Abstract

Background

To determine the relationship between mixed exposure to three types of endocrine-disrupting chemicals (EDCs), namely phenyl hydroxides, polycyclic aromatic hydrocarbons (PAHs), and phthalates (PAEs), and risk of arthritis.

Methods

Participants were selected from National Health and Nutrition Examination Survey (NHANES). The relationships between the urinary concentrations of phenyl hydroxides, PAHs, and PAEs and the risk of arthritis were analyzed by generalized linear regression model. The mixed exposure to these EDCs and the risk of arthritis was analyzed by weighted quantile sums (WQSs) and Bayesian kernel machine regression (BKMR) model.

Results

Our analysis showed that participants with urinary benzophenone-3 and methylparaben concentrations in the highest quartile (Q4) had an increased risk of arthritis compared with those in Q1. For each one-unit increase in the natural logarithm-converted urinary concentrations of 1-hydroxynapthalene and 2-hydroxynapthalene, the risk of arthritis increased by 5% and 8%, respectively. Chemical mixing index coefficients were significantly associated with risk of arthritis in both WQS positive- and negative-constraint models. In the BKMR model, there was a significant positive correlation between mixed exposure and the risk of arthritis.

Conclusion

Mixed exposure to phenyl hydroxides, PAHs, and PAEs increased the risk of arthritis, with exposure to PAHs being the key factor.

Background

Arthritis is a type of joint inflammation characterized by joint swelling, pain, and cartilage destruction [1], and its subtypes include osteoarthritis, rheumatoid arthritis, and ankylosing spondylitis. The prevalence of arthritis increases with age, and thus due to population aging, arthritis is an increasing global health problem. For example, the age-standardized point prevalence of osteoarthritis worldwide was 3.75% in 2017, and the age-standardized prevalence of osteoarthritis in the United States increased by 23.20% from 1990 to 2017 [2, 3]. The pathogenesis of arthritis has not been fully clarified, but many influencing factors have been identified, such as autoimmune reactions, infection, susceptibility genes, age and other factors such as environmental factors, which have increasingly attracted attention [4].

In recent years, studies increasingly have focused on the relationship between environmental factors and arthritis. There is evidence that occupational exposure to asbestos, textile dust, and other substances is significantly associated with an increased risk of rheumatoid arthritis [5, 6], and that exposure to heavy metals such as cadmium is independently associated with an increased prevalence of rheumatoid arthritis [7, 8].

EDCs encompass a diverse range of naturally occurring and synthetic compounds, including phenyl hydroxides, phthalates (PAEs), and polycyclic aromatic hydrocarbons (PAHs). Fan et al. discovered a significant correlation between exposure to low levels of EDCs and mortality [9]. Furthermore, exposure to PAHs, the initial known chemical carcinogens, has been identified as a risk factor for conditions such as heart disease and high blood pressure [10]. EDCs exposure has been linked to increased inflammation and oxidative damage in arthritis [11], with population studies indicating that high levels of PAH metabolites in urine elevate the risk of rheumatoid arthritis [12, 13]. The association between PAHs and the risk of osteoarthritis, potentially attributed to the endocrine-disrupting properties of PAHs, has also been observed [14]. PAHs have been shown to exacerbate arthritis by affecting Th17 generation both in vivo and in vitro [15]. A review suggests that PAEs may contribute to arthritis development by enhancing the production of inflammatory cytokines [16]. Research has revealed that perinatal exposure to benzyl butyl phthalate can increase the prevalence and severity of rheumatoid arthritis in offspring by disrupting genes or signaling pathways associated with the condition [17]. Additionally, a study based on the National Health and Nutrition Examination Survey has indicated a connection between phthalate exposure and a higher prevalence of rheumatoid arthritis in adults [18]. Furthermore, evidence suggests that both single and combined exposures to phthalates may be linked to an increased incidence of rheumatoid arthritis [19]. Despite this, there remains a scarcity of population-based studies investigating the potential relationship between simultaneous exposure to phenyl hydroxides, PAHs, and PAEs and the risk of arthritis, highlighting the need for further research in this area.

The National Health and Nutrition Examination Survey (NHANES) 2005–2014 cohort was selected due to the comprehensive nature of the data available for this time period. Utilizing weighted generalized linear regression, the impact of individual exposure to phenyl hydroxides, PAHs, and PAEs on arthritis risk was assessed. Additionally, WQS regression and BKMR were employed as supplementary methods to investigate the combined effect of exposure to these EDCs on arthritis risk. The study also explored the association between combined exposure to phenyl hydroxides, PAHs, and PAEs and arthritis risk, contributing new epidemiological evidence to this area.

Materials and methods

Sample

The NHANES is a national cross-sectional study that has been conducted regularly since the 1960s to assess the health and nutritional status of children and adults in the United States. All procedures and content in the NHANES have been approved by the National Center for Health Statistics Ethics Review Board, and informed consent has been received from all participants. The present study used publicly available data of participants in the NHANES recruited from 2005 to 2014, namely 10,106 participants whose urinary concentrations of phenyl hydroxides, PAHs, and PAEs had been measured. The NHANES 2005–2014 cohort was chosen because of the relative comprehensiveness of the data for this cycle. After the exclusion of 5016 participants for missing covariates, 114 participants for pregnancy, and eight participants for missing arthritis outcomes, 4,878 participants remained for analysis.

Participants with arthritis

Participants with arthritis were diagnosed by a healthcare professional, and data was gathered through questionnaires. Individuals aged 20 and older were queried with the question: ‘Have you ever received a diagnosis of arthritis from a doctor or other medical professional?’ Those who answered ‘yes’ were categorized as having arthritis, while those who answered ‘no’ were categorized as not having arthritis.

EDC analysis

Humans are exposed to phenyl hydroxides, PAHs, PAEs, and other chemicals in a variety of ways, and this exposure can be detected in blood or urine. The NHANES collects biological samples from participants for laboratory testing and thus obtains detailed information on their recent chemical exposures and health and nutritional status [20, 21]. The present study used data from the NHANES on participants’ urinary concentrations of the following 20 EDCs, which were phenyl hydroxides, PAHs, or PAEs: benzophenone-3 (BP3), bisphenyl hydroxide A (BPA), methylparaben (MPB), propyl paraben (PPB), triclosan, mono (carboxynonyl) phthalate (CNP), mono (carboxyoctyl) phthalate (COP), mono-(2-ethyl-5-carboxypentyl) phthalate (ECP), mono-n-butyl phthalate, mono-(3-carboxypropyl) phthalate (MC1), mono-ethyl phthalate (MEP), mono-2-ethylhexyl phthalate, mono-isobutyl phthalate (MIB), mono-2-octylhexyl phthalate (MOH), mono-benzyl phthalate, 1-hydroxynaphthalene (P01), 2-hydroxynaphthalene (P02), 3-hydroxyfluorene (P03), 2-hydroxyfluorene (P04), and 1-hydroxyphenanthrene (P06). Those of the aforementioned EDCs that had a detection rate of greater than 50% were selected for subsequent analysis. Urine samples were stored at -20Ìý°C, and urinary creatinine was used as a correction factor (i.e., to account for the dilution effect of urine) when calculating the urinary concentrations of the 20 EDCs.

Covariates

The literature [8, 22] was reviewed to select covariates that may be associated with exposure to phenyl hydroxides, PAHs, PAEs and the risk of arthritis. The continuous variables selected as covariates were age, body mass index (BMI), serum concentration of cotinine, and energy intake. The categorical variables selected as covariates were sex (male or female), race/ethnicity (Mexican American, other Hispanic, non-Hispanic white, non-Hispanic black, other race), literacy [below 9th grade, 9th–11th grade (including 12th grade without a diploma), high school or equivalent, college or Associate of Arts degree, master’s degree or above], household poverty-to-income ratio (≤ 1.30, 1.31–3.50, > 3.50), alcohol consumption (more than 12 drinks in the past year, grouped with drinking or not), physical activity (never or rarely, moderate, vigorous), high blood pressure (yes or no), and diabetes (yes or no).

Statistical analysis

Descriptive statistics were used to summarize the participants’ demographic characteristics and urinary concentrations of EDCs. Categorical variables are reported as numbers of cases (percentages) and were compared using chi-square tests. Non-normal continuous variables are expressed as medians (quartiles) [M (Q1, Q3)] and compared using Wilcoxon rank sum tests. Normal continuous variables are expressed as means ± standard deviations, and were compared using independent sample t-tests. Urinary creatinine-corrected EDC concentrations were subjected to a natural logarithm (ln) conversion to afford data with a normal distribution. Spearman rank correlations were used to analyze the correlations between EDCs.

EDC concentrations were grouped into quartiles, and a weighted multivariate logistic regression model was used to examine the correlation between these quartiles and the risk of arthritis. A WQS regression model was used to analyze the relationship between mixed EDC exposure and the risk of arthritis, and to further determine which EDCs had the greatest influence on the risk of arthritis. A WQS regression model is a statistical model that assesses the contribution of each component of environmental exposure to an overall effect based on a weighted index and is widely used to examine the effects of mixed exposures [23]. It is based on quantiles of concentration of the chemicals of interest, with these being weighted according to their relative importance in a mixture, and thus helps to identify potentially toxic chemicals. A Bayesian kernel machine regression model (BKMR model) is used to explore the combined effects and possible interactions between a mixture of chemicals [24]. In addition, as BKMR modeling is a semi-parametric approach, it can discern nonlinear and non-additive relationships and measure the importance of variables. Therefore, it allows correlations between specific chemical exposures to be calculated based on the exposure–response relationship between these chemicals and the outcome of interest. Adjusting for confounders by including covariates in the weighted multivariate logistic regression model, WQS regression model, BKMR model.

Statistical analysis was performed using R (ver.3.4.3), and a bilateral P value of less than 0.05 was regarded as indicating a statistically significant difference.

Results

Baseline characteristics of sample

The general characteristics of the 4,878 participants are shown in TableÌý1 below. Among them, 1,316 (26.98%) had been diagnosed with arthritis and 3,562 (73.02%) had not. Statistical analyses revealed that there were significant differences between those with arthritis and those without arthritis in terms of age, sex, race, education level, marital status, family income, alcohol consumption, BMI, physical activity, energy intake, and pre-existing hypertension or diabetes status (P &±ô³Ù; 0.05).

Table 1 Population characteristics of arthritis status in NHANES adults from 2005 to 2014

EDC exposure determination and correlation analysis

As mentioned, the participants’ exposures to 20 EDCs, which were phenyl hydroxides, PAHs, or PAEs, were determined. In all participants, the detection rates of 18 EDCs were greater than 50%, whereas those of MIB and MOH were less than 50%. The detection rates and distributions of the 20 EDCs are listed in Supplementary Tables 1 and 2. Spearman correlation analysis of the 18 EDCS with a detection rate of greater than 50% revealed that the correlation coefficient between P03 and P04 was the largest (0.93), followed by that between PPB and MPB (0.83). In addition, PAHs (P01, P02, P03, P04, and P06) were correlated with each other (coefficients = 0.37–0.93), as shown in Fig.Ìý1.

Fig. 1

Spearman correlation coefficient between pairwise chemical measurement values

Spearman correlation coefficient between pairwise chemical measurement values. The x in this graph indicates that the correlation coefficient between the concentrations of the two substances is not significant, and the larger the value, the stronger the correlation

Weighted generalized linear regression

An adjusted covariate logistic regression model was constructed to analyze the effects of phenyl hydroxides, PAHs, and PAEs on the risk of arthritis. The results showed that for urinary concentrations of BP3, the odds ratio (OR) of the highest-quartile (Q4) concentrations was 1.20 times that of the lowest-quartile (Q1) concentrations [OR = 1.20, 95% confidence interval (CI) = 1.04–1.40]. Similarly, for urinary concentrations of MPB, Q4 concentrations were associated with a higher risk of arthritis than Q1 concentrations (OR = 1.18, 95% CI = 1.01–1.38). However, there was no statistically significant difference between the Q4 and Q1 concentrations of other EDCs and the risk of arthritis. Continuity analysis showed that ln-transformed concentrations of P01 and P02 were significantly associated with the risk of arthritis (OR = 1.05, 95% CI = 1.01–1.09 and OR = 1.08, 95% CI = 1.01–1.16, respectively). Specifically, each unit increase in the ln concentrations of P01 and P02 resulted in the risk of arthritis increasing by 5% and 8%, respectively (P < 0.05), as shown in TableÌý2.

Table 2 Association of individual phenyl hydroxide, polycyclic aromatic hydrocarbons (PAHs), phthalates (PAFs)and arthritis

Further adjustment of the model by the addition of other EDCs showed that there was no significant association between exposure to any of these other EDCs and the risk of arthritis. In addition, the variance inflation factor of nine EDCs (BPA, CNP, COP, MC1, MPB, P03, P04, P06, and PPB) was greater than 10, which suggests that there was multicollinearity between these EDCs (see TableÌý3 for details).

Table 3 Association of phenyl hydroxide, PAFs, PAHs and arthritis

WQS regression model

In the WQS positive constraint model, the EDC mixture index coefficient was significantly associated with the risk of arthritis (OR = 1.46, 95% CI = 1.16–1.84). Moreover, ECP, PPB, and P01 had high weights (0.097, 0.096, and 0.093, respectively, all > 1/18), which indicates their relative importance (Fig.Ìý2).

Similarly, in the WQS negative constraint model, the EDC mixture index coefficient was significantly correlated with the risk of arthritis (OR = 1.40, 95% CI = 1.11–1.77). Moreover, MEP, BPA, and MC1 had high weights (0.065, 0.065, and 0.063, respectively; Fig.Ìý3), which indicates their relative importance.

Fig. 2

Weight of regression index of WQS model on arthritis (positive constraint)

Fig. 3

Weight of regression index of WQS model on arthritis (negative constraint)

BKMR model

The BKMR model was first fitted to assess the combined effects of EDC exposure on the risk of arthritis, with creatinine-corrected and ln-transformed concentrations of each EDC as continuous variables. The results revealed a significant positive association between mixed exposure to EDCs and the risk of arthritis, with exposure to high concentrations of EDC mixtures associated with an increased risk of arthritis. Specifically, compared with the 50th-percentile concentrations of the 20 EDCs, the 60th- and 80th-percentile concentrations of the 20 EDCs were associated with 1.02-times (95% CI = 1.01–1.04) and 1.12-times (95% CI = 1.06 to 1.19) higher risks of arthritis, respectively (SFigure 1).

TableÌý4 summarizes the posterior inclusion probabilities (PIPs) derived from the BKMR model for the three groups of EDCs (group PIPs) and for each EDC (conditional PIPs). In the model, the PAH group had the highest group PIP (0.91). Furthermore, the conditional PIP for the PAH P02 was extremely high (0.97), whereas those for the other PAHs were low. This suggests that the association between these EDCs and the risk of arthritis was primarily driven by P02. Moreover, when all the other EDCs were at medium-level concentrations, only P02 was positively associated with the risk of arthritis, as shown in SFigure 2.

Table 4 Posterior inclusion probabilities (PIPs) of each phenols, polycyclic aromatic hydrocarbons, phthalates for arthritis, using the bayesian kernel machine regression (BKMR) model, NHANES, 2005–2014

Next, with 18 EDCs at constant median concentrations, the exposure–response function of one EDC with the risk of arthritis was explored when the exposure of another EDC was fixed at the 25th, 50th, and 75th percentiles, respectively. As shown in SFigure 3 this analysis revealed that when the other chemicals were fixed at median levels, the slope of the exposure–response function for one chemical was similar across different quantiles of exposure to any other EDC, suggesting that no two EDCs had a combined effect on the risk of arthritis.

Discussion

Environmental EDCs are man-made or natural chemicals that exhibit hormone-like activity that can affect the function of the endocrine system and disrupt the action of hormones, thereby affecting human health [25]. In this study, compared with Q1 urinary concentrations of BP3 and MPB, Q4 concentrations of these EDCs were determined to increase the risk of arthritis. Phenyl hydroxide is an important raw material in industrial processes and thus has become a widespread pollutant. It enters the human body mainly through the skin or mucosa, and is metabolized and detoxified by the liver to produce metabolites that are excreted in urine [26]. Studies have shown that environmental phenyl hydroxides have endocrine-disrupting effects, and long-term exposure to phenyl hydroxide can inhibit hormone biosynthesis, metabolism, and other functions in humans [27, 28]. Moreover, Loffredo et al. reported that environmental phenyl hydroxides such as BPA, a known endocrine disruptor, may be associated with increased risks of arthritis and systemic inflammation [29], but the physiological mechanisms of this process remain to be elucidated.

In the present study, the BKMR model suggests that the risk of arthritis due to PAH exposure was largely attributable to P02 exposure. PAHs are atmospheric pollutants found in oil, coal, and tar deposits, and the human body is primarily exposed to these PAHs through contact with contaminated air or soil. PAH exposure promotes the production of inflammatory cytokines and inhibits the anti-inflammatory pathway, and thus contributes to the development of rheumatoid arthritis. In particular, PAHs upregulate the release of several inflammatory cytokines [e.g., interleukin (IL)-1 beta, IL-6, and IL-8] in synovial cells, leading to the pathophysiological deterioration typical of arthritis [30, 31]. Moreover, Talbot et al. demonstrated that PAHs aggravated arthritis in an antigen-induced mouse model by modulating the differentiation of Th17 cells [15], which play an important role in the development of arthritis. PAH exposure is also a risk factor for diseases such as hypertension [10], heart disease, and depression. Furthermore, a class of PAHs exemplified by benzo[a]pyrene, 2,3,7,8-tetrachlorodibenzo-p-dioxin, and 3-methylcholanthrene are agonists of the polycyclic aromatic receptor (AHR) [32], and AHR has a regulatory effect on immune function [15], bone metabolic balance, intestinal flora, and other biochemical processes that are dysregulated in those with arthritis. In addition, as the AHR signaling pathway plays a key role in regulating immune function, the unregulated activation of AHR may lead to the development of autoimmune diseases [32, 33]. The aryl hydrocarbon receptor (AHR) plays a crucial role in the autoimmune response, particularly in the context of PAH exposure. AHR signaling has the potential to influence the differentiation of regulatory T cells by modulating the expression of regulatory T cell markers within these cells or through dendritic cells (DCs) [33]. Additionally, AHR is involved in the development of T helper 17 (Th17) cells, which play a role in promoting the immune response [33]. In the case of rheumatoid arthritis, an autoimmune disease characterized by chronic synovial joint inflammation and infiltration of activated immunoinflammatory cells leading to progressive cartilage degradation and bone erosion, a subset of CD4 T cells producing interleukin 17 (IL-17), known as Th17 cells, has been identified as a significant contributor to disease development and progression. PAH exposure can activate AHR, potentially regulating the proliferation of Th17 cells through the transcriptional control of miR-132/212 clusters [15]. The IL-17 cytokine family, particularly IL-17ÌýA, has been implicated in the pathophysiology of rheumatoid arthritis in humans [34]. Furthermore, AHR may directly regulate IL-17 expression, leading to the induction of proinflammatory cytokines, such as IL-6 and IL-1β, by synovial macrophages. IL-1β, in particular, can stimulate osteoclastogenesis directly in osteoclast precursor cells, contributing to bone erosion [11, 15]. As such, AHR is a link between exposure to environmental EDCs and the immune response [35, 36]. Overall, PAHs are common environmental EDCs that exhibit estrogenic or anti-androgenic activity and disrupt homeostasis and endocrine hormones [37], especially estrogen and thyroid hormones, which promote the growth and development of human bones and play an important role in joint development [14, 38].

PAEs are widely used as preservatives in cosmetics, drugs, and food and are another type of environmental EDC. They exhibit anti-androgen effects that damage the human endocrine, reproductive, and thyroid systems and the liver, kidney, and other organs [39]. Yang et al. showed that high urinary concentrations of PAEs are related to the human inflammatory response, which may indicate a link between PAE exposure and the risk of arthritis [40], but the physiological mechanism of this putative link remains to be clarified. Exposure to PAEs is also associated with cardiovascular, cerebrovascular, and respiratory diseases, can lead to metabolic diseases such as diabetes, and has carcinogenic effects [10, 41].

Few studies have examined the relationship between mixed exposure to EDCs and the risk of arthritis in the general population. All three regression models in the present study showed that simultaneous exposure to high concentrations of phenyl hydroxides, PAHs, and PAEs was positively correlated with the risk of arthritis. Others have shown that exposure to PAHs may increase oxidative stress in the human body, and thereby cause irreversible oxidative damage to DNA [42, 43]. Oxidative stress has also been confirmed to play a role in the pathogenesis of arthritis [12, 13]. The findings of this study partially support the notion that oxidative stress resulting from PAH exposure can lead to the development of arthritis. Additionally, this study delves into the potential mechanisms through which oxidative stress may contribute to arthritis. Further research is warranted to investigate the impact of combined EDC exposure on human health and elucidate its underlying mechanisms through experimental studies. Furthermore, population-based studies have shown that high urinary concentrations of PAH metabolites such as 2-hydroxynaphthalene, 2-hydroxyfluorene (2-FLU), and 3-FLU increase the risk of developing arthritis [12, 13]. The findings of this study help to enrich existing research while providing a theoretical basis for policy formulation.

There are several limitations of this study. First, it used a large sample of NHANES data and was cross-sectional, which meant that it was not possible to determine the temporal and causal relationships between mixed exposure to the EDCs and the risk of arthritis. Second, the pathogenesis of arthritis is complex, and arthritis subgroups not available in the database in greater detail. Therefore, no further subgroup analyses of arthritis were performed in this study. Moreover, some potential confounding factors, such as heredity, lifestyle, and occupation, were not excluded. Therefore, further control of more potential confounders is needed in future studies to ensure the usability of results. These might be causes of bias. Third, phenyl hydroxides, PAHs, and PAEs are converted into a range of metabolites in the human body, but the potential impact of such metabolites on the risk of arthritis was not examined. Therefore, in a future study, the relationship between EDC metabolites and the risk of arthritis should be explored, and the physiological mechanism linking mixed EDC exposure to the pathogenesis of arthritis should be comprehensively delineated.

Conclusion

This study analyzed data from a large sample to determine the effects of single exposure and mixed exposure, respectively, to three types of EDCs, namely phenyl hydroxides, PAHs, and PAEs, and the risk of arthritis. The results showed that high levels of mixed exposure to phenyl hydroxides, PAHs, and PAEs increased the risk of arthritis. Furthermore, exposure to PAHs, particularly P02, was a key factor affecting the increased risk of arthritis.