|
|
| |||||||||||||||||||||||||||||||||||||||||||||||||
Background It is uncertain whether the rates and causes of early-onset sepsis (that occurring within 72 hours after birth) among very-low-birth-weight infants have changed in recent years, since antibiotics have begun to be used more widely during labor and delivery.
Methods We studied 5447 very-low-birth-weight infants (those weighing between 401 and 1500 g) born at centers of the Neonatal Research Network of the National Institute of Child Health and Human Development between 1998 and 2000 who had at least one blood culture in the first three days of life and compared them with 7606 very-low-birth-weight infants born at centers in the network between 1991 and 1993.
Results Early-onset sepsis (as confirmed by positive blood cultures) was present in 84 infants in the more recent birth cohort (1.5 percent). As compared with the earlier birth cohort, there was a marked reduction in group B streptococcal sepsis (from 5.9 to 1.7 per 1000 live births of infants weighing 401 to 1500 g, P<0.001) and an increase in Escherichia coli sepsis (from 3.2 to 6.8 per 1000 live births, P=0.004); the overall rate of early-onset sepsis was not significantly changed. Most E. coli isolates from the recent birth cohort (85 percent) were resistant to ampicillin, and mothers of infants with ampicillin-resistant E. coli infections were more likely to have received intrapartum ampicillin than were those with ampicillin-sensitive strains (26 of 28 with sensitivity data vs. 1 of 5, P=0.01). Infants with early-onset sepsis were more likely to die than uninfected infants (37 percent vs. 13 percent, P<0.001), especially if they were infected with gram-negative organisms.
Conclusions Early-onset sepsis remains an uncommon but potentially lethal problem among very-low-birth-weight infants. The change in pathogens over time from predominantly gram-positive to predominantly gram-negative requires confirmation by ongoing surveillance.
Methods
Population Data Base
We studied infants who were born at the 15 neonatal centers that belong to the Neonatal Research Network of the National Institute of Child Health and Human Development (NICHD). The network maintains a registry of all very-low-birth-weight infants (those weighing between 401 and 1500 g) who are born or cared for at participating centers.17 Information on infants weighing between 401 and 500 g began to be included in the data base in 1993.1 Trained research nurses collect data on mothers at the time of delivery and on infants from birth to 120 days, hospital discharge, or death, whichever comes first. The registry includes information on early-onset sepsis as documented by positive blood cultures, the type of infecting organism, the mother's age and marital status, any complications of the pregnancy, characteristics of the labor and delivery, characteristics of the infant, other medical conditions in the infant, and the infant's final status. In September 1998, surveillance for infection was expanded to include information on maternal antibiotic use during the hospitalization for delivery, the dates and results of all blood and cerebrospinal fluid cultures in the infant, C-reactive protein levels, any indwelling lines, use of antibiotics in the newborn, and the susceptibility patterns of the pathogens causing cases of early-onset sepsis. During the hospitalization for delivery, all antibiotic use in the mother within seven days before delivery was recorded. We have no information on the use of antibiotics by the mother before admission.
Outcomes
Early-onset sepsis was defined by a positive culture of blood drawn within 72 hours after birth. Cultures that were positive for organisms that are generally considered to be contaminants — including corynebacterium, propionibacterium, and diphtheroids — were excluded from analysis. Cultures that tested positive for coagulase-negative staphylococci were reviewed to distinguish definite or possible coagulase-negative staphylococcal infections from the effects of contaminants. Definite infection was defined by two positive cultures of blood specimens drawn within two days of each other or one positive culture and a blood C-reactive protein level greater than 1 mg per deciliter18 within two days after the blood culture; possible infection was defined by one positive culture and treatment for at least five days with vancomycin, oxacillin, a semisynthetic antistaphylococcal agent, or another drug to which the organism was susceptible. Contamination was considered probable if there was one positive culture without an elevated C-reactive protein level or antibiotic therapy as outlined above. Infants with definite or possible coagulase-negative staphylococcal sepsis were included in the analysis. For all other pathogens, sepsis was defined by the presence of the organism in cultured blood. To identify possible changes in the distribution of pathogens over time, we compared the pathogen-specific infection rates among infants enrolled in the very-low-birth-weight registry of the NICHD Neonatal Research Network between 1991 and 19931 with those in the recent birth cohort.
Gestational age was either determined by obstetrical methods including dating from the last menstrual period and the use of prenatal ultrasonography or estimated by a neonatologist on the basis of physical and neurologic criteria. Intraventricular hemorrhage was graded according to the method of Papile et al.19 Necrotizing enterocolitis was classified according to the system of Bell et al.20 For this analysis, respiratory distress syndrome was defined by the following: the need for supplementary oxygen at 6 to 24 hours of life; clinical features of respiratory distress (including grunting, nasal flaring, and chest retractions) within the first 24 hours; the need for respiratory support (continuous positive airway pressure or ventilation) during the first 24 hours; and an abnormal chest radiograph within the first 24 hours. Chronic lung disease or bronchopulmonary dysplasia was defined by the need for supplementary oxygen at 36 weeks of postmenstrual age. The study was approved by the institutional review board at each study site, and informed-consent procedures were followed as required at each site.
Statistical Analysis
Early-onset sepsis was treated as a binary variable. We assessed associations between early-onset sepsis and maternal and neonatal variables, characteristics of the hospital course, and death. Chi-square tests or Fisher's exact tests were used to compare the rates of early-onset sepsis among centers, to evaluate changes between the two birth cohorts, and to assess the relations between maternal ampicillin use and Escherichia coli sepsis and between the type of pathogen and the risk of death. Logistic-regression models were used to assess associations between early-onset sepsis and maternal variables, neonatal characteristics, complications, and the risk of death. Linear regression models were used to compare infants with and without early-onset sepsis in terms of the duration of umbilical catheter use, mechanical ventilation, and the hospital stay. Gestational age was included as a variable in all models.
Infants from both single and multiple gestations were included. Because of the possible correlation of findings between twins or among higher-order multiples, the association between maternal variables and early-onset sepsis was explored in two ways by logistic regression. First, a model was used in which all observations among infants were assumed to be independent. Next, infants were clustered according to the mother in models that used the generalized-estimating-equation method to account for the correlation between siblings. Robust estimators of variance were used, with the working correlation structure first specified as independent and in the second model as exchangeable. Analysis of clustered data was performed with the software package SUDAAN.21 All other analyses were completed with SAS software.22
Results
Rates of Early-Onset Sepsis
Between September 1, 1998, and August 31, 2000, 6204 very-low-birth-weight neonates were born at network centers; 757 of these infants (12 percent) did not have an early blood culture performed and were excluded. Over half (51 percent) of the neonates without blood cultures had extremely low birth weights (401 to 750 g) and received no or limited care. We studied the 5447 infants who had one or more blood cultures during the first three days of life. Eighty-four infants had early-onset sepsis — a rate of 15.4 per 1000 live births of infants weighing 401 to 1500 g. The incidence of early-onset sepsis in a given center ranged from 4 to 27 per 1000 live births (differences were not significant). On the basis of the findings in an earlier published study involving network centers,1 we determined that there was a nonsignificant decline in the rate of early-onset sepsis between 1991 to 1993 and 1998 to 2000 (from 19.3 per 1000 very-low-birth-weight infants to 15.4 per 1000 very-low-birth-weight infants). Fewer infants in the earlier cohort (255 of 7861 [3 percent]) were excluded from the analysis because they had not been evaluated for early sepsis.1 Demographic characteristics of the infants in the two cohorts are presented in Table 1.
|
The majority of early-onset infections in the recent cohort (60.7 percent) were caused by gram-negative organisms (Table 2). E. coli was the single most common pathogen (accounting for 44.0 percent of all infections). Gram-positive pathogens accounted for 36.9 percent of infections. Fungal infections were rare.
|
|
Information on sensitivity to antibiotics was available for 33 of the 37 isolates of E. coli from infants in the recent birth cohort: 28 of the 33 isolates (85 percent) were resistant to ampicillin and 1 of the 33 (3 percent) was resistant to gentamicin. All 26 E. coli isolates tested for sensitivity to selected third-generation cephalosporins were sensitive to these agents. Although the numbers were small, the mothers of infants with ampicillin-resistant strains of E. coli were significantly more likely to have received intrapartum ampicillin than were those with ampicillin-sensitive strains (26 of 28 [93 percent] vs. 1 of 5 [20 percent], P=0.01). Sensitivity data were not collected between 1991 and 1993, so it is not possible to compare the two cohorts.
Maternal Antibiotic Use
Twenty percent of the infants born in 1998 to 2000 were the products of multiple gestations, and a total of 4887 mothers were included in the analysis. A total of 65 percent of these women (3186 of 4887) received antibiotics during the hospitalization for delivery; 26 percent of these received only one dose of antibiotics, and 74 percent received two or more doses. Mothers who did not receive antibiotics were significantly more likely to have delivered within two hours after hospital admission than were those who received antibiotics (25 percent vs. 4 percent, P<0.001).
During the 72 hours before delivery, 63 percent of mothers received antibiotics; 41 percent of these women were treated less than 6 hours before delivery, 24 percent were treated 6 to 24 hours before delivery, 15 percent were treated 25 to 48 hours before delivery, and 20 percent were treated more than 48 hours before delivery. The most frequently prescribed antibiotics were ampicillin (given to 49 percent of mothers), penicillin (14 percent), and erythromycin (13 percent). Most mothers (68 percent) received only one antibiotic, 25 percent received two antibiotics, and 7 percent received three to five antibiotics during the 72 hours before delivery. There were no significant differences in the rates of early-onset sepsis between infants whose mothers received antibiotics during their hospitalization or during the 72 hours before delivery and those whose mothers did not receive antibiotics (Table 4). No association was found between any maternal antibiotic use during the hospitalization for delivery and either the risk of E. coli or group B streptococcal sepsis or the risk of death due to either of these infections (data not shown). The proportion of infants with E. coli sepsis was higher among those whose mothers received ampicillin within 72 hours before delivery than among those whose mothers did not (25 of 2348 [1.1 percent] vs. 12 of 3099 [0.4 percent], unadjusted P=0.004). However, the difference was not significant after the analysis was adjusted for gestational age and the interval between membrane rupture and delivery.
|
Selected characteristics of the mothers and the infants are summarized in Table 4. A longer interval between the rupture of the membranes and delivery was a strong risk factor for early-onset sepsis. However, 41 percent of infected infants were born within six hours after membrane rupture. Mothers with a longer interval between membrane rupture and delivery were more likely to have received antibiotics, but the association between this interval and the risk of early-onset sepsis did not differ between women who received antibiotics and those who did not (data not shown).
Gestational age and birth weight were inversely associated with the risk of early-onset sepsis, but the presence of intrauterine growth restriction was associated with a reduced risk. The race and sex of the infant were not associated with the risk of early-onset sepsis (Table 4).
Neonatal Morbidity and Mortality
After adjustment for gestational age, infants with early-onset sepsis had a significantly higher risk of respiratory distress syndrome (63 percent vs. 43 percent; odds ratio, 1.8 [95 percent confidence interval, 1.1 to 2.9]; P=0.02), severe intraventricular hemorrhage or periventricular leukomalacia (32 percent vs. 12 percent; odds ratio, 3.2 [95 percent confidence interval, 1.9 to 5.5]; P<0.001), and bronchopulmonary dysplasia (62 percent vs. 35 percent; odds ratio, 2.4 [95 percent confidence interval, 1.2 to 4.7]; P=0.01). Although the proportion of infants with these conditions was higher among infants infected with gram-negative organisms than those infected with gram-positive organisms, the differences were not significant (data not shown). There were no significant differences in the duration of use of an umbilical catheter, the duration of mechanical ventilation, or the length of the hospital stay between infants with early-onset sepsis who survived to 120 days or to discharge and those without early-onset sepsis who survived that long (data not shown). Among all infants and among those who survived to 120 days or to discharge, a history of early-onset sepsis was not associated with an increased risk of late-onset sepsis or necrotizing enterocolitis (data not shown).
During follow-up to 120 days, 734 of the 5447 infants (13 percent) died. Death was more common among infants with a lower birth weight and a lower gestational age (data not shown). Infants with early-onset sepsis were significantly more likely to die within 120 days, and in particular to die within 3 days after birth, than were those without early-onset sepsis (Table 5). These differences persisted even after adjustment for gestational age. The proportion of infants who died was higher among those infected with gram-negative organisms (21 of 51 [41 percent]) than among those infected with gram-positive organisms (8 of 31 [26 percent]), but the difference was not significant. Infants with gram-negative sepsis were more likely than infants infected with other agents to die within three days after birth (15 of 51 [29 percent] vs. 2 of 33 [6 percent], P=0.01).
|
Early-onset sepsis remains a potentially lethal but uncommon problem among very-low-birth-weight infants. Although only 1.5 percent of the 5447 very-low-birth-weight neonates in this cohort had early-onset sepsis, 37 percent of those with sepsis died. Extreme prematurity was the greatest risk factor for early-onset sepsis. Although the risk of early-onset sepsis was increased when the membranes ruptured more than 24 hours before delivery, 41 percent of infants with early-onset sepsis were born less than 6 hours after membrane rupture — an interval that many clinicians would not find worrisome. It has been suggested that as many as 85 percent of early preterm births are associated with intrauterine infection before membrane rupture.23 Moreover, there are increased risks of histologic chorioamnionitis, infected amniotic fluid, and a fetal inflammatory response associated with lower gestational age at birth.24,25 Early-onset sepsis is associated with an increased risk of several complications of prematurity, including respiratory distress syndrome, bronchopulmonary dysplasia, and severe intraventricular hemorrhage or periventricular leukomalacia. Further study is required to assess the effect of early-onset sepsis and the cytokine response to infection on these adverse outcomes of prematurity and on long-term neurodevelopmental outcome.
There has been a marked increase in recent years in the intrapartum use of antibiotics. This increase is due in part to national programs designed to reduce the vertical transmission of group B streptococcal infection, but it also reflects an effort to reduce the risk of neonatal illness when there is preterm rupture of the membranes. For the prevention of neonatal group B streptococcal infections, antibiotics are recommended during labor when there has been a positive rectovaginal culture for group B streptococcus before labor or when there are risk factors for neonatal infection (including labor at less than 37 weeks of gestation).6 All mothers of infants in the recent birth cohort we studied had at least one risk factor for neonatal group B streptococcal infection, but in only 65 percent of cases did they receive antibiotics. We have no data on the reasons why antibiotics were or were not given. The observation that some women gave birth soon after hospital admission may partially explain their not receiving antibiotics during labor and delivery.
Recent reports suggest that the intrapartum use of antibiotics has reduced the prevalence of early-onset group B streptococcal infections.6,9,10 We noted a reduction in the incidence of early-onset group B streptococcal sepsis between the period from 1991 to 1993 and the period from 1998 to 2000. However, there was also a significant increase in the incidence of E. coli infections during the same interval. Although more infants weighing between 401 and 500 g were included in the later cohort, this difference did not explain our findings, since no infant weighing less than 500 g was infected with either group B streptococci or E. coli. Historically, the predominant organisms associated with neonatal sepsis have changed over time. Gram-negative organisms, especially E. coli, were the most common causes of cases of neonatal sepsis reported at Yale University from the late 1940s to the mid-1960s.26 Gram-negative organisms remain the most frequently reported cause of neonatal sepsis in developing countries.27
Antibiotic resistance is a major public health threat.28,29,30 Gram-negative organisms have both innate resistance to antibiotics and the ability to acquire resistance through new mechanisms that may be transferred from other pathogens.31 In our study, 85 percent of E. coli isolates were resistant to ampicillin.32,33 We do not have data to assess whether the high rate of resistance reflects antibiotic-resistance patterns in the neonatal intensive care units involved in the study or resistance patterns of genital flora in the mothers in our study or in ambulatory populations more generally. Furthermore, we have no data on antibiotic use during pregnancy but before the hospitalization for delivery.
Early-onset sepsis is an important cause of neonatal death. Gram-negative sepsis is particularly lethal. Further surveillance is warranted to determine whether the observed reduction in the incidence of early-onset gram-positive infections will be mirrored by a continued increase in the risk of infection with more virulent gram-negative organisms.
Supported by grants (U10 HD27851, U01 HD36790, U10 HD21364, U10 HD34216, U10 HD27871, U10 HD27856, U10 HD27853, U10 HD34167, U10 HD21373, U10 HD27904, U10 HD21397, U10 HD21415, U10 HD21385, U10 HD40689, U10 HD27880, and U10 HD27881) from the National Institutes of Health.
We are indebted to Mazie Tinsley for assistance in the preparation of the manuscript.
Source Information
From the Department of Pediatrics, Emory University School of Medicine, Atlanta (B.J.S.); Research Triangle Institute, Research Triangle Park, N.C. (N.H., W.K.P.); the Department of Pediatrics, Case Western Reserve University, Cleveland (A.A.F.); the National Institute of Child Health and Human Development, Bethesda, Md. (L.L.W.); the Department of Pediatrics, University of Alabama, Birmingham (W.A.C.); the Department of Pediatrics, Yale University School of Medicine, New Haven, Conn. (R.A.E.); the Department of Pediatrics, Indiana University School of Medicine, Indianapolis (J.A.L.); the Department of Pediatrics, College of Medicine, University of Cincinnati, Cincinnati (E.F.D.); the Department of Newborn Medicine, Brigham and Women's Hospital, Boston (A.R.S.); the University of Texas at Houston, Houston (J.E.T.); Women and Infant's Hospital, Providence, R.I. (W.O.); the Department of Pediatrics, University of Miami, Miami (C.R.B.); the Newborn Center, University of Tennessee–Memphis, Memphis (S.B.K.); the Division of Neonatal and Perinatal Medicine, Wayne State University, Detroit (S.S.); the Department of Pediatrics, University of Texas Southwestern Medical Center, Dallas (A.R.L.); the Division of Neonatology, Stanford University Medical Center, Palo Alto, Calif. (D.K.S.); and the Department of Pediatrics, University of New Mexico School of Medicine, Albuquerque (L.-A.P.).
Address reprint requests to Dr. Stoll at the Department of Pediatrics, Emory University School of Medicine, 2040 Ridgewood Dr., NE, Atlanta, GA 30322.
References
The following centers belonged to the Neonatal Research Network of the NICHD between 1996 and 2000, with the principal investigator and study coordinator (and the number of patients studied, in parentheses): Brown University (396): W. Oh and A. Hensman; Case Western Reserve University (298): A.A. Fanaroff and N. Newman; Emory University (312): B.J. Stoll and E. Hale; Harvard University (429): A.R. Stark and K. Fournier; Indiana University (547): J.A. Lemons and D. Appel; Stanford University (279): D.K. Stevenson and B. Ball; University of Alabama (476): W.A. Carlo and M. Collins; University of Cincinnati (509): E.F. Donovan and M. Mersmann; University of Miami (381): C.R. Bauer and A.M. Worth; University of New Mexico (262); L.-A. Papile and C. Backstrom; University of Tennessee (372): S.B. Korones and T. Hudson; University of Texas–Dallas (337): A.R. Laptook and S. Madison; University of Texas–Houston (305): J.E. Tyson and G. McDavid; Wayne State University (316): S. Shankaran and G. Muran; Yale University (228): R.A. Ehrenkranz and P. Gettner; NICHD: L.L. Wright and B.B. McClure; Research Triangle Institute: W.K. Poole and B. Hastings. The Steering Committee chairman was A.H. Jobe.
| |||||||||||||||||||||||||||||||||||||||||||||||||
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2009 Massachusetts Medical Society. All rights reserved. |
Previous
