, while meropenem is endorsed as the first-line drugs target on EPE meningitis by general guidelines [33,34].
Nevertheless, carbapenems should be used with caution because its overuse will generate acquisition of carbarpenem intermediate/resistant isolates by degrees. Also, resistance to third-generation cephalosporin in Enterobacteriaceae is always associated with ESBL, whereas resistance to carbapenems can be caused by production of an ESBL or plasmid mediated coded AmpC cephalosporinase combined with an efflux pump system or a decrease in outer membrane permeability [35][36][37]. It is reported that the mutation rate of the membrane permeability of Enterobacteriaceae is relatively high [38,39], so when dealing with meningitis caused by EPE, even if it is treated with carbapenem antibiotics, good clinical outcomes may not be obtained. All of the 220 cases of Enterobacteriaceae in this study were sensitive to polymyxin B, which maybe the last line of defence against EPE or Carbapenem-resistant Enterobacteriaceae (CRE). It is reported that polymyxin B can be injected intrathecally in patients with severe Gram-negative bacteria [40], which maybe the ultimate choice for Enterobacteriaceae infection. However, polymyxin B causes nephrotoxicity, neuromuscular blockages and other adverse reactions, such as respiratory depression, and its application is limited to patients with critical disease. Therefore, to better deal with EPE postoperative meningitis, prevention of infection is still one of the main choices.
At present, several studies have explored the risk factors of postoperative meningitis. However, most of them focused on the incidence and total risk factors of neurosurgical meningitis. For example, a case-control study reports that patients with EVD, LD and diabetes were risk factors of meningitis after neurosurgery [3]. Another study showed that, hydrocephalus, Koos grade IV, operative duration > 3 h and intraoperative bleeding volume >400 ml were significantly were independent risk factors of postoperative meningitis after microsurgery for vestibular schwannoma. Fewer studies have focused on clinical characteristics and risk factors, specifically for multidrug resistant bacterial meningitis. Pintado et al. conducted a comparative cohort study to evaluate the prognostic factors of methicillin-resistant Staphylococcus aureus meningitis and concluded that mortality was related to indwell of cerebrospinal devices (OR 7.9, 95% CI 3.1-20.3, P < 0.001) [2]. Yagel conducted a matched case-control study to evaluate the bacteraemia caused by EPE and NEPE, and the result showed that Gramnegative bacteria have statistically significantly different factors involved in successful and failed treatment, including pathogen types, highest body temperature in the first 24 h of symptoms, CSF glucose content and meropenem susceptibility [41]. In this study, 21 variables were evaluated and binary logistic analysis of risk factors indicating that craniotomy and malignancy are individual independent risk factors of EPE postoperative meningitis. For craniotomy and malignancy patients, surgical operation is complicated, and the prognosis and the patient's own immunity is poor, which may lead to a high incidence of infection in patients. In addition, patients with malignancy have adopted high-level preoperative and empirical treatment [42], which may induce resistance to Enterobacteriaceae. Also, due to poor patient prognosis, the proportion of patients who gradually received highgrade antibiotics is high, and NEPE that cannot infect the patient's central nervous system directly may lead to Enterobacteriaceae meningitis.
Some limitations still exist in this study. First, this is a retrospective study, and the conclusion mostly depends on the accuracy of the data in hospital, which may result in selection bias, although we did a prospective genetic study of Enterobacteriaceae. Second, we did not include all clinical variables related to meningitis, such as grade of the tumors, multiple catheter insertion, and primary clinical laboratory tests were not embedded. In our next study, we will conduct a prospective study to include more neurosurgical centres (6)(7)(8) and patients (300-500) and strive to achieve a universal risk factor assessment algorithm across China, creating a more objective risk factor assessment and prediction of EPE postoperative meningitis.

Conclusion
To our knowledge, this study is the first one to conduct a multi-centre molecular and comparative cohort study of EPE postoperative meningitis. Enterobacteriaceae postoperative meningitis is a relatively serious clinical challenge worldwide. We have determined the characteristics of the Enterobacteriaceae, molecular epidemiology and evaluated the risk factors and treatment of postoperative meningitis. Craniotomy and malignancy were determined to be independent risk factors for EPE postoperative meningitis, and it is necessary to pay attention to prevention and treatment clinically.