Sexual Dimorphisms in the Associations of BMI and Body Fat with Indices of Pubertal Development in Girls and Boys

Abstract

Context:

The effect of obesity and concomitant insulin resistance on pubertal development is incompletely elucidated.

Objective:

To determine how measures of adiposity and insulin resistance are associated with pubertal maturation in boys and girls.

Setting and Design:

Breast and pubic hair Tanner stage and testicular volume by orchidometry were determined by physical examination in 1066 children. Ovarian volume was estimated by trans-abdominal ultrasound. Fat mass, skeletal age, and fasting serum for insulin and glucose, total T, estradiol, estrone, dehydroepiandrosterone-sulfate, and androstenedione were measured at the National Institutes of Health Clinical Research Center. Convenience sample; 52% obese, 59% female.

Results:

Logistic regression identified a significant interaction between sex and obesity for prediction of pubertal development (P ≤ .01). There was a negative association between boys' testicular volume and body mass index (BMI)/fat mass but a positive association between girls' breast stage and BMI/fat mass. Ovarian volume in girls was positively associated with insulin resistance but not with BMI/fat mass. There was a positive association between obesity and measures of estrogen exposure (breast development and skeletal age) in both sexes. Positive correlations were seen for girls between BMI and pubic hair development and between insulin resistance and T production, whereas adiposity was negatively associated with pubic hair in boys.

Conclusions:

Significant sexual dimorphisms in the manifestations of pubertal development are seen in obese girls and boys. Two known effects of obesity, increased peripheral conversion of low-potency androgens to estrogens by adipose tissue-aromatase and increased insulin resistance, may be in large part responsible for these differences.

Analyses of secular trends in the United States (1) suggest that the average age for thelarche, the secondary sexual characteristic most associated with gonadarche in girls, decreased during the latter half of the twentieth century. During this same interval, the prevalence of pediatric obesity, defined as a body mass index (BMI) at or above the 95th percentile for age and sex based on United States national standards, tripled (2), leading to speculation that excessive adiposity might be one of the important causes for earlier pubertal development (3). Many cross-sectional studies of girls have demonstrated a link between more advanced breast development and higher BMI (4–13) or fat mass (8, 9, 11). Longitudinal analyses demonstrate that obesity precedes the earlier pubertal development in girls (14–19) and that obese girls achieve menarche earlier than their leaner counterparts (13, 20–27). In contrast to the unanimity of results among girls, the association between pubertal development and adiposity in boys has been less consistently reported, with studies finding a positive association (5, 28–30), lack of association (7, 31), and a negative association (4, 32) between obesity and testicular volume or onset of testicular enlargement.

Data from adults suggest a strong positive association between body weight and capacity to convert androstenedione to estrone (33–35). Adipose tissue expresses aromatase (36, 37) that catalyzes the conversion of low-potency androgens to higher-potency estrogens. Aromatase activity seems to be stimulated by macrophage-derived proinflammatory factors that are commonly increased in obesity (38). Androgen-to-estrogen conversion rates and estrogen concentrations are proportional to excess body weight among male and female adults (33, 35, 39). Thus, obesity might predispose children to produce estrogens without gonadal axis activation, leading to progression of estrogen-mediated physical growth. Consistent with this hypothesis, skeletal maturation, which seems closely linked to estrogen action (40, 41), has been found to be advanced in obese children as a group (20, 42–45) and when analyzed separately in boys (7, 28, 46–48) and girls (7, 47–49). Another sign of estrogen effect, gynecomastia in males, is also more prevalent among obese than among nonobese males (14, 50, 51), although it may be difficult to distinguish from lipomastia.

The associations between excess body weight and androgen concentrations are less clear. Whether weight is associated with the timing of adrenarche is controversial. Obese females demonstrate higher T (7, 52) and dehydroepiandrosterone-sulfate (DHEA-S) (7, 53, 54) than lean females; obese males also have higher DHEA-S (7, 28, 54) but lower T even at the same Tanner stage as their lean peers (7, 35, 55). Pubic hair, a manifestation of androgen exposure, shows unclear associations with weight. Several studies have found a lack of association between these parameters (7, 9), or even later onset of pubic hair in obese children (56), although some studies have reported earlier development of pubic hair in both obese girls and boys (10, 29, 30, 57, 58). Data from the National Health and Nutrition Examination Survey (NHANES) III showed a positive association in girls but no association in boys (13).

High adiposity typically leads to increased insulin resistance (IR), which may affect sex hormone concentrations. In females such a relationship is seen in polycystic ovary syndrome, where the degree of IR is positively correlated with androgen concentrations (59, 60). Furthermore, in vitro, insulin stimulates ovarian androgen production (61). In adult men, the IR seen in type 2 diabetes mellitus and obesity often is inversely associated with androgen concentrations (62–64), and T therapy may improve IR (65). Whether IR plays a role, independent of adiposity, in pubertal development is unknown.

To examine the relationships of adiposity, IR, and puberty, we analyzed a large cohort of children and adolescents oversampled for subjects with severe obesity to demonstrate the effects of high adiposity more convincingly. We hypothesized that obesity would be associated with alterations in secondary sexual characteristics and gonadal development consistent with elevated nongonadal estrogen production.

Materials and Methods

Participants

A convenience, cross-sectional sample of children (ages 5–17 years) was constructed from subjects studied between July 1996 and September 2010 (ClinicalTrials.gov IDs: NCT00001195, NCT00001522, NCT00001723, NCT00005669, NCT00320177, NCT00263536, NCT00680979, NCT00315172, and NCT00631644). Participants were recruited through notices mailed to children in local school districts, advertisements in local newspapers, and mailings to local family physicians and pediatricians requesting participation of children willing to undergo phlebotomy and imaging assessments for studies investigating hormones and metabolic functioning in children or for studies investigating the prevention or treatment of obesity. For this analysis, data exclusively from visits before initiation of any treatment were included. Participants in nontreatment studies were financially compensated for their time and inconvenience. All studies were approved by the Eunice Kennedy Shriver National Institute of Child Health and Human Development Institutional Review Board. Children provided written assent and parents or guardians gave written consent for study participation. Subjects were eligible if they were 5–17 years old at the time of study entry, generally healthy, and medication free for at least 2 weeks prior to baseline evaluation.

Physical measurements

Participants were seen at the Clinical Research Center at the National Institutes of Health (NIH). An endocrinologist or trained nurse practitioner performed a medical history and a physical examination on all subjects. Height (cm), measured three times, and weight (kg) were obtained using calibrated instruments. Breast development was assigned according to the five stages of Tanner (66) based on inspection that were corrected using the results of direct breast palpation (67), which is suggested as the technique to distinguish lipomastia from true gynecomastia in boys (68). Testicular volume (cc) was measured by using a set of orchidometer beads as standards according to Prader (69). In cases where the stage/size was discordant between the right and left breasts/testes, the higher stage/size was assigned. To facilitate analyses combining girls and boys, testicular volume was further classified into stages as follows: I (1–3 cc), II (>3–6 cc), III (>6–10 cc), IV (>10–15 cc), and V (>15–25 cc). Subjects were considered prepubertal if they were classified as stage I. Ratings of pubic hair also were made in both girls and boys according to the five stages of Tanner (66, 70), but breast/genital measures were used to determine entry into clinical puberty because these are held to be the most accurate single indicators of sexual maturity (71).

Measures of weight and adiposity

Height and weight were used to compute BMI (kg/m²) and BMI SD (BMI z) scores (72). Subjects were considered obese if BMI z was ≥1.65 (BMI ≥95th percentile for age and sex) (73). Total body fat mass (kg) and percentage fat mass (%) were assessed either with air-displacement plethysmography (Life Measurement Inc) after an overnight fast with participants wearing only underclothes or with whole-body dual energy x-ray absorptiometry (DXA) using the QDR200 pencil beam, QDR4500A, or Delphi fan beam densitometers (Hologic). When measurements from both techniques were available, results were averaged. Subjects were considered overfat if the total body fat or percentage body fat was ≥75th percentile for age based on DXA scan data from NHANES 1999–2004 (74). This cut point has been suggested to identify most children with BMI ≥95th percentile as overfat and relatively few children with normal BMI as overfat (75).

Radiologic measures

Skeletal age was determined by a left-hand roentgenogram and assessed by a radiologist according to the technique of Gruelich and Pyle (76). Skeletal age advancement was calculated as skeletal age minus chronological age. Transabdominal ultrasonography was performed on girls and read by a single radiologist to assess ovarian volume, calculated as 0.52 × length × width × depth. Average ovarian volume was calculated as the mean of the right and left ovary if both were visualized, but ovaries with cysts greater than 1 cm in any dimension were not included in the analysis. Ultrasonography was also used to identify uterine length. Ovarian volumes and uterine lengths were converted to z scores for age using previously published standards (77, 78).

Serum measures

Serum estradiol was measured by electrochemiluminescence immunoassay on a Roche Cobas analyzer (sensitivity, 5 pg/mL; interassay CV, 1.6–2.5%; intraassay CV, 3.5–5.7%). Chemiluminescent immunoassays on a Siemens Immulite 2500 analyzer were used to measure total T (sensitivity, 20ng/dL; interassay CV, 10.1–10.9%; intraassay CV, 5.1–10.1%) and DHEA-S (sensitivity, 0.03 ug/mL; interassay CV, 5.87–16.3%; intraassay CV, 4.15–8.82%). High-performance liquid chromatography /tandem mass spectrometry was used to determine estrone (sensitivity, 5 pg/mL; interassay CV, <6%; intraassay CV, <8%) (79) and androstenedione (sensitivity, 13 ng/mL; interassay CV, <8%; intraassay CV, <8%) (80). Fasting plasma glucose (analyzed with a colorimetric assay with glucose oxidase) and insulin (analyzed with a chemiluminescence immunoassay) were collected as previously described (81); participants had been instructed to adhere to at least an 8-h overnight fast prior to testing. Fasting samples were used to estimate IR by the homeostasis model assessment–insulin resistance index (HOMA-IRI), insulin (mU/mL) × glucose (mmol/L)/22.5.

Statistical analysis

All analyses were conducted using SPSS for Windows, 18.0.0 (SPSS, Inc.). The main study hypothesis was that the prevalence of stage II or greater pubertal development (defined by breast stage for girls and testicular volume for boys) would show different patterns for girls and boys based on participant's degree of obesity. A series of logistic regressions was therefore conducted to determine whether there was a significant interaction between sex and measures of weight (BMI z score) or adiposity (fat mass and percentage fat mass) for the prediction of pubertal status (stage I vs. stage II–V). Age and race were used as covariates to account for age-dependent pubertal progression and differences in pubertal timing in Blacks vs. Whites. When this primary hypothesis was confirmed, each sex was then analyzed separately using logistic regression and analyses of covariance (ANCOVAs), retaining age and race as covariates, in order to examine differences between obese and nonobese groups in pubertal development, skeletal maturation, pelvic ultrasound results, and hormone values. Continuous variables were analyzed for skew and kurtosis; estradiol and T were noted to have outliers, which were adjusted to fall 1.5 times the interquartile range below or above the 25th or 75th percentile (82). Percentage fat mass was transformed using the arcsin square root function. Total fat mass and ovarian volumes were log transformed. Height squared (m²) was used as a covariate in analyses that included fat mass to normalize for body size. BMI z, percentage fat mass, total fat mass, skeletal age, skeletal age advancement, testicular size, and all serum values were considered as continuous variables. Breast and pubic hair Tanner stage were evaluated as discrete variables and thus were always entered into models as an independent variable in relevant analyses; however, to obtain β-coefficients, these variables were classified as continuous. For each dependent variable, a family of four comparisons was performed; thus, P values less than .0125 were considered significant.

Results

A total of 1066 children, 59% female, were studied (Table 1). BMI z scores ranged widely, with approximately half the sample considered obese (55% of girls and 47% of boys), defined by a BMI ≥95th percentile for age (BMI z score ≥1.65), or overfat (58% of girls and 46% of boys), defined by total body fat mass or percentage fat mass ≥75th percentile for age. Distribution of pubertal stages within nonobese and obese girls and boys, as well as unadjusted means for skeletal age, skeletal age advancement, and sex hormone concentrations are presented in Table 1.

Table 1.

Characteristics of the Study Population

Characteristic	Females (n = 627)		Males (n = 439)
	Non-Obese	Obese	Non-Obese	Obese
n	285	342	231	208
Age (y)a	11.7 ± 3.3 12.0 (5.0–17.97)	12.4 ± 3.1 13.0 (6.0–17.7)	11.2 ± 3.4 10.5 (5.9–17.99)	11.7 ± 2.9 11.7 (5.5–17.8)
Raceb
NH White (%)	53	40	62	51
NH Black (%)	33	53	28	44
Hispanic (%)	7	5	3	2
Weight (kg)a	46.2 ± 18.0 45.6 (15.6–85.5)	86.7 ± 30.5 84.2 (21.9–199.9)	42.4 ± 17.7 35.3 (17.6–94.9)	89.1 ± 38.9 83.1 (26.9–225.6)
Height (cm)a	147.5 ± 17.3 150.8 (108.3–180.7)	154.8 ± 13.8 158.0 (101.0–182.6)	147.2 ± 20.6 143.3 (107.3–197.1)	155.1 ± 16.6 156.2 (117.7–191.5)
BMI (kg/m2)a	20.3 ± 4.2 19.6 (12.9–30.1)	35.2 ± 8.6 33.9 (20.7–66.9)	18.6 ± 2.8 18.0 (13.9–26.7)	35.5 ± 11.0 33.4 (19.3–88.1)
BMI z Scorea	0.5 ± 0.9 0.8 (−2.5–1.6)	2.4 ± 0.3 2.4 (1.7–3.1)	0.3 ± 0.7 0.4 (−2.0–1.6)	2.5 ± 0.4 2.6 (1.7–3.4)
Total Fat Mass (kg)a,c	13.7 ± 8.2 12.2 (1.7–36.7)	41.5 ± 17.4 38.0 (11.1–109.8)	7.6 ± 3.9 6.7 (1.2–26.5)	42.3 ± 21.3 38.7 (6.8–110.4)
≥75th Percentile (%)	15.7	95.9	2.1	94
Percentage Fat Mass (%)a,c	27.2 ± 8.2 26.9 (8.3–46.7)	44.9 ± 6.0 44.6 (26.8–65.3)	17.8 ± 6.7 17.2 (4.2–35.8)	43.6 ± 8.4 44.3 (16.4–64.5)
≥75th Percentile (%)	12.3	90.2	3.6	92
Breast Tanner Stage (%)d
1	30	13	95	43
2	12	17	4	39
3	19	16	1	12
4	16	13	0	1
5	23	41	0	5
Ovarian Volume (cc)a,e	2.3 ± 3.1 1.3 (0.2–16.0)	5.9 ± 4.1 5.7 (0.2–25.5)
Uterine Length (cm)a,f	3.7 ± 1.6 3.3 (2.2–8.0)	6.5 ± 1.7 7.0 (2.5–10.5)
Testes Stage (%)g
1 (1–3 cc)			52	37
2 (>3–6 cc)			11	22
3 (6–10 cc)			5	15
4 (10–15 cc)			7	10
5 (15–25 cc)			25	16
Pubic Hair Tanner Stage (%)h
1	34	18	54	40
2	12	12	8	21
3	12	13	9	12
4	19	16	12	11
5	23	41	17	16
Skeletal Age (y)a,i	10.2 ± 3.3 9.4 (5.0–18.0)	13.2 ± 3.2 13.2 (4.0–19.0)	10.4 ± 3.1 10.0 (6.0–19.0)	13.0 ± 2.9 13.0 (6.0–19.0)
Skeletal Age Advancement (y)a,i	0.2 ± 1.1 0.1 (−3.7–3.2)	1.2 ± 1.2 1.1 (−1.9–4.7)	0.5 ± 0.9 0.5 (−1.6–3.1)	1.4 ± 1.2 1.3 (−1.8–5.0)
Estradiol (pg/mL)a,j	37.8 ± 66.8 13.5 (3.5–493.0)	54.6 ± 49.5 45.4 (3.4–305.0)	15.0 ± 11.7 9.6 (3.5–53.4)	24.3 ± 18.2 20.0 (2.9–83.6)
Estrone (pg/mL)a,k	32.6 ± 27.3 23.4 (10.0–155.0)	33.0 ± 26.1 25.1 (4.7–130.0)	22.5 ± 18.9 17.0 (10.0–164.0)	20.9 ± 15.3 17.0 (9.0–109.0)
T (ng/dL)a,l	22.8 ± 13.6 20.0 (0.6–70.0)	32.0 ± 24.5 26.0 (3.7–148.0)	96.7 ± 180.2 16.0 (0.6–725.0)	83.0 ± 115.1 20.0 (0.6–612.0)
Androstenedione (ng/dL)a,m	68.9 ± 62.1 44.4 (6.5–404.0)	73.8 ± 54.1 59.0 (10.0–308.0)	47.1 ± 37.2 39.7 (10.0–220.0)	40.9 ± 24.6 32.9 (10.0–120.0)
DHEA-S (mcg/dL)a,n	73.7 ± 67.3 54.6 (5.0–365.0)	111.0 ± 78.6 90.0 (6.0–488.0)	81.1 ± 86.8 49.4 (4.0–632.0)	127.3 ± 87.2 110.0 (13.0–533.0)
HOMA-IRIa,o	1.8 ± 1.8 1.5 (0.2–23.4)	4.9 ± 3.1 4.1 (0.3–25.7)	1.2 ± 0.9 1.0 (0.3–5.2)	4.7 ± 3.4 3.9 (0.4–18.9)

NH, non-hispanic.

^aMean ± sd, Median (range).

^bAvailable for 622 girls (282 non-obese) and 436 boys (229 non-obese).

^cAvailable for 559 girls (261 non-obese) including 172 with DXA, 134 with air-displacement plethysmography, and 253 with both; and available for 369 boys (193 non-obese) including 76 with DXA, 123 with air-displacement plethysmography, and 170 with both.

^dAvailable for 609 girls (273 non-obese) and 402 boys (212 non-obese).

^eAvailable for 190 girls (32 non-obese).

^fAvailable for 185 girls (31 non-obese).

^gAvailable for 417 boys (219 non-obese).

^hAvailable for 591 girls (266 non-obese) and 422 boys (222 non-obese).

ⁱAvailable for 413 girls (142 non-obese of average age 10 y and 271 obese of average age 12.1 y) and 323 boys (143 non-obese of average age 9.9 y and 180 obese of average age 11.6 y).

^jAvailable for 348 girls (137 non-obese) and 274 boys (134 non-obese).

^kAvailable for 202 girls (130 non-obese) and 188 boys (125 non-obese).

^lAvailable for 412 girls (149 non-obese) and 312 boys (131 non-obese).

^mAvailable for 222 girls (124 non-obese) and 189 boys (111 non-obese).

ⁿAvailable for 400 girls (132 non-obese) and 311 boys (132 non-obese).

^oAvailable for 601 girls (273 non-obese) and 422 boys (218 non-obese).

Logistic regression for entry into puberty in girls and boys

Using logistic regression to study factors associated with entry into puberty, and adjusting for age and race, there was a significant interaction between sex and BMI z (P < .001) and between sex and percentage fat mass (P ≤ .01) for the prediction of entry into clinical puberty. Given that these results suggested a different association between measures of adiposity and puberty in each sex, we conducted the remaining analyses separately for girls and boys.

In sex-specific logistic regression models that were adjusted for age and race, BMI z (P < .001, OR = 1.86), total fat mass adjusted for height squared (P = .025, OR = 4.47), and percentage fat mass (P < .001, OR = 110.98) were each positively associated with significantly greater odds of being pubertal (Tanner II–V) vs. prepubertal (Tanner I) in girls. This result suggests that among girls of the same age and race, there is almost 2-fold greater odds of having entered puberty for every one point increase in BMI z. However, no significant association was seen among boys.

Linear models examining measures of adiposity as predictors of gonadal size

Once puberty begins, gonad size increases with age throughout adolescence. We used generalized linear models, adjusting for age and race, to examine measures of weight (BMI z score, total fat mass adjusted for height squared, and percentage fat mass) for associations with gonadal volume. BMI z scores, total fat mass, and percentage fat mass were not significantly associated with ovarian volume, whether expressed as age-adjusted z-score or as raw ovarian volume (Table 2A). In contrast, BMI z scores, total fat mass, and percentage fat mass were negatively associated with testicular volume (P's < 0.01, Table 2B). The average testicular volume in the group of boys considered obese was 7.5 cc (95% confidence interval (CI), 6.9–8.1) compared with 9.2 cc (95% CI, 8.6–9.8) in those boys considered nonobese (P < .001). Testicular volume was plotted against skeletal age, and regressions were determined for obese and nonobese boys (Figure 1). Beyond a skeletal age of 13, obese boys demonstrated significantly smaller testicular size compared with the nonobese boys of the same skeletal age, suggesting more delayed puberty despite similar skeletal maturation.

Table 2.

Associations Between Measures of Adiposity and Markers of Central Puberty, Estrogen Concentration, Estrogenization, Androgen Concentration, and Pubarche

	BMI z Score	Total Fat Mass	Percentage Fat Mass
A) Girls
Ovarian Volume	NS	NS	NS
Ovarian Volume z	NS	NS	NS
Breast Tanner Stage	β 0.41, P < .001	β 0.06, P < .001	β 0.03, P < .001
Uterine Length	β 0.59, P < .001	β 1.66, P = .001	β 3.38, P = .002
Uterine Length z	β 0.51, P < .001	β 1.51, P < .001	β 3.16, P = .001
Skeletal Age	β 0.35, P < .001	β 2.26, P < .001	β 0.03, P < .001
Skeletal Age Advancement	β 0.35, P < .001	β 0.72, P < .001	β 0.03, P < .001
Estradiol	NS	NS	NS
Estrone	NS	NS	NS
Androstenedione	NS	NS	NS
DHEA-S	NS	NS	NS
T	NS	NS	NS
Pubic Hair Tanner Stage	β 0.29, P < .001	NS	NS
B) Boys
Testicular Stage	β −0.44, P = .01	β −4.39, P < .001	β −0.11, P < .001
Breast Tanner Stage	β 0.75, P < .001	β 0.26, P < .001	β 0.11, P < .001
Skeletal Age	β 0.41, P < .001	β 0.10, P < .001	β 0.05, P < .001
Skeletal Age Advancement	β 0.40, P < .001	β 0.09, P < .001	β 0.04, P < .001
Estradiol	NS	β 6.67, P = .01	NS
Estrone	NS	NS	NS
Androstenedione	NS	NS	NS
DHEA-S	NS	NS	NS
T	NS	β −0.002, P < .001	β −0.002, P < .001
Pubic Hair Tanner Stage	β −0.1, P = .01	β −0.13, P < .001	β −0.05, P < .001

NS, no statistically significant association.

All analyses adjusted for age and race; fat mass analysis also adjusted for height².

Figure 1.

Relationship between testicular volume and skeletal age. Lines indicate sigmoidal curve fits. The P value indicates a statistically significant different relationship between skeletal age and testicular volume in obese boys vs. the relationship in nonobese boys after the age of 13 years.

Linear models examining IR as a predictor of gonadal size

We also evaluated the association of IR with gonadal size. IR was positively associated with ovarian volume, whether expressed as raw ovarian volume (β 0.02, P < .001) or age-adjusted z-score (β 0.27, P < .001). This relationship remained statistically significant even when data were adjusted for BMI z (β 0.02, P = .001 for ovarian volume and β 0.26, P = .001 for ovarian volume z-score) or percentage fat mass (β 0.03, P < .001 for ovarian volume and β 0.29, P < .001 for ovarian volume z-score). When adjusted for total fat mass, the relationship remained significant for raw ovarian volume (β 0.02, P = .01), but not for ovarian volume z-score. A higher degree of IR was associated with significantly less advanced testicular development (β −0.34, P < .001). The relationship between HOMA-IRI and testicular development remained significant when adjusted for BMI-z (β −0.08, P = .002) but not when adjusted for total or percentage fat mass.

ANCOVA models examining measures of adiposity as predictors of estrogen concentrations or markers of estrogenization

Estrogen exposure leads to skeletal maturation and breast development in both sexes, as well as uterine development in girls. Among girls, BMI z scores, total fat mass, and percentage fat mass were each found to have significant positive associations with breast development, uterine length (absolute or z score), skeletal age (Figure 2A), and skeletal age advancement, but did not show significant associations with estradiol or estrone concentrations (Table 2A). The mean values, adjusted for age and race, for uterine length, skeletal age, and skeletal age advancement, were higher in the group of girls considered obese compared with the girls considered nonobese (Table 3A). Similarly, body weight and adiposity in boys had significant positive associations with skeletal age (Figure 2B) and skeletal age advancement (Table 2B). In addition, body weight and adiposity were positively associated with boys' Tanner breast stage and the presence of gynecomastia (Table 2B). Total fat mass had a significant positive association with boys' serum estradiol concentrations (Table 2B); the relationships between BMI z scores or percentage fat mass and serum estradiol had nominal P values <.05 that did not remain statistically significant after adjustment for multiple comparisons. The mean values, adjusted for age and race, for skeletal age, and skeletal age advancement were higher in the group of boys considered obese compared with the boys considered nonobese (Table 3B).

Figure 2.

Relationship between BMI z score and skeletal age in girls A), and boys (B). Skeletal age was adjusted for age and race.

Table 3.

Adjusted Mean Values for Markers of Estrogenization Among Non-Obese and Obese Groups

	Non-Obese (BMI z < 1.66)	Obese (BMI z ≥ 1.66)	P Value
A) Girls
Uterine Length (cm)	5.3 (4.8–5.7)	6.2 (6.0–6.3)	.001
Skeletal Age (y)	11.6 (11.4–11.8)	12.5 (12.4–12.6)	<.001
Skeletal Age Advancement (y)	0.2 (0.0–0.4)	1.2 (1.0–1.3)	<.001
B) Boys
Skeletal Age (y)	11.4 (11.2–11.6)	12.3 (12.1–12.4)	<.001
Skeletal Age Advancement (y)	0.6 (0.4–0.8)	1.4 (1.2–1.6)	<.001

All analyses adjusted for age and race.

Adjusted means (95% CI) shown.

ANCOVA models examining measures of adiposity as predictors of androgen concentrations or pubic hair development

Pubic hair development is mediated by androgen exposure. Among girls, higher BMI z scores were significantly associated with a more advanced pubic hair stage (Table 2A). Additionally several of the associations between fat mass and pubic hair, DHEA-S, or T had nominal P values <.05 that did not reach statistical significance after adjustment for multiple comparisons. Androstenedione was not significantly associated with any weight measurement in girls. In boys, higher BMI z scores, total fat mass, and percentage fat mass were negatively correlated with pubic hair stage (Table 2B). Fat mass was associated with lower T concentrations (Table 2B); again, several other associations among androgen concentrations and adiposity did not reach statistical significance when adjusted for multiple comparisons.

Models examining IR as a predictor of androgen concentrations or pubic hair development

Among girls, IR was positively correlated with T concentrations (β 0.04, P < .001), and remained so even after adjusting for BMI z score (β 1.76, P < .001), total fat mass (β 0.92, P = .009), or percentage fat mass (β 1.01, P = .002), suggesting a correlation between IR and T independent of weight. However, IR was not significantly correlated with pubic hair or other androgen concentrations. In boys, there were no significant relationships between IR and androgen concentrations. There was a negative relationship between HOMA-IRI and pubic hair Tanner stage (β −0.040, P = .01), but this remained significant only when adjusting for BMI z score (β −0.22, P = .003), but not for total or percentage fat mass.

Discussion

Our results confirm that there are sexual dimorphisms in the relationship of adiposity with markers of pubertal development. We found that there was a negative relationship between measures of adiposity and testicular volume in boys, but no such relationship between adiposity and ovarian volume in girls. In association with higher degrees of adiposity, both sexes manifested increased signs of estrogenization, including skeletal age advancement, breast Tanner stage advancement, and, in girls, increased uterine length. However, BMI z showed opposing associations with pubic hair development, in the positive direction for girls but negative direction for boys, who also displayed a negative association between fat mass and T concentrations. IR, independent of adiposity, was also found to have sexually dimorphic associations. In girls, higher degrees of IR were associated with increased ovarian volume and T concentrations, even independent of BMI z or fat mass. In boys, also independent of BMI z, there was a negative association between HOMA-IR and both testicular volume and pubic hair Tanner stage.

In girls, the lack of association between adiposity and ovarian volume observed in this study implies that adiposity per se may not lead to significant early activation of the central hypothalamic-pituitary-gonadal pulse generator. Such a hypothesis is consistent with one prior study of pelvic ultrasonography in girls (83) and is also supported by studies demonstrating lower luteinizing hormone secretion in early pubertal obese girls compared with their normal weight counterparts despite earlier breast maturation (52, 84). We suggest that the advanced breast development observed among early pubertal obese girls may be primarily due to peripheral conversion of relatively inactive androgens to more bioactive estrogens by aromatase in adipose tissues (85–88), while the central axis remains relatively inactive. This speculation is not supported by our lack of association between adiposity and estrogen concentrations, although this may be due to limitations of the biochemical assays. However, as many other studies have demonstrated, obese girls do experience somewhat earlier menarche; thus, central puberty eventually is initiated and obese girls progress more rapidly to the onset of menses.

Obesity leads to an earlier increase in leptin concentrations that may permit the onset of central puberty (89), possibly through increased kisspeptin synthesis (90). We propose an additional reason for earlier menses in obese girls: We speculate that, as with other forms of premenarchal hyperandrogenism in girls such as congenital adrenal hyperplasia, the hypothalamic pulse generating unit may eventually be stimulated or derepressed by the elevated androgens produced by obese females. This androgen production may be induced by IR, as suggested by the association between HOMA-IRI and T, even independent of BMI z or fat mass. Such a relationship has been seen in other studies when examining IR and adrenarche (91). In addition, our data show a relationship between IR and ovarian volume, and prior studies suggest the ability of insulin to stimulate ovarian androgen production (92–94) and increase ovarian volume (95). After androgens allow for the onset of central puberty, the insulin-stimulated ovaries and the peripheral estrogen-exposed uterus and breast tissue may respond more robustly, allowing for a more rapid pubertal progression and earlier menarche. In other studies, metformin has slowed the progression of thelarche and onset of menarche in girls with premature adrenarche, suggesting that early insulin resistance and androgen production may be responsible for earlier progression to central puberty (96). Furthermore, the independence of the relationship between insulin and androgen production from adiposity may also account for some of the phenotypic variation in pubarche observed among obese girls.

In boys, excess fat mass seems to delay the onset of central puberty as evidenced by the smaller testicular volumes in obese boys. This effect is similarly believed to be caused by elevations in circulating estrogens produced by adipose tissue aromatase (35) as evidenced by signs of estrogen-mediated changes such as gynecomastia and skeletal advancement. However, as with the girls, the lack of association between adiposity and estrogen concentrations in our data did not lend support to this hypothesis, nor can our data prove that the source of the estrogens was adipose tissue. Increased estrogenic compounds, which we found were not accompanied by an increase in boys' adrenal or gonadal androgens, may inhibit gonadotropin production. As IR was not independently associated with testicular volume when adjusted for adiposity, we speculate that IR may mediate this differential effect of obesity on androgen production observed between boys and girls.

The present study is limited by its cross-sectional nature, which can only establish correlation, not causation. We did not perform interobserver correlations for our assessments of pubertal staging, although our prior study found excellent interobserver correlations for breast and pubic hair staging (97). Our study was also limited by the low sensitivity of some of the hormone assays employed. In particular, estradiol, estrone, and T were not measured with high-sensitivity assays, which may explain why we did not find significant associations between adiposity and these hormone concentrations. We also lacked data on high-sensitivity luteinizing hormone and follicle-stimulating hormone concentrations, which might have provided useful insights into the timing of central pubertal activation. The presence of true breast tissue vs. lipomastia may have been overdiagnosed due to the difficulty distinguishing between these tissue types in obese subjects, even with palpation. Finally, given the retrospective nature of this analysis, we were limited by the data available at the time of collection. We did not have measurements available for hormone concentrations or radiologic data on all subjects, or data on menses onset among girls, limiting the power to detect some of these associations. Future studies could evaluate these questions in prospective, longitudinal analyses. Strengths of the study include its large sample size and the careful measurements of pubertal development and of body composition. This study also adds to the overall understanding of the physiology of obesity and puberty by synthesizing data regarding measures of weight, fat mass, and IR with markers of androgen and estrogen production, and by using gonadal size to distinguish central pubertal activation from peripheral production; all of these elements have not been examined in one study of this nature before.

In sum, there are significant sexual dimorphisms in the signs of pubertal development seen in obese girls and boys. We propose that these differences can largely be explained by humoral changes caused by obesity: increased estrogens via aromatase activity and increased insulin resistance.