Exposure to phytoestrogens in utero and age at menarche in a contemporary British cohort

Abstract

Phytoestrogens are estrogenic compounds that occur naturally in plants. Phytoestrogens can cross the placenta, and animal studies have found associations between in utero exposure to phytoestrogens and markers of early puberty. We investigated the association between in utero exposure to phytoestrogens and early menarche (defined as < 11.5 years of age at onset) using data from a nested case-control study within the Avon Longitudinal Study of Parents and Children, a longitudinal study involving families living in the South West of England. Concentrations of six phytoestrogens were measured in maternal urine samples collected during pregnancy. Logistic regression was used to explore associations between tertiles of phytoestrogen concentrations and menarche status, with adjustment for maternal age at menarche, maternal education, pre-pregnancy body mass index (BMI), child birth order, duration of breastfeeding, and gestational age at sample collection. Among 367 mother-daughter dyads, maternal median (interquartile range) creatinine-corrected concentrations (in μg/g creatinine) were: genistein 62.1 (27.1–160.9), daidzein 184.8 (88.8–383.7), equol 4.3 (2.8–9.0), O-desmethylangolensin (O-DMA) 13.0 (4.4–34.5), enterodiol 76.1 (39.1–135.8), and enterolactone 911.7 (448.1–1558.0). In analyses comparing those in the highest tertile relative to those in the lowest tertile of in utero phytoestrogen exposure, higher enterodiol levels were inversely associated with early menarche (odds ratio (OR) = 0.47; 95% confidence interval (CI): 0.26–0.83), while higher O-DMA levels were associated with early menarche (OR = 1.89; 95% CI: 1.04–3.42). These findings suggest that in utero exposure to phytoestrogens may be associated with earlier age at menarche, though the direction of association differs across phytoestrogens.

1. Introduction

Puberty is a crucial period of growth and development. The timing and patterning of pubertal events, such as age at menarche, can provide information on overall health and some previous exposures, while potentially forecasting future health outcomes (Christensen et al., 2011; Biro et al., 2001; Golub et al., 2008).

Age at menarche, on average, has decreased since the late 19th century (Wyshak and Frisch, 1982; Zacharias and Wurtman, 1969), and a secular trend towards earlier development of secondary sexual characteristics has been reported among girls in the Avon Longitudinal Study of Parents and Children (ALSPAC) based in the United Kingdom (Rubin et al., 2009). In the United States, recent estimates for average age at menarche (12.4 years) are almost a year younger than the average age at menarche of women born in the 1920s (13.3 years), and decreases in average age at menarche have been observed across all racial/ethnic groups (McDowell et al., 2007). While improvements in nutritional status since the 19th century and the increasing prevalence of childhood obesity may be in part responsible for this trend, exposure to endocrine disrupting chemicals (EDCs) may also lead to altered timing and patterning of pubertal development (Christensen et al., 2011; Herman-Giddens et al., 1997; Buck Louis et al., 2008; Blanck et al., 2000; Biro et al., 2012).

An EDC, as defined by the U.S. Environmental Protection Agency, is “an exogenous agent that interferes with synthesis, secretion, transport, metabolism, binding action, or elimination of natural blood-borne hormones that are present in the body and are responsible for homeostasis, reproduction, and developmental processes” (Diamanti-Kandarakis et al., 2009). EDCs can be natural or man-made, and research suggests that EDCs may have the greatest impact during prenatal and early postnatal development when organ and neural systems are forming (National Institutes of Health, 2015). Most EDCs have estrogenic and/or anti-androgenic actions (Daxenberger et al., 2001), which are thought to have puberty-inducing effects in females (Mouritsen et al., 2010). Previous studies have examined the associations of in utero exposure to various EDCs with pubertal development, particularly age at menarche, with some conflicting results (Christensen et al., 2011; Blanck et al., 2000; Vasiliu et al., 2004; Hatch et al., 2011). Most studies were limited by the use of retrospectively-collected age at menarche data.

One potential class of naturally-occurring EDCs of interest is phytoestrogens. Phytoestrogens are estrogenic compounds that occur naturally in plants; two important groups of phytoestrogens are isoflavones (commonly found in legumes like soybeans) and lignans (commonly found in flax seeds, cereal grains, and some fruits) (Kim and Park, 2012; Centers for Disease Control and Prevention, 2008). Although exposure to phytoestrogens is mostly dietary, phytoestrogens can cross the placental barrier in humans (Foster et al., 2002; Todaka et al., 2005).

Phytoestrogens can interact with estrogen receptors, ERα and ERβ, which mediate many downstream activities. The affinities of phytoestrogens for ERα and ERβ are relatively weak compared to estradiol, and the affinities within phytoestrogens for ERα and ERβ vary widely. Additionally, phytoestrogens can have agonist or antagonist activity depending on whether estradiol is also present (Shanle and Xu, 2011; Jefferson et al., 2012; Patisaul and Jefferson, 2010; Cederroth et al., 2012).

Animal studies have reported the effects of phytoestrogens to be quite different according to time, dosage, and route. In utero exposure to phytoestrogens are of particular interest because of the timing of differentiation and development (Casanova et al., 1999; Takashima-Sasaki et al., 2006; Takagi et al., 2004). Studies in rodents have found that isoflavones administered through diet or subcutaneous injection during gestation or early life can lead to early vaginal opening (akin to early menarche in humans), irregular estrous cyclicity, and decreased GnRH (gonadotropin-releasing hormone) activation (GnRH coordinates reproductive maturation and function) (Casanova et al., 1999; Takashima-Sasaki et al., 2006; Takagi et al., 2004; Kouki et al., 2003; Lewis et al., 2003; Bateman and Patisaul, 2008; Lee et al., 2009; Nagao et al., 2001). To date, few studies have examined the effects of lignan metabolites on pubertal development, particularly age at menarche (Kim and Park, 2012).

In humans, the effect of soy-based infant formula on pubertal development has been studied to some extent, though this has yielded mixed results regarding an association with age at menarche (Strom et al., 2001; D’Aloisio et al., 2013; Adgent et al., 2011). One retrospective cohort study found no association between soy-based formula and self-reported age at menarche (Strom et al., 2001), while a much larger cohort study that also relied on self-reported recall of age at menarche found that soy-based formula was associated with both very early (≤10 years) and late (≥15 years) menarche (D’Aloisio et al., 2013). Within the ALSPAC prospective cohort study (n = 2028), researchers found a 53% increased risk of early menarche among those fed soy-based formula, when compared to cows’ milk-based formula (Adgent et al., 2011); it should be noted that this paper examined the ALSPAC cohort as a whole, as opposed to the nested case-control study used in the present study.

There have been no human studies published to date that have investigated the association between in utero phytoestrogen exposure and age at menarche. Our aim was to do so, using maternal gestational levels of phytoestrogen exposure and prospectively-collected age at menarche data in a population-based nested case-control study.

2. Study design and methods

2.1. Study population

The Avon Longitudinal Study of Parents and Children (ALSPAC) is an ongoing prospective birth cohort of 14,541 pregnancies. ALSPAC enrolled pregnant women with an expected delivery date between April 1st, 1991 and December 31st, 1992 from three health districts in the former county of Avon, Great Britain. Information has been collected on these parents and children through interviews, mailed questionnaires, and clinic visits. Details on ALSPAC recruitment and study methods have been described elsewhere (Boyd et al., 2013). A nested case-control study was conducted within the ALSPAC cohort to explore associations of prenatal maternal concentrations of various EDCs and age at menarche among the daughters. A ‘Growing and Changing’ questionnaire was developed to collect information on the offspring’s pubertal development and distributed to participants annually between the ages of 8–17 years (1999–2008), with the exception of age 12 (2003). Menarche was determined through parental- or self-report of menarche status, and, if it had occurred, month and year of occurrence so that age could be computed. From the original base population of 14,062 live births, case and control series were selected from singleton (n = 11,820) female subjects (n = 5756) who had completed at least two puberty staging questionnaires between the ages of 8 and 13 (5 possible questionnaires returned; n = 3682). Girls meeting eligibility criteria were ordered according to reported age at menarche when the 13-year old data became available. A cut-off of 11.5 years was established as defining ‘early’ menarche to satisfy sample size and power needed for the case-control study. Eligible cases could complete any two questionnaires in the series, provided that one was completed after menarche, while controls had to complete the 13-year old questionnaire in order to ascertain that menarche had not occurred by the cutoff of 11.5 years. Of the girls who reported menarche before the age of 11.5 (n = 338), 59.8% (n = 202) had a prenatal maternal urine sample available, and were considered potential cases. Among girls who reported menarche at or after the age of 11.5, a random sample of 394 was chosen as potential controls, and of these, 61.2% (n = 241) had a maternal urine sample available. After evaluating the integrity of the maternal urine samples, 86.1% (n = 174) of potential case and 81.3% (n = 196) of potential control samples were analyzed. Two cases and one control were excluded due to missing creatinine concentrations, leaving a total sample size of 367 mother-daughter dyads.

2.2. Laboratory analysis

Maternal urine samples were stored at −20 °C before being transferred under controlled conditions to the National Center for Environmental Health, Centers for Disease Control and Prevention (Atlanta, GA) for analysis using high-performance liquid chromatography-tandem mass spectrometry. The analytical methods are described elsewhere (Rybak et al., 2008). Phytoestrogens (genistein, daidzein, equol, O-desmethylangolensin (O-DMA), enterodiol, and enterolactone) were measured in maternal first morning void urine samples collected at a median gestational age of 12 weeks (interquartile range 8–17 weeks). The isoflavones under study are genistein and daidzein, as well as the daidzein metabolites, equol and O-DMA. Also under study are the lignan metabolites, enterodiol and enterolactone, which are metabolites of secoisolariciresinol and matairesinol. Phytoestrogen concentrations were creatinine-corrected using the standardization plus covariate adjustment method of O’Brien et al. (2016). Maternal urine concentrations were used as a proxy for fetal exposure (Green and Marsit, 2015; Ahmed et al., 2011; Engel et al., 2006).

2.3. Statistical analysis

Potential confounders to be considered in the analyses were identified a priori based on previously published literature and biological plausibility. Covariates were collected at various time points. We considered the following as covariates: child race (white/non-white); maternal education (ordinally classified as O-level); maternal age at menarche (8–11 years /12–15 years); maternal pre-pregnancy self-reported body mass index (BMI) (kg/m²); prenatal vegetarian diet (yes/no); prenatal smoking (any/none); maternal age at delivery (years); child birth order (first born, second born, or third born or later); child birth weight (< 2500 g/≥2500 g); breastfeeding duration (ordinally classified as not breastfed, < 3 months, 3–5 months, ≥6 months); use of infant soy formula (any/none); vegetarian diet during childhood (yes/no); objectively measured childhood BMI Z-score at age 8 (if missing for age 8, used age 7, 9, or 10); and gestational age at sample collection (weeks).

All data analysis was performed using SAS 9.3 (Cary, NC). Descriptive statistics were calculated for the sample comprised of 367 mother-daughter dyads; chi square and Fisher’s exact tests were used to compare groups by menarche status. Medians and interquartile ranges were calculated for each phytoestrogen for the total sample and by menarche status, and the Wilcoxon rank sum test was used to compare groups by menarche status.

Prior to modeling, phytoestrogen concentrations were log transformed. Phytoestrogen concentrations were also divided into tertiles by using cut points based on the distribution among the controls. To investigate the association between maternal phytoestrogen concentration and earlier age at menarche, unconditional logistic regression models were used. Using the set of potential confounding variables selected a priori for consideration in multivariable regression models, the final model was achieved through hierarchical backwards elimination of insignificant variables (Kleinbaum et al., 1982). Maternal education, maternal age at menarche, maternal vegetarian diet, and childhood BMI were considered as potential effect modifiers. Given that 15% (n = 56) of observations were missing data for covariates included in the model, multiple imputation using the fully conditional specification method was performed to address missing covariate data (Liu and De, 2015).

Please note that the ALSPAC study website contains details of all the data that are available through a fully searchable data dictionary: http://www.bristol.ac.uk/alspac/researchers/access/ (University of Bristol, 2015). Ethical approval for the study was obtained from the ALSPAC Ethics and Law Committee and the Local Research Ethics Committees. The U.S. Centers for Disease Control and Prevention (CDC) Institutional Review Board assessed and approved human subjects protection. Mothers provided informed consent at time of enrollment.

3. Results

In the ALSPAC cohort, girls were predominantly born to white mothers, most of whom had ordinary levels of education or higher (Table 1). Cases were more likely to have mothers who had an earlier age at menarche; one-third of case mothers reported menarche between 8 and 11 years of age, compared to less than 14% of control mothers. Mothers of cases were more than twice as likely to have an overweight or obese pre-pregnancy BMI, and cases were more than twice as likely to have a childhood BMI that was more than one standard deviation above the mean. Cases were more likely to be the first born child (61.6% versus 51.6%) and more likely to have never been breastfed or breastfed for less than 3 months (48.5% versus 39.9%). The median age of menarche among cases was 11.0 years, while the median age of menarche among controls was almost two years later at 12.8 years.

Table 1.

Characteristics of the Avon Longitudinal Study of Parents and Children (ALSPAC) nested case-control study population (N = 367 mother-daughter dyads).

	Menarche < 11.5 years (N = 172)		Menarche ≥ 11.5 years (N = 195)		P-value for differencea
Characteristic	N	%	N	%
Child race					0.61
White	159	95.8	182	96.8
Non-white	7	4.2	6	3.2
Maternal educationc					0.39
< O-level	36	21.7	32	16.8
O-level	56	33.7	62	32.5
> O-level	74	44.6	97	50.8
Maternal age at menarche, years					< 0.0001
8–11	50	33.3	23	13.9
≥12	100	66.7	142	86.1
Maternal pre-pregnancy BMI, kg/m2					0.0004
< 25 (under/normal weight)	113	72.4	159	87.8
≥25 (overweight/obese)	43	27.6	22	12.2
Prenatal vegetarian diet					0.92
Yes	10	6.1	11	5.9
No	154	93.9	177	94.1
Prenatal smoking					0.47
Any	29	17.3	27	14.4
None	139	82.7	160	85.6
Maternal age at delivery, years					0.63
< 25	32	18.7	35	17.9
25–29	71	41.5	73	37.4
≥30	68	39.8	87	44.6
Child birth order					0.03
First born	101	61.6	97	51.6
Second born	36	22.0	66	35.1
Third born or later	27	16.5	25	13.3
Child birth weight, gd					1.00b
< 2500	<5	–	<5	–
≥2500	166	–	188	–
Breastfeeding duration, months					0.04
Not breastfed	26	16.0	38	20.8
<3	53	32.5	35	19.1
3–5	26	16.0	34	18.6
≥6	58	35.6	76	41.5
Use of infant soy formulad					0.43b
Any	<5	–	<5	–
None	163	–	187	–
Vegetarian diet during childhood					0.65
Yes	8	4.8	7	3.8
No	158	95.2	176	96.2
Childhood BMI Z-score					< 0.0001
<0	32	20.9	60	34.5
0–1	50	32.7	77	44.3
≥1	71	46.4	37	21.3
Gestational age at sample, weeks	Median	IQR	Median	IQR	0.04e
	12.0	7.5–17.0	13.0	9.0–18.0
Age at menarche, years
	11.0	10.7–11.3	12.8	12.3–13.4

Abbreviations: g, grams; kg/m², kilograms per meter-squared; IQR, interquartile range.

^aCompared using chi-square tests unless otherwise noted.

^bCompared using Fisher’s exact test.

^c< O-level = none, Certificate of Secondary Education, and vocational education, which are equivalent to no diploma or a GED in the United States. O-levels (ordinary levels) are required and completed at the age of 16. > O-level=A-levels (advanced levels) completed at 18, which are optional, but required to get into university; and a university degree.

^dCounts and percents suppressed due to small cell sizes.

^ep-value for difference between cases and controls using the Wilcoxon Rank Sum Test.

Median enterodiol concentrations were almost 15 μg/g creatinine lower among cases than among controls, while median O-DMA concentrations were more than 2 μg/g creatinine higher among cases than among controls (Table 2). There were no other differences evident. For the most part, phytoestrogen concentrations were moderately correlated (Supplementary Table 1).

Table 2.

Gestational urinary phytoestrogen concentrations among mothers of girls with and without earlier age at menarche in the Avon Longitudinal Study of Parents and Children (ALSPAC) nested case-control study population (N = 367 mother-daughter dyads).

	Total		Menarche <11.5 years (N = 172)		Menarche ≥11.5 years (N = 195)		p-valued
Analytea,b,c	Median	IQR	Median	IQR	Median	IQR
Genistein	62.1	27.1–160.9	60.2	26.6–157.6	65.1	27.2–162.7	0.58
Daidzein	184.8	88.8–383.7	187.8	91.9–369.7	183.0	88.8–423.8	0.79
Equol	4.3	2.8–9.0	4.4	2.8–7.4	4.2	2.6–9.6	0.74
O-DMA	13.0	4.4–34.5	14.0	5.7–43.8	11.7	3.4–31.5	0.06
Enterodiol	76.1	39.1–135.8	69.6	33.6–123.6	84.3	49.4–144.4	0.02
Enterolactone	911.7	448.1–1558.0	901.1	494.8–1478.2	923.8	440.6–1650.9	0.89

Abbreviations: IQR, interquartile range; O-DMA, O-Desmethylangolensin.

^aCreatinine-corrected concentrations in μg/g creatinine.

^bThere are missing concentrations for genistein (n = 1 where menarche <11.5 years), O-DMA (n = 2 where menarche <11.5 years), and equol (n = 1 where menarche <11.5 years and n = 1 where menarche ≥11.5 years).

^cThe limits of detection (LOD) were 0.3 ng/mL (genistein, daidzein, equol, enterodiol, enterolactone) and 0.2 ng/mL (O-desmethylangolensin). There were no samples below the LOD for daidzein and enterodiol, one sample below the LOD for enterolactone, two samples below the LOD for genistein, 16 samples below the LOD for equol, and 30 samples below the LOD for O-DMA; the number of samples below the LOD did not differ by case-control status. The LOD/√2 was substituted for values below the LOD.

^dp-value for difference between cases and controls using the Wilcoxon Rank Sum Test.

In the unadjusted model, decreased odds of early menarche were observed for those in the second and third tertiles of in utero enterodiol exposure compared to those in the lowest tertile (odds ratio (OR)_second = 0.60; 95% confidence interval (CI)_second: 0.36–0.98; p-value_second = 0.04; OR_third = 0.58; 95% CI_third: 0.35–0.96; p-value_third = 0.03; p-trend: 0.03) (Table 3). Increased odds of early menarche were observed for those in the second and third tertiles of in utero O-DMA exposure compared to those in the lowest tertile (OR_second = 1.91; 95% CI_second: 1.12–3.27; p-value_second: 0.02; OR_third = 1.97; 95% CI_third: 1.16–3.35; p-value_third: 0.02; p-trend: 0.02).

Table 3.

Associations of maternal urinary phytoestrogen concentrations with earlier age at menarche in the Avon Longitudinal Study of Parents and Children (ALSPAC) nested case-control study population (N = 367 mother-daughter dyads).

	Unadjusteda			Adjustedab
Analytec	OR (95% CI)	p valuef	p for trendf	OR (95% CI)	p valuef	p for trendf
Genistein
Continuous	0.96 (0.83–1.11)		0.55	0.93 (0.78–1.10)		0.38
Tertile 2 (38.86–118.38)	1.02 (0.62–1.67)	0.94		0.88 (0.50–1.55)	0.66
Tertile 3 (118.38–17916.60)	0.87 (0.52–1.44)	0.58	0.59	0.82 (0.46–1.47)	0.51	0.41
Daidzein
Continuous	1.03 (0.87–1.22)		0.70	1.02 (0.84–1.24)		0.83
Tertile 2 (110.09–319.03)	1.37 (0.83–2.25)	0.21		1.20 (0.69–2.08)	0.52
Tertile 3 (319.03–21880.45)	1.00 (0.59–1.68)	0.99	0.99	0.83 (0.46–1.50)	0.53	0.58
Equol
Continuous	0.97 (0.83–1.13)		0.70	0.98 (0.83–1.16)		0.80
Tertile 2 (3.19–6.93)	1.22 (0.74–1.99)	0.44		1.04 (0.59–1.83)	0.89
Tertile 3 (6.93–9005.85)	0.86 (0.51–1.44)	0.55	0.57	0.88 (0.48–1.62)	0.68	0.63
O-DMA
Continuous	1.12 (1.00–1.26)		0.05	1.14 (0.99–1.30)		0.06
Tertile 2 (4.77–21.04)	1.91 (1.12–3.27)	0.02		1.51 (0.83–2.73)	0.18
Tertile 3 (21.04–1631.58)	1.97 (1.16–3.35)	0.02	0.02	1.89 (1.04–3.42)	0.04	0.03
Enterodiol
Continuous	0.79 (0.64–0.98)		0.03	0.74 (0.58–0.94)		0.01
Tertile 2 (56.49–117.99)	0.60 (0.36–0.98)	0.04		0.45 (0.26–0.80)	0.01
Tertile 3 (117.99–1188.20)	0.58 (0.35–0.96)	0.03	0.03	0.47 (0.26–0.83)	0.01	0.01
Enterolactone
Continuousd	1.04 (0.87–1.26)		0.65	1.21 (0.96–1.52)		0.10
Tertile 2e (551.49–1314.96)	1.39 (0.85–2.28)	0.19		1.75 (0.99–3.09)	0.05
Tertile 3e (1314.96–8718.04)	0.98 (0.58–1.65)	0.93	0.94	1.56 (0.84–2.90)	0.16	0.13

^aUnconditional logistic regression.

^bAdjusted for maternal age at menarche, maternal education, pre-pregnancy BMI, child birth order, duration of breastfeeding, and gestational age at sample collection.

^cCreatinine-corrected concentrations in μg/g creatinine.

^dContinuous represents natural log transformed values of phytoestrogen concentration.

^eTertiles represent the comparison of the higher tertiles, tertiles 2 or 3, to the lowest tertile of phytoestrogen concentration.

^fThe p-value is for the comparison of tertile 2 or 3 to the lowest tertile of phytoestrogen concentration; the p for trend is for the trend across all three tertiles.

The results of the analyses adjusting for maternal age at menarche, maternal education, pre-pregnancy BMI, child birth order, duration of breastfeeding, and gestational age at sample collection were similar to those from the unadjusted analyses (Table 3). In the adjusted model, the odds of early menarche for each unit increase of logged mothers’ enterodiol concentration were OR = 0.74 (95% CI: 0.58–0.94; p-trend: 0.01). When enterodiol was treated categorically, decreased odds of early menarche were observed for those in the second and third tertiles of in utero enterodiol exposure compared to those in the lowest tertile (OR_second = 0.45; 95% CI_second: 0.26–0.80; p-value_second = 0.01; OR_third = 0.47; 95% CI_third: 0.26–0.83; p-value_third = 0.01; p-trend: 0.01). The odds of early menarche for each unit increase of logged mothers’ O-DMA concentration were OR = 1.14 (95% CI: 0.99–1.30; p-trend: 0.06). When comparing those in the third tertile of O-DMA concentration to those in the lowest tertile, an 89% increase in the odds of early menarche was observed (OR_second = 1.51; 95% CI_second: 0.83– 2.73; p-value_second = 0.18; OR_third = 1.89; 95% CI_third: 1.04–3.42; p-value_third = 0.04; p-trend: 0.03). No other significant associations were observed between phytoestrogen concentration and early menarche. There was no evidence of effect modification by maternal education, maternal age at menarche, maternal vegetarian diet, or childhood BMI.

4. Discussion

In this study, we observed strong associations between enterodiol and decreased odds of earlier age at menarche, and some evidence of an association between O-DMA and increased odds of earlier age at menarche.

While enterodiol was associated with decreased odds of earlier age at menarche, the other lignan metabolite under study, enterolactone, was nonsignificantly associated with increased odds of earlier age at menarche. The parent compounds of the lignan metabolites under study, secoisolariciresinol and matairesinol, are not estrogenic by themselves, but are readily converted to enterodiol and enterolactone, which are estrogenic (Setchell et al., 1981; Glitsø et al., 2000; Cornwell et al., 2004). A recent cohort study (n = 1051) found that enterolactone measured in girls before puberty (6–8 years) was associated with later menarche, a result that conflicts with the enterolactone findings in this study, although there are important differences in the timing of exposure assessment (Wolff et al., 2017). It is also possible that the findings in the present study were due to chance, as sample size is small. While in vitro studies have suggested that lignans can exhibit biphasic effects (estrogenic and antiestrogenic effects) dependent on exposure level (Tang et al., 2015; Mousavi and Adlercreutz, 1992; Mueller et al., 2004; Pettersson and Gustafsson, 2001; Waters and Knowler, 1982; Welshons et al., 1987; Adlercreutz, 2007; Wang, 2002), it is difficult to extrapolate in vitro studies to humans, and animal studies of phytoestrogens and pubertal development have focused almost exclusively on isoflavones, excluding lignans (Kim and Park, 2012). Given the scarcity and conflicting results of the few existing studies, additional research is needed in animal models and human populations to better understand the effects of lignans on puberty.

O-DMA is an intestinal bacterial metabolite of daidzein, and about 90% of individuals harbor bacteria capable of metabolizing daidzein to O-DMA (Frankenfeld, 2011). O-DMA is less structurally similar to 17β-estradiol than its parent compound and therefore may exhibit different biological actions than daidzein. Of note, O-DMA is the least estrogenic of the isoflavones and isoflavone metabolites under study; genistein is more estrogenic than daidzein, and daidzein and equol are both more estrogenic than O-DMA (Hopert et al., 1998; Schmitt et al., 2001). However, the underlying bacteria that metabolize daidzein to O-DMA may have a distinct physiological role; urinary excretion of O-DMA is a marker of harboring intestinal bacteria capable of C-ring cleavage, and therefore it is suspected that the role of the phenotype may extend beyond daidzein metabolism (Frankenfeld, 2011).

To our knowledge, this is the first published study of associations of in utero phytoestrogen exposure with age at menarche. Although there were few significant associations found between phytoestrogen levels and age at menarche, there is biological plausibility for such an association. Exposures during pregnancy are extremely relevant to pubertal development, since this represents the period of initial organ development, including the brain, endocrine system, and reproductive tract. Furthermore, the fetus is more susceptible to such exposures due to smaller size, lack of a complete blood-brain barrier, and absence of metabolizing enzymes (Todaka et al., 2005). Studies have found that phytoestrogens can cross the placental barrier in humans, and one study of Californian women undergoing amniocentesis found that 96% of second trimester amniotic fluid samples contained quantifiable amounts of dietary phytoestrogens (Foster et al., 2002). Additionally, as previously described, studies in rodents have found that exposure to high doses of isoflavones in utero and through diet in early life accelerated pubertal onset in female animals (Casanova et al., 1999; Takashima-Sasaki et al., 2006; Takagi et al., 2004; Kouki et al., 2003; Lewis et al., 2003; Bateman and Patisaul, 2008; Lee et al., 2009; Nagao et al., 2001).

To our knowledge, no previous studies have investigated in utero phytoestrogen exposure; therefore, we looked to previous studies on the effect of in utero exposure to other potential EDCs, which produced mixed results, as have previous studies on early life phytoestrogen exposure to soy infant formula. A cohort study (n = 151) assessing in utero exposure to polychlorinated biphenyls (PCBs) and dichlorodiphenyldichloroethlene (DDE) with age at menarche found that increased exposure to DDE was associated with an earlier age at menarche, while exposure to PCBs was not associated (Vasiliu et al., 2004). A nested case-control study (n = 448) found no association with in utero exposure to polyfluoroalkyl chemicals (PFCs) and age at menarche (Christensen et al., 2011). Prior studies of the association between soy-based infant formula and age at menarche are also inconclusive, with one study finding no association, another study finding an association with early menarche (12 years), and a third study finding an association with very early (≤10 years) and late menarche (≥15 years) (Strom et al., 2001; D’Aloisio et al., 2013; Adgent et al., 2011). While it is difficult to compare across classes of EDCs and at different times of exposure, earlier studies have not suggested a clear association between in utero and early life exposure to EDCs and age at menarche.

Strengths of this study are the inclusion of multiple phytoestrogen biomarkers, substantial covariate data available on mothers and children from multiple time points over gestation and childhood, and outcome data generally collected in the year that the outcome occurred. Limitations of this study include a single spot urine measurement of phytoestrogen exposure which was not collected at a uniform time of gestation, missing information on age at menarche among some controls, incomplete information on some covariates, and no assessment of phytoestrogen exposure or early childhood soy consumption (excluding soy formula) in the daughters, which might influence timing of menarche. Additionally, few studies have validated the use of maternal urinary phytoestrogen concentrations as a proxy for fetal exposure to phytoestrogens, though a 2006 study by Engel et al. found that the relative concentrations of phytoestrogens measured in amniotic fluid were identical to those in urine (Engel et al., 2006). Unlike some other EDCs that are estimated to have half-lives on the order of several years, peak rates of urinary excretion of phytoestrogens occur between 6 and 12 h after ingestion (King and Bursill, 1998). Since phytoestrogens are excreted rather quickly, phytoestrogen exposure assessed through urinary excretion at a single time point may under- or overestimate intermittent phytoestrogen exposure. Age at menarche was obtained through self-report on ‘Growing and Changing’ puberty questionnaires completed yearly by parents and/or children, depending on age. There is some potential for misclassification of the outcome, such as the completion of the questionnaire by a parent unaware of the child’s menarche status, or issues in the parent or child’s recall of the month and year menstruation began. Although the absence of a ‘Growing and Changing’ questionnaire at age 12 could diminish the accuracy of recall on the questionnaire completed at age 13, previous studies in middle-aged women have suggested moderate to high recall ability of age at menarche, and it would be expected that teenage girls would have even better recall of age at menarche (Cairns et al., 2011; Cooper et al., 2006; Must et al., 2002).

It is also possible that the girls selected for this nested case-control study were not representative of the base cohort. When comparing ALSPAC girls who returned at least two ‘Growing and Changing’ questionnaires to those who did not return any questionnaires, non-respondents’ parents were more likely to have lower educational attainment. Compared to non-respondents, mothers of respondents were generally older at time of index birth. Further, non-respondents were more likely to be of non-white race. This could have affected our findings since socioeconomic status is related to age at menarche (Braithwaite et al., 2009); however, whether socioeconomic status is related to phytoestrogen concentrations is unclear. Although race is related to age at menarche (Biro et al., 2001, 2006; McDowell et al., 2007; Wu et al., 2002; Freedman et al., 2002; Britton et al., 2004) and was associated with maternal lignan concentrations in this study, we were not able to examine the effect of race due to the small number of non-white girls enrolled in ALSPAC. Furthermore, it has been suggested that several genes in Caucasians code for early menarche (Dvornyk and Waqar-ul-Haq, 2012), and since we did not include genetic data in this study, we do not know if our results could be affected. Last, due to a modest sample size, this study may have been underpowered to detect additional associations between in utero phytoestrogen exposure and age at menarche. Future studies should consider that additional EDCs could contribute to the earlier age at menarche observed among cases; it could also be worthwhile to examine whether a mixture of EDCs are associated with earlier age at menarche.

Urinary phytoestrogen concentrations during 1991–1992 among mothers of girls participating in the ALSPAC cohort were roughly two to three times higher for all phytoestrogens except equol (which was half as high) when compared to 1999–2002 National Health and Nutrition Examination Survey (NHANES) data for white women between 20 and 39 years old (Centers for Disease Control and Prevention, 2008) (data not shown). It should be noted though that these samples were taken almost a decade apart. Compared to phytoestrogen concentrations among other studies published before 2005, phytoestrogen concentrations observed in ALSPAC were notably lower than phytoestrogen concentrations observed in Asian populations and in studies where participants had a special diet (e.g. soy-rich diet) (Valentín-Blasini et al., 2005).

5. Conclusions

In summary, we compared exposure to phytoestrogens during pregnancy among mothers of girls who did and did not have earlier age at menarche in the ALSPAC cohort. We found an association between O-DMA and increased odds of earlier age at menarche, while decreased odds of earlier age at menarche were observed for enterodiol. As demonstrated in our study by the conflicting effects of phytoestrogens, plus the general lack of human studies on the associations between phytoestrogens and pubertal outcomes, there is a need for additional studies to explore these associations in a variety of populations and to describe potential mechanisms of action for O-DMA and enterodiol.

Appendix A. Supplementary material

Supplementary data associated with this article can be found in the online version at doi:10.1016/j.envres.2017.02.030.