Steroids, stress, and body mass index interact to accelerate female pubertal development

Abstract

Context

Estrogens underlie puberty in girls, but the steroid metabolome may also regulate pubertal timing in response to elevated stress and increasing body mass index (BMI).

Objective

Our objective was to identify steroid metabolome patterns linked to accelerated puberty and test whether BMI and stress markers modify this relationship.

Methods

From the LEGACY Girls Study, a longitudinal cohort followed for 6 years, we selected 327 girls aged 5 to 13 years at baseline to measure 36 steroid metabolites of glucocorticoids, androgens, progesterone, and estrogens in 2 urine samples collected before and during puberty. Parents reported the age at onset of breast development (thelarche), which had a high correlation in the subset with clinically assessed Tanner. Study staff measured height and weight and administered questionnaires, including the Internalizing Composite Scale, a parental proxy of child stress. We estimated hazard ratios (HRs) for the association between doubled steroid metabolites and ages at thelarche, pubarche, and menarche using Weibull survival models, testing interactions with stress and BMI z-scores.

Results

Accelerated thelarche was associated with higher prepubertal urinary metabolites of glucocorticoids (HR = 1.9, 95% CI: 1.5-2.5), androgens (HR = 3.9, 95% CI: 2.7-5.6), and progesterone (HR = 6.7, 95% CI: 4.1-10.9). Girls with high glucocorticoid metabolites combined with high BMI and stress reached thelarche 7 months earlier than their counterparts with low measures in these parameters.

Conclusion

Elevated metabolites of glucocorticoids, androgens, and progesterone are associated with accelerated pubertal onset, and BMI and stress modify this association. Previous studies have focused on estrogens, menarche, and BMI; our results suggest that androgens and stress impact the timing of thelarche as well, explaining secular declines in pubertal timing.

Age at menarche is a critical health indicator. Socially, it marks the beginning of the transition to adulthood (1). Biologically, however, menarche is one of the last changes to occur during puberty. One of the first signs of puberty is the onset of breast development, or thelarche, which typically occurs 2 to 4 years before menarche (2). The pubertal window between thelarche and menarche, measured as pubertal tempo (2), is also an important marker of health. For example, earlier ages at thelarche and menarche and a longer tempo have each been associated with 20% to 30% increased risk for breast cancer (3, 4).

A series of underlying hormonal changes occur before and during puberty. While estrogens underlie puberty in girls, other metabolites composing the steroid metabolome also may regulate pubertal timing in response to elevated stress and increasing body mass index (BMI). Biologically weak androgens, dehydroepiandriosterone (DHEA) and its sulfate (DHEA-S) (5, 6), are the first steroids to rise between ages 6 and 8 years, during adrenarche, a maturational process of the zona reticularis in the adrenal gland (7). The adrenals are part of the hypothalamic–pituitary–adrenal (HPA) axis (8, 9). This axis can be activated by stressors, which within a stress reaction trigger the release of glucocorticoids in the zona fasciculata and DHEA in the zona reticularis. The relationship between androgens and glucocorticoids, generally expressed as a ratio, reflects stress reactivity; a higher ratio of androgens to glucocorticoids reflects a hyporeactive stress response, and a lower ratio reflects the hyperreactive response (10). During the pubertal window, adrenal androgens that begin with adrenarche continue to rise together with the ovarian production of androgens, estrogens, and progesterone, reflecting an active “reproductive axis” or hypothalamic–pituitary–ovarian (HPO) axis, responsible for breast development, pubic hair growth (pubarche), and menarche.

Increased BMI (11, 12) and psychosocial stress (2, 13, 14) are associated with earlier puberty, suggesting cross-talk between metabolic, stress, and reproductive processes (15). Despite decades of research demonstrating that stress and BMI are predictors of pubertal onset (16), to date, most studies have not examined these factors simultaneously. There is convincing evidence from experimental and rodent studies that stress and reproductive axes interact during the pubertal window. The pubertal window involves heightened neural plasticity and rapid maturation of neurobehavioral systems, resulting in the recalibration of the HPA axis during puberty (17, 18). After puberty is complete, stress reactivity plasticity declines (19).

Our objective was to identify steroid metabolome patterns linked to accelerated puberty and then test whether BMI and stress markers modify this relationship. We tested the hypothesis that stress during childhood elevates glucocorticoids and androgens, and then adipose tissue converts androgens into estrogens, which promote breast development. We examined this relationship among girls with and without a family history of breast cancer. We predicted girls with elevated BMI and stress to have the earliest puberty. We further conjectured that the stress response changes during the pubertal window.

Materials and Methods

Population

The LEGACY Girls Study is a cohort of 1040 girls, aged 6 to 13 years at baseline, recruited across 5 study sites (New York, Philadelphia, San Francisco Bay Area, Salt Lake City, and Toronto, Canada) (20). Within the cohort, 51% have a breast cancer family history, and 49% are of similar age and race and ethnicity but without breast cancer family history (20). LEGACY followed participants in person every 6 months, collecting questionnaire data from mothers/guardians and girls aged ≥10 years, pubertal assessments, biospecimens, and anthropometric measurements. Institutional Review Board approval was obtained by participating institutions at each site: New York: Columbia University Irving Medical Center; Toronto: Mount Sinai Hospital Research Ethics Board; Philadelphia: Fox Chase Cancer Center, the University of Pennsylvania, and The Children's Hospital of Philadelphia; San Francisco Bay Area: Cancer Prevention Institute of California and Stanford University; and Salt Lake City: Huntsman Cancer Institute. Written informed consent was collected from mothers/guardians, and assent was collected from girls according to institutional standards.

This analysis includes participants who were prepubertal at baseline and had given first-morning urine samples at baseline and at subsequent visits after they had reached puberty. The participants in this study were from the New York, Philadelphia, Toronto, and Salt Lake City study sites.

Pubertal outcomes

Parents assessed puberty through the Pubertal Development Scale (PDS) (21) and the Sexual Maturation Scale (SMS) (22) of breast development (thelarche) and pubic hair growth (pubarche). In New York and Utah, trained research staff performed standardized clinical breast Tanner staging with palpation (23). We previously generated the kappas from 282 participants from the same LEGACY cohort and found maternal reports of breast onset using the PDS to be highly reliable (κ = 0.8) and valid (sensitivity = 86.6%; specificity = 89.6%) in the subset of girls who also had clinical breast Tanner stage. The accuracy and kappas were very similar by BMI status (BMI < 85th percentile: κ = 0.8, accuracy = 89%; BMI ≥ 85th percentile: κ = 0.7, accuracy = 85%). Given its high accuracy with clinical Tanner, we used PDS as our primary form of assessment (24). We defined thelarche and pubarche as the equivalent of Tanner Stage 2 in the PDS. We conducted sensitivity analyses using SMS (n = 316) and a clinical Tanner (n = 144) to assess thelarche. Where applicable, mothers/guardians reported their daughter's age at menarche in half-year intervals.

Steroid metabolites

Participants provided first-morning urine samples at baseline (98% participation rate). We measured the steroid metabolome in prepubertal urine from 327 participants who provided samples at least 12 months before the onset of breast development. We selected an additional urine sample, collected after onset of breast development but before menarche (n = 115). The Steroid Research and Mass Spectrometry Unit of the Center of Child and Adolescent Medicine at Justus Liebig University in Germany analyzed urinary metabolites using gas chromatography–mass spectrometry (GC–MS) (25-27). The GC–MS urinary steroid hormone metabolome assay measured 36 steroid metabolites, including 14 glucocorticoids (including 4 corticosterone metabolites), 10 androgens, 9 intermediates of progesterone (referred to as progesterone metabolites), and 3 estrogens. The list of individual steroids, their abbreviations, and limits of detection can be found in Table S1 (28). Using samples from this same cohort, we have previously shown that urinary steroid metabolites are not all highly correlated with serum concentrations and measure different metabolic processes than what is in circulation (29). For concentrations below the limit of detection, we assigned half the limit of detection. We log₂ transformed each concentration to improve normality and then divided by creatinine. The same laboratory measured creatinine using a Dimension Vista® analyzer (Siemens Healthineers, Eschborn, Germany). For all urinary steroids measured, interassay precision varied between 10% and 24% (26). We examined the metabolites individually and summed them into groupings of androgens (A), estrogens (E), progesterone (P), and glucocorticoids (G), and the total (T) sum of all metabolites, as others have previously done (30). Due to the correlation between steroids, we also created metabolite ratios (A:E and A:G), so that we could investigate the associations of each summed grouping independent of androgens, especially in the case of estrogens (30).

Stress

Mothers/guardians completed the Internalizing Composite Scale, which includes subscales for anxiety, depression, and somatization, with established criteria for clinical at-risk status (t ≥ 60) (31). We used the internalizing composite scale score as in our previous work (32, 33).

Body mass index

Trained research staff measured height and weight twice every 6 months using a stadiometer and a digital scale. We used the averages to calculate BMI, defined as weight (kilogram) divided by the square of height (m²), and calculated the z-scores for age using the 2000 US Centers for Disease Control and Prevention growth charts (34).

Cancer family history

Mothers/guardians provided information on girls' family history of all cancers (1st, 2nd, and 3rd degree) using a pedigree data collection instrument at baseline. From these pedigree data, we calculated a continuous risk score using the Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm (BOADICEA) (35).

Covariates

At baseline, mothers/guardians self-reported information on pregnancy and infancy, including birth weight and their child's race and ethnicity.

Statistical analyses

We estimated whether prepubertal individuals, metabolic groupings, and ratios of metabolites were associated with the age of each pubertal timing outcome—thelarche, pubarche, and menarche—and the duration of pubertal tempo—thelarche to menarche. We used parametric survival (Weibull) models, which account for left censoring, to predict median age of pubertal outcomes and hazard ratios (HRs) and 95% CIs for early puberty. When assessing the association between metabolites and pubertal timing, a higher HR indicates an earlier age at onset.

We built our models, including a priori confounders that temporally precede the exposure of interest (eg, birth weight, race and ethnicity, and study site). We also included an interaction term between the hormone and the age at specimen collection to account for differences in time between hormone measurement and pubertal onset. We tested for a 3-way interaction between hormone, BMI z-score, and stress by including a cross-product term for all 3 continuous variables. We also stratified models comprised of 4 groups: low BMI/low stress, low BMI/high stress, high BMI/low stress, and high BMI/high stress, cutting the continuous variables at the study sample medians (BMI = 15.6 kg/m²; stress = 46). We also formally tested for interaction by BOADICEA risk score.

To test for the association between pubertal hormone levels (collected after thelarche and before menarche) and menarche, we created 4 nested models. Model 1 was the age-adjusted association between the pubertal hormone and menarche, including an interaction with age at collection. Model 2 additionally adjusted for race and ethnicity, study site, and birth weight. Model 3 additionally adjusted for the prepubertal hormone level. Model 4 additionally adjusted for age at breast development onset (calculated as the midpoint between the ages at B1 and B2). We tested for 3-way interactions among pubertal hormone levels, stress, and BMI by adding a cross-product term to Model 3 and graphed the stratified results.

We conducted principal component (PC) analysis on the prepubertal steroid metabolite data and used 12 of these PCs (determined using a scree plot to predict >90% of the variance) as predictors in the Weibull models (36). We also tested for a 3-way interaction of each PC with stress and BMI.

Results

Sample characteristics

Table 1 describes the study sample by key variables of interest for this analysis. The mean age of the study sample was 8.1 years. Most (74%) of the study sample self-reported their race as non-Hispanic White, were not overweight or obese (<85th percentile), and had levels of stress below the clinical threshold (<60). Maternal education was high, with over half the study sample having bachelor's or graduate degrees.

Table 1.

Characteristics of participants in the LEGACY Girls Study (n = 327)

Sample characteristics	Mean (SD) or % (n)	Range
Age, year	8.1 (1.4)	5.2-12.9
Race and ethnicity
White	74 (242)
Black	6.5 (21)
Hispanic	11 (36)
Asian	7.5 (25)
Other	1 (3)
Study site
Philadelphia	18 (60)
New York	27 (87)
Utah	26 (85)
Ontario	29 (95)
Birth weight (g)	3303 (571)	964-4536
BMI (kg/m2)	15.6 (2.1)	11.1-31.0
BMI (z-score)	−0.31 (1.15)	−4.88 to 2.36
Stress score (%)	45.3 (28.7)	1-99
Maternal education
Some college/vocational/technical school or less	22 (73)
Bachelor's degree	46 (149)
Graduate degree	32 (105)

Prepubertal hormones with pubertal outcomes

Total and specific metabolic pathways of prepubertal steroid metabolites were associated with the age at onset of breast development in crude and adjusted models (Fig. 1, Table S1) (28). Participants with a doubling in urinary levels of total (HR = 2.0, 95% CI: 1.5-2.6), glucocorticoid (HR = 1.9, 95% CI: 1.5-2.5), androgen (HR = 3.9, 95% CI: 2.7-5.6) and progesterone (HR = 6.7, 95% CI: 4.1-10.9) metabolites had younger ages at onset of breast development than participants with half the levels, whereas participants with a doubling of urinary estrogens had older ages at onset of breast development (HR = 0.3, 95% CI: 0.2-0.4) than those with half the levels. Patterns of associations between steroid metabolites and age at onset of pubarche were similar to thelarche. Participants with a higher ratio of androgens to glucocorticoids had thelarche (HR = 2.3, 95% CI: 1.8-2.9) and pubarche (HR = 2.2, 95% CI: 1.7-2.9) earlier than those with a higher ratio of glucocorticoids to androgens (Table S2) (28). Only prepubertal estrogens were associated with menarche. Participants with a doubling of estrogens reached menarche at older ages than those with half the levels (HR = 0.2, 95% CI: 0.1-0.5). Results were consistent using SMS and clinical exam to assess the pubertal outcomes (data not shown).

Figure 1.

Hazard rate ratios and 95% CIs of the timing of puberty given a doubling in prepubertal hormone (μg/mL per 10 mg of creatinine), adjusting for race and ethnicity, birth weight, and study site.

Pubertal hormones, menarche, and pubertal tempo

Participants with a doubling of pubertal levels of estrogens experienced menarche later than those with half the levels in models adjusting for both covariates and prepubertal estrogen levels (HR = 0.4, 95% CI: 0.2-0.9; Table 2). Participants with a doubling in pubertal levels of androgen (HR = 0.3, 95% CI: 0.1-0.8) and progesterone (HR = 0.2, 95% CI: 0.1-0.7) metabolites also experienced menarche at older ages when holding the age at thelarche constant, meaning they had a longer pubertal tempo than those with half the levels. Participants with a higher ratio of androgens to glucocorticoids had significantly earlier age at menarche (HR = 1.9, 95% CI: 1.5-3.4) than those with a higher ratio of glucocorticoids over androgens, and the association was stronger when accounting for prepubertal levels (HR = 2.2, 95% CI: 1.2-4.1).

Table 2.

Hazards rate ratios and 95% CIs of timing of menarche per doubling of peripubertal hormone (μg/mL per 10 mg of creatinine)

	Model 1	Model 2	Model 3	Model 4
Total	1.13 (0.85-1.49)	1.19 (0.83-1.7)	1.14 (0.81-1.62)	0.67 (0.42-1.06)
Glucocorticoids	1.16 (0.84-1.61)	1.24 (0.81-1.9)	1.21 (0.81-1.82)	0.67 (0.4-1.12)
Androgens	1.09 (0.75-1.59)	1.14 (0.69-1.88)	0.91 (0.55-1.51)	0.32 (0.13-0.76)
Progestogens	1.24 (0.67-2.3)	1.32 (0.57-3.02)	1.17 (0.5-2.75)	0.23 (0.07-0.73)
Estrogens	0.47 (0.22-1.01)	0.39 (0.17-0.9)	0.38 (0.16-0.92)	1.02 (0.9-1.15)
A:E	0.63 (0.52-0.76)	0.6 (0.46-0.78)	0.56 (0.42-0.74)	0.83 (0.55-1.27)
A:G	1.71 (0.96-3.02)	1.88 (1.05-3.39)	2.19 (1.17-4.1)	0.71 (0.33-1.54)

Model 1 is crude.

Model 2 is adjusted for race and ethnicity, study site, and birth weight.

Model 3: Model 2 + prepubertal hormone level.

Model 4: Model 3 + age at thelarche.

The values in bold are statistically significant.

Interactions among hormones, BMI, and stress

There were interactions between stress, BMI, and prepubertal and pubertal hormone levels for thelarche and menarche, but not pubarche. Girls with high measures of prepubertal glucocorticoids, BMI, and stress reached thelarche 7.2 months earlier than girls with low measures of prepubertal glucocorticoids, BMI, and stress (Fig. 2A). The association of higher prepubertal glucocorticoids with earlier menarche was strongest in participants with high BMI and high stress (Fig. 2B); in contrast, higher glucocorticoids were associated with later menarche in those with high BMI and low stress (Fig. 2B). The association of higher prepubertal progesterone metabolites and early menarche was strongest in those with low BMI and low stress (Fig. 2C); progesterone metabolites had a weaker association with age at menarche in those with high BMI and low stress. The association of higher pubertal estrogens with later menarche was strongest among those with low BMI and low stress (Fig. 2D). Higher pubertal androgen and progesterone metabolites were associated with earlier menarche in those with high BMI and low stress (Fig. 2E and 2F).

Figure 2.

Predicted ages at thelarche and menarche according to pre- and peri-pubertal hormone level, stratified by high and low BMI and Stress. Panels A and B show pre-pubertal glucocorticoids, panel C shows pre-pubertal progesterone, and panels D-F show peri-pubertal estrogens (D), androgens (E), and progesterone (F).

There were no interactions among hormones and breast cancer family history for any of the outcomes (data not shown).

Principal components

The results from the PC analysis in Table 3 confirm the results derived from our analysis based on metabolic pathways. There were 12 primary PCs (Fig. 3) that explained 95% of the variation in steroid metabolites, and 4 of these were associated with age at thelarche. Participants with higher PC1 scores had high glucocorticoid metabolites and other metabolites that were the most abundant in urine. They began thelarche at younger ages (HR = 1.2, 95% CI: 1.0-1.5) and had a longer tempo (HR = 0.45, 95% CI: 0.22-0.91). Participants with high PC2 scores had high levels of androgen and progesterone metabolites and low levels of glucocorticoid metabolites, and they also had thelarche at younger ages (HR = 2.0, 95% CI: 1.3-3.2) and a longer tempo (HR = 0.28, 95% CI: 0.12-0.65) than participants with low PC2 scores. Participants with higher PC2 scores also had pubarche earlier than those with lower scores. Participants with higher PC8 scores, which reflected high 17a-hydroxypregnanolone, pregnenediol, and 11-oxo-androsterone, had thelarche at earlier ages than those with lower PC8 scores. Participants with high PC11 scores had extremely high levels of pregnanediol, and these participants reached thelarche (HR = 0.1, 95% CI: 0.01-0.27), pubarche (HR = 0.1, 95% CI: 0.03-0.29) and menarche (HR = 0.004, 95% CI: 0.0004-0.05) at older ages and had a longer tempo than those with lower scores (HR = 0.02, 95% CI: 0.001-0.6). There were significant 3-way interactions with 2 PCs (3 and 8) with stress and BMI.

Table 3.

Hazard rate ratios (HR) and 95% CI of early thelarche according to 12 steroid hormone PCs, adjusted for race and ethnicity, study site, and birth weight

	Thelarche		Pubarche		Menarche		Tempo
Principal component	HR	95% CI	HR	95% CI	HR	95% CI	HR	95% CI
PC1	1.22	(0.99-1.49)	1.01	(0.84-1.21)	0.58	(0.29-1.16)	0.45	(0.22-0.91)
PC2	2.00	(1.27-3.17)	3.37	(2.06-5.49)	0.96	(0.32-2.91)	0.28	(0.12-0.65)
PC3	1.31	(0.88-1.96)	1.05	(0.70-1.56)	1.21	(0.28-5.27)	0.50	(0.09-2.97)
PC4	1.23	(0.53-2.85)	1.10	(0.49-2.47)	0.27	(0.03-2.64)	0.40	(0.06-2.68)
PC5	1.69	(0.73-3.88)	1.04	(0.45-2.37)	2.17	(0.09-53.83)	0.47	(0.02-10.26)
PC6	1.29	(0.47-3.48)	0.76	(0.28-2.03)	0.25	(0.02-3.33)	0.52	(0.03-8.73)
PC7	1.56	(0.48-5.00)	0.49	(0.18-1.30)	0.38	(0.05-2.75)	0.14	(0.03-0.62)
PC8	3.50	(1.43-8.54)	1.52	(0.65-3.56)	0.72	(0.05-10.28)	0.05	(0.00-1.27)
PC9	0.84	(0.21-3.43)	1.28	(0.36-4.48)	1.31	(0.02-100.05)	0.70	(0.01-62.23)
PC10	0.40	(0.11-1.40)	0.68	(0.18-2.50)	1.43	(0.06-34.76)	3.30	(0.05-227.54)
PC11	0.10	(0.04-0.27)	0.10	(0.03-0.29)	0.004	(0.0004-0.05)	0.02	(0.00-0.60)
PC12	1.44	(0.30-7.03)	0.94	(0.20-4.41)	0.46	(0.004-55.02)	3.67	(0.12-116.40)

Adjusted for race and ethnicity, study site, and birth weight. The values in bold are statistically significant.

Figure 3.

Distribution of steroid pathways for 12 PCs.

Discussion

Higher levels of steroid metabolites were associated with the onset and duration of puberty, but these relationships differed by metabolic groupings, ratios, and PCs. Specifically, elevated glucocorticoid, androgen, and progesterone metabolites were associated with accelerated onset, and elevated androgens and progesterone metabolites were associated with lengthened duration. In contrast, estrogen metabolites were associated with delayed thelarche and menarche but had no association with duration. The association between higher levels of metabolites and timing of thelarche and menarche depended on BMI and stress status of participants. The earliest ages of thelarche and menarche were in those with high levels of glucocorticoids, high BMI, and high stress. Pubertal window also mattered. The association of progesterone metabolites with menarche differed most between those with low vs high BMI in the prepubertal window, whereas the association differed most between those with high vs low stress during the pubertal window.

This longitudinal cohort of girls is the first to examine the metabolism of all steroids in relation to pubertal development. Other puberty cohort studies have examined individual hormones longitudinally (37-45), but none is enriched with girls with a breast cancer family history or follow >250 girls throughout the duration of the pubertal window. The 3 studies that followed participants semiannually found that changes in hormones occur before pubertal characteristics appear (37) and elevated prepubertal serum androgens are associated with accelerated breast development (39) and menarche (38) independent of body size. In the longitudinal DONALD study, urinary estrogens were associated with earlier breast development and menarche and shorter duration, but these results were greatly attenuated after adjusting for androgen metabolites, suggesting that the effect of estrogens might be attributed partly to the conversion from adrenal androgens (46). None of these studies examined stress.

Our study is consistent with another study that found that participants with overall higher serum estrogen and estradiol have delayed thelarche (29). However, they found that participants with PCs with relatively higher serum levels of DHEA-S had later thelarche, and those with higher levels of estradiol had earlier menarche. The differences in findings compared to our study may be due to the timing and type of hormone measures. For example, they measured hormones 6 months before, at the time of, and after the onset of breast development and included only 4 hormones measured in serum. The PCs in our study included multiple metabolites of estrogens, androgens, and progesterone, which led to the identification of the ratio of progesterone to androgen metabolites driving earlier pubertal onset, as well as high concentrations of specific progesterone metabolites, such as pregnanediol, being associated with delayed onset of all pubertal milestones. There are also differences between serum and urine levels of steroids. In urine, steroid hormone metabolites are excreted primarily in conjugated form, either as sulfates or as glucuronides. In our method the total amount of each steroid is determined after hydrolysis. We and others have found weak to moderate positive correlations between urinary and serum steroids, suggesting that urine measures hormone production that is neither the inverse of circulating steroids nor positively correlated with them (9, 47, 48). In children and adolescents specifically, there is evidence that urinary androgens are more reliable predictors of adrenarche than circulating plasma androgens (8, 9). Another difference is that our PCs included glucocorticoids, which may also explain inconsistencies but also generated new evidence about the stress response and pubertal development.

The relationship between stress reactivity hormones and pubertal timing differed between prepubertal and pubertal periods, supporting hypotheses of stress programming during puberty. Higher levels of androgens and glucocorticoid metabolites were associated with earlier thelarche but not with menarche. Participants with a hyporeactive response to stress, which is higher androgens relative to glucocorticoids, before puberty and during puberty, had an earlier onset but longer duration of puberty. Principal components also capturing a hyporeactive stress response were associated with earlier thelarche. Our study confirms similar results derived from experimental studies in children and in animals that suggest that there are changes in activity of both the HPA and HPO axes as well as stress activity during puberty (49). The DONALD study found that higher totals of glucocorticoid metabolites, adjusting for androgens, were associated with delayed breast development and menarche in girls (n = 55) (50). That study, using 24-hour urinary measurements, is in line with our results since it captured the hyperreactive response by adjusting glucocorticoids for androgens. We captured the inverse stress response, stress hyporeactivity, and found it associated with earlier puberty, so the studies are consistent with each other.

Strengths and limitations

A major strength of our study is that we integrate siloed bodies of scientific investigation of stress and BMI as drivers of pubertal timing into 1 comprehensive study. Another strength is that, in contrast to most studies that only examine individual steroids (usually in serum) (51, 52), we examined a comprehensive panel of the steroid metabolome in first-morning urine. Our study also had limitations. There was measurement error of the hormones given the inter- and intraassay coefficients of variation (CVs). We batched samples to minimize differential misclassification, but the strength of the association may be underestimated. Although we used the most comprehensive assay of the steroid metabolome available, it does not measure all metabolites, including many metabolites of estrogen. However, the metabolites included in the different hormone categories in this study are a representative mixture of important metabolites of each parent hormone. Lastly, while our sample includes multiple races and ethnicities, the majority of participants are White, limiting the generalizability of results.

Conclusion

Elevated glucocorticoids and androgens were associated with accelerated pubertal onset and lengthened pubertal window, especially in those with high BMI and stress. These findings have clear implications for pediatric clinical practice as well as public health strategies to prevent breast cancer. In terms of pediatric practice, precocious puberty is defined as breast development or pubic hair growth before age 8 in girls and may be due to underlying endocrine pathology; however, screening for early puberty after age 8 but before 10 may also indicate girls at risk for subsequent menstrual and breast health risks. Stress-reducing interventions may be effective interventions to offer during this time. The LEGACY cohort, which is enriched with participants that have a family history of breast cancer, allows the opportunity to assess whether pubertal hormone patterns may be harbingers of breast cancer risk. We did not observe an interaction with breast cancer family history, but this means that, irrespective of breast cancer family history, similar mechanisms drive pubertal development. This is important for public health messaging because girls with and without a family history of breast cancer can benefit from childhood lifestyle interventions, such as physical activity, that modify BMI, stress, or hormones (53). Androgens and glucocorticoids are associated with earlier pubertal timing, and we previously showed in a nested case-control study of breast cancer, using the same urinary assay, that both these metabolite groupings are associated with up to 2.6-fold higher odds of breast cancer (30). Therefore, these hormones and the phenotypes captured by the steroid metabolome PCs may serve as biomarkers of potential breast cancer risk. Without any current population-based breast cancer screening for adults aged <40 years and with rates of early-onset breast cancer rising more than in other age groups (54, 55), there is great potential for longitudinal hormonal biomarker tracking to be a screening modality.

Abbreviations

BMI

body mass index

BOADICEA

Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm

DHEA

dehydroepiandriosterone

DHEA-S

sulfate

GC–MS

gas chromatography–mass spectrometry

HPA axis

hypothalamic–pituitary–adrenal axis

HPO axis

hypothalamic–pituitary–ovarian axis

HR

hazard ratio

PC

principal component

PDS

Pubertal Development Scale

SMS

Sexual Maturation Scale

Funding

This research was funded by the following National Cancer Institute grants: 5K07CA21816603 to L.C.H., 5R01CA15986804 to M.B.T., 5R01CA138819 to A.R.B. and M.B.D., 5R01CA138638 to E.M.J., and 5R01CA138844 to I.L.A. as well as from the Breast Cancer Research Foundation, which supported the follow-up of the cohort. The Biomarkers Core at Columbia University was also supported by P30CA013696 and P30ES009089.

Disclosures

The authors declare no conflict of interest.

Data availability

Restrictions apply to the availability of some or all data generated or analyzed during this study to preserve patient confidentiality or because they were used under license. The corresponding author will on request detail the restrictions and any conditions under which access to some data may be provided.