Diagnostic value of MiRNAs in endometriosis: a systematic review and meta-analysis

A total of 1,394 publications were initially identified. After removing duplicates, 887 records remained for title and abstract screening. Based on the screening of titles and abstracts, 815 articles were excluded. The full texts of the remaining 72 articles were assessed for eligibility. Of these, 41 studies were excluded for various reasons, as detailed in Fig. 1 . Ultimately, 30 studies involving a total of 3,274 participants were included in this systematic review. The key characteristics of the included studies are summarized in Table 1 [ 24 – 27 , 36 – 61 ]. Fig. 1 Flowchart depicting selection process of included studies Flowchart depicting selection process of included studies Table 1 Clinical and demographic characteristics of studies included in the meta-analysis Included studies Country Patients size (controls) rASRM stage I-II (III-IV) Age (mean \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\pm\:$$\end{document} SD) EMs BMI kg/m 2 (controls) Source of MicroRNA Types of MicroRNA AUC SEN SPE Junmei Wang 2025 [ 36 ] China 100 (80) 59 (41) 32.19 ± 1.43 23.93 ± 1.00 (24.00 ± 1.40) Peripheral blood hsa-miR-375-3p 0.84 0.78 0.75 Shan Jiang 2024 [ 37 ] China 119 (96) 62 (57) 34.01 ± 5.46 23.03 ± 2.60 (22.95 ± 2.25) Peripheral blood has-miR-134-5p 0.82 0.90 0.63 Lucía Chico-Sordo 2024 [ 38 ] Spain 67 (72) 7 (60) 35.69 ± 6.47 22.8 ± 3.1 (24.1 ± 3.0) Peripheral blood has-miR-30c-5p 0.84 0.78 0.91 Izabela Walasik 2023 [ 39 ] Poland 24 (25) 0 (24) 31.20 22.1 (22.5) Peripheral blood hsa-miR-125b-5p, hsa-miR-199a-3p, hsa-miR-451a, hsa-miR-3613-5p 0.83 0.88 0.76 Chunli Lin 2023 [ 25 ] China 80 (80) 32 (48) 31.95 ± 3.85 27.39 ± 5.63 (26.76 ± 6.25) Peripheral blood has-miR-17-5p, has-miR-424-5p 0.94 0.94 0.89 Guansheng Chen 2023 [ 40 ] China 155 (77) 48 (107) 31.77 ± 7.04 20.9 ± 2.9 (20.9 ± 2.7) Peripheral blood hsa-miR-199a-3p, hsa-miR-122-5p, hsa-miR-145-5p, hsa-miR-141-5p, hsa-miR-542-3p, hsa-miR-9-5p 0.87 0.77 0.76 Alexandra Perricos 2022 [ 41 ] Austria 17 (17) 7 (10) 36.2 ± 8.23 22.3 ± 3.5 (25.2 ± 5.1) Peripheral blood hsa-miR-135a 0.81 0.69 0.81 Seyed Danial 2025 [ 42 ] Iran 25 (25) NA 36.8 ± 6.29 24.6 ± 2.99 (25.7 ± 7.6) Peripheral blood hsa-let-7d-3p, hsa-miR-9-5p, hsa-miR-17-5p, hsa-miR-20a-5p, hsa-miR-122-5p, hsa-miR-125b-5p, hsa-miR-135a-5p, hsa-miR-141-5p, hsa-miR-145-5p, hsa-miR-185-5p, hsa-miR-199b-3p, hsa-miR-200c-3p, hsa-miR-224-5p, hsa-miR-340-5p, hsa-miR-342-3p, hsa-miR-451a 0.85 0.88 0.72 Yohann Dabi 2022 [ 43 ] France 153 (47) 80 (73) 31.11 ± 11.52 24.36 ± 8.38 (24.84 ± 11.10) Peripheral blood hsa-miR-548j-5p, hsa-miR-29b-1-5p, hsa-miR-548p-5p, hsa-miR-548 L-5p, hsa-miR-3913-5p, hsa-miR-124-3p 0.68 0.71 0.65 Narges Zafari 2021 [ 44 ] Iran 25 (25) 15 (10) 33.68 ± 5.72 NA Peripheral blood hsa-miR-199b-3p, hsa-miR-224-5p, hsa-let-7d-3p 0.99 0.92 1.00 Lei Wang 2021 [ 45 ] China 80 (66) 60 (20) 32.00 27 (20) Endometrial tissue hsa-miR-451-5p 0.91 0.59 0.97 Sema Misir 2021 [ 46 ] Turkiye 71 (65) 48 (23) 37.01 ± 8,70 NA Peripheral blood miR-34a-5p, miR-200c 0.69 0.79 0.49 Afshin Bahramy 2021 [ 47 ] Iran 30 (30) 0 (30) 32.5 ± 1.61 NA Peripheral blood miR-340-5p, miR-381-3p 0.72 0.70 0.72 Lu Zhang 2020 [ 24 ] China 20 (20) NA 28.64 ± 4.59 NA Peripheral blood hsa-miR-34a-5p, hsa-miR-200c-3p 0.86 0.75 0.90 Wang Huimin 2020 [ 48 ] China 99 (85) 52 (47) 35.37 ± 6.16 35.68 ± 6.23 (35.02 ± 6.11) Peripheral blood hsa-miR-30b-5p, hsa-miR-30d-5p 0.86 0.92 0.68 Mohammad Hossein Razi 2020 [ 49 ] Iran 25 (25) NA 30.3 ± 5.63 21.8 ± 3.1 (25.2 ± 5) Peripheral blood hsa-miR-185-5p 0.92 0.81 0.90 Elahe Papari 2020 [ 50 ] Iran 25 (28) NA 33.00 NA Peripheral blood hsa-miR-199a-3p, hsa-miR-143-3p, hsa-miR-340-5p, hsa-let-7b-5p, hsa-miR-21-5p, hsa-miR-17-5p, hsa-miR-20a-5p, hsa-miR-103a-3p 0.95 0.92 0.86 Sarah Moustafa 2020 [ 51 ] USA 41 (59) 18 (23) 35.75 ± 7.85 28.1 ± 7.5 (30.4 ± 7.5) Peripheral blood hsa-miR-125b-5p, hsa-miR-150-5p, hsa-miR-451a, hsa-miR-3613-5p, hsa-let-7b-5p 0.92 0.90 0.91 Cheng-lei Gu 2020 [ 27 ] China 29 (31) NA 31.00 NA Peripheral blood hsa-let-7a-5p, hsa-let-7b-5p, hsa-let-7d-5p, hsa-let-7f-5p, hsa-let-7i-5p, hsa-miR-199a-3p, hsa-miR-320a, has-mir-320b, hsa-miR-320c, hsa-miR-320d, hsa-miR-320e, hsa-miR-328-3p, hsa-miR-331-3p 0.90 0.92 0.89 A Vanhie 2019 [ 52 ] Belgium 60 (30) NA 31.33 ± 4.09 21.6 ± 2.7 (24.0 ± 3.5) Peripheral blood hsa-let-7d-5p, hsa-miR-21-5p, hsa-miR-28-5p 1 1 1 Wang Fang 2018 [ 53 ] China 80 (51) NA 33.00 NA Peripheral blood has-miR-17 0.75 0.70 0.59 Ahmed M Maged 2018 [ 26 ] Cairo Egypt 45 (35) 15 (30) 29.56 ± 3.9 21.71 ± 2.94 (21.32 ± 2.70) Peripheral blood has-miR-122, has-miR-199a 0.96 0.96 0.91 Warren B Nothnick 2016 [ 54 ] USA 40 (41) NA 33.00 NA Endometrial tissue hsa-miR-451a 0.86 0.85 0.85 Emine Cosar 2016 [ 55 ] New Haven 24 (24) NA 32.62 ± 8.09 NA Peripheral blood hsa-miR-125b-5p, hsa-miR-150-5p, hsa-miR-342-3p, hsa-miR-143-3p, hsa-miR-500a-3p, hsa-miR-18a-5p, hsa-miR-6755-3p, hsa-miR-3613-5p 0.97 0.96 0.92 Wen-Tao Wang 2012 [ 56 ] China 60 (25) 22 (38) 31.28 NA Peripheral blood hsa-miR-199a-3p, hsa-miR-122-5p, hsa-miR-145-3p, hsa-miR-542-3p 0.88 0.96 0.69 Swati 2012 [ 57 ] Pittsburgh 33 (20) NA 46.84 ± 17.24 NA Peripheral blood hsa-miR-16-5p, hsa-miR-191-5p, hsa-miR-195-5p 0.9 0.88 0.6 Shuang-zheng Jia 2012 [ 58 ] China 23 (23) 0 (23) 33.1 ± 6.08 NA Peripheral blood hsa-miR-17-5p, hsa-miR-20a-5p, hsa-miR-22-3p 0.90 0.91 0.78 Darya A. 2021 [ 59 ] Ukraine 64 (24) 29 (35) 33.3 ± 6.03 20.72 + 2.3 (20.72 + 2.3) Endometrial tissue hsa-let-7 0.89 0.93 0.83 Sofiane 2023 [ 60 ] France 159 (41) 99 (60) 33.2 ± 8.17 24.7 ± 5.5 (24.2 ± 4.8) Saliva 109 miRNAs (Random forest model) 0.96 0.96 0.95 Sofiane 2022 [ 61 ] France 153 (47) 80 (73) 31.11 ± 11.52 24.36 ± 8.38 (24.84 ± 11.10) Saliva Model 0.99 0.97 1 Alexandra Perricos 2022 [ 41 ] Vienna 17 (17) 7 (10) 36.2 ± 8.23 22.3 ± 3.6 (25.2 ± 5.1) Saliva hsa-mir-135a 0.80 0.71 0.88 NA Not Available; Clinical and demographic characteristics of studies included in the meta-analysis 100 (80) 59 (41) 23.93 ± 1.00 (24.00 ± 1.40) 119 (96) 62 (57) 23.03 ± 2.60 (22.95 ± 2.25) 67 (72) 7 (60) 22.8 ± 3.1 (24.1 ± 3.0) 24 (25) 0 (24) 22.1 (22.5) 80 (80) 32 (48) 27.39 ± 5.63 (26.76 ± 6.25) 155 (77) 48 (107) 20.9 ± 2.9 (20.9 ± 2.7) Alexandra Perricos 2022 [ 41 ] 17 (17) 7 (10) 22.3 ± 3.5 (25.2 ± 5.1) 25 (25) 24.6 ± 2.99 (25.7 ± 7.6) 153 (47) 80 (73) 24.36 ± 8.38 (24.84 ± 11.10) 25 (25) 15 (10) 80 (66) 60 (20) 27 (20) 71 (65) 48 (23) 30 (30) 0 (30) 20 (20) 99 (85) 52 (47) 35.68 ± 6.23 (35.02 ± 6.11) 25 (25) 21.8 ± 3.1 (25.2 ± 5) 25 (28) 41 (59) 18 (23) 28.1 ± 7.5 (30.4 ± 7.5) 29 (31) 60 (30) 21.6 ± 2.7 (24.0 ± 3.5) 80 (51) 45 (35) 15 (30) 21.71 ± 2.94 (21.32 ± 2.70) Warren B Nothnick 2016 [ 54 ] 40 (41) 24 (24) 60 (25) 22 (38) 33 (20) 23 (23) 0 (23) 64 (24) 29 (35) 20.72 + 2.3 (20.72 + 2.3) 159 (41) 99 (60) 24.7 ± 5.5 (24.2 ± 4.8) 153 (47) 80 (73) 24.36 ± 8.38 (24.84 ± 11.10) Alexandra Perricos 2022 [ 41 ] 17 (17) 7 (10) 22.3 ± 3.6 (25.2 ± 5.1) NA Not Available; This review included a total of 30 publications encompassing 118 diagnostic datasets. These studies were conducted in diverse countries and regions, involving a total of 3,274 participants, including 1,943 patients with endometriosis and 1,331 non-endometriotic controls. The publication years ranged from 2012 to 2025, reflecting the temporal evolution of research in this field. In terms of sample type, 112 datasets used blood specimens, 3 used endometrial tissue, and 3 used saliva samples. A wide range of miRNAs (93 in total) were investigated across studies. Among them, 16 diagnostic datasets evaluated combinations of multiple miRNAs for endometriosis diagnosis, such as hsa-miR-224-5p combined with hsa-let-7d-3p; hsa-miR-224-5p combined with hsa-miR-199b-5p; and a triple panel including hsa-miR-199a-5p, hsa-miR-122-5p, and hsa-miR-145-5p. The remaining studies focused on the diagnostic value of individual miRNAs. Of the 30 included publications, 16 originated from Asian countries and 14 from Europe and North America. The methodological quality of the included studies was evaluated using the QUADAS-2 tool in RevMan 5.4. The results are presented in Fig.  2 . Overall, most studies exhibited a low risk of bias and good applicability across the four QUADAS-2 domains. However, some studies showed either unclear or high risk of bias in specific areas, and their findings should therefore be interpreted with caution. Fig. 2 Risk of bias and applicability concerns summary Risk of bias and applicability concerns summary Spearman correlation analysis was conducted using Meta-Disc software to assess the presence of a threshold effect [ 62 ]. The correlation coefficient between the logit of sensitivity and the logit of (1-specificity) was − 0.143 ( P = 0.123, > 0.05), indicating significant threshold effect. In addition, the summary receiver operating characteristic (SROC) curve (Fig. 3 did not display a typical “shoulder-arm” shape, further confirming the absence of a threshold effect. Fig. 3 Summarized receiver operating characteristic curve of microRNAs in EMs Summarized receiver operating characteristic curve of microRNAs in EMs A random-effects model was applied to pool diagnostic effect sizes due to substantial heterogeneity observed in both sensitivity and specificity. Between-study heterogeneity was substantial (τ²_sensitivity = 0.70; τ²_specificity = 0.70; RE correlation = 0.11). The 95% prediction interval for sensitivity was 0.61–0.93, and for specificity 0.58–0.91. The forest plots for sensitivity and specificity are presented in Fig.  4 . The results of the meta-analysis showed that miRNAs have moderate-good diagnostic accuracy for endometriosis. The pooled sensitivity was 0.82 (95% CI: 0.79–0.84), and the pooled specificity was 0.79 (95% CI: 0.76–0.82). The combined positive likelihood ratio (SPLR) was 3.9 (95% CI: 3.3–4.5), the negative likelihood ratio (SNLR) was 0.23(95% CI: 0.20–0.27), and the diagnostic odds ratio (DOR) was 17 (95% CI: 13–22). The area under the summary receiver operating characteristic (SROC) curve was 0.87 (95% CI: 0.84–0.90), as shown in Fig.  3 . These results collectively suggest that miRNAs demonstrate strong diagnostic performance and hold considerable potential as biomarkers for endometriosis. Fig. 4 Forest plots depicting the sensitivity and specificity of microRNAs in EMs Forest plots depicting the sensitivity and specificity of microRNAs in EMs Due to significant heterogeneity observed in this meta-analysis, univariate meta-regression and subgroup analyses were performed. For the diagnostic accuracy of miRNAs in endometriosis, the following variables were considered: sample size (≥ 100 vs. <100), gene expression pattern (upregulated vs. downregulated), sample type (peripheral blood vs. non-blood), geographic region (Asia vs. non-Asia), age (≥ 40 vs. <40), BMI (≥ 30 kg/m² vs. <30 kg/m²), and miRNA classification (single vs. combinative). Subgroup analysis revealed that among the dysregulated miRNAs, 39 were upregulated and 72 were downregulated in endometriosis. The diagnostic performance of miRNAs was comparable between upregulated and downregulated groups. Combined miRNA panels demonstrated superior diagnostic accuracy compared to single miRNAs, suggesting that multi-marker approaches may be more suitable for the diagnosis of endometriosis. Univariate meta-regression results indicated that gene expression regulation ( p  < 0.001), miRNA classification ( p  < 0.001), sample size ( p  < 0.001), and BMI ( p  < 0.001) were significant sources of heterogeneity (Fig.  5 ). In contrast, ethnicity ( p  = 0.66), sample type ( p  = 0.30), and age ( p  = 0.59) were not significant contributors to heterogeneity. Detailed results are provided in Table  2 . Fig. 5 MiRNA was used to distinguish EMs patients from the healthy control group MiRNA was used to distinguish EMs patients from the healthy control group Table 2 Assessment of diagnostic accuracy and heterogeneity in subgroup analysis Parameter Category Number of studies Sensitivity P value Specificity P value ΔSEN/ ΔSPE Gene expression UP 39 0.81[0.76–0.86] p  < 0.001 0.83[0.79–0.87] p   100 17 0.74[0.66–0.81] p  < 0.001 0.72[0.63–0.81] p  < 0.001 -0.09/ -0.08 ≤ 100 101 0.83[0.81–0.86] 0.80[0.77–0.83] miRNA types single 101 0.80 [0.77–0.83] p  < 0.001 0.78[0.75–0.81] p  30 kg/m 2 6 0.77[0.73–0.82] p  < 0.001 0.75[0.70–0.80] p  < 0.001 -0.12/ -0.07 ≤ 30 kg/m 2 49 0.90[0.83–0.96] 0.82[0.71–0.94] These p-values are unadjusted for the multiple comparisons made. Findings should be interpreted as exploratory due to the potential for Type I error inflation. ΔSEN and ΔSPE represent the difference in values (Category in the second row minus Category in the first row) between subgroups Assessment of diagnostic accuracy and heterogeneity in subgroup analysis 0/ 0.09 -0.09/ -0.08 0.10/ -0.07 -0.12/ -0.07 These p-values are unadjusted for the multiple comparisons made. Findings should be interpreted as exploratory due to the potential for Type I error inflation. ΔSEN and ΔSPE represent the difference in values (Category in the second row minus Category in the first row) between subgroups In the sensitivity analysis of miRNAs for the diagnosis of endometriosis, goodness-of-fit and bivariate normality tests (Fig.  6 a and b) confirmed the robustness of the bivariate model. Influence analysis identified 17 studies with potential impact, and further outlier detection revealed 13 specific outliers (Fig.  6 c and d). Fig. 6 Meta-regression and subgroup analysis. a Goodness of fit, b Bivariate normality, c Influence analysis, d Outlier detection Meta-regression and subgroup analysis. a Goodness of fit, b Bivariate normality, c Influence analysis, d Outlier detection After excluding the 9th study, the recalculated SSEN was 0.81 (95% CI: 0.79–0.84), SSPE was 0.79 (95% CI: 0.76–0.82), SPLR was 3.9 (95% CI: 3.4–4.5), the SNLR was 0.23 (95% CI: 0.20–0.27), DOR was 17 (95% CI: 13–22), AUC remained stable at 0.87 (95% CI: 0.84–0.90). After excluding the 11th study, the recalculated SSEN was 0.82 (95% CI: 0.79–0.84), SSPE was 0.79 (95% CI: 0.76–0.82), SPLR was 3.9 (95% CI: 3.3–4.5), the SNLR was 0.23 (95% CI: 0.20–0.27), DOR was 17 (95% CI: 13–21), AUC remained stable at 0.87 (95% CI: 0.84–0.90). After excluding the 12th study, the recalculated SSEN was 0.82 (95% CI: 0.79–0.84), SSPE was 0.78 (95% CI: 0.75–0.81), SPLR was 3.8 (95% CI: 3.4–4.4), the SNLR was 0.23 (95% CI: 0.20–0.27), DOR was 16 (95% CI: 13–21), AUC remained stable at 0.87 (95% CI: 0.84–0.90). As shown in Table  3 , we summarized the results obtained after sequentially excluding each outlier study along with the rationale for their removal. Overall, the pooled estimates changed only minimally after excluding these studies, indicating the robustness of our meta-analytic findings. Table 3 Summary of data results and reasons after sequentially excluding anomalous data ID SSEN SSPE SPLR SNLR DOR AUC Exclusion Reason 9 0.81 0.79 3.9 0.23 17 0.87 Abnormal standardized residuals 11 0.81 0.79 3.9 0.23 17 0.87 Excessively high Cook’s distance 12 0.82 0.79 3.9 0.23 17 0.87 Excessively high Cook’s distance 13 0.82 0.78 3.8 0.23 16 0.87 Poor model fit 14 0.82 0.79 3.8 0.23 17 0.87 Poor model fit 15 0.82 0.78 3.9 0.23 16 0.87 Poor model fit 44 0.82 0.79 3.9 0.23 17 0.87 Poor model fit 47 0.82 0.79 3.9 0.23 17 0.87 Poor model fit 50 0.82 0.79 3.8 0.23 16 0.87 Excessively high Cook’s distance 72 0.81 0.78 3.7 0.24 15 0.86 Excessively high Cook’s distance 94 0.81 0.78 3.7 0.24 15 0.86 Excessively high Cook’s distance 105 0.81 0.77 3.6 0.24 14 0.86 Abnormal standardized residuals 117 0.81 0.77 3.6 0.25 15 0.86 Abnormal standardized residuals Summary of data results and reasons after sequentially excluding anomalous data Given that several included studies contained zero cells, which may potentially impact the stability of the estimates, we further examined the consistency of the results by refitting the model after applying a continuity correction of 0.5 to all fourfold tables. After recalculation, the SSEN was 0.79 (95% CI: 0.76–0.81), the SSPE was 0.76 (95% CI: 0.73–0.79), the SPLR was 3.3 (95% CI: 2.9–3.7), the SNL R was 0.28 (95% CI: 0.24–0.31), the DOR was 12.0 (95% CI: 9.6–15.0), and the AUC was 0.85 (95% CI: 0.83–0.87). Compared with the estimates from the original model, the diagnostic sensitivity, specificity, and DOR did not exhibit any qualitative changes, further supporting the stability of our results. Collectively, these findings demonstrate that the exclusion of influential studies had only a minimal effect on the overall estimates, thereby confirming the robustness and reliability of our meta-analysis results. Fagan’s nomogram (Fig.  7 ) illustrated that, with a pre-test probability of 20%, a positive likelihood ratio of 3.9 would increase the post-test probability to 49%, indicating that a positive result provides strong support for the diagnosis of endometriosis. Conversely, a negative likelihood ratio of 0.23 would reduce the post-test probability to 5%, suggesting that a negative result can effectively rule out the disease. These findings support the clinical utility of miRNA-based testing in guiding diagnostic decisions and reducing unnecessary interventions. Fig. 7 The fagan’s nomogram The fagan’s nomogram Deeks’ funnel plot asymmetry test yielded a p-value of 0.86 (> 0.05), indicating no significant publication bias (Fig.  8 ). This suggests that the included studies provide an objective representation of the diagnostic performance of miRNAs in endometriosis. However, due to the inherent subjectivity of funnel plot interpretation and the potential influence of small sample sizes, the results should be interpreted with caution. Fig. 8 The results of publication bias The results of publication bias

This meta-analysis was registered in the PROSPERO database ( https://www.crd.york.ac.uk/PROSPERO ) under the ID CRD42025635369. The study was conducted and reported in accordance with the PRISMA 2020 (Preferred Reporting Items for Systematic Reviews and Meta-Analysesc) guidelines [ 29 ]. A comprehensive literature search was conducted in four major databases: PubMed, Cochrane Library, EMBASE, and Web of Science, to identify original studies evaluating the diagnostic value of miRNAs in endometriosis. The search covered all available literature from the inception of each database up to March 1, 2025. Keywords included terms related to “Endometriosis,” “MicroRNA OR miRNA OR miR,” and “Diagnosis.” Appropriate Boolean operators were used to combine search terms. For example, the search strategy used in PubMed was: “(Endometriosis) AND (MicroRNA OR miRNA OR miR) AND (Diagnosis).” Search strategies were adjusted as necessary for each database based on their specific indexing systems and functionalities. In addition to database searches, the reference lists of included articles, relevant review articles, and conference proceedings were manually screened to identify any additional eligible studies. This strategy aimed to ensure comprehensive coverage of all relevant literature. Detailed search strategies are provided in supplementary Table 1. The inclusion criteria for this meta-analysis were defined using a modified PECOS framework, which considers the study’s population, exposure, comparison, outcome, and study design. (1) Studies must include patients with histologically confirmed endometriosis, along with a control group comprising healthy individuals or patients with histologically confirmed non-endometriotic conditions; (2) Studies must evaluate the diagnostic value of miRNAs in endometriosis and provide sufficient data to assess diagnostic performance, including true positives (TP), false positives (FP), true negatives (TN), and false negatives (FN), enabling calculation of sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), and diagnostic odds ratio (DOR); (3) The samples must be of human origin, with clearly specified sample types (e.g., blood, endometrial tissue, saliva); (4) Only original research articles were included; reviews, commentaries, case reports, and conference abstracts were excluded; (5) Studies must be published in the English language. Exclusion criteria were as follows: (1) Non-original articles such as reviews, editorials, meta-analyses, letters, case reports or series, conference abstracts, study protocols, Non-English publications and commentaries; (2) Studies with incomplete or missing data, rendering key diagnostic indices unextractable; (3) Duplicate publications by the same research group with substantially overlapping content; (4) Studies based solely on animal or cellular experiments without human samples; (5) Studies with a total sample size (cases + controls) of ≤ 20; (6) Studies involving patients diagnosed with endometriosis during pregnancy. Two reviewers independently screened all retrieved records according to the predefined inclusion and exclusion criteria. Any disagreements were resolved through discussion with a third reviewer. EndNote 21 software (Clarivate, Philadelphia, PA, USA) was used to manage references. After removing duplicates, the titles and abstracts were screened, followed by full-text review to identify eligible studies. The following data were extracted: basic study characteristics (first author, year of publication, country/region); participant details (sample size, age range, disease stage, BMI, etc.); and diagnostic performance data (TP, FP, TN, FN) to compute sensitivity, specificity, PLR, NLR, and DOR. Additionally, the types of miRNAs investigated, their expression patterns, and sample sources were recorded. If essential information was missing, corresponding authors were contacted via email to obtain clarification. If any of the previously mentioned data were absent from the publications, they were recorded as “Unknown, UN”. If one study reported diagnostic results for two or more different types of miRNA, each miRNA test was considered as an independent study. In addition, the results of multiple miRNA combination assays were considered as independent study data. We conducted the primary meta-analysis using a bivariate random-effects model, which is robust for handling sparse data with zero cells. To assess the robustness of our findings, we performed several sensitivity analyses: (1) refitting the model after applying a continuity correction of 0.5 to all fourfold tables; and (2) excluding studies with standardized residuals exceeding an absolute value of 2. In this sensitivity assessment, these outlier studies were removed one at a time, followed by re-analysis to evaluate the stability of the pooled estimates. For data presented only in graphical form, numerical values were extracted using the Engauge Digitizer software (version 12.1; https://markummitchell.github.io/engauge-digitizer/ ). All miRNA names were standardized according to the miRBase database [ 30 ]. Two reviewers independently assessed the methodological quality of included studies using the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) tool [ 31 ]. Disagreements were resolved by discussion with a third reviewer until a consensus was reached. QUADAS-2 focuses on four key domains: (1) patient selection, (2) index test, (3) reference standard, and (4) flow and timing. Each domain is evaluated for risk of bias, and the first three domains are also assessed for applicability. Each item within the domains is categorized as having “low risk,” “high risk,” or “unclear risk” of bias. A “low risk” rating indicates well-designed and executed study components unlikely to bias results; “high risk” indicates serious flaws; and “unclear” signifies insufficient information to make a judgment. This comprehensive assessment enables an objective evaluation of study quality and provides a basis for the subsequent meta-analysis. Data were analyzed using RevMan 5.4 (The Nordic Cochrane Centre, The Cochrane Collaboration, Copenhagen, Denmark), Stata 18.0 (Stata CorpLLC, College Station, TX, USA), and Meta-Disc 1.4 software (XI Cochrane Colloquium, Barcelona, Spain). Spearman correlation analysis was conducted in Meta-Disc to assess the presence of threshold effects, where a strong positive correlation suggests a threshold effect. If no threshold effect was identified, a bivariate random-effects model [ 32 ] was used to pool diagnostic metrics, including pooled summary sensitivity (SSEN), summary specificity (SSPE), PLR, NLR, DOR, and their corresponding 95% confidence intervals (CIs). The area under the summary receiver operating characteristic (SROC) curve (AUC) was calculated to evaluate overall diagnostic accuracy. AUC was interpreted as follows: <0.7 (low accuracy), 0.7–0.9 (moderate accuracy), and ≥ 0.9 (high accuracy). Forest plots and SROC curves were generated accordingly. Sensitivity analyses, including influence analysis and outlier detection, were performed to evaluate the robustness of the findings. Between-study heterogeneity was assessed using Cochran’s Q test [ 33 ] and Higgins inconsistency index (I²) statistic [ 34 ]. A p-value 50% indicated significant heterogeneity [ 35 ], prompting further exploration through univariate meta-regression and subgroup analyses. Predefined subgroups included sample size (≥ 100 vs. <100), gene expression pattern (upregulated vs. downregulated), sample type (peripheral blood vs. non-blood), geographic region (Asia vs. non-Asia), age (≥ 40 vs. <40 years), BMI (≥ 30 vs. <30 kg/m²), and disease stage (Stage I/II vs. Stage III/IV), miRNA types (single vs. combinative). When the number of included studies for a particular index test was ≤ 10, a formal meta-analysis was not conducted. Instead, 2 × 2 contingency tables were constructed for each study, and sensitivity, specificity, AUC, and 95% CI were calculated individually. Publication bias was assessed using Deeks’ test. A symmetric funnel plot suggests the absence of publication bias, while asymmetry indicates possible bias. Deeks’ test statistically evaluates funnel plot symmetry, with p  > 0.05 suggesting no significant bias, and p  ≤ 0.05 indicating potential publication bias, necessitating cautious interpretation of the results.

In conclusion, miRNAs demonstrate promising potential as diagnostic biomarkers for endometriosis, particularly when used in combination, with moderate to good diagnostic accuracy (AUC ≈ 0.87), especially when used in multi-miRNA panels. Further high-quality studies are needed to validate the clinical utility of circulating miRNAs and to facilitate the development of more effective predictive models.

To our knowledge, this is the meta-analysis to systematically evaluate the diagnostic utility of miRNAs for endometriosis, incorporating a relatively large number of studies and a robust sample size. Our findings demonstrate that miRNAs exhibit strong diagnostic performance as biomarkers for EMs, with an area under the SROC curve of 0.87. The pooled sensitivity and specificity were 0.82 (95% CI: 0.79–0.84) and 0.79 (95% CI: 0.76–0.82), respectively, supporting the feasibility of using miRNAs as diagnostic molecular markers in clinical settings. These results indicate that miRNAs could provide a promising avenue for early and non-invasive detection of EMs. From a molecular perspective, miRNAs play key regulatory roles in the pathogenesis of endometriosis. Previous studies have demonstrated that specific miRNAs regulate genes associated with cell proliferation, invasion, angiogenesis, immune modulation, and hormonal responsiveness [ 63 – 66 ]. For instance, miR-210 suppresses the proliferation and migration of endometrial cells by targeting IGFBP3 [ 67 ]. Similarly, miR-34a-5p inhibits invasion, migration, colony formation, and stemness by directly binding to the 3’-UTR of the MMP-2 gene, thereby suppressing transcription and limiting disease progression [ 68 ]. Despite these advances, the precise mechanisms by which miRNAs contribute to EMs remain unclear. Differences in miRNA expression profiles across studies may reflect population heterogeneity, genetic background, environmental exposures, disease stage, sample processing, and detection methods. Additionally, interactions among miRNAs and the regulatory networks they form introduce additional complexity to mechanistic investigations [ 69 ]. Emerging evidence supports the role of miRNAs as potential non-invasive biomarkers. Several studies [ 70 , 71 ] have found that the expression levels of certain miRNAs are significantly lower in patients with endometriosis than in healthy controls. A recent study involving 67 patients with EMs and 60 healthy controls profiled 130 miRNAs in serum samples using a censored regression model, identifying significant expression differences that may aid diagnosis [ 72 ]. Other studies have identified miR-139-3p, miR-140-3p, and miR-629-5p as anti-mullerian hormone (AMH)-related biomarkers for predicting ovarian reserve recovery following perioperative dienogest (DNG) administration in patients with ovarian endometrioma (OE) [ 73 ]. Moreover, exosomal miRNAs isolated from tubal fluid were shown to be associated with tubal damage in EMs-related infertility, offering insights into reproductive pathophysiology and the mechanisms of tubal ultrastructural injury [ 74 ]. MiR-451 has also been associated with adverse IVF/ICSI-ET outcomes in patients with EMs [ 45 ]. These findings underscore the broader diagnostic and prognostic value of miRNAs not only in disease detection but also in the context of fertility preservation. Our analysis showed a pooled positive likelihood ratio (PLR) of 3.9, indicating that a positive miRNA test increases the probability of EMs by 3.9-fold. The negative likelihood ratio (NLR) was 0.23, suggesting that a negative result substantially reduces the probability of having EMs. The diagnostic odds ratio (DOR) reached 17 (95% CI: 13–21), with a narrow confidence interval not including 1, further supporting the robustness of these results. In 50% of the studies, the false positive rate (FPR) of miRNAs was below 15%, significantly lower than the 28% observed for CA125, indicating a reduced risk of misdiagnosis. This makes miRNAs a suitable tool for large-scale population screening, helping to alleviate unnecessary anxiety and reduce healthcare resource waste. In 94% of the studies, the positive predictive value (PPV) exceeded 85%, meaning that when a test result is positive, there is an 85% probability that the patient actually has the disease. This can effectively reduce unnecessary laparoscopies and optimize clinical decision-making. Additionally, in 42% of the studies, the false negative rate (FNR) was below 10%, suggesting a low risk of missed diagnoses and helping to prevent disease progression due to oversight. Collectively, these findings demonstrate that miRNAs show favorable performance and may complement existing markers such as CA-125; head-to-head comparative DTA studies are warranted. For detailed data, please refer to Supplementary Table 2. Logistic regression analysis showed that combined panels of miRNA offered better diagnostic accuracy than individual miRNAs. However, the composition of diagnostic panels varied across studies, and nearly half of the included reports did not specify clear diagnostic thresholds or used similar cutoff values. These differences may have affected the comparability of diagnostic performance and, consequently, the robustness of the findings. Future studies should aim to standardize threshold reporting, validate diagnostic cutoffs across populations, and identify optimal miRNA combinations that can be applied in diverse clinical settings. Subgroup analyses indicated comparable diagnostic performance for both upregulated and downregulated miRNAs. However, this observation should be interpreted cautiously. Evidence from five studies indicates that the baseline expression level of miR-199a is significantly higher in Asian populations compared to European and American populations [ 26 , 39 , 40 , 50 , 56 ]. Further multi-center, multi-ethnic studies are warranted to validate these preliminary findings and clarify whether ethnicity-related miRNA expression differences truly exist. However, to successfully integrate these promising miRNA biomarkers into routine clinical diagnostic pathways, significant work in assay development and standardization is required. Future efforts must focus on establishing robust, reproducible detection platforms. In this regard, the field can draw valuable parallels from the development of biosensors for autoimmune diseases, which have advanced by focusing on high sensitivity, specificity, and cost-effectiveness for the accurate detection of biomarkers in bodily fluids [ 75 ]. Similarly, developing standardized, high-performance detection systems for miRNA panels, guided by evidence-based guidelines and consensus, is a critical next step toward clinical translation. A higher body mass index (BMI) greater than 30 kg/m² in patients with EMs was associated with greater heterogeneity and potential diagnostic relevance, although further clinical validation is required. Differences among sample types should also be considered in future model optimization. Although histopathology remains the gold standard for the diagnosis of EMs, its invasive nature and high cost limit widespread use [ 76 ]. Imaging techniques such as ultrasound and MRI are suboptimal tools for early diagnosis. Therefore, identifying novel, non-invasive biomarkers with diagnostic, therapeutic, and prognostic utility is of utmost importance. MiRNA profiling enables early detection of pathophysiological changes. For example, one study reported that miRNAs demonstrated higher diagnostic sensitivity in stage I–II EMs than in stage III–IV, with sensitivity reaching up to 80% [ 52 ]. Others studies have applied miRNAs to distinguish EMs from endometrial and ovarian cancers [ 77 ]. Fagan’s nomogram was used to assess clinical applicability. Assuming a pre-test probability of 20%, a positive miRNA result increased the post-test probability to 57%, suggesting the need for confirmatory testing. A negative result reduced the post-test probability to 5%, which aligns with the test’s high negative predictive value (NPV) of 95%, and supports its utility in screening low-risk populations to reduce unnecessary invasive procedures. Our findings reveal no significant diagnostic differences between sample types, although their clinical applicability varies. Peripheral blood samples offer advantages such as ease of collection and minimal invasiveness, resulting in good patient compliance and making them suitable for large-scale preliminary screening. However, their diagnostic specificity remains suboptimal and requires further improvement. Tissue-based histopathological examination is considered the gold standard for diagnosing endometriosis. MiRNA data derived from such samples hold significant scientific value for investigating disease mechanisms and validating diagnostic biomarkers. Nevertheless, the invasive nature of sample collection and relatively low patient compliance limit their broader clinical application. Salivary samples demonstrated low diagnostic heterogeneity (I² = 17%, P = 0.30), although only three studies were included in this meta-analysis. Although saliva samples exhibit no significant diagnostic heterogeneity, related studies remain limited, and their diagnostic value warrants further exploration. The limited number of studies may be the primary source of the low heterogeneity observed in salivary data. In addition, heterogeneity may also result from variations in pre-analytical factors such as the timing of saliva collection and food intake. For instance, Sofiane et al. did not provide a detailed description of the procedures preceding saliva sampling [ 60 ]. This has drawn our considerable attention and interest, primarily due to the high accessibility of saliva samples [ 41 , 60 , 61 ]. Despite the current scarcity of research evidence, recent studies have identified the potential diagnostic utility of salivary miRNAs in other diseases [ 78 , 79 ]. These findings indirectly suggest that salivary miRNAs hold promise as diagnostic biomarkers, offering notable advantages such as convenience and non-invasiveness. Consequently, salivary miRNA analysis may represent a novel and promising direction for future diagnostic applications. Therefore, their diagnostic performance requires further validation. Compared with conventional diagnostic methods, miRNA testing offers unique advantages. With continuous advancements in detection technology, the throughput and sensitivity of miRNA profiling have significantly improved, enabling simultaneous analysis of multiple miRNAs and providing more comprehensive diagnostic information. For instance, Bendifallah and colleagues have developed several predictive models based on multiple miRNAs, which have been applied to both blood and saliva samples [ 60 , 80 ]. Some studies have combined miRNA profiling with ultrasound imaging to enhance diagnostic accuracy [ 37 ], while others have integrated serum CA125 measurements with specific miRNA expression signatures. These multimodal approaches have demonstrated higher diagnostic performance than single-marker strategies and hold promise for improving the early detection rate of endometriosis [ 40 ]. The clinical implementation of miRNA testing can be integrated into existing medical infrastructures, as illustrated in Fig.  9 . Sample collection can be performed through routine venipuncture, which is fully compatible with standard clinical blood collection procedures. Analytical testing can be conducted in hospital-based central laboratories equipped with molecular pathology platforms, employing standardized, high-throughput techniques such as reverse transcription quantitative polymerase chain reaction (RT-qPCR) to ensure the reproducibility and reliability of the results. The key to clinical application lies in establishing a panel of miRNAs with high sensitivity and specificity, thereby providing a robust basis for clinical decision-making. Based on the findings of the present meta-analysis, several miRNA combinations with favorable diagnostic performance were identified, with their specific composition, sensitivity, and specificity summarized in Table  4 . Fig. 9 The clinical scenarios and a practical workflow for integrating miRNA-based diagnostics into existing medical infrastructures The clinical scenarios and a practical workflow for integrating miRNA-based diagnostics into existing medical infrastructures Table 4 Diagnostic performance of serum MiRNA panels for endometriosis Source of Collected Specimens Exact analytic panel (Combined) AUC SEN SPE Peripheral blood hsa-miR-17-5p, hsa-miR-424-5p 0.94 0.94 0.89 Peripheral blood hsa-miR-199b-3p, hsa-miR-224-5p, hsa-let-7d-3p 0.99 0.96 1.00 Peripheral blood hsa-let-7d-5p, hsa-miR-21-5p, hsa-miR-28-5p 1 1 1 Peripheral blood has-miR-122, has-miR-199a 0.96 0.96 0.91 Peripheral blood hsa-miR-125b-5, hsa-miR-451a, hsa-miR-3613-5p 0.97 0.96 0.92 Diagnostic performance of serum MiRNA panels for endometriosis

Despite its strengths, this study has several limitations. First, suboptimal methodological quality in some included studies, particularly in patient selection and miRNA testing protocols, may have introduced bias and compromised reliability, underscoring the need for more rigorous, standardized designs in future research. Second, significant heterogeneity among studies, partly due to inconsistent cut-off values, panel compositions and platform differences, limits the clinical translation of miRNA biomarkers. Third, though spectrum bias was present, the consistent use of a diagnostic gold standard across most studies supports the overall reliability of these tests. Fourth, potential within-study multiplicity—such as inclusion of multiple miRNAs from the same cohort—may have inflated precision and underestimated variance. Fifth, the restriction to English-language publications may have introduced language bias, although most high-quality studies in this field are reported in English. Given the complexity of endometriosis and the potential synergistic effects of miRNAs, future work should prioritize developing reproducible, multi-marker diagnostic models validated in large, diverse, and multi-ethnic cohorts to enhance both accuracy and generalizability.

Endometriosis is a common benign gynecological disorder, affecting approximately 10% of women of reproductive age, with even higher prevalence among those with infertility [ 1 – 3 ]. It is characterized by the ectopic implantation of endometrial tissue that responds to hormonal changes and causes pelvic pain, infertility, and reduced quality of life [ 4 , 5 ]. However, early diagnosis of endometriosis remains highly challenging due to nonspecific clinical symptoms and a lack of reliable early biomarkers, often resulting in delays or misdiagnosis [ 6 – 9 ]. Currently used diagnostic tools, transvaginal ultrasound (TVS) and magnetic resonance imaging (MRI) and serum CA125, lack sufficient sensitivity for superficial or early lesions [ 10 – 14 ]. Laparoscopy though the gold standard, is invasive and costly, limiting its feasibility for early screening [ 15 , 16 ]. These limitations underscore the need for non-invasive, accurate, accessible, and cost-effective diagnostic strategies for the early detection. MicroRNAs (miRNAs) are endogenous non-coding RNAs that regulate gene expression and have emerged as promising as non-invasive biomarkers [ 17 ]. They are involved in key biological processes and have been associated with the pathogenesis of multiple diseases [ 18 – 20 ]. In the context of endometriosis, increasing attention has been paid to the diagnostic potential of miRNAs. Wang et al. were the first to to identified circulating miRNAs as non-invasive diagnostic tools [ 21 ]. They may also contribute to malignant transformation and progesterone resistance, suggesting potential as diagnostic markers and therapeutic targets [ 22 , 23 ]. These interactions suggest that miRNAs may serve as both diagnostic markers and therapeutic targets. An expanding body of evidence indicates that endometriosis is associated with distinct alterations in miRNA expression profiles, providing a theoretical basis for their application as diagnostic biomarkers [ 24 ]. However, current studies exhibit significant heterogeneity in sample sizes, study designs, and technical procedures, resulting in inconsistent conclusions and limiting clinical applicability [ 25 – 27 ]. To clarify the diagnostic utility of miRNAs in endometriosis, a systematic review and meta-analysis is warranted [ 28 ]. Therefore, we conducted a meta-analysis to assess the diagnostic performance of miRNAs—including pooled sensitivity, specificity, and AUC—to explore heterogeneity and provide a more robust evidence base for their potential clinical application.

Supplementary Material 1: Search Strategy. Supplementary Material 1: Search Strategy. Supplementary Material 2: Detailed Value Summary Table. Supplementary Material 2: Detailed Value Summary Table. Supplementary Material 3: Prisma Checklist. Supplementary Material 3: Prisma Checklist.