Exposure to multiple metals and adolescent major depressive disorder: a case-control study in China

Abstract

Background

Major depressive disorder (MDD) in adolescents is a common mental health problem with a complicated cause and increasing prevalence. Previous studies reported the impact of metals on adolescents with depressive symptoms, but limited evidence is available on the relationship between metals and adolescent MDD.

Methods

A total of 203 adolescents with MDD and 203 healthy controls were recruited for this case-control study. Urinary levels of 21 metals were measured using inductively coupled plasma mass spectrometry. The Least Absolute Shrinkage and Selection Operator (LASSO) regression was applied to identify metals for key metals, and conditional logistic regression and Bayesian Kernel Machine Regression (BKMR) were applied to assess individual and joint effects.

Results

In the multi-metal model, higher urinary copper (Cu) and iron (Fe) levels were associated with increased MDD risk, whereas nickel (Ni) and iodine (I) inversely associated. BKMR analysis confirmed positive associations for Cu and Fe and protective effects for Ni and I. The combined metal exposure showed an overall negative association with MDD risk. These associations were consistent across several subgroups.

Conclusions

This research was the first case-control study to explore the relationship between urinary metal levels and adolescent MDD. Our study identified urinary Cu and Fe as risk factors for adolescent MDD, whereas urinary Ni and I were potential protective factors. These findings might provide new light on the future prevention of adolescent MDD.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12888-026-08444-9.

Introduction

Major depressive disorder (MDD) is a leading contributor to the global burden of disease and the major cause of disability among adolescents and young adult [1, 2]. In the United States, the prevalence of adolescent MDD increased from 8.1% in 2009 to 15.8% in 2019 [3]. Approximately 30% of adolescents with MDD report experiencing suicidal ideation, with over 10% attempting suicide [4]. Furthermore, experiencing the first-episode MDD during childhood or adolescence is associated with a higher lifetime prevalence of mental disorders and poorer quality of life in adulthood [5]. Currently, the pathophysiology of adolescent MDD is still poorly understood [6, 7], while there are multiple neurobiological perturbations in MDD, including deficits in monoamine neurotransmitters [8], dysregulation of the hypothalamic-pituitary-adrenal axis [9], and alterations in excitatory and inhibitory neurotransmission [10], likely resulting from complex etiological, psychological, and environmental factors [11]. Identifying and mitigating pre-morbid risk factors is therefore essential for preventing adolescent MDD.

Recently, an increasing number of studies have focused on the impact of environmental factors on neuropsychiatric disorders, including environmental toxins [12], air quality [13, 14], residential green spaces [15], and metals [16]. Méndez-Armenta et al. reported that lead (Pb) affects neurogenesis and apoptosis [17], as well as oxidative stress and glutathione pathways in neuropsychiatric disorders [18]. Additionally, magnesium (Mg) plays a key role in membrane phospholipid synthesis and intracellular signaling, which is crucial for brain function and mood regulation [19, 20]. Previous studies have reported the impact of metals on mental disorders, such as MDD [21], anxiety disorders [22], bipolar disorder [23, 24], and schizophrenia [25]. However, the conclusions of previous studies are inconsistent. For example, it was reported that serum copper (Cu) levels were lower in MDD patients than in healthy controls (HCs) [26], while a case-control study indicated higher serum Cu in MDD patients [27]. A prospective study in China reported a negative correlation between serum iron (Fe) levels and MDD [28], whereas a cross-sectional study in Germany found no significant difference [29]. These inconsistencies may result from variations in exposure contexts, geographic locations, sampling types and time, and detection methods [30, 31]. Therefore, the association between exposure to metals and MDD requires further validation.

Notably, while several studies have explored the relationship between metals and MDD, the specific role of age in modulating this association remains insufficiently understood [32]. Accumulating evidence has demonstrated pathophysiological differences between adolescent and adult MDD, highlighting the need for age-specific studies in this field [33, 34]. Additionally, studies have found correlations between metal levels and age [35]. In a study using data from the National Health and Nutrition Examination Survey, Shan et al. reported that high urinary iodine (I) levels were associated with an increased risk of MDD in adults over 46 years, but not in younger adults [30]. This study further suggested that urine analysis, as a non-invasive method, may detect biochemical imbalances and reflect long-term metal exposure [36, 37]. Consequently, urinary metal levels could serve as potential biomarkers for adolescent MDD and warrant further investigation.

Furthermore, most prior research has focused on the impact of single metals on MDD [38]. However, Shen et al. have highlighted that concurrent, chronic, low-level exposure to multiple metals more accurately reflects real-world exposure scenarios [39]. While some studies have explored the associations between exposure to multiple metals and MDD using simple interactive models [40], these models fail to comprehensively estimate the overall effect of multi-metal exposure and do not adequately account for potential multicollinearity among two or more metals. Recently, Bayesian Kernel Machine Regression (BKMR) models have been recognized as effective for assessing the overall and interactive effects of multiple metal exposures [41], as well as studying the nonlinear relationship between multiple metals and diseases [31, 42]. Additionally, given that the concentrations of metals are typically at the microgram per gram (µg/g) level, the use of Inductively Coupled Plasma-mass Spectrometry (ICP-MS) is essential for their quantification [43]. ICP-MS is characterized by its high sensitivity and low detection limits, thus could enhance the accuracy of quantification and reduce analytical education [44].

In this study, we conducted the first case-control study using ICP-MS to explore the relationship between urinary metal levels and adolescent MDD. BKMR models were employed to analyze the interaction between them and the effect of mixed exposure (see Fig. 1). These findings might provide new insights into the early prevention of adolescent MDD.

Fig. 1.

Workflow for exploring of the association between metals and adolescent major depressive disorder. Urine samples from healthy controls and adolescents with MDD were analyzed for 21 metals using ICP-MS. Associations were assessed using single- and multiple-metal models, and mixture effects were further evaluated with Bayesian Kernel Machine Regression (BKMR), adjusting for relevant covariates

Methods

Study population

This case-control study was approved by the Medical Ethics Committee of the First Affiliated Hospital of Chongqing Medical University (2020 − 864). Written informed consent was obtained from all participants and their legal guardians. Adolescents with MDD were recruited from Department of Psychiatry, the First Affiliated Hospital of Chongqing Medical University between July 2021 and May 2022. Inclusion criteria included: (1) aged 12 to 18 years; (2) diagnosis of MDD according to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, based on a semi-structured interview with the Schedule for Affective Disorders and Schizophrenia for Children-Present and Lifetime version (K-SADS-PL). All interviews were conducted by licensed psychiatrists who were trained in child psychopathology and in the administration of the K-SADS-PL. (3) Participants had a baseline score of 8 or higher on the Hamilton Depression Scale-17 (HAMD-17) [45, 46]. Exclusion criteria for MDD included: (1) lifetime diagnosis of bipolar disorder, schizophrenia, autism spectrum disorder, attention deficit hyperactivity disorder, or other serious psychiatric disorders; (2) neurological diseases such as seizure, stroke, or head injury; (3) somatic diseases, including cancer, organ failure, or serious cardiovascular or cerebrovascular disease; (4) illicit drug use; (5) menstruation, pregnancy, or breastfeeding. During the same study period, matched HCs were recruited from the physical examination center, the First Affiliated Hospital of Chongqing Medical University. Selection criteria for HCs were: (1) aged 12–18 without any psychiatric or unstable medical diagnoses, and (2) a HAMD-17 score less than or equal to 7. Control participants were matched with MDD participants by age (± 2 years) and gender distribution.

Measurement of covariates

We conducted in-person interviews using structured questionnaires to collect data on socio-demographic characteristics and lifestyle factors, which were subsequently used as covariates. The socio-demographic variables included age (years), sex(girls or boys), body mass index (BMI [kg/m²]), residence (rural or urban), average monthly household income (< 3000 RMB, 3000–6000 RMB, 6000–12000 RMB, > 12000 RMB) [47] and family history of mental disorders (Yes/No). Lifestyle factors assessed included leisure time physical activity per week (< 1 h, 1–3 h, 3–5 h, 5–7 h) [48], daily sleep duration (< 8 h, ≥8 h) [49], and smoking or alcohol use over the past 12 months (Yes/No) [50, 51]. Detailed questionnaires are provided in the Supplementary Methods.

Urinary metals and creatinine measurement

First-morning void urine of adolescents was collected in 50 mL sterile tubes. Then all the samples were refrigerated at 4 ◦C, sent to our laboratory within 4 h, and stored at -80 ◦C for subsequent detection. Metal concentrations were measured using ICP-MS (Perkin Elmer NexION 300X, Shelton, CT, USA). Before testing, frozen urine samples were thawed at room temperature. An aliquot of 0.2 mL urine was placed into 10 mL polypropylene centrifuge tubes and diluted with 3.8 mL of 1% (v/v) nitric acid (HNO₃). After vortex mixing, the samples were digested via microwave heating at 90 °C for 1 h, then centrifuged at 3,000 rpm for 10 min at room temperature before instrumental analysis. An internal standard solution (10 µg/L) purchased from PerkinElmer was used for internal online calibration, and the standard curve regression coefficient was maintained above 0.999. The intra-assay and inter-assay coefficients of variation for urine metal levels were all below 10% and accuracy was checked using SeronormTM Trace Element Urine L-2 (Sero, Billingstad, Norway). The recovery rates ranged from 90.8% to 113.7%. The limit of detection (LODs) for these 21 metals ranges from 0.001 ug/L to 8.35 ug/L. Each sample was normalized to urine creatinine levels, which were measured via the sarcosine oxidase method on a Beckman Coulter DxC 800 automatic biochemistry analyzer. Detection rates for all metals exceeded 70%, as shown in Table A.1. Values below the LOD were substituted with LOD/√2, following previous guidelines [52].

Statistical analysis

Descriptive statistics were conducted for demographic and clinical characteristics. Continuous variables were expressed as means with standard deviations (SD), while categorical variables were presented as counts and percentages. Group differences were assessed using the nonparametric Wilcoxon test for continuous variables and the chi-squared (χ²) tests for categorical variables.

For metal levels in both MDD and HCs groups, we initially calculated the mean (SD) and median to describe the distribution of metals (Table A.2). Metal levels were standardized by creatinine (µg/g creatinine) and transformed using the natural logarithm. Conditional logistic regression analysis was employed to evaluate the association between each metal and the risk of adolescent MDD, calculating odds ratios (ORs) with 95% confidence intervals (95% CIs). Metal levels were categorized into quartiles based on the HCs, with the lowest quartile serving as the reference group. Both crude and adjusted models were fitted, with adjustments made for demographic characteristics as covariates. Linear trend p-values were determined by entering each metal concentration (using the lowest quartile as the reference) into the model. Additionally, to examine whether metal concentrations were associated with depression symptom severity, Pearson and partial correlation analyses were performed between each metal and HAMD-17.

To explore the impact of multiple metals on adolescent MDD risk, we created a multi-element model to reflect mixed exposures. Considering the variety and elements and correlation with each other, we used Least Absolute Shrinkage and Selection Operator (LASSO) regression to select representative elements into the multi-element model, for achieving dimensionality reduction [53]. The LASSO regression can regress and penalize all variables so that the coefficients of relatively unimportant independent variables become 0 and are excluded from the model [54]. Multicollinearity among the selected metals was assessed using variance inflation factors (VIF) and Spearman correlation coefficients. Sensitivity analyses were performed by modeling metals as continuous natural log-transformed variables.

To accurately capture the nonlinear effects of the metals on adolescent MDD, we utilized Bayesian Kernel-Machine Regression (BKMR) to elucidate both individual and interactive effects of the metals involved [41, 55]. An exposure matrix was constructed, comprising the four metals identified through LASSO regression, along with covariates for demographic characteristics. The BKMR model was fitted using the Markov Chain Monte Carlo (MCMC) algorithm via the kmbayes function, which allowed for variable selection and estimation of the posterior inclusion probability (PIP) for each metal. The model fitting used 50,000 MCMC iterations, and the trace plots of model parameters were visually examined to confirm model convergence. The relationship between each metal and adolescent MDD risk was explored while holding all other metal levels at their median values. Univariate exposure-response relationships for each metal are illustrated in cross-sectional plots.

To assess the cumulative effect of mixed-metal exposure, we compared scenarios where all metal exposures ranged from the 25th to 75th percentile (in 5-percentile increments) against those where all metals were at their median. Another approach to characterizing exposure effects involved examining the distribution of individual exposures on the cumulative effect, with other exposures held constant at specific quartiles (P25, P50, and P75) and comparing the difference between the P75 and P25 levels for each metal’s risk. Furthermore, we assessed interactions between multiple metals and adolescent MDD risk by visualizing bivariate exposures for metal pairs while keeping the remaining two metals fixed at their median levels.

The data was analyzed using SPSS software (version 26.0. IBM Corp., Armonk, NY, USA) and R software (version 4.2.2, R Foundation for Statistical Computing). The two-sided p-value < 0.05 was considered statistically significant.

Results

Participants characteristics and metal levels

Table 1 presented the demographic and clinical characteristics of the 406 participants. There were no statistically significant differences between adolescent MDD and HCs in terms of age, sex, BMI, residence, or monthly household income. However, adolescents with MDD exhibited significantly higher levels of depression and anxiety, and more frequent alcohol consumption or smoking, less exercise time, shorter sleep duration, and a higher prevalence of family history of mental disorders. Table A.2 displayed the distributions of 21 metal levels adjusted by urinary creatinine between the two groups. Concentrations of magnesium (Mg), nickel (Ni), arsenic (As), iodine (I), and cesium (Cs) were significantly higher in HCs than in adolescent MDD (all p < 0.05). Notably, the metal levels detected in this study were mostly consistent with those reported in the general population (Table A.2).

Table 1.

Demographic and Clinical Characteristics of Participants

Participants		MDD	HCs	P-value
		n = 203	n = 203
Age, year, mean (SD)		15.42 (0.11)	15.21 (0.14)	0.34
HAMD-17, mean (SD)		19.82 (0.42)	1.59 (0.13)	< 0.001
HAMA-14, mean (SD)		16.38 (0.45)	1.52 (0.12)	< 0.001
Urine creatinine, g/L, mean (SD)		1.53 (0.04)	1.68 (0.05)	0.277
Girls, n (%)		151 (74.38)	151 (74.38)	/
BMI, kg/m2, n (%)				0.409
<18.5		58 (28.57)	59 (29.06)
18.5–23.9		115 (56.65)	102 (50.25)
24.0-27.9		16 (7.88)	24 (11.82)
≥28.0		14 (6.90)	18 (8.87)
Resident, n (%)				0.461
Urban		179 (88.18)	174 (85.71)
Rural		24 (11.82)	29 (14.29)
Household monthly income, RMB, n (%)				0.156
≤3000		19 (9.36)	10 (4.93)
3000–6000		109 (53.70)	103 (50.74)
6000–12,000		59 (29.06)	65 (32.02)
>12,000		16 (7.88)	25 (12.31)
Leisure time physical activity per week, h, n (%)				< 0.001
<1		111 (54.68)	40 (19.71)
1–3		60 (29.56)	93 (45.81)
3–5		23 (11.33)	46 (22.66)
5–7		5 (2.46)	13 (6.40)
≥7		4 (1.97)	11 (5.42)
Daily sleep time, h, n (%)				< 0.001
<8		163 (80.30)	125 (61.58)
≥8		40 (19.70)	78 (38.42)
Smoking status, n (%)				0.002
No		187 (92.12)	200 (98.52)
Yes		16 (8.87)	3 (1.48)
Alcohol status, n (%)				0.001
No		188 (92.61)	201 (99.01)
Yes		15 (3.94)	2 (0.00)
Family history of mental disorders, n (%)				0.001
No		185 (91.13)	200 (98.52)
Yes		18 (8.87)	3 (1.48)
Suicide, n (%)				< 0.001
No		32 (15.76)	196 (96.55)
Suicide ideation		97 (47.78)	5 (2.46)
Suicide attempt		74 (36.46)	2 (0.99)
NSSI, n (%)				< 0.001
No		64 (31.53)	199 (98.03)
Yes		139 (68.47)	4 (1.97)

MDD, major depressive disorder; HCs, health controls; SD, standard deviation; HAMD-17, the 17-item Hamilton depression rating scale; HAMA-14, the 14-item Hamilton anxiety rating scale; BMI, body mass index; NSSI, non-suicidal self-injury

Quantitative presented as Mean (SD) or percentage

P-value of HCs and MDD by Mann-Whitney U Test and Chi-Square Test, respectively

Correlation analyses

Pearson correlation analyses in the full sample showed that copper (Cu) and selenium (Se) were positively correlated with HAMD-17 scores (r = 0.176 and 0.161, both P < 0.01), whereas I, Rb, Ni, and Cs were negatively correlated (r = − 0.196 to − 0.115, P < 0.05) (Table A.3). No significant correlations were found within either the HC or MDD subgroup (all P > 0.05). After controlling for covariates, partial correlation results in the full sample remained consistent with the simple Pearson correlations, yet again no significant associations were detected within either subgroup (Table A.3).

Metal exposure model for adolescent MDD

The associations between single metal exposure and the risk of adolescent MDD were detailed in Fig. 2 and Table A.4. In the adjusted models, we observed a negative association between the trends of Mg, Ni, As, I, Cs and the risk of adolescent MDD (all p-trend < 0.05). In the highest quartiles, negative correlations were found for exposures to Ni and I when compared to the lowest quartile (quartile 4 vs. quartile 1: OR = 0.44, 95% CI: 0.20–0.97; OR = 0.38, 95% CI: 0.18–0.81, respectively). Additionally, for I and Cs, the trends in ORs suggested that higher concentrations reinforced the negative correlations with the risk of adolescent MDD. We used LASSO regression to screen the 21 metals and selected Copper (Cu), iron (Fe), Ni, and I for inclusion in the multi-metal model (Fig. 3A and B). Table 2 showed that Cu and Fe were positively correlated with adolescent MDD, with ORs of 9.84 (95% CI: 2.24, 43.30) and 4.54 (95% CI: 1.42, 14.53), respectively. In contrast, there was a negative association between Ni and I with the risk of adolescent MDD, with ORs of 0.26 (95% CI: 0.08, 0.84) and 0.12 (95% CI: 0.03, 0.41), respectively. The VIF values for Cu, Fe, Ni, and I were all below 2.0 (Table A.5.), and Spearman correlations among these metals ranged from 0.07 to 0.55 (Figure A.1), indicating no substantial multicollinearity. Sensitivity analyses using continuous natural log-transformed concentrations yielded consistent results (Table A.6).

Fig. 2.

The single-metal models for the association between urine metals and adolescent MDD. Odds ratios (95% CIs) for adolescent MDD associated with individual metal exposure across quartiles (Q2–Q4, Q1 as reference), estimated by conditional logistic regression adjusted for age, BMI, residence, income, physical activity, sleep, smoking, alcohol, and family history. Red and blue lines indicate significant positive and negative associations (P < 0.05), respectively. Leftward arrows indicate significant negative P-trend (P < 0.05)

Fig. 3.

The single-metal models for the association between urine metals and adolescent MDD. (A) Trajectory of variable coefficient changes (λ). (B) Cross-validation plot for the penalty term

Table 2.

Multi-metal models for adolescent MDD associated with multiple metals and characteristics of participants

Metals	Quartile 1	Quartile 2	Quartile 3	Quartile 4	P-trend
		OR (95%CI)	OR (95%CI)	OR (95%CI)
Fe	1 (ref)	1.79 (0.60–5.30)	5.79 (2.00-16.81)	4.54 (1.42–14.53)	0.005
Ni	1 (ref)	0.39 (0.16–0.97)	0.24 (0.08–0.68)	0.26 (0.08–0.84)	0.011
Cu	1 (ref)	4.54 (1.55–13.31)	3.94 (1.29–12.04)	9.84 (2.24–43.3)	0.008
I	1 (ref)	0.22 (0.08–0.61)	0.13 (0.04–0.41)	0.12 (0.03–0.41)	< 0.001

The multi-metal models in conditional logistic regression analysis were adjusted for age, BMI, resident, household monthly income, leisure time physical activity, daily sleep time, smoking status, alcohol status and family history of mental disorders

BKMR analyses for adolescent MDD

To assess the influence of exposure to a polymetallic mixture exposure, we further conducted BKMR analyses. Initially, the posterior inclusion probability (PIP) for each metal was calculated (Table A.7) to determine its respective importance in the model. Higher PIP values (closer to 1) indicated greater importance, while lower values (closer to 0) suggested less impact. The univariate exposure-response functions between urinary metal levels and the risk of adolescent MDD were displayed in Fig. 4A, with all other metal levels set to their median values and the analysis adjusted for covariates. Notably, Cu and Fe exhibited positive associations with adolescent MDD, whereas Ni and I showed negative associations. Moreover, the cumulative effects of all metals were set to different percentiles (P₂₅ to P₇₅, and step value = 5th) compared to their median levels. Metal combinations with total metal content above the 50th percentile (P₅₀) exhibited a significantly negative association with adolescent MDD relative to those at the P₅₀ level. When total metal concentrations were below the P50 level, no statistical difference was found, although there was a trend toward decreased risk (Fig. 4B). As illustrated in Fig. 4C, when other metals were fixed at their P₂₅ and P₅₀ levels, the difference value of Ni_P75−P25 was reduced. When the levels of other metals were increased to their P₇₅ levels, although the effect continued to rise, there was no statistical significance for close to 0. Conversely, for I, as the levels of other metals increased, the effect of its P₇₅ level became gradually smaller compared to its P₂₅ level. Additionally, the influences of Cu and Fe at disparate levels were not statistically significant, regardless of the levels of other metals. Figure 4D showed a possible joint effect of Cu and Fe on increasing the risk of adolescent MDD. The positive slope of Cu became steeper at higher concentrations of urinary Fe, while the other two metals reached the median. Conversely, when urinary Ni or I concentration increased, the slope of Cu became flattened.

Fig. 4.

The BKMR models of each metal and mixed exposure effects on adolescent MDD. (A) Univariate exposure-response functions of each metal, along with their respective 95% confidence interval (95% CI), while keeping the concentrations of other metals fixed at their medians (P₅₀). (B) The overall effects of mixed exposure, with metals fixed at different percentiles compared to their concentrations when they were at their medians (P₅₀). (C) The effects of a single exposure when an individual metal was at its 75th percentile, contrasted with the scenario when that exposure is at its 25th percentile. Notably, all other exposures were fixed at P₇₅. (D) The bivariate cross-section effects of the exposure-response function of a single metal, where the second metal was fixed at P₂₅, P₅₀ and P₇₅. This analysis provides insights into the joint effects of these metals at different concentration levels

Sensitivity analysis

To assess the robustness of our findings, we conducted two sets of sensitivity analyses. First, we performed subgroup analyses restricted to non-smokers, non-drinkers, and participants without a family history of mental disorders; LASSO regression in these subgroups selected variables with no significant difference (Figure A.2–4), and the multi-model results showed no significant variations from the primary analysis (Table A.8–10). Second, to evaluate whether LASSO pre-screening omitted relevant mixture components, we reran BKMR using all 11 metals retained at lambda.min. Cu and I maintained absolute dominance (PIP = 1.000) across both models. Fe and Ni showed consistently high PIPs in the sensitivity analysis (0.991 and 0.960, respectively) (Table A.11). The inclusion of the other additional metals did not alter the direction of the overall mixture effect, consistent with the primary model (Figure A.5).

Discussion

This study aimed to explore the association between urinary metals and adolescent MDD, providing new insight into the early prevention of adolescent MDD. Correlation analyses indicated that Cu, Ni, Se, I, Rb and Cs were significantly associated with HAMD-17 in the full sample, but not within the HC or MDD subgroups, suggesting these metals may reflect diagnostic state rather than symptom severity. Results from the single-metal model indicated that urinary levels of Ni, and I exhibited protective effects to adolescent MDD. In the multi-metal model, elevated levels of Fe and Cu were positively associated with the risk of adolescent MDD, while Ni and I were inversely associated with adolescent MDD. The BKMR model confirmed the multi-metal model findings through univariate exposure-response functions. The combined exposure to multiple metals showed the protective effects of I and Ni outweighed the negative effects from Cu and Fe, resulting in an association with lower odds of MDD in adolescent with the comprehensive multi-metal exposure model. Additionally, bivariate exposure-response functions showed a positive association between the combined exposure to Fe and Cu and the risk of adolescent MDD, whereas an inverse interaction may exist between Ni or I and Cu.

In this study, we observed a positive association between urinary Cu levels and the risk of adolescent MDD, which was consistent with existing literature [56]. A review by Ni et al. indicates that increased Cu concentrations have been hypothesized to contribute to the degradation of 5-hydroxytryptamine and modify both the activity and concentration of dopamine and norepinephrine within the central nervous system [57]. This disturbance in the balance of monoamine neurotransmitters is associated with the pathogenesis of major depressive disorder (MDD) [58]. Furthermore, research suggests that Cu has been shown to affect cellular prion proteins, which regulate N-methyl-D-aspartate receptor activity, potentially disrupting brain glutamatergic transmission and causing neurological dysfunction [59]. Studies show that the concentration of copper/zinc superoxide dismutase is elevated in patients with depression and decreases following antidepressant treatment, suggesting a potential link between depression and copper-mediated oxidative stress [60, 61]. Additionally, Cu may influence the pathogenesis of MDD via the immune system [62]. As components of cytochrome oxidase and superoxide dismutase, changes in Cu levels may alter their activity, impact free radical clearance and impairing immune function [57].

Fe is an essential micronutrient, and even minor disturbances in iron homeostasis could impair motor and cognitive function by affecting neurotransmitters such as serotonin, norepinephrine, and dopamine, which have been linked to psychiatric disorders [63, 64]. In the present study, we established a positive correlation between urinary Fe levels and adolescent MDD. Consistent with our findings, Zeng et al. gave mice subject to chronic unpredictable mild stress (CUMS) a high-iron diet and found that the elevated serum Fe levels were significantly correlated with impaired hippocampal functional connectivity, which was associated with CUMS [65]. A study on children with Thalassemia showed that Fe overload induced by transfusion treatment was associated with major depression-like episodes in this population [66]. Another study has reported elevated serum ferritin levels in some patients with depression, and significant improvement in depressive symptoms after treatment with iron chelators [67]. In addition, previous studies have examined brain iron levels in patients with MDD using quantitative magnetic susceptibility imaging (QSM) and observed elevated iron concentrations in certain brain regions in patients with MDD [68, 69]. These findings support the hypothesis that iron overload was associated with depression through mechanisms that affect neurotransmitter synthesis and neuroinflammation. Notably, Liang et al. indicated that both Fe excess and deficiency have been associated with MDD, suggesting that iron metabolism (including excess and deficiency) may affect brain function through different mechanisms, which may contribute to depression [70]. The precise threshold and regulatory mechanisms of iron homeostasis should be explored in the future.

Furthermore, we observed an inverse association between urinary Ni levels and adolescent MDD. A recent study demonstrated that patients with MDD had significantly lower urinary concentrations of Ni than HCs [21], which is consistent with our findings. At low environmental exposure levels, Ni serves as a cofactor for several metalloenzymes and [71, 72] may be involved in pathways related to serotonin and dopamine synthesis [73, 74], which are implicated in MDD mechanisms. However, Ni is generally regarded as a toxic heavy metal and has been linked to oxidative stress and neurotoxicity; therefore, the inverse association observed for Ni should be interpreted cautiously [75, 76]. While we have described potential biological mechanisms, these explanations are speculative given the limited evidence on Ni’s neuroprotective effects at environmental exposure levels. The association may reflect unmeasured confounding by dietary habits, socioeconomic status, or reverse causality [77, 78]. Therefore, further research is needed to validate these findings. Additionally, our study showed a negative correlation between urinary I concentrations and the risk of adolescent MDD. A case-control study among adolescents aged 8 to 16 found that high urinary iodine concentration strongly correlated to decreased risk of depression [79]. The inverse association of I might be attributed to its association with altered enzyme activities enzyme activities in the central nervous system [55, 80, 81]. Additionally, I is involved in the synthesis of thyroid hormone (TH), and the lack of TH has been associated with impaired neurogenesis and reduced newborn neuroblasts and immature neurons, which may be linked to psychological disorder in previous studies [82, 83]. Notably, high seafood intake could simultaneously elevate urinary I and Ni levels. In addition, the results of the BKMR model showed a significant positive interaction between Cu and Fe and a significant negative association between Cu and Ni or I. Nielsen et al. showed complex interactions between Ni, Cu, and Fe [84]. Ni acted synergistically with Cu, and that interaction might be affected by the diet of Fe. However, the mechanism of the combined effect of four metals (Fe, Cu, Ni, and I) is still unclear and needs to be further explored.

It is noteworthy that the study population was predominantly female (74.4%) and urban (88%). Although sex and urban/rural residence were adjusted as covariates in all multivariable models to control for confounding, the relatively small numbers of male and rural participants limit statistical power in these subgroups. Consequently, our findings may be more applicable to female adolescents in urban settings. This demographic pattern is relevant given the well-established female predominance in depression incidence, with women having approximately twice the odds of major depression compared to men, peaking during adolescence [4, 85]. Moreover, sex-specific differences in metal metabolism may modify exposure-outcome relationships; for instance, females with MDD exhibit higher serum copper and lower zinc levels than males [86], and the protective associations of several minerals against depression appear more pronounced in women [87, 88]. Additionally, urban populations may experience unique environmental exposure profiles (e.g., traffic-related metal emissions) that differ from rural areas [89, 90]. Therefore, caution should be exercised when generalizing these findings to male adolescents or rural populations, and future research with more balanced demographic samples is warranted.

However, this study has several limitations. First, due to the cross-sectional design, causality between metal co-exposure and the risk of adolescent MDD could not be ascertained. While previous research has indicated that multiple metals may influence the risk of adolescent MDD [21], there is also evidence that MDD could affect metal absorption and metabolism, suggesting a potential bidirectional communication between metals and MDD [91]. Therefore, it is necessary to conduct longitudinal research in the future to explore the relationship between metal and MDD. Second, urinary metal concentrations may be influenced by changes in renal function and urinary dilution, though we further corrected creatinine levels to minimize the error. Third, although study adjusted for some covariates, we did not assess potential sources of human urinary metal concentrations (possible pathways of metal exposure), such as detailed dietary intake, seafood consumption, iodine supplementation, local air quality, and drinking water, which were found to be significantly associated with MDD. This residual confounding may affect the interpretation of the observed associations for Cu, Fe, Ni, and I. For example, high seafood intake could simultaneously elevate urinary I and Ni levels. Therefore, future studies should incorporate comprehensive exposure assessments, including dietary questionnaires, geocoded environmental data, and biomonitoring of multiple exposure matrices (e.g., blood, hair, and toenails), to better disentangle metal-specific effects from underlying confounding factors. Fourth, the wide confidence intervals for the highest quartiles partly reflect the smaller sample size in Q4 and the broad exposure contrast between extreme quartiles. Future studies with larger cohorts are warranted to improve the precision of these estimates. Finally, the relatively small sample size and race may limit the generalizability of the current findings, which need to be further investigated in the future with a larger, more diverse cohorts to strengthen these conclusions.

Conclusions

To summarize, our study explored the relationship between urinary metal levels and adolescent MDD. Specifically, we found that exposure to Cu and Fe was positively associated with the risk of adolescent MDD, whereas Ni and I were inversely associated with adolescent MDD. These findings provide new insights into the potential role of metal exposure in the development of adolescent MDD. Further studies are needed to determine the long-term effects of exposure to metals on MDD and to explore the mechanism of action, thereby validating our results and providing more comprehensive prevention and potential intervention targets.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

MDD

Major depressive disorder

LASSO

Least absolute shrinkage and selection operator

BKMR

Bayesian kernel machine regression

HCs

Healthy controls

ICP-MS

Coupled plasma-mass spectrometry

HAMD-17

Hamilton depression scale-17

BMI

Body mass index

LOD

Limit of detection

SD

Standard deviations

ORs

Odds ratios

MCMC

Markov chain monte carlo

PIP

Posterior inclusion probability

CUMS

Chronic unpredictable mild stress

QSM

Quantitative magnetic susceptibility imaging

TH

Thyroid hormone

Author contributions

Yuqian He: Writing - original draft, Writing - review and editing, Methodology, Investigation, Formal analysis. Xue Li: Writing - original draft, Software, Formal analysis, Visualization. Dan Zhu: Writing - original draft. Xueer Liu: Writing - Review & Editing, Supervision. Xuemei Li: Writing - Review & Editing. Teng Teng: Writing - Review & Editing. Tingting Lei: Visualization. Jie Li: Resources, Data curation. Yajie Huang: Resources, Data curation. Xieyu Gan: Resources, Data curation. Hang Du: Investigation, Methodology, Validation, Funding Acquisition. Yijue Wang: Resources, Data curation. Jingye Zhou: Resources, Data curation. Chengzhi Chen: Writing - Review & Editing, Investigation. Fen Huang: Writing - Review & Editing, Software, Validation, Investigation, Formal analysis. Xinyu Zhou: Writing - Review & Editing, Investigation, Supervision, Project administration, Funding acquisition.

Funding

This work was supported by STI2030-Major Projects (2022ZD0212900), the National Key R&D Program of China (2024YFC2707800), the National Natural Science Foundation of China (82271565), Natural Science Foundation of Chongqing, China (CSTB2023NSCO-MSX0381), Natural Science Foundation of Chongqing, China (CSSTC2020JCYJ-MSXM3341), Joint Project of Chongqing Health Commission and Science and Technology Bureau (2022MSXM076).

Data availability

Data will be made available on request.

Declarations

Ethics approval and consent to participate

This study was approved by the ethics committees of the First Affiliated Hospital of Chongqing Medical University (2020 − 864). Written informed consent was obtained from every participant and at least one of the parents prior to study initiation. All procedures were undertaken in accordance with Declaration of Helsinki.

Consent for publication

Written informed consent for publication was obtained from all participants and their parents.

Competing interests

The authors declare no competing interests.