Birthweight and Body Size throughout Life in Relation to Sex Hormones and Prolactin Concentrations in Premenopausal Women

Birthweight and attained adult height are positively associated with premenopausal breast cancer risk, whereas being heavy during childhood, adolescence, and adulthood are inversely associated with risk (1-5). One mechanism by which these exposures may alter risk is through sex hormone concentrations. For example, adult testosterone levels have been associated with an increased risk of premenopausal breast cancer (6-10), whereas estradiol levels were positively associated with risk in some (6, 10-12), but not all (7, 8, 13), studies.

Although both birthweight and attained height are thought to reflect in utero and early childhood exposures, some data suggest that these markers of early growth are correlated with sex hormone levels during infancy (14, 15), childhood (16-18), adolescence (19-21), and possibly early adulthood (22, 23). However, few studies have looked at these factors in relation to mid- to late-premenopausal hormone levels. Some studies have examined the associations between childhood (24-27) and adult (28-41) body mass index (BMI) with premenopausal sex hormones, but the results have been inconsistent. This may be due to the small sample size of many studies and because blood samples were collected without regard to timing in the menstrual cycle in many studies, and although some studies accounted for this in the statistical analyses (30, 40), others did not (33, 35, 36, 42).

Therefore, we examined the associations of birthweight, body shape at ages 5 and 10, BMI at age 18 and adulthood, adult waist circumference, waist-to-hip ratio (WHR), and attained height in relation to plasma concentrations of estrogens, androgens, progesterone, prolactin, and sex hormone–binding globulin (SHBG) in nearly 600 premenopausal women, ages 33 to 52 years old at blood draw, from the Nurses' Health Study II. A unique aspect of this study is that timed blood samples were collected in both the early follicular and mid-luteal phases of the menstrual cycle, allowing us to assess phase-specific associations for the estrogens.

The Nurses' Health Study II was established in 1989, when 116,609 female registered nurses, ages 25 to 42, completed a questionnaire. The cohort has been followed biennially since its inception to update exposure variables and ascertain newly diagnosed disease. The racial/ethnic breakdown is 94% White, 2% Asian, 2% African-American, and 2% Hispanic.

Between 1996 and 1999, 29,611 cohort members, ages 32 to 54 years, provided a blood sample (described in ref. 43). Briefly, premenopausal women who had not taken any hormones, been pregnant, or breastfed within 6 months (n = 18,521) answered a short questionnaire and provided timed blood samples on the 3rd to 5th day of their menstrual cycle (follicular sample), and 7 to 9 days before the anticipated start of their next cycle (luteal sample). Follicular plasma was aliquoted by the participant 8 to 24 h after collection and frozen. All other women (n = 11,090) provided a single untimed blood sample. Luteal and untimed samples were shipped via overnight courier, processed by our laboratory, and separated into plasma, RBC, and WBC components. Samples have been stored in continuously monitored, liquid nitrogen freezers since collection.

We restricted the analysis to premenopausal women, who were defined as having provided timed samples, or for women who provided untimed samples, those who reported that her periods had not ceased, or who reported having had a hysterectomy but with at least one ovary remaining, and were ≤47 (for nonsmokers) or ≤45 (for smokers) years of age. Follow-up of the blood cohort was 98% in 2003. Participants in this study were controls from a nested case-control study who were matched to breast cancer cases diagnosed after blood collection and before June 2003 (43), and a subset of women who were included in a reproducibility study (44). The study was approved by the Committee on the Use of Human Subjects in Research at the Brigham and Women's Hospital.

Information on exposures and potential covariates were asked on a questionnaire completed upon blood collection and during the biennial study questionnaires. In 1989, participants recalled their body fatness at ages 5 and 10, using a nine-level figure drawing or somatotype (ref. 44; Fig. 1) originally developed by Stunkard and colleagues (45). Women's recall of body shape in childhood has been validated against weight and height measurements taken in childhood (47). Lifetime oral contraceptive use, age at menarche, cycle regularity between ages 18 and 22, weight at age 18, and height were reported at baseline in 1989; oral contraceptive use was updated on subsequent biennial questionnaires. In 1991, participants were asked to report their birthweight in one of the following categories: unknown, <5.5, 5.5 to 6.9, 7.0 to 8.4, 8.5 to 9.9, or 10+ pounds. Current weight and details about the blood collection date, time, and fasting status were reported on the blood questionnaire. BMI at age 18 and current BMI were calculated as weight in kilograms divided by attained height in meters squared. In 1993, women were asked to measure their waist and hip circumferences, to the nearest 1/4 inch, if they had a tape measure easily available; 64% of women provided these measurements.

We have previously conducted validation studies of self-reported current weight, weight at age 18, waist and hip circumferences, and birthweight (48-50). Mean self-reported weights were ∼1.5 kg lighter than the technician-measured value, compatible with the difference between a random casual weight in clothing and a nude, post-void morning weight. The correlations between measured and self-reported weight were r = 0.96 and did not differ by level of BMI. Correlations between self-report and technician measurements were 0.89 for waist circumference, 0.84 for hip circumference, and 0.70 for WHR (48). The correlation between recalled weight at age 18 and that documented in college/or nursing school physical exam records was 0.84 (49), and the Spearman correlation between self-reported birthweight and that reported on the birth certificate was 0.74 (50).

Hormone assay methods for estrogens and testosterone have been described previously (51). In brief, samples were assayed at Quest Diagnostics (San Juan Capistrano, CA) by RIA following extraction and celite column chromatography. After extraction of estrone, enzyme hydrolysis, extraction, and column chromatography, estrone sulfate was assayed by RIA of estrone. Free estradiol and testosterone were calculated according to Sodergard et al. (52). At the Royal Marsden Hospital, dehydroepiandrosterone (DHEA) and androstenedione were assayed by RIA (Diagnostic Systems Laboratories, Webster, TX), and dehydroepiandrosterone sulfate (DHEAS), SHBG, and progesterone were measured by chemiluminescent immunoassay using the Immulite autoanalyzer (Diagnostic Products, Corp., Llanberis, United Kingdom). Prolactin was measured using a microparticle enzyme immunoassay at the Massachusetts General Hospital, using the AxSYM Immunoassay system (Abbott Diagnostics, Chicago, IL). We measured the hormones on the following sample sets: estradiol, estrone, and estrone sulfate in follicular and luteal samples; testosterone, androstenedione, prolactin, and SHBG in follicular, luteal, and untimed samples; DHEA and DHEAS in luteal and untimed samples; and progesterone in luteal samples.

Follicular and luteal samples from each woman were assayed together; samples were assayed in three batches. The interassay coefficients of variation from masked replicate samples in each batch were 6% to 14% for all hormones except progesterone (coefficient of variation, 17%). Correlations from a subset of 12 samples run in two of these batches were >0.90 for all hormones.

For each analyte, we excluded women with missing values related to assay difficulties or low volume. We also identified and excluded a small number of values (n ≤ 6 per hormone) that were statistical outliers (6, 43, 53). For the estrogens, we examined the associations with follicular and luteal measures separately. For testosterone, androstenedione, prolactin, and SHBG, we averaged the follicular and luteal values, as levels did not vary substantially between the phases (6, 43, 53).

Primary analyses calculated adjusted geometric means by category of exposure, using a general linear model. Exposures consisted of birthweight (<5.5, 5.5-6.9, 7.0-8.4, 8.5+ pounds), somatotype at ages 5 and 10 (1, 2, 3, 4, 5+), BMI at age 18 (<19, 19 to <21, 21 to <23, 23 to <25, 25+ kg/m²), BMI at blood collection (<20, 20 to <22.5, 22.5 to <25, 25 to <27.5, 27.5 to <30, 30+), and quartiles of waist circumference (<60.5, 60.5 to <65.5, 65.5 to <72.6, 72.6+ cm), WHR (<0.73, 0.73 to <0.77, 0.77 to <0.82, 0.82+), and height (<139, 139 to <143, 143 to <147, 147+ cm). Tests for trend were conducted by modeling continuous exposure measures and calculating the Wald statistic (54). In analyses of birthweight and somatotype at ages 5 and 10, we excluded women born preterm or as part of a multiple birth. For all exposures, we conducted secondary analyses restricted to ovulatory cycles (defined as progesterone ≥400 ng/dL), to women who reported having regular menstrual cycles from age 18 to 22 years, and to parous women. Stratified analyses by age and BMI at blood draw used a multiplicative interaction term. Women with missing exposure or hormone information were excluded only for the specific analyses of those exposures or hormones.

Multivariate models adjusted for assay batch (1, 2, 3), age at blood draw (<40, 40 to <45, 45+ years), fasting status (≤10, >10 h) and time of day of the follicular and luteal (or untimed) blood draws (1-8 a.m., 9 a.m.-noon, 1 p.m.-midnight); month of blood draw (continuous); difference between luteal blood draw date and date of the next menstrual period (3-7, 8-21 days, unknown/untimed); duration of past oral contraceptive use (never, <4, 4+ years, missing); and ovulatory status at the blood draw (ovulatory, anovulatory, untimed). In analyses of waist circumference and WHR, we additionally adjusted for BMI (continuous). We also considered other potential confounders including simple hysterectomy, history of benign breast disease, family history of breast cancer, and parity; however, these did not change the results and therefore were not included in the final model. All P values were two-sided and considered statistically significant if ≤0.05.

Among 592 women available for analysis, the mean age was 43 years (Table 1). Eighty-six percent of women provided timed samples, and of those, 91.1% were ovulatory; 77.3% of women reported regular cycles at ages 18 to 22 years. Among women born full-term, 3.4% weighed <5.5 pounds at birth and 15.2% weighed ≥8.5 pounds. Few women had a large body size (somatotype ≥ 5) at age 10 years (9.9%), and, on average, women had a BMI at age 18 equal to 21.0 kg/m². Mean BMI at blood collection (current BMI) was 25.0 kg/m². Sex hormone, prolactin, and SHBG levels were in the expected ranges for premenopausal women (55).

Generally, birthweight was not associated with premenopausal sex hormone levels (Table 2). However, birthweight was inversely associated with luteal estrone and estrone sulfate (P trend = 0.04 for both), with 7% and 30% lower levels, respectively, between women weighing 8.5+ versus <5.5 pounds at birth. There was a suggestive inverse association of birthweight with DHEAS (P trend = 0.12, comparable % difference = 23%). These results were substantially attenuated when excluding women with anovulatory cycles or irregular menstrual cycles (data not shown). The associations did not vary by age or BMI (data not shown).

Somatotype at age 5 was not associated with any hormones (data not shown). However, somatotype at age 10 was significantly inversely associated with follicular and luteal estrone sulfate (P trend = 0.05 and 0.02, respectively; Table 3); we observed 23% and 25% lower levels, respectively, between women reporting the highest versus lowest somatotypes. The associations remained when excluding women with anovulatory cycles or irregular menstrual cycles, or when adjusting for BMI at age 18 or at blood draw (data not shown). In general, we observed few associations with the other hormones overall or when stratified by age (data not shown). Also, among women with a current BMI < 25 kg/m², somatotype at age 10 was positively associated with SHBG (P trend = 0.06; % difference, highest versus lowest category = 20%, P interaction = 0.16) and inversely associated with free testosterone (P trend = 0.05; comparable % difference = −32%, P interaction = 0.21). The average of somatotypes at ages 5 and 10 were not associated with sex hormone levels (data not shown).

Current BMI was strongly inversely associated with SHBG (P trend < 0.001) and positively associated with free testosterone concentrations (P trend < 0.001; Table 4). Although current BMI was not associated with follicular or luteal free estradiol, it was inversely associated with follicular and luteal total estradiol (P trend for both < 0.001). We observed an inverse association between BMI and testosterone (P trend = 0.02), androstenedione (P trend < 0.001), and progesterone (P trend = 0.001). Comparing women with a BMI ≥ 30 versus <20 kg/m², levels were 51% lower for SHBG, 53% higher for free testosterone, 39% lower for follicular estradiol, and 20% lower for luteal estradiol. The results were similar when including only women with ovulatory cycles and regular menstrual cycles (Table 4). Results also were similar among parous women or when stratifying by age at blood draw (data not shown). The results for BMI at age 18 were largely similar to, but somewhat weaker than, those for current BMI; furthermore, these associations were substantially attenuated when adjusting for current BMI (data not shown).

After adjustment for current BMI, we observed few associations between waist circumference (Table 5), WHR, or height with sex hormones overall or in a priori secondary analyses (data not shown). However, height was inversely associated with DHEA (P trend = 0.02; % difference, <139 versus ≥147 cm = −13%), luteal estradiol (P trend = 0.02; comparable % difference = −16%), and luteal free estradiol (P trend = 0.03; comparable % difference = −13%).

Overall, we observed that early life factors, such as birthweight and childhood body size, were not strongly associated with mid to late premenopausal sex hormone concentrations, with the possible exception of estrone sulfate. However, adult BMI, but not late adolescent BMI, waist circumference, WHR, or height, was strongly inversely associated with SHBG, androgen, and total estrogen levels. For current BMI, these associations were robust and remained after excluding women with anovulatory cycles or irregular menstrual cycles. Our results suggest that the inverse association between BMI and breast cancer risk in premenopausal women might possibly be mediated in part by sex hormones.

Several studies have reported alterations in sex hormone concentrations among females with a low birthweight. In infants, girls, and adolescents, low birthweight has consistently been associated with higher follicle-stimulating hormone (14, 15, 17, 18, 20) and androgen levels (16, 19-21), which is consistent with reports of earlier menarche among girls born with low birthweights (56). Results have been inconsistent for estradiol and SHBG (17, 18, 20, 21). Only two studies to date have examined associations of birth size and premenopausal sex hormone levels. One reported no association between intrauterine growth retardation and DHEAS, androstenedione, testosterone, or SHBG (22). The other study of 26- to 35-year-old women (n = 135) reported lower estradiol levels (aligned by day in menstrual cycle) in the lowest birthweight quartile versus the top three quartiles (23). We did not observe a significant association with follicular or luteal estradiol; however, follicular levels were suggestively lower among women with the lowest birthweights. We did observe an inverse association of birthweight with luteal estrone and estrone sulfate, apparently driven by women with anovulatory cycles; however, given the large number of comparisons, this result should be interpreted with caution. Overall, existing data suggest that birthweight may be associated with sex hormone concentrations in females through adolescence, but not into adulthood.

Few studies have examined the associations between childhood body size and postmenarchal sex hormone concentrations. Among four studies of adolescents (24-27), higher childhood body size was generally associated with higher androgen and lower SHBG levels, and possibly, higher estradiol concentrations. In an adult population, we observed very few associations, although both follicular and luteal estrone sulfate were inversely related to body size at age 10, even after adjustment for current BMI or when excluding anovulatory cycles. Again, these results suggest that whereas childhood body size may be associated with sex hormones during adolescence, the effect generally does not extend into adulthood.

Interestingly, we observed strong associations between BMI at age 18 and both estrogens and androgens; however, these associations became null after adjustment for current BMI, suggesting that the association with BMI at age 18 was due to its strong correlation with current BMI (r = 0.56). We and numerous previous investigators (28, 30-40, 42) observed a strong inverse association between SHBG and BMI in premenopausal women. Consistent with this finding, we also observed that free testosterone levels were higher among overweight and obese women compared with normal weight women. Interestingly, we did not observe a similar association with free estradiol, likely due to the lower total estradiol levels in women with high BMI. Although several studies have examined the relationship between current BMI and premenopausal estrogens and androgens, the results are relatively inconsistent (28, 30-40, 42). Possible reasons for this may be the small sample sizes (n < 110; refs. 31-35, 37, 39, 42) and not having collected samples timed in the menstrual cycle (30, 33, 35, 36, 40, 42). Among six studies collecting timed follicular or luteal samples or conducting phase-specific analyses, four smaller studies (n = 28, 48, 50, and 107) reported no association with follicular or luteal estradiol, DHEAS, androstenedione, or testosterone (31, 32, 37, 39). A fifth small study (n = 88) observed an inverse relationship between follicular estradiol and BMI (29). Furthermore, in a study of nearly 3,000 women with follicular blood samples, an inverse association was reported between current BMI and estradiol and DHEAS, and a positive association with testosterone (38). In our study, we observed a similar association with estradiol, but the opposite association for testosterone and no association with DHEAS. When restricted to follicular samples for the testosterone analysis, the inverse trend was somewhat attenuated, and when restricted to luteal samples (i.e., excluding untimed samples) for the DHEAS analysis, we still observed no association with current BMI.

Two hypotheses may explain the possible inverse relationship between BMI and total estradiol levels. First, a high BMI may be associated with ovulatory insufficiency, beyond its known role in increasing anovulatory cycles (55). Our data support this hypothesis in that increasing BMI was associated with lower total estradiol and androgen levels even after excluding women with anovulatory cycles and irregular menstrual cycles. The hypothesis is also supported by epidemiologic data suggesting that a BMI as low as 24 kg/m² is associated with an increased risk of ovulatory infertility (57). In addition, we observed that the association of BMI with androgen levels was stronger for androgens derived primarily from the ovary. Specifically, the strongest inverse association was with androstenedione, 50% of which is ovarian-derived, whereas the inverse association with BMI was weaker for testosterone and DHEA, which are ∼25% ovarian-derived (58). We did not observe an association between BMI and DHEAS, a hormone that is produced exclusively by the adrenal gland (59). Estrogens derive from both the ovaries and adipose tissue in premenopausal women (55), with relatively less ovarian contribution during the follicular versus luteal phase (55). We observed similar associations of BMI with follicular and luteal estradiol, suggesting that reductions in ovarian function due to body fat may outweigh the increased peripheral production. This, in part, may be because peripheral conversion primarily creates estrone rather than estradiol (55).

A second hypothesis for the observed inverse relationship between BMI and total estradiol levels may be through an indirect regulation by SHBG. As SHBG declines, free estradiol should increase. However, the pituitary and hypothalamus maintain strict regulatory control over free estradiol levels in premenopausal women (55). So, in response to decreased SHBG, follicle-stimulating hormone levels may decrease to lower total estradiol production by the ovaries, thus keeping free estradiol relatively constant. Additionally, the molecular clearance rate of estradiol is positively associated with weight, also potentially reducing total estradiol levels (60). There are no clear feedback mechanisms for testosterone (55, 61-63), such that free testosterone levels are higher with increasing BMI.

The relationships in premenopausal women somewhat contrast those in postmenopausal women. In this latter population, increased BMI is associated with decreased SHBG (as in premenopausal women), but higher total and free estradiol levels (30, 40, 64). In fact, the increased risk of postmenopausal breast cancer risk with higher BMI seems to be largely due to increased estrogen levels (64). However, in premenopausal women in whom higher BMI is associated with a slightly lower risk of breast cancer (1), increased adiposity is associated with lower SHBG and total estradiol, but not free estradiol. Although free estradiol levels remain fairly constant, it is possible that non–SHBG-bound estradiol (free plus albumin-bound), the estradiol fraction hypothesized to be most bioavailable to tissue (55), may be decreased in obese women. However, to our knowledge, this association has not been examined previously.

Finally, we observed very few associations with waist circumference or WHR after adjustment for BMI, suggesting that there is little additional value to this measure above BMI in predicting premenopausal sex hormone levels. Also, we observed very few associations with attained height. It is possible that the relationship between height and breast or ovarian cancer risk is primarily through growth factors.

The primary limitation of this study is its cross-sectional nature, precluding the ability to determine causal relationships. However, this is a large study, with nearly 600 premenopausal women, most of whom have carefully timed follicular and luteal blood samples. This collection is unique to our study, allowing phase-specific analyses and reducing misclassification by time in the menstrual cycle of the blood draw. A second limitation is that a single blood sample provides a somewhat imprecise measure of long-term average hormone levels (44), and thus could attenuate our results. Additionally, both prolactin and DHEA are affected by circadian rhythms; however, results were similar after restricting to blood draws conducted before 11 a.m.

In conclusion, our results suggest that early life factors, including birthweight and childhood body shape, do not have clear long-term associations with adult premenopausal hormone levels, but previous data (24-27) do suggest an association with childhood hormone levels. Furthermore, higher adult BMI was strongly associated with lower SHBG, total estrogen, and androgen concentrations. Although this may in part explain why BMI is associated with a reduced risk of breast cancer in premenopausal women (1), the lack of association between BMI and free estradiol suggests that the relationships are complex. Future studies, with appropriately timed samples, should further examine how body size may alter sex hormone levels in childhood, adolescence, and adulthood, to better understand the mechanisms through which adiposity throughout life might affect breast cancer risk.

We thank Elizabeth Folkerd and Deborah Doody for their assistance in performing selected assays.