Total Effective Xenoestrogen Burden in Serum Samples and Risk of Endometrial Cancer in the Spanish Screenwide Case–Control Study

Abstract

Background:

Endometrial cancer is a hormone-dependent cancer, and estrogens play a relevant role in its etiology. However, little is known about the effects of environmental pollutants that act as xenoestrogens or that influence estrogenic activity through different pathways.

Objective:

We aimed to assess the relationship between the combined estrogenic activity of mixtures of xenoestrogens present in serum samples and the risk of endometrial cancer in the Screenwide case–control study.

Methods:

The total effective xenoestrogen burden (TEXB) attributable to organohalogenated compounds ($\text{TEXB-}\mathrm{\alpha }$) and to endogenous hormones and more polar xenoestrogens ($\text{TEXB-}\mathrm{\beta }$) was assessed in serum from 156 patients with endometrial cancer (cases) and 150 controls by combining chemical extraction and separation by high-performance liquid chromatography with the E-SCREEN bioassay for estrogenicity.

Results:

Median $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels for cases (0.30 and $1.25\text{\hspace{0.17em}Eeq pM}/\mathrm{mL}$, respectively) and controls (0.42 and $1.28\text{\hspace{0.17em}Eeq pM}/\mathrm{mL}$, respectively) did not significantly differ ($p=0.653$ and 0.933, respectively). An inverted-U risk trend across serum $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels was observed in multivariate adjusted models: Positive associations were observed for the second category of exposure in comparison to the lowest category of exposure [$\text{odds ratio\hspace{0.17em}}\left(\text{OR}\right)=2.11$ (95% CI: 1.13, 3.94) for $\text{TEXB-}\mathrm{\alpha }$, and $\text{OR}=3.32$ (95% CI: 1.62, 6.81) for $\text{TEXB-}\mathrm{\beta }$], whereas no significant associations were observed between the third category of exposure and the first [$\text{OR}=1.22$ (95% CI: 0.64, 2.31) for $\text{TEXB-}\mathrm{\alpha }$, and $\text{OR}=1.58$ (95% CI: 0.75, 3.33) for $\text{TEXB-}\mathrm{\beta }$]. In mutually adjusted models for $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels, the association of $\text{TEXB-}\mathrm{\alpha }$ with endometrial cancer risk was attenuated [$\text{OR}=1.45$ (95% CI: 0.61, 3.47) for the second category of exposure], as well as estimates for $\text{TEXB-}\mathrm{\beta }$ ($\text{OR}=2.68$; 95% CI: 1.03, 6.99). Most of the individual halogenated contaminants showed no associations with both TEXB and endometrial cancer.

Conclusions:

We evaluated serum total xenoestrogen burden in relation to endometrial cancer risk and found an inverted-U risk trend across increasing categories of exposure. The use of in vitro bioassays with human samples may lead to a paradigm shift in the way we understand the negative impact of chemical mixtures on human health effects. These results are relevant from a public health perspective and for decision-makers in charge of controlling the production and distribution of chemicals with xenoestrogenic activity. https://doi.org/10.1289/EHP13202

Introduction

Endometrial cancer ranks as the most common gynecological tumor in regions with a very high human development index.¹ The burden of this cancer is expected to increase worldwide owing to an increased aging population, given that it is typically a postmenopausal cancer, among others.² Endometrial cancer has been traditionally classified into two subtypes: type I (endometrioid) and type II (non-endometrioid). Type I tumors, the most common subtype, are typically low-grade and good-prognosis cancers.³ Type II tumors constitute 10%–15% of endometrial cancers, are high-grade carcinomas, and show a poor prognosis.³ Surrogates of the classification system developed by The Cancer Genome Atlas (TCGA) Consortium, which is based on molecular features, are now being integrated into clinical practice.^4,5

Endometrial cancer is a hormone-dependent cancer, and an estrogen imbalance plays a crucial role in its development.⁶ Nulliparity, age at last birth, and infertility have repeatedly been associated with endometrial cancer.^7–9 Obesity can also increase the risk of endometrial cancer via the conversion of androgens to estrogens via aromatase activity in adipose tissue and other mechanisms.¹⁰ A past history of breast cancer has also been associated with endometrial cancer, especially in estrogen receptor (ER)-positive breast cancer patients treated with tamoxifen.¹¹ Circulating estrogens and estrogen metabolites have been repeatedly associated with endometrial cancer,^12–14 as have genetic variants in sex hormone-related genes, such as CYP19.^15,16 Mendelian randomization studies have supported the hypothesis that endometrial cancer risk is causally driven by estrogens.^16–18 Both types of endometrial cancer share common risk factors, and to a lesser extent, the etiology of type II tumors has also been shown to have an estrogenic influence.^8,19

Although exposure to estrogens is an established risk factor for endometrial cancer, little is known regarding the effect of endocrine-disrupting chemicals (EDCs); that is, environmental pollutants that have the potential to act as xenoestrogens or influence estrogenic activity through different pathways.²⁰ Most of the literature originates from basic in vitro and in vivo data,²¹ and the scarce observational literature has shown mostly null results,^22–28 with some exceptions.^29–31 However, the available literature suffers from limited sample sizes, has heterogeneous designs and exposure assessments, and focuses on a limited number of pollutants. Exposure assessments were performed using dietary variables,^22–24,32 urinary markers,^26,31 serum biomarkers,^27,28 adipose tissue biomarkers,²⁵ or endometrial and myometrial tissue biomarkers.³⁰ Functional biomarkers measuring the joint estrogenic activity of EDC mixtures offer a validated approach for combined exposure assessment.^33–36 Thus, a case–control study of 186 patients with breast cancer and 196 controls reported a positive association between the joint effect of environmental xenoestrogens extracted from serum samples and breast cancer risk,³⁶ which share certain etiologic similarities with endometrial cancer. However, to our knowledge no study has evaluated the combined effects of xenoestrogens on endometrial cancer. In the present study, we aimed to assess the association between the combined estrogenic activity of mixtures of xenoestrogens present in serum samples and the risk of endometrial cancer in the Screenwide Spanish case–control study.

Materials and Methods

Study Population

Cases were recruited within the Screenwide study, a Spanish case–control study (2017–2021), in which 182 consecutive incident endometrial cancer cases were enrolled, as well as 266 hospital controls (including 190 women with benign gynecological pathology and 76 women with nongynecological diseases) frequency-matched to cases by age. Benign gynecological pathology among controls included conditions such as endometriosis, fibroids, benign cysts, prolapse, and polyps, as well as regular gynecologic controls in healthy women. Inclusion criteria were having an intact uterus and, in addition for cases, having an incident diagnosis of endometrial cancer. Endometrial cancer cases were classified into the classic two broad subtypes, the so-called type I, which included the endometrioid histological subtype, and type II, which included the rest of histological subtypes.³⁷ Exclusion criteria for all participants included pregnancy, puerperium (8 wk), communication difficulties, intellectual disability, and treatment with chemotherapy or radiotherapy during the previous 6 months. Response rates among the eligible participants were 89.7% for cases, 80.4% for controls with benign gynecological pathology, and 76.8% for hospital controls without benign gynecological pathology. Further details have been published elsewhere.³⁸

Participants were identified and enrolled through outpatient visits and visits for anesthesia prior to surgery. Peripheral blood serum was drawn from participants upon enrollment before any cancer treatment, using SST II Advance BD Vacutainers, and was aliquoted and stored at $-80°\mathrm{C}$. Data on all covariates included in model adjustments were self-collected through face-to-face interviews conducted by trained personnel. Reference dates were age at diagnosis for cases and age at enrollment for controls. In cases, the age at diagnosis closely matched the age at enrollment, and the questions referred to typical exposures before the cancer diagnosis. Anthropometric questions included the usual weight 1 y prior to the diagnosis, which was used to compute body mass index (BMI). Lifetime occupational history was assessed for all jobs held for at least 1 y. Prior to the interview, all participants were informed about the purpose of the study and signed an informed consent form. Among those cases who agreed to donate blood ($N=174$), we did not include samples from 6 cases with $<2\mathrm{mL}$ of serum available, and 1 sample from a case enrolled after the shipment to the site of analyses. We further randomly discarded 11 samples owing to budget constraints, yielding 156 samples from cases for analysis. We included 150 controls (75 controls with benign gynecological pathology, and 75 women attending hospital for nongynecological diseases), who were frequency-matched to cases by 5-y age intervals.

The study followed all the requirements established by the Ethics Committee for Clinical Research and was approved by the Ethics Committee for Clinical Research from the Bellvitge University Hospital (references: PR128/16 and PR348/19), as previously described in Peremiquel-Trillas et al.³⁸ Participation in the study was voluntary, and all eligible participants signed an informed consent form. The Screenwide study followed the national and international directives on ethics and data protection (Declaration of Helsinki and subsequent amendments; EU Reglament 2016/679) and the Spanish laws on data protection (Organic Law 3/2018; Law 14/2007 Biomedical Research). The study was registered in the National Register of Biobanks/Collections (C.0004389).

Exposure Mixture Assessment

The total effective xenoestrogen burden (TEXB), which assesses the combined estrogenic effect of chemical mixtures, was assessed in serum samples from cases and controls, collected when the participants signed the informed consent form—in cases before starting oncological treatment—using a previously standardized methodology described by Fernández et al.³⁵ and Pastor-Barriuso et al.³⁶ Briefly, $3\mathrm{mL}$ of methanol were added to each serum sample ($3\mathrm{mL}$), and the mixture was extracted with $7.5\mathrm{mL}$ of hexane:ethyl ether (1:1 vol/vol), using a Bond Elut polychlorinated biphenyl (PCB) cartridge (Varian), previously prepared with $1.5\mathrm{mL}$ of hexane. The dried eluates were reconstituted with $200\mathrm{\mu }\mathrm{L}$ of hexane and injected in duplicate ($100+100\mathrm{\mu }\mathrm{L}$) into a high-performance liquid chromatograph (HPLC). The semipreparative HPLC method was designed and validated to effectively separate persistent halogenated xenoestrogens from endogenous hormones. The method collects two HPLC fractions: the $\mathrm{\alpha }\text{-fraction}$, containing the more lipophilic compounds (PCBs, organochlorine pesticides and metabolites, and halogenated bisphenols, among others), and the $\mathrm{\beta }\text{-fraction}$, containing the endogenous hormones together with more polar xenoestrogens (nonhalogenated bisphenols, phytoestrogens, polyphenols, and mycoestrogens) other than those eluted in the $\mathrm{\alpha }\text{-fraction}$.

Of the first $11\mathrm{mL}$ collected (11 min) corresponding to the $\mathrm{\alpha }\text{-fraction}$, $1\mathrm{mL}$ was reserved for subsequent analysis of the presence and concentration of some persistent compounds. Organohalogenated compounds typically found within the $\mathrm{\alpha }\text{-fraction}$, including the organochlorine pesticides $p,p\prime$-dichlorodiphenyldichloroethylene ($p,p\prime$-DDE), hexachlorobenzene (HCB), hexachlorocyclohexane (HCH), and three PCBs congeners (138, 153, and 180), were quantified by gas chromatography-tandem mass spectrometry (GC-MS/MS), using an Agilent 7890 GC with split-splitless inlet and 7693 ALS autosampler.³⁶ These environmental pollutants were selected, not only for their activity as xenoestrogenic endocrine disruptors through different pathways, but also for their associations with other hormone-dependent tumors such as breast cancer.³⁹ $p$-Chlorodibenzophenone was used as the internal standard and a limit of detection (LOD) for all of the chemicals was set at $0.05\text{\hspace{0.17em}ng}/\mathrm{mL}$. The remaining $10\text{-mL}$ duplicates of each fraction were combined, dried, resuspended in $1\mathrm{mL}$ of experimental steroid-free medium (phenol red-free medium supplemented with $1\mathrm{mL}$ of charcoal–dextran fetal bovine serum), and tested for estrogenic activity in the E-SCREEN bioassay.⁴⁰ The E-SCREEN bioassay is an in vitro assay based on the enhanced proliferation of human breast cancer cells (MCF-7) in the presence of estrogen-active substances. The MCF-7 BUS subtype, characterized as providing the highest proliferative response to estradiol, was used. The $\mathrm{\alpha }$ and $\mathrm{\beta }$ fractions were assayed in triplicate, at three different dilutions (1:1, 1:5, and 1:10), along with negative (cells treated with experimental steroid-free mediums) and positive controls (cells treated with 100 pM of estradiol). The proliferative effect (PE) of $\mathrm{\alpha }$ and $\mathrm{\beta }$ fractions were expressed as MCF-7 cell proliferation–fold over control, transformed into estradiol equivalent units (Eeq) by reading from a dose–response curve with estradiol; and expressed as the estradiol equivalent concentration in picomolar per milliliter of serum (Eeq pM/mL).³⁶ The LODs for $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ were set up at $0.1\text{\hspace{0.17em}Eeq pM}/\mathrm{mL}$, which was the lowest concentration required to induce a significantly different PE from that observed in steroid-free control cells. For quality control, 10 serum samples were analyzed in triplicate through independent extraction, HPLC fractionation, and E-SCREEN bioassay. The interassay coefficients of variation for $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ were 15.9% and 14.4%, respectively. Researchers at the University of Granada and Biosanitary Institute of Granada (ibs.GRANADA) labs, responsible for the extraction, fractionation, and cell culture analysis (i.e., E-SCREEN) were blinded to the characteristics of the study population.

Statistical Analysis

Descriptive analyses were performed using medians and interquartile ranges (IQRs) per participant for continuous data and counts and percentages for categorical data. Chi-square tests were used for categorical variables, and Mann–Whitney tests for continuous variables. Pearson coefficients were employed to assess the correlation between continuous variables. A level equal to half of the first value above the LOD was imputed (single imputation⁴¹) for 30.1% ($n=92$) and 18.6% ($n=57$) of women with $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ determinations below the LOD, respectively, as well as for the individual contaminants (47.4% for $p,p\prime$-DDE, 26.8% for HCB, 0.0% for $\mathrm{\alpha }\text{-HCH}$, 0.0% for $\mathrm{\beta }\text{-HCH}$, 0.3% for $\mathrm{\gamma }\text{-HCH}$, 22.9% for PCB-138, and 0.0% for PCB-180). Participants were grouped into three groups of exposure: Those below the LOD ($<0.1$ Eeq pM/mL) were classified as the lowest exposure category, and those with detectable values were grouped considering the median of serum $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels based on their distributions among the controls. We used unconditional logistic regression models adjusted for potential confounders to estimate odds ratios (ORs) and 95% confidence intervals (CIs). Estimated OR for endometrial cancer risk comparing tertiles of specific organohalogenated compounds [PCB-138, PCB-180, HCB, HCH (isomers $\mathrm{\alpha }$, $\mathrm{\beta }$, and $\mathrm{\gamma }$), and $p,p\prime$-DDE] based on their control distributions, were also calculated. Basic models included adjustment for age at reference date ($<60$, 60–69, $\ge 70$ y) and education level (high school or below, some college, college or above). We further adjusted multivariable models for BMI ($<25$, 25–29.9, $\ge 30{\mathrm{kg}/\mathrm{m}}^2$), use of hormonal contraceptives (ever, never), parity (0, 1–2, $\ge 3$), and past history of other nonendometrial cancers (yes, no). These variables were selected based on their association with the outcome (endometrial cancer) or the exposure (TEXB) in univariate analyses, following the definition of a confounder. Variables associated with endometrial cancer in the bivariate analyses were BMI ($p<0.001$), use of hormonal contraceptives ($p=0.002$), parity ($p=0.036$), and past history of cancer ($p=0.039$; Table 1). BMI was also associated with $\text{TEXB-}\mathrm{\alpha }$ ($p=0.021$; Table 2). Similarly, a third model was further mutually adjusted for serum $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels ($<\text{LOD}$, below median, above median), that is, models evaluating $\text{TEXB-}\mathrm{\alpha }$ levels were adjusted for $\text{TEXB-}\mathrm{\beta }$ levels and vice versa. Given that during premenopause, estrogen levels are higher, we performed sensitivity analyses restricted to postmenopausal women. Similarly, analyses restricted to never users of postmenopausal hormone therapy, and stratified analyses by hormonal contraception (ever, never), were performed to exclude the effect of exogenous hormone use. We conducted analyses exclusively on women with no prior history of cancer to investigate the potential influence of past cancer treatments on hormonal burden. Selection bias due to the inclusion of controls with benign gynecologic conditions was also inspected using stratified analyses by type of control (gynecologic, nongynecologic). Given that xenoestrogens may accumulate in adipose tissue, stratified analyses were additionally performed by BMI (normal, overweight, obese). We performed likelihood-ratio tests for the interaction terms between main exposure variables and the specified variables in logistic models. Results with fewer than five participants per category were omitted, given that they were based on a limited number of participants. In addition, generalized additive models were used to inspect the linearity of associations of log-transformed serum $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels with endometrial cancer risk. To test for linear trend, ordinal variables were treated as continuous variables. Data were missing for $<5\%$ of participants for all variables, except for use of hormone replacement therapy, where data for $<10\%$ were missing. In models, missing covariate values were treated as distinct categories. All tests were two-tailed with a significance level of 0.05. All analyses were conducted by using Stata (version 16.0; StataCorp).

Table 1.

Descriptive characteristics [$n$ or $n$ (%)] of Spanish participants diagnosed with endometrial cancer and controls frequency-matched to cases by age ($N=306$), Screenwide case–control study, 2017–2021.

Characteristics	Controls	Cases	$p$-Value
Participants	150	156
Age at reference date (y)a			0.736b
$<60$	44 (29.3)	43 (27.6)
60–69	41 (27.3)	49 (31.4)
$\ge 70$	65 (43.3)	64 (41.0)
Median (range)	68 (47–88)	68 (42–91)
Education level achieved			0.104b
High school or below	118 (78.7)	114 (73.1)
Some college	23 (15.3)	26 (16.7)
College or above	6 (4.0)	16 (10.3)
Missing	3	0
BMI at reference age (${\mathrm{kg}/\mathrm{m}}^2$)a			$<0.001$b
$<24.99$	44 (29.3)	24 (15.4)
25–29.99	54 (36.0)	48 (30.8)
$\ge 30$	43 (28.7)	78 (50.0)
Missing	9	6
Smoking status			0.223b
Never	94 (62.7)	110 (70.5)
Ever	53 (35.3)	46 (29.5)
Missing	3	0
Live births			0.036b
Nulliparous	9 (6.0)	24 (15.4)
1–2	96 (64.0)	90 (57.7)
$\ge 3$	40 (26.7)	42 (26.9)
Missing	5	0
Age at menarche (y)			0.139b
$<11$	20 (13.3)	13 (8.3)
11–12	48 (32.0)	60 (38.5)
13–14	54 (36.0)	64 (41.0)
$>15$	20 (13.3)	12 (7.7)
Missing	8	7
Menopause status at reference agea			0.127b
Premenopausal	13 (8.7)	7 (4.5)
Postmenopausal	134 (89.3)	149 (95.5)
Missing	3	0
Ever used PMH			0.241b
Never	130 (86.7)	136 (87.2)
Ever	5 (3.3)	10 (6.4)
Missing	15	10
Ever used hormonal contraceptives			0.002b
Never	64 (42.7)	96 (61.5)
Ever	83 (55.3)	60 (38.5)
Missing	3	0
Ever diagnosed with diabetes			0.154b
Never	126 (84.0)	124 (79.5)
Ever	21 (14.0)	32 (20.5)
Missing	3	0
Ever diagnosed with hypertension			0.170b
Never	86 (57.3)	79 (50.6)
Ever	61 (40.7)	77 (49.4)
Missing	3	0
Past history of cancer			0.039b
No	132 (88.0)	132 (84.6)
Yes	11 (7.3)	24 (15.4)
Missing	7	0
Histology			NA
Endometrioid	NA	112 (71.8)
Non-endometrioid	NA	44 (28.2)

Note: BMI, body mass index; NA, not applicable; PMH, postmenopausal hormone use.

^a Reference dates were age at diagnosis for cases and age at enrollment for controls.

^b Chi-squared calculated without missing values comparing cases with controls.

Table 2.

Descriptive characteristics [$n$ (%) or median (IQR range) by serum $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels among Spanish controls ($N=150$), in the Screenwide case–control study, 2017–2021.

Characteristics	Serum $\text{TEXB-}\mathrm{\alpha }$ (Eeq pM/mL)a			$p$-Value	Serum $\text{TEXB-}\mathrm{\beta }$ (Eeq pM/mL)a			$p$-Value
	$<\text{LOD}$ ($n=50$)	$\le 5.72\times {10}^{-1}$ ($n=50$)	$>5.72\times {10}^{-1}$ ($n=50$)		$<\text{LOD}$ ($n=35$)	$\le 1.79$ ($n=57$)	$>1.79$ ($n=58$)
Age at reference date (y)b				0.968c				0.364c
$<60$	14 (28.00)	14 (28.0)	16 (32.0)		7 (20.0)	22 (38.6)	15 (25.9)
60–69	14 (28.00)	15 (30.0)	12 (24.0)		10 (28.6)	14 (24.6)	17 (29.3)
$\ge 70$	22 (44.0)	21 (42.0)	22 (44.0)		18 (51.4)	21 (36.8)	26 (44.8)
Median (range)	69 (47–83)	67 (47–88)	68 (46–82)		70 (48–83)	62 (46–88)	68 (47–82)
Education level achieved				0.165c				0.099c
High school or below	34 (68.0)	42 (84.0)	42 (84.0)		25 (71.4)	41 (71.9)	52 (89.7)
Some college	11 (22.0)	7 (14.0)	5 (10.0)		5 (14.3)	13 (22.8)	5 (8.6)
College or above	3 (6.0)	0 (0.0)	3 (6.0)		3 (8.6)	2 (3.5)	1 (1.7)
BMI at reference age (${\mathrm{kg}/\mathrm{m}}^2$)b				0.021c				0.253c
$<24.99$	10 (20.0)	22 (44.0)	12 (24.0)		7 (20.0)	17 (29.8)	20 (34.5)
25–29.99	23 (46.0)	14 (28.0)	17 (34.0)		17 (48.6)	20 (35.1)	17 (29.3)
$\ge 30$	13 (26.0)	10 (20.0)	20 (40.0)		7 (20.0)	16 (28.1)	20 (34.5)
Smoking status				0.220c				0.531c
Never	26 (52.0)	33 (66.0)	35 (70.0)		19 (54.3)	35 (61.4)	40 (69.0)
Ever	22 (44.0)	16 (32.0)	15 (30.0)		14 (40.0)	21 (36.8)	18 (31.0)
Live births				0.976c				0.567c
Nulliparous	3 (6.0)	3 (6.0)	3 (6.0)		3 (8.6)	3 (5.3)	3 (5.2)
1–2	31 (62.0)	31 (62.0)	34 (68.0)		19 (54.3)	35 (61.4)	42 (72.4)
$\ge 3$	13 (26.0)	15 (30.0)	12 (24.0)		10 (28.6)	18 (31.6)	12 (20.7)
Age at menarche (y)				0.113c				0.530c
$<11$	7 (14.0)	9 (18.0)	4 (8.0)		4 (11.4)	9 (15.8)	7 (12.1)
11–12	10 (20.0)	16 (32.0)	22 (44.0)		7 (20.0)	17 (29.8)	24 (41.4)
13–14	24 (48.0)	17 (34.0)	13 (26.0)		16 (45.7)	20 (35.1)	18 (31.0)
$>15$	6 (12.0)	6 (12.0)	8 (16.0)		5 (14.3)	8 (14.0)	7 (12.1)
Menopause status at reference ageb				0.939c				0.156c
Premenopausal	4 (8.0)	4 (8.0)	5 (10.0)		1 (2.9)	8 (14.0)	4 (6.9)
Postmenopausal	44 (88.0)	45 (90.0)	45 (90.0)		32 (91.4)	48 (84.2)	54 (93.1)
Ever used PMH	0 (0.0)	3 (6.0)	2 (4.0)	0.263c	0 (0.0)	2 (3.5)	3 (5.2)	0.422c
Ever used hormonal contraceptives	26 (52.0)	28 (56.0)	29 (58.0)	0.923c	15 (41.9)	34 (59.7)	34 (58.6)	0.342c
Ever diagnosed with diabetes	3 (6.0)	10 (20.0)	8 (16.0)	0.125c	3 (8.6)	7 (12.3)	11 (19.0)	0.385c
Ever diagnosed with hypertension	20 (40.0)	17 (34.0)	24 (48.0)	0.405c	14 (40.0)	20 (35.1)	27 (46.6)	0.498c
Past history of cancer	3 (6.0)	2 (4.0)	6 (12.0)	0.239c	3 (8.6)	2 (3.5)	6 (10.3)	0.344c
Serum $p,p\prime$-DDE (ng/mL)	0.02 (0.02–0.22)	0.14 (0.02–0.53)	0.02 (0.02–0.17)	0.268d	0.02 (0.02–0.32)	0.02 (0.02–0.24)	0.02 (0.02–0.38)	0.631d
Serum HCB (ng/mL)	0.10 (0.03–0.28)	0.16 (0.03–0.26)	0.12 (0.03–0.22)	0.278d	0.15 (0.07–0.39)	0.11 (0.03–0.22)	0.18 (0.03–0.26)	0.175d
Serum $\mathrm{\alpha }\text{-HCH}$ (ng/mL)	0.09 (0.09–0.10)	0.09 (0.09–0.10)	0.09 (0.09–0.10)	0.571d	0.09 (0.09–0.10)	0.09 (0.09–0.10)	0.09 (0.09–0.10)	0.323d
Serum $\mathrm{\beta }\text{-HCH}$ (ng/mL)	0.19 (0.16–0.31)	0.20 (0.15–0.36)	0.21 (0.16–0.31)	0.787d	0.21 (0.16–0.35)	0.20 (0.15–0.29)	0.21 (0.16–0.32)	0.690d
Serum $\mathrm{\gamma }\text{-HCH}$ (ng/mL)	0.06 (0.05–0.06)	0.05 (0.05–0.06)	0.06 (0.05–0.06)	0.115d	0.06 (0.05–0.06)	0.05 (0.05–0.06)	0.06 (0.05–0.06)	0.101d
Serum PCB-138 (ng/mL)	0.05 (0.03–0.08)	0.06 (0.03–0.09)	0.05 (0.01–0.07)	0.278d	0.06 (0.05–0.08)	0.05 (0.01–0.07)	0.06 (0.01–0.09)	0.323d
Serum PCB-180 (ng/mL)	0.17 (0.12–0.21)	0.18 (0.14–0.24)	0.15 (0.12–0.26)	0.571d	0.17 (0.12–0.23)	0.17 (0.14–0.21)	0.16 (0.13–0.25)	0.794d

Note: BMI, body mass index; HCB, hexachlorobenzene; HCH, hexachlorocyclohexane; LOD, limit of detection; PCB, polychlorinated biphenyl; PMH, postmenopausal hormone use; $p,p\prime$-DDE, $p,p\prime$-dichlorodiphenyldichloroethylene; TEXB, total effective xenoestrogen burden; $\text{TEXB-}\mathrm{\alpha }$, TEXB of the alpha fraction (represents the combined estrogenic effect of mixtures of lipophilic organohalogenated xenoestrogens); $\text{TEXB-}\mathrm{\beta }$, TEXB of the beta fraction (represents the combined estrogenic activity of endogenous hormones and more polar xenoestrogens).

^a Participants were grouped into three groups of exposure, those below the LOD were classified as the lowest exposure category ($<\text{LOD}$; $<0.1$ Eeq pM/mL), and those with detectable values were grouped considering the median of serum $\text{TEXB-}\mathrm{\alpha }$ ($\text{median}=5.72\times {10}^{-1}$ Eeq pM/mL) and $\text{TEXB-}\mathrm{\beta }$ levels ($\text{median}=1.79$ Eeq pM/mL).

^b Reference dates were age at diagnosis for cases and age at enrollment for controls.

^c Chi-squared calculated without missing values.

^d Nonparametric equality-of-medians test.

Results

Compared with controls, cases were more likely to be obese, nulliparous, have a history of cancer and were less likely to have ever used hormonal contraceptives (Table 1). History of cancer included breast cancer in 45.4% ($n=5/11$) of controls and 45.8% ($n=13/24$) of cases, whereas the rest of the past cancers included a variety of types, such as colon, ovary, and skin melanoma and nonmelanoma. Cases had similar concentrations of serum $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels, PCB-138, PCB-180, and HCB and higher $p,p\prime$-DDE and $\mathrm{\beta }\text{-HCH}$ concentrations than controls (Table 3). Descriptive characteristics of controls according to serum $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels are shown in Table 2. Significant differences were only observed in serum $\text{TEXB-}\mathrm{\alpha }$ according to BMI among the controls. Serum levels of $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ were highly correlated among controls [Pearson correlation coefficient for continuous log-transformed variables: 0.84 (95% CI: 0.78, 0.88), and among cases: 0.69 (95% CI: 0.60, 0.77)]. None of the individual contaminants were positively correlated with serum TEXB levels among the controls (Table 2).

Table 3.

Serum levels of TEXB and individual organohalogenated compounds in Spanish participants diagnosed with endometrial cancer and controls frequency-matched to cases by age ($N=306$), Screenwide case–control study, 2017–2021.

TEXB and organohalogenated compounds	Controls ($n=150$)				Cases ($n=156$)				$p$-Value
	$<\text{LOD}$		Median (IQR)	(Range)	$<\text{LOD}$		Median (IQR)	(Range)
	$n$	%			$n$	%
Serum $\text{TEXB-}\mathrm{\alpha }$ (Eeq pM/mL)	50	33.3	0.417 (0.014–0.720)	(0.014–4.500)	42	26.9	0.302 (0.014–0.617)	(0.014–3.900)	0.653a
Serum $\text{TEXB-}\mathrm{\beta }$ (Eeq pM/mL)	35	23.3	1.280 (0.237–2.270)	(0.070–10.400)	22	14.1	1.250 (0.473–1.950)	(0.070–10.70)	0.933a
Serum $p,p\prime$-DDE (ng/mL)	82	54.7	0.023 (0.023–0.278)	(0.023–2.349)	63	40.4	0.098 (0.023–0.392)	(0.023–5.000)	0.021a
Serum HCB (ng/mL)	42	28.0	0.124 (0.025–0.250)	(0.025–0.773)	40	25.6	0.135 (0.025–0.271)	(0.025–0.766)	0.986a
Serum $\mathrm{\alpha }\text{-HCH}$ (ng/mL)	0	0.0	0.094 (0.093–0.096)	(0.092–0.142)	0	0.0	0.094 (0.093–0.104)	(0.091–0.143)	0.232a
Serum $\mathrm{\beta }\text{-HCH}$ (ng/mL)	0	0.0	0.201 (0.157–0.312)	(0.106–1.105)	0	0.0	0.264 (0.178–0.442)	(0.102–1.415)	0.005a
Serum $\mathrm{\gamma }\text{-HCH}$ (ng/mL)	0	0.0	0.056 (0.053–0.058)	(0.045–0.189)	1	0.6	0.056 (0.052–0.060)	(0.021–0.391)	0.232a
Serum PCB-138 (ng/mL)	35	23.3	0.055 (0.019–0.079)	(0.006–0.230)	35	22.4	0.049 (0.019–0.091)	(0.006–0.235)	0.783a
Serum PCB-180 (ng/mL)	0	0.0	0.171 (0.129–0.231)	(0.084–0.635)	0	0.0	0.169 (0.128–0.226)	(0.077–0.813)	0.876a

Note: LOD for $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ were set at $0.1\text{\hspace{0.17em}Eeq pM}/\mathrm{mL}$, and LOD for individual organohalogenated compounds were set at $0.05\text{\hspace{0.17em}ng}/\mathrm{mL}$. HCB, hexachlorobenzene; HCH, hexachlorocyclohexane; IQR, interquartile range; LOD, limit of detection; PCB, polychlorinated biphenyl; $p,p\prime$-DDE, $p,p\prime$-dichlorodiphenyldichloroethylene; TEXB, total effective xenoestrogen burden; $\text{TEXB-}\mathrm{\alpha }$, TEXB of the alpha fraction (represents the combined estrogenic effect of mixtures of lipophilic organohalogenated xenoestrogens); $\text{TEXB-}\mathrm{\beta }$, TEXB of the beta fraction (represents the combined estrogenic activity of endogenous hormones and more polar xenoestrogens).

^a Two-sample Wilcoxon rank-sum (Mann–Whitney) test comparing cases and controls.

Median $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels did not significantly differ among cases ($3.02\times {10}^{-1}$ and $1.25\text{\hspace{0.17em}Eeq pM}/\mathrm{mL}$, for $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$, respectively) and controls ($4.17\times {10}^{-1}$ and $1.28\text{\hspace{0.17em}Eeq\hspace{0.17em}pM}/\mathrm{mL}$; $p=0.653$ and 0.933, respectively, Table 3). In adjusted models, the risk for endometrial cancer did not linearly increase with increasing category levels of both $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ ($p_{\text{linear}\text{\hspace{0.17em}trend}}=0.566$ and 0.679, respectively, Table 4). However, positive associations were observed among those in the second category of exposure compared with the lowest [$\text{OR}=2.11$ (95% CI: 1.13, 3.94) for $\text{TEXB-}\mathrm{\alpha }$, and $\text{OR}=3.32$ (95% CI: 1.62, 6.81) for $\text{TEXB-}\mathrm{\beta }$], whereas no significant associations were observed among those in the third category of exposure [$\text{OR}=1.22$ (95% CI: 0.64, 2.31) for $\text{TEXB-}\mathrm{\alpha }$, and $\text{OR}=1.58$ (95% CI: 0.75, 3.33) for $\text{TEXB-}\mathrm{\beta }$]. In mutually adjusted models for $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels, the association of $\text{TEXB-}\mathrm{\alpha }$ with endometrial cancer risk was attenuated [$\text{OR}=1.45$ (95% CI: 0.61, 3.47) for the second category of exposure compared with the lowest], whereas estimates for $\text{TEXB-}\mathrm{\beta }$ were slightly lower but still statistically significant [$\text{OR}=2.68$ (95% CI: 1.03, 6.99); Table 4]. Generalized additive models revealed an inverted-U risk trend across serum $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels, with a sharp increase in risk for exposures $<0.5$ and $1\text{\hspace{0.17em}Eeq pM}/\mathrm{mL}$, respectively, and a downturn at higher levels (Figure 1).

Table 4.

Associations between TEXB and endometrial cancer risk among participants diagnosed with endometrial cancer and controls frequency-matched to cases by age ($N=303$), Screenwide case–control study, 2017–2021.

TEXB level	Controls ($n$)a	Cases ($n$)	${\text{OR}}_{\text{basic}}$ b	(95% CI)	$p$-Valuec	${\text{OR}}_{\text{multivariate}}$ d	(95% CI)	$p$-Valuec	${\text{OR}}_{\text{mutually}}$ e	(95% CI)	$p$-Valuec
Serum $\text{TEXB-}\mathrm{\alpha }$ (Eeq pM/mL)f
$<\text{LOD}$	48	42	Ref			Ref			Ref
$\le 5.72\times {10}^{-1}$	49	69	1.72	(0.98, 3.02)	0.059	2.11	(1.13, 3.94)	0.019	1.45	(0.61, 3.47)	0.403
$>5.72\times {10}^{-1}$	50	45	1.04	(0.58, 1.86)	0.907	1.22	(0.64, 2.31)	0.543	1.06	(0.42, 2.71)	0.903
				$p_{\text{trend}}=0.941$ g			$p_{\text{trend}}=0.566$ g			$p_{\text{trend}}=0.828$ g
Serum $\text{TEXB-}\mathrm{\beta }$ (Eeq pM/mL)f
$<\text{LOD}$	33	22	Ref			Ref
$\le 1.79$	56	85	2.45	(1.27, 4.69)	0.007	3.32	(1.62, 6.81)	0.001	2.68	(1.03, 6.99)	0.044
$>1.79$	58	49	1.32	(0.67, 2.58)	0.425	1.58	(0.75, 3.33)	0.225	1.33	(0.44, 4.06)	0.616
				$p_{\text{trend}}=0.920$ g			$p_{\text{trend}}=0.679$ g			$p_{\text{trend}}=0.512$ g

Note: BMI, body mass index; CI, confidence interval; LOD, limit of detection; OR, odds ratio; Ref, reference; TEXB, total effective xenoestrogen burden; $\text{TEXB-}\mathrm{\alpha }$, TEXB of the alpha fraction (represents the combined estrogenic effect of mixtures of lipophilic organohalogenated xenoestrogens); $\text{TEXB-}\mathrm{\beta }$, TEXB of the beta fraction (represents the combined estrogenic activity of endogenous hormones and more polar xenoestrogens).

^a Three participants were dropped from the model owing to missing data given that there were no missing values in the case group.

^b Logistic models adjusted for age ($<60$, 60–69, $\ge 70$ y) and education level (high school or below, some college, college or above).

^c Obtained using Wald test from logistic models, representing whether the estimate of the category differ significantly from the reference category.

^d Logistic models adjusted for age ($<60$, 60–69, $\ge 70$ y), education (high school or below, some college, college or above), BMI ($<25$, 25–29.9, $\ge 30{\mathrm{kg}/\mathrm{m}}^2$), use of hormonal contraceptives (ever, never), parity (0, 1–2, $\ge 3$), and past history of cancer (no, yes).

^e Logistic models adjusted for age ($<60$, 60–69, $\ge 70$ y), education (high school or below, some college, college or above), BMI ($<25$, 25–29.9, $\ge 30{\mathrm{kg}/\mathrm{m}}^2$), use of hormonal contraceptives (ever, never), parity (0, 1–2, $\ge 3$), past history of cancer (no, yes), and the other fraction of TEXB ($<\text{LOD}$, below median, above median).

^f Participants were grouped into three groups of exposure, those below the LOD were classified as the lowest exposure category ($<\text{LOD}$; $<0.1$ Eeq pM/mL), and those with detectable values were grouped considering the median of serum $\text{TEXB-}\mathrm{\alpha }$ ($\text{median}=5.72\times {10}^{-1}$ Eeq pM/mL) and $\text{TEXB-}\mathrm{\beta }$ levels ($\text{median}=1.79$ Eeq pM/mL) based on their distributions among the controls.

^g Obtained using Wald test from logistic models treating ordinal variables as continuous variables, representing whether there is a statistically significant linear trend in the estimates across the ordered categories.

Figure 1.

Generalized additive models for serum levels of TEXB of (A) $\mathrm{\alpha }$ and (B) $\mathrm{\beta }$ fractions and endometrial cancer with 2 degrees of freedom among participants in the Screenwide case–control study ($n=303$), 2017–2021. Generalized additive models adjusted for age ($<60$, 60–69, $\ge 70$ y), education (high school or below, some college, college or above), BMI ($<25$, 25–29.9, $\ge 30{\mathrm{kg}/\mathrm{m}}^2$), use of hormonal contraceptives (ever, never), parity (0, 1–2, $\ge 3$), past history of cancer (no, yes), and the other fraction of TEXB ($<\text{LOD}$, below median, above median). Median for $\text{TEXB-}\mathrm{\alpha }=5.72\times {10}^{-1}$ and median for $\text{TEXB-}\mathrm{\beta }=1.79$ Eeq pM/mL. Smooth estimates are represented in solid lines and 95% confidence intervals in dashed lines. Numeric data can be found in Excel Tables S1(alpha) and S2 (beta). Note: BMI, body mass index; LOD, limit of detection; TEXB, total effective xenoestrogen burden; $\text{TEXB-}\mathrm{\alpha }$, TEXB of the alpha fraction (represents the combined estrogenic effect of mixtures of lipophilic organohalogenated xenoestrogens); $\text{TEXB-}\mathrm{\beta }$, TEXB of the beta fraction (represents the combined estrogenic activity of endogenous hormones and more polar xenoestrogens).

Associations of $\text{TEXB-}\mathrm{\alpha }$ by cancer histology in multivariate models revealed significant associations for endometrioid cancers between the second category of exposure compared with the lowest ($\text{OR}=2.37$; 95% CI: 1.19, 4.72), whereas no statistically significant associations were found for non-endometrioid cancers ($\text{OR}=1.36$; 95% CI: 0.50, 3.73). For $\text{TEXB-}\mathrm{\beta }$, significant associations were observed in both cancer groups [$\text{OR}=2.74$ (95% CI: 1.23, 6.09), and $\text{OR}=4.44$ (95% CI: 1.39, 14.24) for endometrioid and non-endometrioid, respectively, comparing the second category of exposure with the lowest; Table 5]. When serum $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels were mutually adjusted, associations for endometrioid cancers were no longer statistically significant [$\text{OR}=1.68$ (95% CI: 0.63, 4.50), and 2.07 (95% CI: 0.70, 6.14) for $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$, respectively], and the association between $\text{TEXB-}\mathrm{\beta }$ and non-endometrioid cancers remained similar [$\text{OR}=4.41$ (95% CI: 1.06, 18.34), comparing the second category of exposure with the lowest].

Table 5.

Associations for endometrial cancer and TEXB, by histological subtype ($N=303$), Screenwide case–control study, 2017–2021.

Histology	Controls ($n$)	Cases ($n$)	${\text{OR}}_{\text{basic}}$ a	(95% CI)	$p$-Valueb	${\text{OR}}_{\text{multivariate}}$ c	(95% CI)	$p$-Valueb	${\text{OR}}_{\text{mutually}}$ d	(95% CI)	$p$-Valueb
Endometrioid cancers
Serum $\text{TEXB-}\mathrm{\alpha }$ (Eeq pM/mL)e
$<\text{LOD}$	48	29	Ref			Ref			Ref
$\le 5.72\times {10}^{-1}$	49	56	1.98	(1.07, 3.66)	0.029	2.37	(1.19, 4.72)	0.014	1.68	(0.63, 4.50)	0.301
$>5.72\times {10}^{-1}$	50	26	0.85	(0.43, 1.67)	0.632	0.95	(0.45, 2.01)	0.898	0.78	(0.25, 2.39)	0.662
				$p_{\text{trend}}=0.649$ f			$p_{\text{trend}}=0.955$ f			$p_{\text{trend}}=0.339$ f
Serum $\text{TEXB-}\mathrm{\beta }$ (Eeq pM/mL)e
$<\text{LOD}$	33	16	Ref			Ref			Ref
$\le 1.79$	56	58	2.14	(1.04, 4.39)	0.038	2.74	(1.23, 6.09)	0.014	2.07	(0.70, 6.14)	0.188
$>1.79$	58	37	1.34	(0.63, 2.81)	0.446	1.59	(0.69, 3.66)	0.271	1.49	(0.40, 5.54)	0.548
				$p_{\text{trend}}=0.796$ f			$p_{\text{trend}}=0.568$ f			$p_{\text{trend}}=0.961$ f
Non-endometrioid cancers
Serum $\text{TEXB-}\mathrm{\alpha }$ (Eeq pM/mL)e
$<\text{LOD}$	48	13	Ref			Ref			Ref
$\le 5.72\times {10}^{-1}$	49	13	1.16	(0.47, 2.89)	0.742	1.36	(0.50, 3.73)	0.551	0.77	(0.22, 2.73)	0.690
$>5.72\times {10}^{-1}$	50	19	1.50	(0.65, 3.49)	0.340	1.50	(0.61, 3.70)	0.379	1.38	(0.38, 5.02)	0.623
				$p_{\text{trend}}=0.333$ f			$p_{\text{trend}}=0.386$ f			$p_{\text{trend}}=0.516$ f
Serum $\text{TEXB-}\mathrm{\beta }$ (Eeq pM/mL)e
$<\text{LOD}$	33	6	Ref			Ref			Ref
$\le 1.79$	56	27	3.65	(1.28, 10.38)	0.015	4.44	(1.39, 14.24)	0.012	4.41	(1.06, 18.34)	0.041
$>1.79$	58	12	1.38	(0.45, 4.20)	0.572	1.53	(0.46, 5.11)	0.491	1.23	(0.23, 6.52)	0.807
				$p_{\text{trend}}=0.952$ f			$p_{\text{trend}}=0.988$ f			$p_{\text{trend}}=0.344$ f

Note: Three participants were dropped from the model owing to missing data given that there were no missing values in the case group. BMI, body mass index; CI, confidence interval; LOD, limit of detection; OR, odds ratio; Ref, reference; TEXB, total effective xenoestrogen burden; $\text{TEXB-}\mathrm{\alpha }$, TEXB of the alpha fraction (represents the combined estrogenic effect of mixtures of lipophilic organohalogenated xenoestrogens); $\text{TEXB-}\mathrm{\beta }$, TEXB of the beta fraction (represents the combined estrogenic activity of endogenous hormones and more polar xenoestrogens).

^a Logistic models adjusted for age ($<60$, 60–69, $\ge 70$ y), and education level (high school or below, some college, college or above).

^b Obtained using Wald test from logistic models, representing whether the estimate of the category differ significantly from the reference category.

^c Logistic models adjusted for age ($<60$, 60–69, $\ge 70$ y), education (high school or below, some college, college or above), BMI ($<25$, 25–29.9, $\ge 30{\mathrm{kg}/\mathrm{m}}^2$), use of hormonal contraceptives (ever, never), parity (0, 1–2, $\ge 3$), and past history of cancer (no, yes).

^d Logistic models adjusted for age ($<60$, 60–69, $\ge 70$ y), education (high school or below, some college, college or above), BMI ($<25$, 25–29.9, $\ge 30{\mathrm{kg}/\mathrm{m}}^2$), use of hormonal contraceptives (ever, never), parity (0, 1–2, $\ge 3$), past history of cancer (no, yes), and the other fraction of TEXB ($<\text{LOD}$, below median, above median).

^e Participants were grouped into three groups of exposure, those below the LOD were classified as the lowest exposure category ($<\text{LOD}$; $<0.1$ Eeq pM/mL), and those with detectable values were grouped considering the median of serum $\text{TEXB-}\mathrm{\alpha }$ ($\text{median}=5.72\times {10}^{-1}$ Eeq pM/mL) and $\text{TEXB-}\mathrm{\beta }$ levels ($\text{median}=1.79$ Eeq pM/mL) based on their distributions among the controls.

^f Obtained using Wald test from logistic models treating ordinal variables as continuous variables, representing whether there is a statistically significant linear trend in the estimates across the ordered categories.

Individual xenoestrogens contained in the $\mathrm{\alpha }\text{-fraction}$ generally showed null associations with endometrial cancer, except for a few positive associations (Table 6). In particular, the ORs for endometrial cancer comparing the third with the first tertile in multivariate models were 1.68 (95% CI: 0.94, 3.00) for $p,p\prime$-DDE, 0.95 (95% CI: 0.50, 1.80) for HCB, 1.11 (95% CI: 0.62, 2.01) for $\mathrm{\alpha }\text{-HCH}$, 2.01 (95% CI: 0.99, 4.09) for $\mathrm{\beta }\text{-HCH}$, 1.25 (95% CI: 0.70, 2.25) for $\mathrm{\gamma }\text{-HCH}$, 0.96 (95% CI: 0.50, 1.88) for PCB-138, and 1.03 (95% CI: 0.56, 1.9) for PCB-180. Models further adjusted for $\text{TEXB-}\mathrm{\beta }$ yielded similar but attenuated associations [1.62 (95% CI: 0.89, 2.94) for $p,p\prime$-DDE, 0.98 (95% CI: 0.51, 1.89) for HCB, 1.16 (95% CI: 0.63, 2.12) for $\mathrm{\alpha }\text{-HCH}$, 1.82 (95% CI: 0.88, 3.76) for $\mathrm{\beta }\text{-HCH}$, 1.38 (95% CI: 0.76, 2.53) for $\mathrm{\gamma }\text{-HCH}$, 0.81 (95% CI: 0.40, 1.61) for PCB-138, and 1.04 (95% CI: 0.55, 1.95) for PCB-180, comparing the third with the first tertile; Table 5].

Table 6.

Associations between individual organohalogenated compounds and endometrial cancer risk among participants in the Screenwide case–control study ($N=303$), 2017–2021.

Organohalogenated compound	Controls ($n$)	Cases ($n$)	${\text{OR}}_{\text{basic}}$ a	(95% CI)	$p$-Valueb	${\text{OR}}_{\text{multivariate}}$ c	(95% CI)	$p$-Valueb	${\text{OR}}_{\text{mutually}}$ d	(95% CI)	$p$-Valueb
Serum $p,p\prime$-DDE (ng/mL)
$\le 0.02$ (Tertile 1)	81	63	Ref			Ref			Ref
0.02–0.15 (Tertile 2)	17	25	2.00	(0.97, 4.13)	0.062	1.53	(0.70, 3.33)	0.285	1.49	(0.67, 3.32)	0.334
$>0.15$ (Tertile 3)	49	68	1.97	(1.15, 3.38)	0.014	1.68	(0.94, 3.00)	0.082	1.62	(0.89, 2.94)	0.113
				$p_{\text{trend}}=0.014$ e			$p_{\text{trend}}=0.080$ e			$p_{\text{trend}}=0.112$ e
Serum HCB (ng/mL)
$\le 0.07$ (Tertile 1)	50	58	Ref			Ref			Ref
0.07–0.22 (Tertile 2)	49	47	0.83	(0.47, 1.47)	0.525	1.10	(0.59, 2.05)	0.755	1.06	(0.56, 2.00)	0.856
$>0.22$ (Tertile 3)	48	51	0.94	(0.52, 1.69)	0.836	0.95	(0.50, 1.80)	0.877	0.98	(0.51, 1.89)	0.950
				$p_{\text{trend}}=0.825$ e			$p_{\text{trend}}=0.887$ e			$p_{\text{trend}}=0.956$ e
Serum $\mathrm{\alpha }\text{-HCH}$ (ng/mL)
$\le 0.093$ (Tertile 1)	49	60	Ref			Ref			Ref
0.093–0.095 (Tertile 2)	49	31	0.52	(0.28, 0.96)	0.037	0.52	(0.27, 1.01)	0.052	0.53	(0.27, 1.03)	0.063
$>0.095$ (Tertile 3)	49	65	1.10	(0.64, 1.90)	0.721	1.11	(0.62, 2.01)	0.719	1.16	(0.63, 2.12)	0.642
				$p_{\text{trend}}=0.656$ e			$p_{\text{trend}}=0.660$ e			$p_{\text{trend}}=0.607$ e
Serum $\mathrm{\beta }\text{-HCH}$ (ng/mL)
$\le 0.17$ (Tertile 1)	49	37	Ref			Ref			Ref
0.17–0.29 (Tertile 2)	49	48	1.44	(0.78, 2.69)	0.247	1.41	(0.72, 2.75)	0.310	1.28	(0.64, 2.55)	0.479
$>0.29$ (Tertile 3)	49	71	2.41	(1.26, 4.62)	0.008	2.01	(0.99, 4.09)	0.054	1.82	(0.88, 3.76)	0.108
				$p_{\text{trend}}=0.007$ e			$p_{\text{trend}}=0.053$ e			$p_{\text{trend}}=0.103$ e
Serum $\mathrm{\gamma }\text{-HCH}$ (ng/mL)
$\le 0.053$ (Tertile 1)	49	53	Ref			Ref			Ref
0.053–0.057 (Tertile 2)	48	39	0.77	(0.43, 1.38)	0.380	0.89	(0.47, 1.67)	0.715	0.95	(0.50, 1.83)	0.887
$>0.57$ (Tertile 3)	50	64	1.17	(0.68, 2.01)	0.578	1.25	(0.70, 2.25)	0.450	1.38	(0.76, 2.53)	0.291
				$p_{\text{trend}}=0.557$ e			$p_{\text{trend}}=0.438$ e			$p_{\text{trend}}=0.284$ e
Serum PCB-138 (ng/mL)
$\le 0.04$ (Tertile 1)	49	61	Ref			Ref			Ref
0.04–0.07 (Tertile 2)	49	42	0.66	(0.37, 1.19)	0.168	0.73	(0.38, 1.40)	0.349	0.72	(0.37, 1.40)	0.332
$>0.07$ (Tertile 3)	49	53	0.87	(0.47, 1.59)	0.647	0.96	(0.50, 1.88)	0.914	0.81	(0.40, 1.61)	0.541
				$p_{\text{trend}}=0.640$ e			$p_{\text{trend}}=0.948$ e			$p_{\text{trend}}=0.566$ e
Serum PCB-180 (ng/mL)
$\le 0.14$ (Tertile 1)	48	50	Ref			Ref			Ref
0.14–0.20 (Tertile 2)	49	54	0.98	(0.56, 1.73)	0.952	1.16	(0.63, 2.15)	0.632	1.06	(0.56, 1.98)	0.865
$>0.20$ (Tertile 3)	50	52	0.96	(0.54, 1.69)	0.884	1.03	(0.56, 1.91)	0.923	1.04	(0.55, 1.95)	0.910
				$p_{\text{trend}}=0.884$ e			$p_{\text{trend}}=0.927$ e			$p_{\text{trend}}=0.911$ e

Note: Three participants were dropped from the model owing to missing data given that there were no missing values in the case group. BMI, body mass index; CI, confidence interval; HCB, hexachlorobenzene; HCH, hexachlorocyclohexane; LOD, limit of detection; OR, odds ratio; PCBs, polychlorinated biphenyls; $p,p\prime$-DDE, $p,p\prime$-dichlorodiphenyldichloroethylene; Ref, reference; TEXB, total effective xenoestrogen burden; $\text{TEXB-}\mathrm{\alpha }$, TEXB of the alpha fraction (represents the combined estrogenic effect of mixtures of lipophilic organohalogenated xenoestrogens); $\text{TEXB-}\mathrm{\beta }$, TEXB of the beta fraction (represents the combined estrogenic activity of endogenous hormones and more polar xenoestrogens).

^a Logistic models adjusted for age ($<60$, 60–69, $\ge 70$ y), and education level (high school or below, some college, college or above).

^b Obtained using Wald test from logistic models, representing whether the estimate of the category differ significantly from the reference category.

^c Logistic models adjusted for age ($<60$, 60–69, $\ge 70$ y), education (high school or below, some college, college or above), BMI ($<25$, 25–29.9, $\ge 30{\mathrm{kg}/\mathrm{m}}^2$), use of hormonal contraceptives (ever, never), parity (0, 1–2, $\ge 3$), and past history of cancer (no, yes).

^d Logistic models adjusted for age ($<60$, 60–69, $\ge 70$ y), education (high school or below, some college, college or above), BMI ($<25$, 25–29.9, $\ge 30{\mathrm{kg}/\mathrm{m}}^2$), use of hormonal contraceptives (ever, never), parity (0, 1–2, $\ge 3$), past history of cancer (no, yes), and fraction of $\text{TEXB-}\mathrm{\beta }$ [$<\text{LOD}$ ($<0.1$ Eeq pM/mL,) below median, above median; $\text{TEXB-}\mathrm{\beta }$ $\text{median}=1.79$ Eeq pM/mL].

^e Obtained using Wald test from logistic models treating ordinal variables as continuous variables, representing whether there is a statistically significant linear trend in the estimates across the ordered categories.

Stratified analyses by type of control and menopausal status yielded no significant interactions, although estimates for the $\mathrm{\beta }$ fraction were higher among nongynecologic controls (Figure 2). Results by BMI revealed lower estimates among those participants with obesity ($\text{BMI}\ge 30{\mathrm{kg}/\mathrm{m}}^2$), although interactions were not statistically significant. Similarly, estimates were generally higher among hormonal contraceptive users than among nonusers, and a 6-fold increase in risk was observed for $\text{TEXB-}\mathrm{\beta }$ comparing the second category of exposure with the lowest, although the interactions were not statistically significant.

Figure 2.

Odds ratios (ORs) for endometrial cancer comparing categories (above and below the median compared with below the LOD) of TEXB of $\mathrm{\alpha }$ and $\mathrm{\beta }$ fractions by subgroups, among participants in the Screenwide case–control study ($n=303$), 2017–2021. Numbers of participants in each category are provided for controls (Co) and cases (Ca). Logistic models adjusted for age ($<60$, 60–69, $\ge 70$ y), education (high school or below, some college, college or above), BMI ($<25$, 25–29.9, $\ge 30{\mathrm{kg}/\mathrm{m}}^2$), use of hormonal contraceptives (ever, never), parity (0, 1–2, $\ge 3$), past history of cancer (no, yes), and the other fraction of TEXB ($<\text{LOD}$, below median, above median). Median for $\text{TEXB-}\mathrm{\alpha }=5.72\times {10}^{-1}$ and median for $\text{TEXB-}\mathrm{\beta }=1.79$ Eeq pM/mL. Results for normal BMI are omitted owing to the limited number of participants in the case categories. All $p$-interactions $>0.05$. Note: BMI, body mass index; CI, confidence interval; HC, hormonal contraception; PMH, postmenopausal hormone use; LOD, limit of detection; Ref, reference; TEXB, total effective xenoestrogen burden; $\text{TEXB-}\mathrm{\alpha }$, TEXB of the alpha fraction (represents the combined estrogenic effect of mixtures of lipophilic organohalogenated xenoestrogens); $\text{TEXB-}\mathrm{\beta }$, TEXB of the beta fraction (represents the combined estrogenic activity of endogenous hormones and more polar xenoestrogens).

Discussion

This study has identified for the first time a positive association between serum TEXB levels and endometrial cancer risk. In particular, an inverted-U risk trend across serum $\text{TEXB-}\mathrm{\alpha }$ levels was observed, and women in the second category of exposure had a 45% increase in risk, but that was not seen among those in the highest exposure categories compared with the lowest exposure category. $\text{TEXB-}\mathrm{\beta }$ values were also positively associated with endometrial cancer, with an almost 3-fold increase in risk among women in the second category of exposure compared with the lowest. Positive associations were somewhat expected, given that circulating endogenous estrogens were included in the $\mathrm{\beta }$ fraction, and that exposure to circulating endogenous estrogens is an established risk factor for endometrial cancer.^13,16,18 Associations by endometrial type revealed that estimates for $\text{TEXB-}\mathrm{\beta }$ were higher among non-endometrioid cancers than for endometrioid cancers, whereas typically non-endometrioid cancers have been considered less driven by estrogens.^8,19 However, our results were limited by small sample sizes for type II cancers. Other stratified analyses yielded estimates generally $>1$ and all $p$-interactions were not statistically significant.

An inverted-U risk trend for endometrial cancer in relation to combined exposure to serum lipophilic xenoestrogens ($\text{TEXB-}\mathrm{\alpha }$) is biologically plausible, given that both endogenous hormones and some EDCs have shown to exert nonclassical dose–response curves, with lower doses exerting more potent effects than higher doses.⁴² Nonmonotonic dose–response curves are common in endocrinology. Thus, in vivo studies in rats have shown that estrogen’s impact on mammary gland development varies with dosage, with low-to-moderate doses promoting terminal end bud formation and ductal elongation, whereas higher doses hinder these processes.⁴³ For example, in the US Food and Drug Administration Clarity study evaluating Sprague–Dawley rats, perinatal exposure to bisphenol A (BPA) was found to induce breast cancer at the lowest tested dose, but not at the highest doses.⁴⁴ Similarly, related studies revealed nonmonotonic effects of BPA on the developing rat mammary gland.⁴⁵ Other nonmonotonic dose–response curves have also been observed in epidemiological studies for different EDCs and several outcomes, including diabetes^46–48 and BMI,^49,50 among others.⁵¹ These nonmonotonic dose–response curves suggest intricate interactions in which estrogen can act as both agonist and antagonist on multiple targets.^43,44,52 The nonmonotonic dose response observed in our study could be attributed to competition to specific receptors between endogenous hormones and xenoestrogens present in serum samples.⁵¹ In situations where endogenous hormones do not fully saturate these receptors, xenoestrogens may bind to the available ones, thereby enhancing the overall cellular response. Under these circumstances, at low-to-moderate TEXB levels, xenoestrogens mimic the actions of endogenous estrogen by promoting endometrial proliferation, as suggested using mathematical models.^51,53 Tumor cell proliferation effects in endometrial cancer are mainly mediated through $\text{ER}\mathrm{\alpha }$, whereas the role of $\text{ER}\mathrm{\beta }$ remains inconclusive.⁵⁴ However, when TEXB levels are high, xenoestrogens may surpass natural ligands in $\text{ER}\mathrm{\alpha }$ binding competition, potentially leading to an attenuation of the overall biological response owing to their lower estrogenic activity and/or potency. This competition could lead to partial antagonistic behavior, reducing the stimulatory impact of xenoestrogen in endometrial tissues and, in turn, potentially diminishing the risk of endometrial cancer.^51,53

Positive estimates were observed for $\mathrm{\beta }\text{-HCH}$ and $p,p\prime$-DDE, whereas for the rest of identified chemical compounds, no clear associations with endometrial cancer were found. The individual chemicals, analyzed in this study, were also not associated with $\text{TEXB-}\mathrm{\alpha }$, consistent with what has been previously described in the literature.^35,36,55 Several factors, including possible interactions established between the chemical compounds involving additive, synergistic or antagonistic mechanisms, and/or unmeasured substances, could explain the absence of agreement. Humans are exposed to both lipophilic and more polar xenoestrogens EDCs, such as PCB, brominated flame retardants, alkylphenolic compounds, polycyclic aromatic hydrocarbons compounds, parabens, phthalates, BPA, pesticides, and certain metals, among others.^21,42,56 Epidemiology literature evaluating xenoestrogens in endometrial cancer is generally scarce. Two human studies evaluated exposure to individual organochlorines and PCBs using serum exposure biomarkers, yielding null associations in accordance with results on individual contaminants.^27,28 Other epidemiological studies have evaluated organochlorines and PCBs using dietary variables,²³ as well as adipose tissue exposure biomarkers among reduced samples sizes,²⁵ which also yielded null results. Dietary assessments have been also used to assess personal exposure to cadmium, with positive results in a meta-analyses of two studies,²⁹ but not in following large prospective studies.^22,24 Urinary concentrations of BPA, parabens, and phthalate metabolites have been recently evaluated in a study among 139 case–control sets, yielding null results for all single chemicals evaluated, except for mono-$n$-butyl phthalate (MnBP) and dibutyl phthalate (DBP) excretion. These had statistically significant associations for the second tertiles of exposure, but not the third tertiles,²⁶ also suggesting an inverted-U risk shape pattern consistent with our results. The null results observed for parabens by Sarink et al.²⁶ contrasted with the positive associations observed by Dogan et al.,³⁰ although Dogan et al. assessed individual chemical exposure in endometrial and myometrial tissue samples in a smaller study sample size (33 cases and 49 controls). Urinary biomarkers have been also used to evaluate exposure to few alkylphenols in 49 endometrial cancer cases, showing positive associations for chemical exposures above the median.³¹ Interestingly, most of the above individual chemical compounds tested elutes in the $\mathrm{\beta }\text{-HPLC}$ fraction. In addition to the heterogeneity of exposure designs and assessments, most epidemiological studies assessing EDCs and endometrial cancer have focused on the concentration of individual chemicals or a limited number of chemicals. Historically, this one-chemical-at-a-time approach to exposure assessment has led to insufficient knowledge of the human health effects caused by exposure to mixtures. There was a need to implement improved tools, such as in vitro bioassays, that allow the analysis of both exposure and biological activity resulting from the mixture of chemicals present in human biological samples.⁵⁷ TEXB of the $\mathrm{\alpha }\text{-fraction}$ represents the combined estrogenic effect of organohalogenated lipophilic xenoestrogens mixtures, whereas the TEXB of the $\mathrm{\beta }$ fraction represents the combined estrogenic activity of endogenous hormones plus more polar xenoestrogens. Thus, the TEXB bioassay represents an effective way to explore the combined impact of these compounds and to overcome the unpredictability of xenoestrogen interactions derived from possible additive, synergistic, or antagonistic mechanisms present in complex mixtures. These interactions would explain the null results found when EDCs are analyzed individually.⁵⁸

The E-SCREEN assay is a cell proliferation assay based on the enhanced proliferation of human breast cancer cells (MCF-7) in the presence of estrogen-active substances; thus, it primarily measures estrogenic activity but no other hormonal activities that could be relevant to the development of endometrial cancer. The assay’s specificity to estrogenic activity may overlook the opposing effect of progesterone or the influence of other hormones, such as androgens.^12,14 To attain a more comprehensive understanding of the hormonal context of endometrial cancer, future research could explore complementary assays or studies targeting other hormonal pathways and their potential interactions.

Similar to findings shown in a previously reported breast cancer case–control study,³⁶ the combined estrogenic activity in serum extracts from cases and controls in this study was not associated with age, or with any other participant characteristic, except BMI. Interestingly, the associations were attenuated among obese women in stratified analyses. Obesity could act as a relevant confounder, given that it is associated with both case–control status and the levels of environmental lipophilic chemicals. The conversion of androgens to estrogens in adipose tissue could contribute, among others, to the role of obesity in endometrial cancer.¹⁰ Aromatase levels increase with age and adiposity, and therefore this conversion results in estrogen-induced endometrial proliferation, as shown in some epidemiological studies.⁵⁹ Obesity, on the other hand, could lead to a lower serum concentration of persistent environmental chemicals, given that these compounds would be diluted in human adipose tissue.⁴⁸ However, obese controls in our study were more likely to have higher serum $\text{TEXB-}\mathrm{\alpha }$ than nonobese controls, and adjusting estimates for BMI yielded stronger associations. Other studies, however, have reported no uniform relationship between chemical levels in serum and adipose tissue, highlighting that the adipose-to-serum ratios of these chemicals vary considerably among women.⁶⁰ Furthermore, obesity could potentially serve as a mediator in the present study, given the influence of certain persistent organic pollutants as obesogens leading to weight gain,⁴⁸ adding complexity to the analysis of the association between TEXB and endometrial cancer.

We observed a positive correlation between serum $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels, and when we accounted for $\text{TEXB-}\mathrm{\beta }$ levels, we noticed a reduction in the strength of the association between $\text{TEXB-}\mathrm{\alpha }$ and endometrial cancer risk. This association suggests that both $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels could be affected by an unspecified common exposure to both lipophilic and more polar xenoestrogens.³⁵ However, previous extensive assessment corroborated that lipophilic organohalogenated xenoestrogens that elute in the $\mathrm{\alpha }\text{-fraction}$ are distinct from more polar xenoestrogens that elute, together with endogenous sex steroids, in the $\mathrm{\beta }\text{-fraction}$.³⁵ $\text{TEXB-}\mathrm{\beta }$ levels are also influenced by the endogenous hormone levels of the participants at the time of blood collection, which could directly modulate the risk of endometrial cancer. Although the adjustment of $\text{TEXB-}\mathrm{\alpha }$ estimates for $\text{TEXB-}\mathrm{\beta }$ can help mitigate the influence of endogenous hormones and correlated xenoestrogens found in the $\mathrm{\beta }$ fraction, this correction can also introduce M-bias or collider bias, resulting in a misleading attenuation of the results.⁶¹ Recent literature using simulated data indicates that when a collider acts as a confounder, giving priority to confounding control should supersede the avoidance of M-bias.⁶² Nonetheless, the effect of $\text{TEXB-}\mathrm{\alpha }$ on endometrial cancer is expected to fall within the estimates with and without adjustment for $\text{TEXB-}\mathrm{\beta }$.⁶¹ Estrogenicity was below the LOD for $\text{TEXB-}\mathrm{\alpha }$ in one-third of the serum samples and in 18.6% for $\text{TEXB-}\mathrm{\beta }$. As stated above, the HPLC methodology was developed to separate lipophilic xenoestrogens ($\mathrm{\alpha }\text{-fraction}$) from sex steroid hormones and more polar xenoestrogens ($\mathrm{\beta }\text{-fraction}$).³⁵ Many of the chemicals that elute in HPLC ($\mathrm{\alpha }\text{-fraction}$) have been subject to restrictive measures on their production and use owing to their persistence and suspected health effects.⁶³ Accordingly, $\text{TEXB-}\mathrm{\alpha }$ levels in controls of the present study were lower than those in a previous similar population that was enrolled about a decade earlier and evaluated using the same methodology.³⁶ Consequently, the general population is exposed to relatively low but frequent doses of these compounds, which would explain the lower sensitivity of the bioassay to obtain a different signal than the negative control. This may justify why the serum TEXB values obtained in these participants are lower than those found in the multicase–control study, MCC-Spain. In that study, healthy women and women with breast cancer were enrolled between 2008 and 2013, and geometric means of 8.32 and $9.94\text{\hspace{0.17em}Eeq pM}/\mathrm{mL}$ were observed among cases for $\text{TEXB-}\mathrm{\alpha }$ and $\text{TEXB-}\mathrm{\beta }$ levels, respectively, and 2.99 and $5.96\text{\hspace{0.17em}Eeq pM}/\mathrm{mL}$, respectively, among controls.³⁶

The strengths of this study include the detailed assessment of covariates that allowed several sensitivity analyses and the first-time use of a reliable biomarker for the combined estrogenic effect of EDC mixtures in endometrial cancer risk. Our sample size was relatively large, considering the previous literature on xenoestrogens, as well as the serum-demanding and time-consuming characteristics of the TEXB bioassay. However, our sample size was too small to evaluate certain associations, especially for stratified analyses. Serum samples were collected at diagnosis for cases owing to the case–control design, which could lead to reverse causation bias if hormones or xenoestrogens levels changed after the onset of the disease. However, we expect this bias to be relatively low given that all serum samples were collected before any surgery, chemotherapy, or hormone therapy for cancer. Moreover, if serum estrogen depletion occurs as a result of tumors’ reliance on estrogen, this would lead to inverse estimates, further bolstering our conclusions regarding the positive associations. However, we used a single serum specimen as a surrogate for usual xenoestrogen exposures, and consequently, these exposures may not accurately represent the etiologic window for a cancer with a lengthy latent period. Endogenous estrogens and hydrophilic xenoestrogens primarily account for the hormonal activity observed in the $\mathrm{\beta }\text{-fraction}$. Sexual hormones exhibit higher potency compared with nonpersistent xenoestrogen components that are detected within the $\mathrm{\beta }\text{-fraction}$, such as BPA, parabens, and phthalates. Human exposure to these nonpersistent substances is usually assessed through repeated urine samples because they all have relatively short half-lives and are swiftly eliminated as metabolites in this biological matrix. Obtaining repeated samples from participants would have been highly beneficial in characterizing this exposure, particularly considering the long latency period of endometrial cancer. Adjustment variables were based on self-reported data, and therefore recall bias can result in residual confounding. The inclusion of controls with benign gynecologic conditions could lead to selection bias; these controls may have a different distribution of risk factors compared with the general population, especially those related to hormonal exposures (because they are often prescribed hormonal compounds), which could lead to attenuated estimates. Restricted analyses to controls without these benign gynecologic conditions, however, yielded consistent estimates. Hospital controls also may not be representative of the population hormonal burden, and further evaluation using prospective designs may yield additional insights.

Conclusions

We evaluated a reliable marker of the combined estrogenic effect of EDC mixtures in relation to endometrial cancer etiology in a Spanish case–control study. Both the combined estrogenic activity of mixtures of persistent xenoestrogens and endogenous hormones together with nonpersistent xenoestrogens in serum samples were positively associated with endometrial cancer risk with an inverted-U dose–response pattern. The use of in vitro bioassays on human samples may provide relevant insights in the way we understand the negative impact of chemical mixtures on human health effects.

Conclusions and opinions are those of the individual authors and do not necessarily reflect the policies or views of EHP Publishing or the National Institute of Environmental Health Sciences.