Serum trace and heavy metal exposure and IVF/ICSI outcomes: modifying effects of maternal diet in a prospective cohort.

This investigation used a prospective cohort design and took place at the Center for Reproductive Medicine, Peking University Third Hospital in Beijing, China. Between March and December 2023, we screened women who were starting their first cycle of in-vitro fertilisation or intracytoplasmic sperm injection (IVF/ICSI). Eligible participants were 20–38 years of age, had experienced at least twelve months of unsuccessful, unprotected intercourse, and provided both a fasting serum sample (cycle days 2–4) and a baseline dietary questionnaire. We excluded women with pre-existing endocrine or metabolic disease, severe systemic or autoimmune disorders, systemic hormone or corticosteroid use within the past three months, prior ovarian or uterine surgery, congenital uterine malformation, adenomyosis, endometriosis, hydrosalpinx, previous assisted-reproduction attempts, or involvement in pre-implantation genetic testing or donor-oocyte cycles. The exclusion of patients with endometriosis and other structural uterine or ovarian pathologies was specifically intended to minimize clinical heterogeneity and ensure the reliability of embryological outcomes. A total of 396 women met all criteria and were enrolled. The study complied with the principles of the Declaration of Helsinki (2013 revision) and was approved by the Ethics Committee of Peking University Third Hospital (approval No. M2022722). Written informed consent was obtained from all participants prior to enrolment. Venous blood was obtained after an overnight fast on cycle days 2 to 4, that is, before the first injection of gonadotrophins. Within two hours the samples were centrifuged at 4 °C, and the serum was transferred to acid-washed polypropylene vials and stored at − 80 °C. The two macro-elements, magnesium and calcium, were measured by inductively coupled plasma atomic-emission spectrometry (ICP-AES; iCAP 6000, Thermo Fisher Scientific). The remaining trace elements were quantified by quadrupole inductively coupled plasma mass spectrometry (ICP-MS; ELAN DRC II, PerkinElmer, and 7700x, Agilent). For ICP-MS analysis, 0.10 mL of serum was diluted to 2 mL with ultrapure water containing one per cent nitric acid; rhenium at 4 ng mL⁻¹ served as the internal standard. For ICP-AES, 0.05 mL of serum was diluted to 5 mL with the same acid matrix and stabilised with yttrium at 10 µg mL⁻¹. Calibration curves with five concentration points were prepared immediately before each analytical run. Every batch contained procedural blanks, reference material (ClinChek Trace Elements Serum, Level II) and at least five per cent blinded duplicates. The limit of detection for each element was defined as three times the standard deviation of ten low-level standards. Intra- and inter-batch coefficients of variation remained below ten per cent. Because concentration distributions were right-skewed, natural logarithms were used in all statistical models. At baseline, participants completed a brief semi-quantitative food frequency questionnaire covering habitual dietary intake over the previous three months. This assessment window was chosen to align with the approximately 90-day duration of human folliculogenesis, capturing the nutritional environment during the critical period of oocyte recruitment and maturation. The questionnaire included 13 food groups, with intake frequencies recorded as times per week. Values > 35 times/week were considered implausible and treated as missing. Dietary variables were standardized and subjected to principal component analysis (PCA) with varimax rotation to identify dietary patterns. The number of components was determined based on eigenvalues > 1, scree plot inspection, and interpretability. Pattern scores were calculated for each participant using the regression method and used in further stratified and interaction analyses. Sensitivity analysis was conducted by excluding three food groups with low sampling adequacy to assess the robustness of the identified patterns. We obtained all outcome data from electronic clinical records. Embryological outcomes comprised the total number of oocytes retrieved, the number of transferable embryos, and the number of good-quality embryos, which were defined by centre criteria (typically six to ten cells with < 20% fragmentation). The fertilisation rate was calculated as the number of normally fertilised two-pronuclear (2PN) zygotes divided by the number of metaphase-II oocytes. The good embryo rate was calculated as the number of good-quality embryos divided by the total number of embryos evaluated. Biochemical pregnancy was defined as a serum β-human chorionic-gonadotrophin concentration of at least 25 IU·L⁻¹ fourteen days after embryo transfer. Clinical pregnancy required ultrasonographic evidence of at least one intra-uterine gestational sac with a heartbeat approximately 30 days after transfer. Live birth was defined as the delivery of at least one live-born infant after 28 weeks of gestation. Laboratory staff who graded embryos were unaware of metal concentrations and dietary scores. All statistical work was performed in R (version 4.3.2, R Foundation for Statistical Computing, Vienna, Austria). A two-sided P value of 0.05 defined statistical significance. Continuous variables are summarised as means ± standard deviations or as medians with inter-quartile ranges, whereas categorical variables are presented as counts and percentages. Serum metal concentrations were natural-log-transformed to reduce skewness. Each transformed value was expressed per inter-quartile-range (IQR) increase, so every effect estimate describes the expected change in the outcome when exposure rises from the 25th to the 75th percentile. Outcome-specific models were selected according to data type. We applied negative binomial regression to count outcomes (total oocytes, transferable embryos, and good-quality embryos), fractional logistic regression for proportion outcomes (fertilization rate, good embryo rate), and logistic regression for binary outcomes. All models were adjusted for maternal age, BMI, infertility type, active smoking, passive smoke exposure, and critical clinical indicators, including anti-Müllerian hormone (AMH), basal FSH, antral follicle count (AFC), total gonadotropin (Gn) dose, and PCOS status. To minimize bias and ensure the maximum use of our cohort data, missing values in clinical covariates were addressed using Multiple Imputation by Chained Equations (MICE). We split each dietary-pattern score at its median into low- and high-adherence groups. Metal–outcome models were fitted within each stratum, and a metal × diet interaction term was included in the full models to test for effect modification using the Wald test. To address the risk of false-positive findings due to multiplicity, we implemented the Benjamini-Hochberg False Discovery Rate (FDR) procedure. We reported both nominal P -values ( P  < 0.05) and FDR-adjusted P-values ( PFDR  < 0.05). Interaction analyses were treated as exploratory. Furthermore, considering that essential elements often exhibit non-monotonic biological effects, we explored potential non-linear or threshold relationships for Mg, Ca, and Se using Restricted Cubic Splines (RCS) with three knots. Model fit for linear versus non-linear associations was compared using the Akaike Information Criterion (AIC).

A total of 396 women undergoing their first IVF/ICSI cycle between March and December 2023 were enrolled. The mean age was 32.3 years (SD: 3.2), and the mean BMI was 23.1 kg/m² (SD: 4.0). Most were non-smokers (97.7%), and 27.5% reported passive smoke exposure. Primary infertility accounted for 65.9% of cases, with tubal factor being the most common cause (50.3%). Regarding clinical and endocrine profiles, the mean AMH level was 4.36 ng/mL (SD: 3.17), the mean basal FSH was 6.32 mIU/mL, and the mean AFC was 13. Additionally, 17.93% of the cohort ( n  = 71) had a diagnosis of PCOS. During the treatment phase, the antagonist protocol was used in 90.4% of cycles, with a mean total gonadotropin dose of 2384.13 IU. Conventional IVF was performed in 69.2%, ICSI in 28.5%, and both in 2.3%. Fresh and frozen embryo transfers were conducted in 47.0% and 51.8% of cases, respectively. Biochemical and clinical pregnancy rates were 61.4% and 56.6%, respectively, and the live birth rate was 35.6%. Thirteen food groups were used to derive maternal dietary patterns through PCA. Bartlett’s test (χ² = 1119.17, p  < 0.001) and a KMO value of 0.68 supported the suitability of the data. Four major dietary patterns were identified by PCA with varimax rotation, explaining 58.4% of the total dietary variance(Figure S1; Table S1). These patterns were interpreted based on food groups with high factor loadings (∣loadings∣>0.4): (1) the Alcohol–Processed Food pattern was characterized by spirits (0.89), red wine (0.84), pickled foods (0.49), and barbecue (0.48); (2) the Traditional Balanced Diet was dominated by staple foods (0.84), vegetables (0.76), and meat (0.63); (3) the Protein-Rich/Modern Nutritive pattern showed high loadings for milk (0.80), soy products (0.76), and eggs (0.70), reflecting a mix of animal- and plant-based protein sources; and (4) the Caffeinated-Beverage pattern was defined by tea (0.85) and coffee (0.61), with a secondary contribution from seafood (0.41). Notably, alcohol intake (spirits and red wine) loaded exclusively on Pattern 1, with negligible loadings (< 0.15) on all other components. Individual pattern scores were used for subsequent stratified and interaction analyses. All ten trace metals were detected in 100% of serum samples (Fig.  1 ; Table S3). The highest concentrations were observed for Fe, with a geometric mean (GM) of 1652.15 ng/mL and a wide range up to 12,632.57 ng/mL, followed by Ca (GM: 103.62 ng/mL) and Se (GM: 117.52 ng/mL). In contrast, the lowest concentrations were noted for Cd (GM: 0.18 ng/mL), Pb (GM: 0.78 ng/mL), and Mo (GM: 1.43 ng/mL). Most metals showed right-skewed distributions, as reflected by high standard deviations and upper percentiles (Table 1 ). Fig. 1 Serum concentrations of ten trace metals in women undergoing IVF/ICSI. Note: Boxplots show serum concentrations (ng/mL) of Mg, Ca, Se, Fe, Sr, Ni, Pb, Cd, Sn, and Mo. Boxes indicate the interquartile range; horizontal lines mark medians; diamonds denote means. All metals were detected in 100% of samples Serum concentrations of ten trace metals in women undergoing IVF/ICSI. Note: Boxplots show serum concentrations (ng/mL) of Mg, Ca, Se, Fe, Sr, Ni, Pb, Cd, Sn, and Mo. Boxes indicate the interquartile range; horizontal lines mark medians; diamonds denote means. All metals were detected in 100% of samples Table 1 Demographic and clinical characteristics of the study population ( n  = 396) Characteristic Data Age, years 32.28 (3.21) BMI, kg/m² 23.08 (4.02) Smoking status  Non-smoker 387 (97.7%)  Smoker 9 (2.3%) Passive smoke exposure 109 (27.5%) Infertility type  Primary 261 (65.9%)  Secondary 135 (34.1%) Infertility cause  Tubal factor 199 (50.3%)  Ovulatory disorder 82 (20.7%)  Pelvic/uterine factor 79 (19.9%)  Male factor 36 (9.1%) Infertility duration, years 2.97 (2.26) AMH(ng/mL) 4.36(3.17) Basal FSH(mIU/mL) 6.32(2.1) AFC 13.04(5.41) Total dosage of Gn(IU) 2384.13(931.89) PCOS 71(17.93%) COS protocol  Antagonist 358 (90.4%)  Long protocol 26 (6.6%)  Ultra-long 9 (2.3%)  Short 1 (0.3%)  Ultra-short 2 (0.5%) Fertilization method  IVF 274 (69.2%)  ICSI 113 (28.5%)  IVF + ICSI 9 (2.3%) Type of cycle  Fresh 186 (47.0%)  Frozen-thawed 205 (51.8%)  No embryo available 5 (1.2%) Oocytes retrieved 13.61 (7.58) Transferable embryos 3.88 (3.09) Good-quality embryos 4.83 (4.02) Fertilization rate 0.62 (0.23) Good-quality embryo rate 0.39 (0.24) Biochemical pregnancy 243 (61.4%) Clinical pregnancy 224(56.6%) Live birth 141(35.6%) Continuous variables are presented as mean (standard deviation), and categorical variables are shown as number (percentage, %). A “No embryo available” refers to cycles in which no transferable embryos were obtained due to failed fertilization or poor embryo development AMH Anti-Müllerian hormone, FSH follicle-stimulating hormone, AFC Antral follicle count, Gn gonadotropin, PCOS Polycystic ovary syndrome, BMI Body mass index, COS Controlled ovarian stimulation, IVF In vitro fertilization; ICSI Intracytoplasmic sperm injection Demographic and clinical characteristics of the study population ( n  = 396) Continuous variables are presented as mean (standard deviation), and categorical variables are shown as number (percentage, %). A “No embryo available” refers to cycles in which no transferable embryos were obtained due to failed fertilization or poor embryo development AMH Anti-Müllerian hormone, FSH follicle-stimulating hormone, AFC Antral follicle count, Gn gonadotropin, PCOS Polycystic ovary syndrome, BMI Body mass index, COS Controlled ovarian stimulation, IVF In vitro fertilization; ICSI Intracytoplasmic sperm injection We evaluated the associations between dietary patterns and IVF outcomes using multivariable models adjusted for the full suite of clinical covariates, including AMH and total Gn dose (Tables  2 and 3 ). After applying the False Discovery Rate procedure to account for multiple testing, we observed that while some associations achieved nominal significance ( P  < 0.05), they did not meet the stringent threshold of P-fdr  < 0.05. Table 2 Associations between maternal dietary patterns and laboratory indicators of IVF outcomes Results are presented as adjusted relative risks (RRs) with 95% confidence intervals (CIs) for count outcomes (total oocytes retrieved, number of transferable embryos, and number of good-quality embryos), and as adjusted regression coefficients (β) for proportion outcomes (fertilization rate and good embryo rate). All models were adjusted for maternal age, BMI, infertility type, smoking status, passive smoke exposure, anti-Müllerian hormone (AMH), basal FSH, antral follicle count (AFC), total gonadotropin (Gn) dose, and polycystic ovary syndrome (PCOS) status. Values in bold indicate nominal significance ( P < 0.05). Asterisks (*) indicate associations that remained statistically significant after False Discovery Rate (FDR) correction (PFDR<0.05). Ref = reference group (Q1, lowest quartile) Associations between maternal dietary patterns and laboratory indicators of IVF outcomes Results are presented as adjusted relative risks (RRs) with 95% confidence intervals (CIs) for count outcomes (total oocytes retrieved, number of transferable embryos, and number of good-quality embryos), and as adjusted regression coefficients (β) for proportion outcomes (fertilization rate and good embryo rate). All models were adjusted for maternal age, BMI, infertility type, smoking status, passive smoke exposure, anti-Müllerian hormone (AMH), basal FSH, antral follicle count (AFC), total gonadotropin (Gn) dose, and polycystic ovary syndrome (PCOS) status. Values in bold indicate nominal significance ( P < 0.05). Asterisks (*) indicate associations that remained statistically significant after False Discovery Rate (FDR) correction (PFDR<0.05). Ref = reference group (Q1, lowest quartile) Despite this, several clinically relevant trends emerged. Adherence to the Caffeinated-Beverage pattern showed a strong nominal association with impaired embryo development; specifically, women in the highest quartile (Q4) had a 22% reduction in the good embryo rate (adjusted OR = 0.78, 95% CI: 0.65–0.93, P  = 0.01, P-fdr  = 0.36) and a lower number of good-quality embryos ( P  = 0.048). Furthermore, the Protein-Rich/Modern Nutritive pattern (Q3) was nominally associated with a substantial decrease in the likelihood of live birth (adjusted OR = 0.46, 95% CI: 0.25–0.86, P  = 0.02, P-fdr  = 0.49). Although these dietary findings should be interpreted as exploratory due to the conservative nature of the FDR adjustment in a cohort of this size, the consistency of the effect directions across multiple laboratory and clinical endpoints suggests potential biological significance. Table 3 Associations between maternal dietary patterns and pregnancy outcomes Results are presented as adjusted odds ratios (ORs) with 95% confidence intervals (CIs) derived from multivariable logistic regression models. All models were adjusted for maternal age, BMI, infertility type (primary or secondary), smoking status, passive smoke exposure, anti-Müllerian hormone (AMH), basal FSH, antral follicle count (AFC), total gonadotropin (Gn) dose, and polycystic ovary syndrome (PCOS) status. The False Discovery Rate (FDR) procedure was applied to account for multiple testing across all outcomes and dietary patterns. Values in bold indicate nominal significance ( P < 0.05). No associations reached the statistical significance threshold after FDR correction (PFDR<0.05). Ref = reference group (Q1, lowest quartile) Associations between maternal dietary patterns and pregnancy outcomes Results are presented as adjusted odds ratios (ORs) with 95% confidence intervals (CIs) derived from multivariable logistic regression models. All models were adjusted for maternal age, BMI, infertility type (primary or secondary), smoking status, passive smoke exposure, anti-Müllerian hormone (AMH), basal FSH, antral follicle count (AFC), total gonadotropin (Gn) dose, and polycystic ovary syndrome (PCOS) status. The False Discovery Rate (FDR) procedure was applied to account for multiple testing across all outcomes and dietary patterns. Values in bold indicate nominal significance ( P < 0.05). No associations reached the statistical significance threshold after FDR correction (PFDR<0.05). Ref = reference group (Q1, lowest quartile) Multivariable regression models, adjusted for maternal age, BMI, smoke exposure, infertility type, and clinical indicators including AMH, basal FSH, AFC, total Gn dose, and PCOS status, revealed several significant associations between serum metal concentrations and IVF/ICSI outcomes (Fig.  2 ; Table S5). Effect estimates are reported as RRs for count outcomes and ORs for proportion and pregnancy outcomes per interquartile range increase in log-transformed serum metal levels. Fig. 2 Multivariable-adjusted associations between ln-transformed serum metal concentrations and IVF/ICSI outcomes ( n  = 396). Effect estimates are presented per interquartile range (IQR) increase in serum metal levels. Ten metals (Mg, Ca, Se, Fe, Sr, Ni, Pb, Cd, Sn, and Mo) were quantified in serum collected on menstrual cycle days 2–4. Models were adjusted for maternal age, BMI, infertility type, smoke exposure, AMH, basal FSH, AFC, total Gn dose, and PCOS status. Negative binomial regression was used for count outcomes (total oocytes, transferable embryos, and good-quality embryos); fractional logistic regression for proportion outcomes (fertilization rate and good embryo rate); and logistic regression for binary pregnancy outcomes. The dashed vertical line represents the null effect (RR or OR = 1.0). Points represent adjusted effect estimates (Relative Risks [RRs] or Odds Ratios [ORs]), and horizontal bars represent 95% confidence intervals (CIs). Colors indicate the direction of association for nominally significant results (orange = higher outcome, green = lower outcome); gray indicates non-significance. Asterisks (*) denote associations that remained statistically significant after False Discovery Rate (FDR) correction (P FDR < 0.05); plus signs (+) denote nominal significance ( P  < 0.05) Multivariable-adjusted associations between ln-transformed serum metal concentrations and IVF/ICSI outcomes ( n  = 396). Effect estimates are presented per interquartile range (IQR) increase in serum metal levels. Ten metals (Mg, Ca, Se, Fe, Sr, Ni, Pb, Cd, Sn, and Mo) were quantified in serum collected on menstrual cycle days 2–4. Models were adjusted for maternal age, BMI, infertility type, smoke exposure, AMH, basal FSH, AFC, total Gn dose, and PCOS status. Negative binomial regression was used for count outcomes (total oocytes, transferable embryos, and good-quality embryos); fractional logistic regression for proportion outcomes (fertilization rate and good embryo rate); and logistic regression for binary pregnancy outcomes. The dashed vertical line represents the null effect (RR or OR = 1.0). Points represent adjusted effect estimates (Relative Risks [RRs] or Odds Ratios [ORs]), and horizontal bars represent 95% confidence intervals (CIs). Colors indicate the direction of association for nominally significant results (orange = higher outcome, green = lower outcome); gray indicates non-significance. Asterisks (*) denote associations that remained statistically significant after False Discovery Rate (FDR) correction (P FDR < 0.05); plus signs (+) denote nominal significance ( P  < 0.05) After applying the FDR procedure to account for multiple testing, the associations for magnesium (Mg), calcium (Ca), selenium (Se), and lead (Pb) remained statistically significant ( P-fdr  < 0.05). Specifically, higher serum concentrations of Mg (OR = 1.10, 95% CI: 1.04–1.15, P-fdr  = 0.021) and Ca (OR = 1.14, 95% CI: 1.05–1.22, P-fdr  = 0.021) were robustly associated with an increased good embryo rate. In contrast, higher Pb levels were significantly associated with a reduced good embryo rate (OR = 0.88, 95% CI: 0.82–0.95, P-fdr  = 0.038). For the laboratory, Se showed a robust inverse association with the total oocyte count (RR = 0.91, 95% CI: 0.86–0.96, P-fdr  = 0.021). Several other metals exhibited nominal associations ( P  < 0.05, marked with ‘+’ in Fig.  2 ), although these did not survive stringent FDR correction. Higher serum nickel (Ni) was nominally associated with a lower total oocyte yield (RR = 0.94, 95% CI: 0.89–0.99, P  = 0.030) and a reduced number of transferable embryos (RR = 0.88, 95% CI: 0.81–0.96, P  = 0.004, P-fdr  = 0.057). Similarly, tin (Sn) and cadmium (Cd) showed nominal inverse associations with oocyte and embryo counts (Sn vs. total oocytes: RR = 0.93, P  = 0.014; Cd vs. transferable embryos: RR = 0.91, P  = 0.012). Regarding clinical outcomes, strontium (Sr) was nominally associated with an increased likelihood of clinical pregnancy (OR = 1.17, 95% CI: 1.01–1.38, P  = 0.047). Furthermore, considering that essential elements often exhibit complex dose-response dynamics, we utilized Restricted Cubic Splines (RCS) to evaluate potential non-linear relationships for magnesium, calcium, and selenium, which showed the strongest primary associations with embryological outcomes (Table S4; Figure S2). A significant non-linear threshold effect was confirmed for serum calcium in relation to the good embryo rate ( P overall  < 0.001, P non−linear  = 0.005). The adjusted predicted probability of producing good-quality embryos increased with calcium concentrations but reached a plateau at approximately 136.05 ng/mL (ln-Ca ≈ 4.90), after which no further clinical benefit was observed. In contrast, for magnesium and selenium, the RCS models suggested that the linear fit remained appropriate ( P non−linear  = 0.125 and 0.262, respectively). Specifically, magnesium exhibited a monotonic positive trend with embryo quality, while selenium showed a steady linear decline in oocyte yield across the observed physiological range. The superior fit of the non-linear model for calcium was further supported by a reduction in AIC (ΔAIC = 6.23) compared to the linear model. Multivariable-adjusted stratified analyses, accounting for AMH, basal FSH, total Gn dose, and PCOS status, confirmed that maternal dietary patterns significantly reshaped the relationship between serum metals and early IVF endpoints (Fig.  3 ; Table S6). Across all models, 13 metal–outcome pairs exhibited nominal interaction effects ( P interaction  < 0.05), although these did not remain statistically significant after conservative FDR correction. Fig. 3 Modifying effects of maternal dietary patterns on the associations between serum trace metal concentrations and IVF/ICSI outcomes. Panels ( a – m ) represent associations where the multiplicative interaction term reached nominal significance ( P-interaction  < 0.05). Red circles represent the high-adherence (High) group, and blue triangles represent the low-adherence (Low) group. All models were adjusted for maternal age, BMI, smoking status, passive smoke exposure, infertility type, AMH, basal FSH, AFC, total Gn dose, and PCOS status. The horizontal dashed line marks the null effect (Relative Risk [RR] = 1.0 for count outcomes and Odds Ratio [OR] = 1.0 for proportion outcomes). Note: None of the interaction associations remained significant after FDR correction; these findings should therefore be interpreted as exploratory evidence of potential effect modification Modifying effects of maternal dietary patterns on the associations between serum trace metal concentrations and IVF/ICSI outcomes. Panels ( a – m ) represent associations where the multiplicative interaction term reached nominal significance ( P-interaction  < 0.05). Red circles represent the high-adherence (High) group, and blue triangles represent the low-adherence (Low) group. All models were adjusted for maternal age, BMI, smoking status, passive smoke exposure, infertility type, AMH, basal FSH, AFC, total Gn dose, and PCOS status. The horizontal dashed line marks the null effect (Relative Risk [RR] = 1.0 for count outcomes and Odds Ratio [OR] = 1.0 for proportion outcomes). Note: None of the interaction associations remained significant after FDR correction; these findings should therefore be interpreted as exploratory evidence of potential effect modification Under the Alcohol–Processed Food pattern, high adherence consistently unmasked or accentuated the adverse impacts of several metals. For oocyte yield, significant inverse associations appeared exclusively among high adherents for molybdenum (RR = 0.775, 95% CI: 0.632–0.950, P int = 0.008), tin (RR = 0.821, 95% CI: 0.730–0.922, Pint  = 0.019), and calcium (RR = 0.702, 95% CI: 0.528–0.932, Pint  = 0.008), while associations in low adherents were null. Similarly, high adherence to this pattern amplified the detrimental effect of cadmium on the good embryo rate (RR = 0.780, 95% CI: 0.623–0.973). Within the Traditional Balanced Diet, a significant interaction was observed for selenium ( Pint  = 0.018); specifically, higher selenium exposure was associated with a more pronounced reduction in oocyte yield among high adherents (RR = 0.386) compared to low adherents (RR = 0.834). For the Protein-Rich/Modern Nutritive pattern, high adherence exacerbated the reproductive toxicity of nickel and cadmium. Among women with high adherence, higher circulating nickel was associated with a substantial loss of both oocytes (RR = 0.711, 95% CI: 0.592–0.853, Pint  = 0.002) and a lower good embryo rate (OR = 0.646). Similarly, the adverse associations for cadmium with transferable embryos (RR = 0.758) and good embryo rate (OR = 0.758, Pint  = 0.008) were concentrated in the high-intake stratum. In the Caffeinated-Beverage pattern, we observed a distinctive reversal of effects for essential minerals. Among low adherents, higher magnesium was associated with a decreased good embryo rate (OR = 0.704) and fewer transferable embryos (RR = 0.517). However, among high caffeine consumers, magnesium was robustly associated with improved outcomes, showing a five-fold increase in the odds of producing good-quality embryos (OR = 5.097, 95% CI: 1.911–13.708, Pint  < 0.001). A similar beneficial interaction was observed for calcium ( Pint  = 0.012), where its positive association with the good embryo rate was significantly stronger in high adherents (OR = 1.711) than in low adherents (OR = 0.851). Overall, while these interaction effects are considered exploratory due to the conservative multiple-testing adjustment, the consistent patterns suggest that dietary contexts significantly modulate environmental reproductive risks. High-protein and caffeinated patterns appeared to either amplify the toxicity of heavy metals or reshape the bioavailability and response to essential elements, independent of the patient’s baseline ovarian reserve and stimulation intensity.

This study provides evidence that circulating concentrations of essential elements like calcium and magnesium, as well as toxic metals such as lead and selenium, are independent predictors of early IVF success. Our application of non-linear modeling further clarifies that elements like calcium exhibit a threshold effect, where reproductive benefits are most pronounced within a specific physiological window. Crucially, we found that habitual maternal diet functions as a significant modifier of these environmental risks. Plant-forward and balanced eating patterns appear to support the beneficial effects of essential minerals, whereas diets high in processed foods or caffeine may exacerbate the detrimental impacts of heavy metals. These findings suggest that preconceptional dietary optimization, ideally starting at least three months prior to stimulation, may improve individual resilience against environmental metal exposure. Future large-scale, multicenter studies are warranted to confirm these modifying effects and to inform clinical guidelines for nutritional intervention in assisted reproduction.

We evaluated whether circulating concentrations of ten serum metals were associated with early IVF/ICSI outcomes and the modifying role of maternal dietary patterns. In this cohort, accounting for multiple testing via the FDR procedure and adjusting for clinical indicators including AMH and stimulation intensity, we identified outcome-specific associations for several metals, particularly calcium, magnesium, lead, and selenium. Additionally, we observed effect modification by habitual diet: balanced and protein-rich patterns appeared to buffer, whereas alcohol-processed and caffeinated patterns tended to accentuate the associations between metal exposures and reproductive endpoints. The relationship between a Traditional Balanced Diet and oocyte yield, while showing a positive trend, was modest after comprehensive adjustment for clinical predictors such as AMH. Nevertheless, diets rich in vegetables, fruits, and whole grains provide one-carbon donors and antioxidants that contribute to granulosa cell integrity and mitochondrial function [ 21 , 23 , 24 ]. Prior reports also emphasized that overall dietary adequacy plays a critical role in early embryogenesis [ 25 ]. Our findings suggest that while such a balanced intake is a practical strategy for preconceptional optimization, its independent contribution to oocyte numbers must be interpreted with caution once clinical confounders are fully controlled. By contrast, the Caffeinated-Beverage pattern (dominated by coffee and tea according to our factor loadings) was significantly associated with impaired embryo development. Women with the highest adherence (Q4) exhibited a significant reduction in the good embryo rate. This may reflect the contribution of sugar-sweetened caffeinated drinks, although our data did not distinguish between sweetened and unsweetened sources. Previous prospective IVF research linked sugared sodas with reduced oocyte yield and embryo quality [ 26 ]. While meta-analyses found no definitive link between caffeine alone and pregnancy success, ultra-processed or high-sugar profiles are consistently linked to impaired development through oxidative stress and endocrine disruption [ 27 ]. The nominal association between the Alcohol–Processed Food pattern (where spirits and red wine are loaded exclusively) and biochemical pregnancy should be viewed as exploratory. Alcohol consumption ≥ 84 g/week is generally associated with reduced pregnancy rates [ 27 , 28 ]. Given that ultra-processed dietary profiles are typically linked to reduced live birth, this observed signal likely represents a non-linear or threshold effect rather than a true clinical benefit [ 25 ]. The Protein-Rich/Modern Nutritive pattern, which captured a mix of animal proteins (milk, eggs) and plant proteins (soy products), was associated with lower success rates. Specifically, higher adherence was linked to a decreased likelihood of live birth. While protein is generally beneficial [ 29 , 30 ]., composite scores may capture diverse protein qualities or preparation methods. This finding may reflect a threshold effect where excessive intake of specific protein sources induces metabolic states that amplify reproductive risk. This unexpected result underscores the need for granular data on protein quality in fertility research. We found that dietary associations with IVF/ICSI outcomes were endpoint-specific and not consistently linear. While certain patterns appeared to influence embryo development and live birth independently of ovarian reserve, these associations reached nominal significance but did not survive conservative FDR correction, and should thus be interpreted as hypothesis-generating. Diets labeled by dominant nutrients often include diverse components with opposing effects. Further studies capturing specific food sources are needed to evaluate the relevance of these patterns to clinical counseling. Multivariable models, adjusted for clinical confounders and multiple testing, revealed stage-specific associations between serum metals and IVF/ICSI outcomes. Notably, after FDR correction, higher serum Mg and Ca remained robustly associated with an increased good embryo rate, while Pb was confirmed as a significant risk factor for poorer embryo quality. In the laboratory, Se exhibited a robust inverse association with total oocyte yield. The roles of Ca [ 31 ]and Mg [ 12 ]in oocyte activation and energy metabolism are well-supported. Our Restricted Cubic Spline (RCS) analysis further refined these findings, identifying a significant non-linear threshold effect for Ca, where the beneficial impact on embryo quality reached a plateau at approximately 136.05 ng/mL. This suggests a specific optimal physiological window for Ca rather than a simple linear benefit. While Mg showed a robust overall association with embryo quality, its non-linear component was less pronounced, suggesting a more monotonic positive trend within the observed range. These robust findings underscore the biological importance of essential element homeostasis in early embryogenesis, independent of maternal ovarian reserve. Pb was negatively associated with embryo quality, a finding that remained robust after FDR correction. This aligns with evidence from follicular fluid studies and likely reflects Pb’s interference with oxidative balance and calcium-dependent signaling pathways [ 15 , 32 ]. These reproducible adverse effects support the clinical relevance of Pb as a reproductive toxicant. In contrast, several other toxic metals, including Ni, Sn, and Cd, showed nominal inverse associations with oocyte or embryo counts, although these did not survive stringent statistical correction. Experimental models suggest Ni can impair ovarian function via inflammation and mitochondrial disruption [ 33 ], and our findings for Sn and Mo echo results from smaller IVF cohorts [ 16 , 34 ]. Se exhibited a robust inverse association with oocyte yield. Given its narrow physiological range, our RCS analysis confirmed that this relationship was primarily linear, suggesting that within the observed concentration range, higher systemic Se consistently correlates with reduced follicle recruitment. This may reflect the pro-oxidative phase of the Se dose-response curve, where excessive levels impair rather than support follicular development [ 17 , 35 ]. Interestingly, the discrepancy between systemic levels and outcomes underscores the biological compartmentalization between systemic circulation and the follicular microenvironment. The blood-follicle barrier, maintained by granulosa cells, regulates the transport of divalent cations via specific channels, which may lead to transport saturation or homeostatic disruption when systemic levels are excessive. Serum Sr was nominally associated with a higher chance of clinical pregnancy, although it showed no clear association with upstream laboratory metrics. Sr²⁺ is capable of mimicking Ca²⁺ in oocyte activation and has been used experimentally in rescue protocols [ 36 ]. However, without parallel evidence at the fertilization or embryo level, this association remains exploratory. Together, these results highlight distinct and stage-specific associations. The consistency of adverse associations for Pb and the beneficial threshold effects for Ca provides a more precise framework for evaluating environmental impacts on assisted reproduction by accounting for both clinical confounding and non-linear biological dynamics. Building on the observed associations, we evaluated whether maternal dietary patterns modified the relationships between serum metals and IVF/ICSI outcomes. After adjusting for clinical indicators, including AMH and total Gn dose, we identified 13 interaction pairs that achieved nominal significance, suggesting that diet remains a key modifier of environmental reproductive toxicity independent of a patient’s clinical baseline. Among women with high adherence to the Alcohol–Processed Food pattern, higher circulating Mo and Sn were associated with a significant reduction in transferable embryos and oocyte yield, whereas no such associations appeared in low adherents. Diet likely modulates this toxicity; ultra-processed foods and alcohol promote a pro-oxidative and pro-inflammatory milieu [ 37 , 38 ], which may lower the cellular defense threshold against metal-induced damage. Experimental and epidemiologic data support this, showing that Mo can impair oocyte quality via dose-dependent regulation of ovarian oxidative stress [ 16 , 39 ]. Our findings suggest that high spirits and processed food intake may exacerbate the adverse expression of these trace metals during early embryo development. Within the Traditional Balanced pattern, the only remaining significant interaction after adjusting for ovarian reserve was for Se. Interestingly, the inverse association between Se and oocyte yield was more pronounced among high adherents. This indicates that while plant-forward, nutrient-rich diets are generally beneficial [ 23 , 40 ], they may not fully counteract the reproductive toxicity of high Se concentrations when they reach the pro-oxidative phase. This underscores the need for broader environmental risk mitigation beyond nutritional optimization alone. In the Protein-Rich/Modern Nutritive pattern, high adherence significantly exacerbated the toxicity of Ni and Cd. Among women in the high-intake stratum, elevated Ni was linked to a substantial loss of oocytes and fewer good-quality embryos. Dietary patterns high in protein can influence metabolic states depending on source; while moderate protein enhances insulin sensitivity, excessive intake—often associated with higher fat in modern patterns—may amplify oxidative stress [ 41 , 42 ]. In our cohort, this pattern included both animal and plant proteins, yet it still appeared to sensitize the follicular environment to the detrimental effects of Cd and Ni. Mechanistic studies suggest these metals interfere with calcium signaling and redox regulation during gametogenesis [ 43 , 44 ], pathways that may be further compromised by the metabolic load of this specific dietary profile. In the Caffeinated-Beverage pattern, dominated by tea and coffee, we observed a striking reversal of effects for essential minerals. Among high caffeine consumers, higher serum Mg and Ca were associated with a robust increase in the good embryo rate, whereas these associations were null or inverse in low adherents. One possible explanation involves the competitive bioavailability of these ions. Caffeine is known to increase the urinary excretion of divalent cations [ 45 , 46 ], potentially creating a relative state of depletion. In this context, higher systemic levels of Mg and Ca may serve as a critical buffer, restoring the ionic balance required for oocyte maturation. Additionally, coffee and tea contain polyphenols with potent antioxidant properties [ 47 , 48 ]that may synergistically enhance the reproductive benefits of essential minerals. Overall, while these interaction associations did not survive conservative FDR correction and should be viewed as exploratory, the consistency in the direction of effects suggests that habitual diet shapes the oxidative and metabolic environment in which metals interact with the reproductive system. Our findings underscore the importance of dietary counseling beginning at least three months prior to ovarian stimulation—a window that aligns with the 90-day cycle of folliculogenesis. Emphasizing balanced, plant-forward eating and reducing alcohol-rich or highly processed products may enhance a woman’s resilience to environmental metal exposures and improve IVF/ICSI success rates. Our study provides a unique perspective by integrating maternal dietary patterns with the analysis of ten essential and toxic metals in a prospective IVF setting. A primary advantage is the rigorous control of clinical confounders. By incorporating anti-Müllerian hormone, basal FSH, and total gonadotropin dose into our models, we ensured that the observed environmental and nutritional effects were independent of the patients’ underlying ovarian reserve and treatment intensity. Furthermore, the three-month assessment window was chosen to align with the approximately 90-day cycle of human folliculogenesis, capturing the nutritional environment during the critical period of oocyte recruitment and maturation. Regarding generalizability, the specific characteristics of our cohort deserve careful consideration. The study population consisted of relatively young women (mean age 32.3 years) undergoing their first IVF/ICSI attempt. This naïve population allowed for a clearer assessment of periconceptional exposures without the interference of previous ovarian stimulation or the cumulative psychological stress associated with multiple failed cycles. However, these findings may not fully extrapolate to older patients, who typically exhibit a decline in oocyte repair mechanisms and antioxidant defenses. It is possible that the adverse impacts of toxic metals like lead and nickel are even more pronounced in older women, whose physiological resilience to oxidative stress is diminished. Similarly, for patients with a history of recurrent IVF failure, environmental exposures might interact differently with pre-existing reproductive pathologies. While our study focused on ART-specific endpoints, the fundamental biological risks identified—such as the impact of lead on early embryogenesis—likely represent universal reproductive concerns applicable to women seeking natural conception, although the magnitude of these effects may vary across different physiological contexts. Several other limitations should be noted. A single serum measurement might not capture long-term variability in metal exposure, potentially leading to some misclassification. While the food frequency questionnaire reflected habitual intake, it lacked granular detail on nutrient composition and specific preparation methods. We also recognize that the observational design precludes definitive causal inferences. Furthermore, while we utilized False Discovery Rate correction to address multiplicity, the conservative nature of this procedure meant that interaction results remained exploratory. The statistical power was also lower for distal outcomes like live birth compared to proximal laboratory markers. Finally, our measurements reflect systemic serum levels rather than direct intrafollicular exposure. While serum serves as a reliable proxy, the blood-follicle barrier maintains a distinct local environment.

Infertility poses a substantial global health burden, affecting approximately 1 in 6 couples of reproductive age [ 1 , 2 ]. Even with modern in vitro fertilisation/intracytoplasmic sperm injection (IVF/ICSI), a single fresh cycle yields a live birth in only about 30–40% of attempts [ 3 ]. This efficacy gap underscores the urgent need to identify and address modifiable preconceptional factors that may influence oocyte quality, fertilization, embryo development, and implantation. Among these factors, environmental exposure to trace and heavy metals is of growing concern in reproductive epidemiology. Trace and heavy metals are pervasive in food, water, and air, and they are readily measurable in serum and follicular fluid collected during IVF procedures [ 4 ]. Toxic metals such as cadmium (Cd), lead (Pb), and mercury (Hg) are ubiquitous environmental contaminants with established systemic toxicity. In the reproductive system, they can act as potent endocrine disruptors and inducers of oxidative stress [ 5 , 6 ]. Pb has been shown to disrupt the hypothalamic-pituitary-gonadal axis, altering gonadotropin-releasing hormone (GnRH) pulsatility and leading to impaired follicular development [ 7 , 8 ]. Cd can accumulate in follicular fluid, and experimental studies link Cd exposure to increased oocyte apoptosis and meiotic spindle defects, thereby compromising embryo quality [ 9 ]. Conversely, essential trace elements are critical for reproductive function. Selenium (Se), a key component of the antioxidant enzyme glutathione peroxidase (GPx), is vital for protecting the developing oocyte from reactive oxygen species (ROS) [ 10 ]. Similarly, zinc (Zn) is indispensable for oocyte maturation and meiotic division [ 11 ], while calcium (Ca) and magnesium (Mg) signaling pathways are fundamental to oocyte activation and fertilization [ 12 , 13 ]. Despite plausible biological mechanisms, epidemiologic findings regarding metal–fertility associations remain inconclusive and often contradictory. A recent prospective cohort study reported a significant inverse relationship between blood Pb levels and IVF outcomes [ 14 ]. However, another prospective IVF study reported inconsistent associations after multivariable adjustment [ 15 ]. Similar inconsistencies exist for essential elements; while some studies link higher Se status to favorable early embryo development, others report null or even adverse associations in IVF settings [ 16 , 17 ]. Diet is a primary pathway for metal exposure but also a key determinant of the body’s physiological response to that exposure. On one hand, dietary choices can significantly increase the body burden of toxic metals. Consumption of large predatory fish is a major source of mercury, while shellfish and organ meats can be high in cadmium [ 18 , 19 ]. On the other hand, diet provides essential nutrients that can profoundly modulate the absorption, transport, and detoxification of metals. It is well-established that dietary calcium and iron competitively inhibit gastrointestinal Pb uptake through shared transporters like divalent metal transporter 1 (DMT1) [ 20 ]. Furthermore, dietary patterns rich in antioxidants—such as the Mediterranean diet, characterized by high intake of fruits, vegetables, and healthy fats—may directly counteract the oxidative damage induced by toxic metals in the follicular microenvironment, a mechanism hypothesized to underlie its observed benefits on IVF success [ 21 , 22 ]. Nevertheless, most studies on metal–fertility relationships have not accounted for these crucial dietary influences, and few have assessed how comprehensive dietary patterns modify these associations in IVF populations. This represents a significant gap in our understanding of how combined environmental and lifestyle factors impact ART outcomes. To address these gaps, we conducted a prospective cohort study among women undergoing their first IVF/ICSI cycle at a tertiary fertility center. We assessed serum concentrations of ten trace and heavy metals, including both toxic and essential elements, alongside maternal dietary patterns. We hypothesized that dietary factors may modify the associations between metal exposure and IVF/ICSI outcomes. By jointly evaluating environmental and nutritional influences, this study aims to identify modifiable factors that could improve reproductive success in assisted conception.

Supplementary Material 1. Supplementary Material 1. Supplementary Material 2. Supplementary Material 2.