Association of Prepubertal and Adolescent Androgen Concentrations With Timing of Breast Development and Family History of Breast Cancer

This cohort study evaluates the association of prepubertal androgen concentrations with timing of breast development in girls with or without a family history of breast cancer.

Meaning If replicated in larger studies, these results may support the finding that elevated androgen concentrations in adolescent girls whose mother has breast cancer may be another factor contributing to the familial clustering of breast cancer.

Introduction
A large body of epidemiologic research supports the role of estrogen in postmenopausal breast cancer 1 ; however, the role of androgens (metabolic precursors to estrogens) in breast cancer risk has received less attention. Higher concentrations of prediagnostic androgens are associated with both premenopausal and postmenopausal breast cancer. 2,3 In a nested case-control study of 1375 postmenopausal women, testosterone concentrations were associated with a 55% increased risk of postmenopausal breast cancer among women in the top quartile vs bottom quartile of concentrations. 2 In the large European Prospective Investigation Into Cancer and Nutrition study, androgens have been associated with a 30% to 55% increase in premenopausal breast cancer risk. 3 During prepubescence, androgen concentrations begin to rise between ages 6 to 8 years, and a study 4 reported that girls with elevated androgen concentrations had earlier onset of breast development. We recently reported a 50% higher rate of earlier breast development in girls with a breast cancer family history (BCFH) compared with girls without a family history (no BCFH). 5 Early breast development, independent of age at menarche, increases the risk of breast cancer by 20%. 6 If and when androgen trajectories can be modified is unknown. In one study, 7 androgen concentrations in adolescent girls with late menarche were either above or below the range of androgen concentrations in girls with early and typical age at menarche. This finding suggests that varying factors in girls with late menarche influence adolescent androgen production. First, these girls may have had early thelarche. Early thelarche is associated with obesity 8 ; thus, childhood body mass index trajectories may be implicated. Stress reactivity may also play a role. [9][10][11] Stress reactivity is sensitive to modulation during puberty 12,13 ; therefore, differences between hyperreactivity and hyporeactivity may explain the wide range in androgen hormone concentrations in girls. Given that stressful conditions have been shown to alter pubertal timing 10 -and specifically at the family level, that a father's absence accelerates pubertal timing 14 -breast cancer in mothers may also pose a similar stressful childhood environment.
We hypothesized that the association between elevated adult androgen concentrations and increased breast cancer risk is programmed during puberty and modified by a BCFH, which represents both shared environment and shared underlying genetic susceptibility. Thus, we examined whether prepubertal concentrations of androgens were associated with breast development in girls with and without a BCFH and whether BCFH was associated with androgen concentrations. If androgen concentrations are different in girls with and without a BCFH, this finding may support the hypothesis that additional factors contribute to familial clustering of breast cancer.

Methods
Our reporting adheres to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cohort studies. Using the New York site of the Lessons in Epidemiology and Genetics of Adult Cancer From Youth (LEGACY) Girls Study, a prospective cohort, we examined girls aged 6 to 13 years at baseline and monitored them longitudinally for pubertal development. 15 We established the cohort in 2011 with continued follow-up. We conducted the present analysis in 2017 using data collected between August 16, 2011, and March 24, 2016. Every 6 months, girls and their mothers had follow-up visits involving anthropometric measurements (height and weight to calculate body mass index) and pubertal development assessment. Mothers reported the race/ethnicity of their daughters. We collected blood samples annually. We assessed breast development using the Pubertal Development Scale; scores range from 1 to 4, and scores of 2 or higher indicate breast development. We obtained institutional review board approval from Columbia University, New York, New York, to conduct the study. Mothers or guardians provided written informed consent, and girls provided assent based on institutional standards.
We measured 3 androgen concentrations in serum at the Reproductive Endocrine Research Laboratory at the Keck School of Medicine, Los Angeles, California, by using radioimmunoassays with preceding organic solvent extraction and celite column partition chromatography 16 to measure testosterone and androstenedione concentrations. Dehydroepiandrosterone sulfate (DHEA-S) and sex hormone-binding globulin were measured by direct chemiluminescent immunoassay. We calculated the free testosterone concentration using a validated algorithm 17 that requires total testosterone and sex hormone-binding globulin concentrations. The within-batch coefficients of variation ranged from 6% to 13%, and the between-batch coefficients of variation ranged from 5% to 7%. All androgen concentrations were log-transformed to improve normality. The precursor to the other androgens, DHEA-S, was correlated with androstenedione (r = 0.65), total testosterone (r = 0.62), and free testosterone (r = 0.71).
We classified girls as BCFH-positive if the participating mother or guardian reported a BCFH in the daughter's first-or second-degree relatives. First-degree BCFH is having a mother with breast cancer, and second-degree BCFH is having a grandmother or aunt with breast cancer. In addition to the categorical BCFH variable, we used the Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm (BOADICEA) risk model to estimate lifetime breast cancer risk for each girl 18 by estimating the absolute risk based on family pedigree information. The model produces a continuous score ranging from 0 to 100 for lifetime and fixed intervals of risk. We used the scores to rank the girls according to their risk of breast cancer and to provide greater precision. The scores cannot be interpreted as absolute risks because BOADICEA has been validated only for use in adult women. 19 We evaluated breast cancer-specific distress at the baseline visits and at follow-up visits 1, 3, and 5 using the 8-item Child Impact of Events Scale, a developmentally appropriate version of the Revised Impact of Event Scale. [20][21][22] Both scales evaluate intrusion and avoidance as indices of breast cancer-specific distress. 23,24 We summed the scores from the intrusion and avoidance indices to create a total distress score. Potential scores for the total Child Impact of Events Scale can range from 0 to 40, and a higher score indicates higher breast cancer-specific distress. When we examined the association of breast cancer-specific distress and the age of breast development, we collapsed the individual visit scores into 1 binary variable (yes vs no). When we examined the association between androgen concentrations by BCFH, we used the score at the same study visit as when we collected the serum sample; if that score was missing, we used the score from the preceding visit.

Statistical Analysis
We conducted 2 primary analyses, which included an examination of whether prepubertal androgens were associated with earlier breast development (analysis 1) and a comparison of serum androgen concentrations in girls with and without a BCFH (analysis 2).
For analysis 1, we used parametric Weibull survival models to estimate the median (95% CI) age at onset of breast development in girls with higher (above the median) vs lower (below the median) androgen concentrations measured at the same study visit as pubertal assessment for the crosssectional analysis (n = 92) and before any breast development for the prospective analysis (n = 36).
We adjusted our main model (model 1) for an interaction between androgen concentration and age compared with lower androgen concentrations (Figure 1). Positive associations between androstenedione, total testosterone, and free testosterone concentrations and early age at thelarche were evident after adjustment for family breast cancer risk as measured by the BOADICEA ( Table 2).

JAMA Network Open | Oncology
Family history score by this algorithm was also positively associated with earlier age at thelarche (model 2 in Table 2). Breast cancer-specific distress did not attenuate the association between androgen concentrations and the timing of breast development (model 3 in Table 2). We also calculated the medians (95% CIs) from the larger cross-sectional cohort (n = 92), in whom we assessed androgen concentrations and pubertal development stage concurrently. Breast Median (95% CI) ages were generated from an interval-censored Weibull regression model using a dichotomous variable based on the median concentration for each androgen and adjusted for race/ethnicity, body mass index, and an interaction term between androgen concentration and age at sample collection. Curves and medians pictured result from the prospective analysis on the subcohort of 36 girls in whom the androgen concentration was measured before any breast development.  Compared with no BCFH, a first-degree, but not second-degree, BCFH was associated with 240% higher geometric mean concentrations of androstenedione (no BCFH, 0.49 ng/mL vs firstdegree BCFH, 1.8 ng/mL vs second-degree BCFH, 1.6 ng/mL; P = .01), 10% higher concentrations of total testosterone (12.7 ng/dL vs 14.0 ng/dL vs 13.7 ng/dL; P = .01), and 92% higher concentrations of free testosterone (1.3 pg/mL vs 2.5 pg/mL vs 0.3 pg/mL; P = .14) (Figure 2). Percentage differences were calculated from the above means using the following formula:
The DHEA-S concentration was higher in girls with breast cancer-specific distress, both unadjusted and adjusted for BCFH (models 2 and 3 in Table 3). This result was specific to DHEA-S and not the other androgens. (To convert androstenedione to nanomoles per liter, multiply by 0.0349; total testosterone to nanomoles per liter, multiply by 0.0347; free testosterone to picomoles per liter, multiply by 3.47; and DHEA-S to micromoles per liter, multiply by 0.027).

Discussion
Using a cohort enriched with BCFH, we found that higher serum androgen concentrations were associated with earlier breast development. These results support another prospective study 4  Before puberty, androgens are primarily produced along the HPA axis, also known as the stress axis.
Stressors may activate the HPA axis prematurely, increasing the pool of androgens, which are then converted peripherally into estrogens to promote puberty.
Our findings support the hypothesis that the elevated androgen concentrations in BCFHpositive girls may be limited to girls with a first-degree BCFH. Because the production and metabolism of androgens differ before, during, and after puberty and because the stress response system is programmed during puberty, both psychosocial and genetic factors may explain elevated androgens in BCFH-positive girls. In an earlier study, BCFH-positive mothers and daughters reported greater breast cancer-specific distress compared with BCFH-negative mothers and daughters. 22 If replicated, it is plausible that early-life stress associated with a mother diagnosed with breast cancer may activate the HPA axis prematurely, thereby producing DHEA-S, which would explain our finding that breast cancer-specific distress was specific to DHEA-S and not the other androgens.
Dehydroepiandrosterone sulfate is metabolized into androstenedione and testosterone; when puberty begins, the hypothalamic-pituitary-ovarian axis is activated and the ovary starts to contribute 25% to 50% of androstenedione and testosterone in circulation. During puberty the cross-talk between the hypothalamic-pituitary-ovarian and HPA axes may result in differential intracellular metabolism of DHEA-S into downstream androgens by 3β-and 17β-hydroxysteroid dehydrogenase enzymes. 27 Whether the production and metabolism of androgens is fixed after puberty, which has been noted for stress-induced adrenal glucocorticoid secretion, or can continue to change after puberty is yet to be determined. Emerging evidence from animal models supports the association of stressful environments with steroid hormone production by showing that both physical and social stress can influence glucocorticoid and androgen production. 28 Furthermore, these models suggest that social stress in adolescence is associated with androgen trajectories, with future implications for disease risk. 29

Limitations
Our study was limited by a small sample size, especially in our prospective subset, which examined androgen concentrations and the timing of breast development. Previous studies 25 have shown that androgen concentrations increase from 18 to 30 months before the onset of puberty; therefore, the prospective study design required the least number of assumptions for statistical analysis. However, to increase our sample size, we also performed a cross-sectional analysis that included girls whose hormone concentrations were measured after the onset of puberty. These measurements were less informative in answering our question as to whether elevated hormone concentrations are associated with earlier breast development. The larger cross-sectional analysis yielded similar results for all of the androgens except DHEA-S. Another limitation was that both the prospective and crosssectional samples were too small to test for interactions between androgen concentrations and