To document the trend of ampicillin-resistant infections in newborns weighing at least 1500 g and to determine factors associated with ampicillin-resistant neonatal early-onset infection in the era of routine group B streptococcal prophylaxis.
Referral hospital with level I through level III nurseries.
Newborns aged 0 to 7 days with cultures positive for bacterial infection, born from January 1994 to August 2002 (n = 53). Random controls were matched to admission year and nursery level (n = 159).
Main Outcome Measures
Trends of and factors associated with ampicillin-resistant infections.
Trends in our institution were the same as those found in some recent reports, a decrease in group B streptococcal early-onset infections without a concomitant increase in gram-negative early-onset infections. Specifically, when stratified by birth weight, newborns weighing at least 1500 g had no increase in gram-negative pathogens in the eras both before and after group B streptococcal prophylaxis (0.8 per 1000 live births to 0.3 per 1000 live births; incidence ratio, 2.3 [95% confidence interval, 0.5-10.9]). No increase in ampicillin resistance was seen during the same 3 periods (50%, 60%, and 50%, respectively; P = .97). Independent risk factors associated with ampicillin-resistant early-onset infection were intrapartum antibiotics for a 24-hour duration or longer (odds ratio, 4.8 [95% confidence interval, 1.0-23.3]) and clinical chorioamnionitis (odds ratio, 9.2 [95% confidence interval, 2.6-32.9]).
No increase in early-onset infections with gram-negative or ampicillin-resistant pathogens was detected. Ampicillin-resistant early-onset infection was associated with intrapartum antibiotics given for 24 hours or longer prior to delivery and with clinical chorioamnionitis. Ampicillin sodium and gentamicin sulfate remain appropriate initial antibiotic therapies for early-onset infection in newborns weighing at least 1500 g and without these risk factors.
Intrapartum antibiotic prophylaxis for neonatal group B streptococcal (GBS) infection has become routine since the 1996 publication of consensus guidelines by the Centers for Disease Control and Prevention and endorsement by the American Academy of Pediatrics, the American College of Obstetrics and Gynecology, and the American Academy of Family Practitioners.1- 5 Subsequently, early-onset GBS infection decreased by 65% between 1993 and 1998 to an overall incidence of 0.6 per 1000 live births.6,7 In contrast to the rate of GBS infection, some studies suggest that the overall rate of early-onset infection has not changed in infants with very low birth weights (<1500 g). In these infants, infection with gram-negative and ampicillin-resistant organisms has increased.8,9 Recent reports suggest that gram-negative, ampicillin-resistant infections have also emerged as the most common pathogens in late-onset sepsis (age 7-90 days) in term infants.10,11
The appearance of gram-negative, ampicillin-resistant pathogens has been associated with the routine use of intrapartum antibiotics for GBS prophylaxis, both in infants with very low birth weights8,9 and in studies with infants of all birth weights.12- 15 However, at least 3 studies from the United States and Australia found that intrapartum antibiotics decreased or had no effect on non-GBS infections in infants of all birth weights.16- 18 No studies have specifically examined the effects of intrapartum antibiotics on newborns who weigh at least 1500 g. Knowing the effect of these antibiotics on pathogens causing sepsis and antibiotic resistance in newborns is essential for optimal prevention and empiric treatment. Risk factors associated with the acquisition of ampicillin-resistant infection would be important information to consider when selecting empiric antibiotic therapy.
Our primary objectives were to describe the trends of early-onset infection in newborns weighing at least 1500 g since the implementation of GBS prophylaxis and to examine risk factors associated with ampicillin-resistant pathogens in these newborns.
The institutional review board of the University of Utah Health Sciences Center (Salt Lake City) approved this study, waving the need for parental consent. We performed a retrospective cohort study to document trends in newborns weighing at least 1500 g. We then performed a case-control study to examine the risk factors associated with ampicillin-resistant and ampicillin-susceptible early-onset infection.
The computerized microbiology database at the University of Utah was searched to identify newborns, both live births at the university hospital and subjects born outside that hospital, aged 0 to 7 days with aerobic and/or anaerobic cultures of blood, cerebrospinal fluid, catheterized urine, placental tissue, or tracheal aspirates positive for bacteria between January 1994 and August 2002. To be included in the study, samples from catheterized urine required more than 100 000 colony-forming units and could not be polymicrobial. Samples from tracheal aspirates were considered true pathogens if collected on the first day of life from patients undergoing intubation who had a chest radiograph consistent with pneumonia. Medical records of identified newborns were reviewed to exclude contaminated cultures. Contaminants included any coagulase-negative Staphylococcus species, Corynebacterium diphtheriae, Bacillus species, and other skin microorganisms. Cases with pathogens isolated were reviewed and data were collected, including histories of the newborn and mother, features of early-onset infection, and antibiotic susceptibility of the pathogens. Chorioamnionitis was defined as clinical uteroplacental infection as diagnosed by the mother's obstetrician and documented in the medical record.
An analysis of the cohort of live births at the University of Utah Hospital from 1994 to 2002 was conducted to examine temporal trends of early-onset infection. This helped to determine if the temporal trends seen in our institution were similar to those seen around the country in the era of GBS prophylaxis. We documented the trend specifically in infants weighing at least 1500 g.
According to the frequency of intrapartum antibiotic use, which correlated with the publication of the Centers for Disease Control and Prevention guidelines, we divided our study period into 3 intervals for analysis. The first period was January 1994 through December 1996, when antibiotic use was at its lowest point in obstetric patients (pre-GBS prophylaxis). The second was January 1997 through December 1998, when an initial rise was seen in antibiotic use, corresponding to the publication of GBS prophylaxis guidelines (transition). The third was January 1999 through August 2002, when antibiotic use was stable and ampicillin sodium was preferentially used owing to a national shortage of penicillin G potassium (routine GBS prophylaxis).19
We also determined the background rate of change in ampicillin resistance of Escherichia coli in our community by surveying a local reference microbiologic laboratory (Associated Regional University Pathologists, Salt Lake City). E coli was used as a marker for changes in ampicillin resistance because it was the most common gram-negative organism seen in early-onset infection. Other gram-negative infections were intrinsically resistant to ampicillin and unlikely to change across time.
A case-control study was performed to determine risk factors associated with ampicillin-resistant and ampicillin-susceptible early-onset infection. Three controls per case were randomly selected from a hospitalwide database of all deliveries from 1994 to 2002, matching for year of delivery and nursery level. Nursery level was used to approximate gestational age and birth weight, 2 factors that were not available in the hospitalwide database. We believe this matching to be accurate because at our institution, all infants younger than 34 weeks' gestation are admitted to the level III nursery.
Newborns with susceptible infections were compared with controls, and newborns with resistant infections were compared separately with the same controls to determine the effect of each potential risk factor on the 2 types of infections. This statistical model was used because antibiotics can have 2 types of effects on infection: one is to reduce the risk of susceptible infection, and the other is to increase the risk of resistant infections. This study design enabled us to highlight these differences and ensured the limitation of overestimating bias, which would have been a significant limitation if susceptible infections had been used as controls. In this way, protective and causal effects could be determined separately for each infection type.20,21
The incidence proportion of early-onset infection was calculated per 1000 live births in the cohort, and incidence ratios and 95% confidence intervals (CIs) were used to detect differences. To determine independent associations between potential risk factors and resistant infection in the case-control study, we calculated odds ratios (ORs) with estimated 95% CIs. We used multivariable conditional logistic regression models containing the variables significantly associated with infection in univariate analysis. P <.05 was considered significant. Stata 8.0 statistical software (Stata Corp, College Station, Tex) was used for the analyses.
Ampicillin resistance of E coli at the reference laboratory serving the University of Utah Hospital and surrounding Salt Lake City area remained stable among periods 1, 2, and 3 (39% vs 48% vs 42%; P = .07). In the early-onset cases, the percentage of gram-negative infections that were ampicillin resistant did not change during the study period: 50% in period 1, 60% in period 2, and 50% in period 3 (P = .97). When comparing newborns with very low birth weights with those who weighed at least 1500 g, we found no significant difference in the trends of ampicillin-resistant organisms between the 2 groups during the 3 periods.
The cohort of live births in the study period numbered 26 288: 8030 in period 1, 6128 in period 2, and 12 130 in period 3. We identified 53 cases of early-onset infection. In all periods, the most frequent causes of early-onset infection in newborns were GBS and gram-negative pathogens (Table 1). There were no significant differences between pathogens identified in newborns weighing 1500 g or less. In both groups, GBS infection and E coli were the most common organisms. The pathogens were principally isolated from blood (34). Other recovery sites included tracheal aspirates (20), cerebrospinal fluid (6), urine (3), and the placenta (3). Many of these infections were documented with cultures positive for GBS infection from multiple sites (12) (Table 2).
Overall, early-onset infection incidence proportion had a decreasing trend between periods 1 and 3, from 2.4 per 1000 to 1.5 per 1000 live births (incidence ratio, 1.6 [95% CI, 0.8-3.2]), whereas no increase in gram-negative pathogens was seen, from 1.1 per 1000 to 0.8 per 1000 live births (incidence ratio, 1.4 [95% CI, 0.5-3.7]). In newborns weighing at least 1500 g, the incidence proportion of gram-negative pathogens showed a decreasing trend from 0.8 per 1000 to 0.3 per 1000 live births between periods 1 and 3, but statistical significance was not attained (incidence ratio, 2.3 [95% CI, 0.5-10.9]).22
We identified 53 newborns with early-onset infection and 159 controls. Two newborns with early-onset infection had polymicrobial infections, one with E coli and GBS infection and the other with 2 Pseudomonas species and an Enterococcus species. Both newborns were included in the analysis because their infections were treated as true pathogens by the attending physician and were not thought to represent contamination. A total of 56 pathogens were identified. The pathogens identified and infections caused in all newborns are listed in Table 1 and Table 2.
Demographic and clinical information for the cases and controls are shown in Table 3. For early-onset cases, the mean ± SD age of newborns at the time of the first culture postive for infection was 0.5 ± 1.2 days. The only statistical difference in demographic characteristics between cases and controls, including both survivors and nonsurvivors, was for a shorter mean length of stay (cases: mean ± SD, 33.2 ± 36.5 days; controls: mean ± SD, 47.2 ± 40.8 days; mean ± SD difference, 14.0 ± 4.4 days) and a higher mortality rate (cases: 17%; controls: 2%; OR, 10.6 [95% CI, 2.5-62.7]). Both of these outcomes were effects of infection rather than causes and therefore were not included in the regression model.
We examined exposures associated with ampicillin-resistant and ampicillin-susceptible gram-negative pathogens in both birth weight categories in our model. In comparison with controls by univariate analysis, exposures associated with acquiring any early-onset infection (resistant or susceptible) included intrapartum ampicillin use when compared with other types of intrapartum antibiotics, intrapartum antibiotic use for a 24-hour duration or longer, mothers with clinical chorioamnionitis, and rupture of membranes for more than 18 hours (Table 3).
These factors as well as nursery level were used in a multivariable conditional logistic regression model to determine independent associations with infection caused by an ampicillin-resistant or ampicillin-susceptible organism. After removing the newborns without susceptibility data (4 cases), 49 newborns with 52 pathogenic isolates were included in this analysis (17 resistant and 35 susceptible). Newborns with polymicrobial infections were treated as 1 case, deemed resistant if any of the organisms were resistant and susceptible if all were susceptible.
In multivariable analysis, factors associated with an increased risk of resistant infection were intrapartum antibiotic use for 24 hours or longer prior to delivery (OR, 4.8 [95% CI, 1.0-23.3]) and chorioamnionitis (OR, 9.2 [95% CI, 2.6-32.9]). Factors related to an increased risk of susceptible infection were chorioamnionitis (OR, 5.6 [95% CI, 1.9-16.4]) and rupture of membranes for more than 18 hours (OR, 5.0 [95% CI, 1.6-15.6]). Intrapartum antibiotic use for 24 hours or longer prior to delivery was not a risk factor for acquiring a susceptible infection (OR, 1.4 [95% CI, 0.4-4.9]). Factors associated with a decreased risk of susceptible infection were any antibiotic use, either ampicillin (OR, 0.3 [95% CI, 0.1-0.9]) or other antibiotics (OR, 0.1 [95% CI, 0.02-0.60]). We found no decreased risk of resistant infection with any antibiotic use (Table 4).
This study, examining trends of early-onset infection following the introduction of intrapartum antibiotics for GBS prophylaxis, yielded results similar to others.16- 18 To our knowledge, this study is the first to look specifically at newborns weighing at least 1500 g. We documented no increase in infection due to gram-negative pathogens in this weight group. In addition, we saw no increase in ampicillin-resistant infections. However, risk factors for ampicillin resistance such as prolonged intrapartum antibiotic use or chorioamnionitis were documented and may help to improve the rational selection of antibiotics in newborns at risk for early-onset infection.
Newborns with ampicillin-resistant infections were nearly 5-fold more likely to have mothers who received antibiotics for 24 hours or longer prior to delivery when compared with controls. Terrone et al15 also documented an increase in ampicillin-resistant E coli in mothers who received prolonged courses of antibiotics. In our model, prolonged duration of antibiotics was not a significant predictor of ampicillin-susceptible infections. However, other well-known risk factors, chorioamnionitis and rupture of membranes for more than 18 hours, were associated with developing susceptible early-onset infections; intrapartum antibiotic use was protective in these cases. These results closely fit our conceptual model of how intrapartum antibiotic use influences both resistant and susceptible infections.
Increasing antibiotic resistance from prolonged antibiotic treatment or use of broad-spectrum antibiotics, as in chorioamnionitis, can be explained by selection pressure exerted during these treatments. The maternal bacterial microorganisms are altered by antibiotic therapy, and if enough exposure occurs prior to delivery, the newborn may have an ampicillin-resistant early-onset infection.23- 29 Short courses of antibiotics, such as those given for GBS prophylaxis, did not seem to cause ampicillin-resistant early-onset infections in our patient population. There may be insufficient time with this treatment for an ampicillin resistance to develop prior to delivery. However, these short antibiotic exposures may be enough to alter maternal microorganisms and place the newborn at increased risk for horizontal acquisition and possibly late-onset infection with resistant pathogens.30
Use of ampicillin rather than other intrapartum antibiotics was a risk factor for early-onset infection in the univariate analysis but did not remain significant in the conditional logistic regression model. This reflects the protective effect seen in susceptible infections. However, health care professionals should still consider the potential risk for ampicillin to select for resistant pathogens and, whenever possible, preferentially choose penicillin for intrapartum prophylaxis because it does not alter the maternal gastrointestinal microorganisms.25,26
This study is limited by its retrospective design. The data are dependent on the medical record and historical reporting. Maternal colonization with group B Streptococcus was not documented in all subjects and was not included as a risk factor for infection in the model. Small numbers of early-onset infections were seen in the study period, making the power to detect small differences in the trend of pathogens low. Episodes of culture-negative sepsis were not captured. Newborns from other institutions who had blood cultures positive for bacteria were also missed. Antibiotic use other than in the intrapartum period was not documented, and we were therefore unable to control for this potentially confounding factor. In addition, we are limited by our ability to detect changes in GBS infection because of the low incidence of these bacteria in our population. Our hospital is nearly 75% white, and no African Americans were included as cases or controls. White populations may have a lower risk of GBS infection, as seen in studies from England and Finland.31,32
Despite these limitations, we believe that this study provides guidance for physicians caring for newborns with suspected early-onset infection. No increase in ampicillin-resistant early-onset infection was documented during the study period, and in newborns weighing at least 1500 g, there was no increased incidence of gram-negative pathogens. In suspected cases of sepsis, ampicillin sodium and gentamicin sulfate are still appropriate empiric antibiotic agents for the larger newborn in the era of GBS prophylaxis. Ampicillin-resistant early-onset infection should be considered in newborns whose mothers received intrapartum antibiotics for longer than 24 hours or had chorioamnionitis. In these cases, initial antibiotic coverage should include agents with activity against ampicillin-resistant pathogens. Penicillin should be used preferentially to ampicillin for GBS prophylaxis whenever possible. Continued surveillance for antibiotic-resistant pathogens and risk factors for transmission is critical.
Intrapartum GBS prophylaxis has been shown to significantly reduce the incidence of neonatal early-onset GBS infection. However, recent reports have raised concerns of an increasing incidence of gram-negative, ampicillin-resistant early-onset infection associated with intrapartum antibiotic use in newborns with very low birth weights. Although our study found no significant increase in the incidence of such pathogens in newborns with birth weights of at least 1500 g in the era of GBS prophylaxis, the results suggest that these pathogens may be associated with prolonged administration of intrapartum antibiotics for other reasons, such as chorioamnionitis. This indicates that ampicillin and gentamicin are still appropriate empiric antibiotic therapies for newborns weighing at least 1500 g whose mothers did not have clinical chorioamnionitis or receive antibiotics for less than 24 hours prior to delivery.
Corresponding author and reprints: Alison C. Rentz, MD, Division of Neonatology, University of Utah School of Medicine, 50 N Medical Dr, Suite 2A 100, Salt Lake City, UT 84132 (e-mail: email@example.com).
Accepted for publication February 5, 2004.
Dr Rentz is supported in part by the Children's Health Research Center, University of Utah, Salt Lake City, and is a Primary Children's Medical Center Foundation Scholar. Dr Byington was supported by the Robert Wood Johnson Generalist Physician Faculty Scholar Program, Princeton, NJ.
This study was presented in part at the Annual Meeting of the Pediatric Academic Societies, May 3, 2003, Seattle, Wash.
Rentz AC, Samore MH, Stoddard GJ, Faix RG, Byington CL. Risk Factors Associated With Ampicillin-Resistant Infection in Newborns in the Era of Group B Streptococcal Prophylaxis. Arch Pediatr Adolesc Med. 2004;158(6):556-560. doi:10.1001/archpedi.158.6.556