Demographics and exposure distributions
Characteristics of participants in the original cohort of 997 women were previously described in detail [70]. Mothers included in our analysis had similar demographic characteristics to the overall ELEMENT population; the mean age of mothers at the time of enrollment was 26.6 (standard deviation=5.3). Mothers had on average 11 years of education, most were married or cohabitating (89%), and all lived within Mexico City. Very few (3%) reported smoking during pregnancy. Average age of the boys at the early-teen and late-teen visits were 10.4 and 13.7, respectively. Mean and standard deviation of the follow up period were 3.5 and 0.5 years. The percent of samples with metal concentrations below the limit of detection, as well as the geometric means, standard deviations, and selected percentiles of sample concentrations from prenatal and peripubertal (early-teen) visits are shown in Table 1. Spearman correlations between the prenatal and peripubertal visit metal concentrations adjusted for SG are also presented in the table. With the exception of Sb and Fe, other metals were detected in > 50% of urine samples, therefore, Sb and Fe were excluded from further analysis. Weak correlations between maternal and peripubertal metal concentrations and weak to moderate correlations between different urinary metal concentrations within maternal and child samples have been previously reported elsewhere [76].
Hormone and Tanner stages distributions
Distributions of reproductive hormones among ELEMENT boys were described previously [89]. Except for 12 total testosterone measurements, all measures were above the LOD. Spearman correlations among hormones were weak to moderate (R=-0.42 to 0.63). Distributions of Tanner stages of sexual maturation and testicular volume among the male children at the two follow up visits are reported (Supplementary Table S1). We additionally provided spaghetti plots depicting Tanner stage and testicular volume progression between two visits (Supplementary Figure S1), and the distribution of measures of sexual maturation for different age groups of ELEMENT boys (Supplementary Table S2). At the early -teen visit, the majority of the boys (n=94, 81.7%) were at Tanner stage 1 for pubic hair development whereas 57 boys (49.6%) were at Tanner stage 1 for genital development. Most boys who were at Tanner stage 1 moved to more advanced Tanner stages at the late-teen visit- only 28 (26.4%) and 8 (7.5%) were still at Tanner stage 1 for pubic hair development and genital development after 3 years on average since the early-teen visits. Boys who were at Tanner stages 2, 3, and 4 at the early-teen visit all progressed to higher stages at the late-teen visit, with 14 (13.2%) and 18 (13.2%) boys reaching full development (Tanner stage=5) for the two measurements. In terms of testicular volume distribution, the percentage of the boys in the prepubertal stage dropped from 14.8% to 0% from early-teen to late-teen visit.
In utero and peripubertal metal exposure and peripubertal hormone concentrations
Associations between in utero and peripubertal metal concentrations and reproductive hormones are presented in Figure 2 and Supplementary Table S4. Positive associations were observed between some urinary essential metal concentrations and estradiol, testosterone, and SHBG. One IQR increase in in utero Zn concentration was associated with 13.7 % higher serum estradiol (95% CI: 0.3, 28.8). In utero Co and Mn were positively associated with SHBG concentrations, with an IQR increase associated with 16% (95% CI:0.4, 34.2) and 14.2 % (95%CI: 1.5, 28.5) higher serum SHBG after adjustment for child age, BMI z-score, and SG, respectively. As shown in Figure 2, effect estimates for the association between both in utero essential and non-essential metal concentrations and testosterone were larger compared to other reproductive hormones, 51.3% for Mo, 35% for As, 38.9% for Cd. In models where reproductive hormones were regressed on concurrent peripubertal exposures, the strongest associations were observed also between metal concentrations and testosterone, particularly with non-essential metals Ni (%△/IQR: 35.1%, 95% CI: 2.6, 77.8) and Ba (%△/IQR: 59.1%, 95% CI: 22.5, 106.8). Peripubertal Ba concentration was also associated with higher estradiol (%△/IQR: 10.2%, 95% CI: 0.5, 20.9). However, no significant associations were detected between peripubertal metal exposures and DHEA-S or inhibin B.
After correcting for multiple testing, the associations of in utero Mo, As, and Cd with testosterone, as well as the association of peripubertal Ba with testosterone had q-values < 0.15 (Figure 2 and Supplementary Table S4), providing greater confidence in these associations.
In utero and peripubertal metal exposure and sexual maturation
We have presented results from multiple ordinal regression models of in utero metal concentrations and Tanner stage and testicular volume in Figure 3 and Supplementary Table S5. Tanner stages and testicular volume at the early-teen visit were not associated with in utero metal concentrations, with the exception of a non-essential metal, Al; an IQR increase in in utero Al concentrations was associated with 3.6 times greater odds (95% CI: 1.67, 7.76) of being at a higher category of testicular volume versus lower categories. In the same figure and table, associations between in utero metal concentrations and pubertal development over time in boys are also presented. During the follow-up, an IQR increase in in utero concentrations of non-essential metalloid As was associated with 36% (OR/IQR: 0.64, 95% CI: 0.48, 0.85) lower odds of genital development progression per year, adjusting for age, BMI, and Tanner stage at the early-teen visit (GEE longitudinal model). In utero concentrations of non-essential metal(loid)s Al (OR/IQR: 0.61, 95% CI: 0.45, 0.83) and As (OR/IQR: 0.64, 95% CI: 0.43, 0.97) were associated with lower odds of progressing to a higher testicular volume category (i.e. 39 and 36% lower odds/IQR). Essential metal Zn was also associated with 38% lower odds of testicular volume progression (OR/IQR: 0.62, 95% CI: 0.44, 0.88).
Similarly, Figure 4 and Supplementary Table S5 show the cross-sectional and longitudinal associations between prepubertal metal concentrations and sexual maturation and progression. No significant associations were found between peripubertal urinary metal concentrations with Tanner stage for genital development or progression over follow-up, although several essential and non-essential metals were associated with pubic hair Tanner stages, testicular volume, and progressions. Peripubertal Zn concentration was associated with higher odds of being at a higher developmental stage for pubic hair (OR/IQR: 6.11, 95% CI: 1.89, 19.69) and testicular volume (OR/IQR: 5.39, 95% CI: 1.88, 15.49) at the early-teen visit, as well as slower progression of pubic hair development (OR/IQR: 0.47, 95% CI: 0.34, 0.67) and testicular volume (OR/IQR: 0.58, 95% CI: 0.40, 0.85) during the follow-up. Higher Mn was associated with 34% (OR/IQR: 0.66, 95% CI: 0.46, 0.96) lower pubic hair development progression only. Regarding non-essential metals, an IQR increase in peripubertal Ba concentration was associated with 2.3 times greater odds (95% CI: 1.15, 4.75) of being at higher Tanner stage for pubic hair development at age 8–14 years, but with 34% (OR/IQR: 0.66, 95% CI: 0.53, 0.83) lower odds of progressing to a higher pubic hair Tanner stage per year of follow-up. Higher peripubertal urinary Al was also associated with 47% (OR/IQR: 0.53, 95% CI: 0.40, 0.70) lower odds of pubic hair development progression per IQR increase.
All the associations described above had q-values < 0.15 (Figure 3, Figure 4, and Supplementary Table S5) after correcting for multiple testing, except for the association between in utero As and testicular volume progressions. Output graphics produced from GAM models showed that there was no significant non-linear relationship between essential and non-essential metal concentrations and hormones and sexual maturation outcomes, after adjusting for the same sets of covariates.
Sensitivity analysis
Results from the hormone subset analysis when we restricted to boys who were prepubertal (94 out of 118) at the early-teen visit are similar to the main analysis results (Supplementary Table S6). For models regressing in utero metal concentrations, some effect estimates for DHEA_S, SHBG, and testosterone were no longer significant, potentially due to the small sample size. For models regressing peripubertal exposure, the notable effect estimates for testosterone were similar to those from the main analysis. The association between DHEA_S and Ni (%△/IQR =20.2, 95%CI=1.7, 42.0) and inhibin B and Al (%△/IQR =-19.2, 95%CI=-31.7, -4.4) became stronger and significant in the peripubertal subset analysis.
The magnitude of estimates from GEE models with and without BMI (Supplementary Table S7) were almost identical. Findings from models adjusting for SES were generally consistent with metal and hormone associations observed in our main analyses; in SES-adjusted models, the association between in utero Co concentrations and higher SHBG (%△/IQR=14.2, 95%CI=-2.9, 23.3) was slightly attenuated and no longer significant, while the association between peripubertal Al and inhibin B (%△/IQR=-14.1, 95%CI=-24.6, -2.2) was stronger and significant. In GEE models for Tanner stage or testicular volume status including SES finding remain consistent with the main models.
A high proportion of samples had metal concentrations above the LOD except for Cu, which was below the detection limit in 46% of samples. In a secondary analysis, we categorized urinary Cu concentrations into three groups. The low group consisted of values below the LOD, while the medium and high groups were made up of equalized bins among the detected values. We estimated the model parameters again and found that they were similar to the main parameter estimates.