Adiposity and Sex Hormones in Girls

Greater body fatness during childhood is associated with reduced risk of premenopausal breast cancer, but few studies have addressed the relation of adiposity with sex hormones in girls. We prospectively examined associations between adiposity and circulating levels of sex hormones and sex hormone–binding globulin (SHBG) among 286 girls in the Dietary Intervention Study in Children. Participants were 8 to 10 years old at baseline and were followed for an average of 7 years. Anthropometric measurements were taken at baseline and at subsequent annual visits, and blood samples were collected every 2 years. Concentrations of dehydroepiandrosterone sulfate (DHEAS) during follow-up were higher among girls with greater body mass index (BMI) at baseline. The mean for the lowest BMI quartile was 63.0 μg/dL compared with 78.8 μg/dL for the highest quartile, and each kg/m² increment in baseline BMI was associated with a 4.3% increase (95% confidence interval, 1.6-7.0%) in DHEAS levels during follow-up (P_trend = 0.002). Concentrations of SHBG during follow-up were lower among girls with greater BMI at baseline. The mean for the lowest BMI quartile was 94.8 nmol compared with 57.5 nmol for the highest quartile, and each kg/m² increment in baseline BMI was associated with an 8.8% decrease (95% confidence interval, 7.0-10.6%) in SHBG levels during follow-up (P_trend < 0.0001). Estrogen and progesterone concentrations were similar across BMI quartiles. These findings suggest that adiposity may alter DHEAS and SHBG levels in girls. Whether and how these differences affect breast development and carcinogenesis requires further research. (Cancer Epidemiol Biomarkers Prev 2007;16(9):1880–8)

Adiposity is an established risk factor for breast cancer; overweight and obesity are associated with increased risk of breast cancer in postmenopausal women but decreased risk in premenopausal women (1, 2). Although most studies have focused on obesity in adulthood, growing evidence indicates that adiposity during childhood and adolescence may play an even more important role. Record linkage studies have shown strong inverse associations of measured body fatness between ages 7 and 15 years with breast cancer incidence (3-5). Prospective cohort studies with more detailed information on lifestyle factors in adulthood as well as earlier in life have also shown reduced risk of breast cancer among premenopausal women who were heavier at young ages (6-8). For example, in the Nurses' Health Study II, women who reported being most overweight at ages 5 and 10 years had approximately 50% lower risk of premenopausal breast cancer compared with the most lean, and this association was not affected by adjustment for adult body mass index (BMI) and menstrual cycle characteristics (8). In contrast, the inverse association between adult BMI and premenopausal breast cancer risk was substantially attenuated after adjustment for childhood body fatness. Several studies have also shown a decreased risk of postmenopausal breast cancer with greater adiposity at young ages (6, 9, 10).

These findings suggest that adiposity at young ages, perhaps even before puberty, has a strong and independent effect on breast cancer risk later in life. Endogenous sex hormones, which have been implicated in the development of breast cancer, could mediate this relationship. Postmenopausal women with elevated concentrations of sex hormones, including both androgens and estrogens, have increased risk of breast cancer (11, 12). Studies of the relation between estrogens and breast cancer in premenopausal women have been inconclusive (13), but recent prospective studies have observed increased risk among premenopausal women with higher levels of some androgens (14, 15) and follicular phase estradiol (16). Sex hormone concentrations are also related to adiposity (17); BMI is positively associated with blood levels of estrogens and some androgens in postmenopausal women (18-21). Results have been less consistent in premenopausal women, but they suggest that levels of estrogens and progestins may be lower in heavier women, perhaps due to increased frequency of anovulatory cycles (17, 18, 22).

Only a few studies have examined whether adiposity in childhood and adolescence is related to levels of sex hormones in girls, and the results are inconclusive (23-27). Previous studies have been hampered by small numbers of participants and limited information on other factors, including menarcheal status and day of the menstrual cycle, which could influence hormone levels. Therefore, the purpose of this study was to examine associations between adiposity and circulating levels of sex hormones and sex hormone–binding globulin (SHBG) among girls who participated in the Dietary Intervention Study in Children (DISC).

The DISC was a multicenter, randomized controlled clinical trial to examine the safety and efficacy of a dietary intervention to reduce serum low-density lipoprotein cholesterol in children. The study enrolled 663 prepubescent boys and girls with elevated low-density lipoprotein cholesterol levels who were recruited through schools, health maintenance organizations, and pediatric practices between 1988 and 1990. Participants were randomly assigned to either a behavioral dietary intervention to reduce fat intake or to usual care. The intervention and follow-up were originally designed to continue for 3 years and were subsequently extended until participants reached 18 years of age. However, the study was stopped early in 1997, when the mean age of participants was 16.7 years, because serum low-density lipoprotein cholesterol levels in the two treatment groups were not significantly different (28). Details about the design of the DISC and the dietary intervention have been published elsewhere (28-31).

Only female participants were included in this analysis of adiposity and sex hormones. Girls were eligible to participate in the DISC if they were 7.8 to 10.1 years old, had serum low-density lipoprotein cholesterol levels in the 80th to 98th percentiles, had no major illness and were not taking medications that could affect their blood lipid levels or growth, were in at least the 5th percentile for height and in the 5th to 90th percentiles for weight-for-height, were at Tanner stage 1 for breast and pubic hair development (32), and had normal cognitive and psychosocial development. The Hormone Ancillary Study (HAS) was initiated in 1990 to assess the effect of the reduced fat dietary intervention on serum levels of sex hormones during adolescence. Girls were excluded from the HAS if they were pregnant or had used oral contraceptives within 3 months of blood collection, if they were postmenarcheal and missing the date they started their next menses after blood collection, or if their next menses started more than 33 days after blood collection. Of the 301 girls who were included in the DISC, 286 participated in the HAS at one or more visits. Effects of the DISC intervention on serum sex hormones in girls have been reported previously (31).

Data were collected from participants at screening and baseline visits and at subsequent annual visits by trained study staff who were blinded to treatment assignment. Information on demographic and anthropometric characteristics, medical history, medication use, and smoking history was obtained at every visit, and each participant also underwent a brief physical examination that included Tanner staging to evaluate sexual maturation (32). Date of onset of first menses was ascertained annually until menarche. Girls were asked about use of oral contraceptives and pregnancy beginning with the year 3 visits.

Because growth was the primary safety end point of the DISC, particular attention was paid to the quality of anthropometric measurements. Clinic staff were specially trained in a common protocol and were certified and recertified annually. Height was measured with stadiometers centrally constructed by the Medical Instruments Unit of the University of Iowa. Weight was measured using beam balance or electronic scales that were calibrated weekly against a range of standard weights between 20 and 100 kg. BMI was calculated as weight (kilograms) divided by height squared (meters squared). A spring-loaded anthropometric measuring tape (Gulick) was used to measure waist and hip circumferences. Waist circumference was measured midway between the lower edge of the ribs and the iliac crest, and hip circumference was measured at the level of the greater trochanters. Waist-to-hip ratio (WHR) was calculated as the ratio of the latter two measurements (29).

A single blood sample was collected by venipuncture in the morning after an overnight fast at baseline and at the year 1, year 3, year 5, and last visits. Serum was separated by centrifugation after the blood sample was kept at room temperature for at least 45 minutes to allow complete clotting. Serum was then aliquoted and stored in glass vials at −80°C until it was analyzed. Blood samples were not timed to the menstrual cycles of the postmenarcheal girls, but day of the menstrual cycle corresponding to the date of blood collection was determined from menstrual cycle calendars that were completed for 6 weeks before and 6 weeks after blood collection.

Hormone assays were done by Esoterix Endocrinology, Inc. Steroid hormones were measured by RIA, and SHBG was measured by an immunoradiometric assay. The concentration of non–SHBG-bound estradiol was calculated as the product of the total estradiol concentration and the percentage non–SHBG-bound estradiol, which was measured by ammonium sulfate precipitation. Serum samples collected at the same clinic visit were grouped and assayed together in the same laboratory batches. Quality control samples were included in each batch, and laboratory personnel could not distinguish quality control samples from participant samples. These samples were aliquots from three serum quality control pools that were created by using charcoal-stripped serum to dilute serum from adults to achieve the range of steroid hormone concentrations expected in the participants' samples. The within-visit coefficients of variation, as estimated from quality control samples, were 8% to 29% for estradiol, 12% to 31% for estrone, 12% to 17% for estrone sulfate, 8% to 17% for androstenedione, 9% to 22% for testosterone, 5% to 9% for dehydroepiandrosterone sulfate (DHEAS), 4% to 10% for progesterone, and 15% for SHBG. Low concentrations of hormones in quality control samples may have contributed to the higher coefficients of variation for some hormones. For example, the mean concentrations of estradiol in samples from the three quality control pools were 0.9, 2.8, and 11.3 ng/dL, and their corresponding within-visit coefficients of variation were 29%, 11%, and 8% (31). These high coefficients of variation at low estradiol concentrations are relevant given that estrogen concentrations were quite low for many of the participants in the present study and especially in premenarcheal girls, where approximately 30% of samples had estradiol levels at or below the limit of detection (0.5 ng/dL).

A total of 286 girls had available blood samples for hormone assays, with measurements at between 1 and 5 visits (median, 3). Of these, 257 girls had premenarcheal blood samples at between 1 and 4 visits (median, 2), and 222 girls had postmenarcheal blood samples at between 1 and 4 visits (median, 2). Among girls with postmenarcheal blood samples, there were 161 girls with hormone measurements during the follicular phase (range, 1-4 visits; median, 1) and 146 girls with hormone measurements during the luteal phase (range, 1-3 visits; median, 1).

For this analysis, hormone levels were examined according to BMI and other measures of adiposity in girls. BMI was used as the primary measure of adiposity because it has been shown to correlate well with direct measures of percentage body fat in children and adolescents, such as underwater weighing and dual-energy X-ray absorptiometry, and it is highly reproducible (33, 34). Secondary analyses were also done using weight, waist circumference, and WHR as the measures of adiposity. Because previous studies have suggested that very early body fatness may be more strongly associated with breast cancer risk than body fatness at later ages, BMI at baseline (ages 8-10 years) was used in the main set of analyses to predict hormone levels at subsequent visits during the 7-year follow-up period (i.e., prospective analyses). However, in secondary analyses, BMI at the time of blood collection was also examined in relation to hormone levels at the same visit (i.e., cross-sectional analyses). BMIs at baseline and at the time of blood collection were not included in the same multivariate models because they are too highly correlated with one another.

Median ages at menarche for BMI quartiles were estimated by a modified product-limit method that allowed for different entry ages and the censoring of girls who had not reached menarche at their last visit; a similarly modified log-rank test was used to compare groups (35). All serum hormone data were transformed before analysis by taking the natural logarithm to improve normality. Mixed linear regression models (SAS Proc Mixed) were used to calculate mean hormone concentrations during follow-up according to quartiles of BMI at baseline, accounting for repeated hormone measurements within each girl across multiple visits, and the means were then exponentiated to compute geometric means and 95% confidence intervals (95% CI). BMI and other measures of adiposity were also modeled as continuous variables to estimate percentage differences in hormone levels and to do tests for trend.

A centering technique was used to account for age-related differences in BMI at baseline, when girls were between ages 8 and 10 years; the median BMI for a specific age in months, based on normative population data from the National Health Examination and National Health and Nutrition Examination Surveys (36) collected by the Centers for Disease Control,⁶

was subtracted from each girl's baseline BMI value. In addition, age at the time of blood collection was centered on its mean and included as a continuous variable in all models, and a term for age squared was also included to account for nonlinear associations with hormone levels. Interaction terms between age and baseline BMI (both continuous) were included to examine whether associations between BMI at ages 8 to 10 years and hormone levels during follow-up varied with age. Other factors included in regression models were treatment group (intervention or usual care), time since randomization (days), race (black, white, or other), height (centimeters), and physical activity (hours per week of vigorous and moderate activity). Indicator variables were used for treatment group and race, and all other variables were modeled as continuous. Interactions between baseline BMI and treatment group were examined in secondary analyses. Because the hormone assays were run in multiple laboratory batches over the course of the study, indicator variables for batch were included in preliminary analyses. However, adjustment for batch had little effect on the means and reduced the precision of the estimates; therefore, batch was not included in the final models.

For estrogens, separate regression models were fit for premenarcheal girls, postmenarcheal girls in the follicular phase of the menstrual cycle (days 15-33 before and day of onset of next menses), and postmenarcheal girls in the luteal phase of the menstrual cycle (days 1-14 before next menses). Progesterone was only measured in postmenarcheal girls, and separate models were fit for the follicular and luteal phases. For androgens and SHBG, which do not vary substantially over the course of the menstrual cycle, the follicular and luteal phases were combined, and models were also run combining premenarcheal and postmenarcheal girls. Indicator variables for days before next menses were included in all models for postmenarcheal girls. Statistical analyses were conducted using SAS, version 9.1 (SAS Institute, Inc.). All tests of statistical significance were two sided.

The median BMI at baseline among girls in the study was 16.7 kg/m² (range, 13.4-24.3), and BMI was similar in the intervention and control groups at each visit (31). Baseline characteristics of participants according to quartiles of BMI are presented in Table 1. There were no significant differences in age, treatment group, race/ethnicity, or annual household income across BMI quartiles. Weight, waist circumference, and height increased with greater BMI at baseline (for each, P < 0.0001). Spearman correlations between BMI and weight, waist circumference, and WHR at baseline were 0.88, 0.87, and 0.21, respectively. The correlation between BMI and height was 0.39, consistent with some previous studies that have shown weak-to-moderate positive associations between BMI and height in children but not in adults (37). Participants with greater BMI at baseline had lower levels of moderate and intense physical activity, although this association did not reach statistical significance (P = 0.07). Girls with greater baseline BMI had a younger age at menarche; the median age at menarche was 13.4 years for the lowest BMI quartile and 12.6 years for the highest BMI quartile (P < 0.0001). Menstrual cycle length did not vary according to BMI at baseline (data not shown).

Concentrations of estrogens in premenarcheal and postmenarcheal girls during follow-up were generally similar across quartiles of BMI and waist circumference at ages 8 to 10 years (Table 2). There was some evidence that estradiol levels in premenarcheal girls during follow-up were lower among those with greater BMI at baseline (4.7% decrease per kg/m²; P_trend = 0.02), but the association was not monotonic and the absolute difference in means between the extreme quartiles was small (1.3 ng/dL for the lowest quartile versus 1.1 ng/dL for the highest quartile). A similar association was observed for waist circumference at ages 8 to 10 years and estradiol levels in premenarcheal girls (2.7% decrease per centimeter; P_trend = 0.001), and there was also a weak but significant inverse association between waist circumference and estrone levels (1.3% decrease per centimeter; P_trend = 0.01). Progesterone levels in postmenarcheal girls did not vary according to BMI or waist circumference at baseline.

Concentrations of the adrenal androgen DHEAS during follow-up were higher among girls with greater BMI at baseline (Table 3). When all premenarcheal and postmenarcheal girls were combined, the mean for the lowest BMI quartile was 63.0 μg/dL compared with 78.8 μg/dL for the highest quartile, and each kg/m² increment in BMI at baseline was associated with a 4.3% increase (95% CI, 1.6-7.0%) in DHEAS levels during follow-up (P_trend = 0.002). In addition, there was a significant negative interaction between age and BMI in premenarcheal girls (P = 0.003), indicating that the increase in DHEAS with greater prepubertal BMI was less at older ages. Waist circumference at baseline was also positively associated with DHEAS levels during follow-up in both premenarcheal and postmenarcheal girls, but it was inversely associated with testosterone levels in premenarcheal girls; each centimeter increment in waist circumference at baseline was associated with a 2.4% decrease (95% CI, 0.8-3.9%) in testosterone levels during follow-up (P_trend = 0.003).

In contrast, concentrations of SHBG during follow-up were lower among girls with greater BMI at baseline (Table 3). Combining premenarcheal and postmenarcheal girls, the mean for the lowest BMI quartile was 94.8 nmol compared with 57.5 nmol for the highest quartile, and each kg/m² increment in BMI was associated with an 8.8% decrease (95% CI, 7.0-10.6%) in SHBG levels during follow-up (P_trend < 0.0001). There was also a significant negative interaction between age and BMI among premenarcheal girls (P = 0.0003), indicating that the decline in SHBG with greater baseline BMI was less at older ages. Associations between waist circumference at baseline and SHBG levels during follow-up in premenarcheal and postmenarcheal girls were similar to those for BMI. Because SHBG binds to testosterone as well as estradiol and we did not estimate the concentration of non–SHBG-bound testosterone directly, we tried adjusting testosterone for SHBG in some analyses, either by including SHBG in the multivariate models or by using the residuals of testosterone after regression on SHBG (38). These adjustments, however, had no appreciable effect on observed associations between BMI and testosterone levels (data not shown).

The associations between baseline BMI and levels of DHEAS and SHBG during follow-up were similar for the intervention and usual care groups (data not shown). There was a significant interaction between BMI at baseline and treatment group (P = 0.04) for androstenedione among premenarcheal and postmenarcheal girls together, with a positive association among those in the intervention group (2.5% increase in androstenedione per kg/m² increment in BMI; P_trend = 0.05) but not among those in the usual care group. No other important interactions were observed between baseline BMI and treatment group with respect to hormone levels (data not shown).

The results for both premenarcheal and postmenarcheal girls were very similar in the cross-sectional analyses that examined BMI at the time of blood collection in relation to hormone levels at the same visit (data not shown). In analyses using weight at baseline as the measure of adiposity, the same patterns were observed for DHEAS and SHBG as for BMI, but there were also some additional differences in premenarcheal girls. Weight at baseline was inversely associated with concentrations of estradiol (4.9% decrease per kilogram increment in weight; P_trend < 0.0001), non–SHBG-bound estradiol (3.3% decrease per kilogram increment in weight; P_trend = 0.006), and estrone (2.1% decrease per kilogram increment in weight; P_trend = 0.0007) during follow-up in premenarcheal girls (data not shown). In addition, weight at baseline was inversely associated with androstenedione and testosterone levels during follow-up in premenarcheal girls; for each kilogram increment in weight, there was a 1.4% decrease in androstenedione (P_trend = 0.03) and a 2.8% decrease in testosterone (P_trend = 0.004). No significant associations between WHR at baseline and hormone levels during follow-up were observed in either premenarcheal or postmenarcheal girls (data not shown).

In this longitudinal study, BMI at baseline, between ages 8 and 10 years, was associated with circulating concentrations of sex hormones in girls over more than 7 years of follow-up. Although no consistent associations were observed for estrogens or progesterone, greater BMI at baseline was associated with higher levels of DHEAS and lower levels of SHBG during follow-up in both premenarcheal and postmenarcheal girls. These associations for DHEAS and SHBG were similar for other measures of adiposity that were examined. In addition, there was some evidence that levels of other sex hormones during follow-up in premenarcheal girls, including estradiol, non–SHBG-bound estradiol, estrone, androstenedione, and testosterone, decreased with greater BMI, weight, and waist circumference at baseline.

Few previous investigations have examined associations between adiposity and sex hormone concentrations in girls, and most have been small cross-sectional studies with limited information on other factors, including menarcheal status, which could affect hormone levels. In an early study, obese girls had significantly higher plasma levels of DHEAS, androstenedione, and progesterone, but lower levels of estradiol, compared with nonobese girls (23). Another study found that fat mass, assessed by skinfold thicknesses, was positively associated with levels of DHEAS and non–SHBG-bound testosterone, but girls in the highest WHR quartile had the lowest concentrations of total and non–SHBG-bound estradiol and testosterone (24, 25). In a study of weight reduction, obese postmenarcheal girls had higher serum concentrations of androstenedione, testosterone, and DHEAS, but lower levels of SHBG, compared with normal weight girls; in addition, girls in the highest WHR tertile had higher concentrations of testosterone, a higher free androgen index, and lower concentrations of SHBG compared with those in the lowest tertile (26). Levels of estradiol, testosterone, and DHEAS decreased significantly with weight loss, whereas levels of SHBG increased. In another longitudinal study, BMI and waist and hip circumferences were positively associated with concentrations of androstenedione, testosterone, and DHEAS, but there were no significant correlations between WHR and androgen levels (27). A recent cross-sectional study showed that BMI was positively associated with total and free testosterone levels and inversely associated with levels of SHBG (39).

In general, these studies suggest that obesity may lead to increased levels of androgens in girls, which is consistent with our finding of a positive association between BMI and DHEAS. It has been hypothesized that insulin and insulin-like growth factor could be part of this biological pathway, although we were not able to evaluate this in the present study. Many children experience insulin resistance and hyperinsulinemia around the time of puberty, and this seems to be increased in the presence of obesity (40). Insulin is involved in the regulation of insulin-like growth factor-I and its binding proteins, and higher circulating concentrations of free insulin-like growth factor-I have been observed in some studies of obese adolescent and preadolescent girls (41-43). Both insulin and insulin-like growth factor-I stimulate androgen production in the ovary, and hyperinsulinemia and elevated insulin-like growth factor-I levels also may be associated with increased production of adrenal androgens, including DHEAS and androstenedione (40, 44-46). Furthermore, insulin has been shown to inhibit production of SHBG by hepatocytes, and hyperinsulinemia is associated with decreased plasma levels of SHBG (40, 47-51).

Although the positive association between obesity and androgen levels in girls is fairly consistent, it is difficult to reconcile this finding with the reduced breast cancer risk that has been observed among women who were heavy during childhood and adolescence (3-10). Several studies have shown that high levels of androgens in adult women are associated with increased risk of premenopausal and postmenopausal breast cancer (11, 12, 14, 15), but androgens could have a different effect during earlier periods of life. Hyperinsulinemia and increased androgen production in obese adolescent girls may inhibit the maturation of ovarian follicles and reduce the frequency of ovulatory cycles (41, 44, 52-56). Fewer ovulatory cycles, in turn, could decrease cumulative lifetime exposure to estrogens and progesterone, leading to reduced cellular proliferation in the breast and lower breast cancer risk (17).

We had hypothesized that there might be positive associations between adiposity and estrogen levels in premenarcheal girls comparable with those seen in postmenopausal women (17, 21), due to aromatization of excess androgen to estrogen in adipose tissue. Lower levels of SHBG in obese girls also could increase bioavailable estradiol and testosterone, both of which bind to SHBG. There are several possible reasons why we did not observe these associations in our data. First, estrogen levels in premenarcheal girls are low; mean levels of estradiol, non–SHBG-bound estradiol, and estrone among premenarcheal girls in this study were similar to those among the leanest postmenopausal women (BMI <22.5 kg/m²) in a pooled analysis of nine studies of BMI and endogenous hormones (21). Levels of estrogen precursors (i.e., DHEA and androstenedione) in premenarcheal girls are low, and there also may be reduced aromatase activity and subsequent conversion of precursors to estrogens before menarche (57). In addition, the assays may not have been sensitive enough to detect meaningful differences in estrogens in premenarcheal girls; for example, as mentioned earlier, the lower limit of detection for estradiol was 0.5 ng/dL, and approximately 30% of premenarcheal blood samples in this study were at or below this level. However, the mean BMI was similar for premenarcheal girls whose estradiol levels were above and below the limit of detection (data not shown). Another potential factor is that the ovaries begin to secrete estradiol in a cyclical pattern before the onset of menarche (58-61), increasing within-person variation in serum levels.

This study has several limitations that are worth noting. Estrogen and progesterone concentrations vary substantially throughout the course of the menstrual cycle among postmenarcheal girls, but samples were not timed to the menstrual cycle; this could attenuate associations, although we did adjust for days until next menses in our models. Furthermore, because a large proportion of menstrual cycles in girls during the first few years after menarche are anovulatory (52-54, 62, 63), more follow-up time may be necessary for differences in estrogen and progesterone levels to become apparent. Another potential limitation is reduced generalizability because girls were selected to participate based on having elevated levels of low-density lipoprotein cholesterol and tended to be slightly heavier than girls in the general population; in addition, because girls had to be in the 5th to 90th percentiles for weight-for-height to be eligible to participate in the DISC, extremely lean and extremely obese girls were not included. We also did not adjust for BMI at the time of blood collection, due to its high correlation with BMI at ages 8 to 10 years.

Despite these issues, this study has some important strengths. It was a longitudinal study in which nearly 300 girls were followed for an average of 7 years and repeated blood samples were taken, giving us substantial power to examine associations of adiposity with hormone levels over time. Because the main safety end point of the DISC study was growth, high-quality anthropometric data were collected and we could examine multiple measures of adiposity in relation to hormone levels. Finally, we had detailed information on a variety of other factors that could influence hormone levels, including menarcheal status and day of the menstrual cycle.

In summary, our findings suggest that prepubertal girls who are heavy may have higher levels of DHEAS and lower levels of SHBG during puberty compared with those who are lean. Further research is necessary to examine whether and how these hormonal differences may possibly affect breast tissue development and risk of breast cancer later in life.

We thank the girls who participated in the DISC and their families; DISC Steering Committee: Peter O. Kwiterovich, Jr., M.D., Ronald M. Lauer, M.D., Linda Van Horn, Ph.D., R.D., Norman L. Lasser, M.D., Ph.D., Victor J. Stevens, Ph.D., Bruce A. Barton, Ph.D., Sue Y.S. Kimm, M.D., M.P.H., Alan M. Robson, M.D.; the DISC Data and Safety Monitoring Committee: John C. LaRosa, M.D. (Chair), Phyllis E. Bowen, Ph.D., Allan Drash, M.D., Robert J. Hardy, Ph.D., Judith K. Ockene, Ph.D., Carol Whalen, Ph.D., Richard H. Grimm, M.D., Ph.D., and Melvin M. Grumbach, M.D. (ad hoc member); Eva Obarzanek, Ph.D., R.D. (Project Officer), Jeffrey A. Cutler, M.D., Marguerite A. Evans, M.S., R.D., and Denise G. Simons-Morton, M.D., Ph.D. (National Heart, Lung, and Blood Institute Project Office, Bethesda, MD); and Catherine S. Berkey, Ph.D. for statistical consultation.