Abstract
Background
Behavioral problems affect 10−20% of children worldwide. Conventional pediatric assessment of risk factors for behavioral problems relies largely on single-biomarker analysis, which fails to capture the synergistic or antagonistic interactions among nutritional and metabolic factors and is further hampered by multicollinearity among biomarkers. Whether integrated nutritional-metabolic patterns improve risk identification beyond individual biomarkers remains unclear. We therefore combined multivariable regression, explainable machine learning, and principal component analysis (PCA) to identify integrated nutritional-metabolic risk patterns for behavioral problems in children.
Methods
This cross-sectional study included 5,468 children aged 6−18 years who underwent comprehensive health check-ups at the Health Management Center of a tertiary Children’s Hospital in China between August 2022 and December 2025 and who had complete biomarker and behavioral data. Blood magnesium, iron, copper, zinc, and lead were measured by inductively coupled plasma mass spectrometry (ICP-MS); triglycerides (TG), calcium, and phosphorus by enzymatic or colorimetric assays; and 25-hydroxyvitamin D [25(OH)D] by chemiluminescent immunoassay (CLIA). Behavioral problems were assessed using the Conners’ Parent Rating Scale (CPRS) based on Chinese norms. Covariates (age, sex, and body mass index) were obtained from standardized anthropometric measurements. Multivariable logistic regression, XGBoost with SHapley Additive exPlanations (SHAP), and PCA were applied.
Results
Among the participants (median age 9.41 years; 61.5% males), 624 (11.41%) had behavioral problems. Higher blood lead [odds ratio (OR) 2.26, 95% confidence interval (CI): 1.51–3.37] and TG (OR 1.77, 95% CI: 1.10–2.82) were associated with increased odds of behavioral problems, whereas iron was protective (OR 0.06, 95% CI: 0.01–0.39). SHAP analysis ranked lead, iron, vitamin D, and TG as the most influential predictors. PCA derived four patterns; Pattern 4 (lead exposure-nutritional antagonism: elevated lead with low calcium and vitamin D) showed the strongest, dose-response association (highest vs. lowest tertile: OR 1.65, 95% CI: 1.34–2.04; Ptrend <0.001).
Conclusions
In this health check-up sample, a pattern characterized by elevated lead combined with insufficient calcium and vitamin D was associated with behavioral problems in a dose-response manner. Although causality cannot be inferred, pattern-based nutritional-metabolic profiling during routine pediatric check-ups may help identify children who could benefit from closer behavioral monitoring, nutritional optimization, and lead exposure reduction.
Keywords: Behavioral problems, children, nutritional and metabolic patterns, machine learning, principal component analysis (PCA)
Highlight box.
Key findings
• In a cross-sectional study of 5,468 school-age children (6−18 years), the prevalence of behavioral problems was 11.41%.
• Elevated level of blood lead, and triglycerides as well as reduced level of serum iron and vitamin D, are significantly associated with children’s behavioral problems.
• Principal component analysis (PCA) identified four distinct nutritional-metabolic patterns. Pattern 4, characterized by elevated blood lead co-occurring with reduced serum calcium and vitamin D, showed the strongest association with behavioral problems.
What is known and what is new?
• Environmental toxics exposure (such as lead) and individual nutritional deficiencies (e.g., iron, vitamin D) are each independently linked to adverse neurobehavioral outcomes in children.
• By employing a pattern-based approach using PCA and machine learning approach with SHapley Additive exPlanations explainability, this study reveals that the synergistic combination of elevated blood lead with reduced calcium and vitamin D (Pattern 4) confers a substantially higher behavioral risk than any single factor alone, suggesting a “toxin-nutrient antagonism” mechanism.
What is the implication, and what should change now?
• Clinical pediatric practice should consider integrated nutritional-metabolic profiling rather than single-biomarker assessment for early identification of children at elevated behavioral risk.
• Pattern 4 may serve as a practical screening signature: children presenting with elevated blood lead alongside low calcium and vitamin D warrant prioritized nutritional intervention and closer behavioral monitoring.
Introduction
The rising incidence of behavioral problems in children has become a notable public health concern, affecting 10% to 20% of the pediatric population worldwide (1,2). Childhood behavioral problems are not a single clinical entity; rather, they comprise heterogeneous manifestations, ranging from oppositional behavior and hyperactivity to anxiety and psychosomatic complaints. These outcomes are shaped by a complex interplay of biological, psychological, and social determinants, including family environment, parenting style, adverse childhood experiences, peer relationships, school context, and academic stress. If left unaddressed, behavioral problems may track into adulthood and increase the risk of physical and mental disorders in adulthood such as substance abuse and cardiometabolic disorders (3,4). Within this multi-determinant framework, identifying measurable and modifiable biological factors offers a practical entry point for early prevention and intervention, especially because the brain is in a highly plastic state during childhood.
A growing body of evidence indicates that children’s behavioral problems are associated with nutritional and metabolic factors. For instance, long-term exposure to environmental toxicants, especially lead (Pb), remains a major concern even at low blood levels (5). More specifically, contemporary low-level lead exposure throughout childhood was associated with externalizing problems, such as aggression and hyperactivity (6), and blood lead levels partially mediated the relationship between social adversity and externalizing behavior in youth (7). For iron, a meta-analysis of 11 studies found that children with attention-deficit/hyperactivity disorder (ADHD) had significantly lower serum ferritin levels than controls (8), and a longitudinal study by Doom et al. showed that iron deficiency in infancy predicted both internalizing and externalizing problems in adolescence (9). Zinc deficiency has been linked to impaired attention and impulsivity through its role in dopaminergic neurotransmission (10,11). Regarding vitamin D, deficiency has been linked to an elevated risk of neurodevelopmental disorders (12,13), and a review has emphasized its role in synaptic and circuit functions during neurodevelopment (14). Recent findings further implicate lipid metabolism, indicating that elevated triglyceride (TG) levels are associated with diminished executive function in adolescents (15). Additionally, dyslipidemia has been connected to neuroinflammation and disruptions in brain energy homeostasis (16,17). Collectively, these studies suggest that nutritional and metabolic disturbances significantly contribute to behavioral risk.
However, the majority of research in this area has predominantly focused on individual biomarkers in isolation (18,19), which cannot adequately capture synergistic or antagonistic interactions among nutrients. Moreover, multicollinearity among nutritional biomarkers makes it difficult to disentangle the contribution of each biomarker in conventional regression models (20,21). Pattern-based approaches such as principal component analysis (PCA) are well established in nutritional epidemiology for characterizing combined exposures, but they have rarely been applied to objectively measured nutritional-metabolic biomarkers in relation to children’s behavioral outcomes. As a result, clinicians still lack integrated, interpretable signatures to guide decisions when faced with a complex laboratory report.
To overcome these constraints, this study leveraged a large dataset of 5,468 children undergoing routine health examination at the Health Management Center of a tertiary Children’s Hospital to transcend the “single-nutrient” approach. In this study, “behavioral problems” were operationally defined as parent-reported scores that exceed two standard deviations (SDs) above the age- and sex-specific norms for Chinese populations on any subscale of the Conners’ Parent Rating Scale (CPRS). This definition is based on a screening-oriented, dimensional approach rather than a clinical diagnosis. Combining explainable machine learning with PCA (22-24), we pursued two objectives: first, to objectively evaluate the relative importance of a variety of different nutritional and metabolic parameters of behavioral problems; and second, to uncover “integrated metabolic patterns” that more accurately reflect the complexity of a child’s nutritional status in real world. We explicitly acknowledge that these biological patterns represent only one layer of the multi-determinant model of child behavior; accordingly, our findings are intended to provide complementary screening information rather than deterministic explanations. We present this article in accordance with the STROBE reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0370/rc).
Methods
Study design and participant selection
This retrospective observational study extracted the electronic health records (EHRs) of children who underwent comprehensive health check-ups at the Health Management Center of the Children’s Hospital, Zhejiang University School of Medicine, between August 15, 2022, and December 31, 2025. The inclusion criteria were: (I) 6 to 18 years old; (II) availability of all data on magnesium, iron, copper, zinc, lead; serum calcium, phosphorus, TGs, 25-hydroxyvitamin D [25(OH)D]; and behavioral assessments; and (III) absence of any neurological disorders and physical disabilities. The exclusion criteria were: (I) multiple EHRs per child, retaining only the earliest one; and (II) missing data on any above-mentioned variable in the EHRs. The details of the sample size consideration were provided (Appendix 1).
Ethical considerations
The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments (25). This study was approved by the Medical Ethics Committee of the Children’s Hospital, Zhejiang University School of Medicine (No. 2021-IRB-185). Waivers for informed consent were obtained since the study was a retrospective analysis of anonymized electronic health records without risk to the participants.
Behavioral outcome assessment
The behavioral problems were measured by the 48-item CPRS, a validated scale for children and adolescents between the ages of 3 and 17 years (26). This scale measures five functional behavioral domains: oppositional behavior, learning problems, psychosomatic symptoms, hyperactivity-impulsivity, and anxiety. Each item has a 4-point Likert scale [0−3], and the scores for the subscales are the mean of the items in the scale. According to the Chinese standardized scale, behavioral problems were considered to be more than two SDs above the mean on a subscale (27). “Any behavioral problem” was defined as meeting this threshold on at least one of the five subscales. All the questionnaires were filled in by the parents or primary guardians of the children under the guidance of well-trained health professionals.
Laboratory exposure measurements
Venous blood samples were collected into heparin vacuum tubes in the morning after an overnight fast by trained pediatric nurses under aseptic conditions. Serum levels of five elements (magnesium, iron, copper, zinc, and lead) were determined by inductively coupled plasma mass spectrometry (ICP-MS). An aliquot of 50−100 μL of whole blood was digested with nitric acid/hydrogen peroxide by microwave or thermal digestion. The digest was then diluted to 5−10 mL with ultrapure water containing internal standards for real-time correction (ICP-MS Trace Element Detection Kit, Hangzhou Du’an Baiyu Technology Co., Ltd., Hangzhou, China). Serum TG concentration was measured by an enzymatic colorimetric method using the glycerol phosphate oxidase-peroxidase (GPO-POD) method (Mindray BS series). Serum calcium and phosphorus levels were determined by colorimetric methods using arsenazo III and molybdenum blue, respectively. Serum 25(OH)D levels were determined by chemiluminescent immunoassay (CLIA) (25-Hydroxy Vitamin D Assay Kit, Orienter, Chengdu, China). The laboratory strictly adhered to the manufacturer’s instructions to ensure that the intra- and inter-assay coefficients of variation were within the clinical standards.
Covariate evaluation
Potential confounders such as age, sex, height and weight were selected based on clinical importance and previous studies. Anthropometric measurements such as height and weight were conducted using standardized procedures when the children were in light clothing without shoes. Body mass index (BMI) was calculated as weight divided by height squared (kg/m2).
Statistical analysis
Data normality was checked by the Shapiro-Wilk test. Non-normally distributed continuous variables {reported as median [Interquartile range (IQR)]} and categorical variables [number (%)] were compared by the Mann-Whitney U test and the Chi-squared test, respectively. Right skewed variables, such as trace elements and TGs, were log10-transformed before multivariable modeling to ensure symmetry of the distributions. In order to determine key predictors for behavioral problems, we combined traditional multivariable logistic regression with machine learning-based feature importance. The contribution of each metabolite was determined using SHapley Additive exPlanations (SHAP) values from the XGBoost model to provide a biologically interpretable ranking of feature importance. PCA was used to examine integrated nutrients patterns based on Z-score standardization of the biomarkers. Sampling adequacy was tested by Kaiser-Meyer-Olkin and Bartlett’s test of sphericity. Components were selected based on Kaiser’s rule, scree plot, and Horn’s parallel analysis. Significant factor loadings were those having absolute values higher than 0.30. Tertiles of factor scores were created for each of the factors extracted, and their relationship with behavioral problems was tested by binary logistic regression, both crude and adjusted by covariates. Linear trend was tested by considering the median of each tertile of each factor score as a continuous variable. Sensitivity of the main findings to unmeasured confounding was assessed using E-values (28).
Descriptive statistics, multivariable logistic regression, and PCA were performed using R (version 4.4.1; R Foundation for Statistical Computing, Vienna, Austria). Machine learning models, including XGBoost, and SHAP-based feature importance analysis, were implemented using Python (version 3.10; Python Software Foundation, Wilmington, DE, USA). Statistical significance was defined as a two-tailed P<0.05.
Results
Study population characteristics
Table 1 presents the characteristics of the study population and comparison between children with and without behavioral problems. The study population included a total of 5,468 children, with a median age of 9.41 years, and 61.49% (3,362 out of 5,468) being males. A total of 11.41% (624 out of 5,468) children were identified with behavioral problems, according to the CPRS. Compared with children without behavioral problems, children with behavioral problems had significantly higher levels of blood lead, with a median of 9.00 compared with 8.00 μg/L (P<0.001), and were more likely to have high TGs (16.99% vs. 13.65%, P=0.02). In contrast, children with behavioral problems had lower iron levels, with a median of 6.71 compared with 6.81 mmol/L (P=0.001), and lower levels of magnesium (P=0.048), but no significant differences were observed between the two groups regarding age, sex, BMI, copper, zinc, calcium, phosphorus, and vitamin D (P>0.05).
Table 1. Characteristics and elements distribution of the study population.
| Characteristics | Overall (n=5,468) | Children without behavior problems (n=4,844) | Children with behavior problems (n=624) | P value |
|---|---|---|---|---|
| Age (year) | 9.41 (7.57, 11.66) | 9.49 (7.57, 11.74) | 9.33 (7.66, 11.41) | 0.50 |
| Sex | 0.40 | |||
| Male | 3,362 (61.49) | 2,988 (61.68) | 374 (59.94) | |
| Female | 2,106 (38.51) | 1,856 (38.32) | 250 (40.06) | |
| Height (cm) | 138.80 (127.50, 153.30) | 139.00 (127.50, 154.00) | 137.15 (127.20, 150.00) | 0.08 |
| Weight (kg) | 31.20 (24.20, 44.00) | 31.35 (24.20, 44.10) | 30.55 (24.00, 43.12) | 0.16 |
| BMI (kg/m2) | 16.04 (14.47, 18.92) | 16.05 (14.48, 18.90) | 15.93 (14.41, 18.98) | 0.41 |
| Weight status | 0.62 | |||
| Thin | 486 (8.89) | 422 (8.71) | 64 (10.26) | |
| Normal | 3,741 (68.42) | 3,319 (68.52) | 422 (67.63) | |
| Overweight | 692 (12.66) | 617 (12.74) | 75 (12.02) | |
| Obesity | 549 (10.04) | 486 (10.03) | 63 (10.10) | |
| Magnesium (mmol/L) | 1.50 (1.39, 1.61) | 1.50 (1.39, 1.61) | 1.49 (1.39, 1.60) | 0.048 |
| Iron (mmol/L) | 6.79 (6.28, 7.35) | 6.81 (6.29, 7.36) | 6.71 (6.19, 7.26) | 0.001 |
| Copper (μmol/L) | 12.44 (11.02, 13.93) | 12.45 (11.04, 13.93) | 12.39 (10.90, 13.94) | 0.74 |
| Zinc (μmol/L) | 67.25 (60.40, 74.81) | 67.33 (60.51, 75.02) | 66.70 (59.63, 73.37) | 0.06 |
| Lead (μg/L) | 8.00 (6.00, 11.00) | 8.00 (6.00, 11.00) | 9.00 (7.00, 12.00) | <0.001 |
| Calcium (mmol/L) | 2.46 (2.41, 2.51) | 2.46 (2.41, 2.51) | 2.46 (2.41, 2.51) | 0.29 |
| Phosphorus (mmol/L) | 1.66 (1.56, 1.77) | 1.66 (1.56, 1.77) | 1.67 (1.57, 1.77) | 0.39 |
| Triglycerides (mmol/L) | 0.74 (0.56, 1.02) | 0.73 (0.56, 1.01) | 0.77 (0.57, 1.08) | 0.09 |
| High triglycerides | 0.02 | |||
| No | 4,701 (85.97) | 4,183 (86.35) | 518 (83.01) | |
| Yes | 767 (14.03) | 661 (13.65) | 106 (16.99) | |
| Vitamin D (nmol/L) | 68.70 (53.50, 86.00) | 68.90 (53.50, 86.40) | 66.00 (53.80, 84.00) | 0.07 |
| Nutritional status of vitamin D | 0.09 | |||
| Sufficiency | 4,434 (81.09) | 3,933 (81.19) | 501 (80.29) | |
| Insufficiency | 881 (16.11) | 784 (16.18) | 97 (15.54) | |
| Deficiency | 153 (2.80) | 127 (2.62) | 26 (4.17) |
Values are median (interquartile range) or n (%) unless otherwise indicated. BMI, body mass index.
Multivariable logistic regression analysis
Table 2 presents the results of multivariable logistic regression. Blood lead levels were significantly associated with the risk of behavioral problems [odds ratio (OR) 2.26, 95% confidence interval (CI): 1.51–3.37, P<0.001], as were TGs (OR 1.77, 95% CI: 1.10–2.82, P=0.02). In contrast, serum iron levels were significantly associated with behavioral problems in a protective manner (OR 0.06, 95% CI: 0.01–0.39, P=0.003). No significant associations were found for age, sex, BMI, or other elements including magnesium, copper, zinc, calcium, phosphorus and vitamin D (all P>0.05).
Table 2. Logistic regression analysis of factors associated with behavioral problems in children.
| Variables | OR (95% CI) | P value |
|---|---|---|
| Age | 1.00 (0.96–1.04) | 0.99 |
| Sex | ||
| Male | Ref. | |
| Female | 1.07 (0.89–1.28) | 0.48 |
| BMI | 0.99 (0.96–1.01) | 0.33 |
| Magnesium (log10-transformed) | 0.40 (0.05–3.38) | 0.40 |
| Iron (log10-transformed) | 0.06 (0.01–0.39) | 0.003 |
| Copper (log10-transformed) | 1.40 (0.40–4.86) | 0.60 |
| Zinc (log10-transformed) | 0.68 (0.19–2.47) | 0.56 |
| Lead (log10-transformed) | 2.26 (1.51–3.37) | <0.001 |
| Calcium | 0.77 (0.25–2.35) | 0.64 |
| Phosphorus | 1.11 (0.65–1.90) | 0.70 |
| Triglycerides (log10-transformed) | 1.77 (1.10–2.82) | 0.02 |
| Vitamin D | 1.00 (0.99–1.00) | 0.25 |
BMI, body mass index; CI, confidence interval; OR, odds ratio; Ref., reference.
Feature importance identified by explainable machine learning
Figure 1 illustrates the relative importance of nutritional and metabolic parameters for behavioral problems, as ranked by SHAP values derived from an XGBoost model. Consistent with the logistic regression results, the most important factors were again identified as serum lead (Pb) and iron (Fe). The summary plot of the SHAP analysis indicated that higher levels of lead were related to increased behavioral risk (indicated by the presence of more red dots on the right), while higher iron levels were related to a strong protective effect against behavioral risk (indicated by the presence of more red dots on the left). To be noted, vitamin D and TGs as the third and fourth most important factors, respectively. Although vitamin D was not a statistically significant variable in the logistic regression analysis, the XGBoost analysis indicated that it was a potentially important variable; higher levels were generally related to lower behavioral risk. Other variables were found to have a relatively lower impact on the model’s results.
Figure 1.
Feature importance ranking of nutritional and metabolic parameters based on SHAP values. SHAP, SHapley Additive exPlanations.
Identification of integrated patterns via PCA
Principal components (PCs) were defined based on biomarkers with absolute factor loadings exceeding 0.30. As illustrated in Figure 2, the first four PCs (PC1 to PC4) were retained, explaining 60.8% of the total variability of the biomarkers. Table 3 demonstrates the factor loadings for the four PCs, or the so-called integrated nutritional and metabolic patterns. Pattern 1 (essential elements factor) showed high positive factor loadings for magnesium (0.557), iron (0.530), zinc (0.392) and copper (0.306). Pattern 2 (nutritional-metabolic regulatory factor) showed positive factor loadings for vitamin D (0.582) and copper (0.514) but also showed an inverse relationship for TGs (−0.549). Pattern 3 (mineral metabolism balance factor) showed negative factor loadings for phosphorus (−0.694) and calcium (−0.397), as well as a positive loading for zinc (0.471). Pattern 4 (lead exposure-nutritional antagonism factor) showed high positive loadings for lead (0.800) and negative loading for calcium (−0.435) and vitamin D (−0.386).
Figure 2.
Scree plot illustrating the individual and cumulative variance explained by the PCs. PCs, principal components.
Table 3. Factor loadings for the four integrated nutritional and metabolic patterns derived from PCA†.
| Variables | Pattern 1 | Pattern 2 | Pattern 3 | Pattern 4 |
|---|---|---|---|---|
| Magnesium | 0.557 | |||
| Iron | 0.530 | |||
| Zinc | 0.392 | 0.471 | ||
| Copper | 0.306 | 0.514 | ||
| Vitamin D | 0.582 | −0.386 | ||
| Triglycerides | −0.549 | |||
| Calcium | −0.397 | −0.435 | ||
| Phosphorus | −0.694 | |||
| Lead | 0.800 |
†, only factor loadings with absolute values >0.30 are listed in the table. Pattern 1: essential elements factor; Pattern 2: nutritional-metabolic regulatory factor; Pattern 3: mineral metabolism balance factor; Pattern 4: lead exposure-nutritional antagonism factor. PCA, principal component analysis.
To evaluate the discriminatory capacity of the identified patterns, standardized factor scores were compared between children with and without behavioral problems, and the results are shown in Table 4. Children with behavioral problems exhibited a significantly higher score for Pattern 4 (lead exposure-nutritional antagonism factor) compared to the control group (mean 0.15 vs. −0.02; P<0.001), suggesting a strong association between lead-driven profiles and behavioral risk. In contrast, the behavioral problems group had significantly lower scores for Pattern 1 (essential elements factor) (mean −0.11 vs. 0.01; P=0.04) and Pattern 3 (mineral metabolism balance factor) (mean −0.09 vs. 0.01; P=0.04). No significant difference was observed for Pattern 2 (nutritional-metabolic regulatory factor) between the two groups (mean −0.04 vs. 0.01; P=0.36).
Table 4. Comparison of nutritional and metabolic pattern factor scores between children with and without behavioral problems.
| Patterns | Behavior problems group | Control group | P value |
|---|---|---|---|
| 1st | −0.11 (1.38) | 0.01 (1.38) | 0.04 |
| 2nd | −0.04 (1.25) | 0.01 (1.16) | 0.36 |
| 3rd | −0.09 (1.08) | 0.01 (1.09) | 0.04 |
| 4th | 0.15 (1.02) | −0.02 (1.01) | <0.001 |
Data are presented as mean (SD). Pattern 1: essential elements factor; Pattern 2: nutritional-metabolic regulatory factor; Pattern 3: mineral metabolism balance factor; Pattern 4: lead exposure-nutritional antagonism factor. SD, standard deviation.
Dose-response association between patterns and behavioral problems
Table 5 presents the results of multivariable logistic regression analysis to assess the dose-response associations for behavioral problems across the tertiles of each pattern score. Pattern 4 (lead exposure-nutritional antagonism factor) showed a significant dose-response relationship between pattern score and risk of behavioral problems (Ptrend <0.001). Compared to the reference group of the first tertile of Pattern 4 (T1), the odds of experiencing behavioral problems were 1.41-fold (95% CI: 1.14–1.75) and 1.65-fold (95% CI: 1.34–2.04) higher for the second (T2) and third (T3) tertiles of Pattern 4, respectively. No significant associations or dose-response trends were detected for Pattern 1, Pattern 2, or Pattern 3 in the adjusted models (all Ptrend >0.05).
Table 5. Multivariable logistic regression analysis of the association between integrated nutritional and metabolic pattern tertiles and behavioral problems.
| Patterns | Tertiles | Behavior problems group | Control group | Crude model, OR (95% CI) | Adjusted model, OR (95% CI) |
|---|---|---|---|---|---|
| 1st | T1 | 226 (12.40) | 1,597 (87.60) | 1.15 (0.94–1.41) | 1.13 (0.92–1.39) |
| T2 | 198 (10.87) | 1,624 (89.13) | 0.99 (0.80–1.22) | 0.98 (0.79–1.21) | |
| T3 | 200 (10.97) | 1,623 (89.03) | Ref. | Ref. | |
| Ptrend | 0.18 | 0.18 | 0.25 | ||
| 2nd | T1 | 231 (12.67) | 1,592 (87.33) | 1.15 (0.94–1.40) | 1.24 (0.99–1.56) |
| T2 | 188 (10.32) | 1,634 (89.68) | 0.91 (0.74–1.12) | 0.94 (0.76–1.17) | |
| T3 | 205 (11.25) | 1,618 (88.75) | Ref. | Ref. | |
| Ptrend | 0.18 | 0.17 | 0.055 | ||
| 3rd | T1 | 224 (12.29) | 1,599 (87.71) | 1.15 (0.94–1.41) | 1.14 (0.92–1.41) |
| T2 | 202 (11.09) | 1,620 (88.91) | 1.02 (0.83–1.26) | 1.01 (0.81–1.25) | |
| T3 | 198 (10.86) | 1,625 (89.14) | Ref. | Ref. | |
| Ptrend | 0.18 | 0.18 | 0.24 | ||
| 4th | T1 | 160 (8.78) | 1,663 (91.22) | Ref. | Ref. |
| T2 | 217 (11.91) | 1,605 (88.09) | 1.41 (1.13–1.75) | 1.41 (1.14–1.75) | |
| T3 | 247 (13.55) | 1,576 (86.45) | 1.63 (1.32–2.01) | 1.65 (1.34–2.04) | |
| Ptrend | <0.001 | <0.001 | <0.001 |
Values are n (%) unless otherwise indicated. Pattern 1: essential elements factor; Pattern 2: nutritional-metabolic regulatory factor; Pattern 3: mineral metabolism balance factor; Pattern 4: lead exposure-nutritional antagonism factor. CI, confidence interval; OR, odds ratio; Ref., reference.
Sensitivity analysis for unmeasured confounding
E-values were computed to evaluate the robustness of the key associations to unmeasured confounding (Table S1). For the association between Pattern 4 and behavioral problems (T3 vs. T1: OR =1.65, 95% CI: 1.34–2.04), the E-value was 2.69 for the point estimate and 2.01 for the lower confidence limit; that is, an unmeasured confounder would need to be associated with both the pattern score and behavioral problems by a risk ratio of at least 2.69 each, above and beyond the measured covariates, to fully explain away the observed association. The E-values were 3.95 (lower confidence limit: 2.39) for blood lead, 2.94 (1.43) for TGs, and 4.57 for the lower confidence limit of the protective association of serum iron.
Discussion
This study leveraged a large-scale pediatric health examination dataset of 5,468 children to systematically investigate the associations of individual nutritional and metabolic biomarkers, as well as their integrated patterns, with behavioral problems. By combining multivariable regression, explainable machine learning (SHAP), and PCA in a progressive analytical framework, we identified a specific risk pattern—characterized by elevated blood lead co-occurring with reduced serum calcium and vitamin D (Pattern 4: lead exposure–nutritional antagonism)—that exhibited a robust, domain-general dose-response relationship with behavioral outcomes. By adopting a progressive analytical approach, moving from individual biomarkers to nutritional metabolic patterns, this study has provided a more physiologically accurate reflection of a child’s internal environment. These results contribute to the growing body of literature linking environmental exposures and nutritional status to neurodevelopmental outcomes in children, with important implications for both clinical and public health practice.
Our study identified blood lead, serum iron, and TG as the top three robust predictors of behavioral problems, as they were consistently confirmed by both multivariable logistic regression and SHAP analysis. Among these, lead exposure serves as the most prominent independent predictor within the discovered pattern, and it is noteworthy that the associations were observed at levels far below those that are considered clinically safe, further supporting the fact that there is no safe level of lead exposure in neurodevelopmental processes (29). Mechanistically, lead exposure affects synaptic plasticity and dopaminergic signaling, which are vital for regulating behavior development (30). In contrast, iron levels showed a strong protective effect, with higher levels significantly reducing the risk of behavioral problems. This can be explained by the important role of iron in maintaining myelination and neurotransmitter metabolism (31,32). We also found that higher TG level was significantly associated with higher risk of behavioral problems, which may be explained by the emerging concept of the metabolic-neurodevelopmental axis. Elevated TG levels may serve as a marker of systemic metabolic dysregulation and related inflammatory processes, potentially impacting energy homeostasis in the brain and consequently influencing neurobehavioral development (16,17). However, because our design was cross-sectional, we cannot determine whether TGs themselves exert a direct neurotoxic effect or simply serve as a proxy for broader dietary and lifestyle patterns, such as high intake of saturated fats.
Interestingly, we found that vitamin D was identified as a significant associated factor for behavioral problems in the SHAP analysis, which was not captured by the multivariable logistic regression. This discrepancy may be due to the non-linear or threshold-dependent nature of the association between vitamin D and neurodevelopment. Unlike traditional logistic regression, which is heavily dependent on global linear assumptions, SHAP excels at dealing with complex, non-linear relationships (33). Secondly, only marginal significance at the population level was observed regarding the prevalence of vitamin D insufficiency and deficiency between children with and without behavioral problems (19.71% vs. 18.16%, P=0.07), such subtle group-level differences that might not have been easily captured in the traditional logistic regression approach. On the other hand, SHAP assesses the value of a variable based on its marginal contribution to the model’s predictive power on the individual basis (34,35). This again highlights SHAP’s better sensitivity in identifying clinically significant biomarkers that may not be detected by traditional statistical approaches. From the biological perspectives, vitamin D is essential for neurogenesis, neuronal differentiation, and the regulation of neurotrophic factors such as brain-derived neurotrophic factor (BDNF). Given its role in modulating neuroinflammation and oxidative stress, our findings also supported that vitamin D status may be an important determinant of behavioral health in children (36). Furthermore, in contrast to our team’s earlier findings, which indicated a notably higher prevalence of vitamin D insufficiency/deficiency (approximately 50%) in children over the age of six (12), the current significant reduction may be ascribed to enhanced public health awareness and increased parental engagement in supplementation efforts in recent years.
Our study has identified a specific risk profile of “lead exposure-nutritional antagonism” in predicting behavioral problems in children. From a pathophysiological perspective, this specific combination of risk factors can be viewed as a “triple threat” to the developing nervous system. The molecular basis of this pattern is grounded in a well-characterized physiological pathway that links intestinal calcium absorption, vitamin D signaling, and lead uptake. At the core of this framework lies the phenomenon of ionic mimicry, wherein Pb2+, like Ca2+, acts as a divalent cation surrogate at critical intracellular signaling targets—including calmodulin and protein kinase C (PKC)—thereby disrupting calcium-dependent neurotransmitter release and synaptic plasticity (37,38). Under physiological conditions, this calcium-mediated signaling is tightly regulated by active vitamin D [1,25(OH)2D3], which, upon binding to the vitamin D receptor (VDR), transcriptionally upregulates calcium-selective epithelial channels such as transient receptor potential vanilloid 6 (TRPV6) in the intestine and transient receptor potential vanilloid 5 (TRPV5) in the renal distal tubule, along with the cytosolic calcium-binding protein calbindin-D9k (CaBP) (14,39). This coordinated system ensures highly selective and regulated calcium absorption, effectively limiting the entry of competing divalent metals, with the intestinal epithelium serving as the primary site for both physiological calcium uptake and pathological lead absorption. However, when vitamin D and calcium availability become insufficient, VDR-mediated expression of TRPV6 and CaBP is downregulated. In response to this state of relative calcium insufficiency, the body initiates a compensatory upregulation of the less selective divalent metal transporter 1 (DMT1/SLC11A2) in the intestinal epithelium, which preferentially transports iron but also transports Pb2+ (40-42). Consequently, this compensatory mechanism inadvertently increases intestinal lead absorption, providing the molecular basis for the nutritional “antagonism” captured by PCA Pattern 4. Once absorbed, lead crosses the blood-brain barrier, affects neural stem and progenitor cells, and promotes oxidative stress through mitochondrial dysfunction and lipid peroxidation, ultimately contributing to the neurodevelopmental vulnerabilities observed in affected children. Experimental evidence indicates that early-life lead exposure impairs hippocampal neurogenesis and alters neuronal morphology (43), while disrupting dopaminergic projections from the ventral tegmental area to the prefrontal cortex and interfering with glutamatergic signaling in the hippocampus (44,45), with downstream effects on executive function, impulse control, attention regulation, and emotional processing—the behavioral phenotypes captured by the CPRS in our study. The significant dose-response gradient across Pattern 4 tertiles observed in our study (Ptrend <0.001) is consistent with this mechanistic chain, lending biological plausibility to the Pattern 4-behavior association.
In addition, the present study identified other three PCs. More specifically, Pattern 1, marked by the co-loading of magnesium, iron, zinc, and copper, signifies a reserve of essential elements crucial for neurodevelopment (46). Iron and zinc are vital for dopamine synthesis and maintaining neural balance (11,47), magnesium is a key cofactor and NMDA antagonist (48,49), and copper aids antioxidant defense (50). Lower Pattern 1 scores in children with behavioral issues imply that a deficiency in these elements may undermine the biological foundation necessary for behavioral control. Pattern 2, characterized by high vitamin D and copper levels with low TGs, indicates a metabolic regulatory axis (51). This pattern suggests a synergistic metabolic profile, as vitamin D is linked to lipid metabolism regulation (52). In children, optimal vitamin D and TG levels often signal good metabolic health. Copper’s presence suggests antioxidant support (50). As we did not observe statistically significant differences in Pattern 2 between children with and without behavioral problems, Pattern 2 likely reflects a stable metabolic state across the pediatric cohort, regardless of behavior. Pattern 3 shows negative loadings for phosphorus and calcium and a positive loading for zinc, indicating a state of mineral metabolism homeostasis. While phosphorus and calcium are crucial for bone health and cellular signaling (53), calcium also regulates neurotransmitter release (54). The significantly lower Pattern 3 scores in children with behavioral problems suggest a potential decrease of mineral reserves that could affect neurochemical stability. The inverse relationship of zinc in this pattern highlights a complex metabolic interaction that needs further exploration to understand its impact on behavior.
This study has substantial implications for clinical and public health practice. Firstly, it advocates for a paradigm shift from isolated nutrient assessments to a pattern-based holistic screening approach, highlighting the synergistic interaction between nutritional buffers, such as calcium and vitamin D, and the burden of environmental toxicants. Secondly, it reveals the hidden risk associated with subclinical lead exposure in relatively “healthy” children, demonstrating that blood lead levels within conventional reference ranges can result in significant behavioral issues when combined with reduced calcium and vitamin D levels. This finding highlights the absence of a safe threshold for lead exposure. Thirdly, the concept of nutritional “buffering” emerges as a cost-effective primary prevention strategy, wherein sufficient reserves of calcium and vitamin D act as a biological barrier against lead absorption and its neurotoxic effects. Lastly, the utilization of explainable machine learning and dimensionality reduction techniques (e.g., PCA) provides a methodological basis for the development of clinical decision support systems, facilitating precision interventions in large-scale health screening programs.
Strengths and limitations
Our study leveraged a substantial dataset of 5,468 children who underwent health check-ups in a tertiary children’s hospital in East China. This large-scale cohort provides a relatively high statistical power to detect subtle associations, particularly in the context of routine health screenings. A key strength lies in our integrated analytical approach framework, which transcends traditional “single-nutrient” models. By sequentially employing multivariable regression, SHAP, and PCA, we successfully addressed the challenges of multicollinearity and non-linear interactions among nutritional and toxicant biomarkers. This methodology offers a more ecologically valid representation of a child’s complex internal metabolic environment. Our findings translate complex laboratory parameters into identifiable nutritional-metabolic patterns, such as the “lead exposure-nutritional antagonism” pattern. These integrated indicators provide pediatricians and public health practitioners with actionable screening tools that are more clinically intuitive than isolated biomarkers, facilitating early identification and holistic intervention for children at risk for behavioral problems.
However, several limitations warrant consideration. First, participants were recruited from a Health Management Center rather than from the community. Families attending routine health check-ups may differ from the general pediatric population in health awareness and socioeconomic status, which may limit the generalizability of our prevalence estimates and observed associations. Second, given the cross-sectional nature of the study, causality cannot be concluded. While we identified significant associations between nutritional metabolic patterns and behavioral problems, longitudinal studies are required to confirm the temporal sequence and the long-term impact of these metabolic imbalances on neurodevelopment. Third, the use of a database from the Health Management Center meant that certain potential confounders, such as parental educational level, household income, dietary habits, and physical activity levels, were not fully captured. More importantly, our dataset did not capture major psychosocial determinants of childhood behavioral outcomes—including family structure, parenting practices, adverse childhood experiences, peer relationships, and school-related stressors. Because such disadvantage is plausibly associated with both less favorable nutritional–metabolic profiles (e.g., socioeconomic gradients in lead exposure and diet quality) and more behavioral problems, residual confounding is likely and would generally be expected to bias the observed associations away from the null. Accordingly, our findings should be interpreted as associations rather than causal effects, and the identified patterns are best understood as complementary biological screening information rather than independent determinants of behavior. Future research incorporating detailed lifestyle, socioeconomic and psychological data in a well-designed cohort study would provide a more comprehensive adjustment for these factors. Additionally, the CPRS data in our dataset were binary, not continuous scores; this constrains our ability to detect graded dose-response relationships and may reduce statistical sensitivity for certain comparisons. While the CPRS is a widely recognized screening instrument, it provides a dimensional assessment of behavior rather than a clinical diagnosis. Future studies incorporating standardized clinical evaluations for neurodevelopmental disorders would provide greater diagnostic specificity. Furthermore, because the CPRS does not provide validated broadband scores for internalizing and externalizing domains, our analysis targeted overall behavioral problems as an aggregate screening outcome defined by the Chinese CPRS norms, and did not differentiate internalizing from externalizing subtypes; deriving ad-hoc composites would lack psychometric validity. Consequently, the present study cannot address domain-specific mechanisms. Future studies using instruments with validated broadband scoring [e.g., the Child Behavior Checklist (CBCL)] or standardized clinical interviews are warranted to examine whether the nutritional-metabolic patterns identified here exhibit domain-specific associations and to explore the underlying mechanisms.
Conclusions
By leveraging a large-scale pediatric database and integrating multifactorial analytical methodologies, this study systematically reveals the nutritional and metabolic factors as well as their integrated patterns associated with behavioral problems in children. We identified a critical “lead exposure-nutritional antagonism” pattern, characterized by elevated lead alongside insufficient calcium and vitamin D, that exhibits a significant dose-response relationship with neurobehavioral risk. These findings advocate for a paradigm shift from single-nutrient assessments toward holistic, pattern-based metabolic screening in child health management. Early identification of such integrated risk profiles may offer a window for targeted nutritional optimization, pending confirmation in longitudinal studies with comprehensive psychosocial adjustment.
Supplementary
The article’s supplementary files as
Acknowledgments
None.
Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Medical Ethics Committee of the Children’s Hospital, Zhejiang University School of Medicine (No. 2021-IRB-185). Waivers for informed consent were obtained since the study was a retrospective analysis of anonymized electronic health records without risk to the participants.
Footnotes
Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0370/rc
Funding: None.
Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0370/coif). The authors have no conflicts of interest to declare.
Data Sharing Statement
Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0370/dss
References
- 1. Polanczyk GV, Salum GA, Sugaya LS, et al. Annual research review: A meta-analysis of the worldwide prevalence of mental disorders in children and adolescents. J Child Psychol Psychiatry 2015;56:345-65.. 10.1111/jcpp.12381 [DOI] [PubMed] [Google Scholar]
- 2. Ayano G, Demelash S, Gizachew Y, et al. The global prevalence of attention deficit hyperactivity disorder in children and adolescents: An umbrella review of meta-analyses. J Affect Disord 2023;339:860-6.. 10.1016/j.jad.2023.07.071 [DOI] [PubMed] [Google Scholar]
- 3. Erskine HE, Norman RE, Ferrari AJ, et al. Long-Term Outcomes of Attention-Deficit/Hyperactivity Disorder and Conduct Disorder: A Systematic Review and Meta-Analysis. J Am Acad Child Adolesc Psychiatry 2016;55:841-50.. 10.1016/j.jaac.2016.06.016 [DOI] [PubMed] [Google Scholar]
- 4. Bordelois P, Koenen KC, Elkind MSV, et al. Childhood internalizing and externalizing problems and cardiovascular and diabetes mellitus risk in adolescence. J Affect Disord 2023;335:239-47.. 10.1016/j.jad.2023.04.084 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Lanphear BP, Hornung R, Khoury J, et al. Low-level environmental lead exposure and children's intellectual function: an international pooled analysis. Environ Health Perspect 2005;113:894-9.. 10.1289/ehp.7688 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Sears CG, Lanphear BP, Xu Y, et al. Identifying periods of heightened susceptibility to lead exposure in relation to behavioral problems. J Expo Sci Environ Epidemiol 2022;32:1-9.. 10.1038/s41370-021-00389-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Liu J, Portnoy J, Raine A, et al. Blood lead levels mediate the relationship between social adversity and child externalizing behavior. Environ Res 2022;204:112396.. 10.1016/j.envres.2021.112396 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Wang Y, Huang L, Zhang L, et al. Iron Status in Attention-Deficit/Hyperactivity Disorder: A Systematic Review and Meta-Analysis. PLoS One 2017;12:e0169145.. 10.1371/journal.pone.0169145 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Doom JR, Richards B, Caballero G, et al. Infant Iron Deficiency and Iron Supplementation Predict Adolescent Internalizing, Externalizing, and Social Problems. J Pediatr 2018;195:199-205.e2.. 10.1016/j.jpeds.2017.12.008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Wang B, Fang T, Chen H. Zinc and Central Nervous System Disorders. Nutrients 2023;15:2140.. 10.3390/nu15092140 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Frederickson CJ, Koh JY, Bush AI. The neurobiology of zinc in health and disease. Nat Rev Neurosci 2005;6:449-62.. 10.1038/nrn1671 [DOI] [PubMed] [Google Scholar]
- 12. Shen Y, Gao M, Guan L, et al. Association of Vitamin D and Weight Status With Neurodevelopmental Outcomes in a Large Pediatric Population: Cross-Sectional Study. JMIR Public Health Surveill 2026;12:e89756.. 10.2196/89756 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Hu J, Jia J, Zhang Y, et al. Effects of vitamin D3 supplementation on cognition and blood lipids: a 12-month randomised, double-blind, placebo-controlled trial. J Neurol Neurosurg Psychiatry 2018;89:1341-7.. 10.1136/jnnp-2018-318594 [DOI] [PubMed] [Google Scholar]
- 14. Christakos S, Dhawan P, Verstuyf A, et al. Vitamin D: Metabolism, Molecular Mechanism of Action, and Pleiotropic Effects. Physiol Rev 2016;96:365-408.. 10.1152/physrev.00014.2015 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Naiberg MR, Newton DF, Collins JE, et al. Elevated triglycerides are associated with decreased executive function among adolescents with bipolar disorder. Acta Psychiatr Scand 2016;134:241-8.. 10.1111/acps.12603 [DOI] [PubMed] [Google Scholar]
- 16. Dimache AM, Șalaru DL, Sascău R, et al. The Role of High Triglycerides Level in Predicting Cognitive Impairment: A Review of Current Evidence. Nutrients 2021;13:2118.. 10.3390/nu13062118 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Tiwari V, Simons M. Lipid metabolism and neuroinflammation: What is the link?. J Exp Med 2025;222:e20241232.. 10.1084/jem.20241232 [DOI] [PubMed] [Google Scholar]
- 18. Hu FB. Dietary pattern analysis: a new direction in nutritional epidemiology. Curr Opin Lipidol 2002;13:3-9.. 10.1097/00041433-200202000-00002 [DOI] [PubMed] [Google Scholar]
- 19.Willett W. Nutritional epidemiology. 3rd edn. Oxford: Oxford University Press; 2012. [Google Scholar]
- 20. Jacques PF, Tucker KL. Are dietary patterns useful for understanding the role of diet in chronic disease?. Am J Clin Nutr 2001;73:1-2.. 10.1093/ajcn/73.1.1 [DOI] [PubMed] [Google Scholar]
- 21. Schulz CA, Oluwagbemigun K, Nöthlings U. Advances in dietary pattern analysis in nutritional epidemiology. Eur J Nutr 2021;60:4115-30.. 10.1007/s00394-021-02545-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Lundberg SM, Lee SI. A unified approach to interpreting model predictions. Proceedings of the 31st International Conference on Neural Information Processing Systems 2017;30:4768-77. [Google Scholar]
- 23. Hoffmann K, Schulze MB, Schienkiewitz A, et al. Application of a new statistical method to derive dietary patterns in nutritional epidemiology. Am J Epidemiol 2004;159:935-44.. 10.1093/aje/kwh134 [DOI] [PubMed] [Google Scholar]
- 24. Newby PK, Tucker KL. Empirically derived eating patterns using factor or cluster analysis: a review. Nutr Rev 2004;62:177-203.. 10.1301/nr.2004.may.177-203 [DOI] [PubMed] [Google Scholar]
- 25. World Medical Association. . World Medical Association Declaration of Helsinki. Ethical principles for medical research involving human subjects. Bull World Health Organ 2001;79:373-4. [PMC free article] [PubMed] [Google Scholar]
- 26. Conners CK, Sitarenios G, Parker JD, et al. The revised Conners' Parent Rating Scale (CPRS-R): factor structure, reliability, and criterion validity. J Abnorm Child Psychol 1998;26:257-68.. 10.1023/a:1022602400621 [DOI] [PubMed] [Google Scholar]
- 27. . [Google Scholar]
- 28. VanderWeele TJ, Ding P. Sensitivity Analysis in Observational Research: Introducing the E-Value. Ann Intern Med 2017;167:268-74.. 10.7326/M16-2607 [DOI] [PubMed] [Google Scholar]
- 29. Vorvolakos T, Arseniou S, Samakouri M. There is no safe threshold for lead exposure: Α literature review. Psychiatriki 2016;27:204-14.. 10.22365/jpsych.2016.273.204 [DOI] [PubMed] [Google Scholar]
- 30. Ramírez Ortega D, González Esquivel DF, Blanco Ayala T, et al. Cognitive Impairment Induced by Lead Exposure during Lifespan: Mechanisms of Lead Neurotoxicity. Toxics 2021;9:23.. 10.3390/toxics9020023 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Kulaszyńska M, Kwiatkowski S, Skonieczna-Żydecka K. The Iron Metabolism with a Specific Focus on the Functioning of the Nervous System. Biomedicines 2024;12:595.. 10.3390/biomedicines12030595 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Lozoff B. Early iron deficiency has brain and behavior effects consistent with dopaminergic dysfunction. J Nutr 2011;141:740S-6S.. 10.3945/jn.110.131169 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Zhou W, Yan Z, Zhang L. A comparative study of 11 non-linear regression models highlighting autoencoder, DBN, and SVR, enhanced by SHAP importance analysis in soybean branching prediction. Sci Rep 2024;14:5905.. 10.1038/s41598-024-55243-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Ponce-Bobadilla AV, Schmitt V, Maier CS, et al. Practical guide to SHAP analysis: Explaining supervised machine learning model predictions in drug development. Clin Transl Sci 2024;17:e70056.. 10.1111/cts.70056 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Prendin F, Pavan J, Cappon G, et al. The importance of interpreting machine learning models for blood glucose prediction in diabetes: an analysis using SHAP. Sci Rep 2023;13:16865.. 10.1038/s41598-023-44155-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Menéndez SG, Manucha W. Vitamin D as a Modulator of Neuroinflammation: Implications for Brain Health. Curr Pharm Des 2024;30:323-32.. 10.2174/0113816128281314231219113942 [DOI] [PubMed] [Google Scholar]
- 37. Bressler J, Kim KA, Chakraborti T, et al. Molecular mechanisms of lead neurotoxicity. Neurochem Res 1999;24:595-600.. 10.1023/a:1022596115897 [DOI] [PubMed] [Google Scholar]
- 38. Marchetti C. Role of calcium channels in heavy metal toxicity. ISRN Toxicol 2013;2013:184360.. 10.1155/2013/184360 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Hoenderop JG, Nilius B, Bindels RJ. Calcium absorption across epithelia. Physiol Rev 2005;85:373-422.. 10.1152/physrev.00003.2004 [DOI] [PubMed] [Google Scholar]
- 40. Bannon DI, Abounader R, Lees PS, et al. Effect of DMT1 knockdown on iron, cadmium, and lead uptake in Caco-2 cells. Am J Physiol Cell Physiol 2003;284:C44-50.. 10.1152/ajpcell.00184.2002 [DOI] [PubMed] [Google Scholar]
- 41. Kwong WT, Friello P, Semba RD. Interactions between iron deficiency and lead poisoning: epidemiology and pathogenesis. Sci Total Environ 2004;330:21-37.. 10.1016/j.scitotenv.2004.03.017 [DOI] [PubMed] [Google Scholar]
- 42. I Bannon D, Portnoy ME, Olivi L, et al. Uptake of lead and iron by divalent metal transporter 1 in yeast and mammalian cells. Biochem Biophys Res Commun 2002;295:978-84.. 10.1016/s0006-291x(02)00756-8 [DOI] [PubMed] [Google Scholar]
- 43. Verina T, Rohde CA, Guilarte TR. Environmental lead exposure during early life alters granule cell neurogenesis and morphology in the hippocampus of young adult rats. Neuroscience 2007;145:1037-47.. 10.1016/j.neuroscience.2006.12.040 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Neal AP, Guilarte TR. Molecular neurobiology of lead (Pb(2+)): effects on synaptic function. Mol Neurobiol 2010;42:151-60.. 10.1007/s12035-010-8146-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 45. Mason LH, Harp JP, Han DY. Pb neurotoxicity: neuropsychological effects of lead toxicity. Biomed Res Int 2014;2014:840547.. 10.1155/2014/840547 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Yang R, Zhang Y, Gao W, et al. Blood Levels of Trace Elements in Children with Attention-Deficit Hyperactivity Disorder: Results from a Case-Control Study. Biol Trace Elem Res 2019;187:376-82.. 10.1007/s12011-018-1408-9 [DOI] [PubMed] [Google Scholar]
- 47. Beard JL, Connor JR. Iron status and neural functioning. Annu Rev Nutr 2003;23:41-58.. 10.1146/annurev.nutr.23.020102.075739 [DOI] [PubMed] [Google Scholar]
- 48. Sacaan AI, Johnson KM. Regulation of polyamine and magnesium inhibition of the N-methyl-D-aspartate (NMDA) receptor ionophore complex. Neurosci Lett 1991;121:219-22.. 10.1016/0304-3940(91)90689-q [DOI] [PubMed] [Google Scholar]
- 49. Boyle NB, Lawton C, Dye L. The Effects of Magnesium Supplementation on Subjective Anxiety and Stress-A Systematic Review. Nutrients 2017;9:429.. 10.3390/nu9050429 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50. Gaetke LM, Chow CK. Copper toxicity, oxidative stress, and antioxidant nutrients. Toxicology 2003;189:147-63.. 10.1016/s0300-483x(03)00159-8 [DOI] [PubMed] [Google Scholar]
- 51. Blades B, Ayton S, Hung YH, et al. Copper and lipid metabolism: A reciprocal relationship. Biochim Biophys Acta Gen Subj 2021;1865:129979.. 10.1016/j.bbagen.2021.129979 [DOI] [PubMed] [Google Scholar]
- 52. Asano L, Watanabe M, Ryoden Y, et al. Vitamin D Metabolite, 25-Hydroxyvitamin D, Regulates Lipid Metabolism by Inducing Degradation of SREBP/SCAP. Cell Chem Biol 2017;24:207-17.. 10.1016/j.chembiol.2016.12.017 [DOI] [PubMed] [Google Scholar]
- 53. Couce ML, Saenz de Pipaon M. Bone Mineralization and Calcium Phosphorus Metabolism. Nutrients 2021;13:3692.. 10.3390/nu13113692 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 54. Arjun McKinney A, Petrova R, Panagiotakos G. Calcium and activity-dependent signaling in the developing cerebral cortex. Development 2022;149:dev198853.. 10.1242/dev.198853 [DOI] [PMC free article] [PubMed] [Google Scholar]