Fairlie, Tarayn MD, MPH; Zell, Elizabeth R. MStat; Schrag, Stephanie DPhil
Intrapartum antibiotic prophylaxis to prevent infections caused by group B streptococci (GBS) in the first week of life (early-onset) has led to an 85% decline in the incidence of early-onset GBS disease.1 Despite this dramatic decrease, early-onset GBS disease remains a leading infectious cause of illness and death among newborns in the United States, resulting in approximately 1,050 illnesses annually (http://www.cdc.gov/abcs/reports-findings/survreports/gbs10-orig.html). Current prevention guidelines recommend that health care providers screen all women for GBS colonization late in pregnancy and administer intrapartum prophylaxis (4 or more hours of intravenous penicillin or ampicillin as the first line) to GBS-colonized women or women with unknown GBS status who deliver preterm, have membrane rupture 18 hours or more, or have intrapartum fever 100.4°F or greater.1–3
Administration of appropriate prophylaxis to women with indications lies at the heart of early-onset GBS disease prevention. Several large labor and delivery record reviews identified failure to administer prophylaxis to women with unknown GBS status who delivered preterm and failure to administer cefazolin as opposed to clindamycin to women with penicillin allergy at low risk for anaphylaxis as notable gaps in implementation4–6 Additionally, a multistate review found that among women receiving intrapartum prophylaxis, nearly one-third received durations of less than 4 hours,5 which are thought to be less effective and which trigger more newborn blood draws.1 Although not all women present in time for 4 hours of prophylaxis, a recent evaluation found that 14% of missed opportunities for prevention were the result of failure to administer adequate durations of intrapartum antibiotic prophylaxis despite sufficient time (J. Verani, personal communication, 2012).
Although the efficacy of penicillin and ampicillin intrapartum prophylaxis has been established in clinical trials,7 effectiveness in the setting of clinical practice has not been evaluated. Moreover, both efficacy and effectiveness data are lacking for key subpopulations in which prevention implementation is weak. This lack of evidence may contribute to ongoing prevention lapses.
We addressed this gap by performing a secondary analysis of a multistate cohort sampled from a population of over 600,000 liveborn neonates to evaluate the effectiveness of the following regimens in preventing early-onset GBS disease: β-lactam prophylaxis among preterm neonates; cefazolin prophylaxis; clindamycin prophylaxis; and β-lactam prophylaxis with durations of 4 hours or less before delivery.
PATIENTS AND METHODS
Active Bacterial Core surveillance, a component of the Emerging Infections Program Network, conducts active, population-based surveillance for invasive GBS disease in selected counties in 10 U.S. states. The target population for the Birthnet cohort was neonates born alive to surveillance-area residents in 2003 and 2004 who delivered at surveillance-area hospitals with 10 births per year or more; births at these hospitals accounted for nearly all resident births.
Cases of early-onset, invasive GBS disease, defined by the isolation of GBS from a normally sterile site in a liveborn neonate less than 7 days of age, were identified by routine population-based surveillance. Additionally, a random sample of live births stratified according to surveillance area, year of birth, and birth hospital was selected from birth certificates in all surveillance sites. Further details of the sampling process have been published previously.5
For each selected birth in which a chart could be located (n=7,691), trained abstractors collected standardized information from labor and delivery records on the mother's demographic characteristics, prenatal care, obstetric characteristics, intrapartum antibiotic use, including type of antibiotic used and duration, and screening for GBS.
For analysis of intrapartum prophylaxis with clindamycin, we used additional data from a similar survey of 5,144 births that occurred in Active Bacterial Core surveillance sites in 1998 and 1999; the population for this analysis thus included both neonates born alive in 2003 and 2004 as well as neonates born alive in 1998–1999 to surveillance-area residents.
Preterm delivery was defined as delivery at less than 37 weeks of gestation. Intrapartum was defined as the period between the onset of labor or rupture of membranes and delivery. For cesarean deliveries, intrapartum was defined as the period between admission for labor or delivery and cord clamping. Antibiotics given after cord clamping or less than 15 minutes before delivery were not classified as intrapartum. Antibiotics administered for prophylaxis associated with cesarean delivery were not classified as intrapartum when timing of the administration was unknown.
Screening for GBS before delivery was defined as any documented prenatal test or test at admission that was performed 2 days or more before delivery. We used two categories for race: black and non-black (which included white, Asian or Pacific Islander, Native American, other, and unknown) as a result of the small sample size of races other than black and white, and the documented association between black race and elevated incidence of early-onset GBS disease.8
Intrapartum antibiotic prophylaxis with clindamycin was defined as receiving only clindamycin before delivery. Duration questions were only evaluated for women who received at least 15 minutes of penicillin or ampicillin. Ampicillin and penicillin were classified as β-lactams.
We used propensity score matching to reduce bias with respect to important covariates for each specific type of intrapartum prophylaxis to evaluate the effectiveness of different prophylaxis regimens for prevention of early-onset GBS disease. We chose propensity score matching because key covariates differed between women receiving and not receiving prophylaxis, and standard modeling approaches provide estimates that can be wrong when an appropriate comparison population is absent.9 Propensity score matching allows us to use our observational data to design a pseudorandomized trial within our cohort in which women exposed to prophylaxis are similar to women who did not receive prophylaxis on important measured covariates (an approximation of what randomization would achieve in a prospective trial).
Covariates included maternal age, maternal race, scheduled cesarean delivery (defined as a nonemergent cesarean delivery), gestational age, rupture of membranes 18 or more hours, intrapartum fever 100.4°F or higher, GBS bacteriuria, previous neonate with GBS disease, availability of GBS screening results at the time of delivery, most recent GBS test result, presence of a prenatal GBS screen, state where delivery occurred, and births per hospital.
We built separate propensity score models for each of the following six analytic questions: effectiveness of penicillin or ampicillin given 4 or more hours before delivery for term neonates and separately for preterm neonates; effectiveness of cefazolin; effectiveness of clindamycin; effectiveness of penicillin or ampicillin given less than 2 hours before delivery; and the effectiveness of penicillin or ampicillin given 2 or more to less than 4 hours before delivery. We used multivariable logistic regression to estimate the propensity score for each mother (or neonate). The propensity score is the conditional probability of the mother receiving a specific type of intrapartum antibiotic prophylaxis given the variables in the model. We then matched each mother receiving a specific type of prophylaxis to a mother who received no prophylaxis who had the closest propensity score. For each propensity score-matched analysis, balance for each covariate included in the model was evaluated for the two groups (eg, mothers receiving intrapartum antibiotic prophylaxis and mothers not receiving intrapartum antibiotic prophylaxis). The propensity score model that achieved the best balance was used.
We used a matching procedure based on the greedy algorithm (Parsons LS. Performing a 1:N case–control match on the propensity score. Presented at SAS Users Group International 29 Proceedings, Montreal, Canada; 2004, p. 165–229). Graphic presentations were used to evaluate the bias reduction based on the propensity score matching. Graphical procedures included plotting the difference for each variable in our propensity score model looking at the difference in proportions (or means) between treated and untreated individuals before and after propensity score-based matching. The graphics use love plots,10 before and after line plots looking at the absolute difference in means or proportions, and box plots looking at the mean difference and absolute mean difference for the overall cohort and the matched cohort. Variables included in the propensity score had different levels of importance for matching. In some situations in which bias reduction for the matched subset was not optimal, exact matching was used. For example, we matched on preterm status and state of residence when evaluating the effectiveness of at least a 4-hour duration of penicillin or ampicillin before delivery.
After ensuring that balance had been achieved, we compared risk for early-onset GBS disease among women receiving a specific regimen of intrapartum antibiotic prophylaxis to those receiving no prophylaxis in the analyses. Case neonates were included in the analysis if the treated or untreated case neonate was matched with another neonate in the study based on the propensity score match and if they received the antibiotic under evaluation (or received no antibiotic). We calculated the effectiveness as 1-risk ratio, in which risk ratio is the rate of disease among women receiving intrapartum prophylaxis divided by the rate of disease among women receiving no prophylaxis. We used the Mantel-Haenszel method for 95% confidence interval (CI) estimation.
The 2003–2004 Birthnet was determined by the Centers for Disease Control and Prevention (CDC) to be a program evaluation and, therefore, institutional review was not required. The local institutional review board at each participating site also reviewed the protocol and waived the requirement for informed consent. The 1998–1999 Birthnet protocol was approved by an institutional review board of the CDC, and a waiver of informed consent was granted. Appropriate local institutional review boards also reviewed the protocol and either approved it or determined that it was exempt from the requirement for approval.
We analyzed data from 7,691 live births in 2003 and 2004, which included 7,437 in which the neonate did not have GBS disease and 254 early-onset GBS cases. For analyses of clindamycin effectiveness, we also included 113 deliveries from 1998–1999 that were exposed to clindamycin prophylaxis.
A total of 87% (6,729 of 7,691) of all neonates were delivered at term, and 84% (6,438 of 7,691) were screened for GBS prenatally. Thirteen percent (971 of 7,691) of women had one or more intrapartum risk factors for GBS transmission (Table 1). Overall, 31% (2,373 of 7,691) of mothers received intrapartum antibiotic prophylaxis. Penicillin and ampicillin were the most common agents used (75% [1,787 of 2,373]; Table 1) with most women receiving it at least 4 hours before delivery (59% [1,062 of 1,787]). Smaller proportions received penicillin or ampicillin between 2 and 4 hours before delivery (14% [254 of 1,787]) and less than 2 hours before delivery (25% [439 of 1,787]). Nine percent (331 of 3,711) of women received clindamycin, and 8% (200 of 2,373) received cefazolin.
For each discrete assessment of intrapartum antibiotic prophylaxis effectiveness, we were able to match more than 40% of women receiving a specific prophylaxis regimen to similar women who did not receive intrapartum antibiotic prophylaxis (Table 2) with the exception of cefazolin for which we were unable to create a propensity score-matched set. Within the matched subsets, median duration of prophylaxis exposure before delivery was 6.4 hours (range 2.8–13.5 hours) for women receiving clindamycin, and ranged from 1 to 11 hours for women who received penicillin or ampicillin (Table 2).
The effectiveness of 4 or more hours of intrapartum prophylaxis with penicillin or ampicillin was high among term and preterm neonates (term: 91%, 95% CI +63% to +98%; preterm: 86%, 95% CI +38% to +97%; Table 3). Effectiveness was significantly lower for clindamycin (22%, 95% CI −53% to +60%). Finally, the effectiveness of 2 to less than 4 hours of intrapartum prophylaxis with penicillin or ampicillin before delivery (38%, 95% CI −17% to +67%) and the effectiveness of intrapartum prophylaxis with penicillin or ampicillin less than 2 hours before delivery (47%, 95% CI −16% to +76%) were both markedly lower than the recommended regimen of penicillin or ampicillin 4 or more hours before delivery (Table 3).
Few direct estimates of intrapartum antibiotic prophylaxis effectiveness against early-onset GBS disease are available for the era of widespread GBS prevention. The analyses reported here confirm that 4 or more hours of β-lactam intrapartum prophylaxis is greater than 85% effective at preventing early-onset GBS disease among both term and preterm births in the setting of delivery hospitals in 10 U.S. states. Our results also support limiting use of clindamycin intrapartum prophylaxis to the narrowest possible subset of women as a result of its reduced effectiveness compared with β-lactams and more aggressive management of neonates exposed to clindamycin or to short (less than 4-hour) durations of intrapartum prophylaxis as a result of their increased risk of early-onset GBS disease.
The high effectiveness of 4 or more hours of β-lactam prophylaxis complements a recent evaluation of GBS colonization among mother–neonate pairs at four hospitals that found intrapartum antibiotic prophylaxis was 83.7% effective in preventing GBS colonization among newborns born to colonized mothers.11 Additionally, our observation that intrapartum antibiotic prophylaxis effectiveness among women delivering preterm was high and similar to that of women delivering at term is reassuring and reinforces the need for improved implementation of prophylaxis for this vulnerable group in which GBS prevention efforts have fallen short.5
The limited effectiveness of clindamycin intrapartum prophylaxis is consistent with previous pharmacologic studies, which showed poor penetration of clindamycin in fetal circulation.12–14 Given that the median duration of clindamycin prophylaxis before delivery was 6.4 hours, it is unlikely that short clindamycin durations drove the low effectiveness. However, GBS resistance to clindamycin may have contributed. Clindamycin resistance has been increasing among GBS over the past decade and multistate surveillance from 2006–2009 showed that 13–20% of invasive GBS isolates were clindamycin-resistant.1,15 Resistance during the periods of data collection for the clindamycin analysis (1998–1999 and 2003–2004) was lower than it is currently, suggesting that clindamycin prophylaxis effectiveness against currently circulating GBS may be even lower than the estimates we reported. The 2010 GBS prevention guidelines1 recommend that clindamycin intrapartum prophylaxis only be used for GBS-colonized women at high risk for penicillin anaphylaxis whose isolates are susceptible to clindamycin. For this group of women, a more effective agent choice is not available. However, a multistate labor and delivery record review found that more than 70% of penicillin-allergic women at low risk for anaphylaxis received clindamycin prophylaxis, and less than 1% had documentation of antimicrobial susceptibility testing.5
The low clindamycin effectiveness we observed underscores the importance of administering cefazolin intrapartum prophylaxis to penicillin-allergic women without a history of penicillin or cephalosporin-associated anaphylaxis, angioedema, respiratory distress, or urticaria. Cefazolin has a pharmacologic profile similar to that of penicillin and ampicillin and achieves high intraamniotic concentrations. Additionally, cephalosporin crossreactivity affects approximately 10% of patients with penicillin allergy,16 and maternal anaphylaxis associated with GBS intrapartum antibiotic prophylaxis is rare; four case reports of nonfatal anaphylaxis have been published since 1996.17–20 Vancomycin remains an agent of last resort, reserved for women at high risk for penicillin anaphylaxis colonized with a GBS isolate that is not susceptible to clindamycin and erythromycin or of unknown susceptibility.1
In contrast to a vertical transmission evaluation of GBS colonization and a systematic review that suggested 2 or more hours of intrapartum prophylaxis might confer acceptable protection,4,21 we provided the first direct evidence using a GBS disease end point that shorter prophylaxis durations, even those greater than 2 hours, are markedly less effective than 4 or more hours of prophylaxis. This supports the 2010 newborn management recommendations to perform a more aggressive newborn evaluation (blood draw for blood culture and complete blood count) on the highest risk subset of this group (preterm deliveries or those with prolonged membrane rupture) even if the newborns are well-appearing.1 Although a portion of patients will present without 4 hours between admission and delivery, our finding of reduced effectiveness of shorter intrapartum antibiotic prophylaxis durations underscores the importance of initiating prophylaxis as soon as possible on admission for labor and also emphasizes the value of communicating to prenatal patients the importance of presenting for labor with adequate time.
The observational design of our study is an important limitation. We attempted to overcome this by using propensity score methods, allowing us to approximate the balance of a randomized trial. We were also limited by the smaller sample of women (n=331) who received clindamycin prophylaxis; however, this subset of women was derived from a review of almost 13,000 labor and delivery records, and we are not aware of a data source that would provide a more robust sample. We maximized the available data on clindamycin prophylaxis by using data from the 1998–1999 Birthnet cohort, whose sample design was the same as that of the 2003–2004 Birthnet cohort; we performed propensity score matching within each cohort to make appropriate comparisons. The age of the data sets was another limitation, because this was a secondary analysis of data collected 9–13 years ago. However, during this entire time period, the indications for intrapartum prophylaxis did not change meaningfully, and from 2002 forward, universal antenatal screening was in place. The primary bias of data set age with regard to intrapartum antibiotic prophylaxis effectiveness is the changing susceptibility profile of GBS over time. Group B streptococcus remains universally susceptible to penicillin, ampicillin, and cefazolin, and thus our β-lactam estimates were not affected. However, our estimate of clindamycin effectiveness may be conservative, because resistance of GBS to clindamycin has increased from 12% in 1998 to 15% in 2003 to greater than 20% in recent years.15,22,23 Finally, we were unable to evaluate the effectiveness of cefazolin intrapartum prophylaxis, the preferred agent for women with penicillin allergy at low risk of anaphylaxis, as a result of the rare use of this agent in current practice and differences in baseline characteristics between women who did and did not receive the agent, precluding the creation of a propensity score-matched subset. If implementation of cefazolin intrapartum antibiotic prophylaxis increases through improved adherence to the GBS prevention guidelines, data will accumulate allowing for direct cefazolin effectiveness estimates.
Our findings reinforce the recommendations in the 2010 CDC GBS guidelines and underscore the importance of adherence to the guidance for women delivering preterm, those with allergy to penicillin, and neonates exposed to short intrapartum prophylaxis durations. Recent analysis of ongoing early-onset GBS cases suggests that improved prevention implementation in key areas could result in a further 30–50% reduction in cases to an incidence as low as 0.15 cases per 1,000 live births.24 While accumulating evidence such as we presented may help to achieve this goal, a new health care provider application, Prevent Group B Strep, created by the CDC in conjunction with the same key clinical partner organizations that endorsed the 2010 GBS prevention guidelines, may lead to further implementation improvements. The application will be available online and as a native application for smartphones and tablets and will provide patient-specific recommendations. We hope that accumulating evidence about effectiveness along with the Prevent Group B Strep application will reduce missed prevention opportunities and bring us closer to the full affect possible under universal screening.
1. Centers for Disease Control and Prevention. Prevention of perinatal group B streptococcal disease—revised guidelines from CDC, 2010. MMWR Recomm Rep 2010;59:1–36.
2. Committee on Infectious Diseases; Committee on Fetus and Newborn. Baker CJ, Byinton CL, Polin RA. Policy statement—Recommendations for prevention of perinatal group B streptococcal disease. Committee on Infectious Diseases and Committee on the Fetus and Newborn. Pediatrics 2011;128:611–6.
3. Prevention of group B streptococcal disease in newborns. Committee Opinion No. 485. American College of Obstetricians and Gynecologists. Obstet Gynecol 2011;117:1019–27.
4. Stoll BJ, Hansen NI, Sanchez PJ, Faix RG, Poindexter BB, Van Meurs KP, et al.. Early onset neonatal sepsis: the burden of group B streptococcal and Escherichia coli disease continues. Pediatrics 2011;127:817–26.
5. Van Dyke MK, Phares CR, Lynfield R, Thomas AR, Arnold KE, Craig AS, et al.. Evaluation of universal antenatal screening for group B streptococcus. N Engl J Med 2009;360:2626–36.
6. Matteson KA, Lievense SP, Catanzaro B, Phipps MG. Intrapartum group B streptococcal prophylaxis in patients reporting a penicillin allergy. Obstet Gynecol 2008;111:356–64.
7. Boyer KM, Gotoff SP. Prevention of early-onset neonatal group B streptococcal disease with selective intrapartum chemoprophylaxis. N Engl J Med 1986;314:1665–9.
8. Centers for Disease Control and Prevention (CDC). Perinatal group B streptococcal disease after universal screening recommendations—United States, 2003–2005. MMWR Morb Mortal Wkly Rep 2007;56:701–5.
9. Pattanayak CW, Rubin DB, Zell ER. Propensity score methods for creating covariate balance in observational studies [in Spanish]. Rev Esp Cardiol 2011;64:897–903.
10. Ahmed A, Young JB, Love TE, Levesque R, Pitt B. A propensity-matched study of the effects of chronic diuretic therapy on mortality and hospitalization in older adults with heart failure Int J Cardiol 2008;125:246–53.
11. Lin FY, Weisman LE, Azimi P, Young AE, Chang K, Cielo M, et al.. Assessment of intrapartum antibiotic prophylaxis for the prevention of early-onset group B streptococcal disease. Pediatr Infect Dis J 2011;30:759–63.
12. Philipson A, Sabath LD, Charles D. Erythromycin and clindamycin absorption and elimination in pregnant women. Clin Pharmacol Ther 1976;19:68–77.
13. Philipson A, Sabath LD, Charles D. Transplacental passage of erythromycin and clindamycin. N Engl J Med 1973;288:1219–21.
14. Muller AE, Mouton JW, Oostvogel PM, Dorr PJ, Voskuyl RA, DeJongh J, et al.. Pharmacokinetics of clindamycin in pregnant women in the peripartum period. Antimicrob Agents Chemother 2010;54:2175–81.
15. Castor ML, Whitney CG, Como-Sabetti K, Facklam RR, Ferrieri P, Bartkus JM, et al.. Antibiotic resistance patterns in invasive group B streptococcal isolates. Infect Dis Obstet Gynecol 2008;2008:727505.
16. Kelkar PS, Li JT. Cephalosporin allergy. N Engl J Med 2001;345:804–9.
17. Chaudhuri K, Gonzales J, Jesurun CA, Ambat MT, Mandal-Chaudhuri S. Anaphylactic shock in pregnancy: a case study and review of the literature. Int J Obstet Anesth 2008;17:350–7.
18. Dunn AB, Blomquist J, Khouzami V. Anaphylaxis in labor secondary to prophylaxis against group B streptococcus. A case report. J Reprod Med 1999;44:381–4.
19. Sheikh J. Intrapartum anaphylaxis to penicillin in a woman with rheumatoid arthritis who had no prior penicillin allergy. Ann Allergy Asthma Immunol 2007;99:287–9.
20. Gei AF, Pacheco LD, Vanhook JW, Hankins GD. The use of a continuous infusion of epinephrine for anaphylactic shock during labor. Obstet Gynecol 2003;102:1332–5.
21. Illuzzi JL, Bracken MB. Duration of intrapartum prophylaxis for neonatal group B streptococcal disease: a systematic review. Obstet Gynecol 2006;108:1254–65.
22. DiPersio LP, DiPersio JR. High rates of erythromycin and clindamycin resistance among OBGYN isolates of group B streptococcus. Diagn Microbiol Infect Dis 2006;54:79–82.
23. Phares C, Lynfield R, Farley M, Mohle-Boetani J, Harrison L, Petit S, et al.. Epidemiology of invasive group B streptococcal disease in the United States, 1999–2005. JAMA 2008;299:2056–65.
24. Schrag SJ, Verani JR. Intrapartum antibiotic prophylaxis for the prevention of perinatal group B streptococcal disease: Experience in the United States and implications for a potential group B streptococcal vaccine. Vaccine. 2012 Dec 3 [Epub ahead of print].
Figure. No available...Image Tools