Background
Group B streptococcus (GBS) is a commensal bacterium that is generally found in the gastrointestinal and genitourinary tracts of pregnant women [
1] and can be passed to their baby via maternal rectovaginal colonisation during labour, causing neonatal early-onset GBS disease (EOGBS) [
1,
2]. Maternal rectovaginal GBS colonisation varies between populations but is estimated to occur transiently in approximately 18% of pregnant women worldwide [
3,
4]. GBS is a leading cause of adverse maternal and neonatal outcomes, including maternal and neonatal sepsis, stillbirth, and infant death [
1].
Intrapartum antibiotic prophylaxis (IAP) can prevent EOGBS [
5]. The World Health Organization (WHO) recommends IAP administration to women with GBS colonisation, within the context of local policy and guidance on GBS screening [
4]. Some countries like the Netherlands and the United Kingdom recommend risk-based protocols, giving IAP only in the presence of peripartum clinical risk factors. Other countries, like the United States (US), recommend both risk-based and universal culture-based screening for GBS colonisation, so that IAP can be given in the case of known colonisation [
5]. Culture-based testing remains the standard for antepartum screening [
2]. There is currently no international consensus on whether to recommend risk-based or universal culture-based screening for GBS [
5]. The need for laboratory processing limits testing capabilities in low-resource settings [
6]. A 2020 systematic review and meta-analysis found that screening-based protocols were associated with a reduced risk of EOGBS compared to risk-based protocols, without an associated higher antibiotic administration rate [
5]. If pregnant women do screen for GBS colonisation, it is important that the sampling approach has reasonable test accuracy, as false positives can contribute to overtreatment and resultant risk of antibiotic resistance, as well as having adverse effects on neonatal microbiome development [
5]. Conversely, false negatives present a missed treatment opportunity to reduce the risk of maternal and infant morbidity and mortality [
5].
GBS testing is most sensitive when performed near or at term [
7]. GBS detection rates are higher when a combined vaginal-rectal swab is taken compared to a single vaginal or rectal swab only [
8]. Testing is traditionally done with GBS culture, either by or direct plating and/or incubating the specimen in enriched culture medium. Enriched culture has a higher sensitivity than direct plating alone [
9]. Nucleic acid amplification test (NAAT) methodology for GBS testing is available; however, it has limitations, such as the inability to perform susceptibility testing. NAAT has not yet been universally adopted [
2].
WHO defines self-care as, “the ability of individuals, families and communities to promote health, prevent disease, maintain health and cope with illness and disability with or without the support of a health worker… Self-care interventions are tools that support self-care.” Self-care interventions include self-collection of samples [
10]; this involves an individual taking their own specimen, which is sent to a laboratory for processing [
11]. Systematic reviews comparing self-collected and healthcare provider-collected samples in the general population have found comparable accuracy for reproductive tract infection (RTI) testing, including sexually transmitted infections (STIs) [
12] and human papillomavirus (HPV) [
13]. A 2019 systematic review found that self-collection of samples for STI diagnosis in the general population offers convenience, confidentiality, expanded access, and increased patient autonomy and empowerment [
11]. Self-sampling for STIs is acceptable to patients [
14], and programmes offering self-collection have been found to increase uptake of STI testing [
11,
15,
16] and case finding [
11], without significant adverse outcomes [
15]. This may not necessarily translate into increased uptake of screening for GBS during pregnancy, as barriers specific to STI screening, like stigma, may not be as significant an issue for GBS sampling, which is a routine part of antenatal care in some countries. However, screening enablers such as increased acceptability of and reduced embarrassment associated with self-sampling compared to provider-collection may be transferrable.
The US Centres for Disease Control and Prevention guidelines advise that, when paired with clear patient instructions, self-collected vaginal-rectal specimens in pregnancy have similar GBS culture yield rates to provider-collected specimens [
2]. However, the supporting evidence for this advice are studies that have not been formally synthesised [
17‐
20]. No previous systematic review has assessed the accuracy of self-collected samples for RTIs in pregnancy, including GBS, and whether self-collection of samples for RTIs in pregnancy can improve maternal and perinatal health outcomes.
Ensuring there is high quality evidence that sample self-collection for GBS screening is as accurate as provider-sampling presents an opportunity for a strengthened evidence base to support self-care interventions during pregnancy. Expansion of self-care in this context could improve patient choice, convenience, and autonomy, as well as expand antenatal care coverage, ultimately improving health outcomes [
21]. This review aimed to determine (1) whether self-collected samples are as sensitive and specific as provider-collected samples for detection of GBS colonisation in pregnant individuals and (2) whether a self-collection strategy for detection of GBS in pregnancy compared to provider-collection can improve maternal and perinatal health outcomes.
Methods
Search strategy and assessment of eligibility
This systematic review and meta-analysis is part of a larger systematic review (PROSPERO CRD42023396573) which aims to determine the diagnostic accuracy and health effects of sample self-collection for RTI testing in pregnant individuals, compared to provider-collection. In this paper, we report on studies assessing GBS colonisation (results for other RTIs will be reported separately). We report these findings according to PRISMA-DTA guidelines (see Additional file
1 for PRISMA Checklist) [
22].
Eligible studies included those comparing self-collected to provider-collected samples for GBS testing of pregnant individuals. We included studies whose population were entirely or partially comprised of pregnant individuals, provided that disaggregated data for pregnant participants could be obtained (either from the published article, or by contacting study investigators). Studies were eligible if self-collected and provider-collected samples were taken from the same participant or participants were randomised to either self-collection or provider-collection. To be eligible, studies needed to have all participant samples collected from the same anatomical site, with the same type of sampling device, following the same sample transport process, with the same sample processing and test performed on both samples, using the same test cut off. Eligible studies were those reporting positive and negative test results for both self-collected and provider-collected samples. Randomised or quasi-randomised controlled trials, controlled before-after studies, interrupted-time-series studies, historically controlled studies, cohort studies, cross-sectional studies and case–control studies were eligible. We excluded case reports, case series, conference abstracts, poster presentations, editorials, correspondence, and qualitative studies. For protocols of ongoing trials dated 2019 or later, we contacted authors to see if trial data were available.
The following databases were searched on 18–21 June 2022: CINAHL Plus via EBSCOhost (from 1937), Medline and EMBASE (from 1946), Maternity and Infant Care Database (from 1971), Cochrane Central Register of Controlled Trials (from 1998), and the Cochrane Database of Systematic Reviews (from 1996) via Ovid. The search strategy combined keywords and subject headings on self-care (including self-sampling and self-collection) AND pregnancy AND reproductive tract infections (see Additional file
2 for full search strategy). No date or language restrictions were applied. We conducted a manual search of the reference lists of systematic reviews on similar topics for non-pregnant participants, as well as those of included studies in this review.
Two reviewers independently screened all titles/abstracts and potentially eligible full texts for inclusion using Covidence, according to the eligibility criteria. Any disagreements were resolved through discussion or consulting a third reviewer. When warranted, Google Translate was used for studies not in English. For eligible randomised trials, two independent reviewers assessed trial integrity using an adapted research integrity assessment (RIA) checklist tool, which consists of six domains to assess trial research integrity [
23,
24]. We reported findings from the RIA tool for each study by domain and contacted study authors for further information regarding these concerns.
Two reviewers independently extracted data and performed risk of bias assessment using an Excel-based form. Disagreements between individual judgements were resolved by a third reviewer. When multiple articles reported on the same study, data was combined into a single data extraction.
We extracted data on study characteristics—study location, sample size, eligible participants, sample collection process, including location and timing, anatomical site, specimen type, and sampling device, sample transport, stage medium, type of test, test threshold, and the number of GBS colonisation true positives (TP), false positives (FP), true negatives (TN), and false negatives (FN).
Our review outcomes (see Additional file
3) were diagnostic accuracy, and maternal, perinatal and neonatal outcomes. Additional secondary outcomes were outcome of sampling order, uptake of self-collection, case finding, linkage of positive test to clinical assessment or treatment, feasibility, and patient acceptability and preference. Missing or unclear information was noted as such. Where review outcome data were missing, we contacted study investigators to see if additional data were available. Risk of bias in the included studies and concerns regarding applicability to the review question were assessed independently by two reviewers using the QUADAS-2 checklist [
25]. Any disagreements were resolved by a third reviewer.
Meta-analysis was conducted according to the Cochrane Handbook for Systematic Reviews of Diagnostic Test Accuracy, Version 2.0 [
26]. Trials and non-trials were analysed separately. Revman 5.4 was used to generate forest plots for sensitivity and specificity. Other analyses were performed using Stata SE version 17 (STATA Corp., Texas, USA). Using the Stata command
metandi, we calculated the sensitivity and specificity for each study and pooled these estimates using a bivariate, random-effects, meta-analytic model. This model is required to describe variability in accuracy between studies, due to expected heterogeneity [
26]. A Hierarchical Summary Receiver Operating Characteristic (HSROC) plot was constructed using
metandi, which provides a global summary of both sensitivity and specificity estimates and accounts for between-study variability [
27]. The HSROC plot presented sensitivity and specificity estimates from individual studies, the pooled estimates, 95% confidence interval (CI), and 95% prediction interval, i.e. the confidence region for a forecast of true sensitivity and specificity in a future study [
27].
Since no covariates were included in the meta-analysis, the bivariate and HSROC models were mathematically equivalent [
28] and thus presented together. Stata’s
metadta command was used to calculate the bivariate
I2 value, which measures heterogeneity whilst accounting for the correlation between logit sensitivity and logit specificity [
29]. Meta-regression was not done because there were less than ten studies in each meta-analysis [
26]. There were insufficient or zero studies to undertake subgroup analyses by sample anatomical site, gestational age (GA), culture technique, or NAAT thresholds. Forest plots were instead examined visually for any trends in pre-specified subgroups of interest.
Sensitivity analyses were conducted to investigate the effect of non-randomised studies on the summary estimates which (1) included participants at < 35 weeks’ gestation (n = 3) and (2) did not incubate all culture samples in enriched culture media (n = 1). There were too few studies for further sensitivity analyses.
Conclusions
Expanding GBS screening and task-shifting away from clinical services to home-based and community service testing could expand access and decrease burden on healthcare systems. This is particularly so in limited-resource settings with healthcare infrastructure limitations, although sample processing is dependent on the availability of laboratory services and trained staff, which may not be feasible in all low-income country settings. Nearly all studies we identified were conducted in high-income countries; self-collection could be useful for individuals in rural and remote regions in high-income countries if the option of posting samples was available. The evidence from this meta-analysis supports the option of self-collection of samples for GBS testing for individuals who decline provider-collection, who are hard to reach, or face barriers to antenatal care. Availability of accurate sampling options for screening to suit patient choice can hopefully improve uptake of GBS testing and thus reduce the incidence of EOGBS. In high-income settings with universal GBS screening in pregnancy, prevalence of adherence to antenatal GBS screening has been reported as ranging from 52 to 85.5% in Australian and US studies [
42‐
44].
Whilst sample self-collection, as an additional option to provider-collection for detection of GBS colonisation, is promising, further research is required to determine whether self-collection at home, without same-day instructions on how to self-collect, would be as accurate, and improve uptake.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.