Introduction

Pediatric non-alcoholic fatty liver disease (NAFLD) is the most frequent form of chronic liver disease, with a prevalence of 7.6%, while it increases to 34.2% in obese children1. Mtabolic associated fatty liver disease (MAFLD) and pediatric fatty liver disease (PeFLD) have been proposed to replace NAFLD in children, but are not yet internationally recognized, so NALFD is still used in this study. Diagnosis of NAFLD in children is based on biomarkers, liver imaging (such as liver ultrasound, and magnetic resonance), or liver biopsy2. NAFLD comprises various fatty liver diseases without excessive alcohol intake3. It is a multisystem disease that affects extrahepatic organs, and its long-term health effects can extend into adulthood and lead to significant morbidity and mortality4. Some studies have shown that children with NAFLD are at a higher risk of developing cardiovascular diseases and metabolic disorders, which subsequently impact their quality of life5,6. The increased risks of intrahepatic and extrahepatic adverse outcomes attributed to NAFLD result in huge economic and clinical burden to both individuals and society7. Approved medications are not available for NAFLD, while physical exercise and diet control are the major recommended interventions8. Therefore, identifying the risk factors, especially modifiable risk factors is important for carrying out targeted interventions to reduce the disease burden of pediatric NAFLD.

There are 35 commonly encountered exposure metals, of which 23 are classified as heavy metals9. Heavy metals are metallic elements with high atomic weight and high density, which have adverse effects on human health since they are non-biodegradable and deposited in the body10. Humans are exposed to heavy metals mainly through contaminated air and food, domestic effluents, agriculture, and industrial waste11. The endoplasmic reticulum stress, hypothalamic dopaminergic system, impaired adipogenesis, and adipocytokines secretion can be disrupted by heavy metals, and hence liver inflammation and steatosis occur12.

Smoking is also a common source of heavy metal exposure, affecting the levels of various heavy metals in the human body. Smokers have significantly higher levels of cadmium, chromium, and lead in their urine and serum compared to non-smokers13,14. Elżbieta Kulikowska-Karpińska et al. found that smoking significantly influences the level of copper in the urine15. Heavy metal exposure due to smoking is associated with several diseases, such as nasal and sinus issues16, as well as periodontitis17. In addition, smoking is the leading preventable risk factor for cirrhosis, hepatitis B and C, and hepatocellular carcinoma, while the role of smoking in NAFLD remains controversial18,19. Liu et al. showed a positive association between smoking and NAFLD, while no association was observed in the others20,21,22. In America, children are the primary victims of secondhand smoke exposure (SHS), which has severe negative impacts on their health. Research has shown that secondhand smoke can lead to dental caries in children23 and attention deficit hyperactivity disorder24. This serious problem has attracted the attention of the American Academy of Pediatrics and the American Medical Association, and urged pediatricians and medical professionals to assist parents in protecting their children from SHS25,26.

Nevertheless, little is known about the relationship between blood heavy metals and pediatric NAFLD, especially in the case of smoking exposure. Therefore, we first explored the relationship between NAFLD and blood heavy metals in children whose information was collected from the National Health and Nutrition Examination Survey (NHANES). Our study showed no significantly different distribution of blood lead, cadmium, mercury, and selenium in groups of NAFLD and control; however, the disease group exhibited higher blood manganese (Mn) levels. The inconsistent findings27,28,29 triggered us to further explore the association of Mn with NAFLD the previous research. Then, we continued to explore the relationship between the two under smoking exposure. Notably, self-reported smoking status might have bias and serum cotinine level is an indicator reflecting both direct and indirect smoking quantity30. Thus, we assessed the modification role of cotinine exposure in the association of Mn with pediatric NAFLD.

Materials and methods

Data source and study population

NHANES (https://www.cdc.gov/nchs/nhanes/index.htm) is a database to assess the health and nutritional status of children and adults in the United States performed by the National Center for Health Statistics (NCHS). The database includes information on demographics, examination, laboratory, dietary, and questionnaires from 1999 to 2018. The research received approval from the NCHS Research Ethics Review Board and participants signed informed consent forms.

A total of 9254 samples from the 2017 to 2018 cycle were obtained. Exclusion criteria: age > 18 years since children are persons up to the age of 18 [United Nations Youth. [(accessed on 1 January 2020)];2020 Available online: https://www.un.org/en/sections/issues-depth/youth-0/index.html]; ever having over 4/5 or more drinks every day; serologic positivity for viral hepatitis; missing data of cotinine levels; missing information on heavy metals. Therefore, 866 remaining participants were enrolled in the study: non-NAFLD group (n = 685) and NAFLD group (n = 181). Figure 1 shows the screening process of the study population.

Fig. 1
figure 1

Flow charts of the inclusion and exclusion of participants.

Full size image

Diagnosis of pediatric NAFLD

Based on previous research31, NAFLD in children can be diagnosed by waist circumference, body mass index, or fasting plasma glucose level > 100 mg/dL. In lean children, the presence of at least following two metabolic risk disturbances is also defined as NAFLD: high-density lipoprotein cholesterol (HDL-C) < 40 mg/dL, triglycerides ≥ 150 mg/dL, C-reactive protein level > 2 mg/L, HOMA-IR ≥ 2.5, diastolic blood pressure ≥ 85 or systolic blood pressure ≥ 130 mmHg.

Blood heavy metal measurements

The concentrations of blood lead, cadmium, mercury total, selenium, and Mn were determined by inductively coupled plasma mass spectrometry. All QC procedures recommended by the manufacturers were followed32. Limits of detection for blood lead, cadmium, mercury total, selenium, and Mn were 0.07, 0.1, 0.28, 24.48, and 0.99, respectively.

Serum cotinine levels

Serum cotinine is measured by an isotope dilution, high-performance liquid chromatography, and tandem mass spectrometry method (https://wwwn.cdc.gov/Nchs/Nhanes/2017-2018/COT_J.htm). The limit of detection of serum cotinine is 0.015.

Covariates

Based on the previous research and our interest, information on gender, race, age, education, poverty income ratio (PIR), waist circumference, total bilirubin, anion gap (AG), albumin, alanine transaminase (ALT), aspartate aminotransferase (AST), blood urea nitrogen (BUN), creatinine, and HDL-C, and triglycerides were collected. Gender (male and female), race (Hispanic, non-Hispanic White, non-Hispanic Black, and other races), age (≤ 18 years), education level (less than 9th grade and over 9th grade), and PIR were collected from the “Demographic variables and sample weights” file. Waist circumference data were acquired from the “Body measures” file in the “Examination data”. We retrieved data on total bilirubin, AG, albumin, ALT, AST, BUN, creatinine, HDL-C, and triglycerides from the “Laboratory data”. Detailed information is accessible from the official website (https://www.cdc.gov/nchs/nhanes/index.htm).

Statistical analysis

Baseline characteristics of the study population in the non-NAFLD and NAFLD groups were compared using the Chi-square test and Mann-Whitney U test for the categorical and continuous variables, respectively. The categorical data were expressed as counts (percentages), and the continuous data were represented as median [interquartile range] due to the skewed distribution. To avoid potential bias due to a confounding effect, we performed an analysis of collinearity in which the variance inflation factors (VIF) were below 3. Next, restricted cubic spline (RCS) combined with logistic regression models were employed for the exploration of the relationship between blood heavy metal levels as a continuous variable and NAFLD. Mode l was the crude model. Model 2 was adjusted for the demographic characteristics and body measures including age, race, PIR, and waist circumference. Model 3 was additionally corrected for the total bilirubin, AG, albumin, ALT, AST, HDL-C, and triglycerides except for the variables in Model 2. For sensitivity analysis, the heavy metal level was coded into three groups by tertiles to assess its association with NAFLD. Besides, the P for trend test was conducted to observe the trend of NAFLD risk across increasing blood heavy metal levels.

Then, an RCS analysis stratified by cotinine exposure was evaluated to investigate the modification of cotinine exposure in the relationship between blood heavy metals and NAFLD. The same three models were also adopted to analyze the association of heavy metals both as a continuous and categorical variable with NAFLD in two cotinine groups (unexposed and exposed groups). Finally, the relationship between the heavy metal and NAFLD was evaluated in different subgroups. P < 0.05 indicated statistical significance.

Results

Baseline characteristics

The characteristics of the participants are demonstrated in Table 1. The characteristics of the study population across three blood Mn groups are shown in Table S1. Older age, lower PIR, and higher waist circumference were observed in the NAFLD group with statistical significance (all P < 0.05). Total bilirubin, albumin, and HDL-C were decreased, while AG, ALT, AST, and triglycerides were increased in the NAFLD compared to those in the non-NAFLD (all P < 0.05). Particiapnts with NAFLD are more likely to be Hispanics. There was no remarkable difference in gender, education, cotinine concentration, BUN, and creatinine between the two groups. Moreover, except for blood Mn with a significant elevation in the NAFLD groups (P < 0.05), the other four blood heavy metals including blood lead, cadmium, mercury total, and selenium were not differentially distributed in the two groups (P > 0.05). Therefore, we focused on the association of blood Mn with NAFLD. There was no collinearity since the VIFs of all the significant variables were less than 3 (Table 2).

Table 1 Baseline characteristics of the study participants.
Full size table
Table 2 Collinearity analysis.
Full size table

The association of blood mn level and NAFLD

First, RCS analysis showed a linear relationship between blood Mn level and NAFLD both in the crude model (P for nonlinear = 0.078) and adjusted model (P for nonlinear = 0.677) (Fig. 2A and B). The blood Mn level as a continuous variable was significantly related to NAFLD in three different models (P < 0.05) (Table 3). The models were adjusted in an additive way, starting from no covariate-adjusted model 1, adding for age, race, PIR, and waist circumference (model 2), total bilirubin, AG, albumin, ALT, AST, triglycerides, and HDL-C (model 3). In model 1, participants with higher blood Mn levels were 1.035 times more likely to be NAFLD [95% CI: 1.001–1.069]. Age, race, PIR, and waist circumference led to an increase in the odds ratio (OR) of NAFLD (OR [95% CI]: 1.051 [1.003, 1.101]). Laboratory variables enrolled in model 3 further increased the odds of NAFLD (OR [95% CI]: 1.071 [1.004, 1.142]). A similar trend was observed when blood Mn was taken as a categorical variable (P < 0.05). The positive relationship between blood Mn level and NAFLD was further confirmed by the trend test (P for trend < 0.05).

Fig. 2
figure 2

Restricted cubic spline analysis of the association of blood manganese level and nonalcoholic fatty liver disease. (A) A crude model with no covariate-adjusted. (B) An adjusted model with race, age, poverty income ratio, waist circumference, total bilirubin, anion gap, albumin, alanine transaminase, aspartate aminotransferase, high-density lipoprotein cholesterol, and triglycerides adjusted.

Full size image
Table 3 The association of blood mn levels with NAFLD.
Full size table

Cotinine exposure enhanced the association of blood mn level with NAFLD

Although we didn’t find a connection between cotinine concentration and NAFLD (Table 1), there was an obvious association of blood Mn level with NAFLD in different cotinine groups through an RCS model (Fig. 3). Blood Mn level had a positive relation with NAFLD and this relationship strengthened with the exposure of cotinine.

Fig. 3
figure 3

The association of blood manganese and non-alcoholic fatty liver disease stratified by cotinine exposure group.

Full size image

The association of blood Mn level and NAFLD in two cotinine groups is exhibited in Table 4. In the cotinine unexposed group, the relationship between blood Mn and NAFLD was insignificant in three models (all P > 0.05). However, their relationship became significant in the cotinine-exposed group (all P < 0.05). When blood Mn was coded into three groups, the risk of NAFLD in the Tertile 3 group was positively related to blood Mn by taking the Tertile 1 groups as a reference. In the cotinine unexposed group, the OR of NAFLD in the tertile 3 group was 4.452 [95% CI: 1.283, 15.445] (P < 0.05), but this relationship was not statistically significant after controlling multiple variables in model 2 and model 3. In the cotinine-exposed group, the OR of NAFLD in the Tertile 3 group was higher in model 2 (OR [95% CI]: 2.410 [1.183, 4.911]) than the model 1 (OR [95% CI]: 1.823 [1.087, 3.057]). The association had a slight decrease when laboratory variables were added (OR [95% CI]: 2.236 [1.050, 4.762]). Whereas, blood Mn was notably linked to NAFLD independent of all the covariates in three models. These results indicated that the risk of NAFLD increased with blood Mn, which was further strengthened by the cotinine exposure.

Table 4 Cotinine exposure regulated the association of Mn with NAFLD.
Full size table

Subgroup analysis

To further evaluate the role of blood Mn in NAFLD, we performed a subgroup analysis according to gender, race, PIR, and education (Table 5). There was an obvious relationship between blood Mn and NAFLD in subgroups of non-Hispanic White, and less than 9th grade (all P < 0.05). Non-Hispanic White had the highest OR of NAFLD with 1.090 in contrast to the Hispanic, Non-Hispanic Black, and other races. In addition, the participants in the lower education level group were more likely to suffer from NAFLD with statistical significance (P < 0.05) (Table 5).

Table 5 Subgroup analysis of the relationship between manganese and NAFLD in different subgroups.
Full size table

Discussion

Emerging epidemiological evidence suggested a global surge in the prevalence of NAFLD among children, posing significant public health challenges2,33,34. Recent advancements in research have revealed that an accumulation of liver fat and steatosis occurred as a result of disruptions in lipid metabolism, insulin resistance, and the synergistic influence of genetic and environmental factors, ultimately fostering the progression of non-alcoholic steatohepatitis (NASH) and cirrhosis34. Additionally, a substantial body of statistics indicated that most NAFLD patients are afflicted with obesity, implying a potential association between obesity and the development of NAFLD35. Investigations have further demonstrated that obesity-related complications affect oxidative stress in children, stimulating the advancement of NAFLD through the modulation of pro-inflammatory and anti-inflammatory factors36,37. Animal experiments have substantiated that exposure of the liver to heightened levels of proinflammatory cytokines can induce histological changes that closely mimic NASH38. Consistent with earlier studies, our observations have revealed a significantly larger waist circumference in the NAFLD group compared to the non-NAFLD group.

Furthermore, our findings have unveiled a positive correlation between blood Mn levels and NAFLD. Mn is an essential trace element vital for human development39. However, excessive Mn levels may elevate the risk of immune toxicity across various vertebrate species40. An animal study has demonstrated that Mn injections promote inflammation in mice41. Moreover, Nakata et al. have asserted that high occupational exposure to Mn typically led to a decrease in T-lymphocytes among adult men42. The reduced presence of T regulatory cells (Tregs) has also been observed in animal models and humans with NAFLD43. On the other hand, Mn also plays a crucial role in metabolism. Elevated Mn levels may disrupt mitochondrial function by promoting the production of reactive oxygen species, reducing adenosine-triphosphate production, and modulating membrane permeability44, thereby contributing to oxidative stress and the progression of NAFLD45. These outcomes indicated that Mn exposure may facilitate the development of NAFLD through mediating immune responses and metabolism.

In addition, we also found that cotinine may enhance the association between Mn and NAFLD. Cotinine, a nicotine metabolite with an extended half-life, is frequently employed as an indicator for measuring smoking exposure22. Smoking increases the levels of heavy metals in a patient’s body. SHS also releases heavy metals into the air, including Mn, leading to heavy metal exposure in non-smokers and potentially raising their body levels of heavy metals46. Elevated blood Mn levels can lead to an increase in reactive oxygen species, inhibit adenosine triphosphate production, and alter membrane permeability, thereby disrupting normal mitochondrial function44. Mitochondrial dysfunction causes oxidative stress, which increases the levels of inflammatory cytokines in the body, triggering the “multiple hit” hypothesis and potentially leading to NAFLD8,34,47.

Nevertheless, this study possesses several limitations. Firstly, the data utilized is derived exclusively from the NHANES database, which primarily represents the American population. Secondly, the inherent limitation of NHANES renders no causality. Thirdly, some inflammatory markers are not recorded in the NHANES database, thereby preventing the inclusion of additional relevant indicators for a more comprehensive analysis. However, our study has several advantages. Firstly, the reliability and representativeness are bolstered by the utilization of large sample size and appropriate covariate adjustments. Secondly, the NAFLD diagnosis used in this study was specific to children, fostering credibility. This study paves the way for investigating the underlying mechanism of the modification of cotinine exposure on the relationship between blood Mn levels and pediatric NAFLD.

In conclusion, our findings establish a close and positive association between blood Mn levels and pediatric NAFLD, a relationship that can be intensified by cotinine exposure. However, the causal mechanisms and potential pathways necessitate further validation through subsequent studies.