Mode of delivery and other risk factors for Escherichia coli infections in very low birth weight infants

The Polish Neonatal Surveillance Network (PNSN), whose director is Prof. Ewa Helwich from the Institute of Mother and Child in Warsaw, covered six Neonatal Clinics within the territory of Poland: Clinic of Neonatology and Intensive Neonatal Care, Institute of Mother and Child, Warsaw, Clinic of Neonatology, Jagiellonian University Medical College, Krakow, Clinic of Neonatology and Intensive Neonatal Care, Warsaw Medical University, Warsaw, Clinic of Neonatology, Polish Mothers’ Memorial Hospital-Research Institute, Lodz, Department of Neonatal Diseases, Pomeranian Medical University, Szczecin and Gynaecology and Obstetrics Hospital Medical University Poznan. The Polish Neonatal Surveillance Network was established on 6th of September 2007 (Agreement; see attachment). The Chair of Microbiology cooperated within the program of the PNSN. Utilization of data collected in the PNSN for scientific purposes was approved by the Bioethics Committee of Jagiellonian University Medical College (Chairperson Prof. Piotr Thor) – no. KBET/221/B/2011 ) on 27 October 2011 (Appropoval; see attachement). All data entered into the electronic database and analysed during this study were previously anonymised and de-identified.

A prospective surveillance of infections was conducted between 1 January 2009 and 31 December 2012 in five from six NICUs that participated in the activities of the PNSN.

The study covered 1,768 newborns. All episodes of infection were subjected to registration, regardless of the time of occurrence of the first symptoms. Case patients were defined according to Gastmeier et al. [11], with modifications for neonates with birth weight <1.5 kg. Early-onset infection (EOI) was defined as infection diagnosed within 3 days of delivery. The occurrence of chorioamnionitis was based on clinical data (all cases) and histopathology examination of the placenta (~65% of cases).

All isolates were tested using disk diffusion antimicrobial susceptibility methods on Mueller–Hinton agar plates according to the current EUCAST guidelines (European Committee on Antimicrobial Susceptibility Testing. Clinical breakpoint tables v. 3.1; http://​www.​eucast.​org v.3.1, accessed: 11.02.2013). ESBL activity was detected with a modified double disk synergy test using a combination of cefotaxime, ceftazidime, cefepime and aztreonam discs, placed 20 mm apart around the disc containing amoxicillin/clavulanic acid [12].

PFGE was used to determine the possible horizontal transfer of E. coli strains among patients. All isolates were analysed using the standardized PFGE protocol developed at the Centers for Disease Control and Prevention by the PulseNet program http://​ttp://​www.​cdc.​gov/​pulsenet/​pathogens/​ecoli.​html (accessed: 11.02.2013). Genomic DNA was digested with 10 U XbaI (ThermoScientific, ABO, Gdansk, Poland). The resulting DNA fingerprinting was analysed using the CHEF III PFGE system (BioRad, Warsaw, Poland) in 0.5 Tris–borate–EDTA buffer at 14°C at 6 V for 20 h with a ramped pulse time of 2.2–54.2 s. The GelCompar (Applied Maths) was used for cluster analysis using the Dice coefficient and the unweighted pair group method with arithmetic mean.

The influence of type of care and sociodemographic characteristics on the epidemiology of E. coli/other infection was analysed with several statistical techniques, depending on type and distribution of analysed variables. The relation between probability of E. coli and continuous parameters (age, length of stay) was based on simple analysis of variance (ANOVA). If the distribution of continuous variables significantly differed from normality, the nonparametric alternative to ANOVA, the Kruskal–Wallis test, was used. For the contingency of nominal characters frequency test: Pearson’s chi-square (χ²) and likelihood ratio was used. The common influence of risk factors on E. coli identification was analysed with a generalised linear model. Because of the categorical character of the effect and combined — numerical as well as categorical — types of the predictors, the model was constructed for binominal distribution of dependent variables and logit-linked function. A p value <0.05 was considered significant. All analyses were performed using JMP version 9.3 (SAS Institute Inc., Cary, NC, USA).

Neonates with infections of non-E. coli aetiology were characterised by better general condition at birth (Table 1) than neonates with E. coli infections. Significantly fewer infants had adverse perinatal outcomes, such as PROM or chorioamnionitis, while significantly more newborns were delivered by caesarean section. The fatality rate of newborns with non-E. coli infections was two times lower than in those with E. coli infections.

The incidence of E. coli infections was 5.4% and 2.0/1,000 patient-days. EOIs accounted for 12.2% of all infections and late-onset infections (LOIs) accounted for 10.5%. The incidence of EOIs was 1.7% and LOIs was 3.7%. The occurrence of E. coli infections (both EOI and LOI) depended significantly on the NICU (χ² = 73,836; p < 0.0001) and varied between 3.9% and 17.9%. The most common E. coli infection was pneumonia (53.1%) and bloodstream infection (40.6%). Escherichia coli infections were diagnosed at day 17 on average (median: 12 days).

The mean length of hospitalisation (from birth to discharge, or until a weight of 1.8 kg was reached) of newborns with non-E. coli infections was significantly shorter: 48 versus 55 days (infants with E. coli infection).

The gestational age and birth weight of newborns with E. coli infections were lower than in those without infection or with other infections (Table 1). The other risk factors of E. coli infections were: low Apgar 1 and Apgar 5 scores, total parenteral nutrition, chorioamnionitis, or PROM diagnosed during pregnancy or delivery (Table 1).

Caesarean section reduced the risk of E. coli infection, although perinatal antibiotic prophylaxis (PAP) (no standardised protocol) did not reduce the risk of infection; that is, despite the use of PAP, E. coli infections were observed significantly more frequently (Table 1).

Multivariate analysis that took into account the combined effect of intrapartum antibiotic prophylaxis (IAP), PROM, chorioamnionitis and delivery mode (caesarean section) showed no significant effect on the risk of E. coli infection for factors other than the mode of delivery. However, multivariate analysis that took into account the combined effect of demographic data (sex, gestational age and birth weight) and place of birth (NICU) showed that only the place of hospitalisation had a significant effect on E. coli infection risk.

Finally, a multivariate analysis of maternal influences on IAP, PROM, amniotic inflammation, type of labour, gestational age, NICU and the likelihood of specific E.coli infections (vs. other infections) showed a significant statistical relationship with the complete model.

Analysis of the isolates showed that the highest levels of resistance among all E. coli isolates were observed against ampicillin (88.8%) and amoxicillin/clavulanic acid (62.2%), trimethoprim/sulfamethoxazole (34.4%) and aztreonam (33.3%). The ESBL phenotype was found among 25 isolates (27.7%). Sixteen of the ESBL-positive strains (17.7%) were also reported as resistant to at least two other groups of antibiotics (fluoroquinolones, aminoglycosides or trimethoprim–sulfamethoxazole), which allows us to consider these strains as MDR organisms (Table 2).

Escherichia coli isolates showed that different pulsotypes and dominant epidemic clones were not detected. Cluster analysis based on PFGE of the 90 isolates showed 71 unique types, some of which were <70% similar, suggesting a genotypically variable population (Figure 1). Isolates that have identical pulsotypes usually were derived from the same patient (as in the case of 11 isolates) or were isolated from different patients of the same NICU in the same period of time (in the case of seven isolates). The location of the NICU and the site of the isolation did not appear to have a correlation in the cluster analysis.

Electronic database created as the result of continuous prospective targeted surveillance of infections was used in the study.

Participation of hospitals in PNSN was voluntary and confidential. Utilization of data collected in PNSN for the scientific purpose was approved by the Bioethics Committee of Jagiellonian University Medical College – no. KBET/221/B/2011. All data entered into the electronic database and analyzed during preparing this article were previously anonymized and de-identified. Those data were obtained under routine treatment and diagnostic procedures performed during patients’ hospitalization. No additional samples were collected for testing. According to Polish law, utilization this kind of data for scientific purpose does not demand patients’ agreement or even information that data are collected in the database.