Participant characteristics
The current study began by evaluating the demographic data (Table 1), including information on assisted reproductive technology (ART). Interestingly, in contrast to BMI, the ages of the case and control subjects were significantly different (P < 0.001). Considering that some of the ART information had missing values, it was used here only as a reference and was not included in the subsequent statistical analysis.
Metal Element Analysis
The detection rates of the listed metals were all 100% in both groups. As shown in Figure S1, the difference between groups was significant for all the elements except for molybdenum (Mo), titanium (Ti) and ferrum (Fe), indicating that the metallic element may be a potential influencer of PCOS. Of these, only Mg levels were significantly lower in the PCOS group than in the control group (median 35.81 and 39.54 mg/L in PCOS and control groups, respectively), while the levels of the remaining element were higher in the PCOS group (P < 0.001).
As shown in Table 1, age was an influencing factor on PCOS, so it was included in a logistic regression model to assess the association between metal elements and PCOS, with adjustments for covariates. Regression analysis (Fig. 1 & Table S2) revealed that the levels of Cu, Se, Sr, Zn, Ca, and Mg in follicular fluid were significantly associated with the risk of PCOS relative to the first quartile. Notably, three minerals, As, V and Cr, were significantly and positively associated with an increased risk of PCOS at high levels (fourth quartile), with AORs of 2.77 (95% CI 1.59 to 4.82), 4.47 (95% CI 2.44 to 8.29) and 3.37 (95% CI 1.91 to 5.94), respectively. Additionally, Fig. 1 indicates that most metals had a high probability of being risk factors for PCOS (P-trend < 0.001), and only Mg was a protective factor (P-trend < 0.001).
In the PCA, for all twelve elements, the first component explained 33.8% of the total variance, and the second component explained 15.3% (Fig. 2). All variables, except Mg and Ca, were loaded in the same direction on the first component, with loadings ranging from 0.01 to 0.41. The main loadings of the first component came from Se (0.41), V (0.39), As (0.39) and Cu (0.39), with moderate loadings coming from Zn (0.35). The main loadings of the second component come from Mg (0.61), with medium loadings coming from Ti (0.43) and Fe (0.42). However, the adjusted linear regression analysis showed that Ti and Fe were not associated with PCOS.
The correlations between the metal elements are shown in Table S3. According to the rule of thumb for the strength of the correlation coefficient (Schober et al. 2018), different levels of correlation are indicated by different color shades, wherein a value of 0.4 is considered to represent a weak effect. Moreover, the results of the cluster analysis showed that the ten elements were clearly divided into two categories, with the first containing Cu, V, As, Cr, Sr, Se and Zn and the second containing Ca, Ti, Mg and Fe (Figure S1).
Interestingly, after weak correlations (correlation coefficients less than 0.4) were eliminated, similar clustering by elements was also observed in the network analysis, which was aligned with the PCA results (Fig. 3A). Additionally, based on the average weighted degree of nodes, Cr and As were the two most important elements in the network (Fig. 3B).
Hormone Analysis
Regarding hormones, between-group analysis of performance differences showed significantly higher increases in LH, T, and AMH in PCOS patients than in controls, while the opposite was true for FSH (Table S5). Based on the above analysis, we also performed a univariate logistic regression analysis to explore hormones related to PCOS, and the odds ratio (OR) estimates were adjusted for potential confounders (age). Furthermore, the results of the linear-by-linear association chi-square test support a linear trend for the prevalence of PCOS as the hormone changes (i.e., FSH, LH, T, and AMH) (Fig. 4.). Among them, FSH was significantly negatively associated with PCOS risk, and the AORs were 0.58 and 0.36 for the third and fourth quartiles, respectively, with 95% CIs of 0.34 to 0.99 and 0.20 to 0.65 (Fig. 4 & Table S5). In contrast, LH, T and AMH were significantly higher in the fourth quartiles than in the first quartile (Table S5). Overall, FSH was regarded as a protective factor for PCOS, while LH, T and AMH were risk factors (P-trend < 0.001).
Association Between Elements And Hormones
Based on the dose-effect relationship between hormones and PCOS, OPLS models were constructed to investigate the metal elements that influence hormone levels. The models were validated by randomized permutation tests (Figure S2), and finally, two valid models were obtained for FSH and AMH, as shown in Fig. 3.
In regard to FSH, five elements (Mg, Zn, Fe, Se and V) with VIP values greater than 1 could have more significant effects on FSH. Among them, Mg was the most influential element, with which it was negatively correlated. On the other hand, age was negatively correlated with AMH, while As, Se, V, Cu and Zn were positively correlated. Among the five elements, Cu had a relatively large VIP value (1.52) and was the most susceptible to fluctuations in AMH levels. However, the effect of Mg on AMH was negligible in comparison to FSH. Overall, the elements that had significant effects on both hormones were Zn, Se and V, but the effects were not as strong as those of Mg and Cu.
Mediation analysis was used to further reveal the underlying mechanisms of both hormones and elements. The results showed (Table S6) that there was little mediating effect between FSH and its associated elements. Fe was the exception; nevertheless, previous analyses showed no significant association between this element and PCOS. In contrast, AMH had a significant mediating effect between the five related elements and PCOS. Notably, AMH, as a mediating variable, explained 63% of the association between V and PCOS (95% CI 0.46 to 0.83, P < 0.001).
To explore the hypothesis that FSH and exogenous metals affect PCOS through combined actions, FSH was grouped with four elements (risk factors Zn, Se, and V and protective factor Mg) according to their respective median values. Regression analysis was performed in the element-FSH two-dimensional space, using the low Mg-FSH concentration level (Lo-Lo) as a reference, adjusted for age (Figure S3). The results showed that the risk of suffering from PCOS was significantly lower at higher levels of both Mg and FSH (AOR = 0.30, 95% CI 0.16 to 0.56). For the three risk elements, low levels of elements and high levels of FSH (Lo-Hi) were considered references. High levels of FSH do have a protective effect in PCOS; however, the level of exogenous metals has a greater impact on PCOS. As an example, the risk of developing PCOS increased stepwise with V-FSH levels of Lo-Lo, Hi-Hi and Hi-Lo compared to the reference group, and a positive trend was observed (P-trend < 0.001) (Figure S4).
Changes in hormone levels between multielement coexposure and PCOS development suggest that female exposure to As, Se, V, Cu, and Zn may affect PCOS progression by perturbing AMH levels. Both FSH and Mg at low concentrations tend to induce PCOS; while higher levels are protective against PCOS (Fig. 6).