Citizen science reveals comorbidities in endometriosis with no shift in vaginal microbiome composition

Abstract

Endometriosis is a chronic inflammatory condition affecting 2–10% of reproductive-aged women, most commonly presenting with pelvic pain and subfertility. While its impact on reproductive health is increasingly recognized, the vaginal microbiome’s role in the pathogenesis of endometriosis is still unclear and a comprehensive map of potential co-occurring conditions remains underexplored. Leveraging data from the Isala citizen-science platform in Flanders (Belgium), we analysed vaginal microbiome profiles obtained through 16S rRNA sequencing and health data from 95 women with self-reported endometriosis and 2,279 without. While no differences were observed in vaginal microbiome composition or diversity, we identified significant associations between endometriosis and polyendocrine metabolic ovarian syndrome (previously named polycystic ovarian syndrome; OR = 2.92, 95% CI 1.71-4.78, p < 0.001), migraine (OR = 3.75, 95% CI 1.38-8.60, p = 0.025), irritable bowel syndrome (OR = 2.57, 95% CI 1.43-4.36, p = 0.008) and dyspareunia (OR = 1.67, 95% CI 1.14-2.40, p = 0.033). These findings suggest that the vaginal microbiome composition plays at most a limited role in endometriosis and highlights how citizen science can effectively complement clinical research by capturing underrecognized comorbidities.

Introduction

Endometriosis is a chronic inflammatory condition that most commonly affects reproductive-aged women, with an estimated prevalence of 2–10%^1–5. It is defined by the presence of endometrial-like tissue outside the uterus, with lesions varying in size, depth and location, affecting both reproductive and non-reproductive organs. Pelvic pain⁶, subfertility⁷, gastrointestinal issues^8,9 and mental health problems¹⁰ have been frequently reported in different endometriosis cohorts across the world (with sizes between 174 and 188,461), imposing a significant burden on both patient and society with substantial healthcare costs^11–13. Yet, a comprehensive map of the full range of symptoms and co-occurring conditions remains lacking. The pathogenesis of endometriosis is multifactorial and remains incompletely understood. Beyond the widely accepted theory of retrograde menstruation¹⁴, current understanding embraces several complementary hypotheses including coelomic metaplasia¹⁵, vascular or lymphatic dissemination of endometrial cells¹⁶, hormonal imbalances involving oestrogen and progesterone¹⁷, immune dysregulation¹⁸ and (epi)genetic predisposition¹⁹ – with chronic low-grade systemic inflammation as a central driver of disease progression^20–23. This heterogeneous pathophysiology is mirrored in the challenges of diagnosis and treatment. Current guidelines recommend transvaginal ultrasound and/or MRI as first-line imaging modalities due to their high sensitivity and specificity for deep and ovarian endometriosis, while laparoscopy followed by histologic confirmation, an invasive and expensive procedure, is only recommended for individuals with negative imaging or unsuccessful empirical treatment²⁴. However, imaging techniques require expertise and frequently miss superficial peritoneal endometriosis, which accounts for approximately 80% of endometriosis subtypes²⁵, highlighting the need for improved non-invasive diagnostic tests. Current therapeutic strategies include hormone-suppressive therapy, surgical removal of ectopic endometrial lesions and in some cases neuromodulation²⁴. However, these approaches are not curative and can lead to various side effects; in the case of surgery, high risk of complications related to laparoscopic surgery and a risk of persisting chronic postsurgical pain^26,27; in the case of hormonal treatment, different side effects can be noted, most prevalent are depressive symptoms, vasomotor symptoms, weight gain, headaches and bleeding irregularities^28–30. Together, this underscores the urgent need for non-invasive diagnostics and novel treatment strategies.

Recently, the microbiome has emerged as a potential source for diagnostic biomarkers. Bacterial DNA- and metabolite-based signatures have been shown to be implicated in the development and progression of several inflammatory conditions such as irritable bowel disease^31,32 and polyendocrine metabolic ovarian syndrome (PMOS, previously named polycystic ovarian syndrome)^33,34. Emerging research from small-scale studies (n = 35 in Jiminez et al., 2024; n = 19 in Ata et al. 2019) indicate that endometriosis is associated with compositional differences in the microbiome across anatomical sites, including the lower and upper reproductive tracts, as well as the gastrointestinal system³⁵ using 16S rRNA amplicon sequencing. At a functional level, reduced concentrations of gut microbiota-derived n-butyrate in feces were found in a mouse injection model of endometriosis compared to mice without endometriosis using targeted metabolomics³⁶. Others have reported altered levels of 4-hydroxyindole in stool samples of individuals with endometriosis (n = 18) compared to healthy controls (n = 33) through untargeted metabolomics³⁷.

A growing area of interest is the vaginal microbiome, which plays an essential role in reproductive health and is involved in several reproductive inflammatory conditions^38,39. In most women of reproductive age, the vaginal microbiome is predominantly composed of Lactobacillus species, such as Lactobacillus crispatus and Lactobacillus iners, but also a substantial proportion of Gardnerella⁴⁰. Lactobacilli appear beneficial for maintaining a healthy vaginal environment based on various epidemiological^41,42 and lab-based studies^43–47, where their absence has been linked to an increased risk for various reproductive issues such as preterm birth^48–50, bacterial vaginosis⁵¹, endometritis³⁸ and aerobic vaginitis³⁹, mostly using 16S rRNA amplicon sequencing. Reduced levels of Lactobacillus and shifts towards a non-optimal vaginal microbial community have been associated with inflammation^52–54, potentially contributing to the chronic inflammatory environment characteristic of endometriosis. However, it remains unclear whether and how the vaginal microbiome is associated with endometriosis.

Recently we have adopted a citizen-science approach to investigate the vaginal microbiome and general vaginal health of women in Flanders (Belgium)⁴⁰. This catalogue has enabled us in our present study to analyse associations between the vaginal microbiome and endometriosis based on 16S rRNA amplicon sequencing of samples from 95 women with endometriosis and 2,279 women without endometriosis. Furthermore, to characterize the multifactorial nature of endometriosis and identify conditions that may reflect shared underlying mechanisms, we assessed a broad range of infections, clinical conditions, and vaginal symptoms through questionnaires in individuals with endometriosis (n = 144) and compared these with women without endometriosis (n = 3337).

Results

Subfertility affects up to 47% of individuals with endometriosis in a population-based cohort

To assess whether and how endometriosis symptom status influenced the vaginal microbiome composition and general health, participants with endometriosis were stratified into two groups based on their symptom status: individuals experiencing current and past endometriosis-related symptoms (n = 66, n = 78). ‘Endometriosis-related symptoms’ in our study refer to the participant’s self-perceived disease burden, as the survey did not request specific symptom types and was not designed to capture endometriosis-related symptomatology. Current endometriosis-related symptoms were defined as burden experienced at the time of study participation, and past symptoms as burden that had been experienced previously but was no longer present when the questionnaire was completed. This subset of individuals with endometriosis is not representative of the broader endometriosis population as the Isala project targeted a predominantly healthy population and likely includes milder disease presentations. The control group (n = 3,337) consisted of participants who reported not having endometriosis. To minimize the presence of potential undiagnosed endometriosis cases in the control group, we excluded individuals who reported dysmenorrhoea and requiring non-steroidal anti-inflammatory drugs (NSAIDs) during menstruation. Menopausal individuals were also excluded from this study. Information on the location of endometriotic lesions and treatment history was not available because health data were only collected via general surveys in the Isala platform⁴⁰. A detailed overview of the participants’ demographics and additional variables are summarized in Table 1. Within the endometriosis group, individuals with past symptoms were significantly older with a median age of 36 (IQR 31-41) than those with current symptoms (31 (IQR 27-35), p < 0.001). Additionally, compared to the control group (28 (IQR 24-33)), individuals with endometriosis were significantly older regardless of symptom status (p < 0.001 for past symptoms group, and p = 0.006 for current symptoms group). After adjusting for age, demographic, reproductive, and lifestyle characteristics did not significantly differ across the groups, except for subfertility and self-reported health. Subfertility was significantly more prevalent in individuals with endometriosis compared to the control group (p < 0.001). Self-reported health was assessed through a five-point Likert-scale (1=very poor, 2=poor, 3=fair, 4=good, 5=very good) with a poorer perceived health in individuals with current and past endometriosis-related symptoms compared to the control group (p < 0.001; p = 0.003). Overall, after adjustment for age, endometriosis symptom status was primarily associated with subfertility and poorer self-reported health, while other demographic and lifestyle characteristics were comparable across groups.

Table 1.

Demographic characteristics of participants with and without self-reported endometriosis

	(2)
	(1)		(3)
Characteristic	Control (N = 3337)	Endometriosis, past complaints (N = 78)	Endometriosis, current complaints (N = 66)	P-value
Age (years) Median [IQR]	28 [24–33]	36 [31–41]	31 [27–35]	< 0.001 (1;3) 0.006 (2)
BMI (kg/m2) Median [IQR]	23.1 [21.0-26.2]	23.0 [21.3-26.4]	23.6 [21.4-27.5]	0.311
Birth delivery mode				0.766
Vaginal	2964 (88.8%)	72 (92.3%)	57 (86.4%)
Caesarean section	332 (10.0%)	4 (5.1%)	7 (10.6%)
Missing	41 (1.2%)	2 (2.6%)	2 (3.0%)
Contraception				0.142
Combined oral contraceptives	805 (24.1%)	12 (15.4%)	13 (19.7%)
Progesterone only pill	31 (0.9%)	4 (5.1%)	3 (4.5%)
Hormonal intrauterine device	294 (8.8%)	10 (12.8%)	9 (13.6%)
Other or none	2207 (66.2%)	52 (66.7%)	41 (62.2%)
Marital status during last 3 months1				0.063
No partner	757 (22.7%)	15 (19.2%)	8 (12.1%)
One partner	2510 (75.2%)	55 (70.5%)	55 (83.3%)
Multiple partners	70 (2.1%)	8 (10.3%)	3 (4.6%)
Subfertility	85/1108 (7.7%)	17/47 (36.2%)	17/36 (47.2%)	< 0.001 (1;2)
Number of pregnancies Median [IQR]	0 [0-1]	2 [0-2]	0 [0-2]	0.285
Biological child(ren)	1080 (32.4%)	47 (60.7%)	25 (37.9%)	0.813
Allergies and intolerances
Gluten allergy	65 (2.0%)	3 (3.9%)	4 (6.1%)	0.236
Lactose intolerant	253 (7.6%)	11 (14.1%)	9 (13.6%)	0.142
Antibiotics during last 3 months	690 (0.7%)	20 (25.6%)	17 (25.8%)	0.467
Current smoker	298 (8.9%)	10 (12.8%)	2 (3.0%)	0.213
Current drug user	319 (9.6%)	3 (3.9%)	4 (6.1%)	0.467
HPV vaccinated	1416 (42.4%)	14 (18.0%)	17 (25.8%)	0.236
Self-rated health	4.12 ± 0.64	3.85 ± 0.72	3.79 ± 0.65	0.003 (1) < 0.001 (2)

The former group is further categorised into participants experiencing current and past endometriosis-related symptoms. Percentages were calculated based on the total number of participants within each group. Significantly different variables between the three pairwise comparisons: control vs. past symptoms (1), control vs. current symptoms (2) and current symptoms vs. past symptoms (3) are represented in the last column with an asterisk. Significant results are indicated with an asterisk. IQR: interquartile range, BMI: body mass index, HPV: humanpapillomavirus. Self-rated health is a five-point ordinary scale (very good = 5, good = 4, fair = 3, poor = 2, very poor = 1).

¹Without implying cohabitation.

Endometriosis shows no association with Lactobacillus dominance, genus-level diversity or abundance of individual taxa in the vagina

To assess whether the vaginal microbiome composition is associated with endometriosis and its symptom status, vaginal microbiome profiles were obtained for 95 of the 144 participants with self-reported endometriosis, including 43 profiles from individuals with current endometriosis-related symptoms and 52 from individuals with past symptoms. In comparison, 2279 vaginal microbiome profiles were analysed from the control group (Fig. 1a). Overall, 79.1% of participants with current endometriosis-related symptoms had a Lactobacillus-dominated microbiota, compared to 71.2% in participants with past endometriosis-related symptoms and 80.1% in the control group, with no significant differences (p = 0.371). We applied different embedding methods including t-distributed stochastic neighbour embedding (t-SNE, Fig. 1b), uniform manifold approximation and projection (UMAP, Supplementary Fig. 1a) and principal coordinates analysis (PCoA, Supplementary Fig. 1b), but no clear grouping of bacterial profiles was observed based on current or past symptom status compared to the control group. In addition, we examined whether symptom status in participants with endometriosis was associated with the (sub)genera present in at least 10% of all participants. Each test (Maaslin, Limma, DESeq, ANCOM-BC89, and a linear regression on the centered log-ratio (CLR) transformed abundance data) was adjusted for technical confounders, age, recent sexual intercourse, number of pregnancies, phase of the menstrual cycle and use of hormonal contraceptives, and associations were considered significant only if consensus was reported by at least three differential abundance tools as described in Methods and Materials. No significant associations were observed between endometriosis symptom status and relative abundance of specific microbial taxa. Similarly, no significant associations were observed between endometriosis and genus-level diversity, including both alpha (Shannon) diversity and beta diversity, irrespective of symptom status (Fig. 1d). These findings confirm that the vaginal microbiome composition and its key features play only a limited role – if any – in the pathogenesis of endometriosis.

Fig. 1. Overview of cohort size, microbiome composition, and bacterial diversity in vaginal samples of individuals with self-reported endometriosis (Endo) and controls.

a Overview of participants, collected metadata and vaginal microbiome profiles. b t-SNE analysis. c Bar plot showing the vaginal microbiome profiles of participants with self-reported endometriosis (current versus past endometriosis-related symptoms), and without self-reported endometriosis (control group) on genus level. d Associations on the level of beta diversity between the samples; Associations on the level of alpha diversity of the samples; Associations on the level of abundance of specific taxa analysed by five different differential abundance testing methods (Limma, Maaslin2, DESeq2, ANCOM-BC, and a linear regression on the centered log-ratio transformed abundance data).

Polyendocrine metabolic ovarian syndrome shows strong co-occurrence with endometriosis, alongside other comorbid conditions

Complementing the vaginal microbiome analyses and drawing on the uniquely rich survey data collected within the Isala programme, we next assessed comorbidities to better understand the broader health profile associated with endometriosis. To characterize the spectrum of comorbidities associated with endometriosis, we analysed extensive health questionnaires from individuals with endometriosis reporting past or current endometriosis-related symptoms and compared them with controls with no reported endometriosis. The prevalence of several vaginal symptoms, urogenital infections and a range of conditions related to reproductive, metabolic and gastrointestinal health was compared between these groups, with statistical significance determined after correction for multiple testing using the Benjamini–Hochberg procedure. A summary of the results is shown in Fig. 2 with the corresponding counts provided in Supplementary Table 1.

Fig. 2. Associations between endometriosis and a range of infections, clinical conditions, and vaginal symptoms.

Forest plot shows adjusted odds ratios (ORs) with 95% confidence intervals derived from multivariable logistic regression models. All outcomes refer to ever having had the infection, condition, or vaginal symptom, as self-reported by participants. Analyses were adjusted for age, body mass index, and educational level. P-values were corrected for multiple testing using the Benjamini–Hochberg false discovery rate procedure. *P-value < 0.05, **P-value < 0.01, ***P-value < 0.001

Dyspareunia was significantly more common among individuals with endometriosis, with an odds ratio of 1.67 (95% CI 1.14–2.40; p = 0.033), as was irritable bowel syndrome (OR 2.57, 95% CI 1.34–4.36, p = 0.008) and migraine (OR 3.75, 95% CI 1.38-8.60, p = 0.025). Interestingly, a significant co-occurrence of PMOS in individuals with endometriosis was observed (OR 2.92, 95% CI 1.71–4.78, p < 0.001). By contrast, no significant associations were observed between endometriosis and cardiovascular, dermatological, or respiratory conditions, nor with urogenital infections or self-reported vaginal symptoms. None of the comorbid conditions associated with endometriosis showed a significant association with vaginal microbiome composition or diversity (Supplementary Fig. 3). Together, these findings underscore that, beyond the well-recognised menstrual-related pain, individuals with endometriosis experience a higher prevalence of several inflammatory comorbidities. This complexity highlights the multifaceted nature of the disease and supports the need for integrated, multimodal approaches to clinical management.

Discussion

Despite growing recognition of endometriosis as a complex and multifactorial condition affecting reproductive health, the potential role of the vaginal microbiome in the pathogenesis of endometriosis remains largely unexplored. In this study, we leveraged data from the citizen science-driven Isala platform to investigate whether endometriosis and its symptom status was associated with differences in vaginal microbiome composition. Investigating the broad range of conditions in this population is a critical step toward understanding potential shared pathologies with comorbidities and the underlying mechanisms of endometriosis.

Compared with population‑based studies reporting endometriosis prevalence of approximately 8-10%^55,56, the Isala cohort showed a lower overall prevalence of 3.5%. This may be explained by the setup of the Isala project, which was designed to capture a general, predominantly healthy population, as defined by the World Health Organization (i.e., a state of complete physical, mental, and social well-being, not merely the absence of disease or infirmity). Interestingly, despite the call for healthy women, a considerable number of participants with self-reported endometriosis enrolled in the study. Among these individuals, 1.6% experienced a significant physical burden of disease compared with controls (p < 0.001), indicating that disease burden is present even within this nominally healthy cohort. This highlights the complexity of self‑perceived health and may reflect broader societal factors such as the normalization of pain and stigma surrounding reproductive health, which can lead to prolonged diagnostic delays.

In our study, we analysed microbiome profiles from 95 individuals with self-reported endometriosis. Since lactobacilli are documented to have an anti-inflammatory and antimicrobial activity in the vagina^47,57,58 and because endometriosis is clearly linked to inflammation^20–23, we hypothesized that lactobacilli would be decreased in our Isala endometriosis cohort. Of interest, we observed no significant differences in Lactobacillus dominance between the endometriosis group and controls (p = 0.371). Importantly, the presence of a Lactobacillus-dominated microbiome does not necessarily imply the absence of an underlying inflammatory condition. Elevated oestrogen levels, characteristic of the hormonal milieu in endometriosis^59–61, promote glycogen accumulation in vaginal epithelial cells⁶². The breakdown of glycogen releases simple sugars that serve as a substrate for Lactobacillus species⁴⁵, potentially supporting their dominance even in the presence of underlying inflammation or disease. This could explain why, despite the presence of endometriosis, the vaginal microbiome may still appear Lactobacillus-dominated in our study. Similarly, we found no significant associations between endometriosis – regardless of symptom status – and alpha or beta diversity, nor were there significant differences in the relative abundance of specific bacterial taxa when compared to controls. Similar findings have also been observed by a prior study with 21 individuals with endometriosis using 16S rRNA amplicon sequencing⁶³. However, several other studies using the same technique have reported associations between specific bacterial taxa and endometriosis. For instance, a recent study comparing the vaginal microbiota of 35 individuals with endometriosis and chronic pelvic pain to 23 individuals with chronic pelvic pain alone observed an increased relative abundance of Streptococcus anginosus in the endometriosis group⁶⁴. Additionally, a larger study of 78 individuals with endometriosis found a higher abundance of Fusobacterium nucleatum in vaginal samples⁶⁵. Other studies, with sample sizes ranging from 10 to 37 individuals, have described elevated levels of Anaerococcus^66,67, Escherichia^35,68, Streptococcus^35,64,69, Blautia⁷⁰, as well as conflicting findings on the presence of Fannyhessea³⁵ (formerly classified as Atopobium) in vaginal samples of individuals with endometriosis. While these studies provide valuable insights, many are limited by methodological constraints, including a lack of adjustment for confounding factors such as age and correction for multiple testing. Reliance on laparoscopy or imaging for diagnosis, while ensuring clinical accuracy, restricts sample sizes in these studies, limiting the statistical power. Furthermore, while a few studies^35,67,71 have stratified vaginal microbiota profiles based on the revised American Society for Reproductive Medicine (rASRM) classification, it is well established that endometriosis disease stage does not correlate with symptom status^72,73. Given the substantial symptom burden experienced by individuals with endometriosis, stratifying cohorts based on symptom status – rather than disease stage – may facilitate the identification of clinically relevant microbiome-based biomarkers. Conversely, our present study includes a relatively large sample size, accounts for key confounders of the vaginal microbiome, and stratifies participants with endometriosis based on symptom status, offering a more rigorous and comprehensive assessment of microbial taxa potentially associated with endometriosis. Through extensive health questionnaires, we observed significant differences in the prevalence of certain comorbid conditions between individuals with and without self-reported endometriosis. For instance, irritable bowel syndrome was reported more than twice as often among individuals with endometriosis (11.1%) compared to controls (4.8%, OR 2.57, 95% CI 1.43 - 4.36, p = 0.008), which is consistent with ranges reported in previous studies and might reflect shared pathogenic mechanisms, such as mast cell activation^74,75. In addition, migraine was found to be significantly associated with endometriosis (OR = 3.75 95% CI 1.38-8.60, p = 0.025), exceeding the effect sizes reported in previous nationwide studies (OR 1.70, 95% CI 1.59-1.82 in Yang et al. 2012; OR 1.50, CI 95% 1.29-1.74 in Gete et al. 2023)^76,77. Dyspareunia affected 34.0% of individuals with endometriosis compared to 25.1% among controls (OR 1.67, 95% CI 1.14-2.40, p = 0.033), which aligns with previous findings from Singh et al. (2020) (38.3% vs. 17.7%)⁷⁸. Interestingly, our data showed a nearly three-fold higher odds ratio of PMOS among participants with endometriosis (16.9%) compared to controls (4.7%, 95% CI 1.71-4.78, p < 0.001), with the latter falling within the globally reported prevalence range of 3–11%, depending on the diagnostic criteria used⁷⁹. Notably, the association we observed was approximately twice as high as that reported in both a population-based cohort (n = 127) and an operative cohort (n = 473)⁸⁰. Recent findings reported a slightly higher PMOS co‑occurrence of 24.5% among individuals with endometriosis, further supporting the relevance of this association⁸¹. Shared risk genes and overlapping inflammatory pathways, characterized by dysregulated cytokines and chemokines, between endometriosis and PMOS may partly explain the observed comorbidity, as has been reported before^82–84. At the same time, the observed association should be interpreted with caution, as differences in healthcare utilization and diagnostic opportunities may also contribute. PMOS is commonly diagnosed using vaginal ultrasound, and individuals undergoing imaging as part of endometriosis evaluation may therefore have increased opportunities for PMOS detection. Furthermore, we found that among women trying to conceive, 47.2% with current endometriosis-related symptoms and 36.1% with past endometriosis-related symptoms reported requiring fertility treatment compared to 7.7% controls (p < 0.001). While these rates fall within ranges described in literature, it is important to highlight that our findings are derived from a population-based cohort, unlike most previous studies with subfertility cohorts^85,86, which may overestimate the true co-occurrence of subfertility in individuals with endometriosis. In addition, endometriosis is often diagnosed during fertility evaluations, which may lead to increased detection among individuals presenting with subfertility and thus partially inflate the observed association.

This study has several limitations. First, the control group may include asymptomatic individuals with undiagnosed endometriosis, which could lead to an underestimation of the true prevalence of endometriosis in our cohort and distort the group comparisons. Vice versa, the endometriosis group may contain participants without a clinically confirmed diagnosis, some of whom may be false positives. Second, individuals with severe forms of endometriosis, both in terms of clinical extent and impact on quality of life, may be underrepresented in our cohort considering the setup of the Isala project which was intentionally designed to characterize the vaginal microbiome in a large and predominantly healthy population. Third, the available data lacked sufficient resolution to further characterize the endometriosis subgroups, and information on treatment history or symptom management was not available. Fourth, this study did not account for the potential influence of prior or current treatments on the reported symptom burden and healthcare utilization, which may have affected our findings. Finally, although we did not observe a significant association between the vaginal microbiome composition and endometriosis, functional differences at the metabolic or transcriptomic level may still be present and relevant.

Despite inherent limitations, self-reported data remains a valuable tool in endometriosis research, particularly given the challenges and delays in obtaining formal diagnoses. Many individuals experience symptoms for years before receiving clinical confirmation, and self-reporting can offer a more immediate and inclusive snapshot of lived experiences across diverse populations. Moreover, capturing self-reported symptom severity and comorbid conditions can help identify underrecognized patterns and generate hypotheses for further study in clinical cohorts. Taken together, our findings do not support an association between the vaginal microbiome composition and endometriosis. However, we identified a pronounced co-occurrence of PMOS with endometriosis, in addition to other comorbid conditions. These results suggest potential shared mechanisms worth further investigating and demonstrate the value of citizen science in advancing endometriosis research. We suggest that follow-up research should move beyond microbial composition focusing on functional aspects such as host-microbiome interactions, immune profiling, and metabolomic signatures, as well as explore anatomical sites beyond the vagina, which may offer more promising pathways toward identifying biomarkers or therapeutic targets.

Methods

Study cohort and data collection

This study was conducted in accordance with the ethical principles of the Declaration of Helsinki. The study protocol received ethical approval from the Ethical Committee of the Antwerp University Hospital/University of Antwerp (reference number B300201942076) and is registered on clinicaltrials.gov under the unique identifier NCT04319536. Informed consent was obtained from all participants.

This cross-sectional study draws its study cohort from the Isala citizen science project (https://isala.be/en/). For this study, 144 participants with self-reported endometriosis were included and stratified into two groups based on their symptom status: individuals experiencing current endometriosis-related symptoms (n = 66) and individuals experiencing symptoms in the past but without active symptoms during the study (n = 78). This was complemented by a group of 3,337 participants without reported endometriosis, excluding individuals who reported dysmenorrhoea and requiring NSAIDs during menstruation. Menopausal individuals were excluded from this study. All participants filled in a health questionnaire with General Data Protection Regulation-compliant questions on the Qualtrics platform. Vaginal swabs were collected and processed as described before and stored in the in-house biobank decentralized hub to comply with the most recent GDPR-regulations in Belgium on biobanking human samples (KB 2018/30209)⁴⁰. A total of 95 microbiome profiles were obtained from the endometriosis cohort, including 43 profiles from individuals with current endometriosis-related symptoms and 52 profiles from individuals with past symptoms. Additionally, 2,279 microbiome profiles were collected from participants without self-reported endometriosis.

16S rRNA amplicon sequencing, reference database and quality control

Vaginal swabs were collected and processed as previously described by Lebeer et al. ⁴⁰. Briefly, DNA was extracted using the DNeasy PowerSoil Pro Kit (Qiagen), followed by amplification of the V4 region of the 16S rRNA gene using standard barcoded primers adapted for dual-index sequencing. PCR products were purified, quantified, and pooled in equimolar concentrations to generate sequencing libraries. Libraries were sequenced using dual-index paired-end sequencing on an Illumina MiSeq platform, including appropriate negative controls for DNA extraction and PCR. For the construction of the custom 16S reference database and processing of amplicon sequencing data, a refined taxonomic framework was developed to improve resolution within the Lactobacillus genus by defining subgenera based on phylogenetic relationships. A custom 16S rRNA reference database was generated using sequences from the Genome Taxonomy Database (GTDB) and adapted for use with DADA2⁸⁷. Sequence quality control and processing were conducted using the DADA2 pipeline, including filtering of low-quality reads, merging of paired-end reads, and removal of chimeras. Taxonomic assignment was performed using a custom 16S rRNA reference database, followed by reclassification steps to align with updated Lactobacillaceae taxonomy and the defined Lactobacillus subgenera. Additional quality control steps included removal of non-bacterial and low-quality amplicon sequence variants, as well as filtering of samples based on read count and normalized sequencing depth. For full methodological details, we refer to Lebeer et al. ⁴⁰.

Embeddings

To explore the vaginal microbiome composition of participants with and without self-reported endometriosis three different embedding methods were applied using Bray-Curtis dissimilarity based on the relative abundances of (sub)genera within samples: t-distributed stochastic neighbour embedding (t-SNE), uniform manifold approximation and projection (UMAP), and principal coordinates analysis (PCOA). Plots were generated with ggplot2 R package.

Statistical analyses

Through survey data, we compared demographic variables between participants with and without self-reported endometriosis, further distinguishing between individuals reporting current versus past endometriosis-related symptoms. Normality of continuous variables including age, BMI, and number of pregnancies was assessed using the Shapiro-Wilk test. None of these variables were normally distributed and were therefore presented as median and interquartile range (IQR). For the ordinary variable ‘self-rated health’ the mean and standard deviation were calculated based on a five-point ordinary scale (very good = 5, good = 4, fair = 3, poor = 2, very poor = 1). Depending on the data type of the variables, an appropriate age-adjusted model (linear, logistic, Poisson, multinomial, or ordinal regression) was implemented. Overall significance was calculated through likelihood ratio-tests for categorical variables followed by correction for multiple testing through the Benjamini-Hochberg procedure.

Alpha and beta diversity measures were compared as previously described by Lebeer et al. ⁴⁰. The differential abundance of taxa was performed using the in-house R package multidiffabundance, version 0.0.1 (publicly available at https://github.com/thiesgehrmann/multidiffabundance), using Maaslin2⁸⁸, Limma⁸⁹, DESeq2⁹⁰, ANCOM-BC⁹¹, and a linear regression on the centered log-ratio (CLR) transformed abundance data, reporting the consensus of the five tools. Each test was adjusted for technical confounders, age, recent sexual intercourse (last 24 hours), number of pregnancies, phase of cycle, use of hormonal contraceptives, and library read concentration. Correction for multiple testing was applied within each tool using the Benjamini–Hochberg false discovery rate procedure. Data was processed in R version 4.2.2 using the in-house developed package Tidytacos⁹² and the tidyverse set of packages. Results were visualised in Python 3, selecting only a subset of taxa based on their known importance in the vaginal microbiome.

To investigate associations between endometriosis and a range of symptoms, infections, and clinical conditions, we conducted multiple regression analyses, with all outcomes defined as lifetime (‘ever’) self-reported occurrence of the respective infection, condition, or vaginal symptom. The selection of potential confounders was determined using a directed acyclic graph (DAG; Supplementary Fig. 2), which allows the identification of the minimum adjustment required, while avoiding inappropriate ‘overadjustment’ for mediator variables⁹³. The following covariates were included in our models: age, BMI and education level. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated and presented in a forest plot. P-values were corrected for multiple testing through the Benjamini–Hochberg procedure.

Author contributions

S.A., S.W., E.O., G.D., V.V., and S.L. designed the study and worked on the conceptualization of the research project. S.A. and S.L. worked on the survey set-up. I.R. and T.G. cleaned the answers. S.A. and S.L. carried out the experimental and logistical work. I.D.B. was responsible for the biobanking of all collected samples. I.R. analysed the health questionnaire data and performed the statistical analyses. T.G., T.V.R., and S.W. processed the sequencing data and performed the biostatistical analyses. T.V.R. adjusted the microbiome data to the most recent taxonomy. I.R. and T.G. worked on the visualisations. I.R., T.G., S.A., I.D.B., C.N.A., T.V.R., S.W., F.D., and S.L. contributed to the interpretation of the results. I.R. and S.L. wrote the original manuscript. All authors contributed to reviewing and editing of the final manuscript.

Data availability

Sequencing data are available at the European Nucleotide Archive (ENA) under bioproject PREB50407. Sample metadata are available with access control through the European Genome–Phenome Archive (EGA) under dataset ID EGAD00001009890. Access is granted as described, upon agreement to the harmonised Data Access Agreement developed by EU-STANDS4PM (European Union standards for in silico models for personalised medicine; https://doi.org/10.6084/m9.figshare.23904300).

Competing interests

S.L. declares to be a voluntary academic board member of the International Scientific Association on Probiotics and Prebiotics (ISAPP, www.isappscience.org), cofounder of YUN, aMylla, and scientific advisor for Freya Biosciences. The team of S.L. declares research funding from YUN, Bioorg, Puratos, DSM I-Health and Lesaffre/Gnosis. None of these organizations or companies were involved in the design or data analysis of this study, which was fully funded by the university, governmental, and European funding. S.A. declares to be a voluntary member of the student and fellows association of ISAPP. The other authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s44294-026-00144-9.