https://orcid.org/0000-0002-1797-2047
Figures
Abstract
Background
Further evidence is required regarding the influence of metal mixture exposure on mortality. Therefore, we employed diverse statistical models to evaluate the associations between eight urinary metals and the risks of all-cause and cardiovascular mortality.
Methods
We measured the levels of 8 metals in the urine of adults who participated in the National Health and Nutrition Examination Survey (NHANES) from 1999 to 2018. Based on follow-up data, we determined whether they died and the reasons for their deaths. We estimated the association between urine metal exposure and all-cause mortality using Cox regression, weighted quantile sum (WQS) regression, and Bayesian kernel machine regression (BKMR) models. Additionally, we used a competing risk model to estimate the relationship between metal exposure and cardiovascular mortality.
Results
Among the 14,305 individuals included in our final analysis, there were 2,066 deaths, with 1,429 being cardiovascular-related. Cox regression analysis showed that cobalt (Co) (HR: 1.21; 95% CI: 1.13, 1.30) and antimony (Sb) (HR: 1.26; 95% CI: 1.12, 1.40) were positively associated with all-cause mortality (all P for trend <0.001). In the competing risk model, Co (HR: 1.29; 95% CI: 1.12, 1.48), lead (Pb) (HR: 1.18; 95% CI: 1.03, 1.37), and Sb (HR: 1.44; 95% CI: 1.18, 1.75) were significantly associated with an increased risk of cardiovascular mortality (all P for trend <0.001). Sb, Pb, cadmium (Cd), and molybdenum (Mo) had the highest weight rankings in the final WQS model. All metals showed a complex non-linear relationship with all-cause mortality, with high posterior inclusion probabilities (PIPs) in the final BKMR models.
Citation: Cheng T, Yu D, Li G, Chen X, Zhou L, Wen Z (2024) Association between exposure to urinary metal and all-cause and cardiovascular mortality in US adults. PLoS ONE 19(12): e0316045. https://doi.org/10.1371/journal.pone.0316045
Editor: Neftali Eduardo Antonio-Villa, Instituto Nacional de Cardiologia Ignacio Chavez, MEXICO
Received: March 19, 2024; Accepted: December 4, 2024; Published: December 27, 2024
Copyright: © 2024 Cheng et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: The data for this study can be obtained from the NHANES official website (https://wwwn.cdc.gov/nchs/nhanes/Default.aspx).
Funding: This research was funded by the National Key Technology R&D Program for the 12th Five-year Plan of Ministry of Science and Technology, China (No. 2013BAI02B10).
Competing interests: none.
Introduction
Metal pollution is considered one of the most common byproducts of industrialization and urbanization. The general public can be exposed to various metals through mediums such as water, air, soil, and food [1]. An increasing amount of epidemiological evidence suggests that exposure to metals increases the risk of various diseases, such as cardiovascular disease, arteriosclerosis, infertility, and kidney stones [2–5]. In addition, some studies have found that the levels of metals in blood and urine in the human body can partly predict the risk of all-cause, cardiovascular, and cancer mortality [6–9]. Most of the early literature is limited to analyzing the effects of individual chemicals on human health or death [10,11]. The effects of different metal exposures on the human body may be cumulative, synergistic, antagonistic, or indirectly complex, requiring more evidence to support each other [12]. Even though a few studies have begun to focus on the effects of metal mixtures on various diseases [12–15], however, extending it to survival analysis is rare. Additionally, due to competitive risk bias, traditional survival analysis methods used in the early stages may overestimate the impact of metals on specific mortality risk [16,17]. The different exposure patterns, high correlations, and complex interactions of metals in the environment require more statistical strategies to evaluate their combined effects [18]. Currently, many new statistical methods have been successfully introduced into the field of environmental health [19,20], previous studies exploring mixture exposure have shown that combining multiple methods can lead to more comprehensive and reliable conclusions [7]. It is difficult to establish a set of standard criteria for determining which statistical method is most suitable for a specific study. Therefore, there is still insufficient evidence on the combined effects of different metals on mortality.
This study employed data from the National Health and Nutrition Examination Survey (NHANES) carried out from 1999 to 2018 to evaluate the connection between the presence of eight heavy metals in the urine and the all-cause mortality rate for the general population in the United States (US). We employed three different models, specifically Cox regression, weighted quantile sum (WQS) regression, and Bayesian kernel machine regression (BKMR), to assess this association. Moreover, we examined the link between metal exposure and the risk of cardiovascular mortality using a competing risk model. Subsequently, we interpreted the results from multiple models jointly to further explore the relationship between mixed metal exposure and mortality risk. These findings may offer novel suggestions for longitudinal epidemiological and experimental investigations.
Materials and methods
Study population
NHANES is an ongoing national cross-sectional survey, and its data can be found on the website of the Centers for Disease Control and Prevention (CDC) in the US (https://wwwn.cdc.gov/nchs/nhanes/Default.aspx) [21]. The research protocol has been approved by the Research Ethics Review Committee of the National Center for Health Statistics (NCHS). All participants provided written consent at the time of recruitment. In this study, we included 101316 data from 10 survey cycles conducted between 1999 and 2018. Exclusion criteria include: (a) age < 20 years; (b) pregnancy status; (c) loss to follow-up; (d) missing data of any covariate. A total of 14,305 participants were ultimately included in the final analysis (Fig 1). As this study is a retrospective analysis and poses no risk of exposure to personally identifiable information, no additional ethical review or informed consent is required.
Measurement of urinary metal and covariation assessment
The data for 8 urine metals, namely cadmium (Cd), cobalt (Co), lead (Pb), barium (Ba), antimony (Sb), cesium (Cs), molybdenum (Mo), and thallium (Tl), were obtained from NHANES 1999–2018. These metals in urine samples were primarily measured by using the Inductively Coupled Plasma Mass Spectrometry (ICP-MS). The final metal concentrations used in the analysis were obtained by dividing the metal concentration (μg/L) by the urine creatinine concentration (mg/dL) [22].
We have taken into account the covariates that are most likely to affect metal exposure and mortality rates, including gender, age, race, body mass index (BMI), marital status, education level, family poverty income ratio (PIR), smoking, alcohol consumption, health insurance [12,23]. PIR refers to the ratio of household income to the federal poverty line, used to measure the relationship between household income and poverty level [24]. Smoking is defined as having smoked at least 100 cigarettes in one’s lifetime. According to the categorization of alcohol consumption levels by the United States Department of Agriculture, drinking is divided into five categories: never, former, light, moderate, and heavy [25]. We also include self-reported cardiovascular diseases (if including any of congestive heart failure, coronary heart disease, angina or myocardial infarction), stroke (yes or no), cancer (if including any type of cancer), hypertension (yes or no) and diabetes (yes or no).
Outcomes
The public use mortality files from NHANES offer mortality data recorded in the survey registration up to December 31, 2019. These files enable tracking of participant deaths and identification of specific causes. Starting from 1998, the death codes adhere to the guidelines of the 10th edition of the International Classification of Diseases, Injuries, and Causes of Death (ICD-10) [26]. Determine alive or deceased based on the variable MORTSTAT (final mortality status). The mortality specifically related to cardiovascular diseases (coded I00-I09, I11, I13 and I20-I51) was identified based on the ICD-10. For participants who did not experience the event during the study period, their follow-up time was considered right-censored. Similarly, all missing follow-up data were treated as right-censored.
Statistical analysis
All analyses were conducted using R 4.2 (the R Foundation for Statistical Computing, Vienna, Austria). Chi-square tests or t-tests were used to assess participants’ baseline characteristics. Categorical variables were represented as numbers (n) and percentages (%), while continuous variables were represented as means and the 25th and 75th percentiles (P25, P75). This study used creatinine-corrected metal concentrations to analyze the relationship between heavy metal exposure levels and mortality risk, in order to reduce the impact of urine dilution on the measurements [12]. Due to the severely skewed concentrations of these eight chemicals, the standardized data were ln-transformed when treated as continuous variables to approximate a normal distribution. Firstly, we analyzed the Pearson correlation coefficients between the ln-transformed concentrations of the eight metals and the trends of individual metal concentrations in each survey period.
Only Co, Mo, Pb, and Sb passed the proportional hazards (PH) assumption (S1 Table). Cox regression analysis was used to examine the relationship between individual metals and all-cause mortality, and individual metals were categorized into four quartiles (i.e., were split by P25, P50 and P75 into Q1, Q2, Q3, and Q4, respectively) to calculate Hazard Ratio (HR) and P for trend.
Competing risks refer to events that may hinder the occurrence of the primary event of interest, as these events can prevent the occurrence of the event of interest or alter the probability of the primary endpoint occurring [16]. For example, when studying cardiovascular mortality, deaths due to non-cardiovascular causes are considered competing risks. When estimating the incidence of the target outcome, analysts should use the cumulative incidence function [27]. Considering that cardiovascular mortality may be masked by other causes, a competing risk model was used to assess the impact of individual metals on the risk of cardiovascular death.
The weighted quantile sum (WQS) regression model constructs a weighted index ranging from 0 to 1, reflecting the contribution of each metal in the mixture [28]. It is used to estimate the combined impact of all environmental chemical exposures on the outcome and can be employed to test the association between the index and the dependent variable or outcome [29]. Consequently, the relative strength of weights assigned to each variable in the model allows for the assessment of each environmental chemical’s contribution to the overall index effect, thus identifying important chemicals within the mixture [12]. It is important to note that the WQS regression limits the effect of a single metal on mortality risk to one direction. In this study, we used the R package “gWQS” with family = binomial to construct the WQS index for heavy metal exposure, analyzing the combined impact of metal mixture components on all-cause risk, and used 40% of the data as a training set and the remaining data as a test set [30]. We performed 1000 bootstrap samples in the multivariable regression model and applied regularization constraints in the optimization function for weight estimation.
To evaluate the combined impact of metal mixtures on all-cause mortality, we used the “Bayesian kernel machine regression (BKMR)” package and adopted default priors for 10,000 iterations [31]. We analyzed the following three exposure-response functions: 1) Univariate exposure-response function: Assessed the impact of individual metal exposure on all-cause mortality risk. 2) Cumulative effect at different percentiles: Compared the cumulative effect of metal mixtures at different percentiles on cardiovascular disease (CVD) risk. 3) Bivariate exposure-response function: Evaluated the impact of one metal on all-cause mortality risk while fixing the other metals at different percentiles [12,28]. Additionally, we calculated the posterior inclusion probability (PIP) for each metal to identify those with the most significant impact on all-cause mortality risk.
For all statistical methods mentioned above, we progressively built four models to assess the potential confounding effects of different covariate combinations. Model 0 was a crude model, Model 1 adjusted for age, sex, and race. Model 2 additionally considered marital status, PIR, education, BMI, smoking, drinking, and health insurance. In Model 3, we further adjusted for the combined presence of diabetes, stroke, cancer, CVD, and hypertension. A two-sided P-value less than 0.025 was considered statistically significant.
Results
Population characteristics
During the 10 survey cycles, a total of 14,305 participants were included, with 2,066 deaths recorded as all causes (Table 1). Among these deaths, 1,429 were due to cardiovascular reasons. The median follow-up time was 101 months, with the longest follow-up period being 249 months. The age range of the participants was between 20 and 85 years, with males accounting for 51.61% of the total. The majority of deaths due to all causes were in individuals over 60 years old, male, white ethnicity and of highly educated. The cumulative incidence rates of all-cause and cardiovascular deaths can be seen in S1 Fig.
Measurement of urinary metals and their correlation
The detection rates of metal content in urine are shown in S2 Table. Except for Sb, the detection rates of other metals are greater than 90%. The Pearson coefficients for the Ln transformation of metals show a moderate correlation between Cs and Tl (r = 0.58), while the correlations with other metals are relatively weak (S2 Fig). The concentration trends of individual metals during each survey period were reported in S3 Fig and S3 Table.
Assess the association between urinary metal and all-cause and cardiovascular mortality
We used Cox regression analysis to examine the relationship between the levels of 4 urinary metals and the risk of all-cause mortality, as shown in Tables 2 and S4. In the crude model, Co, Mo, Pb and Sb were associated with an increased risk of all-cause mortality (all P for trend <0.001). After adjusting for all covariates in Model 3, Co (HR: 1.21; 95% CI: 1.13, 1.30) and Sb (HR: 1.26; 95% CI: 1.12, 1.40) remained positively associated with all-cause mortality (all P for trend <0.001).
Table 3 presents the results of the competing risk model analyzing the relationship between the levels of 8 urinary metals and cardiovascular mortality. There were a total of 552 cardiovascular deaths. In the crude model, 5 metals (Cd, Co, Mo, Pb, Sb) were each positively associated with cardiovascular mortality (all P for trend <0.001), except for Ba, Tl, and Cs. After adjusting for all covariates in Model 3, Co (HR: 1.29; 95% CI: 1.12, 1.48), Pb (HR: 1.18; 95% CI: 1.03, 1.37), and Sb (HR: 1.44; 95% CI: 1.18, 1.75) remained significantly associated with an increased risk of cardiovascular mortality (all P for trend <0.001). These findings suggest that higher levels of Co, Pb, and Sb are robustly linked to an elevated risk of cardiovascular mortality, even when accounting for other competing risks and covariates.
The association between urinary metal and all-cause mortality by WQS regression model
The Model 3 of WQS analysis showed a positive association between metal mixtures (HR: 2.23; 95%CI: 2.00, 2.48) and all-cause mortality (Table 4). The weighted indices of the four metals (Sb, Cd, Pb, and Mo) in WQS are 0.356, 0.297, 0.262, and 0.080, respectively (Table 5). This result remained stable across Models 0, 1, and 2 (Fig 2).
(Ba: Barium; Cd: Cadmium; Co: Cobalt; Cs: Cesium; Mo: Molybdenum; Pb: Lead; Sb: Antimony; TI: Thallium; A: Adjusted for none; B: Adjusted for age, sex, race; C: Adjusted for age, sex, race, BMI, marital status, PIR, education, smoking, drinking, and health insurance; D: Adjusted for age, sex, race, BMI, marital status, PIR, education, smoking, drinking, health insurance, diabetes, stroke, cancer, CVD, and hypertension).
BKMR model to assess the association between urinary metal and all-cause mortality
In the final model, the relationship between all metals and all-cause mortality rate shows a complex non-linear relationship, with positive association within a certain range (Fig 3). Compared to the previous three models, it can be inferred that covariates have influenced the relationship between metal mixtures and all-cause mortality to some extent. We analyzed the overall impact of metal mixtures on all-cause mortality rate. Compared to the 50th percentile, when all metal mixtures are at or above the 40th percentile, there is an increasing trend in all-cause mortality rate (Fig 4). The posterior inclusion probabilities (PIPs) indicate that each metal is associated with all-cause mortality (Table 5).
(Ba: Barium; Cd: Cadmium; Co: Cobalt; Cs: Cesium; Mo: Molybdenum; Pb: Lead; Sb: Antimony; TI: Thallium; A: Adjusted for none; B: Adjusted for age, sex, race; C: Adjusted for age, sex, race, BMI, marital status, PIR, education, smoking, drinking, and health insurance; D: Adjusted for age, sex, race, BMI, marital status, PIR, education, smoking, drinking, health insurance, diabetes, stroke, cancer, CVD, and hypertension).
(A: Adjusted for none; B: Adjusted for age, sex, race; C: Adjusted for age, sex, race, BMI, marital status, PIR, education, smoking, drinking, and health insurance; D: Adjusted for age, sex, race, BMI, marital status, PIR, education, smoking, drinking, health insurance, diabetes, stroke, cancer, CVD, and hypertension).
Discussion
In this research, we conducted a comprehensive comparison of the outcomes obtained from various statistical methods in order to assess the influence of 8 urinary metals on the mortality rate among the general population of the US. Firstly, Cox regression analysis showed a positive association between between Co and Sb with the all-cause mortality rate. Competing risk models provided additional evidence of a positive association between Co, Sb, Pb, and cardiovascular mortality. Sb, Pb, Cd, and Mo had the highest weight rankings in the final WQS model. All metals showed a complex non-linear relationship with all-cause mortality, with high PIPs in the final BKMR models.This study fully considered the strengths and limitations of different statistical methods and comprehensively evaluated the combined impact of urinary metals on mortality.
We observed that the detection rates of most metals increased over time, with all metals showing generally high detection rates except for Sb. Despite the lower detection rate of antimony, we found a significant association between it and mortality. This may suggest that antimony, even at low concentrations, can still have a notable impact on health, potentially due to its toxicity, bioaccumulation effects, or specific effects on certain populations. Therefore, the health risks of antimony should not be overlooked. We found an interaction Sb and certain years with respect to all-cause mortality (S5 Table). This suggests that the health impact of Sb may be modulated by temporal factors. Variations in Sb emissions, differences in detection technologies and methods, as well as socioeconomic factors and policy interventions in specific years may all influence this interaction.
Cox regression is most commonly used to assess the impact of chemical substances on all-cause and cardiovascular mortality. Typically, these analyses involve one or a group of similar chemical substances for ease of interpretation. However, metal exposures in real environments are not singular, and the interactions between mixed exposures may result in false positive or false negative results. After adjusting for all potential confounders, we found a positive association between Co, and Sb with overall mortality, which is consistent with previous research findings [32]. The epidemiological research on the relationship between adult cardiovascular diseases and environmental metals (such as Ba, Sb, etc.) is receiving increasing attention [33]. Some scholars have proposed that the traditional Cox regression overestimates the cumulative incidence rate when evaluating the impact of metals on cardiovascular mortality [34], so this analysis introduced a competitive risk model. We found that cobalt (Co), lead (Pb), and antimony (Sb) are significantly associated with an increased risk of cardiovascular mortality. The competing risk model has a distinct advantage over Cox regression in multiple potential events studies. It can accurately estimate the risk of a specific event while accounting for the impact of other competing events, thereby providing a more realistic risk assessment [16,27]. Previous literature has confirmed that the proportion of Cd and Sb in urine metal mixtures contributes the most to the incidence of cardiovascular diseases [35]. Sb is positively correlated with mean platelet volume (MPV), which contributes to thrombus formation and increases the risk of cardiovascular disease [1]. Additionally, literature has found that Co, Sb, and Pb have a significant impact on cardiovascular mortality rate [6]. The above results were consistent with our competition risk model results.
WQS analysis can better explain the complex impact of exposure on health in real life by considering the weight and correlation of chemical substances [30]. In our analysis, Sb, Cd, Pb, and Mo are considered to have the highest impact weight on overall mortality. On one hand, this may be due to differences in sample size, which results in a slight difference from the previous WQS regression analysis that mentioned cadmium having the highest weight in the mixture [6]. On the other hand, it has been demonstrated that Pb and Mo are significantly associated with all-cause mortality in the WQS regression model, possibly due to interactions with other metals. However, this association is not observed in the Cox regression model. This indicates that WQS analysis may be more sensitive in identifying important factors than individual analyses, but the standard WQS without penalty terms has limitations in simultaneously evaluating the joint effects of chemicals with different directions of action [36]. High Pb exposure has been confirmed to be positively correlated with the all-cause and cancer mortality [37,38]. Mo also has an independent and comprehensive correlation with all-cause mortality rate [7]. The results of this study once again provide additional evidence for the impact of mixed metal exposure on mortality.
The BKMR analysis can flexibly estimate multivariable exposure-response functions, identifying hidden complex patterns and non-linear relationships [31]. Our final model revealed a complex non-linear relationship between 8 types of metal substances and overall mortality rate, with a positive correlation within a certain range. Due to the introduction of kernel functions, the calculation complexity of the BKMR model is high, leading to slower computation speed and higher requirements for computer configuration. It is necessary to choose appropriate kernel functions and related parameters based on actual situations in order to ensure better interpretability of the model.
This study demonstrates that traditional survival analysis can provide a simple relationship between individual metals and mortality. When considering competing biases, the competing risk model is an alternative additional strategy [39]. The WQS model can explore the impact of mixed exposures on the target outcome [40]. The BKMR model can explore the exposure-response effects of each substance and the interactions between substances [41]. Therefore, these models have their own advantages and disadvantages, and their combined application is beneficial for a comprehensive assessment of the effects of mixed substances on the outcome. The main findings of this study are summarized in Table 5, which can provide a reference for future related research.
Our analysis has some limitations. The assumption underlying our analysis is that the sample based on inclusion analysis represents the adult population of the US. Nevertheless, it remains unclear whether the WQS and BKMR models can be applied under this assumption. In addition, we referred to previous designs that used unweighted data as unweighted estimates are preferable to weighted estimates if the covariates used to calculate sample weights are already included in the regression model [42]. Another limitation is the use of cross-sectional survey data. The measured metal levels in this study can only reflect recent exposure, and long-term follow-up is needed to determine the actual impact of metals on mortality outcomes. Additionally, we did not discuss the relationship between mixed metals and cancer mortality. Furthermore, we cannot establish a causal relationship between mixed metals and mortality. To gather more evidence, it will be necessary to conduct large-scale prospective cohort studies or experimental studies in the future.
Conclusion
Combining all models, it is possible that Sb may have a more stable impact on overall mortality and cardiovascular mortality. We recommend using multiple methods to comprehensively explain the effects of mixtures, meaningful metal effects in individual statistical models still require careful attention.
Supporting information
S1 Fig. Cumulative incidence of all-cause and cardiovascular death events.
https://doi.org/10.1371/journal.pone.0316045.s001
(DOCX)
S2 Fig. Pairwise Pearson correlations between 8 metal concentrations.
https://doi.org/10.1371/journal.pone.0316045.s002
(DOCX)
S3 Fig. Concentration trends of 8 metals (ln-transformed) during each survey period.
https://doi.org/10.1371/journal.pone.0316045.s003
(DOCX)
S1 Table. Proportional hazard assumption in the Cox model.
https://doi.org/10.1371/journal.pone.0316045.s004
(DOCX)
S2 Table. Distributions of metals in the study population.
https://doi.org/10.1371/journal.pone.0316045.s005
(DOCX)
S3 Table. The relationship between 8 metals and survey period.
https://doi.org/10.1371/journal.pone.0316045.s006
(DOCX)
S4 Table. The goodness-of-fit metrics for the final Cox model.
https://doi.org/10.1371/journal.pone.0316045.s007
(DOCX)
S5 Table. Pb*cycle interaction in the crude Cox regressions.
https://doi.org/10.1371/journal.pone.0316045.s008
(DOCX)
S1 Data. This file contains the raw data used for statistical analysis in the manuscript.
https://doi.org/10.1371/journal.pone.0316045.s009
(CSV)
Acknowledgments
We express our gratitude for the NHANES official website’s provision of publicly available and free data support.
References
- 1. Liu W, Yu L, Ye Z, Wang X, Qiu W, Tan Q, et al. Assessment for the associations of twenty-three metal(loid)s exposures with early cardiovascular damage among Chinese urban adults with five statistical methods: Insight into assessing health effect of multipollutant exposure. Chemosphere. 2022;307(Pt 2):135969. Epub 2022/08/09. pmid:35940407.
- 2. Wan Z, Wu M, Liu Q, Fan G, Fang Q, Qin X, et al. Association of metal exposure with arterial stiffness in Chinese adults. Ecotoxicology Environmental Safety. 2023;257:114921. Epub 2023/04/21. pmid:37080131.
- 3. Lin J, Lin X, Qiu J, You X, Xu J. Association between heavy metals exposure and infertility among American women aged 20–44 years: A cross-sectional analysis from 2013 to 2018 NHANES data. Frontiers Public Health. 2023;11:1122183. Epub 2023/03/04. pmid:36866101.
- 4. Ye Z, Chen Z, Luo J, Xu L, Fan D, Wang J. National analysis of urinary cadmium concentration and kidney stone: Evidence from NHANES (2011–2020). Frontiers Public Health. 2023;11:1146263. Epub 2023/04/04. pmid:37006548.
- 5. Pan Z, Gong T, Liang P. Heavy Metal Exposure and Cardiovascular Disease. Circulation Research. 2024;134(9):1160–78. Epub 2024/04/25. pmid:38662861.
- 6. Duan W, Xu C, Liu Q, Xu J, Weng Z, Zhang X, et al. Levels of a mixture of heavy metals in blood and urine and all-cause, cardiovascular disease and cancer mortality: A population-based cohort study. Environmental Pollution. 2020;263(Pt A):114630. Epub 2021/02/24. pmid:33618481.
- 7. Zhao S, Wang S, Yang X, Shen L. Dose-response relationship between multiple trace elements and risk of all-cause mortality: a prospective cohort study. Frontiers Nutrition. 2023;10:1205537. Epub 2023/08/03. pmid:37533572.
- 8. Coradduzza D, Congiargiu A, Azara E, Mammani IMA, De Miglio MR, Zinellu A, et al. Heavy metals in biological samples of cancer patients: a systematic literature review. Biometals. 2024;37(4):803–17. Epub 2024/02/13. pmid:38347295.
- 9. Ayuso-Álvarez A, Nuñez O, Martín-Méndez I, Bel-Lán A, Tellez-Plaza M, Pérez-Gómez B, et al. Metal and metalloid levels in topsoil and municipal cardiovascular mortality in Spain. Environmental Research. 2022;204(Pt D):112395. Epub 2021/11/21. pmid:34800529.
- 10. Swaddiwudhipong W, Mahasakpan P, Limpatanachote P, Krintratun S. An association between urinary cadmium and urinary stone disease in persons living in cadmium-contaminated villages in northwestern Thailand: a population study. Environmental Research. 2011;111(4):579–83. Epub 20110126. pmid:21272865.
- 11. Scinicariello F, Buser MC, Feroe AG, Attanasio R. Antimony and sleep-related disorders: NHANES 2005–2008. Environmental Research. 2017;156:247–52. Epub 20170328. pmid:28363141.
- 12. Guo X, Li N, Wang H, Su W, Song Q, Liang Q, et al. Combined exposure to multiple metals on cardiovascular disease in NHANES under five statistical models. Environmental Research. 2022;215(Pt 3):114435. Epub 20220926. pmid:36174761.
- 13. Zhao S, Fan L, Wang Y, Dong S, Han M, Qin Y, et al. Combined exposure to multiple metals on hypertension in NHANES under four statistical models. Environmental Science Pollution Research International. 2023;30(40):92937–49. Epub 2023/07/27. pmid:37498425.
- 14. Wang X, Han X, Guo S, Ma Y, Zhang Y. Associations between patterns of blood heavy metal exposure and health outcomes: insights from NHANES 2011–2016. BMC Public Health. 2024;24(1):558. Epub 2024/02/23. pmid:38389043.
- 15. Shi X, Wang X, Zhang J, Dang Y, Ouyang C, Pan J, et al. Associations of mixed metal exposure with chronic kidney disease from NHANES 2011–2018. Scientific Reports. 2024;14(1):13062. Epub 2024/06/07. pmid:38844557.
- 16. Dutz A, Löck S. Competing risks in survival data analysis. Radiotherapy And Oncology. 2019;130:185–9. Epub 20181009. pmid:30314718.
- 17. Tran TS, Bliuc D, Blank RD, Center JR, Nguyen TV. Fracture risk assessment in the presence of competing risk of death. Osteoporosis International. 2024;35(11):1989–98. Epub 2024/08/15. pmid:39145778.
- 18. Zhu G, Wen Y, Cao K, He S, Wang T. A review of common statistical methods for dealing with multiple pollutant mixtures and multiple exposures. Frontiers Public Health. 2024;12:1377685. Epub 20240509. pmid:38784575.
- 19. Liang YJ, Rong JH, Wang XX, Cai JS, Qin LD, Liu QM, et al. Correlation between Combined Urinary Metal Exposure and Grip Strength under Three Statistical Models: A Cross-sectional Study in Rural Guangxi. Biomedical Environmental Sciences. 2024;37(1):3–18. Epub 2024/02/08. pmid:38326717.
- 20. Sacks D, Baxter B, Campbell BCV, Carpenter JS, Cognard C, Dippel D, et al. Multisociety Consensus Quality Improvement Revised Consensus Statement for Endovascular Therapy of Acute Ischemic Stroke. International Journal Of Stroke. 2018;13(6):612–32. Epub 2018/05/23. pmid:29786478.
- 21. Ren Z, Xue Y, Zhang H, Zhang T, Wang M, Guo T, et al. Association between probiotic consumption and periodontitis: Evidence from NHANES 2009–2014. Journal Of Clinical Periodontology. 2023;50(11):1476–86. Epub 2023/08/19. pmid:37596796.
- 22. Zhang Y, Dong T, Hu W, Wang X, Xu B, Lin Z, et al. Association between exposure to a mixture of phenols, pesticides, and phthalates and obesity: Comparison of three statistical models. Environment International. 2019;123:325–36. Epub 2018/12/18. pmid:30557812.
- 23. Shi L, Yuan Y, Xiao Y, Long P, Li W, Yu Y, et al. Associations of plasma metal concentrations with the risks of all-cause and cardiovascular disease mortality in Chinese adults. Environment International. 2021;157:106808. Epub 20210806. pmid:34365319.
- 24. Fadeyev K, Nagao-Sato S, Reicks M. Nutrient and Food Group Intakes among U.S. Children (2–5 Years) Differ by Family Income to Poverty Ratio, NHANES 2011–2018. International Journal Of Environmental Research And Public Health. 2021;18(22). Epub 2021/11/28. pmid:34831692.
- 25. Rattan P, Penrice DD, Ahn JC, Ferrer A, Patnaik M, Shah VH, et al. Inverse Association of Telomere Length With Liver Disease and Mortality in the US Population. Hepatology Communications. 2022;6(2):399–410. Epub 2021/09/25. pmid:34558851.
- 26. Song Y, Chang Z, Jia L, Song W, Wang H, Dong Q, et al. Better adherence to the MIND diet is associated with lower risk of all-cause death and cardiovascular death in patients with atherosclerotic cardiovascular disease or stroke: a cohort study from NHANES analysis. Food & Function. 2023;14(3):1740–9. Epub 2023/02/02. pmid:36722931.
- 27. Dignam JJ, Zhang Q, Kocherginsky M. The use and interpretation of competing risks regression models. Clinical Cancer Research. 2012;18(8):2301–8. Epub 20120126. pmid:22282466.
- 28. Ma Y, Hu Q, Yang D, Zhao Y, Bai J, Mubarik S, et al. Combined exposure to multiple metals on serum uric acid in NHANES under three statistical models. Chemosphere. 2022;301:134416. Epub 20220428. pmid:35490746.
- 29. Loftus CT, Bush NR, Day DB, Ni Y, Tylavsky FA, Karr CJ, et al. Exposure to prenatal phthalate mixtures and neurodevelopment in the Conditions Affecting Neurocognitive Development and Learning in Early childhood (CANDLE) study. Environment International. 2021;150:106409. Epub 2021/02/09. pmid:33556913.
- 30. Carrico C, Gennings C, Wheeler DC, Factor-Litvak P. Characterization of Weighted Quantile Sum Regression for Highly Correlated Data in a Risk Analysis Setting. Journal Agricultuarl Biological Environmental Statistics. 2015;20(1):100–20. Epub 20141224. pmid:30505142.
- 31. Bobb JF, Claus Henn B, Valeri L, Coull BA. Statistical software for analyzing the health effects of multiple concurrent exposures via Bayesian kernel machine regression. Environmental Health. 2018;17(1):67. Epub 2018/08/22. pmid:30126431.
- 32. Fan Y, Tao C, Li Z, Huang Y, Yan W, Zhao S, et al. Association of Endocrine-Disrupting Chemicals with All-Cause and Cause-Specific Mortality in the U.S.: A Prospective Cohort Study. Environmental Science & Technology. 2023;57(7):2877–86. Epub 2023/02/03. pmid:36728834.
- 33. Nigra AE, Ruiz-Hernandez A, Redon J, Navas-Acien A, Tellez-Plaza M. Environmental Metals and Cardiovascular Disease in Adults: A Systematic Review Beyond Lead and Cadmium. Current Environmental Health Reports. 2016;3(4):416–33. Epub 2016/10/27. pmid:27783356.
- 34. Hageman SHJ, Dorresteijn JAN, Pennells L, van Smeden M, Bots ML, Di Angelantonio E, et al. The relevance of competing risk adjustment in cardiovascular risk prediction models for clinical practice. European Journal Of Preventive Cardiology. 2023. Epub 2023/06/20. pmid:37338108.
- 35. Domingo-Relloso A, Grau-Perez M, Briongos-Figuero L, Gomez-Ariza JL, Garcia-Barrera T, Dueñas-Laita A, et al. The association of urine metals and metal mixtures with cardiovascular incidence in an adult population from Spain: the Hortega Follow-Up Study. International Journal Of Epidemiology. 2019;48(6):1839–49. Epub 2019/07/23. pmid:31329884.
- 36. Renzetti S, Gennings C, Calza S. A weighted quantile sum regression with penalized weights and two indices. Frontiers Public Health. 2023;11:1151821. Epub 2023/08/03. pmid:37533534.
- 37. Rhee J, Graubard BI, Purdue MP. Blood lead levels and lung cancer mortality: An updated analysis of NHANES II and III. Cancer Medicine. 2021;10(12):4066–74. Epub 2021/05/09. pmid:33963676.
- 38. Yao X, Steven Xu X, Yang Y, Zhu Z, Zhu Z, Tao F, et al. Stratification of population in NHANES 2009–2014 based on exposure pattern of lead, cadmium, mercury, and arsenic and their association with cardiovascular, renal and respiratory outcomes. Environment International. 2021;149:106410. Epub 2021/02/07. pmid:33548850.
- 39. Austin PC, Ibrahim M, Putter H. Accounting for Competing Risks in Clinical Research. Jama. 2024;331(24):2125–6. Epub 2024/05/29. pmid:38809526.
- 40. An Q, Wang Q, Liu R, Zhang J, Li S, Shen W, et al. Analysis of relationship between mixed heavy metal exposure and early renal damage based on a weighted quantile sum regression and Bayesian kernel machine regression model. Journal Of Trace Elements In Medicine And Biology. 2024;84:127438. Epub 2024/03/24. pmid:38520795.
- 41. Bather JR, Robinson TJ, Goodman MS. Bayesian Kernel Machine Regression for Social Epidemiologic Research. Epidemiology. 2024. Epub 2024/08/01. pmid:39087683.
- 42. Blount BC, Pirkle JL, Osterloh JD, Valentin-Blasini L, Caldwell KL. Urinary perchlorate and thyroid hormone levels in adolescent and adult men and women living in the United States. Environmental Health Perspective. 2006;114(12):1865–71. Epub 2006/12/23. pmid:17185277.