European Journal of Pediatrics

, Volume 171, Issue 4, pp 681–687

Late-onset neonatal infections: incidences and pathogens in the era of antenatal antibiotics


  • Capucine Didier
    • Service de Pédiatrie 2, Hôpital de HautepierreCentre Hospitalier Universitaire de Strasbourg
  • Marie-Pierre Streicher
    • Service de Pédiatrie, Centre Hospitalier de Haguenau
  • Didier Chognot
    • Service de Pédiatrie, Hôpital le Parc, Centre Hospitalier de Colmar
  • Raphaèle Campagni
    • Service de Pédiatrie, Centre hospitalier de Mulhouse
  • Albert Schnebelen
    • Service de Pédiatrie, Centre Hospitalier de Saverne
  • Jean Messer
    • Service de Pédiatrie 2, Hôpital de HautepierreCentre Hospitalier Universitaire de Strasbourg
  • Lionel Donato
    • Service de Pédiatrie 2, Hôpital de HautepierreCentre Hospitalier Universitaire de Strasbourg
  • Bruno Langer
    • Service de Gynécologie et Obstétrique, Hôpital de HautepierreCentre Hospitalier Universitaire de Strasbourg
  • Nicolas Meyer
    • Service de Biostatistique et MéthodologieCentre Hospitalier Universitaire de Strasbourg
  • Dominique Astruc
    • Service de Pédiatrie 2, Hôpital de HautepierreCentre Hospitalier Universitaire de Strasbourg
    • Service de Pédiatrie 2, Hôpital de HautepierreCentre Hospitalier Universitaire de Strasbourg
    • Médecine et Réanimation néonatale, Service de Pédiatrie 2, Hôpital de HautepierreCentre Hospitalier Universitaire de Strasbourg
Original Article

DOI: 10.1007/s00431-011-1639-7

Cite this article as:
Didier, C., Streicher, M., Chognot, D. et al. Eur J Pediatr (2012) 171: 681. doi:10.1007/s00431-011-1639-7


Widespread use of intrapartum antimicrobial prophylaxis has significantly reduced the incidence of early-onset neonatal infection (EONI); however, little is known about the effects of increased maternal exposure to antibiotics on late-onset neonatal infection (LONI). This study aims to evaluate LONI epidemiology in our region after the application of French recommendations and to determine whether LONI-causing organisms and their antibiotic susceptibility are influenced by peripartum antibiotic exposure. We performed a prospective epidemiologic study of 139 confirmed and possible cases of bacterial LONI in patients treated with antibiotics for at least 5 days of the 22,458 infants born in our region in the year 2007. The overall incidence of LONI caused by all pathogens, Group B streptococcus (GBS) and Escherichia coli (E. coli) were 6.19, 0.36 and 2.72, respectively, per 1,000 live births. Our findings revealed three major types of LONI: E. coli-induced urinary tract infection (UTI) among term infants, coagulase negative Staphylococcus septicemia affecting preterm infants, and GBS infections with severe clinical presentation. Univariable analysis revealed that maternal antibiotic exposure was significantly associated with the risk of amoxicillin-resistant E. coli infection (p = 0.01). Postnatal antibiotic exposure was associated with an increased risk of E. coli LONI (p = 0.048). This link persisted upon multivariable analysis; however, no additional risk factors were identified for LONI caused by antibiotic-resistant E. coli. Conclusion Our findings confirm that despite the benefits of antenatal antibiotics, this treatment can increase the risk of antibiotic-resistant cases of LONI. National and international surveillance of LONI epidemiology is essential to assess benefits and potential negative consequences of perinatal antibiotic exposure.


Late-onset neonatal infectionGroup B streptococcusEscherichia coliStaphylococcusPeripartum antibiotic treatmentAntibiotic resistance


In contrast to the numerous studies of early-onset neonatal infection (EONI), few epidemiological studies have focused on potential late-onset neonatal infection’s (LONI) modifications due to maternal antibiotic exposure. These infections, which typically affect infants at 4 to 90 days of life [18, 28], remain an important public health concern and are associated with high mortality and morbidity, especially among preterm newborns and very low birth weight (VLBW) infants.

Although the incidence of EONI has decreased from 1.7 to 0.24 per 1,000 live births in the USA [2, 3, 11] after widespread use of intrapartum antimicrobial prophylaxis (IAP) became common, the incidence of LONI has remained stable with a medium of 0.35 cases per 1,000 live births [12]. Some studies of EONI have been conducted in France [19, 22, 25], but no epidemiological studies of LONI have been performed to date.

In 2001, the “High Health Authority” published evidence-based recommendations [8, 9, 19] to prevent vertical transmission of Group B streptococcus (GBS). These concerned intrapartum antibiotic prophylaxis as a means to prevent GBS disease (using a combined risk-based and screening-based approach) and antenatal antibiotics for preterm premature rupture of membranes (defined as rupture of membranes before 37 weeks of gestation and before labour). The risk-based strategy indicates the use of intrapartum antibiotic prophylaxis if one of the following risk factors is present: positive bacteriuria for GBS during pregnancy, history of neonatal GBS infection, labour before 37 weeks, membranes ruptured for more than 12 h or a maternal temperature over 38°C during labour. The screening approach consists in a vaginal swab between 34 and 38 weeks of gestation and indicates antibiotic prophylaxis in case of positive smear. The implementation of these recommendations, which are widely used in Alsace, has resulted in increased maternal antibiotic treatment before and during labour. Between 18% and 25% of all American mothers and as many as 30% of European mothers at some centres [25] are exposed to antenatal maternal antibiotics. This practice raises questions about the possible link between antibiotic exposure and LONI. Glasgow et al. reported that infants with serious LONI were more likely than non-infected control infants to have been exposed to intrapartum antibiotics. The pathogens responsible for serious cases of bacterial LONI were more likely to be resistant to ampicillin when the infant had been exposed to broad spectrum IAP [16]. Moreover, postnatal antibiotic treatment could also impact the incidence of LONI. In France, Labenne et al. found that excessive duration of antibiotic therapy for suspected EONI was independently associated with an increased risk of LONI [23].

Therefore, the goals of this prospective observational study were: (1) to collect epidemiological data about LONI in Alsace, a northeastern region of France and (2) to determine whether the organisms causing LONI and their antibiotic susceptibility were influenced by peripartum antibiotic exposure.


We performed a population-based prospective observational study of all infants aged 4 to 90 days who were treated for bacterial LONI at one of the nine level II and III perinatal centres in Alsace, France between January 1, 2007 and December 31, 2007. Intrapartum, clinical and microbiological data were prospectively obtained from medical records and sent monthly by every local investigator to the coordinating centre. At the end of the study period, the completeness of the reporting of LONI was verified using the “Information System of Medicalization Program” at each centre.

Several LONI definitions are available. Although the neonatal period ends at 28 days, some authors considered that the occurrence of neonatal infections is possible until 90 days of life [5, 6, 12, 14]. Therefore, according to literature data, we defined LONI as any bacterial infection acquired between 72 h and 30 days of life, or a GBS or Escherichia coli infection acquired between 31 and 90 days of life. Infants born before 37 weeks of gestational age (GA) were considered preterm. Low and very low birth weight were defined as birth weight less than 2,500 g and 1,500 g, respectively. Infections that occurred more than 48 h after admission were considered to be of nosocomial origin.

We included infants that presented with the three following conditions: (1) two or more clinical signs of infection and/or positive blood, cerebrospinal fluid (CSF) and/or urine cultures revealing the presence of a pathogenic bacteria; (2) biological inflammatory syndrome defined by an increased C-reactive protein level ≥10 mg/l at 24-h intervals; and (3) antibiotic therapy for at least 5 days, except in the event of death. Considering restricting policies about antibiotic exposure, we expected that antibiotic treatment was stopped at 72 h after negativity of bacterial cultures.

Infants were classified as having either confirmed (i.e., bacteriological proven) or possible (i.e., clinical sepsis) LONI, and the infection was confirmed when a bacteriological smear revealed a positive blood, CSF, urine and/or skin lesion culture. Patients were suspected of having possible LONI when they presented clinical signs of infection, an increased level of C-reactive protein 12 to 60 h after the appearance of clinical signs, no alternative diagnosis and a favourable outcome associated with at least 5 days of antibiotic treatment, except in the event of death [4]. Final diagnostic classifications were determined by two senior physicians who were not involved in the care of the newborns (C.D. and P.K.).

Clinical signs of infection included isolated fever or temperature instability, hemodynamic signs (e.g., tachycardia, bradycardia, capillary refilling time >3 s, hypotension, and oliguria), respiratory signs (e.g., grunting, tachypnea or apnea, respiratory distress or an increase in the requirements for ventilator support), neurologic signs (e.g., bulged fontanel, lethargy, hypotonia or seizures), gastrointestinal signs (e.g., abdominal distension or gastric residuals) or cutaneous signs (e.g., vesicle or redness of central line orifice).

According to French guidelines, the microbiological techniques used were as follows:
  • A blood culture was obtained before any treatment initiation. The culture required a blood volume of 1 ml and was incubated for 72 h according to recommendations about time of positivity of neonatal blood cultures [17].

  • A CSF culture was obtained after lumbar puncture in newborns with neurological signs or positive blood cultures of a particular pathogen.

  • Urine cultures were obtained with non-invasive techniques using a urine collection pad after rigorous disinfection of the perineal area. Invasive techniques using a catheter sample or a suprapubic aspiration with ultrasound guidance are not commonly used in France for the first line diagnosis of a Urinary tract infection (UTI). Sterile urine sample was sent to the laboratory for microscopy and culture.

When CoNS was isolated, the diagnosis of LONI was confirmed when clinical or local skin signs of infection were present and/or C-reactive protein levels increased 12 to 60 h after the appearance of clinical signs [15]. UTI was confirmed when a positive urine culture [>10 leucocytes/mm3 and 105 colony-forming units (CFU)/ml] was associated with clinical and biological inflammatory syndromes [10]. Skin infection was confirmed when a skin lesion was associated with a positive skin bacteriological smear. Written parental consent was not considered necessary because data were collected anonymously and because the study was classified as a regional evaluation program approved by the Eastern Neonatology Study Group.

Differences between continuous variables were assessed using the Student or Mann–Whitney test. Differences between categorical variables were assessed using the Chi-square test or the Fisher’s exact probability. Univariable and multivariable analyses were performed to assess the significance of the data. P-values <0.05 were considered statistically significant.


Study population

During the study period, 22,336 mothers delivered and 22,458 neonates were born alive at participating centres in Alsace. A total of 139 infants were included in the study population. Of these, 110 (79.1%) had confirmed infections and 29 (20.9%) possible infections. A total of 53 infants (38.1%) were preterm, with 41 (29.4%) born before 32 weeks GA and 12 (8.6%) born between 32 and 37 weeks GA. The median [range] age, gestational age and birth weight at the time of LONI diagnosis were 20 days [4,89], 33 weeks [24,42] and 3,007 g [525,4760], respectively.

Perinatal data

A total of 58.3% of the mothers of infected infants were screened for GBS, and 21.7% had positive smears. A total of 19.6% of mothers received antenatal antibiotic therapy including 57.1% for maternal GBS carriage and 42.9% for preterm premature rupture of membranes (PPROM). Amoxicillin was used in 81% of mothers, macrolide in 9.5% and penicillin G in the remaining 9.5%.

Among the 119 newborns for whom complete data were available, 33 (27.7%) were born with risk factors for EONI, five were exposed to maternal GBS, eight experienced threatened preterm labor, 11 were born after PPROM, two had suspected chorioamnionitis, and seven experienced other conditions. A total of 17 infants (14.3%) received postnatal antibiotic therapy early after birth. Cefotaxime and amikacin were administered to four infants (23.5%), and the remaining 13 (76.5%) infants were treated with both amoxicillin and amikacin.

Epidemiology, diagnosis and prognosis of LONI

The distribution of the causative pathogens in confirmed cases of LONI 1 (n = 110) and their main diagnoses are presented in Tables 1 and 2, respectively. The remaining 29 infants (20.9%) had possible infections. The incidences of LONI are presented in Table 3.
Table 1

Distribution of pathogens isolated from newborns with confirmed LONI


Total no (%)

Group B streptococcus

8 (7.3%)

Escherichia coli

61 (55.5%)

Staphylococcus aureus

14 (12.7%)

Coagulase-negative staphylococcus

15 (13.6%)


12 (10.9%)

Table 2

Type of infection of term and preterm infants

Type of infection
















Cutaneous infection




Table 3

Incidence rates of confirmed and possible cases of LONI per 1,000 live births

Confirmed LONI overall




 E. colib






Possible LONI overall


Confirmed plus possible LONI overall


aGroup B streptococcus

bEscherichia coli

cStaphylococcus aureus

dCoagulase-negative staphylococcus

The three major types of LONI identified in the study included:
  1. 1.

    E. coli-induced UTI, which presented as an isolated fever in most (76.7%) cases and affected term infants at a median [range] age of 20 [4,89] days (43% under 28 days). Cases of E. coli-induced LONI were statistically more common in term neonates (p < 0.001).

  2. 2.

    CoNS septicemia and cutaneous infections affecting preterm infants presenting with hemodynamic signs in 22.2% of the infants with septicemia. One third of the entire study population had a central venous line with a median time [extreme] between catheter insertion and infection of 11.8 [±12] days. Cases of Staphylococcus-induced LONI were significantly associated with low gestational age (<32 weeks GA) and VLBW (p < 0.001).

  3. 3.

    GBS infections affecting term infants, typically presenting with severe clinical symptoms (n = 8) such as meningitis (n = 4) and/or hemodynamic signs (n = 5).


A total of 129 infants (92.8%) had a favourable outcome and did not experience infectious relapse upon the conclusion of the treatment. Complications were observed in 10 infants. Five infants died (mortality rate: 3.6%) including one term infant infected with GBS and four infants born <32 weeks GA. Of the five infants with complications who survived, three had osteomyelitis, one had an intracardiac thrombus and one had an intestinal infection that required surgery.

Antibiotic resistance

The global antibiotic resistance for all of the pathogens was 45.8%. Among the eight strains of GBS, two (25%) were erythromycin-resistant and one exhibited penicillin tolerance. Among the E. coli strains, 47.1% (33/70) were amoxicillin resistant, mainly represented by a low level penicillinase (n = 24). No extended spectrum beta-lactamase strains were isolated in our study. The Staphylococcus strains were methicillin resistant in 22 of the 37 (59.5%) cases.

Effects of perinatal antibiotic exposure

Univariable analysis revealed that maternal antibiotic exposure was significantly associated with an infant’s risk of developing amoxicillin-resistant E. coli infection (Fisher’s t-test, p = 0.01). The rate of amoxicillin-resistant E. coli was about 81.8% (n = 9) in exposed versus 35.5% (n = 11) in non-exposed cases, as shown in Fig. 1. The number of E. coli-infected infants born to mothers who received antenatal antibiotics (AA) (n = 9) was too small to conduct any comparative analysis of the type of AA (five mothers received antibiotic treatment for PPROM and four received IAP). This link persisted upon multivariable analysis which did not identify any additional risk factors for the development of LONI caused by antibiotic-resistant E. coli. Our univariable analysis, furthermore, suggested that postnatal antibiotic exposure was also associated with a risk of E. coli LONI (p = 0.048), but this link was not confirmed in a multivariable analysis.
Fig. 1

Maternal antibiotic exposure and LONI due to penicillinases (Penases) producing E. coli. p = 0.01. MatATB + infants exposed to maternal antibiotic treatment, MatATB − infants non exposed to maternal antibiotic treatment, Penases + penicillinases producing E. coli, Penases − penicillinases non-producing E. coli

Our univariable analysis suggested that antibiotic treatment early after birth was associated with an increased risk of Staphylococcus LONI (p = 0.02). We observed a trend between postnatal antibiotic exposure early after birth and methicillin-resistant staphylococcal LONI (p = 0.07). A multivariable analysis that included preterm birth did not confirm these results and so postnatal antibiotic exposure did not appear to be an independent risk factor for staphylococci-induced LONI.


Our study provides epidemiological data at a regional level for LONI. We used a large definition of LONI that included all infants infected between 3 and 30 postnatal days and all newborns younger than 90 postnatal days who were affected by E. coli or GBS infections [5, 6, 12, 14, 20, 28, 32]. This definition allowed us to account for the real burden of infection and helped to evaluate the possible impact of the increasing use of antenatal antibiotics. Consequently, our study population is heterogeneous and composed of term infants affected by community-acquired infection as well as preterm infants suffering from nosocomial infections. We limited infections from organisms other than GBS and E. coli to the first 30 days of life because in very preterm infants, the prolonged duration of hospitalisation increases the independent risk of nosocomial infections. It thus seemed difficult to relate antibiotic resistance and antenatal exposition in infants exposed to many other risk factors for infection and antibiotic resistance.

Here, we provide the first description of the global incidence of LONI. The incidences of confirmed, possible and global LONI that required treatment for more than 5 days illustrate the significant burden of this issue. Owing to the broad definition used in this study, comparisons with data from other countries may not be appropriate.

The incidence of 0.36 cases of GBS LONI per 1,000 live births was higher than the incidence of 0.1/1,000 reported in France between 1997 and 2006; however, that study used a restricted definition wherein subjects were only included if they had acquired GBS infections between 7 and 28 days of life [21]. Nevertheless, other European countries have also reported lower incidences between days 7 and 89 of life including incidence rates of 0.23/1,000 in Italy [5], 0.19/1,000 in Germany [14], 0.24/1,000 in the United Kingdom [20], and 0.14/1,000 in the Netherlands [32]. The United States Centres for Disease Control reported incidence rates of 0.3 per 1,000 live births in 1997 and 0.29 per 1,000 live births in 2006 (i.e., including infections acquired between days 7 and 89) [12]. The latter data confirm that IAP generalisation did not affect this incidence.

In our study, more than half of all cases of LONI were caused by E. coli UTI, which illustrates the high incidence of this infection and is consistent with other reports (64% in Glasgow et al.’s study and 67% in Byington et al.’s report) [10, 16]. High incidences of CoNS also affect VLBW infants. Stoll reported a rate of 21% of LONI in VLBW neonates hospitalised after 72 h of life, with 48% of infections caused by CoNS [28]. This diagnosis is either based on the collection of two positive blood cultures with the same pathogen or based on one positive blood culture with clinical signs and an increased level of CRP [15, 28]. We have chosen the latter criteria and acknowledge that LONI CoNS diagnosis is sometimes debatable. The distinction between infection and colonisation is difficult because clinical signs are often non-specific and blood samples are not easy to obtain. We chose to include infected infants despite negative culture (possible LONI) because we knew that sufficient samples are sometimes difficult to collect in this population of premature infant or newborn. The probability of bacterial isolation can be reduced, especially in case of small bacterial inoculum. Non-invasive method of urine collection, which is the standard reference of UTI diagnosis in France, is also exposed to the risk to include false positive cultures. In these situations, we considered the practitioner’s decision to treat the infant with antibiotics for at least 5 days even if possible contaminant bacteria were isolated.

Several studies of EONI have observed an increase in antibiotic-resistant Gram-negative bacterial infections after widespread use of antenatal antibiotics [26, 27]. We and others have reported that this link is primarily found in infants exposed to prolonged antibiotic treatment, particularly for PPROM, rather than in infants treated for IAP [22, 25, 30, 31]. Actually, the duration of antibiotic exposure seems to play a major role [6, 7, 22, 25, 29].

Few studies have focused on LONI. To our knowledge, Glasgow et al. performed the first case–control study to evaluate whether a potential link between intrapartum antibiotics and LONI caused by resistant pathogens exists [16]. Our observational population-based study used a different design to confirm their results in a regional population of infected infants. Maternal antibiotic treatment, either for PPROM or for IAP, was significantly associated with the risk of resistant E. coli LONI, even on a multivariable analysis. Longitudinal studies that include a follow-up cohort of exposed or non-exposed infants should be performed to determine the respective rates of these infections in each sub-groups. Our analysis did not identify a relationship between AA and LONI due to other pathogens. Labenne et al. performed a population-based study and found that prolonged postnatal antibiotic treatment duration was independently associated with a higher risk of LONI [23]. Our study design did not allow us to assess this relationship. Moreover, in our study, we did not find a significant association between postnatal antibiotic exposure and staphylococcal LONI. The discrepancy between our univariable and multivariable analyses may reflect the fact that postnatal antibiotic exposure was more common in preterm and VLBW infants, a group that was more often exposed to staphylococcal infection. Even if postnatal antibiotic exposure did not seem to be an independent risk factor for staphylococcal LONI, this point should be further evaluated and the possible association between postnatal antibiotic treatment and LONI due to methicillin-resistant staphylococci should be examined. In France, methicillin-resistance of CoNS in paediatric patients does not represent a real problem in clinical practice and rarely exposes to therapeutic failure because of the CoNS susceptibility to vancomycin. However, we think that antibiotic resistance still should be a constant preoccupation, in particular in NICU.

Although the completeness of the reported infections was controlled a posteriori by the “Information System of Medicalization Program” of each centre, some data, particularly data regarding the maternal obstetrical period, were difficult to obtain for a few patients. We used several methods to limit the missing data as much as possible (e.g., referring to the maternal obstetrical record, making phone calls to parents and postal contacts). We are aware that these missing data expose to some limitations and decrease the power of our results. As already mentioned, questions about LONI risk factors and antibiotic resistance may be better examined via an additional study of controlled healthy and infected infants. Although these studies are difficult to perform owing to the relative rarity of LONI, a case–control design should probably be used in future studies evaluating the possible link between perinatal antibiotic exposure and LONI.

Our findings confirm that the use of antenatal antibiotics increases the risk of antibiotic-resistant LONI. In a practical point of view, our results will not allow modification of the empirical treatment of UTI which consists already in the initial use of third generation cephalosporin. To establish the real burden of neonatal infections, EONI and LONI should be considered together. The benefits of IPA are real and should not be challenged; however, physicians should make every effort to administer narrow spectrum antibiotics such as penicillin G which should still be the treatment of reference [1, 13, 19, 24]. In our study, this antibiotic was used in only 9.5% of all treated mothers. National and international surveillance of LONI epidemiology and ecology is needed to assess the benefits and potential negative consequences of perinatal antibiotic exposure.


The following investigators also participated in this trial. All investigators are members of the Eastern Neonatology Study Group and/or the regional perinatal network, “Naître en Alsace”: Centre Hospitalier de Haguenau—André Geraudel, Centre Hospitalier de Colmar—Michel Kretz, Centre Hospitalier de Mulhouse—Marc Benoît, Centre Hospitalier de Saverne—Houria Demil, Clinique Adassa Strasbourg—Ziad Mansour, Clinique Sainte Anne Strasbourg—Mohamed Jernite, Centre Hospitalier de Wissembourg—Izzat Mikaïl, and Centre Médico-Chirurgical et Obstétrical de Schiltigheim—Michèle Weil. The authors declare that they have no conflict of interest.

Copyright information

© Springer-Verlag 2011