Association between metal mixture exposure and the risk of POI: a case control study.

This case-control study used a sample of all participants were recruited from the International Peace Maternity and Child Health Hospital (IPMCH), School of Medicine of Shanghai Jiao Tong University. The criteria for idiopathic POI included secondary amenorrhea for at least 4months, and serum basal FSH > 40IU/l (on two occasions separated by >1month) before age 40. The level of female sex hormone and AMH of all participants measured by the laboratory department of the IPMCH using a chemiluminescent immunoassay. Women with regular menstrual cycles and normal FSH level (< 10 IU/l) who came for regular examination were recruited as controls. Exclusion criteria included women with a chromosomal abnormality, known gene mutations (such as fragile X messenger ribonucleoprotein 1 (FMR1) pre-mutation), history of ovarian or uterus surgery, Radiotherapy and chemotherapy, endometriosis or autoimmune disease, and history of cardiovascular, cerebrovascular, liver, kidney, and hematopoietic system or mental illness. Patients who had been diagnosed with POI and had been taking hormone replacement therapy were also excluded. The recruitment took place from July 2017 to November 2018, recruiting thirty patients with idiopathic POI and 31 control women with normal ovarian function, with all the patients with POI were newly diagnosed. Maternal pre-pregnancy body mass index (BMI) was calculated by dividing her self-reported body weight in kilograms prior to gestation by her height in square meters. The categorization of maternal pre-pregnancy BMI followed the recommended reference for Chinese adults: underweight (BMI < 18.5 kg/m2), normal weight (BMI between 18.5 and 23.9 kg/m2), and OWO (BMI ≥ 24.0 kg/ m2) [ 20 ]. Although not individually matched, the two groups were comparable in terms of age and pre‑pregnancy BMI are shown in Table 1 , with no statistically significant differences. Table 1 Characteristics of patients with premature ovarian insufficiency ( n  = 30) and controls ( n  = 31) a Characteristics Control ( n  = 31) POI ( n  = 30) P -value b Age at enrollment (years) 33.58 ± 0.86 35.83 ± 0.81 0.650 a Amenorrhea duration (months)  None 31 (100.0) 14 (46.7)  12 5 (16.7) BMI (kg/m 2 ) 0.125 b  24.0 4 (12.9) 4 (13.3) Marital status 0.747 b  Unmarried 6 (19.4) 4 (13.3)  Married 23 (74.2) 23 (76.7)  Divorced 2 (6.5) 3 (10.0) Education 0.215 b  Elementary school 3 (9.7) 0 (0.0)  Middle and high school 1 (3.2) 3 (10.0)  College and above 27 (87.1) 27 (90.0) Monthly family income (RMB yuan) 0.238 b  20,000 9 (29.0) 3(10.0) Water intake (mL) 0.610 b  2000 1 (3.2) 1 (3.3) Smoking 0.321 b  No 30 (96.8) 30 (100.0)  Yes 1 (3.2) 0 (0.0) Alcohol consumption 0.053 b  No 18 (58.1) 10 (33.3)  Yes 13 (41.9) 20 (66.7) Data presented as mean ± standard error of the mean or N (%) Abbreviations : BMI body mass index, POI premature ovarian insufficiency a Student’s t-test of independent samples b Chi-square test Characteristics of patients with premature ovarian insufficiency ( n  = 30) and controls ( n  = 31) a Data presented as mean ± standard error of the mean or N (%) Abbreviations : BMI body mass index, POI premature ovarian insufficiency a Student’s t-test of independent samples b Chi-square test Ethical approval for this study was granted by the Medical Ethics Committee of the IPMCH ((GKLW) 2018-43). The study was approved in accordance with the ethical review of clinical trials of drugs guidance’ issued by the decree of the People’s Republic of China State Food and Drug Administration in 2010, the ‘Helsinki declaration’, and the International Medical Science Organization Committee of ‘the international ethical guidelines for biomedical research ethical principles’. Informed consent forms were signed by all participants before their inclusion in this study. At the first visit, participants were subject to complete an interviewer-administered questionnaire and excluded from the analysis if any of the following data were missing, including age, medical, gynecological history, amenorrhea duration, current health status, marital status, family income, educational status, smoking status, alcohol consumption, and water intake and lifestyle factors. Peripheral blood samples were collected from participants in commercially available trace-element-free tubes following a standardized protocol. Plasma samples was separated from cells by centrifugation for 10 min at 1000 g, and then preserved in polypropylene tubes and stored at − 20℃ until subsequent analysis. Fifteen trace elements (Fe, Zn, Se, Co, Mg, Sr, Li, Cu, Al, Cr, Mn, As, Ti, V, I) in the plasma samples were measured for metallic elements using inductively coupled plasma mass spectrometry (ICP-MS) by Baichen Medical Laboratory Co. (Hangzhou, China) with an Agilent Technologies 7800 ICP-MS (Agilent, Tokyo, Japan), based on a standardized protocol that incorporated sample processing, storage, and transportation to the analysis center. The intra- or inter-batch imprecision coefficient of variations (CVs) ≤ 15.00%. The simple and accurate pretreatment procedure for routine analyses is depicted as follow. After collection, two hours before sample preparation, the plasma samples were brought to room temperature. They were mixed gently for homogenisation and 500 µL of the blood sample were diluted with 100 µL 0.1% (V/V) Triton-X-100 solution (Sigma-Aldrich, Seelze, Germany) and 500 µL of the internal standard solution. This solution was filled up to 5 mL with a 0.5% (V/V) NH4OH solution in suprapure quality (Merck, Darmstadt, Germany) in a 10 mL autosampler polypropylene tube using a 5 mL bottle-top dispenser (Brand, Wertheim, Germany). Finally, the samples were homogenised on the magnetic stirrer React 2000 (Heidolph, Kelheim, Germany). Element concentrations were loge transformed due to right-skewed distributions. The average of three replicate readings for each individual sample was reported as the final sample concentration used in the results. The limit of detection (LOD) was calculated as 3 times the average of 10 consecutive measurements of the blank diluent (0.1% [v/v] Triton X-100, 1% [v/v] HNO3 plus 10 µg/L internal standards including 6Li [No gas], 103Rh [No gas], 103Rh [He], 115In [No gas], 115In [He], 187Re [No gas], and 187Re [He]. The detection rate of all elements was above 90%, and values below the LOD were imputed with the LOD/ \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\sqrt{2}$$\end{document} (Table S1). Continuous variables were first tested for normality using the Shapiro–Wilk test. Variables that followed a normal distribution ( p  ≥ 0.05) in both groups are presented as mean ± standard deviation (SD) and compared using the independent two‑sample t‑test. Variables that deviated from normality ( p  < 0.05) are presented as median (interquartile range, IQR) and compared using the Mann–Whitney U test. Categorical variables are expressed as counts (n) and percentages (%). Baseline characteristics between the POI and control group were compared using chi-square tests, while differences in metal concentrations were assessed using t-tests or Mann-Whitney U tests, as appropriate. Spearman correlation coefficients (ρ) were calculated to evaluate the strength of linear relationships between pairs of metals. Logistic regression was performed to assess associations between individual metal concentrations and risk of POI. Metals were categorized into low and high groups based on median values across all subjects, and crude odds ratios (ORs) with 95% confidence intervals (CIs) were calculated. To explore potential non-linear relationships, restricted cubic splines (RCS) were applied to model the risk of POI as a function of metal concentrations. The number of knots for the RCS models was determined using the Akaike information criterion (AIC) minimization principle. Both logistic regression and RCS models were adjusted for age and BMI as potential confounders. The BKMR model [ 21 ] was employed to evaluated the joint and individual associations between elements and risk of POI. This model estimates the overall mixture effect, the exposure-response function for each element, and the posterior inclusion probabilities (PIPs), which quantify the relative contribution of each element to the outcome. Z-score transformations were performed on the concentrations of elements. A variable selection method with 20,000 iterations in Markov Chain Monte Carlo (MCMC) algorithm was conducted. Dose-response curves for individual metals were generated while fixing the remaining six metals at their 50th percentile. Additionally, BKMR was used to assess potential interactions between mixture components by evaluating non-linear effects at exposure levels fixed at the 25th, 50th, and 75th percentiles. A two-tailed P value < 0.05 was considered statistically significant. All statistical analyses were conducted using SPSS version 25.0 (IBM, New York, US) and R version 4.1.3 (R Foundation for Statistical Computing). Age and body mass index (BMI) were included as covariates in all regression models. Age was adjusted because it is a primary determinant of ovarian reserve and a well‑known risk factor for POI. BMI was adjusted because it influences both reproductive hormone profiles and metal metabolism (e.g., through dietary intake, adipose tissue storage, and excretion), and may therefore confound the association between plasma metal levels and POI risk.

Table  1 presents the demographic and baseline characteristics of the control group ( n  = 31) and the POI group ( n  = 30). There were no significant differences between the two groups in terms of age at enrollment ( p  = 0.650), BMI ( p  = 0.125), marital status ( p  = 0.747), education level ( p  = 0.215), monthly family income ( p  = 0.238), water intake ( p  = 0.610), smoking status ( p  = 0.321), or alcohol consumption ( p  = 0.053). Notably, 53.3% of the POI group reported amenorrhea, with durations ranging from less than 6 months (33.4%) to over 12 months (16.7%). In contrast, none of the control participants reported amenorrhea. The majority of participants in both groups had a BMI within the normal range (18.5–24.0 kg/m²), were married, and had attained a college education or higher. Monthly family income and water intake patterns were also similar between the two groups. While smoking was rare in both groups, a higher proportion of POI participants reported alcohol consumption (66.7%) compared to controls (41.9%). The serum reproductive hormonal parameters of the case and control groups are shown in Table  2 . As expected, all participants in the POI case group had FSH levels over 25 IU/l (93.38 ± 34.04), whereas the mean level of the control group was 6.79 IU/l (6.79 ± 1.29) ( p  < 0.001). Women with POI also had significantly higher LH ( p  < 0.001), lower AMH ( p  < 0.001) and estradiol ( p  < 0.001) levels. Table 2 Biochemical parameters of patients with premature ovarian insufficiency ( n  = 30) and controls ( n  = 31) Variables Control ( n  = 31) POI ( n  = 30) p -Value FSH (IU/L) 6.79 ± 1.29 93.38 ± 34.04 p <0.001 a LH (IU/L) 4.74 ± 2.32 40.18 ± 17.75 p <0.001 a Estradiol (pmol/L) 174.00 (142.00-311.00) 79.00 (37.50–123.00) p <0.001 b AMH (ng/ml) 4.51 (3.41–5.61) 0.06 (0.00-0.06) p <0.001 b Data are presented as mean ± standard deviation or median (P 25 -P 75 ) Abbreviations : POI premature ovarian insufficiency, FSH follicle stimulating hormone, LH luteinizing hormone, AMH anti-Mullerian hormone a Student’s t-test of independent samples b Mann-Whitney U test Biochemical parameters of patients with premature ovarian insufficiency ( n  = 30) and controls ( n  = 31) Data are presented as mean ± standard deviation or median (P 25 -P 75 ) Abbreviations : POI premature ovarian insufficiency, FSH follicle stimulating hormone, LH luteinizing hormone, AMH anti-Mullerian hormone a Student’s t-test of independent samples b Mann-Whitney U test Table S1 summarizes the detection rate and distribution of all the quantified 15 plasma metals, which had a detection rate higher than 90% except for Ti. Table  3 compares plasma metal element concentrations between POI patients ( n  = 30) and controls ( n  = 31).Importantly, Mn levels were higher in the POI group (median [IQR]: 2.20 [1.58–2.81] µg/L) compared to controls (1.44 [1.04–2.57] µg/L; p  = 0.050). Similarly, V levels were significantly elevated in the POI group (mean ± SD: 1.19 ± 0.32 µg/L) versus controls (1.00 ± 0.38 µg/L; p  = 0.049). No significant differences were found for the remaining elements, including Fe, Zn, Se, Co, Mg, Sr, Li, Cu, Al, Cr, As, Ti, and I (all p  > 0.05). Table 3 Element concentrations of patients with premature ovarian insufficiency ( n  = 30) and controls ( n  = 31) Elements Control ( n  = 31) POI ( n  = 30) P -value Fe (µmol/L) 938.71 (709.48-1344.57) 930.20 (685.70-1203.77) 0.676 b Zn (µmol/L) 656.18 ± 115.20 647.17 ± 91.21 0.737 a Se (µg/L) 123.28 (111.27-136.71) 127.35 (110.73-135.12) 0.897 b Co (µg/L) 0.22 (0.12–0.38) 0.14 (0.07–0.26) 0.086 b Mg (mmol/L) 20294.40 (19263.45-22163.47) 20432.73 (19256.46-21541.84) 0.729 b Sr (µg/L) 29.04 (24.55–34.71) 26.11 (22.79–32.66) 0.428 b Li (µg/L) 1.33 (0.94–2.37) 1.31 (0.94–1.83) 0.644 b Cu (µmol/L) 891.08 ± 133.32 948.26 ± 137.11 0.104 a Al (µg/L) 1.40 (0.79–2.12) 1.69 (1.16–2.96) 0.100 b Cr (µg/L) 1.66 (1.32–2.62) 1.89 (1.45–3.44) 0.545 b Mn (µg/L) 1.44 (1.04–2.57) 2.20 (1.58–2.81) 0.050 b As (µg/L) 1.56 (0.83–2.58) 1.11 (0.67–1.78) 0.215 b Ti (µg/L) 8.08 (6.21–10.67) 9.27 (8.06–11.49) 0.237 b V (µg/L) 1.00 ± 0.38 1.19 ± 0.32 0.049 a I (µg/L) 60.00 (51.00–79.00) 66.00 (47.75–82.50) 0.740 b Data are presented as mean ± standard deviation or median (P 25 -P 75 ) Bold values indicate statistically significant ( P  < 0.05) Abbreviations : Fe Iron, Zn Zinc, Se Selenium, Co Cobalt, Mg Magnesium, Sr Strontium, Li Lithium, Cu Copper, Al Aluminum, Cr Chromium, Mn Manganese, As Arsenium, Ti Titanium, V Vanadium, I Iodine a Student’s t-test of independent samples b Mann-Whitney U test Element concentrations of patients with premature ovarian insufficiency ( n  = 30) and controls ( n  = 31) Data are presented as mean ± standard deviation or median (P 25 -P 75 ) Bold values indicate statistically significant ( P  < 0.05) Abbreviations : Fe Iron, Zn Zinc, Se Selenium, Co Cobalt, Mg Magnesium, Sr Strontium, Li Lithium, Cu Copper, Al Aluminum, Cr Chromium, Mn Manganese, As Arsenium, Ti Titanium, V Vanadium, I Iodine a Student’s t-test of independent samples b Mann-Whitney U test Table  4 presents the associations between plasma metal concentrations and the risk of POI. Significant associations were observed for Co, Cu, and V. In unadjusted models, higher Co levels were associated with a 98% reduced risk of POI (OR: 0.02; 95% CI: 0.00–0.72; p  = 0.032), while higher Cu and V levels were associated with increased POI risk (Cu: OR: 1.01; 95% CI: 1.00–1.01; p  = 0.032; V: OR: 8.65; 95% CI: 1.09–68.98; p  = 0.042). After adjusting for age and BMI, the associations for Co (aOR: 0.01; 95% CI: 0.00–0.53; p  = 0.021) and V (aOR: 7.85; 95% CI: 1.00–61.55; p  = 0.050) remained significant, while the association for Cu was attenuated (aOR: 1.01; 95% CI: 1.00–1.01; p  = 0.063). No significant associations were observed for the remaining metals. The RCS analysis revealed a non-linear relationship between element concentrations and the prevalence of POI (Fig.  1 ), showing that higher plasma Mn levels were associated with an increased risk of POI (P non-linear = 0.041). Table 4 Association between metal elements and risk for premature ovarian insufficiency Elements OR (95% CI) a p -Value aOR a, b (95% CI) p -Value Fe 1.00 (1.00–1.00) 0.638 1.00 (1.00–1.00) 0.618 Zn 1.00 (0.99-1.00) 0.476 1.00 (0.99–1.01) 0.953 Se 0.98 (0.94–1.02) 0.273 0.98 (0.94–1.02) 0.322 Co 0.02 (0.00-0.72) 0.032 0.01 (0.00-0.53) 0.021 Mg 1.00 (1.00–1.00) 0.660 1.00 (1.00–1.00) 0.624 Sr 0.94 (0.83–1.06) 0.303 0.93 (0.83–1.05) 0.257 Li 1.04 (0.49–2.20) 0.916 1.01 (0.47–2.15) 0.984 Cu 1.01 (1.00-1.01) 0.032 1.01 (1.00-1.01) 0.063 Al 1.32 (0.74–2.35) 0.343 1.26 (0.67–2.40) 0.473 Cr 0.96 (0.81–1.14) 0.632 0.97 (0.80–1.16) 0.714 Mn 0.86 (0.51–1.44) 0.570 0.85 (0.48–1.49) 0.566 As 0.63 (0.31–1.27) 0.197 0.66 (0.30–1.42) 0.285 Ti 2.25E + 14 (0.43-1.17E + 29) 0.056 4.56E + 14 (0.31-6.62E + 29) 0.058 V 8.65 (1.09–68.98) 0.042 7.85 (1.00-61.55) 0.050 I 0.99 (0.95–1.03) 0.607 0.98 (0.94–1.34) 0.464 Abbreviations : OR odds ratio, aOR adjusted odds ration, CI confidence interval, Fe Iron, Zn Zinc, Se Selenium, Co Cobalt, Mg Magnesium, Sr Strontium, Li Lithium, Cu Copper, Al Aluminum, Cr Chromium, Mn Manganese, As Arsenium, Ti Titanium, V Vanadium, I Iodine a Calculated using binary logistic regression b Model adjusted for age and BMI Association between metal elements and risk for premature ovarian insufficiency Abbreviations : OR odds ratio, aOR adjusted odds ration, CI confidence interval, Fe Iron, Zn Zinc, Se Selenium, Co Cobalt, Mg Magnesium, Sr Strontium, Li Lithium, Cu Copper, Al Aluminum, Cr Chromium, Mn Manganese, As Arsenium, Ti Titanium, V Vanadium, I Iodine a Calculated using binary logistic regression b Model adjusted for age and BMI Fig. 1 Dose–response relationships between plasma metals and risk of POI in the RCS analysis. The model was adjusted for age, BMI and other metals. Abbreviations: Fe, Iron; Zn, Zinc; Se, Selenium; Co, Cobalt; Mg, Magnesium; Sr, Strontium; Li, Lithium; Cu, Copper; Al, Aluminum; Cr, Chromium; Mn, Manganese; As, Arsenium; Ti, Titanium; V, Vanadium; I Iodine Dose–response relationships between plasma metals and risk of POI in the RCS analysis. The model was adjusted for age, BMI and other metals. Abbreviations: Fe, Iron; Zn, Zinc; Se, Selenium; Co, Cobalt; Mg, Magnesium; Sr, Strontium; Li, Lithium; Cu, Copper; Al, Aluminum; Cr, Chromium; Mn, Manganese; As, Arsenium; Ti, Titanium; V, Vanadium; I Iodine The pairwise correlations among the 15 plasma metals ranged from − 0.31 to 0.55, indicating modest inter-element associations (Fig. S1). To address potential multicollinearity arising from these correlations, we employed BKMR, a robust method capable of handling high correlations among exposure variables. The overall association between combined exposure to the 15 plasma metals and POI risk showed an increasing trend (Fig.  2 ), which was consistent with the primary results and the Qgcomp analysis results (Fig.S3A), but did not reach statistical significance Since elements, including Mn, Co, Cu, and V, showed significant correlations with POI risk in the single factor association analysis, we evaluated the joint and individual effects of the four metals by performing BKMR. Conditional posterior inclusion probabilities (condPIPs) identified Mn (0.819) as the most influential metal, followed by Co (0.738), Cr (0.735), and V (0.707) (Table S2). Fig. 2 Effects of the four-metal mixture on POI risk estimated using Bayesian kernel machine regression. A Joint effect of the metal mixture (estimates and 95% confidence intervals) showing changes in POI risk when all metals were set at specific percentiles compared to their 50th percentile. B Single-metal effect (estimates and 95% confidence intervals) showing changes in POI risk when an individual metal varies from its 25th to 75th percentile, while other metals are fixed at their 25th (red), 50th (green), or 75th (blue) percentiles. C Univariate exposure–response trends and 95% confidence intervals for each metal as a continuous variable, with other metals held at their median values. D Bivariate exposure–response functions for metal interactions: V when Co is fixed at the 25th (red), 50th (green), or 75th (blue) percentile and Cu and Mn are fixed at their medians (top right panel); other panels follow the same interpretation accordingly. Models were adjusted for age and BMI. Abbreviations: Cu, copper; Mn, manganese; Co, cobalt; V, vanadium Effects of the four-metal mixture on POI risk estimated using Bayesian kernel machine regression. A Joint effect of the metal mixture (estimates and 95% confidence intervals) showing changes in POI risk when all metals were set at specific percentiles compared to their 50th percentile. B Single-metal effect (estimates and 95% confidence intervals) showing changes in POI risk when an individual metal varies from its 25th to 75th percentile, while other metals are fixed at their 25th (red), 50th (green), or 75th (blue) percentiles. C Univariate exposure–response trends and 95% confidence intervals for each metal as a continuous variable, with other metals held at their median values. D Bivariate exposure–response functions for metal interactions: V when Co is fixed at the 25th (red), 50th (green), or 75th (blue) percentile and Cu and Mn are fixed at their medians (top right panel); other panels follow the same interpretation accordingly. Models were adjusted for age and BMI. Abbreviations: Cu, copper; Mn, manganese; Co, cobalt; V, vanadium When all four metals were at or above their 55th percentiles, a joint risk effect on POI was observed (Fig.  2 A), consistent with the analysis results of Qgcomp (Fig.S3B). Univariate analysis revealed that Mn had a significant effect on POI risk, with its effect size increasing as the concentrations of the other three metals rose from their 25th to 75th percentiles (Fig.  2 B). Although no statistically significant interactions were detected due to overlapping confidence intervals, Mn’s effect remained significant when the other metals were fixed at their 75th percentiles, suggesting potential interactions with Co, Cu, or V. Exposure-response curves demonstrated a non-linear relationship for Mn and linear relationships for Co, Cu, and V with POI risk (Fig.  2 C). Bivariate analysis further suggested no significant interactions between Mn and Co, Cu, or V in their association with POI risk (Fig.  2 D).

Female infertility is a significant reproductive health issue that impacts not only individuals and families but also broader societal population growth. Over the past three decades, the number of female infertility cases has risen by 7.06 million in China and 56.71 million worldwide [ 1 ]. Recent experimental and epidemiologic studies have linked exposure to essential and non-essential trace elements to ovarian aging [ 2 – 4 ], a process characterized by the progressive decline in both the quantity and quality of ovarian follicles, which rises concerns about its effects on female fertility and ovarian endocrine function [ 5 ]. When ovarian aging occurs prematurely or is accelerated, it manifests clinically as diminished ovarian reserve (DOR) and/or premature ovarian insufficiency (POI) [ 6 – 8 ]. POI, a condition of aberrant ovarian aging [ 6 ], leads to accelerated menopause and subfertility or infertility before the age of 40 [ 9 ]. This condition arises from a reduction in ovarian follicles and the cessation of ovarian function, which normally serves both reproductive and endocrine roles [ 10 ]. Beyond its reproductive implications, premature loss of ovarian function is associated with long-term health risks, including an increased lifetime risk of cardiovascular and neurocognitive disease, skeletal fragility, and reduced overall life expectancy [ 11 ]. Although the prevalence of POI is influenced by various factors, the underlying cause of most cases remain poorly understood. Emerging research highlights the critical role of trace elements (TEs) exposure in POI. Imbalances in TEs can contribute to ovarian aging, marked by reduced enzyme activity, hormonal imbalances, ovulatory disorders, and decreased fertility [ 2 ]. Studies have shown that elevated urinary thallium levels are positively associated with POI risk, correlating with increased follicle-stimulating hormone and luteinizing hormone levels, while negatively impacting anti-Müllerian hormone and estradiol [ 12 ]. Cadmium exposure, a toxic heavy metal, has been linked to adverse reproductive health outcomes, including POI [ 13 , 14 ]. Additionally, exposure to a mixture of tributyltin (TBT) and mercury (Hg) disrupts the hypothalamic-pituitary gonadal (HPG) axis, exacerbating POI symptoms and reducing fertility in female rats [ 15 ]. Higher levels of copper (Cu) and selenium (Se) have also been associated with an increased risk of POI [ 16 ]. However, previous studies have primarily focused on the effects of isolated or single metal on POI, leaving the critical metals responsible for its onset poorly defined. Importantly, most individuals are exposed to multiple metals simultaneously, and the synergistic or antagonistic interactions among these metals may alter the impact of any single metal. Therefore, assessing the combined effects of metal mixtures is more clinically relevant. Despite this, the collective impact of plasma metals on POI risk remains understudied, and traditional statistical methods may limit the ability to discern the relative importance of different metals. In the past decade, advanced statistical methods have been developed to evaluate the combined effects of exposure mixtures on health outcomes. These include weighted quantile sum regression (WQS), quantile-based g-computation (QGC), and Bayesian kernel machine regression (BKMR). These methods have been widely applied to study the effects of metals on conditions such as hypertension [ 17 ], cardiovascular diseases [ 18 ] and hyperuricemia [ 19 ]. These tools enable researchers to investigate both the overall and individual effects of metal mixtures on health outcomes. In this study, we aim to employ both traditional logistic regression and the advanced BKMR method to assess the association between single and mixed plasma metal exposures and POI risk, and identify the key elements contributing to the risk of POI. For metals with prior epidemiologic evidence (Cu, Se), our analysis was hypothesis‑driven. For the remaining metals ((Iron (Fe), Zinc (Zn), Cobalt (Co), Magnesium (Mg), Strontium (Sr), Lithium (Li), Aluminum (Al), Chromium (Cr), Manganese (Mn), Arsenium (As), Titanium (Ti), Vanadium (V), Iodine (I)), the analysis was exploratory and hypothesis‑generating, based on general toxicological literature suggesting potential roles in oxidative stress, endocrine disruption, or ovarian toxicity. Significant findings from this exploratory component require validation in future studies.

This study investigated the association between plasma metal mixtures and the risk of POI using advanced statistical methods, including BKMR. Key findings revealed that Mn, Cu, and V were significantly associated with an increased risk of POI, while Co exhibited a potential protective effect. Mn emerged as the most influential metal, with higher plasma levels strongly linked to POI risk. These results highlight the complex interplay between environmental metal exposures and ovarian function, providing new insights into the role of trace elements in reproductive health. Mn is one of the essential trace elements with unique functions in the body and is widely involved in a variety of physiological processes, including antioxidant defense, enzyme activity regulation and cell signaling [ 22 ]. Mn is a cofactor of many enzymes, especially superoxide dismutase (MnSOD), which plays a key role in scavenging free radicals and reducing oxidative stress [ 23 ]. Being in agreement with the findings that, exposure to Mn can cause reduced sperm count, sperm shape deformities and lead to abnormal secretion of sex hormones, especially, which disturbs the EP1 and EP2 receptors of paracrine prostaglandin E2 (PGE 2 ) in neurons to abnormal levels of Gonadotropin-releasing hormone (GnRH) secretion [ 24 ]. In addition, ovarian tissue is highly sensitive to oxidative stress, and excessive free radicals will damage the mitochondria and DNA of oocytes and granulosa cells, resulting in decreased egg quality and hormone secretion disorders [ 25 , 26 ]. As a component of MnSOD, Mn can effectively remove exceed free radicals and protect ovarian granulosa cells from oxidative damage [ 27 ]. Studies have shown that moderate Mn intake can increase the antioxidant capacity of the ovaries, thereby improving follicle development and ovulation function [ 28 ]. In addition, Mn is involved in the metabolism of carbohydrates, amino acids and cholesterol, and has important effects on the synthesis and regulation of hormones [ 28 , 29 ].Mn modulates the synthesis of cholesterol synthesis as a cofactor for several enzymes involved in cholesterol synthesis including mevalonate kinase, geranyl pyrophosphate synthetase, and farnesyl pyrophosphate synthase [ 30 ]. Cholesterol is a precursor for synthesis of steroid hormones including progestogens, androgens, and estrogens [ 31 ], all essential for the optimal functioning of ovarian function and is of fundamental importance to ensuring reproductive fertility and maintaining physiological homeostasis [ 32 ]. Additionally, Mn plays an important role in cell proliferation and differentiation and may directly or indirectly affect follicle development. Animal studies have shown that Mn deficiency can lead to abnormal follicular development, while moderate Mn supplementation can promote follicular maturation and ovulation [ 28 ]. These findings suggest the potential importance of Mn in ovarian function through mechanisms such as antioxidant, hormone regulation and follicle development, and maintaining an appropriate intake of Mn is essential for ovarian health. Nevertheless, disturbed homeostasis of Mn in the body has been connected with susceptibility to decreased fertility, causing mitochondrial dysfunction and apoptosis [ 33 ]. In our study, median concentrations of Mn were significantly higher in cases than in controls and higher plasma Mn levels were related to increased POI risk, consistent with a trend of a “U-shaped” association of essential metals. However, more large-scale clinical studies are needed to clarify the specific role and mechanism of Mn in human ovarian function and provide new intervention strategies for female reproductive health. Cu is an essential trace element required for various physiological processes, excessive Cu can lead to toxicity, oxidative stress, and hormonal imbalances [ 34 ], all of which can negatively impact ovarian function. The ovaries are particularly vulnerable to oxidative damage due to their high metabolic activity and the presence of delicate structures like follicles and oocytes [ 26 ]. An excess of Cu ions can catalyze the production of ROS, leading to oxidative stress (OS) [ 35 ].OS Impaired follicular development and reduced oocyte quality, leading to accelerated ovarian aging or POI [ 26 , 36 ].One research reported that high Cu exposure increased risk of ovarian tissue damage and fibrosis by triggering both intrinsic and extrinsic apoptotic pathways and regulating key ovarian genes in oxidative stress-mediated ovarian dysfunction [ 37 , 38 ]. Cu plays a role in the synthesis and metabolism of hormones. A study showed that Cu had been shown to be a potential regulator of the hormone secretion function of Granulosa cells (GCs) [ 30 ]. Excess Cu also disrupts steroidogenesis in ovarian granulosa cells through the FSHR/CYP19A1 pathway [ 39 ] and impaired luteal formation and progesterone production in the mutant females by increasing intracellular ROS [ 40 ], which can affect fertility. Some studies suggest that women with Polycystic ovary syndrome (PCOS) may have elevated Cu levels [ 41 – 43 ]. In addition, except for affecting ovarian follicle development by causing oxidative stress and then interfere with hormone signaling, high Cu levels may impair function of other aspects, including endometrial receptivity [ 44 ], fallopian tube [ 45 ], systemic immune system [ 46 ] and neurological system [ 47 ], further effecting fertility reproductive health. In short, elevated Cu levels can negatively impact ovarian function and overall reproductive health through oxidative stress, hormonal imbalances, and inflammation. These effects may manifest as menstrual irregularities, ovulation disorders, reduced fertility, or exacerbation of conditions like PCOS. Consistent with the above, In the analysis of the univariate regression and the BKMR model of this study and BKMR model, Cu were significantly associated with POI risk. Therefore, maintaining balanced Cu levels through diet, lifestyle, and medical management is essential for preserving ovarian health and reproductive function. Recent researches have showed that occupational and environmental exposure to V has been associated with toxicities in reproductive, respiratory, and cardiovascular systems [ 48 – 50 ]. As a redox-active element, V can act as a strong prooxidant and induce OS at certain levels [ 51 ], which is a key mechanism underlying ovarian dysfunction. OS damages cellular components, including lipids, proteins, and DNA, leading to impaired ovarian function [ 36 ]. Studies have shown that maternal exposure to PM2.5, which contains V, can decrease ovarian reserve and impairs ovarian follicular development in offspring mice, involving the activation of the PI3K/AKT/FoxO3a pathway and reactive oxygen species (ROS)-dependent NF-κB pathway [ 52 ]. In V-treated mice, anestrous, irregular estrous cycles, low serum 17β-estradiol and progesterone concentrations, decreased secondary and pre-ovulation follicle diameter, as well as a thickening of the myometrium and endometrial stroma were observed [ 53 ].Studies have shown elevated V concentrations in urine and seminal plasma may be adversely related to male semen quality and the reproductive toxicity [ 54 ].Elevated V levels, often associated with environmental pollution, can cause ovarian dysfunction through multiple mechanisms, including oxidative stress, epigenetic modifications, inflammation, and endocrine disruption. These effects are particularly concerning for women of reproductive age, as they may lead to reduced fertility and other reproductive health issues and the related research is few. In our study, compared with the controls, Median concentrations of Mn and V were significantly higher in POI group and higher plasma V levels were related to increased POI risk. Thus, Further research is needed to fully understand the extent of V’s impact and to develop strategies for mitigating its effects. Co, as a component of vitamin B12, is an essential element and crucial for some biological processes, including DNA synthesis and regulation, nervous system function, red blood cell formation, etc [ 55 ]. Deficiency of Co can affect pituitary LH level, oestrogen level and stopping ovulation [ 56 ]. In animals, Co, as a probable factor affecting the secretory activity of ovarian granulosa cells, would be expected to affect progesterone and insulin-like growth factor 1 (IGF-1) concentrations [ 57 ]. In addition, a study on porcine ovarian granulosa cells, that the release of IGF-1 by these cells varies according to the concentration of Co administered [ 58 ]. Maternal Co in the concentration range of 0.34–0.90 ng/mL played a protective role against spontaneous preterm birth [ 59 ]. The present study’s results showed that the addition of small amounts of Co promoted the synthesis of E2, but the exact mechanism is still unclear [ 30 ]. Meanwhile, although progesterone secreted by granulosa cells was inhibited by Co at a concentration of 0.09 mg/mL, but no such relationship was shown at higher Co concentrations [ 57 ]. A research reported that blood Co concentrations of 700–800 µg Co/L and higher may pose a risk of more serious neurological, reproductive, or cardiac effects [ 60 ]. In our study, In the content interval of 0.07–0.38 µg/L, median concentration of Co was slightly higher in the POI group than in the control group, although there was no significant difference. Also, Co may play a protective role in mixed exposure according to poi risk BKMR model analysis. Yet, little is known about Co’s effects on the ovaries and we may need to design more scientific studies to explore this issue. This study has several strengths. First, the use of BKMR allowed us to evaluate the combined effects of metal mixtures, addressing the limitations of traditional single-metal analyses. Second, the case-control design and demographic similarity between POI and control groups minimized confounding factors. However, several limitations should be acknowledged. The small sample size may limit the generalizability of our findings and the ability to detect weak associations. Moreover, plasma metal levels may not fully reflect the ovarian microenvironment, as follicular fluid provides a more direct measure of exposure to developing oocytes. Although BKMR allows for the exploration of potential interactions among metals, the underlying mechanisms of such interactions in the context of POI remain incompletely understood. Based on existing literature, several plausible mechanisms could explain synergistic or antagonistic effects. For instance, manganese and iron share common divalent metal transporters; excessive manganese may interfere with iron homeostasis, thereby exacerbating oxidative stress in ovarian cells. Similarly, lead and cadmium may act additively to disrupt steroidogenesis by activating oxidative pathways and inhibiting antioxidant enzymes. However, given the exploratory nature of our study and the limited sample size, these proposed mechanisms require confirmation in future experimental models and larger cohort studies. Therefore, future research should include larger, more diverse populations and assess metal concentrations in follicular fluid to validate our findings. Furthermore, while our study focused on the overall risk of diminished ovarian reserve associated with plasma metals, we did not examine the direct effects of metals on each reproductive hormone. Future studies are warranted to perform mediation analysis or structural equation modeling to clarify whether metal‑induced hormone alterations mediate the increased risk of POI. A potential limitation of this study is the hospital‑based design. Both cases and controls were recruited from a single hospital, which may introduce selection bias, including Berkson’s bias, if hospital attendance is associated with metal exposure levels. Although our controls were selected from women attending routine gynecological check‑ups (a reason unlikely to be related to metal exposure), we cannot completely exclude the possibility that their exposure distribution differs from that of the general population. Therefore, our findings should be interpreted with caution and validated in population‑based or community‑based prospective cohorts. In our study, we also analyzed the combined exposure to the 15 plasma metals in relation to POI risk. The observed increasing trend, although not statistically significant, was consistent with our preliminary findings. This might be due to sample size limitations (resulting in insufficient detection power), more complex interactions among the more kinds of metals, and biological mechanisms such as antagonistic effects (e.g., Se counteracting Cr [ 61 ]), non-linear dose-response relationships, and exposure levels below toxic thresholds. Additionally, a single time‑point measurement of plasma metals may not capture cumulative ovarian exposure over the etiologic window. These hypotheses remain speculative and require validation in experimental models and longitudinal studies with larger sample sizes and repeated exposure measurements. For the remaining plasma metals (Fe, Zn, Se, Co, Mg, Sr, Li, Cu, Al, Cr, As, Ti and I), we did not observe statistically significant associations with POI risk in either the single‑metal models or the BKMR mixture analysis. This lack of significance does not necessarily imply that these metals are biologically inert with respect to ovarian function. Possible explanations include the relatively low exposure levels in our study population, the limited sample size that may have reduced power to detect modest effects, or the presence of unmeasured confounding. Additionally, the effect of a given metal may be modified by other co‑exposed metals or by individual susceptibility factors such as genetic polymorphisms in metal metabolism pathways, and their potential biological relevance therefore cannot be ruled out. Given the limited sample size of this study, the results of BKMR may be particularly prone to overfitting and instability in estimating nonlinear or interaction effects, while the estimation of overall mixed effects by QGComp may also be unstable. Both of these methods’ results should be interpreted with caution. Therefore, we emphasize that all the results of the mixed models are exploratory and preliminary. Larger sample size studies with broader exposure ranges and strict FDR correction are needed to validate our findings and to more reliably assess the potential impact of these non-significant metals on the risk of premature ovarian failure.

This study revealed significant links between plasma metal concentrations and POI risk, underscoring Mn’s potential as a biomarker and risk factor for POI in both individual and combined assessments. Conversely, essential trace elements, notably Co, showed inverse associations trend with POI risk. Our findings facilitate exploring and understanding the mechanisms of combined trace element exposure on POI risk, paving the way for preventive strategies.

Supplementary Material 1. Supplementary Material 1.