Association between exposure to polycyclic aromatic hydrocarbons and endometriosis: data from the NHANES 2001–2006

Aim To evaluate the association between urinary polycyclic aromatic hydrocarbon (PAH) metabolites and the risk of endometriosis. Methods This cross-sectional study obtained data on women aged 20–54 years from the National Health and Nutrition Examination Survey (NHANES) 2001–2006. The weighted multivariate logistic regression model was established to assess the association between the eight urinary PAH metabolites and the risk of endometriosis. In this multivariate analysis, the eight urinary PAH metabolites were adjusted with urinary creatinine, and were divided into three groups according to tertiles: Tertile 1, Tertile 2 and Tertile 3. To evaluate the overall association of mixed PAH metabolites with endometriosis, the Bayesian kernel machine regression (BKMR) model was applied. Results Totally 1,291 women were included, of which 90 (6.97%) had endometriosis and 1,201 (93.03%) did not have endometriosis. After adjusting for age, race, smoking, age at menarche, hysterectomy, ovary removed, female hormone use, and menopause, compared with the Tertile 1 group, the Tertile 2 and Tertile 3 groups of all PAH metabolites demonstrated no significant risk of endometriosis. A positive tendency was found between mixed PAH metabolites and endometriosis when all the metabolites were at their 60th percentile levels or above compared with their median levels. When all the other metabolites were fixed at their median levels, 1-hydroxynaphthalene was positively correlated with endometriosis. Potential interactions existed between 1-hydroxynaphthalene and 2-hydroxynaphthalene and between 2-hydroxyfluorene and 3-hydroxyfluorene. Conclusion No significant association was found between individual PAH metabolites and endometriosis. A positive association existed between mixed PAH metabolites and the risk of endometriosis.


Introduction
Endometriosis, defined as the presence of endometrial tissue outside the uterus, is a chronic gynecological disease with an increased incidence, affecting 5-10% of women of childbearing age and causing pelvic pain and infertility (1)(2)(3).Painful periods, deep dyspareunia, dyschezia, dysuria, and abdominal pain are the most common symptoms (4,5).Endometriosis is associated with a 50% increase in the risk of epithelial ovarian cancer (6), and is also correlated with gastrointestinal and immunological diseases and thyroid cancer (7).Endometriosis is an estrogen-dependent disease, which can be affected by genetics, environmental factors, and the immune system (8,9).Environmental endocrine-disrupting chemicals (EDCs) are a kind of exogenous chemicals that interfere with the endocrine system, which can affect the endocrine system and reproductive health (10,11).Exposure to EDCs has been suggested to be associated with the increasing incidence of endometriosis (12,13).Evidence shows that EDCs such as phthalates, perfluoroalkyl substances and heavy metals are related to endometriosis (14)(15)(16).Polycyclic aromatic hydrocarbons (PAHs), as a kind of EDC, comprise two or more condense aromatic rings and have a toxic nature, threatening to public health (17).PAHs mainly exist in polluted air, soil, water and food (18,19).PAHs can enter the human body through different occupational, environmental, medical, and nutritional sources, such as exposure to aluminum and graphite electrode production or road construction, dietary intake of meat, dairy products or smoked foods, etc. (20,21).Humans are exposed to PAHs through inhalation, ingestion and skin contact (22).PAH metabolites in human urine can be utilized as internal dosage indicators to measure PAH exposure (23).At present, the associations between PAH exposure and sex hormone levels (e.g., reproductive hormones, estradiol, and serum testosterone) have been reported (24)(25)(26)(27), whereas no study has assessed the association between PAH exposure and the risk of endometriosis.Of note, individuals are exposed to various chemicals at the same time, resulting in a mixture exposure pattern.It is necessary to explore the individual and combined association between urinary PAH metabolites and the risk of endometriosis, so as to promote understanding of the relationship between PAH exposure and endometriosis, and provide reference for the prevention and management of endometriosis.
The objective of this study was to evaluate the association of urinary individual and mixed PAH metabolites with the risk of endometriosis, using the data from the National Health and Nutrition Examination Survey (NHANES) 2001-2006.

Endometriosis assessment
Whether participants had endometriosis was assessed according to the question "Has a doctor or other health professional ever told {you/SP} that {you/she} had endometriosis?"If the response was "yes, " the participant was regarded to have endometriosis.

Statistical analysis
Measurement data with normal distribution were described by mean ± standard deviation (Mean ± SD), and the independentsamples t test was used for comparisons between two groups.Measurement data with skewed distribution were shown as median and quartile [M (Q 1 , Q 3 )], and between-group comparisons were subject to the Wilcoxon rank sum test.Enumeration data were expressed as the number of cases and the composition ratio [n (%)], and the Chi-square test or Fisher's exact test was applied for comparisons between groups.The missing data were interpolated by the random forest imputation method.The sensitivity analysis was carried out by comparing the data before and after the imputation to exclude the influence of the random forest imputation method on the results (Supplementary Table S2).
The participants were divided into endometriosis and non-endometriosis groups.After the difference analysis, the variables with significant differences between the two groups were taken as confounding factors.The Pearson correlation was employed to calculate the correlation coefficients among these PAH metabolites.Then the weighted multivariate logistic regression model was established to assess the association between the eight urinary PAH metabolites and the risk of endometriosis, and the confounding factors, age, race, smoking, age at menarche, hysterectomy, ovary removed, female hormone use, and menopause, were adjusted for in the model.Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated.In this multivariate analysis, the eight urinary PAH metabolites were adjusted with urinary creatinine, and were divided into three groups according to tertiles: Tertile 1, Tertile 2, and Tertile 3, and then the study population was grouped based on the three groups of the urinary PAH metabolites.To evaluate the overall association of mixed PAH metabolites with endometriosis, the Bayesian kernel machine regression (BKMR) model was applied with a hierarchical variable selection method with 10,000 iterations.BKMR model assumptions include linear relationship assumption, Gaussian noise assumption, additive assumption, and prior distribution assumption.One of the BKMR model's limitations is the kernel algorithm.Fixing other chemicals at certain levels in order to extrapolate the exposure-response function does not allow for estimation of the effects of co-exposure patterns with both high and low levels of chemicals.The effects of co-exposure patterns with both high and low levels of chemicals cannot be estimated when other chemicals are fixed at certain amounts to extrapolate the exposureresponse function.Besides, the total risk is evaluated using the BKMR model with all of the chemicals at different percentiles relative to their median exposure levels, whereas the actual exposure pattern is a mixture of chemicals with varying levels of exposure.2-hydroxyfluorene and 3-hydroxyfluorene had a strong correlation (r = 0.87), and were grouped into the first group (group 1); 1-hydroxyphenanthrene, 2-hydroxyphenanthrene, 3-hydroxyphenanthrene had similar exposure sources, and 1-hydroxyphenanthrene and 2-hydroxyphenanthrene had a strong correlation (r = 0.81), so these three metabolites were grouped into the second group (group 2); the remaining 1-hydroxynaphthalene, 2-hydroxynaphthalene and 1-hydroxypyrene had weak correlations with each other, and were grouped into the third group (group 3).The joint effect was calculated by comparing all the PAH metabolites at their 60th percentile levels or above with all of them at their 50th percentile.The group posterior inclusion probability (groupPIP) and conditional PIP (condPIP) indicated the probability of the group and the PAH metabolite in each group included in the model and represented their contribution to the overall association.By fixing all the other PAH metabolites at their median levels, the exposureresponse relationship was explored in the BKMR model.BKMR bivariable analysis was conducted to explore the relationship between individual PAH metabolites and endometriosis, while fixing another PAH metabolite at the 10th, 50th, and 90th quantiles (and keeping the remaining PAH metabolites at their median levels), so as to assess the interactions between any two of the PAH metabolites.Prior studies showed that when variables used to calculate sampling weights were included in the analysis, the weighted estimation could cause over-adjusted bias (31,32).In addition, it was unclear whether sampling weights were suitable for BKMR, a complex statistical model.Thus, sampling weights were not applied in the BKMR model.Data cleaning (including missing value statistics) and missing value imputation were completed using Python 3.7.4(Python Association between individual PAH metabolites and endometriosis.Models were adjusted for age, race, smoking, age at menarche, hysterectomy, ovary removed, female hormone use, and menopause.All the urinary PAH metabolites (ng/L) were adjusted/divided by urinary creatinine (mg/dL).PAH, polycyclic aromatic hydrocarbon; Naphthalene1, 1-hydroxynaphthalene; Naphthalene2, 2-hydroxynaphthalene; Fluorene2, 2-hydroxyfluorene; Fluorene3, 3-hydroxyfluorene; Phenanthrene1, 1-hydroxyphenanthrene; Phenanthrene2, 2-hydroxyphenanthrene; Phenanthrene3, 3-hydroxyphenanthrene; Pyrene1, 1-hydroxypyrene; OR, odds ratios; CI, confidence interval.

Participant characteristics
After excluding participants without the assessment of endometriosis (n = 707) and not examined for urinary PAH concentrations (n = 2,955), 1,291 women were included in this study, of which 90 (6.97%) had endometriosis and 1,201 (93.03%) did not have endometriosis.Figure 1 presents the flow chart of participant selection.The average age of the included women was 35.66 years.There were significant differences between the endometriosis and non-endometriosis groups in age, race, smoking, age at menarche, hysterectomy, ovary removed, female hormone use, and menopause (all p < 0.05).The characteristics of the included women are shown in Table 1.The endometriosis group had significantly more 1-hydroxynaphthalene, 2-hydroxyfluorene, 3-hydroxyfluorene, 1-hydroxyphenanthrene, 2-hydroxyphenanthrene, and 3-hydroxyphenanthrene than the non-endometriosis group (all p < 0.05).Overall association between mixed PAH metabolites and endometriosis when all the metabolites were at their 60th percentile levels or above compared with their median levels.Models were adjusted for age, race, smoking, age at menarche, hysterectomy, ovary removed, female hormone use, and menopause.All the urinary PAH metabolites (ng/L) were adjusted/divided by urinary creatinine (mg/dL), and underwent log conversion.PAH, polycyclic aromatic hydrocarbon.

Association between individual PAH metabolites and endometriosis
After adjusting for age, race, smoking, age at menarche, hysterectomy, ovary removed, female hormone use, and menopause, compared with the Tertile 1 group, the Tertile 2 and Tertile 3 groups of all PAH metabolites demonstrated no significant risk of endometriosis (Figure 3).

Overall association between mixed PAH metabolites and endometriosis
The eight urinary PAH metabolites were grouped into three groups according to their similar exposure sources and correlations.After adjusting for the parameters, a positive tendency was found between mixed PAH metabolites and endometriosis when all the metabolites were at their 60th percentile levels or above compared with their median levels (Figure 4).Table 3 exhibits the contribution of all the PAH metabolites to the overall association.The group 1 had the highest groupPIP (groupPIP = 0.25), and 2-hydroxyfluorene made the most contribution in the group 1 (CondPIP = 0.64).The trends of exposure-response functions of the eight metabolites are illustrated in Trends of exposure-response functions of PAH metabolites when all the other metabolites were fixed at their median levels.Models were adjusted for age, race, smoking, age at menarche, hysterectomy, ovary removed, female hormone use, and menopause.All the urinary PAH metabolites (ng/L) were adjusted/divided by urinary creatinine (mg/dL), and underwent log conversion.PAH, polycyclic aromatic hydrocarbon; Naphthalene1, 1-hydroxynaphthalene; Naphthalene2, 2-hydroxynaphthalene; Fluorene2, 2-hydroxyfluorene; Fluorene3, 3-hydroxyfluorene; Phenanthrene1, 1-hydroxyphenanthrene; Phenanthrene2, 2-hydroxyphenanthrene; Phenanthrene3, 3-hydroxyphenanthrene; Pyrene1, 1-hydroxypyrene.  .When all the other metabolites were fixed at their median levels, 1-hydroxynaphthalene was positively correlated with endometriosis.According to BKMR bivariable analysis, potential interactions existed between 1-hydroxynaphthalene and 2-hydroxynaphthalene and between 2-hydroxyfluorene and 3-hydroxyfluorene (Figure 6).

Discussion
To our knowledge, this study was the first to evaluate the association between urinary PAH metabolites and endometriosis.The BKMR model was applied herein to estimate the overall association of mixed PAH metabolites with endometriosis.No significant association was found between individual PAH metabolites and the risk of endometriosis.There was a positive tendency between mixed PAH metabolites and endometriosis when all the metabolites were at their 60th percentile levels or above compared with their median levels, and when all the other metabolites were fixed at their median levels, 1-hydroxynaphthalene was positively correlated with endometriosis.
Increasing evidence showed an association between PAH metabolites and reproductive diseases, such as infertility and urogenital tract abnormalities (33,34).Several studies has connected endometriosis with organic pollutants (12,35,36).For the relationship between PAH metabolites and endometriosis, individual PAH metabolites were reported to be not associated with the risk of endometriosis, according to multivariate logistic regression analysis in this paper.Since chemicals are exposed at the same time, and usually exposure sources and metabolism pathways can lead to high collinearity, the association between a chemical and health outcomes may be masked or exaggerated by other relevant chemicals (37, 38).Interaction between urinary PAH metabolites, while fixing one of two PAH metabolites at the 10th, 50th, and 90th quantiles (and keeping the remaining PAH metabolites at their median levels).Models were adjusted for age, race, smoking, age at menarche, hysterectomy, ovary removed, female hormone use, and menopause.All the urinary PAH metabolites (ng/L) were adjusted/divided by urinary creatinine (mg/dL), and underwent log conversion.PAH, polycyclic aromatic hydrocarbon; Naphthalene1, 1-hydroxynaphthalene; Naphthalene2, 2-hydroxynaphthalene; Fluorene2, 2-hydroxyfluorene; Fluorene3, 3-hydroxyfluorene; Phenanthrene1, 1-hydroxyphenanthrene; Phenanthrene2, 2-hydroxyphenanthrene; Phenanthrene3, 3-hydroxyphenanthrene; Pyrene1, 1-hydroxypyrene.Thus, chemical mixtures should be considered in assessing the association between PAH metabolites and endometriosis.
The BKMR model was applied to evaluate the overall association between mixed PAH metabolites and endometriosis.The current study found a positive trend between mixed PAH metabolites and endometriosis when all the metabolites were at their 60th percentile levels or above versus their median levels.Exposure to PAHs could affect female reproduction, and luteinizing hormone, follicle stimulating hormone, gonadotrophin releasing hormone, and aromatase seem to be influenced by PAHs (39).As endometriosis is estrogen-dependent (40), PAH exposure might be associated with an increased risk of endometriosis through adversely impacting the above hormonal regulators.We also demonstrated that when all the other metabolites were fixed at their median levels, 1-hydroxynaphthalene was positively correlated with endometriosis, which needs future evidence for validation due to the lack of studies.Notably, in the BKMR model, the overall association was assessed with all the PAH metabolites at their 60th percentile levels or above versus their median levels, but the exposure pattern in a real-world setting is a combination of PAH metabolites with different exposure levels.Thus, the results from BKMR analysis should be interpreted cautiously.In addition, there may be interactions between 1-hydroxynaphthalene and 2-hydroxynaphthalene and between 2-hydroxyfluorene and 3-hydroxyfluorene on the risk of endometriosis.As a possible implication, exposure to PAHs should be reduced to prevent or lower the risk of endometriosis, for example, via raising public awareness of PAHs and environmental measures.
The NHANES uses a complex, multistage, probability sampling design, and assesses a nationally representative sample of around 5,000 persons annually.This study included 1,291 women for analysis, representing 20,579,272 women after weight calculation.Some limitations should be mentioned in this study.First, due to the crosssectional design, the causation between PAH metabolites and endometriosis could not be determined, and the relationship between PAH exposure time and endometriosis could not be estimated.Second, many samples without urinary PAH examinations were excluded (n = 2,955), which may cause selection bias.However, PAHs were detected in 1/3 of the respondents from the population aged 6 and older surveyed by the NHANES, and we considered the corresponding weight in the analysis, so the bias is relatively small.Third, this study focused on the American population, the findings of which may have limited generalization.More investigations are warranted to certify these findings.Additionally, external data validation was not performed since the limited sample size.Future large-sample datasets are required for external validation.Because information on occupations was not available in the NHANES, which occupations have exposure to PAHs is unknown.Future studies could investigate this point.

Conclusion
No significant association was found between individual PAH metabolites and the risk of endometriosis.A positive tendency existed between mixed PAH metabolites and endometriosis when all the metabolites were at their 60th percentile levels or above compared with their median levels, and when all the other metabolites were fixed at their median levels, 1-hydroxynaphthalene was positively correlated with endometriosis.
. Since information on endometriosis assessment was only available in the NHANES 1999-2006, and data on urinary PAH metabolites (1-hydroxynaphthalene, 2-hydroxynaphthalene, 2-hydroxyfluorene, 3-hydroxyfluorene, 1-hydroxyphenanthrene, 2-hydroxyphenanthrene, 3-hydroxyphenanthrene, 1-hydroxypyrene) were only available in the NHANES 2001-2006, this study chose the timeline 2001-2006.The NHANES is a cross-sectional survey that collects demographic, clinical, behavioral, dietary, social, and laboratory data on the health and nutritional status of the non-institutionalized population in the United States, employing a multilayered probability sampling design (28).The survey was approved by the Institutional Review Board of the National Center for Health Statistics (NCHS), and all participants provided informed consent.Since the data of the NHANES were

FIGURE 1 Flow
FIGURE 1 Flow chart of participant selection.NHANES, the National Health and Nutrition Examination Survey; PAH, polycyclic aromatic hydrocarbon.

TABLE 1
Characteristics of the included women.

TABLE 2
Distribution of urinary PAH metabolites.

TABLE 3
Contribution of urinary PAH metabolites to the overall association.