Infant Group B Streptococcal Disease Incidence and Serotypes Worldwide: Systematic Review and Meta-analyses

Abstract Background Group B Streptococcus (GBS) remains a leading cause of neonatal sepsis in high-income contexts, despite declines due to intrapartum antibiotic prophylaxis (IAP). Recent evidence suggests higher incidence in Africa, where IAP is rare. We investigated the global incidence of infant invasive GBS disease and the associated serotypes, updating previous estimates. Methods We conducted systematic literature reviews (PubMed/Medline, Embase, Latin American and Caribbean Health Sciences Literature [LILACS], World Health Organization Library Information System [WHOLIS], and Scopus) and sought unpublished data regarding invasive GBS disease in infants aged 0–89 days. We conducted random-effects meta-analyses of incidence, case fatality risk (CFR), and serotype prevalence. Results We identified 135 studies with data on incidence (n = 90), CFR (n = 64), or serotype (n = 45). The pooled incidence of invasive GBS disease in infants was 0.49 per 1000 live births (95% confidence interval [CI], .43–.56), and was highest in Africa (1.12) and lowest in Asia (0.30). Early-onset disease incidence was 0.41 (95% CI, .36–.47); late-onset disease incidence was 0.26 (95% CI, .21–.30). CFR was 8.4% (95% CI, 6.6%–10.2%). Serotype III (61.5%) dominated, with 97% of cases caused by serotypes Ia, Ib, II, III, and V. Conclusions The incidence of infant GBS disease remains high in some regions, particularly Africa. We likely underestimated incidence in some contexts, due to limitations in case ascertainment and specimen collection and processing. Burden in Asia requires further investigation.

factors [10,11]. However, this strategy will not reduce late-onset infant GBS disease (LOGBS; onset on days 7-89 of life), and in LMICs and LICs, where there are more home deliveries and women present later for delivery, IAP may be less feasible and effective than other potential strategies for prevention, such as a maternal GBS vaccine.
This article therefore aims to examine the incidence of invasive GBS disease among young infants and the associated CFR and serotypes causing GBS invasive disease (Figure 1). It is part of a supplement estimating the burden of GBS disease among pregnant women, stillbirths, and infants [12]. The supplement includes systematic reviews and meta-analyses on GBS colonization, and adverse outcomes associated with GBS around birth [10,[13][14][15][16][17][18][19], which provide data inputs for estimating the worldwide burden of GBS [20]. OBJECTIVES 1. To provide a comprehensive, systematic literature review and meta-analyses on the burden of infant invasive GBS disease to include: a. Incidence of infant GBS disease: overall incidence risk, including stratification by EOGBS and LOGBS. b. CFR for EOGBS and LOGBS (7-89 days) and neonatal disease (7-27 days).
c. Serotype distribution: prevalence of GBS serotypes causing GBS disease among infants. 2. To generate parameters to be used as data inputs in a compartmental model estimating the burden of GBS in pregnancy for women, stillbirth, and infants; including a. EOGBS to LOGBS ratio. b. Clinical syndrome (proportion of neonatal disease that was meningitis or sepsis). 3. To evaluate data gaps and recommend improvements for the data regarding GBS disease in young infants.

METHODS
This article is part of a protocol entitled "Systematic estimates of the global burden of GBS in pregnant women, stillbirths and infants, " submitted for ethical approval to the London School of Hygiene & Tropical Medicine (reference number 11966) and approved on 30 November 2016. The general methods are described elsewhere [12]; here we present details specific to estimates related to the incidence of invasive GBS disease among infants. We included studies that described incidence risk, deaths, or serotypes of bacterial isolates among infants aged 0-89 days with invasive GBS disease. Eligible studies were those reporting data published or unpublished between 1 January 2000 and 31 January 2017, limited to humans and with no language restrictions. We identified data through systematic review of the published literature and through an investigator group that sought unpublished data from clinicians, researchers, and relevant professional institutions worldwide.

Definitions
Invasive GBS disease was defined as laboratory isolation of S. agalactiae from any normally sterile site using conventional microbiological methods together with any signs of clinical disease. EOGBS was defined as invasive GBS disease in infants aged 0-6 days after birth and LOGBS in infants 7-89 days after birth. Incidence risk was defined as cases per 1000 live births and CFR as number of deaths in GBS cases divided by total GBS cases.

Search Strategy
We undertook systematic literature searches of PubMed/ Medline, Embase, Literature in the Health Sciences in Latin America and the Caribbean (LILACS), the World Health Organization Library Information System (WHOLIS), and Scopus databases using the search terms ("Streptococcus agalactiae" [Medical subject headings (MeSH)] OR "Streptococcus Group B" OR "Group B streptococcal") AND "infant, " "outcome, " "death, " "mortality, " "case AND fatality AND rate. " We limited searches to humans and publications from 1 January 2000 to 31 January 2017 (see Supplementary Table 1 for the full list of search terms). For consistency, we used the same search terms as a previous systematic review [9]. We did not apply date or language restrictions; texts were translated to English when published in other languages. An additional search for reports with serotype data was performed, using the search terms ("Streptococcus agalactiae serotype" [MeSH] OR "Streptococcus Group B serotype" OR "Group B streptococcal serotype") using the same limits above. We used snowball searches of article reference lists including reviews to identify additional studies One investigator performed the database search, screened for duplicates, and screened titles and abstracts to assess eligibility for inclusion. Two independent investigators (L. M. and M. K. L.) assessed the full-length articles associated with selected abstracts to confirm eligibility and extract data. Where there was discrepancy between the 2 reviewers, a third investigator (A. S.) made the final decision.

Study Selection
We included studies with original data on GBS disease in infants who were aged 0-89 days at onset of infection episode, with clinical specimens obtained from a sterile site, which had a population denominator (total live births). We excluded studies focusing on very high-risk groups (such as only human immunodeficiency virus [HIV]-infected infants or only preterm infants), where data were not representative of live births in the population. Where countries had multiple or duplicated publications or systematically collected surveillance data, we included the most recent data. Studies reporting GBS disease in infants aged 0-90 days that did not specify age at onset for the individual cases were included with the 0-89 day studies as the probability of a case on day 90 is negligible. For full details of inclusion and exclusion criteria, see Supplementary Table 2.

Data Abstraction
We used a standardized data abstraction tool to capture information on the study design (prospective or retrospective), setting (health facility or not), use of IAP, timing of clinical disease (onset in the first 24-48 hours, EOGBS, and LOGBS), outcomes (survived or died), sample type (cerebrospinal fluid, blood, or other sterile site) and GBS serotype. For facility-based studies limited to babies born at the facility, facility live births was used as denominator. Where studies included inborn and outborn babies, a population denominator of all live births in the catchment area of the health facility was used. Data on study location were also abstracted including country and town. These data were imported into Stata version 14 software.

Analysis
We used random-effects meta-analyses to estimate overall infant disease incidence, EOGBS and LOGBS incidence, the EOGBS to LOGBS incidence ratio, and CFRs using the DerSimonian and Laird method [21]. In addition to worldwide estimates, estimates by United Nations regions and/or subregions were obtained when sufficient data were available.
To assess bias, we performed the following sensitivity analyses: 1. Invasive disease: a. Infant invasive disease limited to facility-based studies where denominator was facility births. b. EOGBS estimates limited to studies including data for days 0-6 after birth. c. LOGBS estimates limited to studies including data for days 7-89 after birth. d. Late-onset neonatal incidence limited to studies with data for days 7-27 after birth. 2. The ratio of early-onset disease to late-onset disease, including only studies considered to be less subject to case finding bias resulting from low access to care, nonsystematic sampling, or suboptimal laboratory detection methods [22][23][24][25] as considered by the expert advisory group.

Literature Search and Study Selection
We identified 7535 articles for consideration from database searches, 318 additional records from expert groups in neonatal care and reference lists, and 7 datasets from an investigator group [22,23,26]  . One hundred thirty-five articles (reporting data from 57 countries) met our inclusion criteria (search strategy of study selection in Figure 2). Of these, 90 reported incidence   Figure 2).

Study Characteristics
There were more data from HICs (77 studies) compared to LMICs (18 studies), of which 12 were from Africa (11 in sub-Saharan Africa, 1 in North Africa). Data inputs are illustrated in Figure 3A and 3B. Data inputs of the previous systematic review [9]

Serotype Distribution
A total of 6500 bacterial isolates were included in the meta-analysis of serotype prevalence (data inputs are illustrated in Supplementary Figure 11). Five serotypes (Ia, Ib, II, III, and V) accounted for 97% of invasive isolates in all regions with serotype data ( Figure 5 Figure 12).

Sensitivity Analyses to Assess Bias
Among facility-based studies with facility births denominator (n = 71), the incidence among infants 0-89 days of age was slightly higher than the main analysis  When we limited LOGBS estimates to studies with complete data for days 7-89 after birth (11/33 studies), the incidence estimate was 0.40 per 1000 live births (95% CI, .27-.53), higher than the main analysis (meta-analysis in Supplementary Figure 20). Including only studies with data for days 7-27 after birth (3/32 studies), the incidence was 0.16 per 1000 live births (95% CI, .08-.24) (meta-analysis in Supplementary Figure 21).

DISCUSSION
Our comprehensive review and meta-analyses represent an important update to the previous global invasive infant GBS disease burden estimates [9], and most notably include new data from LMICs (18 new studies from 10 LICs and LMICs). Infant invasive GBS disease incidence and case fatality is high in every world region, yet likely considerably underestimated in settings with limited access to care and diagnostics as <10% of neonates with suspected serious infection have a positive blood culture [139,140]. The overall estimated incidence of infant GBS disease, 0.49 per 1000 live births, is slightly lower than the previous estimate of 0.53 per 1000 live births (95% CI, .41-.62) [9]. While fewer studies in this review reported IAP use compared to the previous review (66% vs 77.0%), more weight (>50%) was applied to the data from Europe and the Americas where IAP is in use. The reduction in overall incidence is likely driven by lower incidence of invasive infant GBS disease in the Americas (0.43/1000 live births here vs 0.67/1000 live births in the previous review), and Europe (0.53 vs 0.57/1000 live births). This difference, especially in the United States where infant GBS rates declined notably during the study period, reflects the use of more recent data in our analysis.
Similarly, the incidence of invasive infant GBS disease in Africa (1.12/1000 live births) was also slightly lower than previously reported, although >2 times higher than in developed countries (0.46/1000 live births). This is the result of broader incidence data from Africa, including large studies in South Africa [24,42,96] Mozambique "(Sigaúque et al, Unpublished data)", and Gambia [68] reporting a high incidence of invasive GBS disease, in contrast to studies in Nigeria [28,77] and Zambia [62] reporting a very low incidence. Our point estimate for EOGBS incidence for Africa was higher compared to that reported in the most recent worldwide review [9], although LOGBS incidence for the same region was similar in both reviews [9]. There are many possible reasons for the increase in EOGBS incidence in Africa, which could be due to true emergence, increases in comorbidities such as HIV [42], or improved data collection to detect early disease. This high incidence is important in terms of total burden, as CFRs in Africa were also 4 times higher than in developed countries (18.9% and 4.7%, respectively); thus the greatest burden of cases, and deaths, is in Africa. However, our data are limited to few African studies mostly in Southern and Eastern Africa.
There are other important regional differences. The incidence of infant GBS disease was strikingly low in Asia at 0.31 per 1000 live births, with the lowest incidence in Southeast Asia (0.21/1000 live births). This may reflect a true regional difference, which could be related to differences in lower overall prevalence of maternal colonization and/or lower prevalence of serotype III [13], which is more commonly associated with the most virulent clone, clonal complex 17. Some of the difference may also be due to incomplete case ascertainment, being in Asia more challenging as they have more home births than Africa. For the earliest-onset cases (<24 hours of birth), differences in access to care and rapid and high case fatality can reduce case ascertainment. Cerebrospinal fluid sampling is infrequently performed in many parts of this region and that would reduce the apparent incidence of LOGBS disease, which is more frequently associated with meningitis. However, the lack of late-onset cases in this region does not fully align with those reasons and also suggests there may be more at play, potentially related to strain differences, level of natural acquired protective maternal antibody, or other host, environmental, or behavioral factors that may affect disease burden.
Difficulties in case ascertainment in LICs likely contribute to the higher incidences observed when the analysis was limited to facility-based studies, particularly in Africa. Studies in contexts where access to care, particularly for home deliveries, is difficult are likely to underestimate EOGBS disease incidence, due to the preponderance of cases with onset on day zero, which can be as high as 90% in studies with high-quality ascertainment but was 68% among the studies we included where this information could be extracted. Late-onset disease is likely underestimated too, due to studies that did not capture cases for the full 7-89 days; the sensitivity analysis showed the incidence of late onset disease to be almost twice as high (0.40 vs 0.26) when only studies with data for days 7-89 were included. This may result in an underestimation of the burden of GBS meningitis in particular, a significant concern given the morbidity associated with this condition.
Differences in the early-to late-onset disease ratios in different regions in the main analysis may also reflect (and reveal) biases in the data. Asia had the highest ratio of EOGBS disease to LOGBS disease, with the lowest in Africa (5.99 vs 1.02). It is possible that Asia has less LOGBS disease, consistent with the lower prevalence of maternal colonization with serotype III, a serotype commonly associated with LOGBS disease. Interestingly, when limiting the EOGBS:LOGBS ratio to high-quality studies, the estimate was similar across regions and lower than the overall EOGBS:LOGBS ratio. This is likely to be influenced by those countries with widespread IAP as well as South Africa, which may have a uniquely low ratio due to the high prevalence of maternal HIV infection, which predisposes to late-onset disease. However, no study from Asia was included in this analysis.
In terms of serotypes causing invasive disease, serotype III accounted for over half of all disease-causing isolates followed by serotypes Ia, V, and II, consistent with previous work [9,141]. Disease-causing serotypes were similar in prevalence across different regions, with some slight variations. The lowest prevalence of serotype III (where data were available) was found in South America and Southeast Asia, 2 of the regions with the lowest prevalence of serotype III among colonized pregnant women [13]. Serotype distribution was similar to that reported in the previous review [9], suggesting stability over time.
Our findings suggest GBS disease is an important cause of infant disease, despite the limitations in the data and uncertainties about the low incidence in Asia. In Africa, where the incidence is highest, the CFR is also highest, suggesting this is the region where prevention strategies are most critical to introduce. Existing preventive strategies using IAP are not usually available in low-income contexts and, with a higher number of home deliveries and late presentation to health facilities for delivery, IAP may be more difficult to implement. Maternal vaccination offers an alternative strategy, and the data we have suggest that a pentavalent conjugate vaccine (including Ia/Ib/II/ III/V) would cover almost all disease-causing serotypes (96%) in young infants worldwide ( Table 2).

Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.  What's new about this? • Most comprehensive review to date of published data, and supplemented by unpublished data, including more data especially from LICs and LMICs with total of 18 studies from 10 LICs and LMICs. • Extensive attempts to standardize inputs and to assess biases through a set of sensitivity analyses.
What was the main finding? • GBS is an important cause of invasive disease among infants worldwide, despite widespread use of intrapartum antibiotic prophylaxis in many countries in developed regions. • Highest incidence is in Africa; the lowest incidence is in Asia and not fully explained by these data.
How can the data be improved? • Lower incidence in Asia may be partially explained by lower maternal colonization rates and less-virulent serotypes, but more data are required to better understand regional differences. • Improved reporting of studies to better understand the biases in the data reported, for example, if low case ascertainment in the first 24 hours after birth, may be reducing reported GBS incidence rates, and if this occurs more frequently in some regions.
What does it mean for policy and programs?
• Prevention strategies are needed in all settings and particularly in the highest-burden settings (in Africa) • Higher proportion of meningitis is among LOGBS cases, for which there are currently no preventive strategies.
• A pentavalent vaccine (serotypes Ia/Ib/II/III/V) would cover the GBS serotypes causing almost all (96%) invasive disease in infants.
of Health. Members of the Expert Advisory Group received reimbursement for travel expenses to attend working meetings related to this series. A. S.-t. M. works for the Bill & Melinda Gates Foundation. C. J. B. has served as a member of the Presidential Advisory Committee for Seqirus Inc and of the CureVac Inc Scientific Advisory Committee, as well as undertaken consultancy work for Pfizer Inc. C. C. has received institutional compensation from Novartis for conducting GBS studies. P. T. H. has been a consultant to Novartis and Pfizer on GBS vaccines but received no funding for these activities. M. I. has undertaken sponsored research from Pfizer on pneumococcal disease in adults and from Belpharma Eumedica (Belgium) on temocillin antimicrobial susceptibility in Enterobacteriaceae. K. L. D. has received funding by the Bill & Melinda Gates Foundation to work on research on GBS serocorrelates of protection to inform vaccine trials, and travel expenses from Pfizer to attend a meeting on an investigator-led project on GBS. S. A. M. has collaborated on GBS grants funded by GlaxoSmithKline and by Pfizer and received personal fees for being member of its advisory committee; he has also collaborated on a GBS grant funded by Minervax. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.