Escherichia coli Is Overtaking Group B Streptococcus in Early-Onset Neonatal Sepsis

The widespread use of intrapartum antibiotic prophylaxis (IAP) to prevent group B streptococcus (GBS) early-onset sepsis (EOS) is changing the epidemiology of EOS. Italian prospective area-based surveillance data (from 1 January 2016 to 31 December 2020) were used, from which we identified 64 cases of culture-proven EOS (E. coli, n = 39; GBS, n = 25) among 159,898 live births (annual incidence rates of 0.24 and 0.16 per 1000, respectively). Approximately 10% of E. coli isolates were resistant to both gentamicin and ampicillin. Five neonates died; among them, four were born very pre-term (E. coli, n = 3; GBS, n = 1) and one was born full-term (E. coli, n = 1). After adjustment for gestational age, IAP-exposed neonates had ≥95% lower risk of death, as compared to IAP-unexposed neonates, both in the whole cohort (OR 0.04, 95% CI 0.00–0.70; p = 0.03) and in the E. coli EOS cohort (OR 0.05, 95% CI 0.00–0.88; p = 0.04). In multi-variable logistic regression analysis, IAP was inversely associated with severe disease (OR = 0.12, 95% CI 0.02–0.76; p = 0.03). E. coli is now the leading pathogen in neonatal EOS, and its incidence is close to that of GBS in full-term neonates. IAP reduces the risk of severe disease and death. Importantly, approximately 10% of E. coli isolates causing EOS were found to be resistant to typical first-line antibiotics.


Introduction
Early-onset sepsis (EOS) is still a serious and potentially fatal neonatal complication, particularly among neonates of the lowest gestational age; as such, identifying and managing infants at risk for EOS is part of the routine care provided by neonatal clinicians.
The epidemiology of neonatal sepsis is a changing landscape [1,2]. In particular, the widespread implementation of intrapartum antibiotic prophylaxis (IAP) to reduce the perinatal transmission of group B streptococcus (GBS) has resulted in a substantial decline in the incidence of EOS in high-income countries, to approximately 0.3-1 per 1000 live births (LBs) [3][4][5][6][7][8][9][10]. However, concerns have been raised regarding the potential of a subsequent increased risk of EOS due to non-GBS pathogens. In fact, previous studies have revealed increased rates of EOS due to Escherichia coli (E. coli) among pre-term and very-low birth weight (VLBW) infants [2,[11][12][13][14][15][16][17][18]. Such a change would be alarming, as Gramnegative infections in newborns may present with high disease severity and case fatality rates [11,[19][20][21]. The increase in IAP exposure of term and pre-term neonates has also raised concerns regarding the potential emergence of antimicrobial resistance in pathogens (including Gram-negative bacteria) to first-line, highly effective beta-lactam antibiotics. Finally, perinatal antibiotics may alter the development of the gut microbiota [22], and the consequent intestinal dysbiosis may affect the susceptibility later in life to diseases mediated by the immune system. Indeed, previous studies have demonstrated that the use of antibiotics during infancy correlates with the occurrence of asthma, atopic dermatitis, multiple sclerosis, and intestinal bowel diseases [23].
Thus, a better understanding of the changing epidemiology of EOS is needed to optimize prevention and treatment strategies. Changes in the burden of GBS and E. coli EOS-the two main pathogens causing EOS in high-income countries [3,5,9]-should be monitored over time. However, the number of cases at individual facilities remains low, preventing an accurate assessment of trends. Prospective area-based Italian data on EOS due to GBS or other pathogens are still lacking, although a large amount of data have been obtained on GBS invasive infections from a Northern region of Italy (Emilia-Romagna), a region where active area-based surveillance of GBS has been promoted since 2003 [24]. Starting from January 2016, the pre-existing GBS surveillance was associated with that of E. coli. For this reason, an area-based, five-year prospective exploratory study, including a birth cohort of approximately 160,000 LBs, was conducted to describe GBS-and E. colispecific EOS incidence trends from 2016 to 2020. In addition, the characteristics of GBS and E. coli EOS, including antibiotic susceptibilities and outcomes, were compared.

Study Design
An antenatal screening-based strategy for the prevention of GBS EOS [25,26] has been implemented in the entire region since 2003. The surveillance network includes eight microbiological laboratories, ten level 1 centres (<1000 LBs/year; inborn criteria: ≥2000 g, ≥35 weeks), four level 2 centres (>1000 LBs/year; inborn criteria: ≥1500 g, ≥32 weeks), and eight neonatal intensive care units (no restrictions for in-and out-born neonates). Prospective area-based surveillance for GBS and E. coli was carried out among LBs of all gestational ages (GAs) in a Northern region of Italy (Emilia-Romagna,~four million people and 35,000 LBs/year). Patients were included if they had a blood or cerebrospinal fluid culture yielding GBS or E. coli within the first 3 postnatal days. To identify all cases, the prospective surveillance involved their reporting through a dual source. Microbiological laboratories were asked to report prospectively (from 1 January 2016 to 31 December 2020) each positive isolate from live-born infants. Clinicians were asked to complete a standardized form for each case of culture-proved EOS identified by the laboratory. In addition, to retrieve missed cases, all laboratories in the network were contacted at the end of each year for a detailed list of all registered cases. Incidence rates were calculated by using the number of live births, preterm deliveries (23-36 weeks' gestation) and very low birth weight neonates (BW, <1500 g, 1501-2500 g, and >2500 g) provided by the Regional Health Agency.
The Ethical Committee of the coordinating centre (Azienda Ospedaliero-Universitaria Policlinico di Modena; Prot. 910/2020) approved the project. Case reporting and isolate collection were determined to be non-research public health surveillance. To maintain patient confidentiality, spreadsheets submitted to the principal investigator were fully anonymous and did not include any identifiable data of patients or caregivers. Therefore, according to the policy of our ethics committee review board, patient consent was not required. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines were followed for this study.

Clinical and Microbiological Practices
According to the individual centre policy, a single (1 mL) or double blood culture was obtained before antibiotic treatment from each neonate with clinically suspected EOS; some centres also obtained blood culture in asymptomatic infants due to maternal risk factors (RFs) for EOS (delivery at <37 weeks' gestation; GBS bacteriuria; previous infant with GBS infection; rupture of membranes (ROM) ≥ 18 h before delivery; and intrapartum fever > 38 • C). Cerebrospinal fluid (CSF) culture was recommended in any case of suspected EOS but was deferred in unstable infants. Isolates were identified by MALDI-TOF MS, using Maldi Biotyper (Bruker Daltonics, Bremen, Germany) or Vitek-MS platforms (BioMérieux, Marcy l'Etoile, France). The capsular type of GBS isolates was identified by serological and molecular methods [27]. Antimicrobial susceptibility testing was performed using Vitek2 (BioMérieux, Marcy l'Etoile, France), MicroScan Walkaway (Beckman Coulter, Brea, CA, USA), or Phoenix (Becton Dickinson, Franklin Lakes, NJ, USA). Results were interpreted according to EUCAST clinical breakpoints.
Ampicillin plus an aminoglycoside (usually gentamicin) was the initial empirical antimicrobial therapy recommended for suspected EOS. For each case of culture-proven EOS, maternal and neonatal information (both from delivery and case medical records) were detailed by surveillance officers using a standardized form. Maternal information included data on pre-natal GBS screening, delivery, IAP exposure, and RFs for EOS. Collected newborn data included infecting organism, antimicrobial susceptibility, antibiotic treatment, and outcome (e.g., death, brain lesions, age at discharge from hospital).

•
Asymptomatic bacteraemia: positive blood culture obtained due to maternal RFs for EOS in an infant who remained asymptomatic. • Pneumonia: positive blood culture associated with respiratory distress syndrome and a characteristic radiographic appearance [9].
• Brain lesions: lesions confirmed by brain ultrasound or magnetic resonance imaging (MRI) in a newborn with clinical indications.

•
Severe disease: any of the following: apnoea at onset; need for fluid resuscitation, catecholamine support, mechanical ventilation, or exchange transfusion; pneumonia; meningitis; brain lesions at hospital discharge; death. • EOS-related death: death occurring within the first 7 postnatal days.

Statistical Analysis
Continuous variables are expressed using mean and standard deviation (SD) or median and interquartile range (IQR); binary and categorical data are reported as frequencies and percentages. The Student's t-test was used for unadjusted comparisons of continuous variables between groups; Pearson's χ 2 test and Fisher's exact test (when one or more cells had expected frequency < 5) were used for unadjusted comparisons of categorical variables between groups. Odds ratios (OR) and relative 95% confidence intervals (95% CI) were computed based on estimated parameters and variances from logistic regression models. Multi-variable logistic regressions were performed to adjust the results for GA.
During the 5-year surveillance period, the GBS-and E. coli-specific incidence rates, and their cumulative incidence, did not change (p = 0.99, p = 0.17, and p = 0.40, respectively; Figure 1). Overall, the rates of EOS among VLBW neonates were 40-fold higher compared to those with BW > 2500 g (OR 39.44, 95% CI 21.01-70.98; p < 0.001). EOS rates among VLBW neonates varied according to the pathogen, and E. coli rates were 16-fold higher than that of GBS among neonates with BW < 1500 g (p < 0.001). E. coli rates were also higher in neonates with BW 1501-2500 g (p = 0.03). In contrast, rates of E. coli and GBS EOS were comparable among infants with a BW > 2500 g (p = 0.35; Figure 2). Similarly, there was evidence of a trend toward increasing incidence among the low GA groups (      The demographics and characteristics of the entire population (overall, E. coli, and GBS cases) are summarized in Table 1. Cases of E. coli EOS were more likely to have a lower median BW and GA, as well as to be delivered with a VLBW. Their mothers exhibited a higher occurrence of histological chorioamnionitis, whereas prolonged ROM was borderline significant. Mothers of GBS EOS cases were more likely to present with a positive antenatal GBS screening, while cases of E. coli were more likely to be IAP-exposed and to receive an adequate IAP. Pre-term infants were more likely to be exposed to adequate IAP (pre-term neonates 50.0% vs. full-term neonates 21.1%; p = 0.018) and to be affected by E. coli. However, after adjustment for GA, neonates exposed to adequate IAP remained nine-fold more likely to be infected by E. coli, compared to GBS (OR 9.32, 95% CI 1.76-49.50; p = 0.009).
Among the overall cases of EOS, culture-proven meningitis was rare (two cases), but less than half of neonates in the entire cohort underwent a lumbar puncture. Three cases of E. coli EOS developed brain lesions. Two of these three (at 25 and 31 weeks of gestation, respectively) developed post-haemorrhagic hydrocephalus and focal periventricular white matter injuries, respectively. One infant was full-term (40 weeks of gestation) and  The demographics and characteristics of the entire population (overall, E. coli, and GBS cases) are summarized in Table 1. Cases of E. coli EOS were more likely to have a lower median BW and GA, as well as to be delivered with a VLBW. Their mothers exhibited a higher occurrence of histological chorioamnionitis, whereas prolonged ROM was borderline significant. Mothers of GBS EOS cases were more likely to present with a positive antenatal GBS screening, while cases of E. coli were more likely to be IAP-exposed and to receive an adequate IAP. Pre-term infants were more likely to be exposed to adequate IAP (pre-term neonates 50.0% vs. full-term neonates 21.1%; p = 0.018) and to be affected by E. coli. However, after adjustment for GA, neonates exposed to adequate IAP remained nine-fold more likely to be infected by E. coli, compared to GBS (OR 9.32, 95% CI 1.76-49.50; p = 0.009). Among the overall cases of EOS, culture-proven meningitis was rare (two cases), but less than half of neonates in the entire cohort underwent a lumbar puncture. Three cases of E. coli EOS developed brain lesions. Two of these three (at 25 and 31 weeks of gestation, respectively) developed post-haemorrhagic hydrocephalus and focal periventricular white matter injuries, respectively. One infant was full-term (40 weeks of gestation) and developed a focal frontoparietal ischemic lesion. Four of five case fatalities were due to E. coli, three of which were ampicillin-resistant. Four of five who died were born pre-term (at 23, 23, 24, and 25 weeks of gestation, respectively). Among infants with GA ≥ 34 weeks, the risk of death was significantly lower (1/45, 2% vs. 4/19, 21%; p = 0.02). When adjusted for GA, the risk of death was 96% lower in IAP-exposed compared to unexposed infants, both including the whole cohort (OR 0.036, 95% CI 0.00-0.70; p = 0.03) or when considering only cases of E. coli EOS (95% lower in IAP-exposed compared to unexposed infants; OR 0.05, 95% CI 0.00-0.88; p = 0.04).

Factors Associated with Severe Disease
RFs associated with severe disease are shown in Table 2. From univariable models, we selected variables with p-value < 0.20. In the multi-variable analysis, BW and prolonged ROM were excluded due to potential problems of multi-collinearity, as suggested by the variance inflation factors (VIFs) (Table 3, Figure 4). The lack of IAP administration remained associated (p = 0.03) with severe disease (area under ROC curve = 0.80, H-L = 0.99, sensitivity 78%, specificity 71%).    Table 4 provides information concerning the antimicrobial susceptibility of pathogens. GBS isolates were universally susceptible to penicillin/ampicillin, vancomycin, and cefotaxime, while approximately one-third of them were resistant to clindamycin and erythromycin. Only eleven GBS isolates out of 64 EOS cases were available for further typing analysis. This small set comprised three strains resistant to erythromycin and clindamycin; two of them were serotype III and one was serotype Ib.

Antimicrobial Susceptibility
Most E. coli isolates (70%) were resistant to ampicillin. No significant differences were identified in ampicillin resistance between infants who received IAP or not (17 of 23 tested, 74% vs. 6 of 10 tested, 60%; p = 0.44). All E. coli isolates were susceptible to amikacin, 10% were resistant to gentamicin, and all were resistant to ampicillin. All E. coli isolates tested were susceptible to meropenem, while a minority were resistant to cefotaxime.    Table 4 provides information concerning the antimicrobial susceptibility of pathogens. GBS isolates were universally susceptible to penicillin/ampicillin, vancomycin, and cefotaxime, while approximately one-third of them were resistant to clindamycin and erythromycin. Only eleven GBS isolates out of 64 EOS cases were available for further typing analysis. This small set comprised three strains resistant to erythromycin and clindamycin; two of them were serotype III and one was serotype Ib. Most E. coli isolates (70%) were resistant to ampicillin. No significant differences were identified in ampicillin resistance between infants who received IAP or not (17 of 23 tested, 74% vs. 6 of 10 tested, 60%; p = 0.44). All E. coli isolates were susceptible to amikacin, 10% were resistant to gentamicin, and all were resistant to ampicillin. All E. coli isolates tested were susceptible to meropenem, while a minority were resistant to cefotaxime.

Discussion
The incidence, microbiology, and mortality associated with GBS and E. coli EOS using Italian area-based surveillance data from 2016 to 2020 were considered in this study, thus updating information given in 2009-2012 [9] for the same area, where universal antenatal screening for GBS prevention has been greatly implemented over the years [30,31]. Compared to 2009-2012, cases of GBS EOS have significantly decreased (from 0.28 to 0.16 per 1000 LBs), as well as fatality rates associated to E. coli EOS. Cases of GBS slightly outweighed E. coli among full-term infants. Declines in GBS EOS have also been previously reported from other countries with robust screening policies in place [3,6,8,13,[32][33][34]. However, unlike the result of a large surveillance study carried out in the USA [3], we observed a rise in the burden of E. coli EOS. We wonder whether rates of E. coli EOS have actually increased, as the prospective active surveillance in 2009-2012 did not cover E. coli, and cases were retrieved only from laboratory databases, through a retrospective study design [9].
The proportion of EOS cases due to GBS and E. coli tends to vary across studies, associated with the prevention strategy in place. GBS rates exceeded those of E. coli by at least two times in three recent European multi-centre studies from countries implementing a risk-based strategy [35][36][37]. In contrast, GBS slightly outweighs E. coli in countries (e.g., USA and Switzerland) where a screening-based strategy for GBS prevention has been greatly enhanced over the years [3,6,8]. With respect to our previous study [9], we found that E. coli is overtaking GBS as the leading pathogen causing EOS, although the 95% CIs around the incidence rates for GBS and E. coli overlapped. A similar finding has also been reported in a recent multi-centre non-area-based study in the USA [5]. As a result of the continuous decline in GBS EOS, E. coli is now emerging as an important pathogen among term infants. Our data reveal the urgent need for novel preventive strategies for E. coli EOS, addressing both pre-term and term infants. Similarly to GBS, high-risk women should be identified as a target for prevention [8]. Other investigators have found that ROM, especially in pre-term mothers, is more common in E. coli EOS than GBS [3]; thus, when EOS is suspected in pre-term neonates, prolonged ROM should immediately raise suspicion of E. coli.
Neonates exposed to adequate IAP had a nine-fold lower risk of GBS EOS, compared to E. coli EOS. Exposure to adequate IAP remained associated with E. coli EOS after controlling for GA. The overrepresentation of adequate IAP exposure among infants with E. coli, compared to those with GBS EOS, confirmed that beta-lactam IAP regimens (most typically penicillin or ampicillin) are effective in preventing GBS EOS, but are far less effective in preventing E. coli EOS. Furthermore, considering limitations due to the small number of fatal cases, exposure to IAP of any duration reduced the risk of death by more than 95% in the whole population of EOS cases, as well as when considering only E. coli infections, independently of GA. Finally, IAP of any duration remained associated with a lower risk of severe disease in the multi-variable analysis, after adjusting for GA, ethnicity, positive GBS screening, histological chorioamnionitis, and Apgar score. Apparently, IAP exposure reduces the severity of both GBS and E. coli EOS. This finding is in conflict with a previous systematic review, which excluded the significant effects of IAP on mortality from GBS and non-GBS infections [34]. Given the rarity of EOS-related deaths, the effects of IAP are difficult to assess. At present, it seems unethical to conduct randomized controlled trials comparing cohorts of neonates exposed or unexposed to IAP administration, as IAP is now a standard of care. However, the effects of IAP may be revealed through large cohort studies carried out in settings where adherence to IAP is strict.
Despite widespread use of IAP, GBS isolates remained universally susceptible to ampicillin, while E. coli resistance to ampicillin remained stable when compared to that in the period 2009-2012 [9]. Our data were consistent with those recently reported in the USA, where two-thirds of E. coli isolates responsible for EOS were ampicillin-resistant, particularly among IAP-exposed neonates [3]. In contrast, ampicillin resistance was comparable between IAP-exposed and -unexposed neonates in our study. Resistance to gentamicin was rare (10%) but notable, being always associated with ampicillin resistance. Our data were also in agreement with European surveillance data [38], where antimicrobial resistance was shown to remain at high levels (especially for gram-negative bacteria) during 2016-2020. Nevertheless, all our E. coli isolates tested were susceptible to meropenem.
As all GBS isolates tested were susceptible to ampicillin and 90% of E. coli isolates tested were susceptible to gentamicin, the recommended first-line antibiotics for EOS (ampicillin plus gentamicin) [39,40] seem appropriate. Nonetheless, it is a major concern that approximately 10% of E. coli isolates causing EOS were resistant to both ampicillin and gentamicin. Such resistance to both first-line antibiotics was consistent with data from the USA [5,41] and should be taken into consideration when neonates become worse despite the suggested empirical treatment. Broad-spectrum empirical antibiotics should be considered in selected, critically-ill neonates until culture results are available (i.e., VLBW infants born after prolonged ROM and exposed to prolonged courses of antepartum antibiotics) [5,9,40]. As all E. coli isolates tested were susceptible, amikacin could replace gentamicin in certain cases. Continued surveillance of trends in the pathogens responsible for EOS and their antimicrobial susceptibilities is warranted, in order to ensure that ampicillin and gentamicin remain appropriate empirical therapies.
This study had several limitations. First, cases of EOS were relatively few, possibly reducing the significance of some results. Indeed, the present study was intended to be exploratory research. Even though our results succeeded in identifying key associations and suggest further investigations, the results displayed were not corrected for multiple testing, potentially exposing the data to alpha error inflation. Secondly, due to its retrospective design, we were unable to retrieve complete information for all cases of EOS, and some data may have been biased. Finally, EOS cases were defined only in terms of positive blood cultures; this definition may be useful in epidemiological studies for benchmarking among centres, but gives insufficient information on the severity of sepsis.

Conclusions
E. coli was found to now be the leading pathogen in pre-term neonatal EOS, and its incidence was also close to that of GBS in full-term neonates. IAP was associated with a reduced risk of death and severe disease in GBS and E. coli EOS. Strategies for the prevention and management of E. coli EOS, especially among pre-term infants, should be considered. As the landscape of EOS microbiology and antimicrobial susceptibilities continues to evolve, ongoing surveillance of cases of E. coli and GBS infections is crucial. Funding: This research received no external funding.

Data Availability Statement:
The data that support the findings of this study are openly available (attachment to this submission).