{"paper_id":"3bf8131c-1bb2-41f2-af08-6d054e16ca2e","body_text":"Yang and Chen  BMC Women’s Health          (2023) 23:261  \nhttps://doi.org/10.1186/s12905-023-02381-5\nRESEARCH Open Access\n© The Author(s) 2023. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which \npermits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the \noriginal author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or \nother third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line \nto the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory \nregulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this \nlicence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco \nmmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.\nBMC Women’s Health\nUrinary phytoestrogens and the risk \nof uterine leiomyomata in US women\nFang Yang1 and Youguo Chen1* \nAbstract \nBackground Uterine leiomyomata (UL) is a common gynecological disease in women. Studied on the relationship \nbetween single metabolites of urinary phytoestrogens and UL, especially for the combined effects of mixed metabo-\nlites on UL still are insufficient.\nMethods In this cross-sectional study, we included 1,579 participants from the National Health and Nutrition Exami-\nnation Survey. Urinary phytoestrogens were assessed by measuring urinary excretion of daidzein, genistein, equol, \nO-desmethylangolensin, enterodiol and enterolactone. The outcome was defined as UL. Weighted logistic regres-\nsion was used to analyze the association between single metabolites of urinary phytoestrogens and UL. Notably, \nwe adopted the weighted quantile sum (WQS) regression, Bayesian kernel machine regression (BKMR), and quantile \ng-computation (qgcomp) models, to investigate the combined effects of six mixed metabolites on UL.\nResults The prevalence of UL was approximately 12.92%. After adjusting age, race/ethnicity, marital status, drinking \nstatus, body mass index, waist circumference, menopausal status, ovary removed status, use of female hormones, \nhormones/hormone modifiers, total energy, daidzein, genistein, O-desmethylangolensin, enterodiol, and enterolac-\ntone, the association of equol with UL was significant [Odds ratio (OR) = 1.92, 95% confidence interval (CI): 1.09–3.38]. \nIn the WQS model, mixed metabolites of urinary phytoestrogen had a positive association with UL (OR = 1.68, 95%CI: \n1.12–2.51), with the highest weighted chemical of equol. In the gpcomp model, equol had the largest positive \nweight, followed by genistein and enterodiol. In the BKMR model, equol and enterodiol have positive correlation on \nUL risk, while enterolactone has negative correlation.\nConclusion Our results implied a positive association between the mixed metabolites of urinary phytoestrogen and \nUL. This study provides evidence that urinary phytoestrogen-metabolite mixture was closely related to the risk of \nfemale UL.\nKeywords Uterine leiomyomata, Urinary phytoestrogens, NHANES, Weighted quantile sum, Relationship\nBackground\nUterine leiomyomata (UL) are the most common solid \ntumors in women [1, 2]. It is estimated that up to 80% of \nwomen will develop UL during their lifetime [1, 3], with \n25–30% of them experiencing significant symptoms, \nincluding chronic pelvic pain, dysmenorrhea, abnormal \nvaginal discharge, and abnormal menstruation [3, 4]. UL \ncontinues to pose a serious disease burden for women.\nAlthough the underlying pathology of UL is not par -\nticularly clear, it has been suggested to be an estrogen-\ndependent tumor [5 ]. Phytoestrogens are a group of \nplant compounds that are similar in chemical structure \nto mammalian estrogens, and they can be absorbed \nfrom food, circulate in the blood, and are excreted in the \nurine [6– 8]. Previous studies have reported the effects of \n*Correspondence:\nYouguo Chen\nygchensu@163.com\n1 Department of Obstetrics and Gynecology, The First Affiliated Hospital \nof Soochow University, No.899 Pinghai Road, Suzhou 215006, P . R. China\n\nPage 2 of 11Yang and Chen  BMC Women’s Health          (2023) 23:261 \nphytoestrogens on UL [9 , 10]. For example, a case–con -\ntrol study included 328 eligible subjects from the Diag -\nnostic Unit of the University Hospital of the West Indies, \nfound that there was no association of urine daidzein, \ngenistein, equol, enterolactone, total phytoestrogens and \nuterine fibroid (diagnosed by abdominal and/or vaginal \nultrasonography) using binary logistic regression analy -\nsis [9 ]. A cross-sectional study contains 1,204 partici -\npants performed by Zhang Y, et  al., implied that equol \nwas significantly associated with the risk of UL after \nadjusting for age, race, pregnant status, ovary removed \nstatus, use of female hormones, body mass index (BMI), \nmenopausal status and urinary creatinine levels [10]. \nThere are still contradictions regarding the effects of \nphytoestrogens on uterine fibroids. Importantly, these \nstudies on the association of phytoestrogens with UL \nhave focused on the effects of single chemicals [9 , 10]. \nGenerally speaking, humans are often exposed to many \nchemicals simultaneously, and the cumulative effect of \nmultiple chemicals is of concern [11]. Nevertheless, little \nis known about the mixed effects of multiple chemicals \nin phytoestrogens on UL.\nHerein, this study aimed at investigating the relation -\nship between single metabolites of urinary phytoestro -\ngens and UL in US women, and the combined effects of \nmixed metabolites on UL risk.\nMethods\nPopulation selection\nIn this cross-sectional study, all data were drawn from \nthe National Health and Nutrition Examination Survey \n(NHANES) database. The NHANES is a cross-sectional \nsurvey conducted by the National Center for Health \nStatistics (NCHS) of the Centers for Disease Control \nand Prevention using a multilayer probability sampling \ndesign, which aim to assess the health and nutritional \nstatus of adults and children in the United States [12]. \nThe NHANES survey combines interviews and physical \nexaminations [13]. The requirement of ethical approval \nand informed consent of the subjects for this was waived \nby the Institutional Review Board of The First Affiliated \nHospital of Soochow University, because the data was \naccessed from NHANES (a publicly available database). \nAll methods were carried out in accordance with relevant \nguidelines and regulations.\nIn this study, we used data from four cycles of the \nNHANES database (NHANES 1999–2000, 2001–\n2002, 2003–2004, 2005–2006). For participants in the \nNHANES database, only women aged 20–54 were asked \ndiagnostic questions about UL (n = 6,508). Participants \nwho met one of the following criteria were excluded: \n(1) Women without measurement of urinary phytoes -\ntrogen concentrations; (2) Women without assessment \nof UL; (3) Women with missing information of covari -\nates related to UL. Ultimately, 1,579 participants were \nincluded in this study (Fig. 1).\nAssessment of urinary phytoestrogen\nUrinary phytoestrogens were assessed by measur -\ning urinary excretion of isoflavones (including daid -\nzein, genistein, equol, and O-desmethylangolensin) \nand enterolignans (including enterodiol and enterolac -\ntone) [14]. The collection of urine specimens was car -\nried out in the Mobile Examination Centers, and stored \nat -20  °C until analyzed [14]. The analyses of urinary \nexcretion were accomplished by using the high-perfor -\nmance liquid chromatography (HPLC)-tandem mass \nspectrometric (MS) detection in the survey 1999–2004 \nand HPLC-atmospheric pressure photoionization- MS \nin the survey 2005–2006 [15]. For 1,579 participants of \nthis study, 1 participant were below the lower limit of \ndetection (LOD) for daidzein (0.40  ng/mL), 9 partici -\npants were below the lower LOD for genistein (0.20 ng/\nmL), 2 participants were below the lower LOD for equol \n(0.06 ng/mL), 29 participants were below the lower LOD \nfor O-desmethylangolensin (0.20  ng/mL), 0 participants \nwere below the lower LOD for enterodiol (0.04  ng/mL) \nand 0 participants were below the lower LOD for entero -\nlactone (0.10 ng/mL) [16]. In the case of results below the \nLOD, the value of this variable is the LOD divided by the \nsquare root of two (https:// wwwn. cdc. gov/ Nchs/ Nhanes/ \n1999- 2000/ PHPYPA. htm# URXDAZ). The concentration \nof daidzein, genistein, equol, O-desmethylangolensin, \nenterodiol, and enterolactone in urinary phytoestrogens \nwas corrected by creatinine in this study. Geometric \nmean and tertiles of each phytoestrogen metabolite (ug/g \ncreatinine) were presented in Supplemental Table 1.\nAssessment of uterine leiomyomata\nThe outcome was considered as UL. Participants in the \nNHANES database were classified as patients with UL \nwhen they answered “Yes” to the question “Has a doctor \nor other health professional ever told you that you had \nuterine fibroids?” .\nPotential covariates\nWe extracted some characteristics of participants from \nthe NHANES database, including age (years), race/eth -\nnicity (non-Hispanic White/ non-Hispanic Black/ oth -\ners), marital status (married/ never married/ others), \neducation level [high school and below/ high school \ngrad/ general educational development (GED) or equiv -\nalent/ some college or associate of arts (AA) degree/\ncollege graduate or above], poverty-to-income ratio \n(PIR, < 1.0/ ≥ 1.0), smoking status (yes/no), drinking sta -\ntus (yes/no), BMI (kg/m 2), waist circumference (cm), \n\nPage 3 of 11\nYang and Chen  BMC Women’s Health          (2023) 23:261 \n \ncotinine (ng/mL), age at menarche (years), menopausal \nstatus (yes/no), ovary removed status (yes/no), hyster -\nectomy (yes/no), use of female hormones (yes/no), hor -\nmones/hormone modifiers, pregnancy status (yes/no), \nnumber of gravidities, fiber (gm) and total energy (kcal). \nPIR was classified as in the NHANES database ≥ 1.0 \n(meaning household income was above the poverty line) \nand < 1.0 (meaning household income is at or below the \npoverty line). Smoking status and drinking status in \nthe NHANES database was based on participants’ self‐\nreport. BMI was calculated as weight (kg) divided by \nheight squared  (m2). Cotinine was assessed measured \nin serum using isotope dilution-high performance liq -\nuid chromatography/atmospheric pressure chemical \nionization tandem mass spectrometry. Similarly, when \nthe result is below the LOD, the value of cotinine is the \nLOD divided by the square root of two. Information on \nage at menarche, menopausal status, ovary removed \nstatus, use of female hormones, hormones/hormone \nmodifiers, pregnancy status and number of gravidities \nwas obtained from the reproductive health question -\nnaire. Use of female hormones was judged by self-report \n\" Have you/Has SP ever used female hormones such as \nestrogen and progesterone?\" and drug code 97–101 in the \nNHANES database. Hormones/hormone modifiers was \ndefined according to drug codes [97–98, 97–103, 97–288, \n97–295, 97–377, 97–411, 97–413, 97–414, 97–416, \n97–417, 97–418, 97–420, 97–422, 97–423, 97–426, \n97–495].\nStatistical analysis\nGiven the nature of the complex sampling of the \nNHANES database, we used a weighted analysis: \nweight variables for the urinary metabolites measure -\nment (WTSB2YR and WTSPH2YR) and study design \nvariables (SDMVPSU and SDMVSTRA). The measure -\nment data were tested for normality using Kolmogo -\nrov–Smirnov, and normally distributed measurement \ndata were described as mean (standard error) [Mean \n(SE)] and compared between two groups using inde -\npendent samples t-test; non-normal data were described \nas median and quartiles [M (Q1, Q3)] and compared \nbetween groups using Mann–Whitney U rank sum test. \nCategorical data were described as number of cases and \ncomposition ratio N (%) and compared between groups \nusing chi-square test and rank data using rank sum test. \nIn the present study, we adopted chain equation multiple \ninterpolation method based on random forest for some \nmissing data of the variables. The miceforest package in \npython is used for interpolation processing (https:// pypi. \nFig. 1 Flowchart of population selection. NHANES = National Health and Nutrition Examination Survey; UL = uterine leiomyomata\n\nPage 4 of 11Yang and Chen  BMC Women’s Health          (2023) 23:261 \norg/ proje ct/ micef orest/). A sensitivity analysis was per -\nformed on the data before and after interpolation (Sup -\nplemental Table  2). SAS (version 9.4), Python (version \n3.9) and R (version 4.0) software were used for statistical \nanalyses. P < 0.05 was considered as statistically signifi -\ncant difference.\nFirst, we performed weighted univariate logistic regres-\nsion to screen covariates. Then, weighted logistic regres -\nsion was used to analyze the association between single \nmetabolites of urinary phytoestrogens and UL. Odds \nratio (OR) and 95% confidence interval (CI) were calcu -\nlated in the study. Last, we adopted three statistical mod -\nels: weighted quantile sum (WQS) regression, Bayesian \nkernel machine regression (BKMR), and quantile g-com -\nputation (qgcomp) models, to investigate the effects of \nsix mixed metabolites on UL.\nWeighted quantile sum (WQS) regression\nWQS regression was used to investigate the effects of \nsix mixed metabolites on UL and identify the predomi -\nnant metabolite. The study sample was randomly divided \ninto training dataset (30%, n = 474) and validation data -\nset (70%, n = 1,105). Exposure to each metabolite in the \ntraining dataset was first divided into tertiles. The tertiles \nwere then added together to generate an overall tertiles \nscore for each metabolite. An empirical weight for each \nmetabolite in the mixture was estimated using the boot -\nstrapping method [17]. The WQS score is a combination \nof six mixed metabolites, representing the whole-body \nburden of six urinary phytoestrogens [10]. The weight \nof each metabolite in the WQS score indicates the con -\ntribution of each metabolite to the overall result [18]. \nMetabolites with an estimated weight greater than 0.333 \n(1/3) were considered to be significant contributors to \nthe WQS score. Using 10,000 bootstrap samples from \nthe training dataset (30%), we calculated the weights \nfor WQS scores. Using the validation dataset (70%), we \nassess the statistical significance of WQS scores [19]. In \naddition, WQS regression requires that all exposure-\noutcome associations be focused in the same direction. \nTherefore, we estimated the positive and negative effects \nof the six metabolites on UL separately. R package gWQS \nwas adopted to perform the analysis.\nQuantile g‑computation (qgcomp) model\ngqcomp is a parameterized and generalized linear \nmodel based on application of g-computation, aimed to \nassess the effect of increasing all exposures in the mix -\nture by one quatile simultaneously [20]. In this study, \nthe gqcomp.noboot function was applied to estimate \nexposure effects, which divides six mixed metabolites \ninto tertiles, assigns a positive or negative weight to \neach metabolite. If a metabolite has multiple effects in \ndifferent directions, a positive or negative weight is inter -\npreted as the proportion of exposure effects that have a \nnegative (or positive) effect on UL, with a total weight \nof up to 2. The relationship of each metabolite endpoint \nand the mixed metabolites was assessed separately, and \nthe finding models were used to estimate the scaled effect \nsizes, variable-specific coefficients, and overall model fit \np-values. Metabolites with an estimated weight greater \nthan 0.05 were considered to be significant contributors \nto the gqcomp scores. R package qgcomp was adopted to \nperform the analysis.\nBayesian kernel machine regression (BKMR)\nBKMR is a supervised approach, which could identify \nnonlinear and nonadditive associations of exposure-out -\ncome [21]. In this study, the BKMR model with 10,000 \niterations was adopted. Genistein, equol and enterodiol \nwere divided into two groups according to their posi -\ntive correlation with UL, while daidzein, O-desmeth -\nylangolensin, and enterolactone were divided into one \ngroup according to their negative correlation with UL. \nThe combined effect was calculated by comparing mixed \nmetabolites at or above the 60th percentile with the 50th \npercentile. Group posterior inclusion probability (Group-\nPIP) and Conditional posterior inclusion probability \n(CondPIP) represent the probability of each group and \nmetabolite in each group included in the model, repre -\nsenting their contribution to the overall effect. R package \nbkmr was adopted to perform the analysis.\nResults\nPopulation characteristics\nTable 1 presents the general characteristics of 1,579 eli -\ngible participants. The average age was 37.81  years. \nApproximately 69.00% of participants reported a his -\ntory of drinking, and 32.14% of participants indicated \nthat they were menopausal. In addition, all participants \nwere divided into UL group (n = 204) and non-UL group \n(n = 1,375). Age, race/ethnicity, marital status, drink -\ning status, BMI, waist circumference, menopausal sta -\ntus, ovary removed status, use of female hormones, \nhormones/hormone modifiers, number of gravidities \nand total energy were significantly different between UL \ngroup and non-UL group (P < 0.05).\nCorrelation between single metabolites of urinary \nphytoestrogens and UL\nAs shown in Supplemental Table  3, the result of univari -\nate logistic regression indicated that age, race/ethnicity, \nmarital status, drinking status, BMI, waist circumference, \nmenopausal status, ovary removed status, use of female \nhormones, hormones/hormone modifiers and total energy \nmight be covariates for this current study. The weighted \n\nPage 5 of 11\nYang and Chen  BMC Women’s Health          (2023) 23:261 \n \nTable 1 The general characteristics of included participants\nVariables Total (n = 1579) UL group (n = 204) Non‑UL group (n = 1375) P\nAge, years, Mean (S.E) 37.81 (0.31) 44.54 (0.59) 36.68 (0.36)  < 0.001\nRace/ethnicity, n (%)  < 0.001\n Non-Hispanic White 745 (69.30) 90 (67.05) 655 (69.67)\n Non-Hispanic Black 314 (11.94) 76 (20.13) 238 (10.57)\n Other  racea 520 (18.76) 38 (12.82) 482 (19.75)\nMarital status, n (%) 0.028\n Married 913 (58.62) 124 (65.13) 789 (57.54)\n Never married 299 (17.72) 24 (9.98) 275 (19.01)\n  Other b 367 (23.66) 56 (24.90) 311 (23.45)\nEducation level, n (%) 0.783\n High school and below 362 (13.79) 33 (11.91) 329 (14.10)\n High school Grad/ GED or Equivalent 334 (22.32) 42 (22.70) 292 (22.26)\n Some College or AA degree/College Graduate or above 883 (63.89) 129 (65.40) 754 (63.64)\nPIR, Mean (S.E) 0.113\n < 1.0 314 (15.39) 23 (11.19) 291 (16.09)\n ≥ 1.0 1265 (84.61) 181 (88.81) 1084 (83.91)\nSmoking status, n (%) 0.129\n No 986 (56.77) 115 (50.34) 871 (57.85)\n Yes 593 (43.23) 89 (49.66) 504 (42.15)\nDrinking status, n (%) 0.045\n No 611 (31.00) 68 (24.70) 543 (32.06)\n Yes 968 (69.00) 136 (75.30) 832 (67.94)\nBMI, kg/m2, Mean (S.E) 28.27 (0.28) 29.63 (0.58) 28.04 (0.29) 0.007\nWaist circumference, cm, Mean (S.E) 92.67 (0.59) 95.54 (1.32) 92.20 (0.64) 0.020\nCotinine, ng/mL, Mean (S.E) 59.55 (4.30) 61.99 (10.57) 59.14 (4.71) 0.808\nAge at menarche, years, Mean (S.E) 12.55 (0.06) 12.40 (0.14) 12.58 (0.06) 0.240\nMenopausal status, n (%)  < 0.001\n No 1030 (67.86) 86 (42.13) 944 (72.16)\n Yes 549 (32.14) 118 (57.87) 431 (27.84)\nOvary removed status, n (%)  < 0.001\n No 1455 (90.64) 132 (62.53) 1323 (95.33)\n Yes 124 (9.36) 72 (37.47) 52 (4.67)\nHysterectomy, n (%)  < 0.001\n No 1432 (88.85) 118 (55.22) 1314 (94.47)\n Yes 147 (11.15) 86 (44.78) 61 (5.53)\nUse of female hormones, n (%)  < 0.001\n No 1270 (73.74) 128 (58.70) 1142 (76.26)\n Yes 309 (26.26) 76 (41.30) 233 (23.74)\nUse of other hormonal drugs, n (%) 0.035\n No 1479 (91.46) 180 (86.89) 1299 (92.22)\n Yes 100 (8.54) 24 (13.11) 76 (7.78)\nUse of non-steroidal anti-inflammatory drugs, n (%) 0.844\n No 1515 (95.06) 194 (94.74) 1321 (95.11)\n Yes 64 (4.94) 10 (5.26) 54 (4.89)\nPregnancy status, n (%) 0.086\n No 1297 (95.11) 193 (97.62) 1104 (94.69)\n Yes 282 (4.89) 11 (2.38) 271 (5.31)\nNumber of gravidities, n (%) 0.002\n 1 216 (13.19) 21 (10.12) 195 (13.70)\n > 1 1125 (67.32) 165 (79.12) 960 (65.34)\n Unknown 238 (19.50) 18 (10.76) 220 (20.96)\n\nPage 6 of 11Yang and Chen  BMC Women’s Health          (2023) 23:261 \nlogistic regression was used to assess the individual effect \nof each metabolite on UL (Table  2). After adjusting for \nage, race/ethnicity, marital status, drinking status, BMI, \nwaist circumference, menopausal status, ovary removed \nstatus, use of female hormones, hormones/hormone mod-\nifiers and total energy, equol in the tertile 3 showed sig -\nnificant association with UL (Model 1: OR = 1.92, 95%CI: \n1.07–3.43, P = 0.029). After further adjusting for age, race/\nethnicity, marital status, drinking status, BMI, waist cir -\ncumference, menopausal status, ovary removed status, use \nof female hormones, hormones/hormone modifiers, total \nenergy, daidzein, genistein, O-desmethylangolensin, enter-\nodiol, and enterolactone, the association of equol in the \ntertile 3 with UL remained significant (Model 2: OR = 1.92, \n95%CI: 1.09–3.38, P = 0.024; Fig. 2).\nWQS, qgcomp and BKMR models to assess the combined \nassociation between six metabolites and UL\nThe WQS model was employed to estimate the combined \neffect of six metabolites of urinary phytoestrogen on UL. \nIn the adjusted model (Table  3), mixed metabolites of \nurinary phytoestrogen had a positive association with UL \n(P = 0.011), and a tertile increase in the WQS index was \nrelated to a 68% increased risk of UL (95%CI: 1.12–2.51). \nWe also calculated the estimated chemical weights of for \neach WQS index (Fig.  3). The highest weighted chemical \nin the WQS model was equol, followed by enterodiol and \ngenistein.\nSimilar to the WQS model, a tertile increase in the \ngpcomp index was associated with risk of UL in the \nadjusted model (Table  4, OR = 1.51, 95%CI: 1.05–2.18, \nP = 0.027). Figure  4 shows the estimated weight of each \nmetabolite on the UL risk. Equol had the largest positive \nweight, followed by genistein and enterodiol, respectively.\nTable 1 (continued)\nVariables Total (n = 1579) UL group (n = 204) Non‑UL group (n = 1375) P\nDaidzein, ug/g, Mean (S.E) 307.31 (38.08) 472.64 (148.81) 279.70 (36.63) 0.214\nGenistein, ug/g, Mean (S.E) 141.77 (16.38) 164.75 (51.92) 137.93 (19.27) 0.655\nEquol, ug/g, Mean (S.E) 68.03 (20.65) 88.37 (33.50) 64.63 (23.77) 0.579\nO-desmethylangolensin, ug/g, Mean (S.E) 90.21 (10.03) 162.26 (48.08) 78.18 (10.23) 0.107\nEnterodiol, ug/g, Mean (S.E) 111.83 (13.33) 117.46 (22.60) 110.89 (14.79) 0.800\nEnterolactone, ug/g, Mean (S.E) 845.38 (84.86) 877.75 (119.89) 839.97 (92.97) 0.783\nTotal energy, kcal, Mean (S.E) 1920.02 (22.88) 1793.11 (55.60) 1941.21 (26.49) 0.024\nFiber, gm, Mean (S.E) 13.69 (0.22) 13.92 (0.76) 13.66 (0.24) 0.748\nGED General Equivalent Diploma, AA Associate of Arts, PIR poverty-to-income ratio, BMI body mass index, SE standard error, UL uterine leiomyomata\nOther  racea = Mexican American, other Hispanic and other race- Including Multi-Racial\nOtherb = widowed, divorced, separated and living with partner\nTable 2 The individual effect of each metabolite on UL by using \nweighted logistic regression\nUL uterine leiomyomata, Ref reference, OR odds ratio, CI confidence interval\nModel 1: adjusted age, race/ethnicity, marital status, drinking status, body mass \nindex, waist circumference, menopausal status, ovary removed status, use of \nfemale hormones, hormones/hormone modifiers and total energy\nModel 2: further adjusted for other metabolites of urinary phytoestrogen on the \nbasis of Model 2\nMetabolites \nof urinary \nphytoestrogen\nModel 1 Model 2\nOR (95% CI) P OR (95% CI) P\nDaidzein\n Tertile 1 Ref Ref\n Tertile 2 0.93 (0.52–1.68) 0.816 0.91 (0.46–1.80) 0.787\n Tertile 3 1.13 (0.69–1.85) 0.625 1.26 (0.57–2.77) 0.565\nGenistein\n Tertile 1 Ref Ref\n Tertile 2 1.24 (0.77–1.98) 0.369 1.21 (0.68–2.15) 0.507\n Tertile 3 1.15 (0.69–1.91) 0.594 1.00 (0.50–1.98) 0.989\nEquol\n Tertile 1 Ref Ref\n Tertile 2 1.17 (0.71–1.94) 0.533 1.19 (0.70–2.02) 0.506\n Tertile 3 1.92 (1.07–3.43) 0.029 1.92 (1.09–3.38) 0.024\nO-desmethylangolensin\n Tertile 1 Ref Ref\n Tertile 2 1.21 (0.67–2.18) 0.528 1.11 (0.61–2.00) 0.729\n Tertile 3 1.03 (0.63–1.69) 0.891 0.81 (0.47–1.40) 0.449\nEnterodiol\n Tertile 1 Ref Ref\n Tertile 2 0.78 (0.43–1.41) 0.407 0.75 (0.41–1.37) 0.347\n Tertile 3 1.18 (0.74–1.90) 0.476 1.07 (0.66–1.73) 0.772\nEnterolactone\n Tertile 1 Ref Ref\n Tertile 2 0.65 (0.37–1.15) 0.139 0.61 (0.34–1.11) 0.107\n Tertile 3 1.16 (0.68–2.00) 0.578 1.06 (0.59–1.89) 0.852\n\nPage 7 of 11\nYang and Chen  BMC Women’s Health          (2023) 23:261 \n \nSupplemental Table  4 summarizes the GroupPIP and \nCondPIP derived from the BKMR model for six metabo -\nlites. The GroupPIP of two group (genistein, equol and \nenterodiol; 0.34) was higher than one group (daidzein, \nO-desmethylangolensin, and enterolactone; 0.04). Entero-\ndiol (CondPIP = 0.89) contributed most to the model for \nthe UL risk. Figure  5 indicates the overall associations \nbetween six metabolites and UL risk. Although the high \nconcentrations of all metabolites were not statistically dif-\nferent compared to their 50th percentile, the overall effect \non UL of the mixture of exposures at the 60th and above \nquantiles showed an upward trend. As all other metabo -\nlites were at their median levels, equol and enterodiol have \npositive correlation on UL risk, while enterolactone has \nnegative correlation (Supplemental Fig. 1). In addition, we \nalso found that there may be an interaction between enter-\nodiol and enterolactone on UL risk (Supplemental Fig. 2).\nDiscussion\nIn this study including 1,579 US women, we assessed \nthe relationship of urinary phytoestrogens and UL risk \nby using a number of statistical models. Overall, the \nweighted multivariate logistic regression indicated a \ncorrelation between equol and UL risk. By the WQS \nand gpcomp models, we observed a positive association \nbetween mixed metabolites of urinary phytoestrogen and \nUL risk. WQS model further identified that equol made \nFig. 2 The association between single metabolite of urinary phytoestrogens and uterine leiomyomata in women in the multivariable logistic \nregression model. Other metabolites of urinary phytoestrogen were further adjusted for age, race/ethnicity, marital status, drinking status, body \nmass index, waist circumference, menopausal status, ovary removed status, use of female hormones, hormones/hormone modifiers and total \nenergy\nTable 3 WQS model to estimate association between six \nmetabolites and UL\nWQS weighted quantile sum, UL uterine leiomyomata, CI confidence interval, \nOR odds ratio, OR estimates represent the odds ratios of UL when the WQS \nindex was increased by one tertile; The positive and negative association was \nestimated respectively. Model was adjusted for age, race/ethnicity, marital \nstatus, drinking status, body mass index, waist circumference, menopausal \nstatus, ovary removed status, use of female hormones, hormones/hormone \nmodifiers and total energy\nOutcome OR (95% CI) P\nPositive weight\n UL 1.68 (1.12–2.51) 0.011\nNegative weight\n UL 1.14 (0.81–1.62) 0.448\n\nPage 8 of 11Yang and Chen  BMC Women’s Health          (2023) 23:261 \nthe most contribution in the association between metab -\nolite mixture of urinary phytoestrogen and UL risk. In \nthe BKMR model, there was no significant association \nbetween overall mixed metabolites and UL appeared, \nbut there was a trend towards an increase. Additionally, \nequol and enterodiol also showed a positive correlation \nwith UL risk in gpcomp and BKMR models.\nPrevious studies have focused on the relationship \nbetween individual chemicals and health outcomes, \nbut in fact, humans are often exposed to mixtures of \nmultiple pollutants/chemicals [19, 22]. In recent years, \nseveral novel statistical methods have been developed \nto assess the impact of exposure to chemical mixtures \non health outcomes, including WQS regression [17–\n19], gpcomp [20] and BKMR [21]. A review assessed \nthe relationship between exposure to mixtures of per- \nand polyfluoroalkyl substances and adverse health \noutcomes, and highlighted the importance of WQS \nand BKMR for assessment of the effects of exposure \nto mixtures [23]. In addition, a cross-sectional study \nperformed in US population found a positive associa -\ntion between combined exposures to mercury, arsenic, \ncadmium and lead measured in urine and higher esti -\nmated glomerular filtration rate using WQS regression \n[24], and they also indicated that there might be influ -\nence for exposure to multiple metals on kidney func -\ntion. In the study of Zhang Y, et al., they reported that \nmixed exposure of ten commonly exposed endocrine-\ndisrupting chemicals had a significant positive associa -\ntion with UL in WQS and BKMR models, the weight \ndistribution showed the highest weights for mercury \n(weight = 0.35) and equol (weight = 0.29) [10]. How -\never, to our knowledge, the association between the \nmixed metabolites of urinary phytoestrogen and UL \nhas not been studied so far.\nFig. 3 WQS model regression index weights for uterine leiomyomata. Model was adjusted for age, race/ethnicity, marital status, drinking status, \nbody mass index, waist circumference, menopausal status, ovary removed status, use of female hormones, hormones/hormone modifiers and total \nenergy\nTable 4 Qgcomp model to assess the combined association \nbetween six metabolites and UL\nCI confidence interval, OR odds ratio; Model was adjusted for age, race/\nethnicity, marital status, drinking status, body mass index, waist circumference, \nmenopausal status, ovary removed status, use of female hormones, hormones/\nhormone modifiers and total energy\nOutcome OR (95% CI) P\ng-computation index’s 1.51 (1.05–2.18) 0.027\n\nPage 9 of 11\nYang and Chen  BMC Women’s Health          (2023) 23:261 \n \nFig. 4 gqcomp model regression index weights of the mixture on uterine leiomyomata risk. Model was adjusted for age, race/ethnicity, marital \nstatus, drinking status, body mass index, waist circumference, menopausal status, ovary removed status, use of female hormones, hormones/\nhormone modifiers and total energy\nFig. 5 Combined effects of six metabolites of urinary phytoestrogens on uterine leiomyomata risk. Model was adjusted for age, race/ethnicity, \nmarital status, drinking status, body mass index, waist circumference, menopausal status, ovary removed status, use of female hormones, \nhormones/hormone modifiers and total energy\n\nPage 10 of 11Yang and Chen  BMC Women’s Health          (2023) 23:261 \nUnlike previous study [5 , 10, 23], this study consid -\nered the mixed effect of six metabolites of urinary \nphytoestrogen (daidzein, genistein, equol, O-desmeth -\nylangolensin, enterodiol, and enterolactone) on UL \nrisk by three approaches (WQS regression, qgcomp, \nand BKMR). These results also indicated that mixed \nmetabolites of urinary phytoestrogen were positively \nlinked to the UL risk, with the greatest effect being \nfrom equol. Equol was related to an increased risk of \nUL. Our results are also consistent with previous study \n[10]. Equol, a metabolite of soy isoflavone daidzein, has \nestrogenic and antioxidant activity [25]. Several stud -\nies have showed that equol has a beneficial impact on \nmetabolic diseases [26, 27]. But, estrogen-dependent \ndiseases such as UL, are likely to be exacerbated by the \nestrogenic effects of equol. As described in an animal \nstudy, equol may trigger uterine tissue hyperplasia by \nincreasing luminal epithelial cell height and myometrial \nand stromal thickness, which further lead to UL [28]. \nOur results agree with a previous study that estradiol \ncould stimulate growth of UL, and was considered to be \nassociated with increased risk of UL [29]. Although we \nfound a combined effects of mixed metabolites on UL \nrisk, the molecular mechanism related to the relation -\nship of phytoestrogen and UL remains unclear. Further \nexploration is needed regarding the potential mecha -\nnisms in the association.\nThe main strength of this study was the use of WQS \nregression, qgcomp, and BKMR, which allowed us to \nassess the mixed metabolites of urinary phytoestrogen \nand UL risk. Some limitations for this study should be \nconsidered. First of all, because of the design of this \ncross-sectional study, there was a limitation in the \ncausal relationship between urinary phytoestrogens \nand UL. Second, some possible confounders were lack -\ning in this NHANES database, such as family history \nof UL. We did not adjust for history of hysterectomy \nbecause they may be a consequence of the outcome \n[30]. Third, for participants in the NHANES database, a \nsingle spot urine sample was only collected for metabo -\nlites analysis. The concentrations of metabolites of phy -\ntoestrogens may vary over time. Fourth, we excluded \n4,587 women who were not measurement of urinary \nphytoestrogen concentrations. Urinary phytoestrogens \nwere tested in 1/3 of the participants aged 6 years and \nolder in the NHANES database. However, this study \nconsidered the weights in the analysis, so the bias was \nrelatively small. Prospective studies with large sam -\nple size are warranted to further analyze the relation -\nship of urinary phytoestrogens and UL, and the related \nmechanisms.\nConclusion\nIn summary, our results implied an association of \nequol and UL. Importantly, WQS regression, qgcomp, \nand BKMR models was adopted to analyze the com -\nbined effects of mixed metabolites on UL risk. A posi -\ntive association between the mixed metabolites of \nurinary phytoestrogen and UL was also identified, with \nthe greatest contribution from equol. This study pro -\nvides evidence that urinary phytoestrogen-metabolite \nmixture was closely related to the risk of female UL \nand further research is needed to explore the detailed \nmechanism.\nAbbreviations\nUL  Uterine leiomyomata\nNHANES  National Health and Nutrition Examination Survey\nNCHS  National Center for Health Statistics\nHPLC  High-performance liquid chromatography\nMS  Mass spectrometric\nBMI  Body mass index\nSE  Standard error\nOR  Odds ratio\nCI  Confidence interval\nWQS  Weighted quantile sum\nSupplementary Information\nThe online version contains supplementary material available at https:// doi. \norg/ 10. 1186/ s12905- 023- 02381-5.\nAdditional file 1: Supplemental Table 1. Distribution of urinary phytoes-\ntrogen levels. Supplemental Table 2. Sensitivity analysis of data before \nand after interpolation. Supplemental Table 3. The Selection of covari-\nates by univariate logistic regression. Supplemental Table 4. GroupPIP \nand CondPIP of six metabolites.\nAdditional file 2: Supplemental Fig. 1. Univariate exposure–response \nfunctionbetween metabolite exposure and UL with fixing all the other \nmetabolites at their median level. Model was adjusted for age, race/eth-\nnicity, marital status, drinking status, body mass index, waist circumfer-\nence, menopausal status, ovary removed status, use of female hormones, \nhormones/hormone modifiers and total energy.\nAdditional file 3: Supplemental Fig. 2. Bivariate exposure–response \nfunction for metabolites in UL, with exposure 1 metabolite at its 10%, 50%, \nand 90% levels and other metabolites fixed at their median levels. Model \nwas adjusted for age, race/ethnicity, marital status, drinking status, body \nmass index, waist circumference, menopausal status, ovary removed status, \nuse of female hormones, hormones/hormone modifiers and total energy.\nAcknowledgements\nNot applicable.\nAuthors’ contributions\nFY and YC designed the study. FY wrote the manuscript. FY and YC collected, \nanalyzed and interpreted the data. YC critically reviewed, edited and approved \nthe manuscript. All authors read and approved the final manuscript.\nFunding\nNot applicable.\nAvailability of data and materials\nThe datasets generated and/or analyzed during the current study are available \nin the NHANES database, https:// wwwn. cdc. gov/ nchs/ nhanes/.\n\nPage 11 of 11\nYang and Chen  BMC Women’s Health          (2023) 23:261 \n \n•\n \nfast, convenient online submission\n •\n  \nthorough peer review by experienced researchers in your field\n• \n \nrapid publication on acceptance\n• \n \nsupport for research data, including large and complex data types\n•\n  \ngold Open Access which fosters wider collaboration and increased citations \n \nmaximum visibility for your research: over 100M website views per year •\n  At BMC, research is always in progress.\nLearn more biomedcentral.com/submissions\nReady to submit y our researc hReady to submit y our researc h  ?  Choose BMC and benefit fr om: ?  Choose BMC and benefit fr om: \nDeclarations\nEthics approval and consent to participate\nThe requirement of ethical approval and informed consent of the subjects \nfor this was waived by the Institutional Review Board of The First Affiliated \nHospital of Soochow University, because the data was accessed from NHANES \n(a publicly available database). All methods were carried out in accordance \nwith relevant guidelines and regulations.\nConsent for publication\nNot applicable.\nCompeting interests\nThe authors declare no competing interests.\nReceived: 17 January 2023   Accepted: 20 April 2023\nReferences\n 1. Baird DD, Dunson DB, Hill MC, Cousins D, Schectman JM. High cumulative \nincidence of uterine leiomyoma in black and white women: ultrasound \nevidence. Am J Obstet Gynecol. 2003;188:100–7.\n 2. Brasky TM, Bethea TN, Wesselink AK, Wegienka GR, Baird DD, Wise LA. \nDietary fat intake and risk of uterine leiomyomata: a prospective ultra-\nsound study. Am J Epidemiol. 2020;189:1538–46.\n 3. Lewis TD, Malik M, Britten J, San Pablo AM, Catherino WH. A comprehen-\nsive review of the pharmacologic management of uterine leiomyoma. \nBiomed Res Int. 2018;2018:2414609.\n 4. McWilliams MM, Chennathukuzhi VM. recent advances in uterine fibroid \netiology. Semin Reprod Med. 2017;35:181–9.\n 5. Alset D, Pokudina IO, Butenko EV, Shkurat TP . The effect of estrogen-\nrelated genetic variants on the development of uterine leiomyoma: \nmeta-analysis. Reprod Sci. 2022;29:1921–9.\n 6. Atkinson C, Lampe JW, Scholes D, Chen C, Wähälä K, Schwartz SM. Lignan \nand isoflavone excretion in relation to uterine fibroids: a case-control \nstudy of young to middle-aged women in the United States. Am J Clin \nNutr. 2006;84:587–93.\n 7. Liu T, Li N, Yan YQ, Liu Y, Xiong K, Liu Y, et al. Recent advances in the \nanti-aging effects of phytoestrogens on collagen, water content, and \noxidative stress. Phytother Res. 2020;34:435–47.\n 8. Jackson MD, McFarlane-Anderson ND, Simon GA, Bennett FI, Walker SP . \nUrinary phytoestrogens and risk of prostate cancer in Jamaican men. \nCancer Causes Control. 2010;21:2249–57.\n 9. Simon GA, Fletcher HM, Golden K, McFarlane-Anderson ND. Urinary \nisoflavone and lignan phytoestrogen levels and risk of uterine fibroid in \nJamaican women. Maturitas. 2015;82:170–5.\n 10. Zhang Y, Lu Y, Ma H, Xu Q, Wu X. Combined Exposure to Multiple \nEndocrine Disruptors and Uterine Leiomyomata and Endometriosis in US \nWomen. Front Endocrinol (Lausanne). 2021;12:726876.\n 11. Chiu YH, Bellavia A, James-Todd T, Correia KF, Valeri L, Messerlian C, \net al. Evaluating effects of prenatal exposure to phthalate mixtures on \nbirth weight: a comparison of three statistical approaches. Environ Int. \n2018;113:231–9.\n 12 Karia PS, Joshu CE, Visvanathan K. Association of oophorectomy and fat \nand lean body mass: evidence from a population-based sample of U.S \nWomen. Cancer Epidemiol Biomarkers Prev. 2021;30:1424–32.\n 13 Karanth SD, Washington C, Cheng TD, Zhou D, Leeuwenburgh C, Braith-\nwaite D, et al. inflammation in relation to sarcopenia and sarcopenic \nobesity among older adults living with chronic comorbidities: results \nfrom the National Health and Nutrition Examination Survey 1999–2006. \nNutrients. 2021;13:3957.\n 14. Reger MK, Zollinger TW, Liu Z, Jones J, Zhang J. Association between \nurinary phytoestrogens and C-reactive protein in the continuous national \nhealth and nutrition examination survey. J Am Coll Nutr. 2017;36:434–41.\n 15. Xiong G, Huang C, Zou Y, Tao Z, Zou J, Huang J. Associations of urinary \nphytoestrogen concentrations with nonalcoholic fatty liver disease \namong adults. J Healthc Eng. 2022;2022:4912961.\n 16. Martínez Steele E, Monteiro CA. Association between dietary share of \nultra-processed foods and urinary concentrations of phytoestrogens in \nthe US. Nutrients. 2017;9(3):209.\n 17. Nguyen HD, Kim MS. Effects of heavy metals on cardiovascular diseases \nin pre and post-menopausal women: from big data to molecular mecha-\nnism involved. Environ Sci Pollut Res Int. 2022;29:77635–55.\n 18. Carrico C, Gennings C, Wheeler DC, Factor-Litvak P . Characterization of \nweighted quantile sum regression for highly correlated data in a risk \nanalysis setting. J Agric Biol Environ Stat. 2015;20:100–20.\n 19. Duc HN, Oh H, Kim MS. The effect of mixture of heavy metals on obesity \nin individuals ≥50 years of age. Biol Trace Elem Res. 2022;200:3554–71.\n 20. Keil AP , Buckley JP , O’Brien KM, Ferguson KK, Zhao S, White AJ. A quantile-\nbased g-computation approach to addressing the effects of exposure \nmixtures. Environ Health Perspect. 2020;128:47004.\n 21. Bobb JF, Claus Henn B, Valeri L, Coull BA. Statistical software for analyzing \nthe health effects of multiple concurrent exposures via Bayesian kernel \nmachine regression. Environ Health. 2018;17:67.\n 22. Yu L, Liu W, Wang X, Ye Z, Tan Q, Qiu W, Nie X, Li M, Wang B, Chen W. A \nreview of practical statistical methods used in epidemiological studies \nto estimate the health effects of multi-pollutant mixture. Environ Pollut. \n2022;306:119356.\n 23. Rosato I, ZareJeddi M, Ledda C, Gallo E, Fletcher T, Pitter G, Batzella E, \nCanova C. How to investigate human health effects related to exposure \nto mixtures of per- and polyfluoroalkyl substances: a systematic review of \nstatistical methods. Environ Res. 2022;205:112565.\n 24. Sanders AP , Mazzella MJ, Malin AJ, Hair GM, Busgang SA, Saland JM, et al. \nCombined exposure to lead, cadmium, mercury, and arsenic and kidney \nhealth in adolescents age 12–19 in NHANES 2009–2014. Environ Int. \n2019;131:104993.\n 25 Mayo B, Vázquez L, Flórez AB. Equol: a bacterial metabolite from the \ndaidzein isoflavone and its presumed beneficial health effects. Nutrients. \n2019;11:2231.\n 26. Liu J, Mi S, Du L, Li X, Li P , Jia K, et al. The associations between plasma \nphytoestrogens concentration and metabolic syndrome risks in Chinese \npopulation. PLoS One. 2018;13:e0194639.\n 27. Takahashi A, Kokubun M, Anzai Y, Kogre A, Ogata T, Imaizumi H, et al. \nAssociation between equol production and metabolic syndrome in \nJapanese women in their 50s–60s. Menopause. 2022;29:1196–9.\n 28. Brown NM, Lindley SL, Witte DP , Setchell KD. Impact of perinatal exposure \nto equol enantiomers on reproductive development in rodents. Reprod \nToxicol. 2011;32:33–42.\n 29. Nowak RA. Fibroids: pathophysiology and current medical treatment. \nBaillieres Best Pract Res Clin Obstet Gynaecol. 1999;13:223–38.\n 30 Marsh EE, Al-Hendy A, Kappus D, Galitsky A, Stewart EA, Kerolous M. \nBurden, prevalence, and treatment of uterine fibroids: a survey of U.S. \nwomen. J Womens Health (Larchmt). 2018;27:1359–67.\nPublisher’s Note\nSpringer Nature remains neutral with regard to jurisdictional claims in pub-\nlished maps and institutional affiliations.","source_license":"CC0","license_restricted":false}