Article Text

Download PDFPDF

Group B streptococcal disease in infants: a case control study
  1. P T Heath1,
  2. G F Balfour1,
  3. H Tighe1,
  4. N Q Verlander2,
  5. T L Lamagni3,
  6. A Efstratiou3
  1. 1
    Child Health and Vaccine Institute, St George’s, University of London, London, UK
  2. 2
    Statistics Unit, Health Protection Agency, Colindale, London, UK
  3. 3
    Centre for Infection, Health Protection Agency, Colindale, London, UK
  1. Correspondence to Dr P T Heath, Division of Child Health, St George’s, University of London, Cranmer Terrace, London SW17 0RE, UK; pheath{at}


Objectives: To describe and quantify the maternal and neonatal factors associated with Group B streptococcus (GBS) disease in infants <90 days of age.

Setting: Neonatal Units in London, Oxford, Portsmouth and Bristol.

Patients: Cases were infants <90 days of age with invasive GBS disease diagnosed between 2000 and 2003, and controls were healthy infants born in the same hospital and in the same birth weight category.

Main outcome measures: Demographic and clinical data on the mother, baby, birth and neonatal stay.

Results: 138 cases and 305 controls were recruited. The majority of cases (74%) presented in the first week of life (early onset, EO); most on day 1 (89%). 65% of EO cases had one or more clinical risk factors (prematurity, prolonged rupture of membranes (PROM), known maternal GBS carriage, fever during labour). A multivariable logistic regression analysis found that the strongest independent associations with GBS disease were known maternal carriage of GBS (odds ratio (OR) 6.9), maternal infection in the peripartum period (OR 4.2) and maximum temperature in labour (OR 2.2 per °C). GBS disease was associated with twice the likelihood of PROM and fetal tachycardia (p = 0.05 and 0.07 respectively). EO cases had lower Apgar scores and were more likely to have respiratory distress and convulsions, and to require tube feeding than controls. They spent longer in hospital than controls, requiring longer stays at all levels of care.

Conclusions: Independent of birth weight, a number of maternal, birth and neonatal factors are significantly associated with GBS disease. The management of babies with GBS disease results in an appreciable use of hospital resources.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Streptococcus agalactiae (Group B streptococcus (GBS)) is an important cause of serious infection in neonates, infants and immunocompromised or pregnant adults. The published risk factors for developing early-onset (EO) disease are based on a large number of observational and case control studies performed predominantly in the USA. Major risk factors include maternal GBS carriage, prematurity, low birth weight, prolonged rupture of membranes, intrapartum fever, young maternal age, black ethnic group, previous delivery of an affected baby and low levels of anticapsular antibody.1 2 3 The risk factors for late-onset (LO) GBS disease are less clear, although maternal, nosocomial and breast milk sources have all been described.2 4 5 6 7 8

In 1998, a working group was established by the Public Health Laboratory Service (now the Health Protection Agency) with a remit to assess the burden of group B streptococcal disease and to produce recommendations for its prevention. A national surveillance study of invasive group B streptococcal disease in infants younger than 90 days was therefore performed between 2000 and 2001. This identified 568 cases (a total incidence of 0.72 per 1000 live births) with an incidence of 0.48 per 1000 for EO disease (<7 days of age) and 0.24 per 1000 for LO disease (7–90 days of age). Fifty-three infants died (9.7%).9 Clinically identifiable antenatal risk factors for early-onset disease were present in 218 (58%) of affected pregnancies. In a separate study, we sought to describe the factors associated with disease in more detail and to quantify their importance by comparison with a control group.

Patients and methods

Cases were defined as infants <90 days of age who were hospitalised in the Greater London, Oxford, Portsmouth and Bristol areas, and from whom GBS was isolated from a normally sterile site, for example blood, cerebrospinal fluid or joint fluid, in the period between 1 February 2000 and 28 February 2003. Cases were identified through routine laboratory reporting and isolate referral of cases to the Health Protection Agency Communicable Disease Surveillance Centre or Respiratory and Systemic Infection Laboratory respectively. The case information was then passed on to the study team. Direct referral was also sought from microbiologists and paediatricians caring for cases. Control infants were identified by local hospital staff. They were asked to review the hospital birth records and identify at least two infants, one born as soon as possible before the case and the other as soon as possible after the case, where both infants fell in the same birth weight category as the case (categories: ⩽749 g, 750–999 g, 1000–1499 g, 1500–1999 g, 2000–2499 g, 2500–2999 g, 3000–3499 g, ⩾3500 g). The control infants were subsequently eligible for inclusion if there was no clinical evidence of sepsis during the first 7 days of life. Matching of cases and controls by birth weight was proposed in recognition of the importance of birth weight/gestation on the clinical assessment of neonates, need for neonatal support and short-term sequelae.

Written permission was sought from parents of cases and controls to be included in the study. When permission was gained, the research nurses obtained the hospital notes of the infant, and a standard proforma was completed. One nurse extracted data from the notes of the infants (cases and controls). The other nurse extracted data from the mothers’ notes, blinded to whether the notes were of the mother of the case or the mother of the control. The status of “GBS carrier” was given if a positive GBS culture from a vaginal swab was recorded in the maternal records.

Data were transferred to a custom-made MS ACCESS database and analysed using SPSS version 10.05 (SPSS, Chicago). Single- and multivariable analyses were performed using STATA 8.2.

Cases were classified as EO (0–6 days of age) or LO (7–90 days of age). Proportions were compared using the χ2 or Fisher exact test. Continuous data were compared using the Mann–Whitney test for independent samples.

All episodes of GBS infection were included in a multivariable analysis of maternal and labour factors with the exception of second episodes. As there were a number of multiple births, this meant that not all observations were independent. This was allowed for by performing an adjustment to the standard errors of the odds ratio estimates. Initial analysis with unconditional logistic regression (ignoring matching) with and without taking into account clustering was performed. Subsequently, conditional logistic regression ignoring clustering (non-independence) was performed.

Those variables with p values of 0.2 or less from a single variable analysis were included in an initial multivariable model with the exception of ethnic origin, which had a relatively large number of missing observations. A series of models were then estimated, at each step removing the least significant variable, but retaining substantial confounders and ending with a final model when all p values were equal to 0.1 or less. Age was kept in the model, regardless of its p value. The ethnic origin variable was added to the final model and tested for significance. This final model was also estimated by unconditional logistic regression, with and without clustering, to see whether clustering affected the results. Separate analyses were performed to see whether the risk factors were different or of different magnitude between EO and LO GBS disease. The analysis was similar to that described above. Single-variable analysis involved interaction of each variable with onset category (early/late) separately. The initial multivariable model contained variables with p values of 0.2 or less from the all cases single variable analysis or with p values of 0.1 or less from the early-/late-onset single variable analysis, with each of these variables in an interaction with onset category. Interactions were then removed in a backwards stepwise procedure, ending with testing the interaction between ethnic origin and onset category.

Approval for the study was given by the multicentre research ethics committee of Scotland (MREC/99/0/86).


During the 37-month period of the study, 242 cases of invasive GBS infection were reported from the participating hospitals, and the parents of 138 cases consented to participate (57%). The parents of 1220 matched controls were then approached, and consent was obtained from 305 (25%). Cases and controls were well matched with respect to birth weight (cases median 3246 g, range 500–4840; controls 3200 g, 677–4680 g). There were two sets of affected twins among the cases and six sets of twins among the controls. Three cases (2.2%) had a second episode of GBS during the surveillance period.

The majority of infants with GBS (102, 74%) presented in the first week of life (EO), most on day 1 of life (91 of 102, 89%) (table 1). Of those developing clinical features on day 1, the majority (81 of 83, 97.6%) were noted by 12 h of age where the hour of onset was given. LO cases were twice as likely to be preterm (<37 weeks, p = 0.006) and two and half times as likely to be of low birth weight (<2500 g, p = 0.001) than EO cases.

Table 1

Group B streptococcus cases

Among EO cases, 25 of 101 (24.8%) were born prematurely (<37 weeks), the mothers of 14 of 99 (14%) were known to have GBS bacteruria or to be GBS carriers before delivery of the infant, 15 of 100 (15%) had a fever >38°C during labour, and 49 of 101 (48.5%) had prolonged rupture of membranes (⩾18 h). In total, 66 of 101 cases (65%) had one or more of these risk factors. Intrapartum antibiotics were known to have been administered to 11 of 65 (17%) mothers with risk factors.

The maternal and labour factors associated with cases and controls are shown in table 2. Young maternal age, black ethnic group, multiple pregnancy and a previous baby with GBS were not shown to be significantly over-represented among mothers of GBS cases. Cases were more likely to be born to primigravida mothers, mothers known to carry GBS or to have had GBS isolated from urine, mothers with intrapartum fever (with a risk that correlated with the height of maximum temperature) and mothers with prolonged rupture of membranes. During labour, mothers of cases were more likely to have had an epidural placed, more likely to have vaginal examinations performed (with a risk that correlated with the number of examinations), more likely to have evidence of fetal tachycardia and fetal distress and more likely to need fetal blood sampling. Mothers of cases were less likely to have a normal vaginal delivery and more likely to require an emergency delivery as compared with control mothers. After delivery, mothers of cases were more likely to have an infection documented and to be prescribed antibiotics.

Table 2

Maternal and labour factors using conditional univariate analysis

A multivariable logistic regression analysis (table 3) found that the strongest independent associations with being a case were: known carriage of GBS (odds ratio (OR) 6.9), having a documented maternal infection in the peripartum period (OR 4.2) and maximum temperature in labour (OR 2.2 per °C). Prolonged rupture of membranes and fetal tachycardia were associated with twice the likelihood of GBS disease although just above the level of statistical significance.

Table 3

Multivariable conditional logistic regression analysis using duration of ruptured membranes of 18 h*

An analysis was performed to see whether the risk factors were different or of different magnitude according to the age at onset of disease (table 4). The only factor that discriminated between EO and LO cases was number of vaginal examinations, with a higher number of examinations being associated with increased likelihood of EO GBS and a lower likelihood of LO GBS.

Table 4

Multivariable conditional logistic regression analysis: discrimination between early and late onset*

Table 5 presents clinical details of the EO cases and their respective controls. Cases had lower Apgar scores at 1, 5 and 10 min of age than their controls. Cases were more likely to have respiratory distress and to require oxygen supplementation, continuous positive airways pressure, assisted ventilation and surfactant therapy. Cases were more likely to have convulsions and to require tube feeding during the neonatal period. Infants with GBS infection spent longer in hospital than controls, requiring longer stays at all levels of care (intensive care, high dependency care and special care).

Table 5

Clinical details of cases and controls (early-onset cases only)

With regard to the clinical management of the infants with GBS infection, 109 of 138 (79%) had a lumbar puncture performed, and the most common antibiotic combination used to treat GBS infection was penicillin and gentamicin (39%) followed by cefotaxime alone (13%) and cefotaxime and penicillin (7%). In total, however, 21 different combinations of antibiotics were used. The mean duration of antibiotic therapy was 9 days (median 8, range 4 to 21) for cases of sepsis and 15.7 days (median 14, range 7 to 35) for cases of meningitis.


These data suggest that the identification of a number of clinical risk factors in pregnant women, that is prolonged rupture of membranes, genital or urinary isolation of GBS, intrapartum fever, or prematurity may identify those women whose infants include 65% of those with EO GBS infection. This is consistent with figures from the USA in the era before widespread use of intrapartum antibiotic prophylaxis10 and with other UK studies.9 11 12 13 14 15 To evaluate these and other maternal factors in more detail, we matched controls to cases on the basis of birth weight. As low birth weight/prematurity is universally acknowledged as a risk factor for GBS disease,16 their exclusion might allow clearer interpretation of other risk factors, many of which are also likely to be associated with birth weight and gestation. For example, a cohort study on population-based risk factors for GBS disease conducted in metropolitan Atlanta concluded that infants with EO GBS disease were more often black, less than 2500 g in birth weight, born to teenage mothers and born to mothers with a history of miscarriage.2 In a study from London, women of black ethnic origin and those with multiple pregnancies were shown to have an increased risk of delivering an infant with EO GBS.11 In our analysis, however, black ethnic group, young maternal age (<20 years), multiple pregnancy and previous fetal death were not significantly over-represented among cases. The known association of these factors with low birth weight may explain why they do not emerge in our study as independent risk factors. This is supported by another US study, in which 41 cases of EO GBS were also matched with non-infected controls on the basis of gestational age.17 Black ethnic group, young maternal age and previous miscarriage were not significantly more associated with cases than controls. Intrapartum fever and having ⩾6 vaginal exams in labour were, however, associated with case status, consistent with the findings in our study.

In the only other UK-based case control study of GBS infection, each of 36 cases of EO GBS was matched with four healthy controls born at the same hospital.12 Antenatal isolation of GBS from maternal culture, prematurity, prelabour rupture of membranes, PROM and intrapartum fever were all significantly associated with cases.12 In our study, isolation of GBS, PROM and intrapartum fever remain strongly associated with GBS infection after multivariable analysis and thus reinforce the importance of these factors, even when birth weight is accounted for. The rationale for inclusion of these factors in risk-based prevention strategies is therefore clear.

It is concerning to note that only 17% of mothers with one or more risk factors received intrapartum antibiotics in this study. National guidelines on GBS prevention have subsequently been published by the Royal College of Obstetricians and Gynaecologists (RCOG) and recommend a risk-based approach.18 A recent audit of GBS protocols from 171 UK obstetric units suggests that current recommendations are broadly consistent with the RCOG guidelines. Some significant variation is noted, however, with a lower proportion of protocols recommending antibiotics for fever during labour, preterm labour and prolonged rupture of membranes (61%, 49% and 52% respectively). Reported practice is also noted to be variable.19 Surveillance data from the Health Protection Agency show that rates of EO GBS bacteraemia have increased between 2005 and 2006 in England and Wales (from 0.31 to 0.37/1000 live births),20 suggesting that a reappraisal of these guidelines and their implementation is urgently required.

Our study, as with others,11 12 17 emphasises that the majority of EO infections actually present in the first 12–24 h of life. Corresponding with this, we found that affected infants were more likely to need an emergency intervention in labour, to have fetal tachycardia and other evidence of fetal distress on monitoring, to require fetal blood sampling and to have lower Apgar scores than birth-weight-matched controls. It is likely that for many EO cases, the onset of GBS infection is in utero.

Late-onset GBS infection appears to be different from EO infection in a number of ways. LO cases are more likely to be of lower birth weight and gestation, and more likely to present with meningitis and focal infection than EO cases. A US study of 122 cases of LO infection matched with 122 controls found prematurity, black race and a mother with a positive antenatal GBS culture to be independently associated with LO disease.4 In our study, black ethnic group and young maternal age were not associated with LO disease. This again may reflect our matching strategy based on birth weight, although numbers were small. The strongest association with LO disease, as with EO disease, was with a positive antenatal GBS culture. A possible explanation for this is that maternal GBS carriage may persist for many months and thus carrier mothers may remain a source of infection for their infants. It is also known that infants colonised at birth may remain colonised for some time. In one prospective study, 10 of 21 infants with LO infections were colonised at birth with the same GBS type that subsequently caused disease.21

The findings that maternal infection was more often diagnosed and antibiotics more often prescribed for mothers of cases than for mothers of controls emphasise that GBS may also cause morbidity in mothers as well as their babies.22 Potential interventions to prevent neonatal GBS infection, such as intrapartum antibiotics or GBS vaccination, may also provide benefit to their mothers.

An important rationale for matching cases and controls on the basis of birth weight was to be able to distinguish the morbidity attributable to GBS infection separately from that due to underling low birth weight and prematurity. It is clear from these data that the management of GBS infection results in an appreciable use of hospital resources with a median total hospital stay at least three times greater than that of similar birth weight controls. A formal economic analysis of healthcare use has recently been published.23

There are several potential limitations of our study to consider. Antenatal swabbing for carriage of GBS was not a routine procedure, and the indications for performing this test were not given and may in themselves be risk factors for GBS disease. This may therefore be a source of bias; however, it reflects current clinical practice. Additionally, the ethics committee approval stipulated that information concerning cases and controls could not be sought until after parental consent had been obtained. Thus, we are not able to compare the features of cases and controls that were and were not recruited to the study to ensure that the latter group was representative. We can, however, compare cases from this study with those in the national study, as the latter attempted to define all cases and did not require parental consent.9 This analysis reveals that the cases included in this study appear to be representative of all UK GBS cases (table 6).

Table 6

Comparison between cases in this study with those in the national study9

The statistical analysis employed included all cases and controls in a multivariable analysis of maternal factors. As there were a number of multiple births, this meant that not all observations were independent. However, when clustering was taken into account, this made little difference to the results, and it was felt that this could safely be ignored. Excluding cases would have resulted in loss of case-control sets and decreased power, especially when considering interactions. Similarly, the subanalysis which sought factors that might discriminate between EO and LO cases should be viewed with caution, as the study was not designed for this comparison. We believed it was preferable to present the analysis of EO and LO data in this way, rather than as separate analyses, because of the small number of LO cases.

What is already known on this topic

  • Neonatal Group B Streptococcal (GBS) disease is associated with a number of maternal and neonatal clinical factors.

  • Many of these factors are inter-related, so their specific relevance to GBS disease may be hard to define.

What this study adds

  • We have quantified the importance of these factors, by matching GBS cases to controls by birth weight and using a multivariable analysis.

  • The strongest independent associations with neonatal GBS disease were maternal GBS carriage, maternal infection in the peripartum period and fever in labour.

  • The maternal risk factors that form the basis of current risk-based GBS prevention strategies are independently associated with neonatal GBS disease.


We thank all paediatricians, microbiologists, maternity staff and parents who contributed to the study.



  • Funding Meningitis Research Foundation.

  • Competing interests None.

  • Ethics approval Ethics approval was provided by the multicentre research ethics committee of Scotland (MREC/99/0/86).

  • Patient consent Obtained from the parents.

  • The HPA GBS Working Group comprises A Bedford-Russell, A Efstratiou, C McCartney, P Heath, R Hughes, M Jokinen, T Lamagni, A Mifsud, S Petrou, E Price, A Reynolds and L Schroeder.