Heavy metals and their relationships with lung function and airway inflammation: insights from a population-based study

Abstract

Background

Humans are commonly exposed to various heavy metals, but their effects on human respiratory health, especially lung function and airway inflammation, remain poorly understood.

Methods

This study included data from the 2011–2012 National Health and Nutrition Examination Survey (NHANES) and utilized multivariable linear regression, subgroup analyses, interaction tests, Bayesian kernel machine regression (BKMR), and weighted quantile sum (WQS) regression to explore the relationship between heavy metals, airway inflammation, and lung function.

Results

This study included 3576 adult participants. In the fully adjusted model, a positive relationship was observed between serum mercury (Hg) and fractional exhaled nitric oxide (FeNO) [0.20 (0.02, 0.37). Serum cadmium (Cd) had a significant negative connection with FEV1 [-106.22 (-143.64, -68.80)], FVC [-74.94 (-119.22, -30.66)], and FEV1/FVC [-1.35 (-1.82, -0.88)], serum lead (Pb) also showed a significant negative association with FEV₁ [-17.85 (-27.48, -8.21)], FVC [-14.84 (-26.22, -3.45)], and FEV₁/FVC [-0.14 (-0.26, -0.02)], while serum manganese (Mn) exhibited a significant positive relationship with FEV1/FVC [0.09 (0.02, 0.15)]. Selenium (Se) exposure showed a positive association with FEV₁ and FEV₁/FVC in Model 2, although these associations were not significant in the fully adjusted model. Subgroup analyses revealed that Body Mass Index (BMI) influenced the relationship between serum Hg, Mn, Se, Pb, and FeNO. BKMR analysis suggested a negative one-way exposure-response association among Cd exposure and FeNO, FEV1, and FEV1/FVC. The overall effect of co-exposure to the five heavy metals on FeNO levels was inhibitory. WQS analysis identified Cd exposure as the most significant negative associated factor on FeNO, FEV1, and FEV1/FVC. In contrast, Hg exposure was the most significant positive factor associated with FeNO, Se contributed the strongest positive weight to FEV₁ and FEV₁/FVC.

Conclusions

We found inconsistent associations between heavy metals, lung function, and airway inflammation. Cd and Pb exposure was associated with reduced lung function, whereas Hg exposure was positively associated with airway inflammation. Se contributed the strongest positive weight to FEV₁ and FEV₁/FVC in the WQS analysis, although Se was not significantly associated with FEV₁ and FEV₁/FVC in the fully adjusted model. These findings underscore the imperative for strengthened environmental regulation of heavy metal exposure. Future studies should employ more prospective methodologies to clarify these relationships.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-06538-8.

Introduction

Respiratory diseases include a range of acute and chronic diseases, with chronic respiratory diseases being the third most prevalent globally, accounting for approximately 7.00% of all deaths and causing a total of 112.32 million disability-adjusted life-years [1, 2]. The most common chronic respiratory diseases include asthma, chronic obstructive pulmonary disease (COPD), and chronic bronchitis, which contribute to the global burden of non-communicable diseases [3].

Early warning indicators of respiratory damage are necessary for effective respiratory illness prevention and control. Respiratory airway inflammation is widely recognized as the key pathophysiologic basis of respiratory injury [4]. Fractional exhaled nitric oxide (FeNO) is an important marker of inflammation in the respiratory airways, which can suggest respiratory health. In recent years, FeNO has served as a non-invasive biomarker of airway inflammation, primarily reflecting eosinophilic inflammation and oxidative stress in the respiratory system [5]. It is widely used in the assessment of asthma and other inflammatory airway diseases, aiding in disease monitoring and treatment response evaluation [6–8]. Lung function parameters, including forced expiratory volume in 1 (FEV1), forced vital capacity (FVC), and FEV1/FVC ratio, are critical indicators for assessing respiratory health and diagnosing conditions such as COPD and asthma [9, 10]. These parameters provide essential insights into airflow limitation, pulmonary compliance, and overall lung function status [11]. By integrating FeNO and lung function measures, a more comprehensive understanding of respiratory impairment can be achieved.

Environmental pollution by heavy metals is becoming an increasingly serious problem, and its adverse effects around the world are causing widespread concern [12]. Heavy metals are metals with a density greater than 5 g/cm3, including cadmium (Cd), mercury (Hg), manganese (Mn), lead (Pb), and selenium (Se) [13]. Heavy metals are necessary for various biological processes, but excessive amounts lead to heavy metal toxicity. Heavy metal poisoning is a widespread health hazard caused by mining, smelting, industry, agriculture, sewage pollution, and rubbish incineration [14, 15], and it can accumulate in humans through a variety of sources, including air, drinking water, food, and skin contact [13, 16]. Studies have found that heavy metal exposure increases oxidative stress and inflammatory responses [17, 18], heavy metals may cause damage to the respiratory system by inducing an inflammatory response in the airways [19]. An animal study based on rats found that multiple heavy metals can accumulate in the lungs, leading to impaired lung function [20]. Upon entering the body, heavy metals disrupt intracellular homeostasis and inhibit protease activity, resulting in excessive superoxide radicals, reactive oxygen species (ROS) production, and DNA damage, which in turn adversely affects the respiratory system [21–23]. Current research studies are limited to the association of few heavy metals or other chemicals with lung function or airway inflammation [24, 25], and there are no studies on the association of multiple heavy metal exposures with airway inflammation and lung function. Furthermore, we selected Cd, Hg, Mn, Pb, and Se because these metals are among the most commonly encountered in environmental pollution and occupational exposure [26, 27]. Additionally, they have well-documented effects, particularly on the respiratory system [28, 29]. Previous studies have shown that mice exposed to systemic Cd and Pb for six months exhibited decreased lung compliance, progressing from emphysema to pulmonary fibrosis [30]. Urinary Mn levels have been negatively correlated with lung function, even when within normal ranges, with females potentially being more susceptible to manganese exposure [31]. A prospective cohort study in China found a significant association between blood Hg levels and lung function decline, particularly among young individuals, especially males [32]. Another study indicated that selenium compounds (selenates) play a role in bleomycin-induced interstitial lung disease [33].

In this study, we selected data from the National Health and Nutrition Examination Study (NHANES) for 2011–2012. We used multivariable linear regression models, weighted quantile sum (WQS) regression models, and Bayesian kernel machine regression (BKMR) models to explore the associations between multiple heavy metal exposures (Cd, Hg, Mn, Pb, and Se) and airway inflammation with lung function in US adults, in addition to subgroup analyses and sensitivity analyses with trend tests to further confirm our results.

Methods

Study population

Data for this study are based on NHANES, a national survey, and a cross-sectional survey conducted by the Centers for Disease and Prevention in collaboration with the National Center for Health Statistics (NCHS) [34–36]. A multi-stage complex sampling approach, a combination of questionnaires, laboratory tests, and physical examinations were used to obtain large samples of scientific clinical data. The NCHS Research Ethics Review Board authorized the study procedure, and each participant signed an informed consent form [37]. This study used data from the 2011–2012 survey of the NHANES database, with a total of 9756 people were initially included. We excluded participants under the age of 20 (N = 4196), individuals with missing pulmonary function data (N = 1220), subjects with missing Fractional exhaled nitric oxide data (N = 220), individuals with missing heavy metal data (N = 172), and participants with missing data on other variables (N = 372). The final analysis included 3576 eligible adult participants (Fig. 1).

Fig. 1.

Flow chart of participants selection. NHANES, National Health and Nutrition Examination Survey

Heavy metal measurement

Whole blood specimens are processed, frozen at -30 degrees, and delivered to Laboratory Sciences, the National Center for Environmental Health, and the Centers for Disease Control and Prevention for research. The researchers employed inductively coupled plasma mass spectrometry (ICP-MS) with quadrupole ICP-MS technology to determine whole blood Pb, Cd, Hg, Mn, and Se concentrations. The detection limits were constant for all analytes, with a lower limit of detection of 0.25 µg/dL for Pb, 0.16 µg/L for Cd, 30 ug/L for Se, 1.06 ug/L for Mn, and 0.16 µg/L for Hg. If the result is below the detection limit, the value of the variable is the detection limit divided by the square root of two.

Measurement of fractional exhaled nitric oxide

FeNO was measured with the Aerocrine NIOX MINO^®, a portable handheld NOx analyzer certified by the FDA in 2008 (Aerocrine AB, Solna, Sweden). The device uses an electrochemical sensor to detect exhaled NOx levels and provides integer values ranging from 5 to 300 ppb. Participants in the study who performed the FeNO test were between the ages of 20 to 79 years old, and those with current chest pain or physical problems that caused forceful exhalation or who were using supplemental oxygen were excluded [38]. NHANES protocols require two repeatable FeNO measurements according to the manufacturer’s suggested testing procedures, and if the repeatable standard is not met on the first two breaths, the participant has an additional two breaths to meet the standard (up to a total of four testing trials). The Respiratory Health ENO Procedures Manual provides a detailed description of the FeNO testing process [39].

Lung function measurements

Ohio 822/827 dry-rolling seal volume spirometers were used for lung function measurements, the indicators measured include FEV1 (mL), FVC (mL), FEV1/FVC (%), peak expiratory flow (PEF) (mL/s), forced expiratory flow between 25 and 75% of vital capacity (FEF 25–75%) (mL/s). Participants between the ages of 20 and 79 years old took spirometry. Participants with current chest pain or physical problems, taking supplemental oxygen, recent eye, chest, or abdominal surgery, heart attack, stroke, tuberculosis exposure, or recent hemoptysis were excluded from the test. Individual spirometry was continued until reproducible spirograms could be obtained from the participant, with the overall goal of having the participant achieve three acceptable expiratory maneuvers according to the American Thoracic Society (ATS) standards, with the two highest values of FVC and FEV1 showing the least amount of variability, full details of the Lung Function Testing Program can be found at wwwn.cdc.gov/Nchs/Data/Nhanes/Public/2011/DataFiles/SPX_G.htm.

Other covariates

Based on previous studies [40], we included clinically significant covariates, mainly sociodemographic characteristics, lifestyle behaviors, and laboratory indicators. The main sociodemographic characteristics included age (years), gender (Male, Female), race (Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black, other races), family income-to-poverty ratio (PIR), marital status (Married, Others), education level (Less than high school, High school or GED, Above high school), and Body Mass Index (BMI) (kg/m2) (BMI was calculated by dividing weight in kilograms by height in meters squared). The lifestyle behaviors included smoking status [never smokers (Smoked < 100 cigarettes in life and does not smoke now), former smokers (Smoked at least 100 cigarettes in life and does not smoke now), and current smokers (Smoked at least 100 cigarettes in life and smoke now)]. The laboratory indicators included serum Cotinine (ng/mL). All data and definitions related to these variables are publicly accessible at https://www.cdc.gov/nchs/nhanes/.

Statistical analysis

All statistical analyses were performed using R (version 4.3) and Empowerstats (version 4.2). WQS regression and BKMR models were performed using the “gWQS” and “BKMR” packages. Data for continuous or categorical variables were analyzed by t-test or ANOVA, and participants were described statistically as mean ± standard deviation (SD) and percentage (%).

Firstly, associations between the five heavy metals and FeNO or lung function were analyzed using multivariable linear regression, where the first quartile (Q1) was designated as the control group, and a trend test was performed based on the quartiles of heavy metal concentrations. No covariates were adjusted in Model 1. Model 2 adjusted for age, gender, and race. Age, gender, race, education level, marital status, pack-years of smoking, diabetes, PIR, BMI, smoking status, and serum cotinine were adjusted in Model 3. To assess potential publication bias, funnel plots, and Egger’s test were performed for the association between heavy metals and lung function.

Secondly, we conducted subgroup analyses by sex, age, race, BMI, and smoking status to explore the sensitivity and interactions between heavy metal exposure and FeNO or lung function in different subgroups, where age was divided into three groups (20 ≤ Age < 40, 40 ≤ Age < 60, Age ≥ 60), and BMl was classified as four groups (BMI < 18.5 kg/m², 18.5 kg/m² ≤ BMI ≤ 24.9 kg/m², 25 kg/m² ≤ BMI ≤ 29.9 kg/m², BMI > 30 kg/m²) according to World Health Organization standards, corresponding to underweight, normal weight, overweight, and obesity, respectively. In the subgroup analysis, the model is not adjusted for the stratification variable itself.

Thirdly, we used weighted quantile and WQS regression models [41]. The overall and individual effects of the five heavy metals on FeNO and lung function were assessed by calculating weighted linear indices and assigning weights to the responses. We employed the R package (‘gWQS’) to compute the Weighted WQS index, a statistical model designed for multivariable regression in high-dimensional datasets. The WQS index is a weighted sum of the individual metal concentrations, where each exposure variable is assigned a weight to estimate its relative contribution to the outcome. This index, ranging from 0 to 1, reflects the burden of heavy metals about the exposure risk. In this study, we used a 10,000-iteration bootstrap to construct positive and negative WQS indexes to estimate each heavy metal’s relative contribution to FeNO and lung function. The dataset was randomly divided into two parts: 40% for training and 60% for validation. The training set was used to estimate the weights, while the validation set was used to assess the significance of the WQS index. For each bootstrap sample, a new dataset was created by performing resampling with replacement from the training set. Model parameters were then estimated using an optimization algorithm. The final WQS index was obtained by averaging the weights across multiple bootstrap samples. After constructing the WQS index, the association between the weighted mixture of exposures and the outcome was tested in the validation set. This approach allowed us to estimate the relative contribution of each heavy metal to the overall exposure risk, providing a comprehensive evaluation of their effects on the outcome.

Finally, given the potential non-linear and non-additive dose-response relationships among the mixture exposures, we employed the BKMR model to assess the joint effects of the heavy metals on lung function and FeNO risk [42]. The model used a threshold of 0.5 to determine the significance of the contribution, in addition to univariate and bivariate exposure-response functions to assess the single effects and interactions of the heavy metals while considering the 25th, 50th, and 75th percentile heavy metals, which was performed after adjusting for all the covariates using the Markov Chain Monte Carlo algorithm for 10000 iterations. This approach combines Bayesian statistical methods with machine learning techniques to estimate the non-linear relationships and interactions within exposure-outcome associations. A distinguishing feature of the BKMR model is its flexibility in modeling exposure-response functions, which facilitates the visualization of both individual and combined exposure effects. Bivariate exposure-response functions were constructed to assess the interactions between heavy metals. In our analysis, we used the ‘bkmr’ package in R to evaluate the combined effects of metal co-exposure on the risk of adverse outcomes. Additionally, we obtained the exposure-response functions and dose-response curves for each metal with respect to the outcome risk.

Previous studies have demonstrated that weighted estimating can cause over-adjustment bias when the major variables used to determine sampling weights are included as covariates in the model [43, 44]. Therefore, this study did not use weighted estimates when adjusting for factors such as demographics. Data with missing covariates were addressed using multiple interpolation [45]. A two-tailed p-value < 0.05 was considered statistically significant.

Results

Baseline characteristics

Table 1 summarizes the FeNO levels and lung function indices of the 3576 adult participants in this study. FeNO levels were higher in older, male participants with a higher level of education (above high school) and higher PIR levels. Conversely, participants with current smokers, higher BMI, and serum cotinine levels had lower FeNO levels. Furthermore, older adults, women, non-Hispanic Black individuals, those with lower education, lower PIR levels, and lower cotinine levels had lower FEV1 and FVC values, whereas older adults, men, non-Hispanic White individuals, those with less than high school, and those who were married had lower FEV1/FVC ratios. Participants with reduced FEV1/FVC ratios were more likely to have higher PIR levels, BMI levels, and serum cotinine levels and were more likely to be former or current smokers.

Table 1.

Weighted characteristics of participants with fractional exhaled nitric oxide and lung function parameters

Characteristic	Overall 3576 (100%)	FeNO (ppb)	FVC (mL)	FEV1 (mL)	FEV1/FVC (%)	PEF (mL/s)	FEF25%-75% (mL/s)
	n (%)	Mean ± SD
Age(years)
20–39	1474 (41.22%)	15.72 ± 14.79	4507.42 ± 1033.90	3687.67 ± 836.45	0.82 ± 0.07	9033.01 ± 2082.68	3819.11 ± 1187.90
40–59	1265 (35.37%)	16.13 ± 14.16	4094.86 ± 1028.17	3138.35 ± 796.55	0.77 ± 0.07	8392.92 ± 2144.19	2852.85 ± 1114.62
≥ 60	837 (23.41%)	18.01 ± 11.66	3523.28 ± 1025.54	2538.47 ± 752.69	0.72 ± 0.09	7252.48 ± 2200.26	1889.70 ± 923.55
P value		< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001
Gender
Male	1828 (51.12%)	17.92 ± 15.47	4856.68 ± 937.09	3757.41 ± 855.43	0.77 ± 0.09	9765.18 ± 2044.35	3429.20 ± 1427.42
Female	1748 (48.88%)	14.76 ± 12.13	3438.54 ± 704.73	2723.19 ± 624.26	0.79 ± 0.08	7084.71 ± 1465.17	2671.48 ± 1084.10
P value		< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001
Race/ethnicity, n (%)
Mexican American	361 (10.1%)	15.35 ± 12.26	4134.83 ± 934.74	3353.67 ± 740.86	0.81 ± 0.06	8714.73 ± 2118.62	3562.54 ± 1178.32
Other Hispanic	356 (9.96%)	18.61 ± 15.87	3965.61 ± 1020.66	3230.81 ± 868.48	0.81 ± 0.07	8296.99 ± 2173.40	3393.17 ± 1262.97
Non-Hispanic White	1336 (37.36%)	16.19 ± 14.03	4297.55 ± 1096.73	3309.37 ± 934.98	0.77 ± 0.09	8504.25 ± 2248.97	2976.38 ± 1349.00
Non-Hispanic Black	925 (25.87%)	15.91 ± 13.40	3565.29 ± 950.00	2851.76 ± 804.99	0.80 ± 0.08	7921.01 ± 2193.70	2899.87 ± 1236.12
Other Races	598 (16.72%)	17.36 ± 14.16	3777.65 ± 1038.49	3058.83 ± 839.08	0.81 ± 0.07	8240.52 ± 2093.38	3133.28 ± 1211.87
P value		0.052	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001
Education level
Less than high school	689 (19.27%)	15.09 ± 13.34	3913.47 ± 1046.15	3033.26 ± 878.13	0.77 ± 0.09	7823.77 ± 2265.47	2884.48 ± 1348.68
High school or GED	725 (20.27%)	14.64 ± 10.56	4071.00 ± 1070.39	3159.53 ± 922.04	0.78 ± 0.09	8320.24 ± 2319.62	2961.95 ± 1411.68
Above high school	2162 (60.46%)	17.08 ± 14.89	4217.56 ± 1097.59	3305.83 ± 904.78	0.78 ± 0.08	8578.14 ± 2168.73	3109.58 ± 1286.57
P value		< 0.001	< 0.001	< 0.001	0.005	< 0.001	< 0.001
Marital status
Married	1762 (49.27%)	16.21 ± 11.70	4151.69 ± 1075.78	3206.31 ± 873.47	0.77 ± 0.08	8493.16 ± 2184.60	2926.79 ± 1245.17
Others	1814 (50.73%)	16.49 ± 16.28	4143.09 ± 1108.55	3280.50 ± 949.70	0.79 ± 0.09	8345.20 ± 2273.25	3195.87 ± 1394.88
P value		0.560	0.814	0.015	< 0.001	0.048	< 0.001
PIR
<1	824 (23.04%)	15.91 ± 16.23	4052.55 ± 1089.75	3225.91 ± 934.95	0.80 ± 0.09	8053.97 ± 2226.30	3194.20 ± 1366.29
1-3.9	1768 (49.44%)	15.67 ± 13.62	4085.42 ± 1085.02	3197.15 ± 914.74	0.78 ± 0.09	8292.27 ± 2203.50	3045.45 ± 1366.60
≥ 4	984 (27.52%)	17.34 ± 13.36	4265.25 ± 1088.53	3300.07 ± 889.99	0.77 ± 0.08	8747.05 ± 2214.90	2995.62 ± 1242.24
P value		0.004	< 0.001	0.008	< 0.001	< 0.001	0.011
BMI(kg/m 2 )
< 18.5	55 (1.54%)	20.21 ± 28.59	3511.09 ± 758.01	2969.31 ± 804.34	0.84 ± 0.11	6996.44 ± 1857.76	3178.53 ± 1270.23
18.5–24.9	1085 (30.34%)	15.37 ± 13.19	4209.31 ± 1048.45	3290.60 ± 892.95	0.78 ± 0.09	8248.64 ± 2095.02	3060.54 ± 1331.39
25-29.9	1141 (31.91%)	16.68 ± 13.61	4329.92 ± 1078.89	3354.39 ± 906.89	0.77 ± 0.08	8778.51 ± 2221.08	3086.89 ± 1356.90
> 30	1295 (36.21%)	16.68 ± 14.09	3942.88 ± 1106.76	3097.05 ± 910.60	0.79 ± 0.07	8290.52 ± 2299.65	3001.05 ± 1281.73
P value		0.016	< 0.001	< 0.001	< 0.001	< 0.001	0.365
Smoking status
Former	768 (21.48%)	18.41 ± 15.92	4129.79 ± 1079.09	3143.60 ± 912.88	0.76 ± 0.08	8394.27 ± 2240.53	2763.26 ± 1354.29
Now	750 (20.97%)	10.01 ± 8.18	4309.70 ± 1069.99	3239.54 ± 930.65	0.75 ± 0.10	8141.65 ± 2263.71	2800.18 ± 1418.36
Never	2058 (57.55%)	17.70 ± 14.12	4098.02 ± 1097.83	3281.55 ± 898.27	0.80 ± 0.07	8538.55 ± 2198.61	3260.15 ± 1233.27
P value		< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001
Serum Cotinine (ng/mL)
≤ 10	2689 (75.20%)	17.88 ± 13.89	4087.56 ± 1089.04	3226.72 ± 900.44	0.79 ± 0.08	8461.90 ± 2192.51	3108.62 ± 1296.58
> 10	887 (24.80%)	11.46 ± 13.18	4337.60 ± 1075.17	3283.52 ± 938.35	0.75 ± 0.10	8309.36 ± 2328.61	2866.76 ± 1386.70
P value		< 0.001	< 0.001	0.111	< 0.001	0.080	< 0.001

Mean + SD for continuous variables: the P value was calculated by the weighted linear regression model

FeNO: Fractional exhaled nitric oxide; FVC: Forced vital capacity; FEV1: Forced expiratory volume in 1; PEF: Peak expiratory flow; FEF25%-75%: Forced expiratory flow between 25 and 75% of vital capacity; BMI: Body mass index; PIR: Family income-to-poverty ratio

Participants’ FeNO and lung function characteristics by heavy metal quartile are shown in Table 2. Subjects with higher serum Cd levels had lower FeNO, FEV1, FVC, FEV1/FVC, and PEF values compared to those in the lowest quartile. FeNO values for serum Se, Hg, Mn, and Pb tended to increase with higher heavy metal concentrations. As serum Mn levels increased, FEV1 and FVC values decreased, while FEV1/FVC ratios increased. In contrast, FVC and FEV1/FVC ratios decreased with higher serum Pb levels.

Table 2.

Weighted basic characteristics of participants by heavy metals

Heavy metals	FeNO (ppb)	FVC (mL)	FEV1 (mL)	FEV1/FVC (%)	PEF (mL/s)	FEF25%-75% (mL/s)
Serum cadmium (µg/L)
Quartile 1	18.0 ± 14.16	4454.3 ± 1069.90	3566.8 ± 894.06	0.80 ± 0.08	9147.7 ± 2100.18	3550.0 ± 1318.76
Quartile 2	18.1 ± 16.55	4181.1 ± 1060.94	3301.5 ± 868.75	0.79 ± 0.07	8657.7 ± 2184.94	3168.51 ± 123.51
Quartile 3	16.9 ± 13.03	3800.7 ± 1024.31	2943.9 ± 817.94	0.78 ± 0.08	7845.1 ± 2099.13	2719.8 ± 1207.72
Quartile 4	11.21 ± 9.19	4051.8 ± 1102.85	3031.6 ± 916.69	0.75 ± 0.10	7764.0 ± 2223.13	2574.1 ± 1286.28
P value	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001
Serum selenium (µg/L)
Quartile 1	15.1 ± 11.33	3958.5 ± 1048.84	3080.63 ± 884.50	0.78 ± 0.09	8043.80 ± 2188.04	2907.51 ± 1312.6
Quartile 2	15.6 ± 12.30	4164.25 ± 1130.44	3264.11 ± 949.75	0.78 ± 0.08	8498.99 ± 2311.39	3096.19 ± 1360.7
Quartile 3	17.3 ± 14.56	4227.15 ± 1090.26	3294.25 ± 905.56	0.78 ± 0.09	8480.04 ± 2176.53	3088.99 ± 1336.8
Quartile 4	17.0 ± 16.66	4221.49 ± 1070.16	3306.06 ± 881.40	0.78 ± 0.08	8636.03 ± 2187.03	3095.28 ± 1272.8
P value	0.002	< 0.001	< 0.001	0.3924	< 0.001	0.006
Serum manganese (µg/L)
Quartile 1	15.53 ± 11.26	4300.24 ± 1156.00	3306.30 ± 976.69	0.77 ± 0.09	8572.22 ± 2354.07	2995.92 ± 1382.97
Quartile 2	17.15 ± 15.05	4283.46 ± 1075.86	3309.59 ± 908.67	0.77 ± 0.09	8522.94 ± 2275.80	3007.45 ± 1320.01
Quartile 3	15.87 ± 12.40	4096.13 ± 1102.29	3235.85 ± 918.12	0.79 ± 0.07	8474.43 ± 2221.18	3126.25 ± 1320.15
Quartile 4	16.79 ± 16.86	3862.54 ± 948.79	3082.11 ± 795.91	0.80 ± 0.08	8072.42 ± 1969.77	3075.60 ± 1252.70
P value	0.044	< 0.001	< 0.001	< 0.001	< 0.001	0.117
Serum lead (µg/dL)
Quartile 1	15.26 ± 13.29	4108.16 ± 1012.95	3339.82 ± 846.43	0.81 ± 0.07	8462.55 ± 2109.13	3428.00 ± 1229.75
Quartile 2	16.46 ± 13.91	4232.19 ± 1131.69	3350.89 ± 927.39	0.79 ± 0.07	8653.43 ± 2150.46	3241.87 ± 1280.95
Quartile 3	17.51 ± 13.48	4132.11 ± 1133.64	3196.34 ± 937.39	0.77 ± 0.08	8407.58 ± 2277.19	2924.38 ± 1307.74
Quartile 4	16.23 ± 15.28	4116.96 ± 1082.76	3043.50 ± 897.63	0.74 ± 0.10	8141.74 ± 2359.62	2515.63 ± 1298.51
P value	0.007	0.055	< 0.001	< 0.001	< 0.001	< 0.001
Serum total mercury(µg/L)
Quartile 1	14.41 ± 11.36	4172.06 ± 1072.95	3258.75 ± 916.54	0.78 ± 0.09	8223.53 ± 2212.54	3079.69 ± 1340.78
Quartile 2	15.77 ± 14.16	4171.84 ± 1101.61	3259.33 ± 924.86	0.78 ± 0.08	8460.87 ± 2249.48	3088.37 ± 1357.91
Quartile 3	16.38 ± 12.63	4129.76 ± 1077.98	3244.49 ± 902.08	0.79 ± 0.08	8504.48 ± 2202.71	3085.61 ± 1316.07
Quartile 4	19.11 ± 17.17	4112.89 ± 1112.03	3193.26 ± 892.64	0.78 ± 0.08	8523.54 ± 2230.26	2933.99 ± 1261.44
P value	< 0.001	0.576	0.396	0.342	0.015	0.043

Mean + SD for continuous variables: the P value was calculated by the weighted linear regression model

FeNO: Fractional exhaled nitric oxide; FVC: Forced vital capacity; FEV1: Forced expiratory volume in 1; PEF: Peak expiratory flow; FEF25%-75%: Forced expiratory flow between 25 and 75% of vital capacity

Relationship between heavy metals and FeNO

Table 3 presents the results of the multivariable linear regression analysis across three models, examining the five heavy metals and FeNO. A significant positive relationship was found between serum Hg and FeNO, which remained statistically significant in all three models, with an increase of 0.20 ppb in FeNO values for each unit increase in serum Hg in the fully adjusted model [0.20 (0.02, 0.37)]. Negative associations between serum Cd and FeNO were significant in Models 1 and 2 but not in the fully adjusted model. Positive relationships between serum Se and FeNO were also observed only in Models 1 and 2, while negative associations between serum Pb and FeNO were significant only in Model 2. No association was found between serum Mn and FeNO in any of the three models.

Table 3.

Association between cd, hg, Mn, Se, Pb, and fractional exhaled nitric oxide

Exposure	Model1:β(95%CI), p	Model2:β(95%CI), p	Model3:β(95%CI), p
Cd	-3.67(-4.38, -2.97) < 0.001	-3.67(-4.38, -2.97) < 0.001	-0.60(-1.49, 0.30) 0.193
Cd classification
Quartile 1	Reference	Reference	Reference
Quartile 2	-0.37(-1.65, 0.91) 0.572	-0.38(-1.66, 0.90) 0.558	0.18(-1.09, 1.46) 0.780
Quartile 3	-1.15(-2.44, 0.13) 0.079	-1.52(-2.84, -0.20) 0.024	-0.01(-1.37, 1.35) 0.991
Quartile 4	-6.07(-7.35, -4.79) < 0.001	-6.46(-7.76, -5.16) < 0.001	-1.34(-3.01, 0.32) 0.114
P for trend	< 0.001	< 0.001	0.060
Hg	0.44(0.27, 0.61) < 0.001	0.31(0.13, 0.48) < 0.001	0.20(0.02, 0.37) 0.027
Hg classification
Quartile 1	Reference	Reference	Reference
Quartile 2	1.25(-0.04, 2.54) 0.058	1.07(-0.22, 2.36) 0.104	0.58(-0.69, 1.84) 0.371
Quartile 3	2.12(0.83, 3.41) 0.001	1.72(0.43, 3.02) 0.009	0.92(-0.36, 2.20) 0.159
Quartile 4	3.64(2.36, 4.93) < 0.001	2.62(1.26, 3.97) < 0.001	1.56(0.20, 2.92) 0.025
P for trend	< 0.001	< 0.001	0.035
Mn	0.06(-0.05, 0.18) 0.274	0.12(-0.01, 0.24) 0.063	0.10(-0.02, 0.22) 0.104
Mn classification
Quartile 1	Reference	Reference	Reference
Quartile 2	1.34(0.05, 2.64) 0.042	1.54(0.25, 2.84) 0.020	1.13(-0.13, 2.39) 0.080
Quartile 3	1.27(-0.02, 2.65) 0.054	1.51(0.19, 2.83) 0.025	1.16(-0.12, 2.44) 0.075
Quartile 4	1.13(-0.05, 0.28) 0.088	1.67(0.29, 3.05) 0.018	1.40(0.05, 2.74) 0.042
P for trend	0.173	0.044	0.072
Se	0.03(0.01, 0.04) 0.003	0.02(0.00, 0.04) 0.024	0.01(-0.01, 0.03) 0.216
Se classification
Quartile 1	Reference	Reference	Reference
Quartile 2	0.93(-0.36, 2.23) 0.158	0.70(-0.58, 1.98) 0.286	0.35(-0.90, 1.60) 0.581
Quartile 3	1.98(0.69, 3.28) 0.003	1.70(0.42, 2.99) 0.009	0.98(-0.27, 2.24) 0.126
Quartile 4	1.03(-0.26, 2.33) 0.117	0.65(-0.64, 1.94) 0.326	-0.05(-1.32, 1.21) 0.934
P for trend	0.074	0.226	0.952
Pb	-0.12(-0.35, 0.11) 0.306	-0.32(-0.55, -0.09) 0.007	-0.06(-0.29, 0.17) 0.601
Pb classification
Quartile 1	Reference	Reference	Reference
Quartile 2	1.46(0.16, 2.75) 0.028	0.04(-1.29, 1.38) 0.950	0.92(-0.39, 2.23) 0.169
Quartile 3	1.19(-0.11, 2.48) 0.073	-0.83(-2.24, 0.58) 0.250	0.63(-0.76, 2.02) 0.377
Quartile 4	0.50(-0.79, 1.80) 0.449	-2.13(-3.61, -0.65) 0.005	0.42(-1.08, 1.92) 0.580
P for trend	0.958	< 0.001	0.985

Model 1: No covariates were adjusted

Model 2: Age, gender, and race were adjusted

Model 3: Age, gender, race, education level, marital status, pack-years of smoking, diabetes, PIR, BMI, smoking status, and serum cotinine were adjusted

Cd: Cadmium; Hg: Mercury; Mn: Manganese; Se: Selenium; Pb: Lead; BMI: Body mass index; PIR: Family income-to-poverty ratio

Relationship between heavy metals and lung function

Table 4 shows the results of the multivariate linear regression analysis of heavy metals and FEV1. All three models indicated significant negative relationships between serum Cd, serum Pb, and FEV1. In the fully adjusted model, FEV1 decreased by 106.22 ml [-106.22(-143.64, -68.80)] for each unit increase in serum Cd and by 17.85 ml [-17.85 (-27.48, -8.21)] for each unit increase in serum Pb. Furthermore, we observed a positive relationship between serum Se and FEV1 in models 1 and 2. However, this association was not maintained in the fully adjusted model. Additionally, only the unadjusted model showed a negative relationship between Mn and Hg exposure and FEV1.

Table 4.

Association between cd, hg, Mn, Se, Pb, and FEV1

Exposure	Model1: β(95%CI), p	Model2: β(95%CI), p	Model3: β(95%CI), p
Cd	-160.79(-206.36, -115.22) < 0.001	-115.95(-145.44, -86.46) < 0.001	-106.22(-143.64, -68.80) < 0.001
Cd classification
Quartile 1	Reference	Reference	Reference
Quartile 2	-300.23(-381.24, -219.22) < 0.001	-24.52(-78.38, 29.35) 0.372	-25.97(-79.27, 27.33) 0.340
Quartile 3	-551.97 (-633.31, -470.63) < 0.001	-58.47 (-114.18, -2.77) 0.040	-52.97 (-109.86, 3.91) 0.068
Quartile 4	-511.65 (-592.59, -430.71) < 0.001	-161.31 (-216.07, -106.56) < 0.001	-122.33 (-191.93, -52.73) <0.001
P for trend	< 0.001	< 0.001	< 0.001
Hg	-13.25(-24.08, -2.42) 0.017	2.95(-4.41, 10.31) 0.432	-3.18(-10.48, 4.13) 0.394
Hg classification
Quartile 1	Reference	Reference	Reference
Quartile 2	-19.15 (-102.43, 64.12) 0.652	75.90 (22.37, 129.43) 0.006	51.82 (-1.01, 104.65) 0.055
Quartile 3	-94.68 (-177.56, -11.80) 0.025	89.23 (35.40, 143.06) 0.001	41.86 (-11.61, 95.33) 0.125
Quartile 4	-96.71 (-179.68, -13.74) 0.022	114.45 (58.01, 170.89) < 0.001	48.64 (-8.48, 105.76) 0.095
P for trend	0.023	0.002	0.319
Mn	-13.55(-20.97, -6.13) < 0.001	-0.43(-5.51, 4.65) 0.867	-0.29(-5.26, 4.68) 0.908
Mn classification
Quartile 1	Reference	Reference	Reference
Quartile 2	71.43 (-11.39, 154.26) 0.091	40.08 (-13.80, 93.96) 0.145	23.76 (-29.07, 76.58) 0.378
Quartile 3	8.86 (-73.94, 91.66) 0.834	20.96 (-33.83, 75.74) 0.454	17.33 (-36.29, 70.94) 0.527
Quartile 4	-134.40 (-217.18, -51.63) 0.002	3.85 (-53.66, 61.36) 0.896	1.65 (-54.65, 57.95) 0.954
P for trend	< 0.001	0.793	0.872
Se	2.65(1.54, 3.76) < 0.001	0.75(0.03, 1.47) 0.042	0.61(-0.10, 1.32) 0.093
Se classification
Quartile 1	Reference	Reference	Reference
Quartile 2	157.89 (75.08, 240.70) < 0.001	82.24 (28.87, 135.61) 0.003	66.40 (14.09, 118.72) 0.013
Quartile 3	177.89 (95.15, 260.63) < 0.001	54.76 (1.33, 108.20) 0.045	38.75 (-13.77, 91.27) 0.148
Quartile 4	205.44 (122.66, 288.23) < 0.001	56.44 (2.67, 110.22) 0.040	44.67 (-8.25, 97.60) 0.098
P for trend	< 0.001	0.112	0.220
Pb	-25.14(-39.86, -10.43) < 0.001	-23.61(-33.30, -13.92) < 0.001	-17.85(-27.48, -8.21) < 0.001
Pb classification
Quartile 1	Reference	Reference	Reference
Quartile 2	-48.35 (-130.68, 33.99) 0.250	-20.48 (-75.68, 34.73) 0.467	-21.89 (-76.35, 32.56) 0.431
Quartile 3	-215.05 (-297.25, -132.85) < 0.001	-69.75 (-127.98, -11.52) 0.019	-68.33 (-126.32, -10.35) 0.021
Quartile 4	-342.85 (-425.07, -260.63) < 0.001	-197.22 (-258.41, -136.03) < 0.001	-182.15 (-244.62, -119.67) < 0.001
P for trend	< 0.001	< 0.001	< 0.001

Model 1: No covariates were adjusted

Model 2: Age, gender, and race were adjusted

Model 3: Age, gender, race, education level, marital status, pack-years of smoking, diabetes, PIR, BMI, smoking status, and serum cotinine were adjusted

FEV1: Forced expiratory volume in 1; Cd: Cadmium; Hg: Mercury; Mn: Manganese; Se: Selenium; Pb: Lead; BMI: Body mass index; PIR: Family income-to-poverty ratio

Table S1 presents the results of the multivariable linear regression analysis for heavy metals and FVC. A negative relationship between Cd exposure and FVC was also found in all three models, with the FVC decrease of 74.94 ml per unit increase in serum Cd in the fully adjusted model [-74.94 (-119.22, -30.66)]. In the unadjusted model, Hg and Mn were negatively associated with FVC, while Se exposure showed a positive correlation with FVC. Furthermore, we found Pb exposure was negatively associated with FVC in model 3 [-14.84 (-26.22, -3.45)].

The results of the multivariable linear regression analysis between the five heavy metals and lung function (FEV1/FVC) are shown in Table S2. A significant negative relationship was found between serum Cd and FEV1/FVC, which remained statistically significant in all three models. In the fully adjusted model, each unit increase in serum Cd was associated with a 1.35% decrease in FEV1/FVC values[-1.35 (-1.82, -0.88)]. In all three models, there was a strong positive connection between serum Mn and FEV1/FVC, with a 0.09% increase in FEV1/FVC values for every unit increase in serum Mn in the fully adjusted model [0.09 (0.02, 0.15)]. A positive association was also identified between serum Se and FEV1/FVC in model 2 [1.10 (0.24, 1.95)], although this positive correlation did not persist in the fully adjusted model. The negative relationship between serum Pb and FEV1/FVC was significant in model 3, with each unit increase in serum Pb resulting in a 0.14% decrease in FEV1/FVC values [-0.14 (-0.26, -0.02)]. In contrast, no association between serum Hg and lung function was found in any of the three models. The funnel plot and Egger’s test results showed no significant publication bias, as the p-value for Egger’s test was greater than 0.05 (Fig. S4).

Sensitivity analysis

We conducted sensitivity analyses to examine the impact of potential confounders and adjustments for pack-years of smoking, BMI, diabetes, and other variables. Additionally, we removed extreme values of heavy metal exposure, defined as the top 5% and bottom 5% of the distribution, to ensure the robustness of the results. The sensitivity analysis showed that, with the exception of the non-significant associations between Hg and FeNO [0.30 (-0.20, 0.80)], as well as between Mn and FEV1/FVC [0.10 (-0.02, 0.22)] in the fully adjusted model, all other results were consistent with our previous findings (Tables S6, S7, S8, and S9).

Subgroup analysis

We performed subgroup analyses and interaction tests (Fig. S1) to assess the relationships between heavy metals and FeNO across different groups based on sex, age, race, BMI, and smoking status. Subgroup analysis revealed contradictory relationships between certain metals and FeNO among groups. Specifically, the relationship between serum Se and FeNO was modified by age (p for interaction = 0.02). BMI-stratified analysis showed an interaction between serum Hg, serum Mn, serum Se, serum Pb, and FeNO (p for interaction < 0.05). Subgroup analysis stratified for smoking revealed a significant interaction between serum Cd, serum Hg, serum Se, and FeNO (p for interaction < 0.05).

Furthermore, subgroup analyses and interaction results between the five heavy metals and FEV1/FVC are shown in Fig. S2. The associations between serum Hg, serum Se, and FEV1/FVC were moderated by gender. Specifically, serum Hg and FEV1/FVC were positively associated in females, whereas serum Se and FEV1/FVC were positively related in males. In the BMI-stratified analysis, an interaction between serum Se, serum Pb, and FEV1/FVC was observed (p for interaction < 0.05). Smoking status influenced the relationship between serum Se and FEV1/FVC, with current smokers showing a beneficial relationship between serum Se and lung function.

BKMR analysis

BKMR analysis revealed a negative one-way exposure-response association between Cd exposure and FeNO levels (Fig. 2A) and a positive one-way relationship between Hg and Mn exposure and FeNO levels. As shown in Fig. 2C, when heavy metals were fixed at different percentiles (25th, 50th, and 75th), serum Cd was negatively associated with FeNO, while serum Hg and Mn were positively associated with FeNO. Moreover, the overall effect of co-exposure to the five heavy metals on FeNO levels was inhibitory (Fig. 2B). Serum Pb was found to interact with other heavy metals (Cd, Hg, Mn, and Se) on FeNO levels (Fig. 2D).

Fig. 2.

Associations of Pb, cd, hg, Se, and Mn with FeNO using the BKMR model. (A) The univariate exposure-response association between exposure to Pb, Cd, Hg, Se, Mn and FeNO was calculated by the BKMR model. (B) The joint effect of Pb, Cd, Hg, Se, and Mn mixture on FeNO was calculated by the BKMR model. (C) Single heavy metal effect (95% CI) to FeNO when other heavy metals were fixed at a specific (25th, 50th, 75th). (D) Bivariate exposure-response relationship between five heavy metals and FeNO (a visualization for evaluating interactions). The models were adjusted for age, gender, race, education level, marital status, PIR, BMI, smoking status, and serum cotinine. Cd: Cadmium; Hg: Mercury; Mn: Manganese; Se: Selenium; Pb: Lead; FeNO: Fractional Exhaled nitric oxide; PIR: Family income-to-poverty ratio; BMI: Body mass index

Similarly, our BKMR analysis of the five heavy metals with FEV1 and FEV1/FVC showed a negative one-way exposure-response association between Cd exposure and both FEV1 and FEV1/FVC (Figs. 3A and 4A). In contrast, a positive one-way exposure-response relationship was found between Hg and Mn exposure and FEV1/FVC. The overall effect of co-exposure to these five heavy metals on FEV1 was initially enhancing, followed by inhibition (Fig. 3B), while the overall impact on FEV1/FVC levels was inhibitory at first, then enhancing, with a net enhancement effect (Fig. 4B). Univariate effects analysis revealed a significant negative relationship between Cd exposure and FEV1/FVC, while Hg exposure was positively associated with FEV1/FVC (Fig. 4C). However, no significant association between the five heavy metals and FEV1 was found in univariate analysis (Fig. 3C). Furthermore, bivariate exposure dose-response curves suggested significant interactions between heavy metals and FEV1/FVC, with results showing an interaction between Cd and Pb, Hg, Se, and Mn, an interaction between Hg and Mn, Pb, and Se, an interaction between Mn and Pb, and an interaction between Pb and Se (Fig. 4D). The bivariate exposure-response effects regarding heavy metals and FEV1 as shown in (Fig. 3D), revealed significant interactions between Cd and Hg, Mn, Pb; between Hg and Mn, Pb, Se; and between Mn and Pb.

Fig. 3.

Associations of Pb, cd, hg, Se, and Mn with FEV1 using the BKMR model. (A) The univariate exposure-response association between exposure to Pb, Cd, Hg, Se, Mn and FEV1 was calculated by the BKMR model. (B) The joint effect of Pb, Cd, Hg, Se, and Mn mixture on FEV1 was calculated by the BKMR model. (C) Single heavy metal effect (95% CI) to FEV1 when other heavy metals were fixed at a specific (25th, 50th, 75th). (D) Bivariate exposure-response relationship between five heavy metals and FEV1 (a visualization for evaluating interactions). The models were adjusted for age, gender, race, education level, marital status, PIR, BMI, smoking status, and serum cotinine. Cd: Cadmium; Hg: Mercury; Mn: Manganese; Se: Selenium; Pb: Lead; FEV1: Forced expiratory volume in 1; PIR: Family income-to-poverty ratio; BMI: Body mass index

Fig. 4.

Associations of Pb, Cd, Hg, Se, and Mn with FEV1/FVC using the BKMR model. (A) The univariate exposure-response association between exposure to Pb, Cd, Hg, Se, Mn, and FEV1/FVC was calculated by the BKMR model. (B) The joint effect of Pb, Cd, Hg, Se, and Mn mixture on FEV1/FVC was calculated by the BKMR model. (C) Single heavy metal effect (95% CI) to FEV1/FVC when other heavy metals were fixed at a specific (25th, 50th, 75th). (D) Bivariate exposure-response relationship between five heavy metals and FEV1/FVC (a visualization for evaluating interactions). The models were adjusted for age, gender, race, education level, marital status, PIR, BMI, smoking status, and serum cotinine. Cd: Cadmium; Hg: Mercury; Mn: Manganese; Se: Selenium; Pb: Lead; FEV1: Forced expiratory volume in 1; FVC: Forced vital capacity; PIR: Family income-to-poverty ratio; BMI: Body mass index

WQS analysis

We employed the WQS regression model to investigate the positive and negative effects of five heavy metal exposures on FeNO, FEV1, and FEV1/FVC. Figure 5 demonstrates the negative associations between different heavy metals and FeNO (Fig. 5A) and FEV1 (Fig. 5B ) FEV1/FVC (Fig. 5C) using the WQS regression model. Regardless of FeNO, FEV1, or FEV1/FVC, Cd and Pb exposure were the predominant negative correlations affecting FeNO, FEV1, and FEV1/FVC.

Fig. 5.

Negative directions of heavy metals with FeNO (A), FEV1 (B), and FEV1/FVC (C) using weighted quantile sum (WQS) regression models. The models were adjusted for age, gender, race, education level, marital status, PIR, BMI, smoking status, and serum cotinine. Cd: Cadmium; Hg: Mercury; Mn: Manganese; Se: Selenium; Pb: Lead; FEV1: Forced expiratory volume in 1; FVC: Forced vital capacity; FeNO: Fractional Exhaled nitric oxide; PIR: Family income-to-poverty ratio; BMI: Body mass index

Tables S3, S4, and S5 show the average weights of the five heavy metals in the WQS regression model for FeNO, FEV1, and FEV1/FVC, where Cd is the most dominant adversely related exposure factor, with a negative impact weight of 0.90 on FeNO, 0.622 on FEV1, and a negative FEV1/FVC weight of 0.85. Fig. S3 depicts the positive relationship between different heavy metals and FeNO, FEV1, and FEV1/FVC. Overall, Hg and Mn were the most dominant exposure factors in the positive correlation with FeNO, as confirmed by the BKMR analysis, while Se exposure was the most dominant factor in the positive relationship with FEV1 and FEV1/FVC, with weights of 0.455 and 0.339, respectively.

Discussion

To our knowledge, this is the first study to comprehensively investigate the relationship between exposures to several heavy metals (Cd, Hg, Mn, Pb, and Se) with airway inflammation and lung function in adults. In the fully adjusted model, we found a significant positive relationship between serum Hg and FeNO, a significant negative association between serum Cd and Pb and FEV1, FVC, and FEV1/FVC, and a significant positive relationship between Mn and FEV1/FVC. BKMR analysis revealed a negative association between co-exposure to multiple heavy metals and FeNO, and a one-way exposure-response relationship between Cd exposure and FeNO, FEV1, and FEV1/FVC. According to WQS analysis, Cd exposure was the most significantly associated with lower airway inflammation and reduced lung function, whereas Hg exposure was the most important positive relationship with FeNO, and Se exposure contributed the strongest positive weight to FEV₁ and FEV₁/FVC. These findings suggest that the relationships between different heavy metals, airway inflammation, and lung function are inconsistent, with Cd and Pb exposure potentially associated with adverse respiratory health outcomes.

Heavy metals are prevalent in the environment, and overexposure to them can lead to heavy metal pollution, primarily resulting from human activities such as metal mining, smelting, foundries, and other industrial processes [12]. Metal toxicity produces free radicals, which can cause DNA damage, altered sulfonyl homeostasis, and lipid peroxidation, culminating in carcinogenicity or oxidative damage [46]. Previous studies have found that heavy metal exposure is related to disorders such as sarcopenia [27], pan-cancer [47], osteoporosis [48], and chronic kidney disease [49]. In this study, we focused on the relationship between Cd, Hg, Mn, Pb, and Se exposure and airway inflammation and lung function. In a cross-sectional study of 13,888 U.S. adult men, Yang et al. [50] found that Cd exposure was associated with an increased risk of asthma. Serum Cd levels were negatively correlated with FEV1% predicted, FEV1/FVC% predicted, and FeNO levels, with a stronger negative relationship in never-smokers and former smokers. Similarly, in a cross-sectional study of 7813 adults, Min et al.³⁷ found that higher blood Cd levels were associated with lower FeNO levels regardless of smoking or serum cotinine levels after adjusting for potential confounders. The above studies are partially consistent with the findings of the current study, which showed a significant negative association between blood Cd exposure and lung function (FEV1, FVC, and FEV1/FVC). In Models 1 and 2, we also observed a negative correlation between Cd levels and FeNO, although this association was not maintained in the fully adjusted model. Furthermore, WQS analysis showed that Cd exposure was most strongly associated with lower FeNO levels and reduced lung function, which had not been previously reported. Cd exposure increases oxidative stress by inhibiting antioxidant enzyme activity and promoting the generation of ROS [51]. This leads to the activation of immune cells and an increase in pro-inflammatory cytokine levels, ultimately contributing to a decline in lung function [52]. We have learned that FeNO, a biomarker of type II inflammation, is associated with airway eosinophilia and that higher FeNO levels tend to indicate higher levels of inflammation and are frequently employed in specific asthma phenotypes [53]. However, FeNO levels are lower in patients with neutrophilic asthma and COPD [54]. Therefore, the negative relationship between serum Cd and FeNO levels in this research could be attributed to the fact that the participants with decreased lung function were primarily neutrophilic asthma and COPD. A study of 382 subjects from a rural area in northwest China showed that heavy metal exposure was associated with decreased lung function regardless of whether the exposure was single or mixed and that FEV1% predicted and FVC% predicted were negatively correlated with Mn, Hg, and Stibium (Sb) [26]. Chen et al. [40] employed a multi-pollutant model to investigate the relationships between 12 urinary metals (including Cd, Pb, and others) and four key lung functions in 1227 US children aged 6 to 17 years. We found a decreasing trend in lung function as metal combination concentrations increased, with Pb having the strongest negative correlation weight, in contrast to the current study, which revealed Cd exposure to be the most important factor affecting lung function. It may be because Chen’s study population was children, and the difference in exposure between children and adults led to the difference. Secondly, the exposure factor in their study was derived from urine, whereas the exposure factor in the present study was heavy metals in the blood. A multicenter cross-sectional study by Pan et al. [55] selected 221 healthy children to investigate the effects of Pb, Hg, and Cd exposure on children’s lung function and found that serum Pb levels were significantly negatively associated with lung function (FEV1, FVC, and FEV1%), Another prospective cohort study from China, which included 1,628 coke oven workers, found that increased Pb exposure was significantly associated with lower FEV1 [56]. The above results are consistent with a negative relationship between Pb and FEV1 in our investigation. Previous studies have demonstrated that Pb exposure may eventually lead to morphological and functional alterations in lung epithelial cells through modifying lipid peroxidation and redox imbalance [57, 58]. Some studies have suggested that Pb may mediate its effects through the extracellular signal-regulated kinase (ERK) 1/2 and p38 signaling pathways, stimulating inflammatory factors such as cyclooxygenase-2 (COX-2) and nicotinamide adenine dinucleotide phosphate (NADPH) oxidase [59].

We also found that Hg exposure was positively associated with FeNO levels and was the most significant correlation among the five heavy metals. A prior study of 3296 adult participants in the United States investigated the relationship between exposure to a mixture of metals, parabens, and phthalates and the proportion of exhaled nitric oxide. The mixture’s overall effect was found to be positively connected with FeNO, and the WQS analysis revealed that Hg was the most important variable (40.2%) [60] positively correlated with FeNO, confirming the results of this study. Although Hg is known for its immunosuppressive effects at high doses, evidence suggests that low-to-moderate chronic exposure may paradoxically promote pro-inflammatory responses in the airways. Specifically, Hg exposure induces ROS, triggering immune responses such as B and T cell activation, increased immunoglobulin levels, and higher autoantibody production [61, 62]. Additionally, in individuals with susceptible genotypes, Hg exposure may stimulate the immune system, leading to the production of anti-cilia protein nucleolar antibodies and the deposition of systemic immune complexes [63].

Se exposure was identified as a potential protective factor for lung function in this study using multivariable linear regression and WQS analyses, although the positive association between the two was not maintained in the fully adjusted model. Fan et al. [29] used a cross-sectional study of 3520 adult individuals in the United States and found a U-shaped connection between Se exposure and lung function, with only moderate Se levels attenuating the harmful effects of Cd and Pb on lung function. Another study showed a significant correlation between selenium intake and lung function in asthmatics. It is recommended that asthmatics consume between 137.65ug and 200ug per day to improve lung function while avoiding the adverse effects of selenium overdose [64]. The protective mechanism of Se on lung function may be by inhibiting the formation of ROS and ameliorating mitochondrial dysfunction to antagonize oxidative stress [65, 66]. Another study suggested that selenium supplementation improved the quality of life in asthmatic patients but had limited impact on lung function [67]. This might be due to the fact that most of the participants were using inhaled corticosteroids, which could interfere with the effects of selenium supplementation. Furthermore, these studies focused exclusively on asthmatic populations, lacking insights into its effects on lung function in the general population. However, high concentrations of Se can induce endothelial cell dysfunction by releasing ROS and decreasing NO production [68]. The literature suggests that additional selenium intake may benefit individuals with low selenium levels, but for those with adequate to high levels, it could have detrimental effects, and selenium supplements should be avoided [69]. Therefore, the effects of Se on lung function may be two-fold, and appropriate selenium supplementation may help improve lung function.

Our study found that the positive association between serum Hg and FEV₁/FVC was more pronounced in females, whereas the positive association between serum Se and FEV₁/FVC was more significant in males. This may be attributed to differences in the bioaccumulation and detoxification of Hg, which could be influenced by sex hormones. It has been suggested that females generally have higher levels of glutathione (GSH) [70], which may enhance Hg clearance. Additionally, Se, as a key antioxidant element, exerts its protective effects mainly through the glutathione peroxidase (GPx) family and selenoprotein P, both of which play crucial roles in antioxidant defense and anti-inflammatory mechanisms. Some studies have indicated that GPx activity may be higher in males than in females [71], potentially allowing Se to more effectively mitigate oxidative stress and exert a stronger protective effect on lung function in males. In smokers, we observed significant interactions between serum Cd, Hg, Se, and both FeNO and FEV₁/FVC. This may be attributed to the presence of pre-existing airway inflammation and oxidative stress in smokers, which could modify their response to heavy metal exposure. Notably, we found a beneficial association between serum Se and lung function in current smokers. Previous studies have demonstrated that, in animal models exposed to waterpipe smoke, Se supplementation restores the activity of antioxidant enzymes such as superoxide dismutase (SOD), GPX-1, and catalase, thereby reducing lipid peroxidation and oxidative damage [72]. This mechanism may help explain the observed protective effect of Se on lung function in smokers in our study. Furthermore, The stratified analysis by BMI showed an interaction between serum Hg, serum Mn, serum Se, serum Pb, and FeNO. BMI may play a key regulatory role in the relationship between heavy metal exposure and FeNO by influencing oxidative stress levels, inflammatory responses, and the absorption and metabolism of heavy metals.

FeNO serves as a biomarker of eosinophilic inflammation, with elevated FeNO levels often indicating increased Type 2 (T2) inflammation [73]. This elevation may suggest a higher risk of exacerbation in T2-driven asthma or a subset of T2-dominant COPD [54]. However, both asthma and COPD encompass various phenotypes, some of which are not driven by T2 immune responses [74]. In such cases, FeNO levels may not be significantly elevated during acute exacerbations. Therefore, while FeNO is a valuable tool for assessing airway inflammation, its interpretation should be considered alongside other clinical and inflammatory markers to accurately characterize disease activity and guide management. Pulmonary function tests (PFTs) are the standard method for assessing respiratory health [75]. Obstructive lung diseases, such as asthma and COPD, are characterized by a reduced FEV₁/FVC ratio due to airflow limitation, where FEV₁ decreases while FVC tends to remain relatively preserved [9, 76]. Restrictive lung diseases, such as pulmonary fibrosis and interstitial lung diseases, are primarily characterized by a reduction in FVC, with a normal or even increased FEV₁/FVC ratio [77]. In these conditions, FeNO levels may remain unchanged and may not serve as a reliable marker for disease progression [78].

There are several drawbacks to this study. Firstly, this study is a cross-sectional study and cannot assess the causal relationship between heavy metals, airway inflammation and lung function, and more cohort studies are needed to verify the temporal causality between them. Secondly, only a single indicator, FeNO, reflects the level of airway inflammation, and a combination of multiple indicators can better reflect the actual airway inflammation [79]. Thirdly, this study adjusted for as many confounders as possible but did not include other potential confounders not measured by NHANES, such as ambient air pollution and genetic factors. Fourth, the absence of data on fine particulate matter (PM2.5 ) exposure limits our ability to fully adjust for air pollution as a potential confounder in this study. Finally, this study did not employ weighted analysis, which has previously been demonstrated to induce an over-adjustment bias when the variables used to calculate sampling weights are included in the model control variables. Despite these shortcomings, this study also has the following strengths. To our knowledge, this study is the first to comprehensively assess the association of multiple heavy metal exposures with airway inflammation and lung function in adults. Secondly, this study utilized multiple statistical methods and a large nationally representative sample of data to explore their relationship, identifying the major contributors to the metal mixture.

In future research, we aim to explore the potential reverse causality between airway inflammation and heavy metal absorption. Although there is currently no direct literature supporting the hypothesis that airway inflammation promotes the absorption of heavy metals, this remains a complex and intriguing area of study. Given that this study is a large-scale cross-sectional analysis, future studies may involve longitudinal designs or methodologies like Mendelian Randomization (MR) to further clarify the directionality of these relationships. Given the practical and methodological considerations, such as hydration status and renal function, urinary metal concentrations were not included in this study. However, future research will incorporate different exposure indicators, including urinary metal concentrations, to further investigate their potential impact on lung function. This study suggests a preliminary positive association between selenium exposure and better lung function. However, this association did not hold in the fully adjusted model. Therefore, prospective studies are needed to better clarify selenium’s role in lung function and its potential therapeutic benefits. Additionally, incorporating air pollution exposure data, such as PM2.5, in future studies could help better account for its potential confounding effects. Research could also explore the combined impact of heavy metal exposure and air pollution on lung function in diverse populations. Finally, large-scale longitudinal and cohort studies should be conducted to establish clinical safety exposure limits and further clarify the relationship between heavy metal exposure levels and changes in lung function over time.

Conclusions

We found inconsistent associations of different heavy metals with lung function and airway inflammation. Overall, Cd exposure was the most significant negative factor for both lung function and airway inflammation, Pb exposure was also negatively associated with lung function, while Hg exposure was the most important positive associated with airway inflammation, and Se contributed the strongest positive weight to FEV₁ and FEV₁/FVC in the WQS analysis. Therefore, we recommend that enhanced pulmonary function monitoring for populations with Cd and Pb exposure may facilitate early identification of potential respiratory impairment. Populations with significant Hg exposure should receive prioritized monitoring for airway inflammation. Given the ongoing controversy regarding Se’s effects on respiratory health, particularly concerning dosage variations and population-specific responses. Future studies should employ more prospective methodologies to clarify these relationships. These findings underscore the imperative for strengthened environmental regulation of heavy metal exposure.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Abbreviations

ATS

American Thoracic Society

BKMR

Bayesian kernel machine regression

BMI

Body Mass Index

COPD

Chronic obstructive pulmonary disease

COX-2

Cyclooxygenase-2

ERK

Extracellular signal-regulated kinase

FEV1

Forced expiratory volume in 1

FVC

Forced vital capacity

GPx

Glutathione peroxidase

GSH

Glutathione

FeNO

Fractional exhaled nitric oxide

ICP-MS

Inductively coupled plasma mass spectrometry

MR

Mendelian Randomization

NADPH

Nicotinamide adenine dinucleotide phosphate

NCHS

National Center for Health Statistics

NHANES

National Health and Nutrition Examination Survey

PEF

Peak expiratory flow

PFTs

Pulmonary function tests

PM2.5

Fine particulate matter

ROS

Reactive oxygen species

SOD

Superoxide dismutase

T2

Type 2

WQS

Weighted quantile sum

Cd

Cadmium

Hg

Mercury

Mn

Manganese

Pb

Lead

Sb

Stibium

Se

Selenium

Author contributions

ZJ: Conceptualization, Project administration, Data curation, Methodology, Formal analysis, Software, Visualization, Writing– original draft; WS: Supervision, Project administration, Validation, Writing– review & editing; JH: Supervision, Validation, Writing– review & editing; GW: Conceptualization, Supervision, Project administration, Validation, Writing– review & editing. All authors contributed to the article and approved the submitted version.

Funding

This study was supported by the Capital Health Development Special Research Program (No.2022-1G-4073) and the Beijing Municipal Science and Technology Program Project: Research and Translational Application of Capital Clinical Specialty Diagnosis and Treatment Technologies (Z221100007422040).

Data availability

The survey data are publicly available on the internet for data users and researchers throughout the world ( www.cdc.gov/nchs/nhanes/).

Declarations

Ethics approval and consent to participate

The study involving human participants was carried out by the Helsinki Declaration and was approved by the NCHS Ethics Review Board. Patients/participants provided written informed consent to participate in this study.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.