OBJECTIVE. The goal was to evaluate risk factors for invasive Escherichia coli infections in the first week of life (early onset), focusing on the role of intrapartum antibiotic use.
METHODS. We conducted a retrospective case-control study. Between 1997 and 2001, case infants, defined as infants <7 days of age with E coli isolated from blood or cerebrospinal fluid, were identified in selected counties of California, Georgia, and Connecticut by the Active Bacterial Core Surveillance/Emerging Infections Program Network. Control infants (N = 1212) were identified from a labor and delivery record review of a stratified random sample of live births at the same hospitals in 1998 and 1999.
RESULTS. Surveillance identified 132 E coli cases, including 68 ampicillin-resistant cases. The case fatality rate was 16% (21 of 132 cases). Two thirds of case infants were preterm, and 49% (64 of 132 infants) were born at ≤33 weeks of gestation. Fifty-three percent of case mothers (70 of 132 mothers) received intrapartum antibiotic therapy; 70% of those received ampicillin or penicillin. Low gestational age (≤33 weeks), intrapartum fever, and membrane rupture of ≥18 hours were associated with increased odds of early-onset E coli infection. Results were similar when case subjects were limited to those infected with ampicillin-resistant strains. Exposure to any intrapartum antibiotic treatment, β-lactam antibiotic treatment, or ≥4 hours of intrapartum antibiotic therapy was associated with increased odds of E coli infection and ampicillin-resistant infection in univariate analyses. Among preterm infants, intrapartum antibiotic exposure did not remain associated with either outcome in multivariable models. Among term infants, exposure to ≥4 hours of intrapartum antibiotic therapy was associated with decreased odds of early-onset E coli infection.
CONCLUSIONS. Exposure to intrapartum antibiotic therapy did not increase the odds of invasive, early-onset E coli infection. Intrapartum antibiotic therapy was effective in preventing E coli infection only among term infants.
Neonatal sepsis is a leading infectious cause of infant morbidity and death. Group B streptococcal (GBS) sepsis emerged in the 1970s as a predominant cause of infections in the first week of life. Development of prevention policies in the 1990s and widespread uptake of the primary prevention strategy, namely, intrapartum antibiotic therapy for women at risk, led to a >70% decrease in perinatal invasive GBS disease incidence in the United States.
Decreases in GBS sepsis rates have shifted the focus to Escherichia coli, which was a leading cause of infection among newborns before GBS disease emerged. Now that the incidence of invasive GBS disease in the first 1 week of life (early onset) approaches 0.3 cases per 1000 live births,1 the incidence of early-onset E coli infection is similar to and in some populations higher than that of GBS disease.2–4E coli sepsis is an important cause of death among newborns, particularly among those of very low birth weight.5 Increased use of intrapartum antibiotic therapy has raised the concern that rates of early-onset sepsis attributable to organisms other than GBS might increase, particularly Gram-negative organisms such as E coli that are associated with high case fatality ratios and emerging antibiotic resistance. Although most population-based studies suggested that incidence rates of early-onset E coli sepsis remained stable in the era of decreasing GBS sepsis rates,6,7 there is evidence of a sustained increase in incidence among very low birth weight infants.2,8 Ampicillin resistance, which has a prevalence in some populations of ≥85% among invasive isolates,2 is of particular concern because penicillin and ampicillin are first-line agents for GBS intrapartum chemoprophylaxis and ampicillin also has been a first-line agent for empiric treatment of suspected sepsis among newborns.
Characterization of risk factors for perinatal GBS sepsis contributed importantly to the development of effective prevention strategies that targeted at-risk women. Risk factors for E coli sepsis have not been well characterized. In particular, the relationship between intrapartum antibiotic exposure and early-onset E coli infection has not been elucidated. On one hand, intrapartum antibiotics may have some efficacy in prevention of early-onset E coli infections, as seen for GBS sepsis9 and for all-cause, early-onset, non-GBS sepsis.10 On the other hand, it is possible that intrapartum antibiotic prophylaxis, which is now administered for an estimated 25% to 30% of deliveries, may increase the risk of ampicillin-resistant E coli infection. Several analyses of neonatal sepsis case series showed an association between exposure to intrapartum ampicillin and infection with ampicillin-resistant E coli.11–14 In this report, we use multistate, hospital-based surveillance for neonatal invasive infections to describe a large cohort of infants with invasive, culture-confirmed E coli infection. We also conduct a retrospective, case-control analysis of early-onset E coli infection risk factors in the era of widespread perinatal GBS disease prevention, with a particular focus on intrapartum antibiotic exposure.
Cases were defined on the basis of isolation of E coli from blood or cerebrospinal fluid collected from a neonate ≤6 days of age (early onset) between January 1, 1997, and December 31, 2001. This time period was chosen because the first national guidelines for perinatal GBS disease prevention were released in 1996, and evaluations of intrapartum antibiotic use in 1998 and 199915 and GBS prevention policies among providers16 demonstrated that intrapartum antibiotic prophylaxis was widespread during this time. We identified cases through active, population-based surveillance for invasive neonatal sepsis in the Active Bacterial Core Surveillance (ABCs)/Emerging Infections Program Network.3,4 Catchment populations included births in 3 counties in the San Francisco area in California, 11 hospitals capturing >85% of births in 8 metropolitan counties in Atlanta, Georgia, and 20 hospitals capturing >80% of births in Connecticut. Surveillance methods were described previously.3,4 In brief, surveillance officers in each area reviewed microbiology records at clinical laboratories serving the catchment populations on a regular basis and collected demographic and clinical information from medical records with a standardized form. Susceptibility testing was not performed centrally, but results were recorded from the medical records.
Control subjects (N = 1212) were selected from a labor and delivery record abstraction of a sample of live births in the ABCs areas, as has been described.15,17 Births to surveillance area residents in 1998 and 1999 were selected with random proportional allocation stratified according to surveillance area and birth hospital. Demographic and clinical information was abstracted from labor and delivery records with a standardized form. Neonates born at hospitals that did not participate in neonatal sepsis surveillance and neonates with invasive, early-onset, GBS infections captured by ABCs surveillance were excluded.
Variables available for evaluation as potential sepsis risk factors included maternal race and ethnicity, mode of delivery, gestational age at delivery, duration of membrane rupture, intrapartum fever, and intrapartum antibiotic administration. Case subjects with Hispanic ethnicity and unknown race (n = 14) were assumed to be white, because >85% of Hispanic children <1 year of age in our catchment areas in the years under evaluation were of white race. For women who received intrapartum antibiotic therapy, data were available on the total duration of therapy and the agents and number of doses administered.
Analyses were conducted with SAS version 9.12 (SAS Institute, Cary, NC). We evaluated 2 primary outcome variables, namely, invasive, early-onset, E coli infection and invasive, ampicillin-resistant, early-onset, E coli infection. Although these outcomes are not mutually exclusive, we evaluated each independently. Ampicillin-susceptible E coli cases were excluded from analyses of risk factors for ampicillin-resistant E coli sepsis. Univariate analyses for variables with >2 levels and multivariate analyses were performed with logistic regression. Multivariate models were evaluated starting with all factors that were significant at P < .15 in univariate analyses and dropping nonsignificant factors with stepwise backward selection. Colinearity of independent variables was evaluated. Control subjects were not matched with case subjects. However, to determine whether surveillance area influenced model outcomes, we also performed conditional logistic regression analysis stratified according to state. All 2-way interactions in final multivariate models were evaluated. Adjusted, generalized, population-attributable risk estimates and confidence intervals (CIs) were estimated with jackknife procedures.18 Throughout, 2-sided P values of <.05 were considered statistically significant and 95% CIs were calculated.
Case Patient Characteristics
Surveillance identified 132 early-onset E coli cases during the 5-year period; 61% had onset on day 0 of life and 80% had onset within the first 3 days. Case characteristics are summarized in Table 1. Two thirds of case subjects were born at <37 weeks of gestation. The median length of hospitalization was 19 days (interquartile range: 7–50 days). The case fatality ratio was 16% (21 of 132 infants). Among surviving infants, 22% (24 of 111 infants) had medical sequelae associated with the hospitalization, although these could not always be linked to the sepsis episode. More than one half of case subjects (53%; 70 of 132 infants) were exposed to intrapartum antibiotic therapy. The most common regimen was ampicillin alone, followed by penicillin alone (28% and 16%, respectively, of case subjects who were exposed to intrapartum antibiotic therapy). Among infants exposed to intrapartum antibiotics, 71% received a regimen including either penicillin or ampicillin; the median dose of penicillin or ampicillin received among infants exposed to intrapartum antibiotics was 1 (range: 0–87 doses).
Comparison of Ampicillin-Resistant and Ampicillin-Sensitive Cases
Among case patients whose isolates had documented ampicillin susceptibility testing results, 59% (68 of 116 patients) were infected with an ampicillin-resistant strain. Compared with case patients with ampicillin-sensitive infections, case patients with ampicillin-resistant infections were more likely to have prolonged membrane rupture (Table 2). Case patients with ampicillin-resistant infections were not more likely to have been exposed to intrapartum antibiotic therapy than were case subjects with ampicillin-sensitive infections. They were, however, exposed to more intrapartum doses of penicillin or ampicillin (mean doses: ampicillin-sensitive: 0.4 doses; ampicillin-resistant: 3.7 doses; Kruskal-Wallis χ12 = 6.6; P = .01). Among case subjects, exposure to ≥2 penicillin or ampicillin doses remained associated with increased odds of ampicillin-resistant infection in multivariable analysis (Table 2). The case fatality rate was higher among case patients with ampicillin-resistant infections than among those with sensitive infections, but the difference was not statistically significant (10 of 68 infants, 15%, vs 4 of 48 infants, 8%; Fisher's exact test, P = .39); 7 of 21 case patients who died did not have ampicillin susceptibility testing results documented.
Case-Control Risk-Factor Analysis
Factors associated with early-onset E coli infection and with early-onset, ampicillin-resistant, E coli infection were similar (Table 3). In univariate and multivariate models, very low gestational age (≤33 weeks of gestation) was the strongest risk factor (all infections: adjusted odds ratio [aOR]: 26.5; 95% CI: 15.0–46.8; ampicillin-resistant infections: aOR: 35.8; 95% CI: 16.8–76.3). Gestational age between 34 and 36 weeks, intrapartum fever, and prolonged membrane rupture were also risk factors for E coli infection and ampicillin-resistant infection in both univariate and multivariate models (Table 2). The population-attributable risk associated with preterm delivery, adjusted for intrapartum fever and prolonged membrane rupture, was 59% (95% CI: 49%–70%).
In univariate analysis (Table 3) and controlling for intrapartum fever, exposure to intrapartum antibiotic prophylaxis was associated with elevated odds of E coli infection and ampicillin-resistant E coli infection. When case patients were limited to those with ampicillin-sensitive infections, exposure to intrapartum antibiotic prophylaxis remained associated with elevated odds of E coli infection in univariate analysis (odds ratio [OR]: 2.3; 95% CI: 1.3–4.2) and controlling for intrapartum fever (OR: 1.7; 95% CI: 0.9–3.2).
In multivariate analysis, none of the intrapartum antibiotic-related variables approached statistical significance for either the E coli or ampicillin-resistant E coli infection outcomes (Table 3). When included in the final multivariate model, the associated ORs were close to 1 (Table 3). Multivariate results remained similar when case and control subjects were pooled according to surveillance area.
Because gestational age was associated so strongly with odds of E coli infection, we also evaluated potential risk factors separately for term infants and preterm infants. Among preterm infants, prolonged rupture of membranes (aOR: 5.2; 95% CI: 2.7–10.0) and intrapartum fever (aOR: 7.6; 95% CI: 2.3–25.0) were the significant risk factors in multivariate analysis. Among term infants, in addition to prolonged membrane rupture (aOR: 5.3; 95% CI: 2.4–11.6) and intrapartum fever (aOR: 8.2; 95% CI: 3.6–18.8), exposure to ≥4 hours of intrapartum antibiotic therapy (aOR: 0.3; 95% CI: 0.1–0.8) was associated with reduced E coli infection odds. For the outcome of ampicillin-resistant infection, prolonged membrane rupture and intrapartum fever were the only factors associated significantly with infection among both preterm and term infants.
Despite the common use of intrapartum antibiotic prophylaxis during our study period, we found no evidence that exposure to intrapartum antibiotics increased the odds of early-onset E coli infection. Moreover, intrapartum antibiotics had a protective effect for term infants. These findings are reassuring and suggest that perinatal GBS disease prophylaxis recommendations do not need modification currently. However, because antibiotic use practices and sepsis pathogens are both dynamic, continued monitoring for unintended consequences of GBS prophylaxis remains important. To date, an increased incidence of early-onset E coli infection has been reported for very low birth weight infants.2,8 Decreases in GBS sepsis rates have been observed widely, including among very low birth weight infants, and early-onset E coli infection incidence and overall non-GBS sepsis incidence seem stable in several countries that have implemented GBS disease prevention guidelines.2–4,8,19–22
The lack of association between intrapartum antibiotic exposure and early-onset E coli sepsis reported here is consistent with similar findings for all-cause sepsis among very low birth weight infants.8 The lack of a robust protective effect of intrapartum antibiotic prophylaxis is also noteworthy. Among preterm infants, the group affected predominantly by this infection, we found no evidence that intrapartum antibiotic exposure reduced sepsis odds. Among term infants, 70% efficacy was found. A multicenter case-control study conducted >1 decade ago estimated 63% effectiveness of intrapartum antibiotic therapy in preventing early-onset, non-GBS sepsis.10E coli was the leading cause of non-GBS sepsis in that study but made up only 28% of cases. Interestingly, the proportion of preterm infants10 was similar to that we report here. The study by Schuchat et al10 was conducted in the era before widespread GBS prevention, when intrapartum antibiotic therapy was used primarily for women with signs of chorioamnionitis. It may be that antibiotics have some benefit for earlyonset E coli sepsis prevention among these high-risk deliveries.
It is also possible that ampicillin and penicillin, the agents administered most commonly in the intrapartum period, are no longer as effective in preventing vertical transmission of E coli. More than one half (59%) of the infections in our case series were caused by ampicillin-resistant strains. However, when our analyses were limited to ampicillin-susceptible cases, there was no evidence of a protective effect of intrapartum antibiotic exposure.
The emergence of ampicillin-resistant E coli has been well documented.2,8,23–25 Although it is likely that antibiotic use contributed to this emergence, our results suggest that intrapartum antibiotic treatment is not the driving force behind early-onset, ampicillin-resistant infections. When we compared case patients infected with ampicillin-resistant strains with those infected with ampicillin-sensitive strains, we found an association between antibiotic exposure and resistance, as did several other studies.8,11–14 Such evaluations, however, often overestimate the impact of antibiotic exposure because they exclude sensitive infections that were prevented. When we compared E coli case patients with uninfected control subjects and when we controlled for other sepsis risk factors, evidence for a lack of association with intrapartum antibiotic exposure was compelling. This highlights the importance of case-control or cohort approaches to elucidation of the relationship between antibiotic exposure and resistant infections. It also highlights the dangers of overinterpreting associations based on data for case patients alone or analyses that do not control for major risk factors.6
Our study has several limitations. Most importantly, this was a retrospective observational study rather than a randomized trial. As a result, the population of patients exposed to intrapartum antibiotic prophylaxis might have differed from the unexposed population in ways that could not be controlled for in the analysis. In addition, our case subjects were drawn from a broader time period than were the control subjects. Because GBS prevention practices remained fairly stable during this time period, decreases in GBS disease incidence plateaued, and guidelines for intrapartum antibiotic use for other reasons (eg, threatened preterm delivery) did not change from 1998 to 2001, we do not think that the differing case and control time periods introduced an important bias.16,26,27 Because we focused on factors associated with disease and not with infection outcome, any changes in neonatal treatment during this time period would not have influenced our primary results. Medical record abstraction did not include information on treatments and procedures administered to case patients; therefore, more-detailed clinical characterization of this case series was not possible. Susceptibility testing drug panels and methods were not standardized, and test ordering was at the discretion of the treating physicians. Fortunately, ampicillin susceptibility results were available for a majority of case patients (88%); however, we were not able to evaluate susceptibilities to other drugs. Lastly, we were not able to evaluate some potential sepsis risk factors that were identified in previous studies (eg, young maternal age, number of vaginal examinations), and we could not evaluate the reason for intrapartum antibiotic administration.
Among the risk factors for early-onset E coli infections we were able to evaluate, preterm delivery was by far the strongest risk factor. This is consistent with previous studies of neonatal sepsis risk factors.8,28 A growing body of literature suggests that vaginal infections may be an important cause of prematurity.29,30 It is possible that the association we observed reflects the fact that low-grade vaginal infection or inflammation caused by E coli results in both preterm delivery and early-onset infections. It is also possible that infants delivered prematurely, because of compromised skin barriers and incomplete transfer of maternal antibodies, are more vulnerable to these infections. Because two thirds of case patients were preterm, efforts to identify strategies to prevent preterm delivery may contribute to prevention of early-onset E coli infections.
Funding for this activity was provided by the Emerging Infections Program of the National Center for Infectious Diseases Office of the Director and by the National Center for Infectious Diseases Antimicrobial Resistance Working Group. Funders played no role in the design and conducting of the study; the collection, management, analysis, and interpretation of the data; or the preparation, review, and approval of the manuscript.
The ABCs neonatal sepsis surveillance team included Pamala Daily, MT, MPH, Joelle Nadle, MPH, and Gretchen Rothrock, MPH (California); Robert Baltimore, MD, Sharon Huie, and Susan Petit (Connecticut); Wendy Baughman, MSPH, and Monica Farley, MD (Georgia); and Terri Hyde, MD, MPH, Ellen Lee, MD, Tami Skoff, MS, Elizabeth Zell, MStat, Sandy McCoy, MPH, and Carolyn Wright (Centers for Disease Control and Prevention).
- Accepted February 27, 2006.
- Address correspondence to Stephanie J. Schrag, DPhil, Mailstop C23, Centers for Disease Control and Prevention, 1600 Clifton Rd, NE, Atlanta, GA 30333. E-mail:
The authors have indicated they have no financial relationships relevant to this article to disclose.
- ↵Baltimore RS, Huie SM, Meek JI, et al. Early-onset neonatal sepsis in the era of group B streptococcal prevention. Pediatrics.2001;108 :1094– 1098
- ↵Hyde TB, Hilger TM, Reingold A, et al. Trends in the incidence and antimicrobial resistance of early-onset sepsis: population-based surveillance in San Francisco and Atlanta. Pediatrics.2002;110 :690– 695
- ↵Schuchat A, Zywicki SS, Dinsmoor MJ, et al. Risk factors and opportunities for prevention of early-onset neonatal sepsis: a multicenter case-control study. Pediatrics.2000;105 :21– 26
- ↵Kahn MJ, O'Fallon WM, Sicks JD. Generalized Population Attributable Risk Estimation. Rochester, MN: Mayo Clinic; 2000. Technical report 54
- Daley AJ, Issacs D, Australasian Study Group for Neonatal Infections. Ten-year study on the effect of intrapartum antibiotic prophylaxis on early-onset group B streptococcal and Escherichia coli neonatal sepsis in Australasia. Pediatr Infect Dis J.2004;23 :630– 634
- ↵Karlowsky JA, Kelly LJ, Thornsberry C, et al. Trends in antimicrobial resistance among urinary tract infection isolates of Escherichia coli from female outpatients in the United States. Antimicrob Agents Chemother.2002;46 :2540– 2545
- Kerrn MB, Klemmensen T, Frimodt-Moller N, et al. Susceptibility of Danish Escherichia coli strains isolated from urinary tract infections and bacteraemia, and distribution of sul genes conferring sulphonamide resistance. J Antimicrob Chemother.2002;50 :513– 516
- ↵Oteo J, Campos J, Baquero F. Antibiotic resistance in 1962 invasive isolates of Escherichia coli in 27 Spanish hospitals participating in the European Antimicrobial Resistance Surveillance System (2001). J Antimicrob Chemother.2002;50 :945– 952
- ↵Centers for Disease Control and Prevention. Early-onset group B streptococcal disease: United States, 1998–1999. MMWR Morb Mortal Wkly Rep.2000;49 :793– 796
- ↵Goldenberg R, Andrews W, Guerrant R, et al. The Preterm Prediction Study: cervical lactoferrin concentration, other markers of lower genital tract infection, and preterm birth: National Institute of Child Health and Human Development Maternal-Fetal Medicine Units Network. Am J Obstet Gynecol.2000;182 :631– 635
- Copyright © 2006 by the American Academy of Pediatrics