Safety of anti-infective agents use during pregnancy on offspring: protocol for an umbrella review incorporating a novel three-stage data extraction strategy Authors: Wenling Wang, Long Ge, Yiquan Xiong, Jingwen Chen, Chen Chen, Yan Ren, Lehana Thabane, Xin Sun, Jing Tan Published: 2026 ABSTRACT AbstractBackgroundAnti-infective agents are widely used during pregnancy to prevent, treat, and mitigate the transmission of infections, and their utilisation has significantly increased over the past decade. However, concerns persist regarding their potential embryotoxicity, compounded by inconsistencies in safety data that complicate medication decision-making. This protocol details an umbrella review that aims to provide a comprehensive and methodologically sound synthesis of evidence from systematic reviews (SRs) on the adverse effects of prenatal anti-infective exposure on offspring. The review will incorporate a standardised, three-stage data extraction strategy and an R-based algorithm to efficiently identify and manage overlapping evidence.MethodsWe will conduct a comprehensive search across six databases (PubMed, Embase, Scopus, Web of Science, Cochrane Library, and Epistemonikos) from inception to June 22, 2024. The search will cover three key concepts—pregnant women, anti-infective agents, and systematic reviews—without restrictions on controls or outcomes. We will include SRs, with or without meta-analyses, that evaluate the adverse effects of prenatal anti-infective exposure and report either neonatal outcomes (e.g. birth defects) or long-term offspring outcomes (e.g. neurodevelopmental disorders). To address overlapping evidence, a structured three-stage data extraction process, supported by an R-based algorithm, will be implemented, sequentially handling PICO definition extraction, overlapping evidence management, and quantitative data extraction. At least two independent reviewers will screen studies, extract data, and assess methodological quality and evidence strength using AMSTAR-2 and GRADE, respectively. Evidence synthesis will be conducted narratively, drawing on meta-analytic associations from direct comparative analyses and organising findings by anti-infective agents, indications, outcomes, effect estimates, certainty of evidence, and methodological quality. Anti-infectives with safety concerns will be summarised in tables and evidence maps.DiscussionBy adopting a standardised process that incorporates an automated algorithm for managing overlapping evidence, the planned umbrella review will provide a comprehensive synthesis of existing evidence on the adverse effects of anti-infective use during pregnancy. The findings are expected to inform clinical practice, guide policy decisions, and support infection management in pregnant patients. Furthermore, the review will identify key knowledge gaps, help prioritise future research, and ultimately contribute to improving maternal and offspring health outcomes.Systematic review registrationPROSPERO CRD42024577013Supplementary InformationThe online version contains supplementary material available at 10.1186/s13643-026-03235-5.Keywords: Umbrella review, Overlapping, Anti-infective agents, Antibiotics, Antiviral medicines, Birth defects, Adverse offspring outcomes Background The intrauterine environment plays a pivotal role in shaping offspring health outcomes postnatally. Medication safety during pregnancy is a major global concern [1–3], with 63 to 97% of pregnant women taking at least one drug during this period [4, 5]. Anti-infective agents—including antibacterials, antivirals, and antiparasitics—are the most commonly used medications, with 28 to 40% of pregnant women exposed and a rising trend observed over the past decade [5, 6]. Nonetheless, prescribing these agents to pregnant women with infections remains clinically challenging. Maternal infections can significantly affect offspring outcomes. Conditions such as malaria and HIV during pregnancy increase the risk of stillbirth [7, 8]. Infections are common among women planning pregnancy or those already pregnant [9, 10]; therefore, prenatal medication is often recommended to prevent, treat, and reduce the transmission of infections [11–14]. However, anti-infective agents may pose embryotoxic or teratogenic risks, as they can cross the placenta, enter foetal circulation, or interfere with placental function [15]. Physiological changes during pregnancy can also alter drug metabolism and disposition [16, 17], raising concerns that inappropriate dosing may affect offspring outcomes [18], particularly in chronic infections requiring prolonged exposure [11]. Beyond increasing the risk of birth complications, including birth defects (BDs) and preterm birth (PTB) [19, 20], anti-infective agents may have long-term effects on neonatal health, such as impacts on immune and neurological development (e.g. atopic march and autism spectrum disorder) [21–23]. These complexities, together with inconsistent and sometimes unreliable evidence, make it difficult to draw definitive conclusions about the safety of anti-infective use during pregnancy and complicate clinical decision-making [24–27]. Due to the diversity of anti-infective agents and potential adverse outcomes, no single systematic review (SR) or overview comprehensively assesses the effects of perinatal anti-infective use on offspring health to guide evidence-based decisions. Numerous SRs focusing on specific anti-infectives or outcomes have been published [13, 14, 28], often with substantial topical overlap [29, 30]. An umbrella review (UR), which synthesises multiple SRs addressing related questions, is therefore an optimal approach [31]. However, the absence of standardised methodology guidelines—particularly for managing overlapping evidence across such broad topics—poses a major challenge and can undermine research quality and credibility [32–34]. Decisions regarding the inclusion or exclusion of overlapping evidence substantially influence review conclusions, thereby increasing the risk of selective reporting. Existing overlap processing approaches struggle to efficiently remove overlapping evidence while maintaining comprehensiveness. For instance, URs typically address overlap at the SR level during screening by retaining only one review per topic, often without accounting for differences in PICO elements across quantitative analyses [35, 36]. Some URs manually calculate the corrected covered area (CCA)—a quantitative measure of overlap—and subsequently exclude highly overlapped evidence, which substantially increases workload [37]. Moreover, recently developed automated tools primarily automate the calculation of CCA and therefore do not meet practical needs, such as identifying and excluding overlapping evidence in URs addressing broad topics [38, 39]. Accurately identifying overlap must align with the research question and is essential for effective overlap management. Unlike URs that define overlap solely by identical drug-outcome pairs, perinatal pharmacology requires consideration of primary study characteristics, particularly control design and study type. Notably, head-to-head comparisons are especially valuable, as they minimise indication bias more effectively than blank-control studies and help determine whether observed associations reflect the medication itself rather than underlying maternal conditions [30, 40, 41]. However, a recent UR on antenatal antibiotic use relied exclusively on blank-control comparisons, overlooking this nuance [41]. Ethical constraints further limit the number of randomised controlled trials (RCTs) that can evaluate medication use during pregnancy, and many key safety outcomes cannot be adequately captured within RCTs. Therefore, evidence from non-randomised studies (NRSs) should not be arbitrarily excluded, as it provides indispensable complementary information. Recognising the necessity of partial data extraction for overlap management, we integrated it into the data-extraction workflow, establishing a structured three-stage process. The core principle is to first identify overlapping meta-analyses (MAs) based on PICO information and then extract quantitative data only from the de-duplicated MAs. This strategy will avoid unnecessary extraction—such as detailed primary study data from non-overlapping MAs used solely for CCA calculation—and will efficiently eliminate redundant evidence. Moreover, we established explicit criteria for identifying overlaps to prevent the omission of important evidence, including that derived from active-control studies, and to ensure comprehensive coverage. Building on this framework, we have refined a standardised process for conducting URs on complex topics. This method, supported by an automated algorithm, will enable an efficient and comprehensive evaluation of the scope, quality, and strength of associations between prenatal anti-infective use and adverse offspring outcomes, with an emphasis on causal inference. This protocol provides detailed methodological guidance to support future research and enhance evidence synthesis in this field. Methods Registration and reporting information The protocol for this UR is registered on the PROSPERO platform (CRD42024577013) and adheres to the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) 2015 checklist (Additional File 1) [42]. It has been developed in accordance with the Cochrane Handbook guidelines for SRs [31] and other relevant published studies [35, 37, 43]. Upon completion, we will report our findings in accordance with the Preferred Reporting Items for Overviews of Reviews (PRIOR) guidelines for overviews of reviews of healthcare interventions [44]. Eligibility criteria This UR aims to identify studies that explore the association between prenatal use of anti-infective agents and adverse offspring outcomes. The eligibility criteria are detailed below. Population/participants We will include SRs that specifically focus on pregnant women, regardless of maternal conditions such as infections or other complications. For SRs conducted in the general population that report subgroup analyses for pregnant women, only data pertaining to the pregnant subgroup will be included in our analysis. Interventions/exposures We will include SRs investigating the use of anti-infective agents during pregnancy, without restrictions on indication, route of administration, dosage, or duration. Based on the WHO Model List of Essential Medicines (23rd List, 2023) [45], three major categories of anti-infective agents will be considered: (1) antibacterials, including antibiotics, antifungals, antileprosy, and antituberculosis medicines; (2) antivirals, including antiretrovirals, antiherpes, antihepatitis, and COVID-19 treatments; and (3) antiparasitic agents, including antiprotozoal medicines, anthelminthics, and medicines for ectoparasitic infections. Studies on vaccines and intrapartum antibiotic prophylaxis will be excluded, as vaccines are not classified as anti-infective agents, and intrapartum exposures are intentionally distinguished from sustained prenatal medication use. Comparators/control We will consider any type of control group, including placebo, no treatment, alternative anti-infective agents, or the absence of a comparator. Outcomes We will consider all potential adverse offspring outcomes, excluding those related to the vertical transmission of maternal infections. Specifically, we will focus on BDs diagnosed via systematic ultrasound or genetic testing during pregnancy, as well as malformations identified by clinicians from pregnancy through the postnatal period. This includes overall BDs, organ system–specific malformations (e.g. congenital heart defects), and specific types of BDs (e.g. cleft lip). We will also consider both adverse birth outcomes (e.g. miscarriage, stillbirth, and PTB) and long-term outcomes during childhood, including neurodevelopmental impairments, immune system disorders, respiratory diseases, and other chronic health conditions, as defined by the authors of the included SRs. SRs investigating any of these outcomes will be eligible for inclusion. Study types We will include all types of SRs, such as those with meta-analyses (SR-MAs), SRs without MAs, network meta-analyses, and individual participant data (IPD) meta-analyses. No restrictions will be applied to the design of the primary studies included in the SRs, encompassing both NRSs and RCTs. Each SR must be formally documented and must demonstrate four essential methodological characteristics, regardless of whether the included primary studies have been assessed [31, 46]: (1) explicitly defined inclusion and exclusion criteria, (2) a reproducible search strategy, (3) a transparent literature selection process, and (4) clearly described methods for quantitative or qualitative data extraction and synthesis. These criteria, adapted from the Cochrane Handbook, will ensure alignment with our review objectives. SRs that do not meet these fundamental criteria will be excluded. Reviews pooling studies identified without a systematic search and ongoing SRs will also be excluded. Other review types—such as scoping reviews, rapid reviews, and narrative reviews with partial systematic elements—will be excluded unless they explicitly satisfy all four SR criteria and examine the association between prenatal anti-infective use and adverse offspring outcomes. To ensure the availability of sufficient information for eligibility assessment, data extraction, and quality appraisal, full-text access will be required for all included studies. Therefore, publications available only as conference abstracts will be excluded. In contrast, other forms of grey literature—such as academic theses and preprints—will be eligible for inclusion, provided that full texts are accessible and all eligibility criteria are met. Search strategy Six databases (i.e. PubMed, Embase, Scopus, Web of Science, Cochrane Library, and Epistemonikos) will be searched from inception through June 22, 2024, without restrictions on language or document type. The search will cover three key concepts: pregnant women (population), anti-infective agents (intervention/exposure), and systematic reviews (study type), with no restrictions on comparators or outcomes, to maximise search scope. The search strategy, combining Medical Subject Headings (MeSH) and free-text terms, has been developed by experienced evidence-based medicine experts and reviewed by an information specialist. The full search strategy is provided in Additional File 2. In addition, we will screen the reference lists of included SRs and relevant reviews. A systematic search of grey literature will not be conducted. Study selection After de-duplication using Endnote X9, studies will be uploaded to the Rayyan platform (https://rayyan.ai/) [47] and will be screened by at least two reviewers based on titles and abstracts. Studies will not be excluded at this stage if they are explicitly identified as SRs by their authors or if their abstracts demonstrate compliance with the predefined SR criteria. Reviewers will then assess the full text of included or conflicted studies and will document reasons for exclusion (Fig. 1). Studies that do not meet our definition of an SR will be excluded and will be labelled as ‘non-systematic reviews’. Figure 2 presents the decision flowchart for literature screening. Disagreements at both screening stages will be resolved through discussion or consultation with a third reviewer. Additionally, all reviewers will receive standardised training and coaching to achieve consensus on eligibility criteria before formal screening begins. Adjusted PRIOR flow chart for literature screening Decision flowchart for including eligibility studies Data extraction and overlap processing To address the lack of systematic and efficient methods for identifying and managing overlaps, we will implement a three-stage data extraction strategy, driven by an R-based algorithm (Fig. 3). This approach integrates overlap processing directly into the data-extraction workflow and leverages PICO information to identify and manage overlaps, thereby efficiently retrieving the most comprehensive and recent evidence from eligible studies. The three-stage data extraction strategy of data extraction and overlap processing Following uniform training, at least two reviewers will independently extract data using standardised forms and will resolve any discrepancies through discussion. Data management will be handled using Microsoft Access, and overlap processing will utilise R 4.4.1. The R-based algorithm and an example dataset (provided to guide data formatting) are available in our GitHub repository (https://github.com/WANGDALINGLING/OP-UR.git). Before data extraction, all eligible SRs will be classified into two groups—SR-MAs and SRs without MAs—based on whether they have performed MAs, defined as the synthesis of more than one independent effect estimate for the population, interventions, controls, and outcomes (PICO) of interest. Crucially, we differentiate between SR-MA and MA: SR-MA refers to the entire systematic review, while MA denotes the quantitative synthesis conducted within it. A single SR-MA may include multiple MAs due to PICO variations. Phase one: identifying and managing overlaps Phase one: identifying and managing overlaps This phase will focus on identifying and managing overlapping MAs to ensure the next stage acquires the most recent and comprehensive quantitative data while minimising redundant extraction. Overlap identification and management will be accomplished using our R-based algorithm, which includes six user-friendly functions. Unlike existing tools that only quantify overlap, these functions import standardised data from SR-MAs to automatically categorise PICO clusters, quantify overlap, remove overlapping MAs, and generate data location tables. The following four steps detail this process. Step one: extracting and standardising data Basically, we will extract key information from all included SR-MAs, including study characteristics (e.g. author, publication year, number of primary studies included), search details (e.g. databases searched, search dates), eligibility criteria, and risk of bias assessments of the included primary studies. We will also extract the PICO definition of each MA conducted within SR-MAs and will standardise them by harmonising the terminology based on the original descriptions provided in the studies. These PICO data will form the core for identifying overlapping MAs and will enable a clinically problem-driven approach. Step two: categorising PICO clusters and identifying overlaps Overlapping occurs when multiple MAs address the same PICO question. Therefore, we will first categorise MAs into distinct clusters (i.e. PICO clusters) based on the extracted PICO definitions, using the R function PICO_cluster() to streamline the categorisation process. We anticipate generating a substantial number of clusters; for example, with four categories each for population, intervention, control, and outcome, a total of 256 clusters could be produced. Next, we will use the identify_overlappingMA() function to identify clusters containing more than one MA and will extract the identifiers of these overlapping MAs. To further quantify the degree of overlap among MAs within each overlapping cluster, we will extract only the basic information of the primary studies cited in the overlapping MAs, thereby avoiding extraction from non-overlapping MAs. Step three: managing overlaps We will use the CCA to assess overlap between MAs within each overlapping cluster. CCA accounts for the number of shared primary studies across MAs while correcting for the total number of studies and MAs, thus providing a standardised index of overlap [48–50]. Higher CCA values indicate greater overlap; in this study, a 5% threshold will be applied. We will use two functions—creat_citationmatrix_CCA() and remove_overlappingMA()—to quantify overlap and manage redundancy. Specifically, these functions first create a citation matrix, then calculate CCA for each pair of MAs within the cluster (i.e. pairwise CCA), and finally remove duplicates according to predefined rules. If the pairwise CCA exceeds 5%, only the most recent and comprehensive MA is retained; otherwise, both MAs are kept (Fig. 3, middle right). If a cluster retains multiple MAs, further analyses will examine differences in eligibility criteria or search strategies that may explain the discrepancies. Step four: constructing data location table Next, we will document the retained MAs and their corresponding SR-MAs using the collect_needextractMA() function to generate a data location table. In this table, rows correspond to SR-MA identifiers, while columns represent PICO cluster identifiers. Each cell indicates the MA derived from a specific SR-MA that has been assigned and retained for the corresponding PICO cluster, which will require extraction in phase two. This table will facilitate efficient extraction of the necessary quantitative data from SR-MAs. To verify which MAs have been removed, if needed (e.g. to check for inadvertent omissions), the collect_needextractMA_S() function will be used to generate a supplementary table that retains all MAs without removing overlapping ones. Phase two: extracting and synthesising pooled estimates Phase two: extracting and synthesising pooled estimates Using the data location table, we will extract information from the retained MAs in each PICO cluster, including synthesis methods and results. For network meta-analyses, only direct estimates will be extracted. For IPD analyses, only primary data obtained through systematic searches will be included. For main and subgroup analyses (focusing on factors relevant to our topic), we will prioritise pooled estimates derived separately from RCTs and NRSs that have adjusted for confounding factors. Furthermore, where available, we will extract results from sensitivity analyses to assess the robustness of the findings. Phase three: supporting a scoping survey Phase three: supporting a scoping survey To support a scoping survey on the safety of prenatal anti-infective agent use, we will extract and synthesise qualitative data from all eligible SRs. Since qualitative data from SR-MAs will have already been extracted in phase one, we will focus on extracting relevant study characteristics, PICO information, and risk-of-bias assessments from SRs without MAs. Assessment of methodological quality The methodological quality of all included SRs will be evaluated using the AMSTAR-2 (A MeaSurement Tool to Assess systematic Reviews, version 2) (Fig. 4) [51]. AMSTAR-2 assesses 16 domains covering critical aspects of review methodology, including protocol registration, comprehensiveness of the literature search, inclusion and exclusion criteria, quality appraisal, and results synthesis. Each domain will be rated as ‘Yes’, ‘Partial Yes’, or ‘No’. Following AMSTAR-2 recommendations, the overall quality of each SR will be categorised as ‘High’, ‘Moderate’, ‘Low’, or ‘Critically Low’, taking into account weaknesses in both critical and non-critical items. Methodological quality ratings will not be used to exclude SRs but will instead guide recommendations for improving the quality of future research. Two reviewers will independently assess all SRs after standardised training, and any disagreements will be resolved by a third reviewer. Assess the study quality and evidence strength Assessment of evidence strength Given the need for caution in medication use during pregnancy, all available evidence (i.e. pooled estimates from MAs) will be considered highly valuable, regardless of statistical significance. The strength of evidence, reflecting its credibility and certainty, will be assessed using GRADE (Grading of Recommendations, Assessment, Development, and Evaluation) guidelines [52] (Fig. 4), ensuring comprehensiveness and comparability. Evidence will be graded as ‘high’, ‘moderate’, ‘low’, or ‘very low’ according to GRADE recommendations. This grading will consider five downgrading factors—risk of bias, inconsistency, indirectness, imprecision, and publication/reporting bias—and two upgrading factors—large effect size and all plausible residual confounding [53]. Additionally, for evidence derived from the synthesis of RCTs and NRSs, which is not explicitly defined in GRADE, we will assign an initial quality rating of ‘high’ for the main analysis. Two independent reviewers will evaluate the evidence following standardised training, and any disagreements will be resolved through discussion. Data synthesis We will narratively synthesise meta-analytic associations from direct comparative analyses, including the exposed anti-infective agents, indications in pregnant women, offspring outcomes, control design, number of participants, number of events, effect metrics, effect estimates, certainty of evidence, and methodological quality. These findings will be summarised in tables and visualised using evidence maps, with separate presentations for anti-infective agents associated with safety concerns and remaining evidence gaps. All data organisation and visualisation will be performed using R 4.4.1. We will report the effect estimates and directions of associations as presented by the original authors. If necessary, effect estimates will be converted to a common metric—for example, transforming log odds ratios to odds ratios—to enhance interpretability. When the control group in a MA consists of alternative anti-infective agents (i.e. active comparators), we will adjust the direction of associations when the control exhibits substantially higher risk than the exposure, ensuring all results consistently reflect which anti-infective agent is associated with higher risk. Given the variability in PICO definitions across MAs—Asen the variability in PICO definitions across n of association ‘Tenofovir Disoproxil Fumarate for prevention and treatment of HIV’ versus ‘Association between macrolides and malformations’)—findings will be standardised and categorised according to the descriptions provided by the authors. Missing data will be reported as stated in each SR, without consulting the primary studies. After identifying and managing overlapping evidence based on PICO combinations, we anticipate that multiple relevant estimates may remain for a given drug–outcome association. This may arise from differences in the indications of the pregnant populations, control group design, types of primary studies incorporated, or limited overlap in primary studies across MAs. If all retained estimates are consistent (e.g. all indicating a significantly increased risk), this consensus will be used to strengthen the robustness of final conclusion. Conversely, if estimates are inconsistent, two scenarios will be distinguished: partial inconsistency (i.e. some results show a significant risk increase while others are non-significant) and conflicting findings (i.e. some results indicate increased risk while others suggest decreased risk). For both scenarios of inconsistency, the affected agents will be flagged with risk indicators. Potential underlying causes (e.g. active comparators or adjusted estimates) will be systematically examined and incorporated into the interpretation of the conclusion. Furthermore, conflicting findings will be specifically prioritised for future investigation. Crucially, all available evidence for the drug–outcome association will be comprehensively reported, regardless of the consistency of the findings. We will prioritise evidence based on specific indications, head-to-head comparisons, or analyses restricted to RCTs, as these designs better mitigate indication bias and provide stronger insights into the causal relationship between prenatal anti-infective exposure and adverse offspring outcomes. Risk signals arising from such evidence will require particular attention. Analysis of subgroups or sensitivity Considering that the route of administration (e.g. oral or vaginal), dosage, or timing (e.g. first, second, or third trimester) of anti-infective exposure during pregnancy may influence offspring outcomes, we will summarise subgroup analyses extracted from the original MAs following the main analyses. This approach will allow us to identify safety signals related to specific anti-infective agents or medication regimens—for example, significant risks associated with first-trimester exposure. When evidence exhibits hierarchical relationships among maternal characteristics (e.g. infectious disease subtypes), drug categories (e.g. macrolides versus azithromycin), and outcome types (e.g. specific BD subtypes), findings will be organised with the highest-level category taking precedence. Interpretations will remain independent; for instance, evidence showing a significant risk for azithromycin will not automatically imply risk for all macrolides, and vice versa. For sensitivity analyses, we will summarise results from the original MAs that differ from the main analysis—for example, analyses excluding primary studies with a high risk of bias—to examine the robustness of the pooled estimates. Discussion This umbrella review will be the first to comprehensively assess the association between prenatal use of anti-infective agents and adverse offspring outcomes. Leveraging a standardised, algorithm-supported process for overlap management, we will systematically and efficiently identify and synthesise evidence from multiple SRs using both qualitative and quantitative approaches. We will apply AMSTAR-2 and GRADE to rigorously evaluate the methodological quality of included SRs and the certainty of the evidence, thereby ensuring a clear and structured presentation of findings. The results are expected to inform clinical decision-making, shape evidence-based policy, and improve the management of infections and anti-infective treatments during pregnancy. Furthermore, by highlighting key evidence gaps, this review will guide future research and address unmet needs in maternal and offspring health. A major challenge in perinatal pharmacology, particularly for assessing offspring safety, stems from indication bias and ethical constraints. Even existing RCTs often lack adequate follow-up to detect long-term risks. Based on prior research, we anticipate that few NRSs and even fewer SRs will adequately address indication bias and other confounders. We therefore will prioritise evidence with a lower risk of indication bias, such as MAs with clearly defined maternal indications, head-to-head comparisons, or RCTs. To ensure comprehensive overlap handling, we will assess consistency across MAs based on study populations, anti-infective agents, comparators, and outcomes, moving beyond reliance on drug–outcome matching at the SR level. As a result, multiple evidence sources may remain for the same drug–outcome association, producing potentially inconsistent findings. Our methodology will explicitly address this by flagging agents with conflicting evidence as ‘risk drugs’ and examining potential reasons for these discrepancies. A key innovation is our three-stage data extraction strategy, designed for efficient and clinically coherent overlap management. Unlike conventional URs that assess overlap solely at the SR level or rely on data-driven checks requiring full primary-study extraction, our method accounts for the complexity of multiple MAs within SR-MAs and will identify overlap at the MA level using PICO definitions. This will enhance both the transparency and interpretability of overlap decisions. Crucially, our strategy also will minimise unnecessary data extraction. Existing tools (e.g. GROOVE and ccaR) require extracting detailed primary-study information from all SRs or MAs, resulting in labour-intensive processes [38, 39]. By contrast, we will first extract only PICO definitions, identify overlaps, and then extract primary-study and quantitative data only for the overlapping or retained MAs. Moreover, our approach will support transparent and automated implementation. The developed R functions will streamline overlap identification, calculation of the degree of overlap, and application of decision rules (e.g. excluding evidence with substantial overlap > 5%). This automation will ensure consistency and scalability across numerous PICO combinations. To facilitate adoption and reproducibility, we have provided sample R code and an example dataset. Several limitations warrant attention. First, as secondary research, our findings will depend on available SRs and will inherit their inherent limitations. The reliability of our results may be affected by the methodological and reporting quality of included SRs, and differing synthesis methods across reviews may lead to divergent conclusions. In complex evidence landscapes with numerous SRs, re-analysing primary studies is challenging and may introduce information bias. Limited access to individual participant data restricts individual-level subgroup analyses, potentially obscuring safety signals from variations in medication regimens. To minimise evidence loss, our protocol will prioritise comprehensiveness by extracting both main and subgroup analyses from existing reviews. We will transparently report any limitations stemming from the quality of included SRs. Second, while the exclusion of non-indexed grey literature (e.g. via OpenGrey) may limit the detection of emerging safety signals, this is justified because our unit of analysis is SRs or MAs, not the primary reports that constitute most grey literature. Furthermore, due to its lack of unified indexing and global coverage, systematic retrieval would be infeasible. The likely contribution, therefore, does not justify the marked increase in workload alongside our comprehensive search of six major databases. Third, we will use GRADE—which already incorporates risk-of-bias domains—to evaluate certainty at the MA level, and AMSTAR-2 to appraise methodological quality at the SR level. Consequently, we will not apply the Risk Of Bias In Systematic reviews (ROBIS) tool. Empirical studies show substantial overlap between AMSTAR-2 and ROBIS (> 70%) and indicate that both tools face similar limitations in safety-focused reviews dominated by non-randomised evidence [54–56]. Moreover, ROBIS requires substantial domain expertise and frequent consensus discussions, which would unduly increase the operational burden in this complex context. This protocol offers a methodological model for future umbrella reviews in other areas of perinatal medicine and complex evidence fields. Key transferable elements include a structured, three-stage approach to managing overlap at the MA level; a systematic strategy for accounting for variations in PICO definitions across included reviews; and a clear plan for mitigating indication bias. Researchers can adopt these principles to enhance the efficiency and completeness of evidence synthesis, supporting robust clinical and policy decision-making in related fields. In conclusion, this umbrella review, grounded in a well-structured protocol, aims to provide valuable insights into ongoing discussions regarding the safety of prenatal anti-infective agent use. We will strictly adhere to our protocol, documenting any amendments on the PROSPERO platform and discussing them in the publication. Ultimately, we hope that this protocol will serve as a useful reference for future researchers conducting umbrella reviews. Supplementary Information Additional file 1. PRISMA-P checklist. Additional file 2. Search strategy across six databases. Acknowledgements Not applicable. Abbreviations A MeaSurement Tool to Assess systematic Reviews, version 2 Birth defects Corrected covered area Excerpta Medica Database Grading of Recommendations Assessment, Development and Evaluation Human immunodeficiency virus Individual participant data Medical Literature Analysis and Retrieval System Online Meta-analyse Medical Subject Headings Non-randomised studies Population, Interventions, Controls and Outcomes Preferred Reporting Items for Overviews of Reviews Preferred reporting items for systematic review and meta-analysis protocol International Prospective Register of Systematic Reviews Premature birth Public Library of Science Biology Randomised controlled trials Risk Of Bias In Systematic reviews Systematic reviews Systematic reviews with meta-analyses Systematic reviews without meta-analyses Umbrella review World Health Organization Authors’ contributions Research conceptualization was contributed by JT, YX, LG, and XS. Research methodology was designed by JT, WW, LG, JC, CC, and YR. First draft of the manuscript was written by WW, further refined by JT, LG, LT, and XS. All authors approved the final manuscript. Funding This protocol was supported by National Natural Science Foundation of China (72174132); National Natural Science Foundation of China (82474335); The Key Research and Development Project of Sichuan Provincial Department of Science and Technology (2024YFFK0152). None of the funders had any role in the study. Data availability The example data for process overlapping evidence by an R-based algorithm can available in the GitHub repository (https://github.com/WANGDALINGLING/OP-UR.git). Declarations Ethics approval and consent to participate Ethical review is not required for umbrella review. Consent for publication Not applicable. Competing interests LG is an associate editor of Systematic Reviews. The remaining authors declare that they have no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Wenling Wang and Long Ge contributed equally to this work. Contributor Information Xin Sun, Email: sunxin@wchscu.cn. Jing Tan, Email: tanjing84@outlook.com. References Associated Data Supplementary Materials Additional file 1. PRISMA-P checklist. Additional file 2. Search strategy across six databases. Data Availability Statement The example data for process overlapping evidence by an R-based algorithm can available in the GitHub repository (https://github.com/WANGDALINGLING/OP-UR.git).