 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
PEDIATRICS Vol. 103 No. 6 June 1999, p. e77
, and
From the Departments of * Pediatrics and § Gynecology and
Obstetrics, Stanford University, School of Medicine, Stanford,
California 94305; and Maternal and Child Health Program, School of
Public Health, University of California, Berkeley, California.
 |
ABSTRACT |
|---|
Top
Abstract
Methods
Results
Discussion
References
|
|---|
Objective. To identify and to establish the prevalence of ORs for factors associated with increased risk for early-onset group B streptococcal (EOGBS) infection in neonates.
Study Design. Literature review and reanalysis of published data.
Results. Risk factors for EOGBS infection include group B
streptococcal (GBS)-positive vaginal culture at delivery (OR: 204),
GBS-positive rectovaginal culture at 28 (OR: 9.64) or 36 weeks
gestation (OR: 26.7), vaginal Strep B OIA test positive at delivery
(OR: 15.4), birth weight 
2500 g (OR: 7.37), gestation <37
weeks (OR: 4.83), gestation <28 weeks (OR: 21.7), prolonged rupture of
membranes (PROM) >18 hours (OR: 7.28), intrapartum fever >37.5°C
(OR: 4.05), intrapartum fever, PROM, or prematurity (OR: 9.74),
intrapartum fever or PROM at term (OR: 11.5), chorioamnionitis (OR:
6.43). Chorioamnionitis is reported in most (88%) cases in which
neonatal infection occurred despite intrapartum maternal antibiotic
therapy. ORs could not be estimated for maternal GBS bacteriuria during pregnancy, with preterm premature rupture of membranes, or with a sibling or twin with invasive GBS disease, but these findings seem to
be associated with a very high risk. Multiple gestation is not an
independent risk factor for GBS infection.
Conclusions. Mothers with GBS bacteriuria during pregnancy, with another child with GBS disease, or with chorioamnionitis should receive empirical intrapartum antibiotic treatment. Their infants should have complete diagnostic evaluations and receive empirical treatment until infection is excluded by observation and negative cultures because of their particularly high risk for EOGBS infection. Either screening with cultures at 28 weeks gestation or identification of clinical risk factors, ie, PROM, intrapartum fever, or prematurity, may identify parturients whose infants include 65% of those with EOGBS infection. Intrapartum screening using the Strep B OIA rapid test identifies more at-risk infants (75%) than any other method. These risk identifiers may permit judicious selection of patients for prophylactic interventions. Key words: group B streptococcus, neonatal sepsis, early-onset sepsis, risk factors, prevention.
Early-onset group B streptococcal sepsis (EOGBS) has been
the leading cause of death attributable to infection in newborn infants
for nearly 3 decades,1 with >6000 cases a year in the
United States.2 The attack rate has not changed over the
past 20 years, but the case-fatality rate has declined from ~50% to
between 10% and 15%.3-5 Long-term morbidity among
survivors, particularly neurodevelopmental disabilities in those with
meningitis, remain distressingly common.6-9 EOGBS disease
may be rapidly progressive and many infants (especially at term) do not
exhibit clinical signs of infection initially5; therefore,
empirical therapy is often initiated because of clinical risk or
minimal signs of disease and continued until infection has been
excluded by laboratory studies and a period of observation. This
aggressive approach has been associated with improved outcomes for
infants with this disease,10,11 but also may lead to
acute-care hospitalization and treatment of >100 000 newborn infants
yearly in the United States. Social impacts are substantial, ranging
from neonatal death to impaired mother-infant bonding and delayed or
ineffective establishment of breastfeeding in healthy infants who are
separated from their mothers for empirical therapy. Economic costs of
empirical treatment of a large number of infants at risk, treatment of
infants with confirmed infection, and long-term care for survivors were
estimated to exceed $700 million/year in 1985.12
Therefore, effective methods for prevention have been sought for >2
decades.13
In July 1992, the American College of Obstetricians and Gynecologists
(ACOG) issued a technical bulletin on group B streptococcal infections
in pregnancy,14 and in November of 1992, the American
Academy of Pediatrics (AAP) published its guidelines for prevention of
group B streptococcal infection by chemoprophylaxis.15 Both statements endorsed intrapartum antimicrobial chemoprophylaxis to
reduce the incidence of early-onset neonatal disease. The AAP advocated
screening for GBS colonization at 28 weeks gestation and giving
prophylaxis to GBS-colonized mothers who have clinical risk factors.
ACOG contended that antepartum screening cultures are not useful, and
advocated prophylaxis for all mothers with clinical risk factors.
Several commentaries and clarifications have addressed these
differences,16-20 but both obstetric19,21
and pediatric22 practices remained inconsistent. Seeking a
consensus, the Centers for Disease Control and Prevention (CDC)
convened a multidisciplinary conference in 1995 and issued
recommendations for prevention of perinatal GBS disease in May of
1996.23 ACOG endorsed these recommendations, but expressed
concerns about lack of clinical trials or experience with the proposed
strategies and potential sequelae of widespread intrapartum antibiotic
therapy.24 The AAP Committees on Infectious Disease and on
the Fetus and Newborn also ratified those
recommendations.25 However, the recommendations include
two strategies for identifying candidates for intrapartum prophylaxis,
and yet another strategy has been proposed by Gotoff and
Boyer,26 indicating that efforts to achieve consensus have
not been entirely successful.27 Consensus regarding
management of infants, particularly those whose mothers have received
intrapartum prophylaxis, has been even more elusive. The sample
algorithm for management of such infants23 that was
provided by the CDC and modified slightly by the AAP25 is
only weakly endorsed, ending with admonitions that "other management
approaches, developed by individual physicians or institutions, may be
appropriate alternatives"23 and it is "suggested but
not an exclusive approach to management."25
Uncertainty about and inconsistent perceptions of clinical factors that
may identify women who are at risk of having an infant with EOGBS
disease are major barriers to achievement of the desired consensus. The
medical literature regarding these subjects, which is extensive and
highly technical in many respects, has been reviewed extensively,128-31 discussed,20,2332-34 and debated.1835-37 Despite metaanalyses,38
critical reviews,39 cost-benefit studies,3640-43 and decision modeling,44
misapprehensions of risk endure, significant observations are often
overlooked, and the essential requirements for high efficacy of
interventions intended to reduce the prevalence of neonatal GBS disease
have not been articulated clearly. Because of the low prevalence of
this disease, the number of enrollees that would be required for
randomized trials to evaluate potential treatment or prevention
regimens has been described as prohibitive.20 For the
present, at least, rational approaches to prevention of GBS disease in
neonates will have to be based on existing knowledge.
Accurate ascertainment of risk to enable selection of a target
population for intervention is a crucial first step in development of
prevention programs. Effective programs for prevention of neonatal GBS
disease should address populations that include the fewest possible
mothers along with the greatest number of infants who will have GBS
infections. Intervention strategies that address a very small but
extremely high-risk segment of the population may have little impact on
the overall incidence of disease. If both the prevalence of a risk
factor and the associated OR for GBS disease are known, the percentage
of infants with GBS disease born to mothers identified by that risk
factor can be calculated. Then, the distribution of GBS cases can be
used to assess the efficiency with which GBS prevention strategies
identify infants at risk. Evaluation and comparison of risk
ascertainment strategies depends on knowledge of the identity,
prevalence, and ORs for clinical findings associated with EOGBS, which
must be both reliable and internally consistent. We have reviewed
critically and reevaluated published data with meticulous attention to
technical details and mathematical rigor to identify and establish the
prevalence and ORs for risk factors of EOGBS disease.
Clinical conditions associated with EOGBS were identified
through a search of MEDLINE, as well as reference lists for those articles and recent reviews of neonatal GBS
disease.1,17,30,34,45 Prevalence rates for prematurity and
low birth weight (LBW) were abstracted from the 1991 national46 or 1992 California birth cohorts. All available
data were evaluated to establish the prevalence of each identified risk
factor. Because the attack rate observed in association with a risk
factor is dependent on the overall attack rate in the population
included in that particular study, as well as the associated OR for GBS
sepsis, raw attack rates should not be extrapolated from such samples
to other populations. Although relative risk is easy to understand and
apply to clinical practice, this risk parameter is not suitable for
application to other populations or distribution of risk among
subgroups within populations. Therefore, the OR, which is the most
robust parameter for such extrapolations,47 is used
throughout this analysis. All risk predictions are based on the same
demographics (GBS colonization and attack rates, prevalence of
prematurity, etc); therefore, risk estimates for different subgroups
could be compared directly. Whenever possible, multiple sources were
consulted to ensure external validity for each prevalence and OR
estimate. Infants were considered to have GBS infection only if GBS
were recovered from nonpermissive cultures, ie, blood or cerebrospinal
fluid. Infants with clinical findings consistent with sepsis or with
positive tests for urinary GBS antigen, who did not have positive
nonpermissive cultures, were not considered to have GBS infection. CIs
for ORs were calculated using the method of Cornfield,48
and CIs for prevalence rates were calculated using the
binomial49 or normal50 distributions, as
appropriate. Ratios were compared using the G test of
independence, a preferred alternative to the
METHODS
Top
Abstract
Methods
Results
Discussion
References
2 or Fisher
exact tests for contingency tables in which the marginal totals are not
fixed by experimental design.51 Gradients of risk in
groups ordered by density of GBS colonization, gestational age, or
birth weight were evaluated using the method of
Bartholomew.52 Formulas for probabilities of GBS disease
in the presence or absence of a risk factor were derived from the
definition of OR.53 For factors with >2 possible values,
eg, gestational age, the probability of falling into the lowest-risk
category for that factor and having GBS disease (x = PF1+D) was calculated from the equation:
solving for x numerically using an iterative method.
(PFi is the population fraction with risk factor
i, ORi is the corresponding OR,
PD is the prevalence of disease, and the
subscript 1 denotes the lowest risk category.) For dichotomous factors,
eg, presence or absence of intrapartum fever, the probability that an
infant has both the risk factor and GBS disease (x = PF+D) was calculated by solving the equation:
for x using the quadratic formula.
(PF is the prevalence of the risk factor,
OR is the corresponding OR, PD is the
prevalence of disease.) Other elements of each contingency table were
derived from these values. Sensitivity, specificity, and positive and predictive negative values were calculated using standard
formulas.54
![(<UP>1 −OR</UP>)<UP>x<SUP>2</SUP> + </UP>[<UP>OR·P<SUB>F</SUB> + </UP>(<UP>1 − P</UP><SUB><UP>F</UP></SUB>)<UP> + </UP>(<UP>OR − 1</UP>)<UP>P</UP><SUB><UP>D</UP></SUB>]<UP>x − OR·P<SUB>D</SUB>·P<SUB>F</SUB> = 0</UP>](/content/vol103/issue6/fulltext/e77/img002.gif)
| |
RESULTS |
|---|
|
Top
Abstract
Methods
Results
Discussion
References
|
|---|
Population Attack Rates for EOGBS
Reported population attack rates for EOGBS disease in the United States range from .76 to 5.46 cases per 1000 live births (Table 1). Data gathered since 1985 suggest a typical attack rate of ~1.8 cases per 1000 deliveries (.0018). In a population with an intrapartum vaginal colonization rate of 14.7%,55 the risk among infants born to GBS carriers would be ~12.2/1000 live births (.0122), which is at the lower end of empirically determined attack rates among such infants (.011-.092).56-63 Because this attack rate is based on culture-proven GBS disease, and because single small-volume blood cultures routinely used in neonates do not yield an organism in many infants with invasive bacterial disease,64,65 this undoubtedly underestimates the true incidence of GBS disease. For decision analysis, the attack is estimated to be 3 cases per 1000 live births (1.8 culture-proven cases per 1000 live births divided by a culture sensitivity of 60%).65
|
Maternal GBS Colonization
Because the association between the presence of the organism in
the birth canal and invasive neonatal disease was recognized soon after
the emergence of GBS as a major cause of neonatal infections in the
late 1960s, the epidemiology of maternal colonization has been
investigated extensively. These studies have defined clearly the
requirements for optimal ascertainment of maternal colonization. To
maintain viability of the organism en route to the laboratory, swabs or
washes should be placed immediately into transport medium, eg, Amies'
or Stuart's medium.66 Specimens should be inoculated into
selective broth medium, eg, Todd-Hewitt medium supplemented with
colistin and nalidixic acid, and subcultured 18 to 24 hours later on
blood agar plates.5867-70 Direct plating of
specimens onto blood agar plates may underestimate colonization
rates by up to 50%.67,6871-76 Swabs or washes of the
lower third of the vagina are more likely to reveal colonization than
those obtained from the vaginal fornices or endocervix.77
Rectovaginal colonization rates exceed vaginal colonization rates by
50%57,78,79 (Table
2). Both vaginal and rectal colonization
may be persistent, transient, or intermittent,5779-82 limiting the predictive value of screening cultures for colonization on
any future date.
|
Reported colonization rates in pregnant women range from 2% to 35%.1,45 This variation has been attributed to use of different culture methods, particularly culture of different anatomic sites and use of selective broth versus direct inoculation of agar plates. Geographic differences, with low colonization rates reported from countries in which neonatal GBS infection is uncommon, also contribute. Even with consistent technique, cross-sectional studies from the United States in which vaginal or rectovaginal specimens from pregnant women were cultured in selective broth medium show substantial variation in colonization rates (Table 2). Ethnicity, maternal age and parity, marital status, education, smoking, and frequent intercourse with multiple partners80,83 also may influence the prevalence of colonization, but relationships among these factors and GBS colonization may be inconsistent. For example, Hispanic patients had the highest colonization rates in New York but the lowest rates in Washington, Oklahoma, and Louisiana, compared with other ethnic groups.83 These risk factors may raise clinical suspicion for GBS colonization but do not identify high-risk women for whom selective screening might be appropriate.83
Maternal vaginal colonization with GBS is essentially a prerequisite for both early colonization of the newborn infant and EOGBS.56,5984-87 Data demonstrating the relationship between intrapartum colonization of the vagina and EOGBS sepsis (Table 3) are quite compelling (OR: 204; 95% CI: 100-419). Data that address the relationship between the density of maternal colonization and neonatal GBS sepsis have become available only recently.87,88 Compared with infants whose mothers are lightly colonized, infants whose mothers have heavy genitourinary GBS colonization are more likely to become colonized,57,60,85,89 and the risk for GBS infection is much greater in heavily colonized neonates.56 Yancey et al88 stratified maternal colonization within the last 2 weeks before delivery from 0 to 4+ according to the number of quadrants in which GBS colonies were observed on blood agar plates inoculated directly with vaginal swabs and streaked in a standardized sequence; a ranking of 0 was assigned for those women for whom GBS was isolated only in enriched broth.90 Regan et al87 defined heavy colonization at the time of delivery as isolation of GBS by direct culture on nonselective solid media and light colonization as isolation of GBS only by subculture of selective broth media; this definition has been adopted for this discussion. As shown in Table 4, these studies show gradients in risk with increasing colonization density (P < .005 by Bartholomew's test52). With heavy colonization in 71.8% of women with GBS-positive vaginal cultures76,91,92 and odds of infection that is 2.54 times greater in infants of heavily colonized, compared with lightly colonized, mothers,87 the ORs for EOGBS in infants of lightly and heavily colonized women are 97.1 and 247, respectively.
|
|
Results of cultures are not available for 36 to 48 hours, thus cultures obtained on presentation for delivery cannot guide intrapartum interventions or early initiation of neonatal treatment. Prediction of colonization status at delivery by performing cultures during pregnancy is imperfect, because women may lose or acquire GBS colonization between screening and delivery,57,59,84,93 producing false-positive or false-negative results. In 12 935 women screened for vaginal and endocervical colonization at 23 to 26 weeks, the OR for infection associated with positive cultures was only 1.65 (95% CI: .70-3.91), and only 30% of neonates with EOGBS disease were born to mothers with positive cultures.87 Yancey et al88 recently reported an OR for neonatal sepsis of only 4.36 (95% CI: 1.59-11. 9) in association with GBS-positive cultures of vaginal swabs obtained within 2 weeks before delivery,88 confirming that even vaginal cultures obtained late in gestation are much less effective than cultures obtained at delivery. Because the rectum may provide a reservoir from which the genitourinary tract may be intermittently recolonized,79,94 addition of rectal specimens or anorectal specimens to vaginal specimens has been recommended. This identifies up to twice as many colonized women,78,79 and increases the predictive value of antenatal cultures for vaginal colonization at delivery.72 Boyer et al55 found that positive rectovaginal screening cultures performed late in the first to early in the third trimester were associated with an OR of 29.4 (95% CI: 7.44-116) for EOGBS sepsis, which compares quite favorably with the OR associated with vaginal colonization identified at 23 to 26 weeks (1.65)83 or during the last 2 weeks of gestation (4.36).88 The AAP and CDC recommendations emphasize correctly that antepartum screening cultures must be performed using swabs from both the vaginal introitus and anorectum.15,23
The predictive values of screening cultures performed at different
stages of pregnancy are critical to selection of the optimal time for
antepartum screening. The predicted colonization rate at delivery,
C, is determined by the equation C = S × P + (1 
S) × (1 N), where S is the prevalence of a positive
screening culture and P and N are its positive
and negative predictive values for colonization at delivery,
respectively. These predictive values and the prevalence of vaginal
colonization are determined by the rates at which GBS rectovaginal
colonization is lost or acquired by colonized and noncolonized women,
respectively, and by the proportion of women with rectovaginal
colonization who have vaginal colonization. Arbitrary selection of
positive and negative predictive values of antepartum cultures
implicitly selects values for rates of loss and acquisition that are
almost certain to be different and inconsistent with the colonization
rate, which is determined by these conversion rates (colonization
rate = acquisition rate 
sum of acquisition and loss
rates). The resulting artifactual shifts in the vaginal colonization
rate expected at delivery distorts the predicted GBS attack rate. Only
one previously published model recognizes this significant source of
bias in decision analysis models43 that is a serious flaw
in the model on which the recent CDC recommendations are
based.44 For this analysis, predictive values for
antepartum cultures were derived from the most comprehensive single
dataset available.55 The rates of loss and acquisition of
rectovaginal colonization were determined to be 2.46% and .72%/week,
respectively,57 by linear regression of the log of
conversion fraction against time between cultures. The prevalence of
vaginal colonization among women with rectovaginal colonization in that
study was 64.5%. These rates of loss and acquisition of colonization
were used to calculate internally consistent predictive values for
antepartum rectovaginal cultures. Cultures at 28 weeks have positive
and negative predictive values for vaginal colonization at delivery of
48.7% and 95.3%, respectively; at 36 weeks the predictive values are
58.2 and 98.2% (Table 5). These
estimates are near the middle of the ranges of reported values
(corrected for different vaginal and rectovaginal colonization rates,
as appropriate),57,59,82,84 and closely approximate
recently reported empirical values (56.1% and 97.4% for vaginal
colonization).95
|
The association between vaginal colonization at delivery and EOGBS infection has been a powerful impetus for development of prevention strategies based on ascertainment of maternal GBS colonization. This approach is central to the strategies advocated by the AAP15 and the CDC,23 but there are barriers to implementation. Practical as well as epidemiologic considerations will influence selection of methods for identification of women who have or are likely to have vaginal GBS colonization at delivery. Among the greatest barriers to general acceptance of antenatal screening strategies is inability to ensure that screening cultures are performed consistently and that results are available at the site and time of delivery.23,96 A policy of antepartum screening may expose obstetricians to an untenable medicolegal risk, particularly in cases of EOGBS infection in an infant whose mother was not screened on schedule or for whom the screening culture result was not available in the labor and delivery suite, leading to omission of intrapartum prophylaxis.97,98 Antepartum screening cultures require collection of rectal, as well as vaginal, samples, which many obstetricians believe to be unacceptable to their patients.99 Therefore, a diagnostic method that permits rapid, easy, and accurate identification of GBS-colonized parturients on presentation for delivery would be ideal for ensuring immediate and consistent ascertainment of GBS colonization. Rectovaginal sampling is not necessary at delivery, because virtually all infants with EOGBS disease are born to mothers from whom GBS can be recovered from vaginal cultures. In fact, ascertainment of a larger pool of mothers with rectovaginal colonization would not identify additional infants who are likely to become infected but would enlarge and dilute the pool of women identified as being at risk for that neonatal outcome. Of the currently available methods, only the BioStar Strep B OIA test (Table 6) has performance characteristics sufficient for clinical utility in this application (Tables 5 and 6). Another recent comparison of rapid immunoassays is not included in this summary because a nonstandard sampling method was used.100 In a population in which 14.7% of women have vaginal colonization (heavy in 71.8%), 52.6%, 10.0%, and 37.3% of parturients with positive intrapartum Strep B OIA tests would be expected to have heavy, light, or no GBS colonization, respectively, and 2.2%, 3.0%, and 94.9% of those with a negative test result are expected to have heavy, light, or no colonization, respectively. Because this test must be performed by skilled laboratory personnel rather than at the bedside, it may be necessary to train a number of laboratory personnel so that the test will be available 24 hours/day. As with any test, such an increase in the number of personnel who perform it may be associated with performance inferior to that described in research protocols.
|
Prematurity and LBW
The excess risk of EOGBS disease in preterm and LBW infants has
been well-recognized for many years. Early reports noted that preterm101,102 and LBW101-103 infants were
overrepresented among infants with early-onset disease, a finding
confirmed by more recent case-control studies.4,35 Increased risk for infants with birth weights 2500 g has been recognized,86,104 and a gradient of increasing risk with
decreasing birth weights <2500 g has also been
delineated.3,5,55,105,106 Yancey et al88 have
recently demonstrated a progressive increase in risk for neonatal
sepsis in general with decreasing gestational age. ORs associated with
birth weight and gestational age are summarized in Table
7. This analysis confirms a statistically
significant gradient of risk with either decreasing gestational age or
birth weight. These variables are correlated, but risk is low for
infants with birth weights >2500 g and for infants born after 37 weeks
gestation. In a multivariate analysis, only the combination of LBW and
prematurity was correlated with neonatal pneumonia and
sepsis,107 thus preterm infants of birth weight >2500 g
and term infants (including those with birth weights 2500 g) may be
at a comparably low risk.
|
The prevalence of prematurity or LBW varies
regionally46 and locally106 and is an
important determinant of the population risk for GBS
disease.106 Because our objective was to evaluate the
impact of potential interventions on an entire population, demographic
distributions were determined using data from the most recent national
(1991) and California (1992) birth cohorts.46 In these
cohorts, the prevalence of prematurity (<37 weeks) was 10.3%, which
is the same as the prevalence of prematurity reported in a multistate
surveillance study of neonatal GBS disease.2 Only .8% of
all deliveries occurred before 28 weeks gestation. The rates of
prematurity were 62.5% among infants with birth weights 2500 g and
6.86% among larger infants. Birth weight was 2500 g in 37.5% of
preterm infants. Allocating the increased risk for preterm infants, as
described by Boyer et al,55 to those infants who weigh
2500 g permits calculation of an OR of 11.4 for preterm infants
2500 g in their population. Applying this relative risk to a
population reflecting California and national demographics yields an
overall relative risk for all preterm infants of 4.83. Distribution of the risk of GBS infection among the preterm infants according to the
ORs defined for each gestational age interval by the data of Yancey et
al88 allows estimation of the ORs and attack rates shown
in Table 10.
|
Onset of labor before term is correlated strongly with subclinical maternal infection,108 which may contribute to the high prevalence of EOGBS disease in premature infants. The separate contributions of preterm labor and preterm delivery, which are obviously strongly covariant, have not been determined, but spontaneous onset of labor before term should be considered to be a clinical indicator of possible intrauterine infection.
Prolonged or Premature Rupture of the Amniotic Membranes
Prolonged rupture of the amniotic membranes for >1855,88 to 20105 hours before delivery substantially increases the risk of neonatal GBS disease55,88,105,107 (Table 8). Because rupture of membranes for >18 hours was observed in <10% (2 of 21) of mothers whose infants developed GBS sepsis, a recent case-control study has suggested that prolonged rupture of membranes (PROM) >10 hours would be more sensitive.109 The significance of this observation remains uncertain, because criteria for sepsis were not limited to recovery of GBS from nonpermissive cultures, only term infants were included, and the sample size was small. The largest published series indicates that PROM >18 hours occurs in 12.5% of deliveries and is associated with an OR of 7.28 (95% CI: 4.42-12.0).55 Premature rupture of the amniotic membranes before the onset of labor occurs in 20% to 25% of pregnancies.110,111 Premature rupture of membranes is more common among mothers who are GBS-colonized110 or whose infants have GBS infection,111 particularly among preterm infants,5,112 but neither the attack rate nor the OR for these infants can be calculated from available data. Preterm premature rupture of membranes occurs in 1% to 2% of pregnancies113 and is a significant univariate risk factor for neonatal pneumonia or sepsis (OR: 5.2; 95% CI: 2.4-11.6), even after controlling for prematurity and LBW.107 Neonates born to GBS-colonized mothers after preterm premature rupture of membranes apparently are at extremely high risk with attack rates of 33% to 50%.114,115
|
Intrapartum Fever
Intrapartum temperature >37.5°C,55
>38.0°C,112 or fever without additional
definition45 are associated with an increased risk of
neonatal GBS infection. Two studies describe a greater likelihood of
intrapartum fever in mothers of infants with confirmed GBS
infection,111,112 but only one study55
provides data sufficient to calculate an OR for an intrapartum
temperature >37.5°C (OR: 4.05; 95% CI: 2.17-7.56). Although
pediatricians generally believe that higher maternal fever confers a
higher risk, there are no objective data to quantitate that
relationship. Cimolai and Roscoe reported GBS bacteremia in 8 of 13 infants born to mothers with a temperature >37.5°C but <38.0°C
and in 13 of 30 of infants born to mothers with fevers
38.0°C.116 These attack rates are remarkably high and
not significantly different. The prevalence of fever (1/1000
deliveries) in that population was much lower than that reported by
others (36-92/1000 deliveries),55,111,117 thus this study
does not refute the hypothesis that higher fevers are associated with
greater risk. Obstetricians may diagnose chorioamnionitis, which is
associated strongly with fetal and neonatal infection, in any women
with a high fever during labor.118 Women who receive
epidural analgesia are more likely to have a
fever,119-124 and their infants are more likely to be
evaluated for sepsis and to receive antibiotic therapy.124
During epidural analgesia, the maternal temperature increases at
.08120,122,123 to .14°C/hour,119 exceeding
38.0°C in 10% to 15% of women.121,122,124 However,
fevers exceeding 38.5°C seem to be quite rare. Use of a higher fever threshold for women who have had epidural analgesia may be
appropriate.124 Interventions may be appropriate for
infants born to mothers who have fevers during labor, but available
data do not identify specific temperature thresholds above which
prophylaxis, diagnostic screening, or empiric therapy are appropriate
or imperative.
Chorioamnionitis
Intrapartum fever accompanied by two or more additional signs, including fetal tachycardia, uterine tenderness, foul-smelling vaginal discharge, or maternal leukocytosis, occurs in 1.0% to 3.8% of parturients and is associated with neonatal GBS attack rates ranging from 6% to 20%.88,118125-128 The risk is presumably even greater in GBS-colonized women with chorioamnionitis. The only study that provided sufficient data for estimation of an OR found that chorioamnionitis, as defined above, occurs in 10.0% (95% CI: 9.0-11.0)88 of pregnancies and is associated with an OR for all early-onset neonatal sepsis of 6.42 (95% CI: 2.32-17.8). This may overstate the prevalence and underestimate the associated risk. Chorioamnionitis was almost universally present in mothers of infants who became septic despite intrapartum prophylaxis with an appropriate intravenous antibiotic (44 of 50 reported cases5,88,129,130). Mecredy et al118 reported bacteremia in 1 of 63 asymptomatic infants and 3 of 23 symptomatic infants whose mothers have received intrapartum treatment for chorioamnionitis. These observations suggest that chorioamnionitis is a marker of high risk for invasive GBS disease, even in infants whose mothers receive antibiotic prophylaxis.
Combined Intrapartum Factors
Seeking a collection of factors associated with a preponderance of
EOGBS cases, Boyer et al55 found that the presence of
intrapartum maternal fever (>37.5°C), PROM (>18 hours), and/or birth weight 2500 g identified a group including <20% of all infants but nearly 75% of those who develop GBS sepsis. Multistate risk assessments identified intrapartum fever, PROM, and/or prematurity in 75% of mothers of GBS cases in 1991 and 1992,111 but
in only 54% in 1995. Based on the data of Boyer et al,55 the probability of having one or more of these high-risk factors is
18.3%55 and the associated OR is 12.6 (95% CI:
7.17-22.2). Both Boyer et al55 and Rouse et
al44 observed PROM or maternal intrapartum fever in 7.5%
of term infants. Excluding infants 2500 g, the OR for infants >2500
g whose mothers had fever or PROM is 11.5 (95% CI: 5.78-23.1).
Applying this risk estimate to a population reflecting national
prevalences for prematurity and LBW, prematurity, PROM, or intrapartum
fever is expected in 17.1% of women and is associated with an OR of
9.74; the prevalence of birth weight 2500 g, PROM, or intrapartum
fever is 13.3% and the associated OR is 12.3.
Maternal GBS Bacteriuria
Infants born to women with GBS bacteriuria during pregnancy are
more frequently and more heavily colonized with GBS131 and
may be at increased risk for invasive GBS disease.56,132 In 1500 consecutive deliveries, Liston et al133 reported a
high prevalence of pneumonia (39%), suspected sepsis (44%), and GBS
bacteremia (6%) among 55 infants whose mothers had GBS in urine
cultures routinely obtained after delivery. Although this suggested an
association between maternal GBS bacteriuria and neonatal sepsis, this
attack rate was not significantly different from that in controls (0 of
41; P = .09). Wood and Dillon prospectively identified
a 2.5% prevalence of significant (105 organisms/mL) GBS
bacteriuria during pregnancy.134 Among 14 pregnancies with
this finding, 2 ended in intrauterine fetal demise and 2 were followed
by neonatal GBS infections (1 EOGBS scalp abscess and 1 late-onset
sepsis). In the only prospective study that provides data permitting
direct comparison of attack rates in infants born to mothers with and
without GBS bacteriuria, Møller et al135 also found a
2.5% prevalence of GBS bacteriuria among women prospectively screened
between 12 and 38 weeks. They reported 5 cases of confirmed GBS sepsis
among 68 infants born to women with compared with 0 cases among 2677 without GBS bacteriuria (P < .001). Persson et
al136 prospectively screened women for asymptomatic
bacteriuria on three occasions during pregnancy and identified 1 infant
with GBS disease among those born to 10 women with 105
GBS colony-forming units/mL in their urine. Six of these women, but not
the mother of the infected infant, received antepartum antibiotics.
These studies vary widely with respect to definitions of significant
bacteriuria, ranging from any GBS in the urine135 to
requiring two sequential urine cultures with 105
organisms/mL,136 but bacteriuria (even if urine obtained by bladder aspiration is sterile) may be an indication of heavy GBS
colonization.136 This dataset is very small, and these
data may be skewed by selection or reporting bias, but the observation
that 7 of 92 infants born to mothers with GBS bacteriuria in these
prospective studies developed GBS infection suggests an attack rate of
76 (95% CI: 51-101) per 1000 such infants.
Sibling With GBS Infection
Although having had a previous infant with invasive GBS disease is accepted widely as placing a mother at high risk in subsequent pregnancies, only four instances have been reported in which neonatal GBS infection followed more than one pregnancy in the same mother.137-139 We have seen other instances of this event, so it probably is not as unusual as is suggested by the rarity of reported cases. Women may remain colonized with the same strain of GBS for prolonged periods140 and may fail to develop protective levels of type-specific serum antibodies despite long-term colonization.140,141 It is likely that the risk in subsequent pregnancies is very high for women who have had a child with EOGBS disease, but this risk cannot be quantitated more precisely.
Multiple Gestation
Several reviews1,17,30,34,36 and the 1992 AAP
guidelines for prevention of group B streptococcal
infection15 state that the prevalence of neonatal GBS
disease is increased in twins, but published data do not support this
belief. The first suggestion of an increased risk in twin gestations
was an incidental observation of twins concordant for disease in a
patient sample collected for another purpose.142 The
prevalence of infection in twins <2500 g birth weight (3 of 56) was
not significantly different from that (7 of 603) in LBW singletons
(P > .05 by both the G test of
independence51 and 2 test with Yate's
continuity correction)143 and there were no cases of GBS
infection in twins of birth weight 2500 g. Edwards et
al144 described 5 sets of twins and 1 set of triplets in
which at least 1 infant had EOGBS disease, but provided no comparison
with singleton pregnancies. In a survey of GBS infections in
Denmark,104 univariate analysis noted a relative risk of
6.9 for twins compared with singletons. Of the 6 twins with EOGBS
disease, 5 were LBW, a group for which the relative risk in the same
study was 14.8, thus apparent increased risk for twins in that
population is attributable to the expected high prevalence of LBW in
multiple gestations. A geographic cohort of >44 000 births in Western
Canada exhibited no association between multiple gestation and EOGBS
infection (P > .25).145 Among 848 products of multiple gestation in that study, no culture-positive EOGBS
case was observed; both cases of presumed GBS infection (infants with
positive urine latex agglutination tests) occurred in preterm (<35
weeks) infants with PROM. A recent case-control study failed to show
any differential risk for twins (OR 95% CI: .24-2.71).111 Other large population studies have also failed to detect an association.3,107 Multiple gestation
is not an independent risk factor for GBS infection.
However, if 1 neonate from a multiple gestation has GBS infection, the risk of disease in the other(s) may be substantial. Concordance of EOGBS disease has been reported in twins,142,144 triplets,144 and quintuplets.146 Pass142 reported EOGBS sepsis in 1 of 2 twins of index cases. In Edwards' series,144 2 of the 5 twins of index cases with early-onset sepsis and 2 of 6 twins of index cases with late-onset infections also developed GBS disease. In the triplets described in their report, 1 infant had EOGBS disease, 1 infant died at 2 hours of age (no bacterial cultures were obtained), and 1 infant remained well. Although these reports may be skewed by preferential reporting of concordant cases, neonatal GBS colonization and lack of protective maternal antibody are likely to be the same in all products of a multiple birth. These data suggest that the prevalence of GBS infection may exceed 40% among other members of a multiple birth group in which 1 infant has EOGBS disease.
Other Factors
Race or ethnicity,3,111 maternal age,3,111 internal monitoring for >12 hours,88 meconium staining,107 asphyxia,4 and fetal acidosis,147 and other factors also may be associated with increased risk. These variables may be covariant with GBS colonization, gestational age at delivery, duration of ruptured membranes, or other factors; therefore, their independent contributions are not delineated readily. Several of these factors are only apparent at or after delivery, and, therefore, are of little utility to strategies for prevention.
| |
DISCUSSION |
|---|
|
Top
Abstract
Methods
Results
Discussion
References
|
|---|
This review of the literature allows classification of clinical risk factors into two groups: factors that are associated with very high attack rates (>50/1000 live births) but that are relatively rare, and other factors that are more prevalent yet associated with less extreme increases in risk (attack rates of 10-25/1000). Clinical factors associated with the highest risks are shown in Table 9. Data related to maternal GBS bacteriuria or an affected sibling or twin are not sufficient for calculation of ORs, but the apparent very high attack rates are appropriately cause for serious concern about these infants. The consensus recommendation for intrapartum prophylaxis of their mothers23 is well-justified, and evaluation and treatment of the infants is probably also appropriate, at least until infection is excluded by a period of observation and negative cultures. The apparently very high risk in infants born to GBS-colonized mothers with preterm premature rupture of membranes justifies empiric treatment of these women at least until cultures have excluded vaginal colonization with GBS. Infants born to women with chorioamnionitis are at high risk for neonatal sepsis, attributable to both GBS and other organisms, and may be uniquely at risk for infection even after maternal intrapartum antibiotic therapy. We concur with Merenstein's recommendation that infants born to women with chorioamnionitis should have a complete evaluation for GBS sepsis and receive empirical treatment until infection is ruled out, regardless of whether their mothers received intrapartum treatment.130
|
Estimates for the prevalence of and ORs for clinical factors associated
with less extreme apparent risk are summarized in Table
10. These parameters were used to
calculate the expected probability of a GBS case among patients with
each risk factor (the positive predictive value of the observation).
From this probability, the sensitivity, specificity, and predictive values, and the proportion of cases associated with each of the risk
factors identified by this review have been calculated (Table 10), based on a presumed population attack rate of 3/1000 live births. For
populations with different attack rates, the estimated risk-specific attack rates can be adjusted proportionately. This risk allocation indicates that the most sensitive and specific tool for identification of infants who are at risk for EOGBS infection is a maternal vaginal culture obtained at the time of delivery. This test identifies a group
that includes <15% of all parturients, only 2% of whom have GBS
infection but who account for >97% of all infants who develop EOGBS
disease. However, the delay in availability of these results make this
test useless for selection of patients for preventive interventions.
Screening with rectovaginal cultures at 36 weeks gestation identifies
just over 20% of the population that includes 57% of EOGBS
cases, missing 35% of the GBS cases because the women deliver before
screening is scheduled. Screening at 28 weeks reduces the proportion of
women who deliver before screening to <1%, but this gain is largely
offset by the lower predictive values associated with earlier
screening. That strategy identifies 23% of the population that
includes only 65% of infants with EOGBS. Currently, the best available method for screening for maternal vaginal colonization is the
Strep B OIA, which can be used for nearly all deliveries (except
deliveries that are precipitous or that occur before arrival in the
obstetrical suite). That test identifies a group of <17% of all
mothers that includes >75% of infants with early-onset disease, thus this should be the preferred method for ascertainment of
risk based on maternal vaginal colonization at delivery. Of the other
clinical risk factors, only PROM and the combination of intrapartum
fever, PROM, and birth weight 2500 g identify subpopulations (of
12.5% and 18.3% of the population, respectively) that include >50%
of the infants who will have EOGBS disease. Although GBS infection is
expected in <2% of infants with any one of these risk factors, they
may identify mothers whose infants are at sufficient risk to justify
preventive interventions.
The CDC has recommended two strategies for identifying women
whose infants are at risk for EOGBS sepsis and advocates intrapartum antibiotic prophylaxis for these at-risk populations.23 These recommendations are based on a projected reduction in the EOGBS
attack rate by 86% with application of the more aggressive of these
approaches. The risk analysis presented in Table 10 indicates that no
group that is defined by a single risk identifier includes
85% of the infants expected to have invasive GBS disease, which would be required to achieve the efficacy target for prevention strategies that has been set implicitly by the recent CDC
recommendations. The ideal risk ascertainment strategy would combine
risk factors to identify a group including the greatest possible number
of infants who will develop GBS disease within the smallest number of
women who were determined to be at risk. Using two or more risk
identifiers in sequence or combination is considered in a companion
article.148
Despite the extensive literature relevant to ascertainment of
risk for EOGBS, many questions remain unanswered. In contrast to the
numerous studies of GBS disease, few studies have sought to identify
correlates of risk for early-onset neonatal sepsis caused by other
organisms, thus it is difficult or impossible to define conditions
scientifically (other than chorioamnionitis) that should dictate use of
broad-spectrum antimicrobial therapy. Although prematurity/LBW,
intrapartum fever, and PROM each increase the risk for GBS infection,
available data do not reveal threshold values for these factors, ie,
gestational age <28 weeks, birth weight <1000 g, fever 40°C, or
PROM >48 hours beyond which empirical treatment, rather than
prophylactic interventions, might be necessary. The criteria for
diagnosis of and risks associated with GBS bacteriuria need to be
defined better, and the implications of diagnosis of GBS disease in a
sibling or a twin should be quantitated in population-based studies.
Because reduced attack rates resulting from effective preventive
strategies may make it more difficult to answer such questions,
systematic collection and publication of such information should be
encouraged strongly.
| |
ACKNOWLEDGMENTS |
|---|
We thank Chad Hellig for abstracting demographic statistics from the 1992 California birth cohort data.
| |
FOOTNOTES |
|---|
Received for publication Oct 1, 1997; accepted Jan 27, 1999.
Reprint requests to (W.E.B.) Division of Neonatal and Developmental Medicine, 750 Welch Rd, Suite 315, Palo Alto, CA 94304. E-mail: benitzwe{at}leland.stanford.edu
| |
ABBREVIATIONS |
|---|
GBS, group B streptococcus;
EOGBS, early-onset
neonatal GBS infection;
ACOG, American College of Obstetricians
and Gynecologists;
AAP, American Academy of Pediatrics;
CDC, Centers
for Disease Control and Prevention;
LBW, low birth weight (2500 g);
PROM, prolonged rupture of membranes (>18 hours).
| |
REFERENCES |
|---|
|
Top
Abstract
Methods
Results
Discussion
References
|
|---|
-hemolytic streptococcal sepsis.
Am J Obstet Gynecol
1986;
154:973-977 [Medline]-haemolytic streptococcal infection in obstetrics?: discussion paper.
J R Soc Med
1984;
77:302-308 [Medline]
