Group B Streptococcus (GBS) is a major contributor to neonatal sepsis worldwide. Late-onset group B Streptococcus disease (LOGBS) and its risk factors remain poorly understood. The isolation of GBS from breast milk has been described in cases of LOGBS. This potential association has raised concerns for mothers and clinicians regarding the safety of ongoing breastfeeding. In this study, we aimed to investigate whether exposure to breast milk is associated with increased risk of LOGBS.
A case-control study of LOGBS was conducted across 4 hospital networks in Victoria, Australia, including the 2 major tertiary pediatric centers in the state, to evaluate 11 years of data (2007–2017). Cases were captured initially from microbiology databases and recaptured with International Classification of Diseases discharge coding. Each case patient was matched with 4 controls to assess feeding status. Patients were matched for chronological age, gestation, discharge status, recruitment site, and calendar year.
We identified 92 cases of LOGBS: 73 cases on initial capture and 76 cases on the recapture analysis. Case patients were matched with 368 controls: 4 controls to each patient. Seventy-two patients were exposed to breast milk at the time of LOGBS (78.3%), compared with 274 controls (74.5%; odds ratio 1.2 [95% confidence interval 0.7–2.3]).
Breastfeeding was not associated with increased risk of LOGBS. Breast milk should not be tested for GBS during a first episode of LOGBS.
Case reports have described children with late-onset group B Streptococcus disease with group B Streptococcus cultured from breast milk. Although controversial, this has led to concern and to breast milk being implicated as a potential source for disease transmission.
Breastfeeding was not associated with an increased likelihood of developing late-onset group B Streptococcus disease in our setting.
Group B Streptococcus (GBS), Streptococcus agalactiae, remains a major contributor to neonatal sepsis throughout the world. The mean incidence globally is described as 0.53 per 1000 live births.1,2 Neonatal GBS infection is classified into 2 main clinical syndromes: early-onset disease, which occurs from birth to the first 6 days of life, and late-onset disease, occurring from day 7 to day 89 of life.2 The incidence of late-onset disease is ∼0.1 to 0.4 per 1000 live births.2–6 A third classification group, very late-onset disease, exists likely as a continuation of late-onset disease occurring from day 90 of life onward.7,8 An increased prevalence of prematurity is usually observed within the late-onset group. It is important to consider immunodeficiency when very late-onset disease occurs in a term infant in the absence of any other apparent risk factor.9 Although well described in the literature, relapse or recurrence of early- or late-onset disease is uncommon, with an estimated incidence of 1%.10
The pathogenesis of early-onset disease is well understood and results from perinatal vertical transmission to the infant from a mother who is colonized from either the anorectal or vaginal tracts. After the introduction of intrapartum antibiotic prophylaxis for women who are colonized with GBS or who have clinical risk factors, the incidence of early-onset disease has considerably reduced.2,11 The incidence of late-onset group B Streptococcus disease (LOGBS), however, remains unaffected by intrapartum antibiotic prophylaxis.12
The mode of transmission for LOGBS, in contrast to early-onset disease, remains less well understood. In addition to vertical transmission of GBS, horizontal transmission is thought to play an important role. Potential sources include the postpartum mother, other household contacts, and even health care professionals within the hospital or community environment.13–17 Although nosocomial outbreaks have been described, the most important known risk factor for LOGBS is maternal colonization.2,18 Serotypes of GBS strains isolated from infants with LOGBS correlate with matched maternal genital isolates. At birth, ∼50% of infants born to a mother who is colonized with GBS are found to be colonized with the maternal serotype.13,19–22 Although the prevalence varies between regions, the overall global prevalence of maternal colonization is ∼20%.2,16,23 Heavy maternal colonization is associated with an increased prevalence of heavy neonatal colonization, which is itself a known risk factor for LOGBS.2,21,22,24 Prematurity and twin pregnancies have also been associated with an increased risk for LOGBS.18,25–29 An asymptomatic twin of an infant who is infected is estimated to be at a 25-fold increased risk for LOGBS.27
Case reports of LOGBS, especially recurrent disease, associated with the isolation of GBS in the breast milk have been described. This has led to the implication that “infected” breast milk may serve as a source of acquisition before invasive disease.25,30–38 This association is controversial, especially given the substantial benefits of breastfeeding.39–52 The implication that infected breast milk may be responsible for LOGBS can lead to maternal distress and has resulted in premature cessation of breastfeeding.26,29,31,34,53–55 There are also reports of preterm infants fed with culture-positive breast milk expressed from a mother with clinical GBS mastitis who do not develop LOGBS.56,57
Breastfeeding is the gold standard of feeding for all infants, and the American Academy of Pediatrics strongly advocates breastfeeding for both term and preterm infants, whenever possible, for at least the first 6 months of an infants’ life.44,58,59 The microbiota of breast milk plays an important role in the development of the neonatal gut microbiome and in providing protection against infection for the infant46,60–62 in addition to the protection afforded by secretory immunoglobulin A obtained through the breast milk.43 The microbiologic testing of expressed breast milk for GBS is not without its risks. One review of 48 mothers of infants with LOGBS or recurrent GBS disease described 20 (42%) who stopped breastfeeding in response. Of these, 11 mothers (23% of total, 31% [11 of 35] of known outcomes) permanently ceased breastfeeding after their infant had a single episode of LOGBS.31
There is a need to further describe the relationship between breastfeeding and the development of LOGBS. This study was designed to review the relationship and association between feeding with breast milk and GBS disease development.
Study Design and Setting
This was a multicenter case-control study conducted across 4 health networks in Victoria, Australia. Study sites included both state tertiary pediatric hospitals, the Monash Children’s Hospital and the Royal Children’s Hospital, both in Melbourne. The NICU at the Monash Children’s Hospital is the largest in Victoria and is located within a tertiary health service with a large obstetric service. The Royal Children’s Hospital is a quaternary center and the largest tertiary pediatric hospital in Australia. Its NICU cares for outborn infants only. Two larger regional Victorian health services, Barwon Health and Bendigo Health, were also included. Data were collected (retrospectively) for the 11-year period from January 1, 2007, to December 31, 2017, across all sites.
A case of invasive GBS was defined as isolation of GBS from a normally sterile site (ie, blood, cerebrospinal fluid, synovial fluid, or pleural fluid). LOGBS was defined as presentation of invasive GBS disease after day 7 of life (we combined late onset, traditionally defined as disease between 7 and 89 days, with very late onset, defined as disease from 90 days onward).2 Recurrent disease was defined as a new invasive GBS episode occurring at any interval after the completion of apparently successful therapy for the initial episode.2,63
For both patients and controls, feeding status was assessed as being fed exclusively with formula, being fed exclusively with breast milk, and mixed feeding. For analysis, however, the latter 2 were considered together.
Cases were identified by using capture-recapture analysis.64 Initial identification was done by using data extracted from each health service’s microbiology database of all positive results for blood, cerebrospinal fluid, and other fluid cultures in children aged <6 months. Recapture methodology was applied by using International Classification of Diseases, 10th Revision discharge coding to identify additional cases in which the initial microbiology was performed at a non–study site.65 The following specific codes were searched at all sites: A40.1, “sepsis due to Streptococcus, group B”; B95.1, “Streptococcus, group B, as the cause of diseases classified elsewhere”; and P36.0, “bacterial sepsis of newborn due to Streptococcus, group B.”66 In addition, the following nonspecific code was searched: P36.1, “sepsis of newborn due to other and unspecified streptococci.” After case identification, corresponding clinical notes were reviewed for confirmation of the case definition; data were subsequently collected by completion of a data collection form into Research Electronic Data Capture.67 Only patients with GBS isolated from a normally sterile site were included.
Given that the anticipated rate of breast milk exposure in patients and controls was >50%, 4 controls were selected as the maximum likely to improve detection rate of a true difference in exposure rates between the groups.68 Controls were matched on the basis of 5 participant characteristics: chronological age, gestational age, whether already discharged or still hospitalized since birth, recruitment site, and calendar year of presentation. Limited information was obtained from controls: demographics, maternal GBS screening status, and feeding status (assessed at time of admission to hospital for controls presenting from home or at the same chronological age of the patient for controls who were still inpatients from birth). Patients aged <2 months were matched to controls within 7 days of chronological age; patients aged 2 months or older were matched within 14 days. Controls were matched according to the gestational age of patients within commonly recognized prematurity definitions: extremely preterm, <28 weeks; very preterm, 28 to <32 weeks; moderate to late preterm, 32 to <37 weeks; and term, 37 to <42 weeks.69 Controls were matched for chronological age according to the patient’s age at time of disease onset. Controls were matched from the same recruitment site as the corresponding patient. If a patient was managed at >1 recruitment site, when possible, controls were matched to the initial site of presentation. Controls were preferentially matched within the same calendar year as patients.
Gestational age, birth discharge status, and recruitment sites were matched strictly, although flexibility was given for calendar year and chronological age if a control was unable to be identified fulfilling all matching criteria. For patients with recurrent disease, controls were matched only once and for the later episode.
Controls were identified by using International Classification of Diseases (ICD) code searching. The conditions were selected on the basis that feeding status did not influence the probability of disease development. For term and moderate to late preterm infants presenting from home with LOGBS, controls were matched by using the following ICD codes: N39.0, “urinary tract infection, site not specified”; J21.0, “acute bronchiolitis due to respiratory syncytial virus”; J21.1, “acute bronchiolitis due to human metapneumovirus”; and J21.8, “acute bronchiolitis due to other specified organisms.” For extremely and very preterm infants presenting from home and moderate to late preterm infants in whom matching left an incomplete match, ICD codes were used to identify all those cared for at that health network that year. The following ICD codes were used to identify extremely preterm infants, very preterm infants, and moderate to late preterm infants, respectively: P07.22, “extreme immaturity of newborn, gestational age 23 completed weeks”; P07.31, “preterm newborn, gestational age 28 completed weeks”; and P07.32, “preterm newborn, gestational age 29 completed weeks.” The clinical record for each preterm infant of matched gestation was reviewed for readmissions occurring within the chronological-age matching window.
For matching of preterm patients in whom the LOGBS episode occurred within their birth admission, controls were identified by using the previously described prematurity ICD codes. Age at discharge was calculated, and for controls, the feeding status at the matched chronological age during admission was reviewed.
Data were exported from Research Electronic Data Capture into Stata (version 14.0; Stata Corp, College Station, TX). Categorical data were presented as proportions, and continuous data were presented as mean or median. We used Fischer’s exact test for comparison of proportions, and for continuous data, we used t tests and the Wilcoxon rank tests for normally and nonnormally distributed data, as appropriate. Odds ratios and 95% confidence intervals (CIs) were calculated by using stratified data with a Mantel-Haenszel test.
Ninety-two LOGBS cases were identified across the 4 health networks during the 11-year study period: 73 cases on the initial capture analysis and 76 cases on the recapture analysis with specific ICD code searching (Fig 1). No additional cases were identified by using the non–GBS-specific P36.1 ICD code. There were 368 matched controls; all patients were matched to 4 controls. Among the 368 controls, 343 (93.2%) were matched completely and 6 (1.6%) were matched within an additional 7 to 14 days. All were matched for gestational age and birth discharge status. For one patient, a very preterm infant, the primary presentation was to one of the regional sites after discharge from the hospital. It was not possible to match for controls at the regional site; however, because the child required transfer to one of the tertiary centers, controls were matched at the tertiary site. Fourteen controls (3.8%) for 6 patients could not be matched for calendar year. Other criteria were preferentially matched, and these controls were identified from the preceding or following 2 calendar years. One control (0.3%) was not matched for chronological age (out by 2 days) or calendar year (preceding year to patient). The baseline demographics for sex, gestational age, and age at time of feeding assessment were all comparable.
Even after removal of missing data, patients (28 of 48) were significantly more likely to be born to a mother who screened positive for GBS than controls (37 of 107; P = .003; Table 1).
Seventy-two infants (78.3%) with LOGBS were fed with breast milk at the time of disease presentation, compared with 274 controls (74.5%; odds ratio 1.2 [95% CI 0.7–2.3]; P = .45; Fig 2).
There were 9 patients (9.8%) with recurrent disease: 5 presented from home for both episodes, whereas in the remaining 4 patients, both episodes happened during their birth admission. The median gestational age at birth for these infants was 29 weeks (range 24–40 weeks), and the median age at onset of the initial LOGBS episode was 48 days (range 4–81 days), with infants treated for the initial episode with antibiotics for a mean duration of 14 days (range 14–22 days). The median age at onset of recurrent disease was 72 days (range 27–104 days).
Eight of these infants were fed with breast milk at the time of disease onset, compared with 64 patients with nonrecurrent disease (P = .43). There was 1 patient with recurrent disease who had not received breast milk at any stage.
With this case-control study, we are the first to explicitly explore the relationship between breastfeeding and the development of LOGBS. The potential etiological role of infected breast milk for LOGBS25,26,31–36 ,55,70 has remained heavily controversial and unproven. The majority of these reports occurred in the setting of clinical mastitis in the mother,14,25,26,35,36,57,70–73 although cases of disease thought to be associated with breast milk have also been described in the absence of maternal illness.29,33,34,74 In our study, LOGBS was not associated with increased risk of breast milk exposure, suggesting that breastfeeding is not a risk factor for LOGBS.
The baseline demographics of the patient and control population groups are comparable. No increased risk of being exposed to breast milk was observed among patients with LOGBS compared with matched controls. Because of small numbers, the CIs were wide for extremely preterm infants and infants with recurrent disease; however, we observed no evidence of increased risk in any subpopulation group.
Infants in this study who developed LOGBS disease were significantly more likely to have been born to a mother who was GBS-positive compared with their matched controls (P = .002). Although the risk factors for LOGBS remain poorly understood in contrast to early-onset disease, maternal colonization remains the strongest known risk factor for disease.2
It is important to place LOGBS within the broader context of neonatal late-onset sepsis resulting from any pathogen. LOGBS makes up a small proportion of the infections observed within the larger context of neonatal late-onset sepsis. LOGBS accounts for only 3.2% of confirmed late-onset sepsis in neonates.75
Breastfeeding is one of the most effective ways to protect against infection in early life, especially among preterm and very low birth weight (<1500 g) infants.47,51,62,76 In one study of 283 preterm infants, 29.3% of breastfed infants, compared with 47.2% of formula-fed infants, developed infection (P = .01); whereas 19.5%, compared with 32.6% (P = .04), respectively, developed sepsis or meningitis.44 Only one infant in this study developed LOGBS, and this infant was exclusively formula fed. The benefits of breastfeeding extend beyond protection against infection, and breastfeeding is associated with reduced rates of allergic disease and even malignancy in later years as well as improved neurodevelopmental outcomes.42,49,77,78 Formula feeding was associated with a 1.3-fold increased risk of infant mortality and a 1.6- to 2.1-fold increased risk of sudden infant death syndrome when compared with breastfeeding.42
Prospective studies reviewing the incidence of GBS in breast milk of both term and preterm infants have described a generally low incidence of between 0.4% and 0.8%,6,76 with the highest incidence being 3.5%.79 Numerous studies describe infants fed with GBS-infected milk samples remaining healthy and free of disease.22,56,57,79,80
Limitations of our study should be noted. First, all information was retrospectively obtained from medical records, and breastfeeding status was reliant on documentation within those records. Second, the sample size precluded any conclusions being drawn regarding specific subpopulations, including preterm infants or those infants with recurrent LOGBS. Third, the ICD codes used to identify controls were all for conditions that are infectious in nature. Given it is known that formula-fed infants are more likely to develop infection over the first year of life, this may have led to overrepresentation of formula feeding among the control group compared to the general population.42 Fourth, because of the small sample size, the results do not allow us to derive clear information for the subpopulation of the very premature infants, who are at higher risk of LOGBS. Further research is still needed in this group. Lastly, although this study was multicentered and spanned an 11-year period, the total study size is still small, <100 patients. It is possible that a larger sample size may have revealed a different result. We agree that the results of this study should be confirmed by further research.
We did not observe an association between LOGBS sepsis and receipt of breast milk feeds. Up to 3.5% of healthy women have GBS detectable in their breast milk, and the vast majority of exposed infants never develops LOGBS. The identification of GBS in breast milk may be associated with a high rate of breastfeeding cessation. The benefits of breastfeeding are broad and clearly described. Recurrence of LOGBS is rare, occurring only 1% of the time. We suggest that breast milk should not be tested after a primary episode of LOGBS and that continued breastfeeding should be supported.
We thank the following pathology services and health information services that assisted in providing the coding and microbiology extraction data: Monash Health, The Royal Children’s Hospital, Barwon Health, Bendigo Health, and Australian Clinical Laboratories. We also thank Jiying Yin, PhD, Senior Research Officer and Data Analyst (Murdoch Children’s Research Institute), for his assistance with the statistical analysis.
Dr Ching conceptualized and designed the study, designed the data collection instruments, collected data, coordinated, and supervised data collection for all sites, conducted initial analyses, drafted the initial manuscript, and reviewed and revised the manuscript; Prof Doherty and Prof Buttery conceptualized and designed the study and critically reviewed the manuscript for important intellectual content; Dr Lai collected data at Monash Children’s Hospital and critically reviewed the manuscript for important intellectual content; Prof Steer coordinated and supervised data collection at The Royal Children’s Hospital and critically reviewed the manuscript for important intellectual content; Dr Standish coordinated and supervised data collection at Barwon Health and critically reviewed the manuscript for important intellectual content; Dr Ziffer coordinated and supervised data collection at Bendigo Health and critically reviewed the manuscript for important intellectual content; Prof Daley collected data at The Royal Children’s Hospital and critically reviewed the manuscript for important intellectual content; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.
FUNDING: Dr Ching’s position was supported by funding from the Department of Paediatrics, Monash University. No direct funding was secured for this study. The funder/sponsor did not participate in the work.
POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.
FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.