Risk Factors Associated with Multi-Drug Resistance in Neonatal Sepsis Caused by Escherichia coli

Introduction

Neonatal sepsis is a major cause of neonatal mortality and therefore a significant global health issue. The reported incidence of neonatal sepsis is about 1–4%,¹ with a higher incidence being reported in developing countries.^1,2 To reduce neonatal mortality and long term adverse consequences caused by bacterial infection, it is vital to select appropriate empirical antibiotics when a neonatal infection is suspected. In most developing countries, including China, Escherichia coli (E. coli) is usually the leading pathogen in neonatal sepsis,^2–7 especially in premature infants.^1,8–10 E. coli plays an important role in causing both early-onset sepsis (EOS) and late-onset sepsis (LOS).^{3,6–8,11–14} E. coli is also one of the most common pathogenic bacteria causing neonatal bacterial meningitis.^15–17 Neonate with E. coli meningitis can have adverse nervous system sequelae, which is an important cause of poor long-term prognosis of newborns with E. coli sepsis.

Selection and timely administration of antibiotics to which the bacteria is sensitive to has important clinical significance for improving the prognosis of neonates with bacterial infections. The first-line empirical antibiotic regimen should be selected to cover E. coli and other common local pathogens based upon their historical antimicrobial sensitivity patterns. Recently, increasing reports show that the proportion of multi-drug resistant (MDR) E. coli is rising,^6,18,19 which brings big challenges to clinicians. Our recent meta-analysis based on a systematic review of the published studies in the Chinese literature demonstrates that in newborn infants hospitalized in Chinese NICUs, roughly 50% of all E. coli bloodstream isolates, whether early or late onset, are MDR due to extended spectrum beta-lactamase (ESBL) production.²⁰ We have further demonstrated that ESBL-producing MDR E. coli has emerged as the major pathogen responsible for EOS, particularly in preterm infants.⁵

The main mechanism of MDR gram-negative Enterobacteriaceae such as E. coli is ESBL production.^21,22 In neonatal intensive care units (NICUs) with a high prevalence of ESBL-producing Enterobacteriaceae infections, carbapenem antibiotics are usually selected for gram-negative bacterial infections before antimicrobial susceptibility test results are available. The abuse of carbapenem antibiotics has led to the emergence of carbapenem-resistant Enterobacteriaceae (CRE) which may be resistant to all available antibiotics.²³ Therefore, it is urgent for most NICUs to have an antibiotic stewardship program to reduce the abuse of broad-spectrum antibiotics. In the current study, we compared the clinical characteristics of a group of hospitalized neonates with culture-confirmed ESBL-producing E. coli sepsis with a group infected with ESBL-negative E. coli in order to determine the potential risk factors associated with ESBL-producing MDR E. coli infection among the neonates with confirmed E. coli sepsis. These results may inform the selection of appropriate empirical antibiotics before antimicrobial susceptibility results are available in the future.

Materials and Methods

Data Collection

This is an observational cohort study based on prospectively collected data on infants with positive blood culture results from the Second Affiliated Hospital and Yuying Children’s Hospital of Wenzhou Medical University. Starting in 2016, all infants with a positive blood or cerebrospinal fluid (CSF) culture have been recorded in our specific neonatal sepsis database. For the current study, all infants with a confirmed pathogen isolation from blood or CSF from 2016 to 2021 were identified and those with E. coli infection were included in the study for further analysis.

The detailed medical history and relevant clinical data of neonates with confirmed E. coli sepsis were retrospectively retrieved from the hospital electronic medical records. The information collected includes basic demographic information, perinatal factors, antimicrobial susceptibility tests, and pre-infection conditions. If a patient had multiple positive cultures during hospitalization, only the first result was included to avoid overrepresentation. Cases with incomplete data were excluded. EOS with E. coli is defined as the presence of E. coli grown from blood or CSF in cultures drawn in any newborn infant within 72 hours of birth. LOS is arbitrarily further divided into community acquired and hospital-acquired. Community acquired late-onset sepsis (CALOS) with E. coli infection is defined as the isolation of E. coli from blood or CSF from outpatients >72 hours after birth or inpatients >72 hours after birth and within 48 hours of re-admission into hospital. Hospital-acquired late-onset sepsis (HALOS) with E. coli infection is defined as the isolation of E. coli from blood or CSF from hospitalized infant >72 hours after birth. The definition of previous antibiotic use was that antibiotics were administered intravenously for more than 48 hours prior to culture collection. The Institutional Ethics Committee of the Second Affiliated Hospital and Yuying Children’s Hospital of Wenzhou Medical University approved the study protocol. A waver for parental consent to review medical records was granted by the Institutional Ethics Committee. The handling of the patient data confidentiality strictly followed the rules set by the institution and were in compliance with the Declaration of Helsinki.

Bacterial Identification and Antimicrobial Susceptibility Testing

The clinical laboratory in our hospital employs routine microbiological tests according to the standard set by the Clinical & Laboratory Standards Institute (CLSI, USA). Bacterial identification was accomplished using a MALDI-TOF/MS mass spectrometer (Bruck, Germany). Bacterial drug sensitivity testing was conducted using the Vitek2 Compact automated drug susceptibility analyzer (Mérieux, France) or disk diffusion method (30ug, British OXOID). The Gram-Negative Susceptibility card (BioMerieux, France) was utilized to determine antibiotic susceptibility of bacterial isolates. ESBL-producing strains were identified by the double-disk synergy method.²⁴ Results were interpreted following the recommendations and definitions of the CLSI.

Statistical Analysis

Statistical analysis and production of graphs were performed using SPSS Statistics (version 24; IBM Corp, Armonk, NY) and GraphPad Prism (version 9.0; GraphPad Software, La Jolla, CA, USA), respectively. The normality of continuous variables was tested using the Kolmogorov–Smirnov test. Normally distributed data were presented as the mean ± standard deviation (M±SD) and were analyzed using the Student’s t-test. Non-normally distributed data were described as median (25–75 percentile) and were analyzed using the Mann–Whitney U-test. Homogeneity test was used to check whether the variance of the data is homogeneous. For more than two groups, analysis of variance (ANOVA) was used for normal and homogeneous data, and Kruskal–Wallis H-test was used for non-normal data. When ANOVA results identified the statistical difference among the groups, post hoc tests were used to determine the significance between the particular pairs of means. Categorical variables were compared using the chi-square test or Fisher’s exact test. The level of significance was set at P < 0.05. Multivariable logistic regression was conducted to determine the risk factors associated with infection by an ESBL+ strains in neonates with confirmed E. coli infection. All relevant variables significantly associated with ESBL+ strains in the univariate analysis were entered into the model. A stepwise selection was used with an alpha-to-enter and alpha-to-exit set at 0.10 and 0.05, respectively. The odds ratio (OR) and 95% confidence intervals (CIs) were reported.

Results

During the study period, a total of 21,439 newborns were admitted to the NICU of the Second Affiliate Hospital and Yuying Children’s Hospital of Wenzhou Medical University, of which 233 were diagnosed with gram-negative bacterial sepsis. Of the 233 cases, 141 (60.5%) were confirmed to be E. coli sepsis. After careful review of the cases, 2 patients were excluded due to incomplete data, and the remaining 139 cases were included in the analysis. Among these, 69 cases were reported to be ESBL-producing and 70 cases were reported to be ESBL-negative E. coli, as shown as flow diagram in Figure 1.

Figure 1 Screening flowchart of research subjects.

Of the total 139 strains of E. coli identified from the blood or CSF samples sent from the neonates in our NICU from 2016 to 2021, the proportion of ESBL producing E. coli among all E. coli isolates each year is presented in Figure 2. It fluctuated and decreased to the lowest value of 31.6% in 2019. Since then, the proportion of ESBL-producing strains has gradually increased, rising to 66.7% in 2021.

Figure 2 The proportion of ESBL+ in all E. coli isolates from 2016 to 2021.

Based on the timing of initial culture and hospitalization status, all enrolled patients were divided into EOS, CALOS and HALOS groups for further analysis. As shown in Table 1, the gestational age and birth weight of infants in the CALOS group were significantly higher than those in the other two groups, while the proportions of in vitro fertilization (IVF), multiples and prolonged rupture of membrane (PROM) in the HALOS group were significantly higher than that in the other two groups (P<0.05). While all E. coli strains isolated from enrolled patients were still sensitive to Amoxicillin/clavulanic acid, amikacin, and carbapenem, the proportion of ESBL-producing stains was significantly higher in the HALOS group, and as the results, the antimicrobial resistance rates to gentamycin, ceftriaxone, ceftazidime and cefepime were also significantly higher in the HALOS group (Figure 3).

Table 1 The Basic Clinical Characteristics of the Three Types of Neonatal E. coli Sepsis

Figure 3 Antimicrobial resistance rate of E. coli isolated from three different types of neonatal sepsis. *P < 0.05.

Comparisons of the basic clinical characteristics between the ESBL-producing and ESBL-negative groups are presented in Table 2. As shown in the table, the birth weight and gestational age (GA) of children in ESBL-producing group were significantly lower than those in ESBL-negative group (P<0.05). There were no statistically significant differences between ESBL-producing and ESBL-negative groups in gender, mode of delivery, multiple births, maternal diabetes, maternal chorioamnionitis, low Apgar scores, EOS ratio, and exclusive breastfeeding (P>0.05). However, the proportion of pregnant women with age >35y, use of prenatal antibiotics, IVF, PROM, proportion of HALOS, and antepartum cephalosporins use in the ESBL-producing group were significantly higher than that in the ESBL-negative group (P<0.05). The proportion of CALOS in ESBL-producing group is lower, while that of HALOS is higher (P<0.05). Characteristics prior to infection indicated that babies in the ESBL-producing group had longer hospitalization prior to infection, were more likely to have central lines, had greater use of antibiotics, were more likely to be on mechanical ventilation, and more often receiving parental nutrition for greater than 1 week. For clinical complications and outcomes, the mortality of ESBL-producing group was significantly higher than that of ESBL-negative group, but there was no significant difference in the proportion of children with meningitis, septic shock between the two groups (Table 3).

Table 2 Comparison of the Basic Clinical Characteristics Between the Two Groups

Table 3 Comparison of Clinical Complications and Outcomes Between the Two Groups

To determine the risk factors associated with an ESBL-producing strain among all infants with confirmed E. coli sepsis, a multivariate logistic regression analysis was conducted. All relevant variables with P<0.1 in univariate analysis were included in the model. As shown in Table 4, antepartum exposure to cephalosporins (OR=25.191, 95% CI: 3.184–199.326, P=0.002) and parenteral nutrition for more than 1 week (OR=4.495, 95% CI: 2.009–10.055, P=0.000) were independent risk factors of neonatal infection with ESBL-producing stain among infants with E. coli sepsis. In this multivariate logistic regression model, all other variables included in the model appear to be covariates and, therefore, not independent risk factors with statistical significance.

Table 4 Logistic Regression Analysis of Risk Factors for Neonatal MDR E. coli Sepsis

Discussion

An increasing number of cases of neonatal sepsis due to ESBL-producing MDR E. coli have been reported worldwide.^5,6,18,19 In this study, we collected data on hospitalized neonates with culture-confirmed E. coli infection and analyzed the risk factors associated with ESBL-producing strains among the neonates with confirmed E. coli infection. We demonstrate that E. coli remains the most common Gram-negative bacterial pathogen for neonatal sepsis, which is similar to the reports both from China and other countries.^2,4 We further demonstrate that antepartum maternal exposure to cephalosporins and parenteral nutrition for longer than 1 week are strong and independent risk factors associated with neonatal infection with ESBL-producing stains among infants with culture-confirmed E. coli sepsis.

ESBL-producing E. coli represents a major worldwide threat among MDR bacteria in both hospital and community settings. ESBLs are often located on large plasmids that also harbor genes resistant to other antimicrobial classes, resulting in MDR isolates.^19,25 Our data show that the overall incidence of neonatal E. coli sepsis in our hospitalized neonates was 0.65% (141/21,439) in the last 6 years and that half of those were ESBL-producing strains. The mortality rate in ESBL+ group was significantly higher than that of the ESBL-negative group (24.6% vs 11.4%). These findings are consistent with other reports that the mortality of neonatal sepsis caused by ESBL producing Enterobacteriaceae is twice as high as that of all other bloodstream infections.^6,26,27 As a consequence, the increase in the prevalence of ESBL-producing Enterobacteriaceae infection is associated with significant increase in neonatal mortality.²⁸

Previous study demonstrated that in EOS, the MDR rate was higher in preterm infants when compared to that of full-term infants.⁵ In the current study, we found that E. coli isolated from HALOS had the highest proportion of ESBL-producing strains, followed by EOS. E. coli isolated from CALOS had the lowest rate of ESBL positivity. By comparing the different types of neonatal sepsis caused by E. coli, we found that the three groups had statistically significant differences in gestational age, birth weight, rates of multiples, IVF, antenatal antibiotic therapy, and PROM. Giannoni et al reported that HALOS occurred in lower gestational age infants and CALOS was more commonly seen in full-term infants and boys,²⁹ similar to the findings in our study. Antenatal antibiotic therapy and PROM were more common in EOS. Multiples and IVF, which usually associated with lower gestational age, were more common in HALOS. A recent multicenter study across China also showed similar ESBL positivity rate and patterns of E. coli strains among hospitalized neonates.³⁰

Univariate statistical analysis showed that there were differences in many variables between the ESBL-producing group and the ESBL-negative group. Babies in ESBL-producing group had a lower gestational age and birth weight, longer hospitalization time before infection, were more likely to have central lines, more often on mechanical ventilation, higher chance be on parenteral nutrition for more than 1 week, and received antibiotics before onset of sepsis. Babies in the ESBL-producing group were also more like to be a product of an IVF, to have mother’s age over 35 years, to be exposed to antepartum cephalosporins, and PROM. Most of the above factors were high-risk factors of nosocomial infection. Therefore, nosocomial infection was more common in the ESBL-producing group, while community infection was more common in the ESBL-negative group. Premature infants, especially those with very low birth weight, usually have prolonged stay in the NICU and are at increased risk for hospital-acquired infections. Most premature infants also have a history of antibiotic exposure. Abuse of third-generation cephalosporins and other broad-spectrum β-lactam antibiotics in NICUs can easily induce the production of ESBL in gram-negative pathogens, which leads to more MDR E. coli infections in premature infants. It has been reported that restricting overall cephalosporin use can significantly reduce the incidence of ESBL-producing gram-negative bacterial infection.³¹

Previous studies on risk factors related to ESBL-producing Enterobacteriaceae mainly focused on hospital-acquired infection or colonization. No comprehensive studies include analysis involving early onset neonatal sepsis and community-acquired sepsis.^32–35 A meta-analysis has revealed that many risk factors were related to infection and/or colonization with ESBL-producing bacteria in NICUs.³⁶ Interestingly, by using multivariate logistic regression analysis, we found that only prenatal cephalosporin use and prolonged use of parenteral nutrition were independent risk factors for MDR E. coli infection, indicating that most other factors that significantly differed between the ESBL-producing and ESBL-negative groups by univariate analysis were covariates. Indeed, the antenatal use of cephalosporins in mothers increases the possibility of colonization of MDR bacteria in the birth canal and increases the chance of infection of ESBL-producing E. coli in neonates both as EOS and LOS. A single-center study in Sri Lanka tested 159 pairs of maternal vaginal swabs and neonatal perianal swabs. Random amplification of polymorphic DNA analysis revealed that Enterobacteriaceae with same strain type was present in 6.9% of the mother-infant pairs, indicating possible transfer between the mother and neonate.³⁷ In addition, vaginal bacteria colonization by E. coli in mothers with PROM was significantly correlated with the microorganisms from the neonatal blood culture.³⁸ For newborns with prolonged need of parenteral nutrition, most of them have a central line and are at much higher risk for infections due to bacterial translocation from the gastrointestinal tract.³⁹ Restricting prenatal antibiotic use especially of cephalosporins and promoting active feeding strategies in premature infants to reduce the duration of parenteral nutrition may protect newborns from MDR E. coli sepsis.

There are certainly some limitations in our study. Limited by being single-center and observational cohort study, the risk factors selected for analysis are not comprehensive enough to include all possible factors. In addition, as a regional perinatal medical center, there are many premature infants and critical newborns which may not be universally representative of NICUs. In addition, the identification of specific drug resistance genes were not performed. A more representative multi-center study is warranted and should be conducted in the future.

Conclusion

E. coli remains the most common Gram-negative bacterial pathogen for neonatal sepsis. Higher proportion of ESBL-producing MDR E. coli are seen in premature infants and those newborns with HALOS. The mortality rate of neonatal sepsis caused by ESBL-producing E. coli is twice as high as that of sepsis caused by ESBL-negative E. coli. Furthermore, antepartum use of cephalosporins and prolonged use of parenteral nutrition may be important factors to consider in the selection of sensitive empirical antibiotics for use in neonatal sepsis caused by gram-negative rods prior to the availability of the results of antimicrobial susceptibility.