Microbiome dysbiosis and endometriosis: a systematic scoping review of current literature and knowledge gaps

Nearly a century after Sampson proposed the retrograde menstruation theory, accumulating evidence suggests that no single mechanism can fully explain the diverse clinical manifestations associated with endometrial tissue at ectopic sites. Many alternative hypotheses have been proposed to explain the development of endometriosis, yet its precise etiology remains unknown ( Abbott et al. , 2024 ). Interestingly, recent findings have implicated microbiota dysfunctions, particularly bacterial infections, as potential causal factors. Thus, this research area has become particularly active. The link between endometriosis and microbiome alterations is supported by a few studies conducted on murine models of the disease ( Yuan et al. , 2018 ; Chadchan et al. , 2023 ; Hu et al. , 2023 ; Muraoka et al. , 2023 ; Wei et al. , 2023 ). Among these, the most interesting breakthrough was reported by Muraoka et al. (2023) , who showed that uterine tissue fragments infected with Fusobacterium nucleatum , when injected into recipient mice, led to larger endometriotic lesions characterized by increased M2 macrophage infiltration, elevated expression of transforming growth factor (TGF)-β1, and a greater number of transgelin-positive myofibroblasts. Importantly, antibiotic treatment reduced lesion weight, underscoring a potential therapeutic avenue. Earlier murine studies linked endometriosis to alterations in the gut microbiome rather than the uterine microbiome ( Yuan et al. , 2018 ; Chadchan et al. , 2023 ). Similarly, treatment with broad-spectrum antibiotics reduced the risk of endometriosis also in these models ( Chadchan et al. , 2019 ). In human studies, four systematic reviews have synthesized evidence regarding gut, endometrial, and vaginal microbiome composition in patients with endometriosis ( Leonardi et al. , 2020 ; D’Alterio et al. , 2021 ; Colonetti et al. , 2023 ; Weber et al. , 2024 ). These reviews, however, highlighted inconsistencies across studies, with no definitive trends indicating whether bacterial changes in patients with endometriosis were predominantly beneficial or harmful. Significant knowledge gaps persist in this field. Understanding whether and how microbiome dysbiosis may influence the development and/or progression of endometriosis could guide or limit future interventions targeting bacterial colonization. To address these gaps, we chose to conduct a scoping review to systematically evaluate the existing literature, assess research methodologies with a particular emphasis on sampling and analytic techniques, and explore potential microbiome-related exposures in affected patients. This way, we have systematically and critically mapped the existing literature addressing the following primary questions: Microbiome characterization: What data are available concerning gut and reproductive tract microbiomes in patients with endometriosis? Methodological assessment: What are the methodological design methods, sequencing platform, analytical pipelines employed in microbiome studies on endometriosis?

This scoping review was conducted in accordance with the PRISMA-ScR guidelines (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) ( Tricco et al ., 2018 ), which provides a widely accepted framework to ensure transparency and methodological rigor in exploratory reviews. The methodological approach was based on the foundational framework originally proposed by Arksey and O’Malley (2005) and later refined by Levac et al . (2010 ) and the Joanna Briggs Institute Reviewers’ Manual ( Aromataris et al . 2024 ). Specifically, the review process adhered to five key stages: identifying the research question, searching the literature, selecting relevant studies, charting the data, and synthesizing and reporting the results. These established protocols helped maintain consistency and reproducibility throughout the review process. The scoping review protocol was recorded a priori and is available at OSF registries (Registration doi: 10.17605/OSF.IO/X6HBT; https://osf.io/registries ). The PubMed, EMBASE, and Web of Science databases were searched to identify studies on endometriosis microbiome characterization, using a combination of MeSH terms (search string available in Supplementary File S1 ). The scoping review included all studies published from January 2016 to November 2024. Papers published before 2016 were excluded because sequencing technologies were in developmental stages and had not yet been widely adopted. Starting in 2016, sequencing methods, particularly next-generation sequencing, became more commonly applied, improving consistency and reproducibility in microbiome analyses due to advancements in depth, accuracy, and standardization ( Wen et al ., 2017 ). The initial literature search was conducted on 5 February 2024, with the most recent update performed on 11 November 2024. Studies were included if they met the following criteria: the study population consisted of women diagnosed with endometriosis and an appropriate control group; the microbiome of the reproductive tract or gastrointestinal system was analyzed using molecular techniques; and the study followed an experimental design and was published in English. The screening process was conducted in two phases. First, titles and abstracts were assessed to exclude clearly irrelevant publications. Subsequently, full-text reviews were carried out to confirm final eligibility. Both stages were independently performed by two reviewers (G.D.S. and D.D.), and any disagreements were resolved by discussion with a third reviewer (M.C.) to ensure objectivity and consensus. The final list of included studies was then reviewed and discussed by the full research team to support the technical and methodological integrity of the selection process. A standardized data-charting form was developed to guide the extraction of key variables. Three co-authors (D.D., G.D.S., and M.C.) independently extracted and charted data from each included study. Discrepancies were resolved through discussion and consensus. The form was iteratively refined as needed to ensure consistency and completeness in data collection. First, the authors collected a combination of qualitative or quantitative data from the final list of publications in a ‘s-charting form’ ( Arksey and O’Malley, 2005 ), reporting: (i) study characteristics, including study design and sample size of study groups; (ii) sample location and (iii) sampling method; (iv) population studies characteristics; (v) inclusion and exclusion criteria; (vi) endometriosis stage/phenotype and diagnostic methods; (vii) technical characteristics of microbiome analysis; (viii) technical characteristics of bioinformatic analysis; and (ix) descriptive results, including alpha and beta diversity (a glossary of microbiome terminology is available in Supplementary Table S1 ). To ensure a structured evaluation of the quality of the studies included in this review, the widely recognized Newcastle–Ottawa Scale (NOS) was employed. This tool was used exclusively to assess the quality of patient selection and group comparability, and not to evaluate methodological or bioinformatic procedures such as sequencing or data analysis. The NOS consists of an 8-item checklist that focuses on three key domains: the selection of study participants, the comparability of groups, and the ascertainment of exposure or outcomes. In accordance with established guidance ( Stang, 2010 ), two independent reviewers applied the NOS criteria to each included study to ensure consistency and minimize subjective bias in the assessment.

The study selection process is summarized in the PRISMA flow diagram ( Fig. 1 ). A total of 2182 studies were initially retrieved from three databases: PubMed, EMBASE, and Web of Science. After removing 659 duplicates and excluding 509 studies published before 2016, 1014 records remained for title and abstract screening. Of these, 181 studies were selected for full-text review. Following the full-text assessment, 145 studies were excluded: 93 did not include a comparison between women with and without endometriosis, and 52 did not enable a molecular-level evaluation of the microbiome. Ultimately, 36 studies met the inclusion criteria and were included in the review. A summary of these studies is provided in Table 1 . PRISMA flow diagram illustrating the study selection process . Overview of the articles included in the review. Eutopic endometrium. Peritoneal fluid. Follicular or luteal. ( P =NS) Stool. Stool; Vaginal fluid. Vaginal fluid. Diverse phases ( P =NS) No, and the use was in different rates between groups ( P =NR) Stool; Vaginal fluid; Oropharyngeal fluid. Vaginal fluid. Stool. No, but the frequency of antibiotics consumption in the last year was not statistically different between groups. ( P =NS) Anal fluid; Vaginal fluid. Stool; Vaginal fluid; Uterine fluid; Eutopic endometrium; Oropharyngeal fluid. No ( P =NR) Peritoneal fluid; Endometrial fluid. Endometriotic tissue. Mainly follicular (>90%) ( P =NR) Stool. Endometriotic tissue; Stool. Cervical fluid; Vaginal fluid. Yes (in the last 30 days) Vaginal fluid; Eutopic endometrium; Endometriotic tissue. Follicular or luteal. ( P =NR) Yes (in the last 30 days) Stool. Vaginal fluid. Yes (in the last 30 days) Cervical fluid. Peritoneal fluid. Follicular or luteal. ( P =NS) Stool; Cervical fluid; Peritoneal fluid. Follicular or luteal. ( P =NR) Anal fluid; Vaginal fluid. Eutopic endometrium. Follicular, luteal, or menstrual. ( P =NS) Eutopic endometrium. Diverse phases in similar rates between groups ( P =NR) Vaginal fluid. Diverse phases ( P =NS) Peritoneal fluid’s EVs. Stool. Stool. OMA, SUP, and DIE No, and the use of hormonal treatments was significantly higher in the endometriosis groups. P  < 0.001 No, but its use was not statistically different between groups. ( P =NS) Vaginal fluid; Cervical mucus; Peritoneal fluid; Endometrial fluid. Vaginal fluid; Eutopic endometrium; Endometriotic tissue. Anal fluid; Vaginal fluid. OMA, SUP, and DIE Cervical mucus. Follicular or luteal, in the same proportion between groups. ( P =NS) Stool; Cervical fluid; Vaginal fluid. OMA, SUP, and DIE Follicular or luteal. ( P =NS) Peritoneal fluid. Cervical fluid; Peritoneal fluid; Endometriotic tissue; Peritoneal tissue. OMA, SUP, and DIE No, the use of hormonal treatments was higher in the endometriosis group ( P =NR) Follicular or luteal. ( P =NS) Endometrial fluid (mixed with cells); Ovarian cyst fluid. Diverse phases in different rates between groups ( P =NR) CCP, chronic pelvic pain; DIE, deep infiltrating endometriosis; Evs, extracellular vesicles; NR, not reported; ns, non-significant; OMA, ovarian endometriosis; rASRM, revised American Society of Reproductive Medicine classification; SUP, superficial endometriosis. Of the 36 studies included in this review ( Table 1 ), the types of biological samples analyzed were distributed as follows: stool/anal fluid: 15 studies (41.7%) ( Ata et al. , 2019 ; Perrotta et al. , 2020 ; Huang et al. , 2021 ; Le et al. , 2021 ; Shan et al. , 2021 ; Svensson et al. , 2021 ; Hu et al. , 2023 ; Pai et al. , 2023 ; Wei et al. , 2023 ; Do et al. , 2024 ; Guo et al. , 2024 ; Jimenez et al. , 2024 ; Marcos et al. , 2024 ; Pérez-Prieto et al. , 2024 ; Hicks et al. , 2025 ); vaginal fluid: 15 studies (41.7%) ( Ata et al. , 2019 ; Hernandes et al. , 2020 ; Perrotta et al. , 2020 ; Wei et al. , 2020 ; Chao et al. , 2021 ; Le et al. , 2021 ; Lu et al. , 2022 ; Muraoka et al. , 2023 ; Yang et al. , 2023 ; Do et al. , 2024 ; Jimenez et al. , 2024 ; MacSharry et al. , 2024 ; Marcos et al. , 2024 ; Sessa et al. , 2024 ; Hicks et al. , 2025 ); cervical mucus/fluid: seven studies (19.4%) ( Campos et al. , 2018 ; Akiyama et al. , 2019 ; Ata et al. , 2019 ; Wei et al. , 2020 ; Huang et al. , 2021 ; Chang et al. , 2022 ; Yang et al. , 2023 ); peritoneal fluid: eight studies (22.2%) ( Campos et al. , 2018 ; Wang et al. , 2018 ; Wei et al. , 2020 ; Huang et al. , 2021 ; Lee et al. , 2021 ; Yuan et al. , 2022 ; Zhu et al. , 2024 ; Malvezzi et al. , 2025 ); uterine/endometrial fluid: four studies (11.1%) ( Khan et al. , 2016 ; Wei et al. , 2020 ; Marcos et al. , 2024 ; Zhu et al. , 2024 ); ovarian cyst fluid: one study (2.7%) ( Khan et al. , 2016 ); oropharyngeal fluid: two studies (5.6%) ( Marcos et al. , 2024 ; Hicks et al. , 2025 ); eutopic endometrium: seven studies (19.4%) ( Hernandes et al. , 2020 ; Khan et al. , 2021 ; Wessels et al. , 2021 ; Muraoka et al. , 2023 ; Marcos et al. , 2024 ; Guo et al. , 2025 ); endometriotic tissue: five studies (13.9%) ( Campos et al. , 2018 ; Hernandes et al. , 2020 ; Hu et al. , 2023 ; Muraoka et al. , 2023 ; Chen et al. , 2024 ); The majority of studies were conducted in East Asian populations: 14 in China ( Wang et al. , 2018 ; Wei et al. , 2020 , 2023 ; Chao et al. , 2021 ; Huang et al. , 2021 ; Shan et al. , 2021 ; Lu et al. , 2022 ; Yuan et al. , 2022 ; Hu et al. , 2023 ; Yang et al. , 2023 ; Chen et al. , 2024 ; Guo et al. , 2024 , 2025 ; Zhu et al. , 2024 ), four in Japan ( Khan et al. , 2016 , 2021 ; Akiyama et al. , 2019 ; Muraoka et al. , 2023 ), two in Taiwan ( Chang et al. , 2022 ; Pai et al. , 2023 ), and one in South Korea ( Lee et al. , 2021 ). European-based studies included one in Ireland ( MacSharry et al. , 2024 ), one in Italy ( Sessa et al. , 2024 ), one in Sweden ( Svensson et al. , 2021 ), one in Turkey ( Ata et al. , 2019 ), and two in Spain, one involving a Spanish population ( Marcos et al. , 2024 ) and the other an Estonian population ( Pérez-Prieto et al. , 2024 ). Eight studies were from the Americas: four from Brazil ( Campos et al. , 2018 ; Hernandes et al. , 2020 ; Perrotta et al. , 2020 ; Malvezzi et al. , 2025 ), three from the USA ( Le et al. , 2021 ; Do et al. , 2024 ; Jimenez et al. , 2024 ), and one from Canada ( Wessels et al. , 2021 ). Only one was conducted in Australia ( Hicks et al. , 2025 ; Table 1 ). This review included only original studies that evaluated the gut or reproductive tract microbiome in women with endometriosis (cases) compared to women without the condition (controls). However, only a small number of included studies explicitly reported using a case-control design ( Hernandes et al. , 2020 ; Pai et al. , 2023 ; Pérez-Prieto et al. , 2024 ; Malvezzi et al. , 2025 ). Four were reported as cross-sectional studies ( Perrotta et al. , 2020 ; Chao et al. , 2021 ; Wessels et al. , 2021 ; Sessa et al. , 2024 ), and three as cohort studies ( Ata et al. , 2019 ; Marcos et al. , 2024 ; Hicks et al. , 2025 ). Most studies, however, did not clearly report their study design ( Khan et al. , 2016 , 2021 ; Campos et al. , 2018 ; Wang et al. , 2018 ; Akiyama et al. , 2019 ; Wei et al. , 2020 , 2023 ; Huang et al. , 2021 ; Le et al. , 2021 ; Lee et al. , 2021 ; Shan et al. , 2021 ; Svensson et al. , 2021 ; Chang et al. , 2022 ; Lu et al. , 2022 ; Yuan et al. , 2022 ; Hu et al. , 2023 ; Muraoka et al. , 2023 ; Yang et al. , 2023 ; Chen et al. , 2024 ; Do et al. , 2024 ; Guo et al. , 2024 , 2025 ; Jimenez et al. , 2024 ; MacSharry et al. , 2024 ; Zhu et al. , 2024 ). Sample sizes varied significantly across studies, ranging from as few as 21 participants ( Hernandes et al. , 2020 ; Wessels et al. , 2021 ; Marcos et al. , 2024 ) to up to 1000 participants ( Pérez-Prieto et al. , 2024 ). Notably, 30 out of 36 studies included 100 participants or fewer. The composition of control groups was also highly heterogeneous. Only 10 studies ( Ata et al. , 2019 ; Chao et al. , 2021 ; Shan et al. , 2021 ; Svensson et al. , 2021 ; Chang et al. , 2022 ; Lu et al. , 2022 ; Wei et al. , 2023 ; Yang et al. , 2023 ; Guo et al. , 2024 ; Hicks et al. , 2025 ) included at least a subgroup of healthy women as controls. In contrast, the remaining studies used control groups composed of women with other gynecological conditions unrelated to endometriosis, such as infertility, uterine fibroids, tubal obstruction, leiomyomas, ovarian cysts, and chronic pelvic pain. Regarding the type of endometriosis evaluated, five studies did not specify either the disease phenotype or the stages ( Chao et al. , 2021 ; Guo et al. , 2024 , 2025 ; Marcos et al. , 2024 ; Pérez-Prieto et al. , 2024 ). In contrast, only 11 studies provided complete information, reporting both the phenotype and stage of endometriosis ( Campos et al. , 2018 ; Ata et al. , 2019 ; Perrotta et al. , 2020 ; Khan et al. , 2021 ; Lee et al. , 2021 ; Yuan et al. , 2022 ; Hu et al. , 2023 ; Wei et al. , 2023 ; Yang et al. , 2023 ; Chen et al. , 2024 ; MacSharry et al. , 2024 ). The methods used to diagnose endometriosis varied across the included studies. Most studies included only participants diagnosed through surgery followed by histological confirmation ( Khan et al. , 2016 , 2021 ; Campos et al. , 2018 ; Akiyama et al. , 2019 ; Ata et al. , 2019 ; Hernandes et al. , 2020 ; Wei et al. , 2020 , 2023 ; Huang et al. , 2021 ; Lee et al. , 2021 ; Shan et al. , 2021 ; Wessels et al. , 2021 ; Chang et al. , 2022 ; Lu et al. , 2022 ; Yuan et al. , 2022 ; Hu et al. , 2023 ; Pai et al. , 2023 ; Yang et al. , 2023 ; Chen et al. , 2024 ; Jimenez et al. , 2024 ; Zhu et al. , 2024 ; Guo et al. , 2025 ; Hicks et al. , 2025 ; Malvezzi et al. , 2025 ). Several important confounders relevant to microbiome research were considered during this review, including the use of hormonal treatments, antibiotic usage, special dietary habits, and the menstrual phase at the time of sample collection ( Table 1 ). Notably, eight studies ( Campos et al. , 2018 ; Hernandes et al. , 2020 ; Chang et al. , 2022 ; Pai et al. , 2023 ; Guo et al. , 2024 ; Marcos et al. , 2024 ; Hicks et al. , 2025 ; Malvezzi et al. , 2025 ) did not report whether hormonal treatment was used as an exclusion criterion or whether its use differed between groups. Similarly, nine studies either failed to report antibiotic use ( Khan et al. , 2016 ; Wang et al. , 2018 ; Le et al. , 2021 ; Wessels et al. , 2021 ; Chang et al. , 2022 ) or did not clarify whether usage differed across groups ( Chen et al. , 2024 ; MacSharry et al. , 2024 ; Hicks et al. , 2025 ; Malvezzi et al. , 2025 ). Only three studies addressed dietary factors ( Table 1 ): two excluded participants with specific eating habits ( Pai et al. , 2023 ; Wei et al. , 2023 ), while one prescribed a standardized diet prior to sampling ( Shan et al. , 2021 ). In addition, several studies ( Hernandes et al. , 2020 ; Le et al. , 2021 ; Svensson et al. , 2021 ; Chang et al. , 2022 ; Lu et al. , 2022 ; Pai et al. , 2023 ; Do et al. , 2024 ; Guo et al. , 2024 , 2025 ; Jimenez et al. , 2024 ; Pérez-Prieto et al. , 2024 ; Zhu et al. , 2024 ; Hicks et al. , 2025 ) did not report the menstrual cycle phase during sample collection, or failed to clarify whether distribution across phases differed between groups ( Khan et al. , 2016 , 2021 ; Huang et al. , 2021 ; Muraoka et al. , 2023 ; Chen et al. , 2024 ). Specific differences among study groups will be addressed in the results for each sample type. Technical characteristics of the 15 studies ( Ata et al. , 2019 ; Perrotta et al. , 2020 ; Huang et al. , 2021 ; Le et al. , 2021 ; Shan et al. , 2021 ; Svensson et al. , 2021 ; Hu et al. , 2023 ; Pai et al. , 2023 ; Wei et al. , 2023 ; Do et al. , 2024 ; Guo et al. , 2024 ; Jimenez et al. , 2024 ; Marcos et al. , 2024 ; Pérez-Prieto et al. , 2024 ; Hicks et al. , 2025 ) providing comprehensive taxonomic evaluations of stool microbiome are detailed in Table 2 . Technical characteristics of microbiome analyses across stool/anal fluid samples in the included studies. Women with endometriosis vs healthy controls 18 vs 18 32.23 ± 5.25 vs 31.40 ± 3.75 § ( P =NR) 22.53 ± 3.15 vs 21.14 ± 1.94 § ( P =NR) TIANamp Stool DNA Kit—TianGen (For extraction of high-quality genomic DNA from various stool samples) Women with endometriosis vs women without endometriosis 33 vs 15 31.6 ± 0.8 vs 33.3 ± 1.8 § ( P =NS) 28.3 ± 1.3 vs 29.5 ± 1.9 § ( P =NS) DNeasy PowerSoil Pro Kit—Qiagen (For the isolation of microbial genomic DNA from all soil types) Women with endometriosis vs (women with other gynecological diseases vs healthy controls) 21 vs (24 vs 19) 35.9 ± 8.1 vs (35.8 ± 7.3 vs 31.5 ± 3.5) § ( P =NS) PSP ® Spin Stool DNA Basic Kit—Invitek (For isolation of bacterial DNA and host DNA from stool samples) Women with endometriosis vs women without endometriosis 136 vs 864 50.0 [40.8–57.9] vs 45.0 [36.0–54.0] † ( P  = 0.005) 25.1 [22.2–29.5] vs 24.2 [21.6; 28.6] † ( P =NS) QIAamp Fast DNA Stool Mini Kit—Qiagen (For isolation of gDNA from stool samples) Women with chronic pelvic pain with endometriosis vs women with other gynecological disorders without chronic pelvic pain 35 vs 15 34.7 ± 8.8 vs 40.6 ± 8.2 § ( P =NS) Reported by three ranges of age ( P =NS). DNaeasy PowerSoil Pro Kit—Qiagen (For the isolation of microbial genomic DNA from all soil types) Infertile women with endometriosis vs women with other infertility-related conditions 8 vs 13 42.7 ± 5.5 vs 39.4 ± 3.7 § ( P =NS) QIAamp Fast DNA Tissue Kit—Qiagen (For rapid isolation of genomic DNA from solid tissue samples) Women with endometriosis vs women without endometriosis 27 vs 24 38.1 ± 1.0 vs 37.7 ± 1.3 § ( P =NS) 24.04 ± 0.87 vs 21.77 ± 0.73 § ( P  = 0.051) QIAamp PowerFecal Pro DNA Kits—Qiagen (For the isolation of microbial DNA from stool and gut samples) Women with endometriotic cysts vs healthy controls 14 vs 24 30.6 [29.3–32.0] vs 29.4 [28.4–30.4] † ( P =NS) 20.29 [19.12–21.46] vs 22.75 [21.23–24.27] † ( P  = 0.01) Cetyltrimethylammonium bromide (CTAB) (Generic method for isolating genomic DNA from different tissues) Infertile women with endometriosis vs infertile and healthy women without endometriosis 35 vs (8 and 22) 32.6 ± 5.7 vs 30.2 ± 5.6 § ( P =NS) 20.52 ± 2.02 vs 19.78 ± 1.55 § ( P =NS) E.Z.N.A. ® Soil DNA Kit—Omega Bio-Tek (For isolation of DNA from soil samples) Women with endometriosis vs healthy controls 21 vs 20 38.3 ± 7.88 vs 34.0 ± 10.8 § ( P =NS) 21.5 ± 2.79 vs 24.3 ± 8.16 § ( P =NS) Quick-RNA Fecal/Soil Microbe Microprep Kit—Zymo Research (For extract of RNA from various soil, fecal, and water samples) Women with endometriosis vs women without endometriosis 20 vs 9 32.5 ± 1.1 vs 32.6 ± 2.0 § ( P =NS) 26.5 ± 1.5 vs 28.1 ± 2.4 § ( P =NS) PowerMag Soil DNA Isolation Kit—MoBio (For isolation of microbial DNA from all types of soil) Women with endometriosis vs healthy controls 12 vs 12 32 ± 2 vs 32 ± 3 § ( P =NS) E.Z.N.A. ® Soil DNA Kit—Omega Bio-Tek (For isolation of DNA from soil samples) Women with endometriosis vs healthy controls 66 vs 198 37.8 [32.8–43.3] vs 37.0 [32.0–44.0] † ( P =NS) 37.8 [32.8–43.3] vs 24.7 [22.1–27.5] † ( P =NS) QIAamp Fast DNA Stool Mini Kit—Qiagen (For isolation of gDNA from stool samples) Women with endometriosis vs women without endometriosis 35 vs 24 34.9 ± 6.8 vs 35.25 ± 6.9 § ( P =NS) 24.8 ± 4.5 vs 24.3 ± 2.7 § ( P =NS) PowerMag Soil DNA Isolation Kit—MoBio (For isolation of microbial DNA from all types of soil) Women with endometriosis vs healthy controls 14 vs 14 28.5 [26.0–31.3] vs 27.5 [25.8–30.0] † ( P =NS) 23.0 [21.0–24.3] vs 21.0 [20.1–24.2] † ( P =NS) QIAamp Fast DNA Stool Mini Kit—Qiagen (For isolation of gDNA from stool samples) Data reported as reported by the original papers, unless otherwise stated. Bp, base pairs; NR, not reported; NS, non-significant; WGS, whole-genome sequencing. Data are expressed as mean±SD. Data are expressed as median [25th–75th percentile]. Sample sizes ranged from n = 8 to n = 136 for cases and n = 9 to n = 864 for controls. Only six studies ( Ata et al. , 2019 ; Shan et al. , 2021 ; Svensson et al. , 2021 ; Hu et al. , 2023 ; Wei et al. , 2023 ; Hicks et al. , 2025 ) recruited healthy women as controls. Most studies reported comparable age between groups; however, the largest one ( Pérez-Prieto et al. , 2024 ) showed a significant age difference ( P  = 0.005), and one study ( Guo et al. , 2025 ) did not report this information. BMI was significantly different in Hu et al. (2023 ; P  = 0.01) and was unreported in four studies ( Shan et al. , 2021 ; Guo et al. , 2024 ; Marcos et al. , 2024 ; Hicks et al. , 2025 ). All others reported BMI comparability between groups. Endometriosis diagnosis was surgical in nearly all studies, except for three that also included diagnoses based on imaging ( Perrotta et al. , 2020 ; Guo et al. , 2024 ; Marcos et al. , 2024 ), generally indicating a focus on moderate to severe forms. Only six studies excluded participants on hormonal therapy ( Ata et al. , 2019 ; Perrotta et al. , 2020 ; Huang et al. , 2021 ; Shan et al. , 2021 ; Hu et al. , 2023 ; Wei et al. , 2023 ) or on antibiotics therapy ( Ata et al. , 2019 ; Perrotta et al. , 2020 ; Huang et al. , 2021 ; Shan et al. , 2021 ; Hu et al. , 2023 ; Wei et al. , 2023 ). Menstrual cycle phase was unreported in eight studies ( Le et al. , 2021 ; Svensson et al. , 2021 ; Pai et al. , 2023 ; Do et al. , 2024 ; Guo et al. , 2024 ; Jimenez et al. , 2024 ; Pérez-Prieto et al. , 2024 ; Hicks et al. , 2025 ) and standardized to early follicular only in four ( Perrotta et al. , 2020 ; Shan et al. , 2021 ; Hu et al. , 2023 ; Wei et al. , 2023 ). One study ( Huang et al. , 2021 ) reported that the collection phase varied but whether there was a statistical difference between groups was not reported. Despite the known influence of diet on gut microbiota ( Flint et al. , 2015 ), only two studies ( Pai et al. , 2023 ; Wei et al. , 2023 ) excluded participants with specific dietary habits, and only one ( Shan et al. , 2021 ) requested participants to follow a specific diet for 3 days before sample collection ( Table 1 ). Microbiome analysis methods are summarized in Table 2 . Notably, several studies used DNA extraction kits designed for soil ( Perrotta et al. , 2020 ; Le et al. , 2021 ; Shan et al. , 2021 ; Wei et al. , 2023 ; Do et al. , 2024 ; Jimenez et al. , 2024 ) or for tissue samples ( Hu et al. , 2023 ; Marcos et al. , 2024 ) rather than stool samples, potentially impacting microbial yield and composition. Nearly all studies employed 16S rRNA sequencing on Illumina or Ion Torrent platforms, though the hypervariable regions targeted varied across studies. In contrast, Pérez-Prieto et al. (2024) employed shotgun metagenomic paired-end sequencing, allowing for broader and deeper taxonomic resolution through whole-genome profiling. Bioinformatics pipelines, including sequence filtering, chimera removal, operational taxonomic unit (OTU) clustering, and taxonomic assignment, differed widely across studies. A complete overview of the specific pipelines used is provided in Supplementary Table S2 . All studies, except one ( Perrotta et al. , 2020 ), assessed alpha diversity, within-sample bacterial diversity, and beta diversity, between-sample bacterial composition differences ( Fig. 2 ). Significant differences in alpha diversity were observed in six studies ( Huang et al. , 2021 ; Svensson et al. , 2021 ; Hu et al. , 2023 ; Do et al. 2024 ; Hicks et al. 2025; Guo et al. , 2024 ), and beta diversity differences were reported in another six ( Huang et al. , 2021 ; Shan et al. , 2021 ; Svensson et al. , 2021 ; Do et al. 2024 ; Hicks et al. 2025 ; Guo et al. , 2024 ). However, each of these studies had at least one major methodologic or demographic variable unreported or significantly different between groups, limiting interpretation. Alpha diversity (within-sample bacterial diversity) and beta diversity (between-sample bacterial composition differences) comparisons between women with and without endometriosis across studies. Red, diversity reported as statistically significant; green, diversity reported as not statistically significant; yellow, mixed findings, with some analyses showing statistical significance and others not; white, diversity not reported in the study. Among the genera reported as significantly differing between women with and without endometriosis, only six genera were identified in more than one study, which were Eubacterium dolichum ( Huang et al. , 2021 ; Shan et al. , 2021 ), Haemophilus sp. ( Marcos et al. , 2024 ; Hicks et al. , 2025 ), Phascolarctobacterium sp. ( Jimenez et al. , 2024 ; Hicks et al. , 2025 ), Prevotella sp. ( Shan et al. , 2021 ; Hu et al. , 2023 ) with increased abundance in the endometriosis groups while Fusicatenibacter sp. ( Wei et al. , 2023 ; Guo et al. , 2024 ), and Lachnospira sp. ( Shan et al. , 2021 ; Svensson et al. , 2021 ; Guo et al. , 2024 ; Hicks et al. , 2025 ) were identified as being more abundant in the control groups. Interestingly, several other genera were reported as differentially abundant between groups but inconsistently since they were found to be increased in the endometriosis group in some studies and in the control group in others, indicating lack of consensus and highlighting variability in study design or populations ( Bacteroides sp. ( Huang et al. , 2021 ; Svensson et al. , 2021 ; Hu et al. , 2023 ; Do et al. , 2024 ; Guo et al. , 2024 ), Bifidobacterium sp. ( Shan et al. , 2021 ; Hu et al. , 2023 ), Blautia sp. ( Huang et al. , 2021 ; Le et al. , 2021 ; Shan et al. , 2021 ; Guo et al. , 2024 ), Coprococccus sp. ( Shan et al. , 2021 ; Svensson et al. , 2021 ), Dialister sp. ( Le et al. , 2021 ; Wei et al. , 2023 ; Guo et al. , 2024 ), Dorea sp. ( Huang et al. , 2021 ; Shan et al. , 2021 ), Escherichia sp. ( Ata et al. , 2019 ; Hu et al. , 2023 ), Eubacterium sp. ( Wei et al. , 2023 ; Jimenez et al. , 2024 ; Hicks et al. , 2025 ), Lactobacillus sp. ( Jimenez et al. , 2024 ; Hicks et al. , 2025 ), Paraburkholderia sp. ( Wei et al. , 2023 ; Guo et al. , 2024 ), Ruminococcus sp. ( Ata et al. , 2019 ; Huang et al. , 2021 ; Guo et al. , 2024 ; Jimenez et al. , 2024 ), and Senegalimassilia sp. ( Ata et al. , 2019 ; Jimenez et al. , 2024 )) ( Fig. 3 ). Bacterial genera identified across stool/anal fluid samples in the included studies. E, genus’s abundance increased in endometriosis; C, genus’s abundance increased in controls; mid-pink, increased in endometriosis in two studies; light pink, increased in endometriosis in one study; grey, inconsistent findings across studies; light blue, decreased in endometriosis in one study; mid-blue, decreased in endometriosis in two studies; dark blue, decreased in endometriosis in ≥3 studies. At higher taxonomic levels (phylum, class, order, family), 11 studies reported differential abundance ( Huang et al. , 2021 ; Le et al. , 2021 ; Shan et al. , 2021 ; Svensson et al. , 2021 ; Hu et al. , 2023 ; Pai et al. , 2023 ; Wei et al. , 2023 ; Guo et al. , 2024 ; Jimenez et al. , 2024 ; Marcos et al. , 2024 ; Hicks et al. , 2025 ), when genus-level resolution was not achieved ( Supplementary Table S3 ). Overall, no consistent dysbiotic signature was identified across studies. Importantly, only Pérez-Prieto et al. (2024) applied Benjamini–Hochberg correction to account for multiple comparisons and reduce the likelihood of type I errors, underscoring a critical gap in statistical rigor in the current literature. Furthermore, among the studies that reported significant differences at the genus level between women with and without endometriosis, only four ( Huang et al. , 2021 ; Shan et al. , 2021 ; Hu et al. , 2023 ; Wei et al. , 2023 ) were considered to be of moderate quality according to the NOS. The remaining studies ( Svensson et al. , 2021 ; Guo et al. , 2024 ; Jimenez et al. , 2024 ; Marcos et al. , 2024 ; Hicks et al. , 2025 ) were assessed as low quality, indicating a higher risk of bias ( Supplementary Table S4 ). A total of 25 studies ( Khan et al. , 2016 ; Campos et al. , 2018 ; Wang et al. , 2018 ; Akiyama et al. , 2019 ; Ata et al. , 2019 ; Hernandes et al. , 2020 ; Perrotta et al. , 2020 ; Wei et al. , 2020 ; Chao et al. , 2021 ; Huang et al. , 2021 ; Le et al. , 2021 ; Lee et al. , 2021 ; Chang et al. , 2022 ; Lu et al. , 2022 ; Yuan et al. , 2022 ; Muraoka et al. , 2023 ; Yang et al. , 2023 ; Do et al. , 2024 ; Jimenez et al. , 2024 ; MacSharry et al. , 2024 ; Marcos et al. , 2024 ; Sessa et al. , 2024 ; Zhu et al. , 2024 ; Hicks et al. , 2025 ; Malvezzi et al. , 2025 ) investigated the microbiome composition in various biological fluids to explore associations with endometriosis. Fluids analyzed included vaginal, cervical, peritoneal, uterine, ovarian cyst, and oropharyngeal samples (see Tables 3–8 for study-specific details). Unlike fecal samples, biological fluids are complex matrices with inherently lower microbial biomass and, typically, their collection is more complex. Protocols for DNA extraction and microbiome analysis from these fluids are so far, less standardized, contributing to methodological heterogeneity across studies. Technical characteristics of microbiome analyses across vaginal fluid samples in the included studies. Women with endometriosis vs women without endometriosis 33 vs 15 31.6 ± 0.8 vs 33.3 ± 1.8 § ( P =NS) 28.3 ± 1.3 vs 29.5 ± 1.9 § ( P =NS) DNeasy PowerSoil Pro Kit—Qiagen (For the isolation of microbial genomic DNA from all soil types) (Women with mild/minimal endometriosis vs women with moderate/severe endometriosis) vs women without endometriosis (11 vs 10) vs 19 (35 [33–37] vs 33 [30–35]) vs 38 [35–40] † ( P =NR) (23.3 [22.6–24.1] vs 25.6 [22.2–26.8]) vs 23.0 [21.1–28.0] † ( P =NR) QIAamp UCP Pathogen Mini Kit—Qiagen (For microbial DNA purification from whole blood, swabs, cultures, and body fluids) Women with endometriosis vs (women with other gynecological diseases vs healthy controls) 21 vs (24 vs 19) 35.9 ± 8.1 vs (35.8 ± 7.3 vs 31.5 ± 3.5) § ( P =NS) QIAamp DNA Kit—Qiagen (For isolation of genomic, mitochondrial, bacterial, parasite or viral DNA from tissues, swabs, CSF, blood, body fluids or washed cells from urine) Fertile women with endometriosis vs fertile women without endometriosis 24 vs 99 27.4 ± 3.2 vs 25 ± 5.7 § ( P =NS) 22.5 ± 3.3 vs 22.7 ± 4.5 § ( P =NS) DNeasy Blood and Tissue Kit—Qiagen (For extraction of total DNA from animal blood and tissues and from cells, yeast, bacteria, or viruses) Women with chronic pelvic pain with endometriosis vs women with other gynecological disorders without chronic pelvic pain 35 vs 15 34.7 ± 8.8 vs 40.6 ± 8.2 § ( P =NS) Reported by three ranges of age ( P =NS) DNeasy PowerSoil Pro Kit—Qiagen (For the isolation of microbial genomic DNA from all soil types) Infertile women with endometriosis vs women with other infertility-related conditions 8 vs 13 42.7 ± 5.5 vs 39.4 ± 3.7 § ( P =NS) QIAamp Fast DNA Tissue Kit—Qiagen (For rapid isolation of genomic DNA from solid tissue samples) Women with endometrioma vs healthy controls 19 vs 21 29 [28–37] vs 37 [34–40] † ( P =NR) DNeasy PowerLyzer PowerSoil Kit—Qiagen (For isolation of DNA from tough soil microbes) Women with endometriosis vs women without endometriosis 10 vs 10 34.5 [31.0–39.0] vs 34.5 [32.0–37.0] † ( P =NR) QIAamp DNA Microbiome Kit—Qiagen (For isolation of bacterial microbiome DNA from swab and body fluids) Women with endometriosis vs healthy controls 16 vs 18 36.75 ± 7.11 vs 35 ± 6.61 § ( P =NS) 20.64 ± 3.04 vs 19.75 ± 1.47 § ( P =NS) TIANamp Bacteria DNA Kit—TianGen (For genomic DNA extraction from Gram-negative, Gram-positive bacteria, and pathogenic bacteria of food) Women with endometriosis vs women with other benign gynecological indications 20 vs 9 32.5 ± 1.1 vs 32.6 ± 2.0 § ( P =NS) 26.5 ± 1.5 vs 28.1 ± 2.4 § ( P =NS) PowerMag Soil DNA Isolation Kit—MoBio (For isolation of microbial DNA from all types of soil) Women with CPP with endometriosis vs (women with CPP without endometriosis vs healthy control) 37 vs (25 vs 66) 39.89 ± 6.24 vs ( 37.56 ± 5.480 vs 38.23 ± 7.80) § ( P =NR) Cetyltrimethylammonium bromide (CTAB) (Generic method for isolating genomic DNA from different tissues) Women with endometriosis vs women without endometriosis with other benign gynecological conditions 36 vs 14 QIAamp DNA Kit—Qiagen (For isolation of genomic, mitochondrial, bacterial, parasite or viral DNA from tissues, swabs, CSF, blood, body fluids or washed cells from urine) Women with endometriosis vs women without endometriosis with other benign gynecological conditions 10 vs 11 QIAamp DNA Blood Kit—Qiagen (For purification of genomic, mitochondrial or viral DNA from blood and other body fluids) Women with endometriosis vs women without endometriosis 35 vs 24 34.9 ± 6.8 vs 35.25 ± 6.9 § ( P =NS) 24.8 ± 4.5 vs 24.3 ± 2.7 § ( P =NS) PowerMag Soil DNA Isolation Kit—MoBio (For isolation of microbial DNA from all types of soil) Women with endometriosis vs healthy controls 14 vs 14 28.5 [26–31.3] vs 27.5 [25.8–30] † ( P =NS) 23 (21–24.3) vs 21 (20.1–24.2) † ( P =NS) QuickGene DNA Extraction Tissue Kit S—Biotec (For isolation of genomic DNA) Data reported as reported by the original papers, unless otherwise stated. bp = base pairs; CPP = chronic pelvic pain syndrome; NR = not reported; NR = not reported; NS = non-significant; nt = nucleotides;. LVFX, levofloxacin; qRT-PCR, quantitative real time-PCR; WGS, whole-genome sequencing. Data are expressed as mean±SD. Data are expressed as median [25th–75th percentile]. Technical characteristics of microbiome analyses across cervical fluid samples in the included studies. Women with endometrioma vs healthy controls 19 vs 21 29 [28–37] vs 37 [34–40] † ( P =NR) DNeasy PowerLyzer PowerSoil Kit—Qiagen (For isolation of DNA from tough soil microbes) Women with endometriosis vs healthy controls 23 vs 10 35 [30–39] † vs NR TC Genomic DNA Isolation Kit—Fair Biotech (For genomic DNA isolation from tissue samples) Women with endometriosis vs women without endometriosis 21 vs 20 38.3 ± 7.88 vs 34.0 ± 10.8 § ( P =NS) 21.5 ± 2.79 vs 24.3 ± 8.16 § ( P =NS) Quick-RNA Fecal/Soil Microbe Microprep Kit—Zymo Research (For extract of RNA from various soil, fecal, and water samples) Women with endometriosis vs women without endometriosis with other benign gynecological conditions 36 vs 14 QIAamp DNA Kit—Qiagen (For isolation of genomic, mitochondrial, bacterial, parasite or viral DNA from tissues, swabs, CSF, blood, body fluids or washed cells from urine) Women with endometriosis vs women without endometriosis 30 vs 39 33.9 ± 5.7 vs 32.5 ± 6.0 § ( P =NS) 21.3 ± 3.2 vs 20.5 ± 2.8 § ( P =NS) NucleoSpin Microbial DNA Mini kit—Macherey‐Nagel (For Isolation of total DNA from Gram-positive and -negative bacteria, yeast, and fungi) Women with endometriosis vs healthy controls 14 vs 14 28.5 [26–31.3] vs 27.5 [25.8–30] † ( P =NS) 23 (21–24.3) vs 21 (20.1–24.2) † ( P =NS) QuickGene DNA Extraction Tissue Kit S—Biotec (For isolation of genomic DNA) Women with endometriosis vs women without endometriosis 73 vs 31 36 [15–49] vs 39 [26–51] † ( P =NS) Reported by three ranges of age ( P =NS) PureLink Genomic DNA Mini Kit—Invitrogen (For genomic DNA purification from blood, tissues, cells, bacteria, swabs, and blood spots) Data reported as reported by the original papers, unless otherwise stated. Bp, base pairs; NR, not reported; NS, non-significant; qRT-PCR, quantitative real time-PCR. Data are expressed as mean±SD. Data are expressed as median [25th–75th percentile]. Technical characteristics of microbiome analyses across peritoneal fluid samples in the included studies. Women with endometriosis vs women without endometriosis 27 vs 23 34[31–41] vs 42[34–46] † ( P  = 0.029) 23[21–28] vs 26[23–29] † ( P =NS) QIAamp DNA Kit—Qiagen (For isolation of genomic, mitochondrial, bacterial, parasite or viral DNA from tissues, swabs, CSF, blood, body fluids or washed cells from urine) (Infertile women with endometriosis stage I/II vs stage III/IV) vs women with tubal obstruction-related infertility (8 vs 18) vs 31 (28.8 ± 4.4 vs 31.1 ± 5.6) vs 31.0 ± 5.3 § ( P =NS) (20.88 ± 2.05 vs 20.60 ± 2.83) vs 23.14 ± 2.98 § ( P  = 0.007) MagPure Soil DNA Kit—Magen (For isolation of high-quality genomic DNA from various soil, stool, and other environmental samples) Women with endometriosis vs women without endometriosis 36 vs 25 35.28 ± 7.24 vs 33.32 ± 8.04 § ( P =NS) 20.9 ± 2.11 vs 21.4 ± 2.03 § ( P =NS) Chloroform/Isoamyl Alcohol (Generic method for purifying DNA from cells and soft tissues) Women with endometriosis vs women without endometriosis 21 vs 20 38.3 ± 7.88 vs 34.0 ± 10.8 § ( P =NS) 21.5 ± 2.79 vs 24.3 ± 8.16 § ( P =NS) Quick-RNA Fecal/Soil Microbe Microprep Kit—Zymo Research (For extract of RNA from various soil, fecal, and water samples) Women with endometriosis vs women without endometriosis 45 vs 45 36.2 ± 1.3 vs 39.4 ± 1.1 § ( P =NS) 36.2 ± 1.3 vs 39.4 ± 1.1 § ( P =NS) PowerMag Soil DNA Isolation Kit—MoBio (For isolation of microbial DNA from all types of soil) Women with endometriosis vs women without endometriosis with other benign gynecological conditions 36 vs 14 QIAamp DNA Kit—Qiagen (For isolation of genomic, mitochondrial, bacterial, parasite or viral DNA from tissues, swabs, CSF, blood, body fluids or washed cells from urine) Infertile women with endometriosis vs infertile women without endometriosis 55 vs 30 37.2 ± 8.2 vs 37.7 ± 7.4 § ( P =NS) 22.5 ± 2.3 vs 22.9 ± 2.1 § ( P =NS) MagicPure Soil and Stool Genomic DNA Kit–TransGen Biotech (For DNA purification from various types of soil and stool samples) Women with endometriosis vs women without endometriosis 54 vs 24 PureLink Genomic DNA Mini Kit—Invitrogen (For genomic DNA purification from blood, tissues, cells, bacteria, swabs, and blood spots) Data reported as reported by the original papers, unless otherwise stated. Bp, base pairs; NR, not reported; NS, non-significant; qRT-PCR, quantitative real time-PCR. Data are expressed as mean±SD. Data are expressed as median [25th–75th percentile]. Technical characteristics of microbiome analyses across uterine fluid samples in the included studies. Infertile women with endometriosis vs women with other infertility-related conditions 8 vs 13 42.7 ± 5.5 vs 39.4 ± 3.7 § ( P =NS) QIAamp Fast DNA Tissue Kit—Qiagen (For rapid isolation of genomic DNA from solid tissue samples) (Infertile women with endometriosis stage I/II vs stage III/IV) vs women with tubal obstruction-related infertility (8 vs 18) vs 31 (28.8 ± 4.4 vs 31.1 ± 5.6) vs 31.0 ± 5.3 § ( P =NS) (20.88 ± 2.05 vs 20.60 ± 2.83) vs 23.14 ± 2.98 § ( P  = 0.007) MagPure Soil DNA Kit—Magen (For isolation of high-quality genomic DNA from various soil, stool, and other environmental samples) Women with endometriosis vs women without endometriosis with other benign gynecological conditions 36 vs 14 QIAamp DNA Kit—Qiagen (For isolation of genomic, mitochondrial, bacterial, parasite or viral DNA from tissues, swabs, CSF, blood, body fluids or washed cells from urine) (Women with endometriosis using GnRHa vs not using GnRHa) vs (women without endometriosis using GnRH analogue vs not using GnRH analogue) (16 vs 16) vs (16 vs 16) (37.5 ± 5.6 vs 35.7 ± 8.3) vs (42.1 ± 8.6 vs 33.6 ± 8.9; P  < 0.01) § ( P =NR) UltraClean ® Soil DNA Isolation Kit—MoBio (For isolate cellular, PCR quality DNA from soil) Data reported as reported by the original papers, unless otherwise stated. Bp, base pairs; NR, not reported; NS, non-significant. Data are expressed as mean±SD. Technical characteristics of microbiome analyses across ovarian cyst fluid samples in the included studies. Women with endometrioma not using GnRH analogue vs women with serous/mucinous cyst adenoma not using GnRH analogue 8 vs 8 UltraClean ® Soil DNA Isolation Kit—MoBio (For isolate cellular, PCR quality DNA from soil) Data reported as reported by the original papers, unless otherwise stated. NR, not reported. Technical characteristics of microbiome analyses across oropharyngeal fluid samples in the included studies. Women with endometriosis vs (women with other gynecological diseases vs healthy controls) 21 vs (24 vs 19) 35.9 ± 8.1 vs (35.8 ± 7.3 vs 31.5 ± 3.5) § ( P =NS) QIAamp DNA Kit—Qiagen (For isolation of genomic, mitochondrial, bacterial, parasite or viral DNA from tissues, swabs, CSF, blood, body fluids or washed cells from urine) Infertile women with endometriosis vs women with other infertility-related conditions 8 vs 13 42.7 ± 5.5 vs 39.4 ± 3.7 § ( P =NS) QIAamp Fast DNA Tissue Kit—Qiagen (For rapid isolation of genomic DNA from solid tissue samples) Data reported as reported by the original papers, unless otherwise stated. Bp, base pairs; NR, not reported; NS, non-significant. Data are expressed as mean±SD. Consequently, it is not surprising that DNA extraction methods varied substantially. The majority of studies ( Khan et al. , 2016 ; Wang et al. , 2018 ; Hernandes et al. , 2020 ; Perrotta et al. , 2020 ; Chao et al. , 2021 ; Huang et al. , 2021 ; Lee et al. , 2021 ; Chang et al. , 2022 ; Lu et al. , 2022 ; Yuan et al. , 2022 ; Yang et al. , 2023 ; Do et al. , 2024 ; Jimenez et al. , 2024 ; Marcos et al. , 2024 ; Sessa et al. , 2024 ; Zhu et al. , 2024 ) employed DNA extraction kits not specifically designed for fluid samples, using kits optimized for soil, feces, or tissues. For instance, one study ( Chao et al. , 2021 ) used cetyltrimethylammonium bromide (CTAB), a general-purpose method traditionally used for plants and tissues, to isolate DNA from vaginal fluid. Similarly, another study ( Yuan et al. , 2022 ) extracted DNA from peritoneal fluid using chloroform/isoamyl alcohol, a standard method for DNA purification from soft tissues and cells. Despite the use of extraction methods not tailored for low-biomass fluid samples or microbiome-specific applications, most studies reported adequate DNA recovery, enabling successful sequencing and subsequent microbiome analysis. Technical characteristics of the 15 studies ( Ata et al. , 2019 ; Hernandes et al. , 2020 ; Perrotta et al. , 2020 ; Wei et al. , 2020 ; Chao et al. , 2021 ; Le et al. , 2021 ; Lu et al. , 2022 ; Muraoka et al. , 2023 ; Yang et al. , 2023 ; Jimenez et al. , 2024 ; MacSharry et al. , 2024 ; Marcos et al. , 2024 ; Sessa et al. , 2024 ; Hicks et al. , 2025 ) analyzing the vaginal fluid microbiome in women with endometriosis are detailed in Table 3 . Sample sizes were generally small, ranging from n = 8 to n = 37 for endometriosis cases and n = 9 to n = 99 for controls. Only five studies recruited healthy women as controls ( Ata et al. , 2019 ; Lu et al. , 2022 ; Yang et al. , 2023 ) or as a subgroup within the control population ( Chao et al. , 2021 ; Hicks et al. , 2025 ). Some studies further limited their scope to specific subtypes of endometriosis. For instance, while Muraoka et al. (2023) enrolled 144 participants, only n = 10 cases of ovarian endometriosis were included for vaginal fluid analysis. Similarly, Yang et al. (2023) and Hernandes et al. (2020) focused solely on deep endometriosis, analyzing n = 19 and n = 10 vaginal fluid samples, respectively. Age was reported and found to be comparable between groups in most studies, though four ( Chao et al. , 2021 ; Muraoka et al. , 2023 ; Yang et al. , 2023 ; MacSharry et al. , 2024 ) did not report whether significant differences existed, and two ( Hernandes et al. , 2020 ; Wei et al. , 2020 ) did not report age data at all. BMI was not reported in seven studies ( Hernandes et al. , 2020 ; Wei et al. , 2020 ; Chao et al. , 2021 ; Muraoka et al. , 2023 ; Yang et al. , 2023 ; Marcos et al. , 2024 ; Hicks et al. , 2025 ). Diagnosis of endometriosis was primarily surgical, although three studies ( Perrotta et al. , 2020 ; Marcos et al. , 2024 ; Sessa et al. , 2024 ) also accepted imaging-based diagnoses, suggesting a focus on moderate to severe cases. However, eight studies ( Wei et al. , 2020 ; Chao et al. , 2021 ; Le et al. , 2021 ; Lu et al. , 2022 ; Do et al. , 2024 ; Jimenez et al. , 2024 ; Marcos et al. , 2024 ; Hicks et al. , 2025 ) did not specify the phenotype of endometriosis and two of them ( Chao et al. , 2021 ; Marcos et al. , 2024 ) also omitted information on disease stage. Hormonal treatment was an exclusion criterion in seven studies ( Ata et al. , 2019 ; Perrotta et al. , 2020 ; Wei et al. , 2020 ; Lu et al. , 2022 ; Muraoka et al. , 2023 ; Yang et al. , 2023 ; MacSharry et al. , 2024 ) while two studies did not report hormonal treatment status at all ( Hernandes et al. , 2020 ; Hicks et al. , 2025 ). Among the remaining studies, several reported hormone usage, but only four ( Chao et al. , 2021 ; Le et al. , 2021 ; Do et al. , 2024 ; Jimenez et al. , 2024 ) confirmed balanced distribution across groups. The other two ( Marcos et al. , 2024 ; Sessa et al. , 2024 ) did not clarify this distribution. Regarding antibiotic use, only three studies ( Le et al. , 2021 ; MacSharry et al. , 2024 ; Hicks et al. , 2025 ) did not consider antibiotic treatments as an exclusion criterion; furthermore, none of these clarify whether its use was similar between groups. There was marked inconsistency in accounting for the menstrual cycle phase at the time of vaginal fluid collection. Six studies ( Hernandes et al. , 2020 ; Le et al. , 2021 ; Lu et al. , 2022 ; Do et al. , 2024 ; Jimenez et al. , 2024 ; Hicks et al. , 2025 ) did not report this information. One study ( Perrotta et al. , 2020 ) collected samples during both menstrual and follicular phases. Among the others, either the distribution of phases was similar between groups ( Ata et al. , 2019 ; Chao et al. , 2021 ; Muraoka et al. , 2023 ; MacSharry et al. , 2024 ), or samples were collected during a uniform phase across participants, through the chosen phase differed across studies: ovulatory phase ( Marcos et al. , 2024 ; Sessa et al. , 2024 ); early follicular phase ( Wei et al. , 2020 ); or follicular phase ( Yang et al. , 2023 ). As with stool and anal fluid studies, dietary factors were not thoroughly considered. No study excluded participants based on special diets (such as vegetarianism), which could influence microbiome composition ( Table 1 ). Microbiome analysis details are summarized in Table 3 . The majority of studies utilized 16S rRNA gene sequencing on Illumina or Ion Torrent platforms ( Ata et al. , 2019 ; Hernandes et al. , 2020 ; Perrotta et al. , 2020 ; Wei et al. , 2020 ; Chao et al. , 2021 ; Le et al. , 2021 ; Lu et al. , 2022 ; Yang et al. , 2023 ; Do et al. , 2024 ; Jimenez et al. , 2024 ; Marcos et al. , 2024 ; Sessa et al. , 2024 ; Hicks et al. , 2025 ). Only one study ( MacSharry et al. , 2024 ) conducted shotgun metagenomic paired-end sequencing, a technique based on whole-genome sequencing. In one case ( Muraoka et al. , 2023 ), the analysis involved bioinformatic reanalysis of previously deposited datasets (The European Nucleotide Archive: PRJEB16013 and PRJEB21098), followed by quantitative real-time polymerase chain reaction (qRT-PCR) to detect a specific bacterial species ( Supplementary Table S2 ). Bioinformatics pipelines, including sequence filtering, chimera removal, OTU clustering, and taxonomic assignment, varied widely across studies. A detailed overview is provided in Supplementary Table S2 . Alpha and beta diversity metrics were reported in nearly all studies ( Fig. 2 ), with the exception of four ( Perrotta et al. , 2020 ; Wei et al. , 2020 ; Muraoka et al. , 2023 ; Marcos et al. , 2024 ). Significant differences in alpha diversity between endometriosis cases and controls were consistently reported in only two studies ( Yang et al. , 2023 ; MacSharry et al. , 2024 ). Four additional studies observed differences only in specific sub-analyses ( Chao et al. , 2021 ; Le et al. , 2021 ; Do et al. , 2024 ; Sessa et al. , 2024 ). Beta diversity findings were similarly variable. Five studies ( Ata et al. , 2019 ; Le et al. , 2021 ; Yang et al. , 2023 ; Jimenez et al. , 2024 ; Hicks et al. , 2025 ) reported no significant difference in community composition between groups, while one study ( Lu et al. , 2022 ) found a significant overall difference. Five other studies ( Hernandes et al. , 2020 ; Chao et al. , 2021 ; Do et al. , 2024 ; MacSharry et al. , 2024 ; Sessa et al. , 2024 ) reported differences only within subgroup analyses. Among the genera found to differ significantly between endometriosis and control groups, only nine taxa were reported in two or more studies with consistent findings ( Fig. 4A ). Alloscardovia sp. ( Chao et al. , 2021 ; Lu et al. , 2022 ; MacSharry et al. , 2024 ), Anaerococcus sp. ( Jimenez et al. , 2024 ; MacSharry et al. , 2024 ), Clostridium sp. ( Chao et al. , 2021 ; Le et al. , 2021 ), Corynebacterium sp. ( Jimenez et al. , 2024 ; MacSharry et al. , 2024 ), Escherichia sp. ( Ata et al. , 2019 ; Sessa et al. , 2024 ; Hicks et al. , 2025 ), Fusobacterium sp. ( Muraoka et al. , 2023 ; Jimenez et al. , 2024 ), Streptococcus sp. ( Yang et al. , 2023 ; Jimenez et al. , 2024 ), and Veillonella sp. ( Chao et al. , 2021 ; Yang et al. , 2023 ; MacSharry et al. , 2024 ) had increased abundance in the endometriosis groups while Pseudomonas sp. ( Sessa et al. , 2024 ; Hicks et al. , 2025 ) was identified as being more abundant in the control groups. Other genera exhibited inconsistent trends, showing increased abundance in endometriosis in some studies and in controls in others, such as Atopobium sp. ( Ata et al. , 2019 ; Le et al. , 2021 ; Lu et al. , 2022 ), Bifidobacterium ( Yang et al. , 2023 ; Jimenez et al. , 2024 ; Sessa et al. , 2024 ), Gardnerella sp. ( Ata et al. , 2019 ; Hernandes et al. , 2020 ; Le et al. , 2021 ; Lu et al. , 2022 ; Yang et al. , 2023 ; Marcos et al. , 2024 ), Lactobacillus sp. ( Hernandes et al. , 2020 ; Wei et al. , 2020 ; Chao et al. , 2021 ; Le et al. , 2021 ; Lu et al. , 2022 ; Yang et al. , 2023 ; MacSharry et al. , 2024 ; Sessa et al. , 2024 ), Limosilactobacillus sp. ( Yang et al. , 2023 ; Jimenez et al. , 2024 ; Sessa et al. , 2024 ), Megasphaera sp. ( Yang et al. , 2023 ; Sessa et al. , 2024 ), Prevotella ( Hernandes et al. , 2020 ; Wei et al. , 2020 ; Le et al. , 2021 ; Yang et al. , 2023 ; Jimenez et al. , 2024 ; Sessa et al. , 2024 ; Hicks et al. , 2025 ), Sneathia sp. ( Chao et al. , 2021 ; Yang et al. , 2023 ; Sessa et al. , 2024 ; Hicks et al. , 2025 ). One study ( Perrotta et al. , 2020 ) conducted a subgroup analysis by endometriosis stage and found that Anaerococcus sp. was significantly increased in stage III-IV disease compared to stage I-II. Furthermore, among studies that reported significant differences at the genus level between women with and without endometriosis, half of them ( Le et al. , 2021 ; Lu et al. , 2022 ; Muraoka et al. , 2023 ; Jimenez et al. , 2024 ; Hicks et al. , 2025 ) were considered to be of low or moderate quality according to the NOS. In contrast, the remaining studies ( Chao et al. , 2021 ; Yang et al. , 2023 ; MacSharry et al. , 2024 ; Sessa et al. , 2024 ) were assessed as high quality, indicating a low risk of bias ( Supplementary Table S4 ). Bacterial genera identified across vaginal and cervical samples in the included studies. E, genus’s abundance increased in endometriosis; C, genus’s abundance increased in controls; dark pink, increased in endometriosis in ≥3 studies; mid-pink, increased in endometriosis in two studies; light pink, increased in endometriosis in one study; grey, inconsistent findings across studies; light blue, decreased in endometriosis in one study; mid-blue, decreased in endometriosis in two studies. ( A ) Vaginal fluid. ( B ) Cervical fluid. Seven studies also reported microbial differences at higher taxonomic ranks (phylum, class, order, or family) ( Wei et al. , 2020 ; Chao et al. , 2021 ; Le et al. , 2021 ; Lu et al. , 2022 ; Marcos et al. , 2024 ; Sessa et al. , 2024 ; Hicks et al. , 2025 ) ( Supplementary Table S3 ). Technical characteristics of the five studies investigating the microbiome of cervical fluid ( Campos et al. , 2018 ; Ata et al. , 2019 ; Huang et al. , 2021 ; Chang et al. , 2022 ; Yang et al. , 2023 ) and the two in the cervical mucus ( Akiyama et al. , 2019 ; Wei et al. , 2020 ) are detailed in Table 4 . The sample sizes varied considerably: the number of endometriosis cases ranged from n = 14 to n = 73, and controls from n = 10 to n = 39. Only three studies ( Ata et al. , 2019 ; Chang et al. , 2022 ; Yang et al. , 2023 ) included healthy women as controls, whereas the others used symptomatic women with other gynecological conditions. Demographic data such as age and BMI were inconsistently reported. Age was found to be comparable between groups in most studies, except for one ( Yang et al. , 2023 ) that did not report whether differences were significant, and two studies ( Wei et al. , 2020 ; Chang et al. , 2022 ), which did not report age data for at least one group. BMI was not reported in three studies ( Wei et al. , 2020 ; Chang et al. , 2022 ; Yang et al. , 2023 ), while others confirmed no significant difference between groups. All studies confirmed endometriosis diagnosis via surgery and histological analysis, and most included women with moderate to severe disease. However, four studies ( Akiyama et al. , 2019 ; Wei et al. , 2020 ; Huang et al. , 2021 ; Chang et al. , 2022 ) did not specify the phenotype of endometriosis. Only Yang et al. (2023) limited their analysis to endometrioma cases ( Table 1 ). Regarding hormonal treatments, five studies ( Akiyama et al. , 2019 ; Ata et al. , 2019 ; Wei et al. , 2020 ; Huang et al. , 2021 ; Yang et al. , 2023 ) excluded participants receiving hormonal therapy. One study ( Chang et al. , 2022 ) did not report on hormonal treatment, while another ( Campos et al. , 2018 ) allowed inclusion of participants undergoing treatment, with reported use in 30.1% of cases and 16.1% of controls, though statistical significance of this difference was not provided. Only one study ( Chang et al. , 2022 ) did not consider antibiotic use as an exclusion criterion and did not report whether usage differed between groups. Menstrual cycle phase at sample collection was not standardized: two studies ( Chang et al. , 2022 ; Yang et al. , 2023 ) did not report the cycle phase; one ( Wei et al. , 2020 ) collected samples in the early follicular phase; another ( Yang et al. , 2023 ) consistently used the follicular phase.The other studies collected samples during either the follicular or luteal phase. Some reported no significant differences between phases within groups ( Campos et al. , 2018 ; Akiyama et al. , 2019 ; Ata et al. , 2019 ) while one study did not specify whether a difference existed ( Huang et al. , 2021 ). No study in this group considered dietary habits as an exclusion criterion. Microbiome analysis details are summarized in Table 4 . All studies used 16S rRNA sequencing on Illumina or Ion Torrent platforms, except for one ( Campos et al. , 2018 ), who applied qRT-PCR to detect specific bacterial species ( Supplementary Table S2 ). However, the 16S regions targeted were not consistent across studies. Bioinformatics pipelines, including sequence filtering, chimera removal, OTU clustering, and taxonomic assignment, varied widely across studies. A detailed overview is provided in Supplementary Table S2 . Alpha diversity ( Fig. 2 ) was reported by five studies ( Akiyama et al. , 2019 ; Ata et al. , 2019 ; Huang et al. , 2021 ; Chang et al. , 2022 ; Yang et al. , 2023 ), but only three ( Akiyama et al. , 2019 ; Chang et al. , 2022 ; Yang et al. , 2023 ) found statistically significant differences in bacterial diversity between cases and controls. Beta diversity was evaluated in six studies ( Campos et al. , 2018 ; Akiyama et al. , 2019 ; Ata et al. , 2019 ; Huang et al. , 2021 ; Chang et al. , 2022 ; Yang et al. , 2023 ), but only two ( Campos et al. , 2018 ; Chang et al. , 2022 ) reported a statistically significant difference in microbial composition between women with and without endometriosis. Microbiome composition at the genus level was reported in all studies ( Fig. 4B ). Only four genera were consistently identified in two or more studies as being more abundant in endometriosis, which were Bifidobacterium sp. ( Chang et al. , 2022 ; Yang et al. , 2023 ), Pseudomonas sp. ( Akiyama et al. , 2019 ; Wei et al. , 2020 ), Streptococcus sp. ( Akiyama et al. , 2019 ; Ata et al. , 2019 ; Chang et al. , 2022 ; Yang et al. , 2023 ), and Veillonella sp. ( Wei et al. , 2020 ; Yang et al. , 2023 ). Megasphaera sp. , Prevotella sp. , Sneathia sp. which were found to be more abundant in endometriosis cases in Yang et al. (2023) , were more abundant in controls in Ata et al. (2019) . Three studies ( Akiyama et al. , 2019 ; Wei et al. , 2020 ; Huang et al. , 2021 ) also reported higher-level taxonomic data at the phylum, class, order, and/or family level ( Supplementary Table S3 ). Furthermore, among studies that reported significant differences at the genus level between women with and without endometriosis, two ( Akiyama et al. , 2019 ; Wei et al. , 2020 ) were considered to be of moderate quality according to the NOS, while the other three ( Ata et al. , 2019 ; Chang et al. , 2022 ; Yang et al. , 2023 ) were assessed as high quality, indicating a low risk of bias ( Supplementary Table S4 ). Technical characteristics of the eight studies ( Campos et al. , 2018 ; Wang et al. , 2018 ; Wei et al. , 2020 ; Huang et al. , 2021 ; Lee et al. , 2021 ; Yuan et al. , 2022 ; Zhu et al. , 2024 ; Malvezzi et al. , 2025 ) investigating the microbiome of peritoneal fluid in women with endometriosis are detailed in Table 5 . The number of endometriosis cases ranged from n = 21 to n = 55. Control groups included n = 14 to n = 45 participants, composed exclusively of women undergoing laparoscopy for gynecological conditions unrelated to endometriosis. In two studies ( Wang et al. , 2018 ; Zhu et al. , 2024 ), controls also included infertile women. Most studies reported comparable age between groups; however, one study ( Malvezzi et al. , 2025 ) observed a statistically significant age difference ( P  = 0.029). Two studies ( Campos et al. , 2018 ; Wei et al. , 2020 ) did not report age data. BMI was significantly different in one study ( Zhu et al. , 2024 ), unreported in two ( Campos et al. , 2018 ; Wei et al. , 2020 ), and comparable across groups in the remaining studies. The diagnosis of endometriosis was confirmed by surgery and histological examination in all studies, except for one ( Wang et al. , 2018 ), which relied on surgical findings alone. An unusual methodological choice was made by Lee et al. (2021) , who analyzed the microbiome in extracellular vesicles isolated from peritoneal fluid, rather than the fluid itself. The phenotypes of endometriosis investigated varied widely across studies, although all the studies included severe presentations of the disease. Regarding hormonal therapy, almost all studies considered it an exclusion criterion, except for Malvezzi et al. (2025) and Campos et al. (2018) , who did not even report whether its distribution differed significantly between groups. Similarly, while most studies excluded participants with recent antibiotic use, two ( Wang et al. , 2018 ; Malvezzi et al. , 2025 ) did not report data on this aspect. Significant heterogeneity was observed in the menstrual cycle phase during which peritoneal fluid samples were collected. Some studies collected samples exclusively during the early follicular phase ( Wang et al. , 2018 ; Wei et al. , 2020 ); while one study ( Lee et al. , 2021 ) collected exclusively during the follicular phase. One study ( Zhu et al. , 2024 ) did not report the timing of sample collection. Other studies collected samples at various phases of the menstrual cycle. Among them, three studies ( Campos et al. , 2018 ; Yuan et al. , 2022 ; Malvezzi et al. , 2025 ) reported that the distribution of cycle phases was similar between study groups. In contrast, two studies ( Huang et al. , 2021 ; Zhu et al. , 2024 ) did not indicate whether the phase distribution was comparable between cases and controls. No study accounted for dietary habits as an exclusion criterion. Microbiome sequencing methodologies are summarized in Table 5 . All studies used 16S rRNA gene sequencing via Illumina or Ion Torrent platforms, except for one ( Campos et al. , 2018 ), who used qRT-PCR to target specific bacterial species. However, the 16S regions amplified differed between studies, and there was no standardized approach to bioinformatic processing. Details on pipelines used for filtering, chimeric sequence removal, identification of OTUs, etc, varied significantly. Notably, some studies provided minimal ( Wei et al. , 2020 ; Zhu et al. , 2024 ) or no ( Wang et al. , 2018 ) bioinformatics details ( Supplementary Table S2 ). Alpha diversity ( Fig. 2 ) was assessed in four studies ( Huang et al. , 2021 ; Lee et al. , 2021 ; Yuan et al. , 2022 ; Zhu et al. , 2024 ). Only one ( Zhu et al. , 2024 ) reported a significant difference, observed only between women with stage III-IV endometriosis and controls. Beta diversity was reported by five studies ( Campos et al. , 2018 ; Huang et al. , 2021 ; Lee et al. , 2021 ; Yuan et al. , 2022 ; Zhu et al. , 2024 ). Significant differences in microbial community structure between endometriosis and control groups were consistently found in three studies ( Campos et al. , 2018 ; Lee et al. , 2021 ; Yuan et al. , 2022 ), and in one study ( Zhu et al. , 2024 ) only when comparing stage III-IV patients with both stage I-II and controls combined. Considering the genera identified in the peritoneal fluid samples, significant differences in microbial composition were observed between women with and without endometriosis. Notably, an increased abundance of Streptococcus sp. was found in the endometriosis groups in two studies ( Lee et al. , 2021 ; Yuan et al. , 2022 ). Pseudomonas was another genus consistently reported with higher abundance in endometriosis cases across multiple studies ( Wei et al. , 2020 ; Huang et al. , 2021 ; Lee et al. , 2021 ; Zhu et al. , 2024 ; Malvezzi et al. , 2025 ), emerging as a predominant and recurring finding. This latter association was further supported by unsupervised analyses using random forest classifiers ( Huang et al. , 2021 ). In contrast, Lactobacillus iners was reported to be more abundant in the control groups in two studies ( Wei et al. , 2020 ; Huang et al. , 2021 ; Fig. 5A ), suggesting a potential protective or non-pathogenic role. Bacterial genera identified across peritoneal, uterine, and oropharyngeal fluid samples in the included studies. E, genus’s abundance increased in endometriosis; C, genus’s abundance increased in controls; dark pink, increased in endometriosis in ≥3 studies; mid-pink, increased in endometriosis in two studies; light pink, increased in endometriosis in one study; grey, inconsistent findings across studies; light blue, decreased in endometriosis in one study; mid-blue, decreased in endometriosis in two studies. ( A ) Peritoneal fluid. ( B ) Uterine fluid. ( C ) Oropharyngeal fluid. Despite these consistent findings, notable inconsistencies were also observed. For instance, Enhydrobacter species ( Lee et al. , 2021 ; Zhu et al. , 2024 ; Malvezzi et al. , 2025 ) and Staphylococcus sp. ( Zhu et al. , 2024 ; Malvezzi et al. , 2025 ) were found to be more abundant in endometriosis cases in some studies, whereas other investigations reported higher levels in the control group. These divergent results may reflect methodological differences or population-specific factors across studies. In addition to genus-level analysis, several studies also provided bacterial identification at higher taxonomic levels, including phylum, class, order, and family ( Supplementary Table S3 ). Furthermore, among studies that reported significant differences at the genus level between women with and without endometriosis, four ( Wei et al. , 2020 ; Huang et al. , 2021 ; Yuan et al. , 2022 ; Zhu et al. , 2024 ) were considered to be of moderate quality according to the NOS. The other two ( Lee et al. , 2021 ; Malvezzi et al. , 2025 ) were assessed as high quality, indicating a low risk of bias ( Supplementary Table S4 ). Technical characteristics of the four studies ( Khan et al. , 2016 ; Wei et al. , 2020 ; Marcos et al. , 2024 ; Zhu et al. , 2024 ) analyzing the uterine fluid microbiome in women with endometriosis are detailed in Table 6 . The sample sizes across these investigations were generally small, ranging from as few as n = 8 to a maximum of n = 36 cases, and from n = 14 to n = 32 controls ( Table 6 ). In all studies, the control groups comprised women with other gynecological conditions. Age was reported to be comparable between groups in two studies ( Marcos et al. , 2024 ; Zhu et al. , 2024 ), whereas one study ( Khan et al. , 2016 ) did not clarify whether age distributions were similar, and another ( Wei et al. , 2020 ) did not provide age data at all. BMI was reported only by one study ( Zhu et al. , 2024 ), which found a statistically significant difference between groups ( P  = 0.007). Endometriosis diagnosis was confirmed by surgery and histology in most of the studies, all of which included patients across different disease stages. Notably, the most recent study ( Marcos et al. , 2024 ) based the diagnosis on imaging or surgical findings but did not report even the disease stage. Regarding hormonal treatment, there was considerable variation. In two studies ( Wei et al. , 2020 ; Zhu et al. , 2024 ), hormonal treatments were an exclusion criterion. In contrast, in one ( Khan et al. , 2016 ), half of the participants in each group were receiving GnRH analogues as part of the study design. Marcos et al. (2024) did not report whether hormonal therapy was an exclusion criterion or whether its use differed between groups. Most studies excluded participants with recent antibiotic use; however, one study ( Khan et al. , 2016 ) did not report any information on this aspect. There was also substantial heterogeneity regarding the menstrual cycle phase during which uterine fluid samples were collected. Marcos et al. (2024) collected samples during ovulation (Days 12–16 of the menstrual cycle), Wei et al. (2020) exclusively during the early follicular phase, while Zhu et al. (2024) did not report the timing of sample collection. In Khan et al. (2016) , samples were collected across different phases of the cycle, but it was not specified whether this variability was distributed similarly between groups. Consistent with the majority of studies included in this review, none of the investigations evaluating the uterine fluid microbiome considered special dietary habits as an exclusion criterion. Microbiome analysis details are presented in Table 6 . All studies employed 16S rRNA sequencing using Illumina or Ion Torrent platforms; however, the amplified regions varied ( Wei et al. , 2020 ; Zhu et al. , 2024 ), or were not reported at all ( Khan et al. , 2016 ; Marcos et al. , 2024 ). Likewise, comprehensive details regarding bioinformatic analyses, such as pipelines used, filtering criteria, chimeric sequence removal, identification of OTUs, etc, were absent in most studies, with the exception of Marcos et al. (2024) , who reported nearly all steps of the computational workflow ( Supplementary Table S2 ). Only one study ( Zhu et al. , 2024 ) assessed microbial diversity measures. In this study, alpha diversity was found to be similar between groups, while beta diversity differed significantly between women with stages III-IV endometriosis and those with stages I-II combined with controls ( Fig. 2 ). Regarding the taxonomic composition, only two studies ( Wei et al. , 2020 ; Zhu et al. , 2024 ) reported the identification of bacterial genera in uterine or endometrial fluid samples. However, no specific taxa were consistently observed between these two studies ( Fig. 5B ). The microbiome of ovarian cyst fluid in women with endometriosis has been examined in only a single study to date ( Khan et al. , 2016 ), as detailed in Table 1 . This study included a limited sample size, comprising just n = 8 cases and n = 8 controls ( Table 7 ). Information regarding patient selection, sample collection, as well as bioinformatic methods and pipelines, is summarized in Tables 1 and 7 and Supplementary Table S2 . Notably, the study did not evaluate alpha or beta diversity metrics ( Fig. 2 ). At the family level, ovarian cyst fluid from endometriosis patients showed an enrichment of Streptococcaceae and Moraxellaceae , alongside a relative decrease of Lactobacillaceae , Enterobacteriaceae , and Staphylococcaceae , when compared to cyst fluid from women with other benign ovarian conditions ( Supplementary Table S3 ). These findings, while suggestive, remain preliminary and underscore the need for further confirmatory studies with larger cohorts and standardized methodologies. The microbiome of oropharyngeal fluid in women with endometriosis has been investigated by only two studies to date ( Marcos et al. , 2024 ; Hicks et al. , 2025 ), with technical characteristics summarized in Table 8 . Marcos et al. (2024) included a small sample of n = 8 endometriosis cases and n = 13 infertile women without endometriosis as controls, while Hicks et al. (2025) evaluated a larger cohort comprising n = 21 women with endometriosis, n = 24 women with other gynecological conditions, and n = 19 healthy controls. As observed in other microbiome studies included in this review, critical variables that could influence microbial composition, such as hormonal use and dietary patterns, were not considered exclusion criteria in either study. Moreover, BMI was not reported, and Hicks et al. (2025) did not provide information on the timing of sample collection within the menstrual cycle or on the distribution of recent antibiotic use across groups. Microbiome sequencing methods are detailed in Table 8 , while the approaches to bioinformatic analyses, including pipelines, filtering, chimeric sequence removal, identification of OTUs, etc, varied widely and are presented in Supplementary Table S2 . Of the two studies, only Hicks et al. (2025) evaluated alpha and beta diversity. While alpha diversity did not differ significantly between groups, beta diversity was found to be statistically different between them ( Fig. 2 ). Regarding microbial composition, certain taxa appeared to differ between women with and without endometriosis. Haemophilus sp. ( Marcos et al. , 2024 ) and Veillonella sp. ( Hicks et al. , 2025 ) were reported in greater abundance in the oropharyngeal fluid of women with endometriosis. Conversely, several genera, including Actinobacillus sp. , Bifidobacterium sp. , Butyrivibrio sp. , Lactobacillus sp. , Lactococcus sp. , Lautropia sp. , Megasphaera sp. , Neisseria sp. , Prevotella sp. ( Hicks et al. , 2025 ), as well as Streptococcus sp. ( Marcos et al. , 2024 ) were found to be more abundant in controls ( Fig. 5C ). At the family level ( Supplementary Table S3 ), endometriosis cases were associated with increased abundance of Actinomycetaceae ( Hicks et al. , 2025 ) and Enterobacteriaceae ( Marcos et al. , 2024 ), along with a reduction in Peptostreptococcaceae and Neisseriaceae ( Hicks et al. , 2025 ). Despite these findings, inconsistencies across studies highlight the need for further research to confirm whether these microbial alterations are reproducible and relevant to the pathophysiology of endometriosis ( Supplementary Table S3 ). The microbiome of eutopic endometrial tissue in women with endometriosis has been investigated in six studies ( Hernandes et al. , 2020 ; Khan et al. , 2021 ; Wessels et al. , 2021 ; Muraoka et al. , 2023 ; Marcos et al. , 2024 ; Guo et al. , 2025 ), with technical characteristics summarized in Table 9 . Technical characteristics of microbiome analyses across eutopic endometrium samples in the included studies. Symptomatic women with endometriosis vs symptomatic women without endometriosis 30 vs 13 37.10 ± 7.30 vs 40.92 ± 7.41 § ( P =NS) 21.27 ± 1.94 vs 23.63 ± 3.37 § ( P  < 0.01) QIAamp DNA Kit—Qiagen (For isolation of genomic, mitochondrial, bacterial, parasite or viral DNA from tissues, swabs, CSF, blood, body fluids or washed cells from urine) Infertile women with endometriosis vs women with other infertility-related conditions 8 vs 13 42.7 ± 5.5 vs 39.4 ± 3.7 § ( P =NS) QIAamp Fast DNA Tissue Kit—Qiagen (For rapid isolation of genomic DNA from solid tissue samples) Women with endometriosis vs women with other gynecological conditions 42 vs 42 34.5 [31.0–39.0] vs 34.5 [32.0–37.0] † ( P =NR) Women with pelvic pain with endometriosis vs women with pelvic pain without endometriosis 12 vs 9 33.8 ± 5.8 vs 35.1 ± 3.3 § ( P =NS) RNeasy Kit—Qiagen (For purification of total RNA from cells, tissues, and yeast) (Women with endometriosis receiving different treatments: untreated vs GnRHa vs LVFX vs GnRHa+ LVFX) vs (fertile women with uterine fibroids receiving different treatments: untreated vs GnRHa vs LVFX vs GnRHa+ LVFX) (21 vs 11 vs 15 vs 6) vs (11 vs 12 vs 10 vs 14) ( P =NR) (36.3 ± 7.7 vs 38.7 ± 5.2 vs 38.2 ± 8.2 vs 35.5 ± 5.6) vs (41.2 ± 8.1 vs 37.5 ± 5.3 vs 43.0 ± 4.5 vs 36.7 ± 4.5) § ( P =NR) UltraClean ® Soil DNA Isolation Kit—MoBio (For isolate cellular, PCR quality DNA from soil) Women with endometriosis vs women without endometriosis with other benign gynecological conditions 10 vs 11 DNeasy PowerSoil Pro Kit—Qiagen (For the isolation of microbial genomic DNA from all soil types) Data reported as reported by the original papers, unless otherwise stated. Bp, base pairs; NR, not reported; NS, non-significant; nt, nucleotides; LVFX, levofloxacin; qRT-PCR, quantitative real time-PCR. Data are expressed as mean±SD. Data are expressed as median [25th–75th percentile]. Sample sizes varied considerably, ranging from as few as n=8 to a maximum of n=53 endometriosis cases, with one study further subdividing its few cases (n=53) based on treatment exposure, including GnRH analogues or levofloxacin. Control groups, generally composed of symptomatic or infertile women, or those with other gynecological conditions, ranged from n=9 to n=51 participants, with similar subgrouping applied in the latter. Only half of the studies reported comparisons of age between cases and controls. One study ( Hernandes et al. , 2020 ) did not provide age-related data, while two ( Khan et al. , 2021 ; Muraoka et al. , 2023 ) reported ages without clarifying whether groups were statistically comparable. BMI was reported in only one study, which found a statistically significant difference between groups. The diagnostic criteria for endometriosis were relatively consistent across studies, with most confirming diagnosis through both surgery and histological examination. While two studies ( Khan et al. , 2021 ; Muraoka et al. , 2023 ) focused exclusively on ovarian endometriosis and one ( Hernandes et al. , 2020 ) on deep disease, the remaining did not specify disease phenotype. Only one ( Wessels et al. , 2021 ) reported that all the stages of endometriosis were included. The use of hormonal treatment varied. In three studies ( Wessels et al. , 2021 ; Muraoka et al. , 2023 ; Guo et al. , 2025 ), its use was an exclusion criterion. In contrast, in Khan et al. (2021) , hormonal treatment with GnRH analogue was integrated into the study design, with participants evenly distributed across treatment groups. Marcos et al. (2024) did not clarify the compatibility of treatment usage between groups, and Hernandes et al. (2020) omitted this information entirely. Regarding use of antibiotics, it was excluded in nearly all studies, except one ( Wessels et al. , 2021 ), who did not report it as an exclusion criterion, and another ( Khan et al. , 2021 ), who incorporated levofloxacin use as part of the intervention. As with other sample types, none of the studies considered dietary habits as an exclusion criterion. Despite the known influence of hormonal cycles on endometrial physiology and microbial communities, the menstrual cycle phase at the time of sample collection was inconsistently addressed. Marcos et al. (2024) were the only authors to collect all the samples during a uniform menstrual phase, the time of ovulation. Muraoka et al. (2023) and Wessels et al. (2021) collected samples at different phases but reported that time did not differ significantly between groups. Khan et al. (2021) documented cycle phases but did not report on their statistical comparability. Guo et al. (2025) , and Hernandes et al. (2020) did not mention the menstrual phase at all. The sampling methods also varied across studies. Curettage was used in three studies ( Hernandes et al. , 2020 ; Wessels et al. , 2021 ; Guo et al. , 2025 ), a seed swab in Khan et al. (2021) , and an endometrial sampler in Marcos et al. (2024) . In the study by Muraoka et al. (2023) , samples were obtained during surgical removal of the uterus, although the methodology was not described in detail. Methodological heterogeneity was also reflected in DNA extraction protocols. Some studies employed kits originally intended for soil samples (such as UltraClean ® Soil DNA Isolation Kit—MoBio ( Khan et al. , 2021 ) and DNeasy PowerSoil Pro Kit—Qiagen ( Hernandes et al. , 2020 )), raising concerns about methodological appropriateness for tissue microbiome profiling. Sequencing platforms were consistent, with nearly all studies employing 16S rRNA sequencing on Illumina or Ion Torrent systems; however, the regions of the 16S gene analyzed varied widely, ranging from the V3 region alone ( Wessels et al. , 2021 ) to V3–V4 ( Hernandes et al. , 2020 ; Guo et al. , 2025 ) and V5–V6 ( Khan et al. , 2021 ), while Marcos et al. (2024) did not specify the region amplified. Muraoka et al. (2023) took a distinct approach, conducting a bioinformatic reanalysis of publicly available datasets (European Nucleotide Archive studies PRJEB16013 and PRJEB21098), followed by targeted qRT-PCR validation. Bioinformatic analyses, including sequence filtering, chimera removal, and OTU assignment, varied across studies, with further details presented in Supplementary Table S2 . Alpha diversity findings were inconsistent ( Fig. 2 ). While Hernandes et al. (2020) reported no differences in microbial richness or evenness between women with and without endometriosis, Guo et al. (2025) , Wessels et al. (2021) , and Khan et al. (2021) observed significant differences. Khan et al. (2021) additionally found that alpha diversity varied among treated and untreated endometriosis patients. Beta diversity analyses yielded similarly inconsistent results. Three studies ( Hernandes et al. , 2020 ; Khan et al. , 2021 ; Wessels et al. , 2021 ) found no significant differences between cases and controls. In contrast, Guo et al. (2025) observed a significant separation in microbial community structure between the groups. Notably, Marcos et al. (2024) and Muraoka et al. (2023) did not report any diversity metrics for eutopic endometrium tissue in their analyses. In terms of taxonomic composition, the genera identified as differing significantly between women with and without endometriosis were not consistent across studies ( Fig. 6A ). No bacterial genus was consistently found to be significantly altered across studies, highlighting the lack of reproducibility and the overall weakness of evidence for an endometriosis-associated microbial signature in eutopic endometrial tissue. These inconsistencies underscore the challenges in identifying a coherent dysbiosis profile for endometriosis. Three studies ( Wessels et al. , 2021 ; Marcos et al. , 2024 ; Guo et al. , 2025 ) extended their analyses to higher taxonomic levels, including phylum, class, order, and family ( Supplementary Table S3 ). Bacterial genera identified across tissue sample types in the included studies. E, genus’s abundance increased in endometriosis; C, genus’s abundance increased in controls; mid-pink, increased in endometriosis in two studies; light pink, increased in endometriosis in one study; grey, inconsistent findings across studies; light blue, decreased in endometriosis in one study. ( A ) Eutopic endometrium. ( B ) Endometriotic tissue. Technical characteristics of the five studies ( Campos et al. , 2018 ; Hernandes et al. , 2020 ; Hu et al. , 2023 ; Muraoka et al. , 2023 ; Chen et al. , 2024 ) that analyzed the microbiome composition of ectopic endometrial tissue compared to control tissues are detailed in Table 10 . Technical characteristics of microbiome analyses across endometriotic tissue samples in the included studies. Ovarian endometriotic tissue from women with endometriosis vs eutopic endometrium from non-endometriosis women with uterine fibroids 23 vs 22 34.8 ± 6.8 vs 37.2 ± 8.2 § ( P =NS) MagPure Soil DNA Kit—Magen (For isolation of high-quality genomic DNA from various soil, stool, and other environmental samples) Endometriotic tissue vs eutopic endometrium from the same women with endometriosis 14 vs 14 Ovarian endometriotic tissue from women with endometriosis vs eutopic endometrium from the same women 42 vs 42 Endometriotic tissue vs eutopic endometrium from the same women with endometriosis 10 vs 11 DNeasy PowerSoil Pro Kit—Qiagen (For the isolation of microbial genomic DNA from all soil types) Endometriotic tissue vs healthy peritoneum from women without endometriosis 68 vs 30 PureLink Genomic DNA Mini Kit—Invitrogen (For genomic DNA purification from blood, tissues, cells, bacteria, swabs, and blood spots) Data reported as reported by the original papers, unless otherwise stated. Bp, base pairs; NA, not applicable; NR, not reported; NS, non-significant; nt, nucleotides; qRT-PCR, quantitative real time-PCR. Data are expressed as mean±SD. Sample sizes ranged from n=10 to n=68 for endometriosis cases and from n=11 to n=42 controls. When evaluating the microbiome of endometriotic tissues, the choice of control tissue is critical. Microbial profiles are expected to differ depending on whether the comparison is made to the eutopic endometrium from the same affected women, from a woman without endometriosis, or from a different tissue altogether. In three studies ( Hernandes et al. , 2020 ; Hu et al. , 2023 ; Muraoka et al. , 2023 ), ectopic tissue microbiomes were compared to eutopic endometrial tissue from the same woman. In contrast, Chen et al. (2024) used eutopic endometrium from women without endometriosis but with uterine fibroids as the control tissue. These control subjects were age-matched to the cases, although BMI data were not reported. Similarly, Campos et al. (2018) used peritoneal tissue from symptomatic women without endometriosis as controls, but data on age and BMI were not reported. In all studies, both eutopic and ectopic tissue samples were obtained intraoperatively ( Table 1 ). Histological confirmation of endometriosis was conducted in all studies except one ( Muraoka et al. , 2023 ), which relied on surgical identification alone ( Table 10 ). Three studies ( Hu et al. , 2023 ; Muraoka et al. , 2023 ; Chen et al. , 2024 ) focused exclusively on ovarian endometriosis, while Hernandes et al. (2020) analyzed deep endometriosis. Campos et al. (2018) included multiple phenotypes and stages of the disease. Hormonal treatment varied across the studies. In three studies ( Hu et al. , 2023 ; Muraoka et al. , 2023 ; Chen et al. , 2024 ), participants were not on hormonal therapies. Hernandes et al. (2020) did not report hormonal use and did not list it as an exclusion criterion. Campos et al. (2018) included participants with varying hormonal use across groups but did not indicate whether differences were statistically significant. Menstrual cycle phase at the time of sample collection was not consistently controlled. Hernandes et al. (2020) did not report this information. Campos et al. (2018) noted similar phases between groups, while Chen et al. (2024) stated that over 90% of samples were in the follicular phase. Regarding antibiotic use, most studies considered it an exclusion criterion, except for Chen et al. (2024) , who did not specify its distribution between groups. None of the studies reported any exclusion criteria regarding dietary habits. Technical aspects and microbiome analysis methods are summarized in Table 10 . Only Campos et al. (2018) reported using a tissue-specific DNA extraction kit (Purelink Genomic DNA Mini Kit, Invitrogen). Other studies used adapted soil DNA extraction kits ( Hernandes et al. , 2020 ; Chen et al. , 2024 ), and two ( Hu et al. , 2023 ; Muraoka et al. , 2023 ) did not report their extraction methods. Three studies used 16S rRNA sequencing on Illumina platforms, targeting the V3-V4 regions. Muraoka et al. (2023) and Campos et al. (2018) did not conduct 16S sequencing but instead used qRT-PCR to detect specific taxa. As previously mentioned, Muraoka et al. (2023) conducted a bioinformatic reanalysis of publicly available datasets followed by targeted qRT-PCR validation. Bioinformatic analyses, including sequence filtering, chimera removal, and OTU assignment, varied across studies, with further details presented in Supplementary Table S2 . Alpha and beta diversity were evaluated by three studies ( Hernandes et al. , 2020 ; Hu et al. , 2023 ; Chen et al. , 2024 ) ( Fig. 2 ). All studies reported no significant differences in alpha diversity between ectopic and control tissues. Similarly, beta diversity did not differ between groups, except in the study by Hernandes et al. (2020) , which reported significant differences when using weighted UniFrac distances, but not when using Bray–Curtis dissimilarity. Regarding the taxonomic composition, the results were heterogeneous. Among the genera identified as significantly differing between ectopic and control tissues ( Fig. 6B ), only Pseudomonas was reported by more than one study ( Hernandes et al. , 2020 ; Hu et al. , 2023 ) as being significantly increased in ectopic tissue compared to eutopic tissue. This limited concordance reflects the weak evidence base, lack of reproducibility, and the absence of evidence of a distinct microbial signature in endometriotic lesions. Furthermore, the two studies ( Hernandes et al. , 2020 ; Hu et al. , 2023 ) that reported significant differences at the genus level between women with and without endometriosis were assessed as being of moderate quality according to the NOS, indicating a moderate risk of bias ( Supplementary Table S4 ).

After analyzing the studies assessing microbiome composition in various districts of women with endometriosis, and according to the information provided by the literature that addresses technical confounders of microbiome analysis, the main conclusion of this review was the mandatory need to standardize study designs. The calculation of power and sample size, the recruiting of an adequate number of participants and samples, the reduction of interpersonal confounders and the use of adequate computational methodological approaches are critical for the advancement of knowledge in this field. Alternatively, all attempts to interpret the tremendously huge and different dimensions of a microbiome will fail to approach its full potential. This review aimed to highlight the existing research on this subject and provide insights that may help guide future directions to enhance the applicability of obtained results in the future.

The association between endometriosis and microbiota is emerging as a promising area of research for understanding pathophysiological mechanisms underlying the disease and for identifying potential biomarkers for diagnosis and management. A healthy microbial balance across different anatomical sites is essential for maintaining mucosal integrity, protecting against pathogens, and regulating physiological processes. Dysbiosis, or an imbalance in the microbiota, can compromise the intestinal barrier, allowing bacteria or endotoxins to pass through, triggering inflammation responses that may contribute to the onset or progression of various diseases. Additionally, some gut bacteria produce enzymes involved in estrogen metabolism, potentially influencing the estrobolome, the collection of microbial genes involved in estrogen metabolism and estrogen-related disorders such as endometriosis ( Pai et al. , 2023 ). In line with these concepts, a quite large number of narrative reviews have been published over the past few years, speculating on the causal relationship between microbiota and endometriosis ( Leonardi et al. , 2020 ; D’Alterio et al. , 2021 ; Colonetti et al. , 2023 ; Weber et al. , 2024 ). Yet, despite these efforts, there remains a critical gap in the literature including a thorough review of study designs and methodologies employed to investigate microbiota alterations in endometriosis. Scoping reviews are particularly valuable in such contexts, enabling researchers to map existing evidence systematically and to discern patterns across studies. Indeed, several studies have utilized scoping reviews to synthesize knowledge in various areas of microbiome research ( Gáspár et al. , 2024 ; Lianos et al. , 2024 ; Nagamine, 2024 ; Sessa et al. , 2025 ). A scoping review is often more suitable for a microbiome research compared to a systematic review due to several reasons: (i) it allows a broad overview of available evidence, including different approaches, trends, and knowledge gaps, rather than answering a specific, narrow question like a systematic review; (ii) contrary to a systematic review which requires strict inclusion/exclusion criteria, it embraces a wider diversity among studies; (iii) it is more appropriate if there is not enough homogeneous data to conduct a quantitative synthesis; microbiome studies vary widely in sample types, sequencing platform, analytical pipelines, and study designs (cross-sectional, longitudinal, interventional). Importantly in our context, scoping reviews also illuminate methodological shortcomings and highlight critical areas where further investigation is needed. Such reviews are particularly well-suited for identifying research gaps and shaping future research agendas, offering a roadmap for addressing limitations and improving the rigor of studies in our field. Recently, multiple studies have focused on characterizing a distinct endometriosis microbiome signature in both humans and animal models. As shown in the results section of this review, several bacterial species have been identified as being associated with endometriosis pathogenesis. However, consistency in the type of bacterial species present and their relative abundances compared to control subjects continues to be elusive for all the anatomical districts investigated. In eutopic endometrial tissue investigated by six studies, no bacterial genus was consistently found to be significantly different between groups across the studies. Some consistency was found for the reduced presence of Lachnospira sp. in stool/anal fluid of endometriosis patients according to 4 of the 15 studies investigating this district microbiome and for the higher presence of Streptococcus sp. in four of the seven studies on cervical samples. Similarly, a higher abundance in peritoneal fluid of patients with endometriosis was found for Pseudomonas sp. in four of the eight studies. Among the genera reported as significantly differing between women with and without endometriosis in vaginal microbiome, three genera ( Alloscardovia sp. , Escherichia sp., and Veillonella sp.) were identified as more abundant in endometriosis in 3 of the 15 studies. Several studies even reported contradicting results between women affected and controls, specifically with regard to abundance of Blautia sp. and Ruminococcus sp. in the stool microbiome and Gardnerella sp., Prevotella sp., and Sneathia sp. in the vaginal district. As shown in Fig. 2 , inconsistencies across studies were also found for alpha and beta diversity for all the districts investigated. Overall, these data hint at weak evidence supporting an endometriosis-associated microbial signature. One recurring issue in microbiome research is the challenge of ensuring adequate sample sizes. Studies have suggested that statistical analyses of microbiome datasets require specialized considerations that differ from traditional sample size calculation methods ( Ferdous et al. , 2022 ), since these studies are exposed to high risks of type I errors. Thus, in the context of case-control studies investigating differences in microbial composition, robust study design requires an a priori assessment of sample size to ensure sufficient statistical power to detect hypothesized effects. Recent guidelines aiming to improve the quality and consistency of microbiome research underscore the importance of power calculations and adequate sample sizes as essential criteria for rigorous study design ( Kelly et al. , 2015 ). Despite this, none of the studies reviewed in this analysis performed an a priori sample size calculation, which represents a significant methodological limitation that compromises the reliability of the findings. A significant limitation in comparing microbiome composition between women with and without endometriosis across available studies is the insufficient attention given to the phase of the menstrual cycle during sample collection. While shifts in microbiome composition throughout the menstrual cycle are described as subtle in fecal samples, they are pronounced in vaginal samples ( Krog et al. , 2022 ). For instance, studies have reported an increase in microbial diversity in the vaginal microbiome during menstruation, followed by a marked dominance of Lactobacillus spp. during the follicular and luteal phases, which strongly correlates with serum estradiol levels. Menstrual-phase samples tend to show a relatively higher abundance of taxa from the phyla Fusobacteria , Proteobacteria , Bacteroidetes , and Actinobacteria . In contrast, follicular-phase samples exhibit a notable reduction or absence of these phyla, with an increased prevalence of Firmicutes taxa ( Kaur et al. , 2020 ). Similarly, hormone-driven variations also affect the gut microbiome. Differences in microbial composition have been linked to testosterone levels in men and estradiol levels in women. Elevated estradiol levels, in particular, have been associated with altered gut microbial communities, characterized by increased abundance of Bacteroidetes and decreased Firmicutes ( Shin et al. , 2019 ). These findings underscore the significant influence of hormonal fluctuations on microbiome composition. Therefore, the lack of menstrual phase standardization across the included studies poses a major methodological concern. Notably, 10 out of 15 studies comparing vaginal fluid microbiome in women with and without endometriosis did not report the menstrual cycle phase during sampling or failed to collect samples consistently within the same phase. This issue severely compromises the validity and comparability of the results obtained. The same problem applies to gut microbiomes, as hormone-related variations in microbial composition prevent reliable aggregation of findings without rigorous control for menstrual cycle phases across cases and controls. Addressing this critical confounding factor in future research is essential to enhance the interpretability and robustness of microbiome studies in endometriosis. Differences in vaginal microbiome composition based on contraception methods have been well-documented ( Shin et al. , 2019 ; Balle et al. , 2023 ). Women who do not use hormonal contraceptives exhibit significantly greater shifts in vaginal microbiome composition across the menstrual cycle compared to those using oral contraceptives. Studies investigating the influence of oral contraceptives on the vaginal microbiome have primarily reported increased levels of Lactobacillus species ( Balle et al. , 2023 ). On the other hand, long-acting progestin-only contraceptives, such as levonorgestrel-implants, have not been shown to significantly alter the vaginal mucosal microbial environment ( Balle et al. , 2023 ). The impact of exogenous sex hormones extends beyond the vaginal microbiome, influencing the gut microbiome as well. Hormonal treatments, including oral contraceptives, have been associated with reduced bacterial richness and diversity in the gut ( Kheloui et al. , 2023 ). Specifically, users of oral contraceptives demonstrate decreased alpha diversity but no significant changes in beta diversity compared to non-users. Given these findings, the inclusion of participants undergoing hormonal treatments alongside those who are not, presents a critical confounding factor in studies comparing the vaginal fluid and gut microbiota between women with and without endometriosis. Standardized protocols controlling for hormonal treatment status are therefore essential to improve the reliability and validity of microbiome research. Studying low-biomass microbial niches such as the endometrium requires meticulously designed experiments to minimize contamination, which can lead to data misinterpretation ( Molina et al. , 2021 ; Odendaal et al. , 2024 ). Contaminant DNA can originate from various sources, including endometrial sampling techniques, the laboratory environment, plastic consumables, researchers, and reagents ( Molina et al. , 2021 ). Furthermore, cross-contamination during microbiome sample processing, such as from other samples or sequencing runs, is a significant concern ( Molina et al. , 2021 ). Recently, methodological considerations and good practice recommendations for studying the endometrial microbiome have been proposed ( Molina et al. , 2021 ). Molina et al. (2021) emphasized that the primary challenge in sampling for endometrial microbiota analysis lies in the high risk of contamination from the lower genital tract. Most studies reviewed utilized conventional endometrial sampling devices. While some authors reported measures to avoid contact with the cervical and vaginal walls, the insertion of sampling devices through the high-microbial biomass cervicovaginal canal without a sheathed tool inevitably increases the likelihood of contamination. None of the studies employed double-sheathed catheters, which are widely used in embryo transfers and are expected to significantly reduce the risk of contamination from cervical or vaginal microbiota ( Reschini et al. , 2022 ). One study ( Muraoka et al. , 2023 ) collected samples during hysterectomy, effectively avoiding contamination from the vaginal or cervical microbiota. However, details of the sampling procedure were not provided, and the study involved peri- and postmenopausal women, limiting the applicability of the findings to women of reproductive age. Additionally, there was a borderline statistical difference in age between women with and without endometriosis in the study, potentially introducing an additional confounding factor. Women with endometriosis often adopt lifestyle changes, including dietary modifications to help manage symptoms. In an online survey of 4087 Italian women with endometriosis, Mazza et al. (2023) reported that 66% of respondents altered their eating habits following diagnosis. Common dietary choices included gluten-free, anti-inflammatory, Mediterranean, and ketogenic diets, indicating that women explore diverse dietary strategies to alleviate symptoms and improve quality of life. Since dietary patterns can significantly influence gut microbiota composition ( Mirhosseini et al. , 2024 ) changes in gut microbial diversity in women with endometriosis may reflect these self-management practices rather than being a direct cause of the disease itself. Studies investigating the gut microbiome should account for dietary factors as potential confounders. However, few studies reviewed in this context controlled these dietary influences, highlighting a critical gap in the existing research. In recent years, substantial efforts have been devoted to developing standardized methods for microbiome research. Initiatives such as the Human Microbiome Project and various methodological guidelines have emphasized the importance of ensuring reliable, reproducible, and robust microbiome data ( Bruggeling et al. , 2021 ; Tourlousse et al. , 2021 ; Szóstak et al. , 2022 ). Large-scale studies of human gut microbiomes have demonstrated significant compositional differences across geographically distinct populations. However, most studies on the relationship between microbiome composition and health outcomes focus on single populations. This raises questions about whether geographical variations in microbiome composition translate to differences in disease susceptibility. Factors such as lifestyle, dietary habits, and antibiotic usage profoundly influence microbiome composition, complicating cross-country comparisons. In this review, most of the included studies were conducted in China ( Wang et al. , 2018 ; Wei et al. , 2020 , 2023 ; Chao et al. , 2021 ; Huang et al. , 2021 ; Shan et al. , 2021 ; Lu et al. , 2022 ; Yuan et al. , 2022 ; Hu et al. , 2023 ; Yang et al. , 2023 ; Chen et al. , 2024 ; Zhu et al. , 2024 ). The reviewed studies employed diverse sampling, DNA extraction, and sequencing techniques. DNA extraction was predominantly performed using commercial extraction kits while a smaller subset of studies used alternative methods ( Chao et al. , 2021 ; Yuan et al. , 2022 ; Hu et al. , 2023 ). Even among studies using commercial kits, subtle differences in protocols across laboratories can introduce variability, particularly in low-abundance samples such as mucus, tissues, or fluids ( Sinclair et al. , 2023 ). Commercial kits, not optimized for a specific tissue or fluid, may lead to bias in the metagenomic analysis due to inefficient extraction, alteration in taxa abundance, and alpha and beta diversities ( Galla et al. , 2024 ; Seethalakshmi et al. , 2024 ). For taxonomic identification, most studies employed 16s rRNA gene sequencing for bacterial detection, while others PCR targeting specific genes of interest ( Campos et al. , 2018 ; Muraoka et al. , 2023 ). Only the two more recent studies used shotgun metagenomics ( MacSharry et al. , 2024 ; Pérez-Prieto et al. , 2024 ) that provides a broader compositional overview of multiple components of human microbiomes ( Joos et al. , 2025 ). To analyze microbial functionality, shotgun sequencing has been reported as ‘ a preferable analytic approach, as it allows for direct functional annotation and analysis of metagenomes compared with amplicon sequencing ’ ( Joos et al. , 2025 ). The choice of sequencing platform often reflects the availability of instruments at the time of the study, which adds another dimension of variability. In downstream statistical analyses, the reviewed studies varied widely in their choice of software tools and, frequently, the specific versions of those tools. Multiple approaches were exploited for filtering of the data, OTU identification and taxonomy and functional annotation, as well as for the computation of alpha and beta diversity scores. However, in many cases, the studies did not specify the versions of the tools used, adding another layer of inconsistency. These methodological discrepancies make reproducibility challenging, and complicate efforts to compare findings across studies. Standardized bioinformatic pipelines and transparent reporting of methods, including software versions, are essential for advancing microbiome research and ensuring comparability of results across different populations and research contexts. The best way to compare the results would be to reanalyze all the data using the same bioinformatic procedure together with a batch effect correction step. However, this is not possible presently since not all the studies uploaded their data and metadata on a repository to make them publicly available. A recent perspective paper published in Nature Reviews Microbiology highlights critical limitations in microbiome research, emphasizing challenges such as defining a ‘healthy’ microbiome and the need to account for confounding factors, niche-specific variables, and population-specific geographical findings ( Joos et al. , 2025 ). These limitations directly impact the reliability and relevance of microbiome studies and may explain the inconsistencies observed in the studies reviewed here. Specifically, the changes in gut, vaginal, or endometrial microbiomes among patients with endometriosis compared to controls were found to be highly variable. Microbiome research in endometriosis often mirrors approaches used in other fields without adequately addressing previously identified limitations. A notable example is the investigation of Fusobacterium nucleatum ( F. nucleatum ) which has been found to have a role in various diseases, particularly those linked to active inflammation and malignancies. Whether F. nucleatum acts as a driver (causative agent) or a passenger (opportunistic colonizer) remains a subject of debate. For instance, F. nucleatum is implicated in placental infections and preterm births, with evidence from human umbilical endothelial cell studies and mouse models suggesting its pathogenic potential. However, its presence in healthy placental microbiota raises concerns about sampling bias, as diseased tissues are more frequently studied. This suggests that F. nucleatum may only become pathogenic under specific conditions or involve particular strains, paralleling its role in oral health. The association between F. nucleatum and colorectal cancer further illustrates the challenges of translating microbiome findings into clinical practice. While multiple studies report its presence across various stages of colorectal cancer, reproducibility issues persist due to differences in detection methods. Prevalence rates range from 3% to 56%, depending on the methodology employed ( Brennan and Garrett, 2019 ). Diagnostic applications are further hindered by the lack of standardization cutoff values and biomarker analysis methods ( Lișcu et al. , 2024 ). In preclinical models, antibiotics targeting F. nucleatum reduced tumor growth and bacterial burden in xenografts ( Brennan and Garrett, 2019 ). However, these findings have seen limited clinical applications, as antibiotics may negatively impact immunotherapy efficacy or exacerbate microbial imbalances, complicating their therapeutic use ( González et al. , 2024 ). Overall, lessons from these studies underline the importance of addressing heterogeneity in: (i) specimen collection, (ii) sequencing methods, (iii) population demographics, and (iv) data analysis tools. Only a single study adjusted for multiple comparisons ( Pérez-Prieto et al. , 2024 ). When the first paper on endometriosis and the gut microbiome was published, experts in other fields underscored that ‘ robust experimental approaches, be it within human cohorts or preclinical models, and reproducible results across microbiota studies are pivotal to bridge the translational gap and ensure that data are neither lost in translation nor mistranslated clinically ’ ( Brennan and Garrett, 2019 ). It seems that we missed the opportunities to follow this suggestion and in the absence of robust data, we are risking being lost in translation.