Mendelian Randomization to Examine the Causal Effects of Cystatin on Ovarian Lesions Authors: Tianqing Yan, Lin Guo, Renquan Lu Published: 2026 ABSTRACT AbstractBackgroundCystatin, a superfamily of cysteine protease inhibitors, are implicated in extracellular matrix remodeling, immune modulation, and tumor progression. Observational studies have reported associations between specific cystatin and gynecological pathologies, including ovarian cancer. However, whether these associations reflect causality remains uncertain due to potential confounding and reverse causation.MethodsWe performed a two-sample Mendelian randomization (MR) study to investigate the causal relationships between genetically predicted levels of seven cystatin subtypes (Cystatin B, C, D, F, M, S, and Cystatin 8) and the risk of various ovarian lesions. Genetic instruments (single nucleotide polymorphisms, SNPs) for cystatin were selected from genome-wide association studies (GWAS) at a significance threshold of P < 5×10−6, with clumping for linkage disequilibrium (LD R2 < 0.001, window = 10,000 kb). Summary-level data for outcomes including ovarian cysts, primary ovarian failure, ovarian torsion, and major histologic subtypes of ovarian cancer (high/low-grade serous, mucinous, clear cell, endometrioid) were obtained from the FinnGen, UK Biobank and Ovarian Cancer Association Consortium (OCAC). The primary analysis used the inverse-variance weighted (IVW) method, complemented by Weighted Median, MR-Egger, Mode-based methods, and sensitivity analyses for pleiotropy and heterogeneity.ResultsGenetically predicted Cystatin-8 was significantly associated with a reduced risk of endometrioid ovarian cancer (OR = 0.898, 95% CI: 0.836–0.965, P=0.003). Cystatin-C showed a suggestive protective effect against high-grade serous ovarian cancer (OR = 0.908, 95% CI: 0.828–0.994, P=0.038). Cystatin-F was associated with a reduced risk of polycystic ovarian syndrome (OR = 0.852, 95% CI: 0.725–0.999, P = 0.049). No significant causal relationships were observed for the other cystatin subtypes with the studied ovarian lesions. Sensitivity analyses were conducted to validate the MR assumptions. Cochran’s Q test was used to assess heterogeneity across the genetic instruments. The MR-Egger intercept test was applied to detect and adjust for directional horizontal pleiotropy. Additionally, a leave-one-out analysis was performed to ensure that no single SNP disproportionately drove the causal estimates.ConclusionThis MR study provides genetic evidence supporting a potential causal, protective role for Cystatin-8 against endometrioid ovarian cancer, and suggestive protective roles for Cystatin-C in high-grade serous ovarian cancer and Cystatin-F in polycystic ovarian syndrome. These findings highlight specific cystatin as potential biomarkers or etiological factors warranting further mechanistic and clinical investigation in ovarian pathophysiology.Keywords: cysteine protease inhibitors, ovarian cancer, ovarian lesions, Mendelian randomization, causal inference Introduction Ovarian lesions encompass a spectrum of conditions from benign disorders, such as ovarian cysts and polycystic ovarian syndrome (PCOS), to malignant neoplasms, which collectively represent a significant cause of morbidity and mortality in women globally.1 Epithelial ovarian cancer (EOC) is the most lethal gynecologic malignancy, with distinct histologic subtypes including high-grade serous (HGSOC), low-grade serous (LGSOC), clear cell (OCCC), endometrioid (ENOC), and mucinous (MOC), exhibiting diverse molecular profiles, clinical behaviors, and prognoses.2 Despite advances in treatment, the 5-year survival rate for advanced-stage disease remains below 50%, underscoring the urgent need for a deeper understanding of etiological factors and early biomarkers.3 The cystatin superfamily comprises endogenous inhibitors of cysteine cathepsins, a class of lysosomal proteases involved in protein turnover, extracellular matrix (ECM) degradation, antigen presentation, and apoptosis.4 By regulating cathepsin activity, cystatin play crucial roles in maintaining tissue homeostasis, modulating inflammatory responses, and influencing tumor microenvironments. Dysregulation of the cystatin-cathepsin axis has been implicated in various cancers, including breast, colorectal, and lung malignancies, where it can affect invasion, metastasis, and immune evasion.5,6 In ovarian pathology, alterations in cystatin expression have been identified, with certain cystatins showing potential links to disease progression.7 However, findings from conventional observational studies in this field are often constrained by methodological limitations, including confounding from environmental and lifestyle factors, as well as the possibility of reverse causality where disease status may itself influence biomarker concentrations. Consequently, establishing a clear causal relationship between circulating cystatin levels and ovarian disease risk remains challenging within the framework of traditional epidemiological approaches. Mendelian randomization (MR) has emerged as a powerful method for causal inference in observational data. By using genetic variants as instrumental variables (IVs) for modifiable exposures, MR mimics the random allocation of a randomized controlled trial, largely avoiding confounding and reverse causation.8 The approach relies on three core assumptions:9 1) the genetic instruments are robustly associated with the exposure; 2) they are independent of confounders; and 3) they affect the outcome only through the exposure (no horizontal pleiotropy). Two-sample MR, which utilizes summary-level data from independent genome-wide association studies (GWAS) for exposure and outcome, has been successfully applied to elucidate causal relationships in oncology,10,11 as demonstrated in recent studies exploring galectins in gynecologic cancers12 and cathepsins in ovarian tumors.13 Inspired by the analytical frameworks of these studies and given the biological plausibility of cystatin involvement in ovarian pathophysiology, we conducted a comprehensive two-sample MR analysis. This study aims to investigate the potential causal effects of seven genetically predicted cystatin subtypes on a broad range of ovarian lesions, from benign conditions to malignant histological subtypes. Our findings aim to clarify the etiological roles of cystatin and identify potential targets for risk stratification and therapeutic intervention. Methods Study Design and MR Assumptions The study examined the causal effects of seven cystatin subtypes (exposures) on various ovarian lesions (outcomes), including benign conditions and malignant neoplasms, while accounting for potential confounders and strictly conformed to the proposed STROBE-MR Statement in the analysis14 (Figure 1). The MR analysis was predicated on three fundamental assumptions:9 1) Relevance: The selected single nucleotide polymorphisms (SNPs) are strongly associated with circulating levels of the respective cystatin. 2) Independence: These SNPs are not associated with any known or unknown confounders of the cystatin-ovarian lesion relationship. 3) Exclusivity: The SNPs influence the risk of ovarian lesions only via their effect on cystatin levels, not through other biological pathways (no horizontal pleiotropy). Study design of the Mendelian randomization analysis. Genetic instrumental variables (IVs) for cystatin subtypes were selected as SNPs associated with each exposure at a genome-wide significance threshold (P < 5×10−6). Clumping was performed with a window of 10,000 kb and a linkage disequilibrium (LD) threshold of R2 < 0.001. Only instruments with an F-statistic > 10 were retained to ensure strong instrumental strength. The red crosses over the dotted lines indicate that potential confounders are blocked or adjusted for in the MR design. Data Sources for Exposures (Cystatin) and Outcomes (Ovarian Lesions) Summary statistics data for SNPs related to plasma Cystatins were extracted from the most comprehensive publicly available genome-wide association studys (GWAS) summary statistics to date (MRC-Integrative Epidemiology Unit (IEU) OpenGWAS database, https://opengwas.io/, accessed on 2025.12.15). The Cystatin-B and Cystatin-C datasets includes 3394 individuals and 437,846 individuals of European ancestry respectively.15,16 Genetic data for the Cystatin-S include a European-ancestry population, covering 501,428 SNPs.17 The datasets for Cystatin-8, Cystatin-D, Cystatin-F, Cystatin-M comprise 3301 individuals of European ancestry.18 Data on ovarian cancer (25,509 cases among 66,450 individuals), HGSOC (13,037 cases among 53,978 individuals), LGSOC (1012 cases among 41,953 individuals), MOC (1417 cases among 42,358 individuals), OCCC (1366 cases among 42,307 individuals), and ENOC (2810 cases among 43,751 individuals) were sourced from the Ovarian Cancer Association Consortium (OCAC).19 The data on ovarian cysts comprise a European-ancestry sample of 462,933 individuals (3877 cases and 459,056 controls) sourced from the UK Biobank via the MRC-IEU consortium. Data on primary ovarian failure (254 cases, 118,228 controls), torsion of ovary/ovarian pedicle/fallopian tube (318 cases, 68,969 controls), and polycystic ovarian syndrome (642 cases, 118,228 controls) were obtained from the FinnGen consortium (www.finnbb.fi). The characteristics of each cystatin dataset, and outcome datasets are detailed in Table 1. Ethical approval and participant informed consent were obtained in all original GWAS. The current analysis utilized fully anonymized summary-level data; therefore, no additional ethical review was required for this study. Table 1Description of Exposures Traits and Outcomes (European)PhenotypeSample Size (Case)SNPsGWAS IDConsortiumExposures Cystatin-8330110,534,735prot-a-706NA Cystatin-B33945,270,646prot-b-3NA Cystatin-C437,8464,232,088ebi-a-GCST90025945NA Cystatin-D330110,534,735prot-a-702NA Cystatin-F330110,534,735prot-a-705NA Cystatin-M330110,534,735prot-a-704NA Cystatin-SNA501,428prot-c-3802_50_1NAOutcomes Ovarian cyst or cysts462,9339,851,867ukb-b-15025MRC-IEU Primary ovarian failureNA16,379,677finn-b-E4_OVARFAILNA Torsion of ovaryNA16,376,184finn-b-N14_OVARTORSNA PCOSNA16,379,676finn-b-E4_POCSNA OC66,450 (25,509)NAieu-a-1120OCAC HGSC53,978 (13,037)NAieu-a-1121OCAC LGSC41,953 (1012)NAieu-a-1122OCAC MOC42,358 (1417)NAieu-a-1123OCAC OCCC42,307 (1366)NAieu-a-1124OCAC ENOC43,751 (2810)NAieu-a-1125OCACAbbreviations: GWAS, genome-wide association studies; PCOS, polycystic ovarian syndrome; OC, ovarian cancer; HGSOC, high-grade serous ovarian cancer; LGSOC, low-grade serous ovarian cancer; MOC, mucinous ovarian cancer; OCCC, ovarian clear cell cancer; ENOC, endometrioid ovarian cancer; OCAC, ovarian cancer association consortium; NA, not available. Description of Exposures Traits and Outcomes (European) Abbreviations: GWAS, genome-wide association studies; PCOS, polycystic ovarian syndrome; OC, ovarian cancer; HGSOC, high-grade serous ovarian cancer; LGSOC, low-grade serous ovarian cancer; MOC, mucinous ovarian cancer; OCCC, ovarian clear cell cancer; ENOC, endometrioid ovarian cancer; OCAC, ovarian cancer association consortium; NA, not available. Selection of Genetic Instrumental Variables (IVs) The SNPs independently associated with its circulating levels at a genome-wide significance threshold of P < 5×10−6 were selected as candidate IVs. A less stringent threshold than the conventional 5×10−8 was used to ensure sufficient numbers of instruments for MR analysis, a common practice in protein quantitative trait locus (pQTL) studies.14 SNPs with a minor allele frequency (MAF) > 0.01 were selected from the analysis. We then performed LD clumping (R2<0.001, window size = 10,000 kb) to ensure the independence of selected SNPs.14 Palindromic SNPs with ambiguous strand orientation were removed. For each SNP, we calculated the proportion of variance explained (R2) and the F-statistic to assess instrument strength, retaining only instruments with an F-statistic > 10 to mitigate weak instrument bias.20 The F-statistic was calculated using the following formula: F = [R2 × (N-1-K)] / [K × (1-R2)]. The variance explained (R2) was derived from the formula: R2 = 2 × MAF × (1-MAF) × β2, where β represents the estimated effect size of the exposure. Statistical Analysis of Mendelian Randomization The primary causal estimates were calculated using the inverse-variance weighted (IVW) method with multiplicative random effects, which provides a meta-analysis of Wald ratios for each SNP.21,22 To ensure robustness and account for potential violations of MR assumptions, four supplementary methods were employed. The Weighted Median remains consistent despite up to 50% invalid instruments,23 whereas MR-Egger regression tests and adjusts for pleiotropy via an intercept term.20 The Weighted Mode clusters SNPs by estimate similarity, in comparison to the more basic Simple Mode estimator. Results for associations are presented as odds ratios (ORs) with 95% confidence intervals (CIs) per unit increase in genetically predicted cystatin level. Sensitivity Analyses We conducted extensive sensitivity analyses to validate the primary results. Cochran’s Q statistic was used to evaluate heterogeneity among the Wald ratios from individual SNPs. A significant Q statistic (P < 0.05) indicates potential heterogeneity, possibly due to pleiotropy. A random-effects IVW model was applied in the presence of significant heterogeneity; conversely, a fixed-effects IVW model was used when no such heterogeneity was detected. The MR-Egger intercept test was performed to assess the horizontal pleiotropy. A non-zero intercept (P < 0.05) suggests the presence of overall directional pleiotropy. Leave-One-Out Analysis was applied to determine if the causal estimate was driven by a single influential variant. All statistical analyses were performed using R software (version 4.5.2) with the TwoSampleMR (version 0.6.29) packages. Visual representations, including scatter plots, funnel plots, and forest plots, were employed to illustrate the analytical results. The scatter plot depicts the effect relationship of IVs on both the exposure and outcome, whereas the funnel plot assesses potential bias, and the forest plot displays the effect estimates of individual SNPs along with their consistency. Results Characteristics of Genetic Instruments and Causal Effects of Cystatin on Ovarian Lesions The selection process yielded strong instrumental variables for each cystatin subtype, with F-statistics all exceeding the threshold of 10, indicating a low risk of weak instrument bias. The number of SNPs used for each cystatin-outcome pair analysis varied and is reported in the forest plots. The IVW MR estimates for all cystatin-ovarian lesion pairs are summarized in the comprehensive forest plot (Figure 2), where statistically significant protective effects of genetically predicted higher Cystatin-8 levels on the risk of endometrioid ovarian cancer (OR = 0.898, 95% CI: 0.836–0.965, P = 0.003). This association was consistent in direction across supplementary methods (Weighted Median, MR-Egger, Mode-based methods) (Figure 3). A suggestive protective association was observed between Cystatin-C and HGSOC (high-grade serous ovarian carcinoma) risk (OR = 0.908, 95% CI: 0.828–0.994, P = 0.038). In addition, genetically predicted higher levels of Cystatin-F were associated with a lower risk of polycystic ovarian syndrome (PCOS) (OR = 0.852, 95% CI: 0.725–0.999, P = 0.049). No significant causal relationships were identified for any cystatin subtype with ovarian cysts, primary ovarian failure, ovarian torsion, or the other ovarian cancer subtypes (LGSOC, mucinous, clear cell) after multiple testing consideration (all PIVW ≥ 0.05). Detailed information for all instruments is listed in Figures S1 and S2. Mendelian randomization estimates for the associations between genetically predicted levels of cystatin subtypes and risks of ovarian lesions. For each cystatin subtype (Cystatin-8 (A), Cystatin-B (B), Cystatin-C (C), Cystatin-D (D), Cystatin-F (E), Cystatin-M (F), Cystatin-S (G)), results are presented as odds ratios (OR) with 95% confidence intervals (CIs) per standard deviation increase in genetically predicted cystatin level. The number of SNPs used as instrumental variables is shown for each outcome. Red squares represent significant associations (P < 0.05). Horizontal lines represent 95% CIs; squares denote point estimates (OR), with size proportional to the precision of the estimate. A vertical dashed line indicates the null effect (OR = 1). The lower section (“Low Risk/High Risk”) provides a reference for interpreting effect direction. Scatter plot of Mendelian randomization analysis. This scatter plot presents a two-sample MR analysis investigating the potential causal effect of Cystatin-8 on endometrioid ovarian cancer (A), Cystatin-C on high-grade serous ovarian carcinoma (B) and Cystatin-F on polycystic ovarian syndrome (C). Each point represents an SNP. The x-axis shows the SNP-exposure association (genetic effect on Cystatin), and the y-axis shows the SNP-outcome association (genetic effect on ovarian lesions). The regression slope represents the causal estimate per unit change in cystatin levels. Sensitivity Analyses for the Main MR Estimates Cochran’s Q tests showed no significant heterogeneity for the associations involving Cystatin-8 and endometrioid ovarian cancer (Q = 10.816, P = 0.902), as well as Cystatin-F and PCOS (Q = 8.127, P = 0.775). Notably, despite significant heterogeneity in the association between Cystatin-C and high-grade serous ovarian cancer as indicated by Cochran’s Q test (Q = 626.874, P < 0.001), the IVW random-effects model yielded a statistically significant causal estimate (OR = 0.908, 95% CI: 0.828–0.994, P = 0.038). The MR-Egger intercept tests yielded intercepts close to zero with non-significant P-values for these associations (all PEgger intercept > 0.1), indicating no evidence of substantial directional horizontal pleiotropy (Table 2). Table 2Assessment of the Heterogeneity and Horizontal Pleiotropy Between Cystatin and Ovarian LesionsExposuresOutcomesHeterogeneityHorizontal PleiotropyCochran’s QPMR-Egger InterceptPCystatin-8Ovarian cyst or cysts14.6010.0670.0010.213Primary ovarian failure29.7960.054−0.0290.634Torsion of ovary28.2290.0790.0670.202Polycystic ovarian syndrome21.3710.317−0.0020.940Ovarian cancer14.7680.6780.0040.608High grade serous ovarian cancer18.5750.418−0.0020.792Low grade serous ovarian cancer23.6500.1670.0150.616Invasive mucinous ovarian cancer13.1800.781−0.0030.905Clear cell ovarian cancer9.0160.940−0.0210.519Endometrioid ovarian cancer10.8160.9020.0030.820Cystatin-BOvarian cyst or cysts0.2680.605NANAPrimary ovarian failure0.9050.824−0.0130.916Torsion of ovary3.1070.3750.0410.760Polycystic ovarian syndrome1.4400.696−0.0680.426Ovarian cancer1.2320.745−0.0170.430High grade serous ovarian cancer2.3810.497−0.0230.370Low grade serous ovarian cancer4.3960.222−0.1280.171Invasive mucinous ovarian cancer0.7020.8730.0000.996Clear cell ovarian cancer0.2840.9630.0180.756Endometrioid ovarian cancer2.5930.459−0.0090.845Cystatin-COvarian cyst or cysts414.7050.7282.122e-050.252Primary ovarian failure469.7670.813−0.0080.331Torsion of ovary480.7890.7020.01290.100Polycystic ovarian syndrome524.7100.197−0.0080.143Ovarian cancer633.642<0.001−1.3610.993High grade serous ovarian cancer626.874<0.0010.0020.386Low grade serous ovarian cancer489.8880.320−0.0100.037Invasive mucinous ovarian cancer451.9710.7800.0020.535Clear cell ovarian cancer496.9770.245−0.0010.843Endometrioid ovarian cancer544.3020.016−0.0030.238Cystatin-DOvarian cyst or cysts3.1290.372−0.0010.532Primary ovarian failure18.5930.017−0.2450.136Torsion of ovary10.8500.210−0.0260.831Polycystic ovarian syndrome3.2100.920−0.0510.469Ovarian cancer14.6310.067−0.0160.461High grade serous ovarian cancer11.0840.197−0.0010.980Low grade serous ovarian cancer9.6180.293−0.0910.133Invasive mucinous ovarian cancer5.2580.730−0.0200.664Clear cell ovarian cancer15.8410.045−0.0770.255Endometrioid ovarian cancer12.2310.1410.0010.978Cystatin-FOvarian cyst or cysts6.4300.4916.556e-050.573Primary ovarian failure10.0680.6100.0160.782Torsion of ovary9.4850.5770.1050.125Polycystic ovarian syndrome8.1270.775−0.0020.958Ovarian cancer13.0100.368−0.0030.779High grade serous ovarian cancer14.5450.2670.0030.765Low grade serous ovarian cancer24.9510.0090.0250.640Invasive mucinous ovarian cancer14.9120.2460.0050.874Clear cell ovarian cancer5.9520.918−0.0210.426Endometrioid ovarian cancer6.4130.894−0.0090.637Cystatin-MOvarian cyst or cysts6.2180.286−0.0020.122Primary ovarian failure2.8270.998−0.0280.728Torsion of ovary18.3590.1440.0360.689Polycystic ovarian syndrome7.9670.846−0.0650.225Ovarian cancer8.7600.7910.0110.333High grade serous ovarian cancer16.8190.2080.0290.056Low grade serous ovarian cancer18.2810.147−0.0500.314Invasive mucinous ovarian cancer9.9700.6960.0290.397Clear cell ovarian cancer8.2610.8260.0140.692Endometrioid ovarian cancer8.4830.811−0.0260.313Cystatin-SOvarian cyst or cysts6.2060.1020.0010.708Primary ovarian failure1.6330.652−0.5870.597Torsion of ovary2.0410.564−0.4120.674Polycystic ovarian syndrome3.1950.363−0.0890.917Ovarian cancer3.1690.3660.2060.284High grade serous ovarian cancer0.1710.9820.0600.757Low grade serous ovarian cancer5.6360.1310.8390.313Invasive mucinous ovarian cancer0.1200.9890.0810.867Clear cell ovarian cancer1.8050.6140.2640.601Endometrioid ovarian cancer6.4620.0910.4150.474Notes: Cochran’s Q test for heterogeneity under the inverse-variance weighted model. MR-Egger intercept test for directional pleiotropy.Abbreviation: NA, not available. Assessment of the Heterogeneity and Horizontal Pleiotropy Between Cystatin and Ovarian Lesions Notes: Cochran’s Q test for heterogeneity under the inverse-variance weighted model. MR-Egger intercept test for directional pleiotropy. Abbreviation: NA, not available. To visually assess potential directional pleiotropy and the overall symmetry of the causal estimates, we generated funnel plots for all significant exposure-outcome pairs (eg, Cystatin-8 on ENOC, Cystatin-C on HGSOC and Cystatin-F on PCOS). In a funnel plot, the precision of each instrumental variable (inverse of the standard error) is plotted against its causal estimate (beta coefficient). In the absence of significant pleiotropy or selection bias, the points should be distributed symmetrically around the IVW estimate line. The funnel plots for our primary analyses showed a largely symmetrical distribution of the individual SNP estimates around the pooled IVW estimate line. No pronounced asymmetry was observed, which suggests the absence of strong directional pleiotropy that could systematically bias our results (Figure 4). This visual assessment corroborates the non-significant intercepts obtained from the MR-Egger regression, further supporting the validity of the chosen genetic instruments and the robustness of the causal inferences. Besides, the results were not driven by any single influential SNP. The causal estimates remained stable and significant in the same direction after sequentially removing each instrument (Figures S3 and S4). The overall consistency and symmetry observed in both the funnel and scatter plots robustly support the reliability of the causal estimates derived from our Mendelian randomization analysis (Figures S5–S8). Funnel plot of Mendelian randomization analysis. Funnel plot assessing the potential directional pleiotropy and heterogeneity across SNPs. (A) Funnel plot for the causal effect of Cystatin‑8 on endometrioid ovarian cancer. (B) Funnel plot for the causal effect of Cystatin‑C on high‑grade serous ovarian cancer. (C) Funnel plot for the causal effect of Cystatin‑F on polycystic ovary syndrome. Symmetry around the pooled causal estimate (central vertical line) suggests the absence of significant horizontal pleiotropy. Asymmetry may indicate bias or invalid instruments. Reverse MR Analysis to Assess Bidirectional Causality To examine the possibility of reverse causality where the disease outcome might influence cystatin levels, we performed bidirectional MR analyses. For each significant association identified in the forward direction (eg, Cystatin-8 on ENOC, Cystatin-C on HGSOC and Cystatin-F on PCOS), we swapped the exposure and outcome. We selected independent genetic instruments (P < 5×10−6, R2 < 0.001) for the ovarian lesion from the respective GWAS summary statistics and estimated their causal effect on the corresponding cystatin level using the IVW method. A lack of significant association in this reverse direction would support the primary causal inference. The results showed that in all the reverse causal paths tested, no statistically significant causal effects were found. Specifically, the IVW analysis results with endometrioid ovarian cancer as exposure and Cystatin-8 as outcome indicated that the genetically predicted risk of ovarian cancer had no significant effect on Cystatin-8 levels (OR = 0.979, 95% CI: 0.890–1.077, P = 0.665). Non-association was observed between HGSOC risk and Cystatin-C (OR = 1.008, 95% CI: 0.957–1.061, P = 0.777). For the MR analysis of PCOS on Cystatin-F, a total of eight SNPs were available after harmonization of effect alleles. However, the GWAS summary statistics for the exposure did not report the per-SNP sample size (N), which is a required parameter for calculating the critical F-statistic to assess instrument strength. Therefore, we cannot evaluate the risk of weak instrument bias. Consequently, in adherence to methodological best practices for MR, we did not proceed with causal inference for this exposure-outcome pair (Figure 5). Besides, the sensitivity analysis revealed no significant evidence of reverse causality. All estimates and their confidence intervals consistently overlapped the null value, indicating that the observed causal effect of the exposure on the outcome is robust and not confounded by reverse causation (Figure S9). Forest plot of bidirectional Mendelian randomization analysis for cystatin levels on ovarian lesion risk. Each row represents a specific exposure-outcome pair, reporting the number of instrumental SNPs, OR with 95% CI, and P-value. No significant associations were observed for endometrioid ovarian cancer or high-grade serous ovarian carcinoma with their respective cystatin exposures. Data for cystatin-F and polycystic ovarian syndrome were unavailable for analysis. Discussion In this comprehensive two-sample MR study, we leveraged large-scale genetic data to investigate the causal roles of seven cystatin subtypes across a spectrum of ovarian lesions. Our primary finding is a robust, genetically predicted protective effect of Cystatin-8 against endometrioid ovarian cancer. We also identified suggestive protective roles for Cystatin-C in high-grade serous ovarian cancer and for Cystatin-F in polycystic ovarian syndrome. These results provide novel genetic evidence highlighting specific members of the cystatin family as potential etiological factors in distinct ovarian pathologies. Cystatin-8 (CRES, or CST8), a member of the cystatin superfamily, is predominantly and highly expressed in the testis and epididymis functioning in spermatogenesis, sperm maturation, hypothalamic-pituitary-gonadal axis regulation and epididymal antimicrobial defense.24–26 Current research on Cystatin-8 is largely confined to male reproductive system disorders, with limited and scattered evidence for its low expression in a few malignancies such as bladder cancer,27 yet no direct original research has reported its expression, functional role or clinical correlation in ovarian lesions, leaving Cystatin-8 a largely unexplored target in ovarian disease. In this study, our findings revealed a significant causal association between Cystatin 8 and endometrioid ovarian cancer risk via the IVW (OR = 0.898, 95% CI: 0.836–0.965, P = 0.003), Weighted Median (OR = 0.908, 95% CI: 0.830–0.993, P = 0.035), and MR Egger (OR = 0.891, 95% CI: 0.804–0.987, P = 0.041) approaches, suggesting that Cystatin 8 may act as a potential protective factor which could pave the way for the development of novel treatments or biomarkers for this cancer subtype. However, further preclinical and clinical studies are needed to elucidate the exact biological mechanisms. Cystatin-C (CysC, CST3) is a ubiquitous cysteine protease inhibitor that potently suppresses cathepsins, thereby modulating extracellular matrix degradation, inflammation, and cell invasion.28–31 As a classic and sensitive biomarker for glomerular filtration rate, its role has recently expanded into the realms of tumor biology and metabolic disorders.32–35 Notably, mounting evidence has clarified the expression pattern, functional role and clinical significance of CysC in ovarian lesions. In ovarian cancer, CysC is dysregulated in tumor tissues, ascites, serum and cyst fluid, and its abnormal expression is closely linked to tumor progression by regulating the cathepsin-protease inhibitor balance in the tumor microenvironment;36–38 it is also involved in the pathological processes of polycystic ovary syndrome (PCOS) by mediating insulin resistance and chronic inflammation.39 These findings confirm CysC as a potential diagnostic or prognostic biomarker and therapeutic target for ovarian diseases, with relevant mechanistic and clinical studies continuously deepening. Our Mendelian randomization analysis identified a significant causal protective effect of CysC on HGSOC based on the IVW method (OR = 0.908, 95% CI: 0.828–0.994, P =0.038) and Weighted Median approach (OR = 0.840, 95% CI: 0.724–0.975, P =0.022), clarifying the genetic causal association between CysC and HGSOC. This finding aligns partially with the observations by Kolwijck et al40 who reported a strong positive correlation (P < 0.001, R = 0.921) between cathepsin B (CatB) and CysC levels in ovarian cyst fluid (oCF) of malignant serous tumors, and proposed that the increase in CysC might counterbalance CatB-mediated proteolysis, thereby restraining tumor invasiveness. However, in their subsequent study,36 CysC levels in oCF were not significantly predictive of patient survival. CysC in ovarian cyst fluid likely reflects local biological activity rather than systemic circulation, which may explain the discrepancy between its lack of prognostic significance in oCF and the protective effect observed for circulating CysC in our Mendelian randomization analysis. The protective effect of Cystatin-F against PCOS is a novel observation (IVW: OR = 0.852, 95% CI: 0.725–0.999, P =0.049). PCOS is a heterogeneous endocrine disorder with chronic low-grade inflammation as a key feature.41 Cystatin-F is primarily expressed in immune cells and is a potent regulator of cytotoxic lymphocyte activity by inhibiting cathepsin C and other proteases.42 Higher circulating Cystatin-F might therefore dampen excessive immune activation or modify inflammatory pathways contributing to ovarian dysfunction and metabolic features of PCOS. This finding suggests a previously unrecognized link between immune protease regulation and PCOS pathogenesis. In contrast, no significant causal relationships were observed between genetically predicted levels of Cystatin B, D, M, or S and any of the examined ovarian outcomes (all P > 0.05). This suggests a distinct functional specificity among cystatin family members in ovarian pathophysiology, with only Cystatin-8, -C, and -F demonstrating clear etiological relevance in the studied phenotypes. The isoform-specific effects highlight the importance of precise molecular targeting in future therapeutic strategies. Our study has several strengths. First, the MR design minimizes residual confounding and reverse causation, providing stronger evidence for causality than observational studies. Second, we analyzed a comprehensive panel of seven cystatin against a wide range of ovarian outcomes, from benign to malignant. Third, the use of large-scale, publicly available GWAS consortia data (OCAC, FinnGen, UK Biobank) ensures substantial statistical power. Fourth, rigorous sensitivity analyses and bidirectional MR assessment confirmed the robustness of our primary findings against pleiotropy and outlier effects. Limitations should also be acknowledged. For example, the genetic instruments were derived from populations of European ancestry, limiting the generalizability of our findings to other ethnic groups. Also, while we used a standard and widely accepted significance threshold (P < 5×10−6), which could marginally increase the risk of including invalid IVs; however, our high F-statistics and consistent sensitivity analyses mitigate this concern. In addition to this, we cannot completely rule out the possibility of undetected horizontal pleiotropy, though the MR-Egger results are reassuring. Conclusion In conclusion, this Mendelian randomization study provides novel genetic evidence that higher genetically predicted levels of Cystatin-8 have a causal protective effect against endometrioid ovarian cancer. It also suggests potential protective roles for Cystatin-C in high-grade serous ovarian cancer and for Cystatin-F in polycystic ovarian syndrome. These findings disentangle correlation from causation and highlight specific cystatin as priority candidates for further functional investigation. Future research should focus on elucidating the precise molecular mechanisms by which these protease inhibitors influence ovarian pathophysiology, with the long-term goal of evaluating their potential as biomarkers for risk stratification or as targets for therapeutic intervention in specific ovarian disorders. Acknowledgments We thank all the participants and investigators of the OCAC, FinnGen, and UK Biobank consortia for making their GWAS summary data publicly available. Funding Statement This work was supported by the National Natural Science Foundation of China (grant number NSFC-82503569 to T. Y., NSFC-82373383 to R. L.), and the “Young Eagle” Program of the Shanghai Anti-Cancer Association (grant number SACA-CY24C01). Abbreviations CI, Confidence intervals; EOC, Epithelial ovarian cancer; ENOC, Endometrioid ovarian cancer; HGSOC, High-grade serous ovarian cancer; IVs, Instrumental variables; LGSOC, Low-grade serous ovarian cancer; LD, Linkage disequilibrium; MOC, Mucinous ovarian cancer; MR, Mendelian randomization; OCCC, Ovarian clear cell cancer; OC, Ovarian cancer; OR, Odds ratios; SNPs, Single nucleotide polymorphisms. Data Sharing Statement All data used in this study are publicly available from the MRC-IEU OpenGWAS database, https://opengwas.io/. Analysis scripts are available from the corresponding author upon reasonable request. Ethics Declaration The study was submitted to the ethics committee of Fudan University Shanghai Cancer Center. The committee waived the need for ethical approval, as the study used publicly available aggregated data that did not permit re-identification of the original participants. This exemption is supported by Article 32 of the Measures for Ethical Review of Life Science and Medical Research Involving Human Subjects (February 18, 2023, China). Disclosure The authors declare no competing interests in this work. References Associated Data Data Availability Statement All data used in this study are publicly available from the MRC-IEU OpenGWAS database, https://opengwas.io/. Analysis scripts are available from the corresponding author upon reasonable request.