2.1 Protocol and registration

This study protocol was registered in PROSPERO (CRD42023439201) and conducted in accordance with the PRISMA 2020 guidelines.

2.2 Eligibility criteria

We included original studies of any design (excluding case reports, conference abstracts, and unpublished studies) that reported associations between maternal BMI assessed before 10 weeks of gestation and the outcomes of interest: EOGBS, intrapartum vGBS (vaginal Group B Streptococcus), or rectovaginal/urinary GBS colonization before term. There were no date or language restrictions.

The PFO structure [13] was used to determine the suitability of article inclusion into the study:

• Population: Pregnant women.
• Factor: Pregestational BMI assessed before the 10^th gestational week.
• Outcome: EOGBS or its proxy outcomes, including intrapartum vGBS (vaginal Group B streptococcus), and rectovaginal or urinary GBS colonization before term.

The prognostic factor was defined as pregestational BMI ≥ 25 kg/m² compared with a reference BMI defined as the traditional “normal” range BMI of 18.5 to 24.9 kg/m². Studies reporting BMI in multiple categories were included to allow assessment of a potential dose–response relationship between increasing BMI and GBS-related outcomes.

2.3 Search strategy

A comprehensive search was conducted in MEDLINE (Ovid), Embase (Ovid), and CENTRAL (Cochrane Library) from database inception to the 28^th of January 2026. Grey literature was identified through Google Scholar, and citation tracking was conducted using Web of Science. The Google Scholar search included the terms “BMI”, “risk”, “GBS”, and “EOGBS”, and screening was conducted through the first 30 pages (300 results). Moreover, references cited by the included studies were also screened for eligibility. The search strategy combined the terms: pregnancy, BMI/obesity, and GBS-related outcomes. Our search strategy included controlled vocabulary Medical Subject Headings (MeSH) and free-text terms, combined in three blocks using the Boolean operator “AND”. In each block, exploded medical subject headings or equivalents, depending on the database, truncations, free-text words, and narrower terms operators were used as appropriate. The following MeSH terms were included in Embase and MEDLINE; high risk pregnancy, pregnancy, pregnancy complications, childbirth, obesity, morbid obesity, body mass, body mass index, body mass, maternal obesity, streptococcal infections and streptococcus agalactiae. The complete search strategy is available in S1 Table.

2.4 Study selection and data extraction

Two reviewers, PN/MB and EG, independently screened titles, abstracts, and full texts using Covidence [14]. Duplicates were removed using Covidence’s automated deduplication algorithm. Eligible studies were selected based on predefined PFO (Population, Factor, Outcome) criteria [13]. Two reviewers, MB and EG, independently extracted data in a predefined, partly adjusted form on study design, setting, BMI categorization, GBS detection methods, population demographics, and outcomes. Inconsistencies in both study selection and data extraction were settled by consensus. There was no need for a third-party adjudication.

Data extraction was based on information reported in the published articles and their supplementary materials. When highly relevant details were missing, attempts were made to contact the corresponding authors to obtain additional information. However, no responses were received. Therefore, the analysis relied on the data available in the published sources.

2.5 Statistical analysis

The association between BMI and the risk of GBS across studies was analyzed using a random-effects meta-regression [15] of the log odds ratio with standard errors estimated from the reported 95% confidence intervals. As the reported BMI intervals varied substantially between studies, we translated the intervals to a continuous scale by estimating the within-interval mean BMI through numerical integration, assuming a universal log-normal distribution [16] with a mean of log (24.6) and a standard error of log (1.2) based on data from the Danish Medical Birth Registry [17,18]. The reference categories representing normal weight were included with a low fixed standard error (0.001 on a log scale). A sensitivity analysis was performed by assuming alternative BMI distributions (normal and uniform) to test the robustness of the log-normal assumption used for interval estimation. Reported adjusted odds ratios were used in the main analysis when available to account for potential confounders. Since there is not yet a strong consensus on the key confounders of the association in question, we did not use a pre-specified approach. Cohort studies reporting hazard ratios were included, assuming that hazard ratios approximate the odds ratio, given the rare incidence of EOGBS. Due to overlapping cohorts in the two Swedish studies [6,7] evaluating BMI as an independent risk factor for EOGBS, the study with the shortest inclusion period, by Håkansson et al. [6] was excluded from the meta-analysis. Subgroup analyses for each proxy outcome were performed, as well as sensitivity analyses restricting results to studies reporting adjusted ORs using only crude OR, and leave-one-out analyses [19]. Further subgroup analyses were conducted according to the GBS testing methodology, including anatomical sampling site (rectovaginal vs mixed/other) and study design. The estimated OR with 95% confidence intervals for the relative risk of outcome per one unit change in BMI was reported, as well as the association between BMI and OR of outcome was plotted with the average normal BMI as reference (BMI = 22.3). Heterogeneity was assessed using Cochrane’s I² statistic [20]. A naive categorization of values for I² would not be appropriate for all circumstances, although we would tentatively assign adjectives of low, moderate, and high to I² values of 25%, 50%, and 75%. Publication bias was evaluated using residual funnel plots. The procedure meta meregress and own programs in Stata Statistical Software: Release 18.5 and College Station, TX: StataCorp LLC were used for the analyses.

2.6 Quality assessment

The quality of the studies was assessed using the Quality of Prognosis Studies in Systematic Reviews (QUIPS) tool [21]. The QUIPS tool is developed for studies on prognostic factors and supports a systematic appraisal of bias across six key domains: study participation, attrition, prognostic factor measurement, outcome measurement, confounding, and statistical analysis. Domain-specific decision rules were predefined prior to the assessment. In each domain, 3–6 subdomains were considered in accordance with guides for prognostic research [21]. Detailed reporting of prognostic factor measurement varied across studies due to their observational design. Given that BMI is a routinely used and standardized measure in epidemiological research, the lack of reporting of the assessment method was judged as moderate rather than high RoB (Risk of Bias).

Particular emphasis was placed on confounding and statistical analysis, as these were considered critical to prognostic research. Adequate statistical adjustment for confounding was required, including multivariable modelling with clearly reported effect estimates. Yet risk factors for GBS and relevant confounders remain incompletely established; however, based on previous research, GDM, diabetes, parity, socioeconomic status, and ethnicity were considered important potential confounders and integrated into the pragmatic RoB evaluation of the confounding domain. Studies reporting only univariable analyses were rated as high risk in the confounding domain.

Studies were classified as:

• Low overall RoB: low risk in all key domains (including confounding and statistical analysis/reporting) and no more than one domain rated as moderate risk.
• Moderate overall RoB: moderate risk in one or more key domains without any domain rated as high risk.
• High overall RoB: high risk in at least one key domain or multiple domains rated as moderate risk.

The assessment was performed by two authors, SK and EG, working independently, followed by evaluation. Disagreements were resolved through discussion until a consensus was reached.

3.1 Main results

The search strategy identified 826 records after duplicate records were removed (Fig 1). The grey literature and citation tracking did not yield any additional studies. Of the 826 records, 807 were excluded for the following reasons: reporting only the mean BMI; comparison of a BMI > 23 (within the normal range) with BMI < 23; BMI categories not meeting the eligibility criteria (BMI < 30 compared to BMI > 35); BMI not reported to GBS status; study population with BMI < 25, EOS (Early Onset neonatal Sepsis) outcomes not differentiating the infectious agent; or inappropriate study design, most commonly unpublished studies.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA flow-chart on the selection process.

https://doi.org/10.1371/journal.pone.0329423.g001

The 19 eligible studies reported data from a total of 3,707,047 women, encompassing 277,887 cases of either EOGBS (two studies [6,7]) or proxy outcomes (17 studies) (see Fig 1 and Table 1). Due to overlap between the EOGBS cohorts, only one study [7] was included in the primary analysis. The studies reporting proxy EOGBS outcomes were categorized according to the timing of GBS assessment: intrapartum (one study (5.3%) [22]); at gestational weeks 35–37, as in universal routine screening (e.g., in the US) (seven studies (36.8%) [12,23–28]); and at other time points during pregnancy (nine studies, (47.4%) [9–11,29–34]). Among studies assessing GBS in gestational weeks 35–37, routine universal screening was implemented in two of seven study settings [12,26]. The studies included nine European, five North American, three Asian, and two African populations, and all had either a retrospective or cross-sectional design. Study characteristics are presented in Tables 1 and 2.

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of included studies, geographical and temporal setup, size, design, and detection methods.

https://doi.org/10.1371/journal.pone.0329423.t001

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Characteristics of the study population.

https://doi.org/10.1371/journal.pone.0329423.t002

In the two Swedish nationwide cohort studies reporting EOGBS outcomes [6,7] the diagnosis was mainly based on culture-confirmed early-onset sepsis. The proxy outcomes used in the remaining studies were based on either intrapartum vaginal GBS-PCR; one study (5.3%) [22], rectovaginal culture; 10 studies (52.6%) [23–30,34,32], vaginal culture; one study (5.3%) [33] or mixed urinary/rectovaginal/vaginal GBS positive samples; five studies (26.3%) [9–12,31]. Common covariates considered potential confounders or effect modifiers such as maternal age, parity, smoking status, diabetes, race, socioeconomic status, hypertension, and preeclampsia (Table 3). BMI categorization varied considerably across studies, as illustrated in Table 3, which presents the studies’ main findings. When available, we based our analyses on adjusted rather than crude outcome measures, as emphasized in Table 3.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Study-reported associations between BMI categories (intervals) and outcomes, with a summary of the study conclusions.

https://doi.org/10.1371/journal.pone.0329423.t003

Our primary analysis, presented in Fig 2, shows the OR of EOGBS or proxy outcomes as a function of BMI with normal BMI at 22.3.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Association between BMI and EOGBS and proxy outcomes based on random-effects meta-regression.

Depicts ORs for EOGBS or proxy outcomes as a function of BMI, with normal BMI at 22.3 as reference, with the shaded region representing 95% confidence interval based on random-effects meta-regression. Bubbles represent individual studies with diameters inversely proportional to the standard errors of the reported associations.

https://doi.org/10.1371/journal.pone.0329423.g002

In the primary meta-analysis 18 studies were included. Assuming a log-linear association, each one-unit increase in BMI was associated with 2.4% higher odds of EOGBS and proxy outcomes (OR 1.024, 95% CI 1.010–1.037; p = 0.001). Sensitivity analyses showed consistent findings. The pooled OR based on crude estimates of all 18 studies was 1.024 (95% CI 1.009–1.041; p = 0.002), whereas the pooled OR based on adjusted estimates, including 14 studies, was 1.030 (95% CI 1.017–1.042; p < 0.001) per BMI unit (Fig 3).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Forest plot.

Forest plot representing the association between BMI and risk of EOGBS or proxy outcomes by odds ratios with 95% confidence intervals based on random-effects meta-regression on 18 studies represented with best linear unbiased predictors (BLUP) of the random-effects. Along with results from various supplementary and sensitivity analyses.

https://doi.org/10.1371/journal.pone.0329423.g003

When translated to clinically relevant BMI differences, these estimates correspond to approximately 20% higher odds at a BMI of 30 compared with 22.3, and about 52% higher odds at a BMI of 40. Estimated ORs across selected BMI levels are presented in Table 4.

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. OR for outcome across selected BMI levels compared to normal level (BMI = 22.3) based on the main analysis including 18 studies and for the sensitivity analysis based on the 14 studies reporting adjusted ORs.

https://doi.org/10.1371/journal.pone.0329423.t004

In S2 Table, absolute risk differences corresponding to the estimated OR for the BMI levels reported in Table 4 are given considering a background level of outcome of 0.43 per 1000 [3] for normal BMI. A BMI level of 40 represents a risk difference of 0.22 (95% CI: 0.07–0.37) in EOGBS or proxy outcomes per 1000.

Two studies, both included in the primary analysis, demonstrated markedly stronger associations (Fig 2). One of these [7], assessed as having a low risk of bias, was based on a large Swedish cohort and used EOGBS as the outcome. It reported an adjusted hazard ratio of 2.4 (95% CI 1.7–3.4) for BMI 35–39.9, with an even stronger association observed among term infants. In absolute terms, the unadjusted equivalent corresponds to a risk of 0.86 per 1000 live births among individuals with BMI 35–39.9 compared with 0.39 per 1000 among those with normal BMI (18.5–24.9), corresponding to an absolute risk difference of approximately 0.47 per 1000 births. The other study [22], assessed as having a moderate risk of bias, examined intrapartum GBS colonization and reported an OR of 2.68 (95% CI 1.26–2.72) for BMI ≥ 30. Overall, findings were broadly consistent but not uniform. Kwon et al. reported an inverse association [30], whereas Rick et al. and Rao et al. reported non-significant or attenuated associations [23,24]. Similarly, Zhou et al. and Nguyen et al. reported non-significant results, although this may reflect the lower BMI threshold used to define exposure (BMI > 25) [27,28]. Despite these differences, leave-one-out analyses indicated that no single study had a decisive influence on the main analysis (S1 Fig).

The influence of sampling method, study design, and model assumptions was examined via a range of sensitivity analyses (Fig 3), which showed minor changes in the pooled ORs. In particular, altering the assumed underlying distribution of BMI led to some changes; however, both distributions deviate markedly from the known skewed population distribution among pregnant women.

The heterogeneity between the studies was substantial with an I² of 75.4% (Fig 3). However, in studies analyzing rectovaginal cultures at gestational weeks 35–37, I² was only 3.5% [12,23–28]. For the proxy outcomes – based on seven studies assessing GBS colonization at or after the 35^th gestational week [12,23–28] and nine studies assessing GBS status at any time during pregnancy [9–11,29–34] – only studies with limited weight in the primary analysis showed results that deviated significantly from the overall findings (Fig 2).

3.2 Risk of bias assessment

The risk of bias (RoB) was evaluated and is available in Table 5.

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. QUIPS study quality assessment.

https://doi.org/10.1371/journal.pone.0329423.t005

According to the QUIPS assessment, seven studies were rated as having an overall low RoB, eight studies as moderate, and four as high [12,27,30,32]. Primarily, RoB issues were related to study attrition, selection bias, risk of confounding, or missing information on exposure and outcome characteristics (Table 5). Reporting of BMI assessment methods was generally limited. Pregestational BMI was most likely frequently based on self-report. Information on GBS assessment procedures—including sampling method, anatomical site, handling of multiple sampling sites, and microbiological analysis—was insufficiently described in several studies.

The regression funnel plot in Fig 4 did not reveal substantial asymmetry, indicating neither publication bias nor between-study heterogeneity.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Funnel plot.

Regression funnel plot including all studies (N = 18), representing 63 data points in the primary random-effects meta-regression analysis. Residuals, expressed on the log odds ratio scale, from the fitted model are plotted against their corresponding observed standard errors. The vertical line indicates zero residual (no deviation from the fitted regression line), and the dashed lines represent the approximate 95% confidence limits. Asymmetry in the dispersion of residuals across levels of precision may indicate small-study effects; however, such patterns may also reflect between-study heterogeneity or model misspecification.

https://doi.org/10.1371/journal.pone.0329423.g004

4.1 Principal findings

This systematic review suggests a dose–response relationship between increasing preconception BMI and a higher risk of EOGBS, intrapartum vGBS, and rectovaginal GBS colonization. Using a BMI of 22.3 as the reference, the association for this composite outcome corresponds to an OR of 1.20 at BMI 30, 1.35 at BMI 35, 1.52 at BMI 40, and 1.71 at BMI 45. For the most clinically important outcome, EOGBS, which was addressed in two overlapping studies, the adjusted hazard ratio in the largest cohort was 2.6 for women with a BMI ≥ 40 [7]. Although absolute risks are low, the estimated association with BMI may represent a clinically relevant difference in risk for higher levels of BMI. In the example described above, a BMI level of 40 represents an excess risk of 0.22 (95% CI 0.07–0.37) per 1000 (S2 Table), although due to the uncertainty as represented in this meta-analysis, this may be as low as 0.07 per 1000 above the level of the normal group.

The biological mechanisms linking increased BMI and EOGBS proxy outcomes are not yet fully understood, but several hypotheses exist. The association between obesity and EOGBS may involve specific pathways, including the interaction of dietary fats, such as palmitate, which can induce inflammatory responses in the gestational membranes and the placenta [35] potentially increasing susceptibility to GBS colonization [36]. Obesity alters immune responses, gut and vaginal microbiota, and inflammatory pathways, all of which may predispose individuals to persistent GBS colonization [37]. Additionally, metabolic dysregulation common in obesity may impair mucosal defences, thereby facilitating GBS persistence and vertical transmission to the infant during labor.

Although BMI is not a standard risk factor in many existing clinical guidelines for EOGBS, both research and clinical practice support our findings that high maternal BMI may increase the risk of EOGBS [38–40]. Besides obesity, GDM and pregestational diabetes are novel risk factors that have been studied. A recent meta-analysis from 2024 found GDM was associated with a 16% elevated risk of maternal GBS colonization, and an even larger risk of 76% reported for pregestational diabetes (OR 1.76, 95% CI 1.27–2.45) [41].

4.2 Clinical implications

When considering the rationale for including BMI as a risk factor in EOGBS prevention programs, it is essential to consider all aspects of a medical technology evaluation. In addition to clinical effectiveness, these include safety (e.g., allergic reactions and microbiome alterations), risks (e.g., bacterial resistance to antibiotics), costs (both short- and long-term), implementability (e.g., training requirements for healthcare staff), patient perspectives, ethical implications (e.g., equity), and alternative approaches (e.g., vaccination or intrapartum assessments of vGBS status by PCR techniques).

Further, one must acknowledge that most of the traditional risk factors in the risk-based screening model are notably stronger than the BMI-related risk found in this review; e.g., prolonged rupture of membranes > 18 hours (OR: 7.3), intrapartum fever (OR 4.1) or PROM at term (OR 11.5 with PROM or intrapartum fever – or both – present) and gestation < 37 weeks (OR: 4.8) [1]. However, including BMI in risk stratification may reduce the number of missed cases. In Denmark, for instance, women with a BMI > 35 account for 5.7% of the birthing population [42]. These women may face a 10% risk of a false-negative risk stratification [43,44].

4.3 Strengths and limitations

This study adheres to rigorous systematic review methodology, including comprehensive search, bias assessment, duplicate screening, structured RoB assessment, and reporting in accordance PRISMA standards. A major strength is the inclusion of diverse populations from high-, middle-, and low-income countries, enhancing external validity and generalizability. Furthermore, including large cohorts across diverse study designs increased the overall sample size and statistical power. However, substantial methodological and clinical heterogeneity must be acknowledged. The included studies differed in design (cohort, cross-sectional, and case-control), population characteristics, GBS screening strategies, BMI ascertainment, and statistical adjustment models. Although the use of a random-effects model accounts for between-study variance statistically, it does not eliminate underlying clinical or methodological heterogeneity. Consequently, pooled estimates should be interpreted as average effects across heterogeneous contexts rather than precise effect sizes applicable to all settings. Design-related heterogeneity is particularly important. Cross-sectional studies do not establish temporality, however subgroup analyses showed no weakening of the association when cross-sectional studies were excluded: OR 1.027 (95% CI 1.017–1.038, p value < 0.001).

Furthermore, case-control studies may be prone to selection bias. Although cohort studies provide stronger temporal inference, the observational nature of all included studies limits causal interpretation. We therefore emphasize that the findings demonstrate association rather than causation.

Risk of bias was primarily driven by potential confounding and measurement variability. Although many studies adjusted for key confounders such as maternal age, parity, diabetes, and smoking, adjustment strategies varied considerably, and several studies did not account for racial composition or behavioral and socioeconomic factors. The present study was not powered to assess potential effect modification by ethnicity [45,46], differences in health care systems, or socioeconomic status [26]. Notably, a stronger association among Black women has been consistently reported across several studies [10,23,45–48]. Stratification by adjustment status (adjusted vs. unadjusted estimates) did not materially alter the pooled results, suggesting that confounding is unlikely to fully explain the observed association. However, given the limited adjustment strategies in several included studies, residual confounding cannot be ruled out (Fig 3 and Table 3).

Measurement heterogeneity further contributes to the risk of bias. GBS ascertainment varied across studies in terms of testing protocols, including differences in culture versus PCR testing, swab collection techniques, and gestational timing. Similarly, the timing of BMI measurement varied across the included studies, introducing potential misclassification bias. Moreover, we assume that BMI assessment was commonly based on self-reporting, which may introduce recall bias. In many registry-based and historical cohort studies, the timing of BMI assessment was insufficiently specified, precluding subgroup analyses and limiting exploration of this source of heterogeneity. Limitations in the statistical strategy used may include the approach to variation between observations, which involved translating reported BMI categories into continuous-scale intervals based on data from the Danish Medical Birth Registry. However, this allowed for a rigorous meta-analysis with results spanning a wider range of BMI values, thereby enabling the proof of dose-response associations, a strength of this review.

In summary, this review is strengthened by methodological rigor, a large aggregated sample, and dose-response modeling. However, substantial clinical and methodological heterogeneity, variability in adjustment strategies, and inherent limitations of observational data introduce risk of bias and restrict causal interpretation. These findings should therefore be interpreted as robust associative evidence, warranting cautious inference and further high-quality prospective research.

4.4 Future directions

Additional research is needed to clarify the biological mechanisms linking obesity and GBS colonization and to refine screening guidelines to optimize the balance between effectiveness and safety [49]. Topics of interest could be the roles of the microbiome, inflammation, and metabolic pathways. Additionally, explore patient perspectives and ethical implications of revised prevention protocols. Moreover, trials are needed to assess whether lifestyle interventions, such as weight optimization or targeted probiotics, reduce GBS colonization and transmission. Lastly, modelling studies may help evaluate the cost-effectiveness and safety of incorporating BMI into GBS screening guidelines.