Longitudinal Study on Pubertal Insulin Resistance Free

Study subjects were part of an ongoing longitudinal study on body composition, energy expenditure, and risk factors for type 2 diabetes and cardiovascular disease in children and adolescents. Findings regarding body fat distribution, aerobic capacity, energy expenditure, insulin action, and dietary factors in this cohort have been previously reported (13,14,15,16). Children were recruited by newspaper and radio advertisements and by word of mouth. No child was taking medications known to affect body composition (e.g., ritalin or growth hormone), diagnosed with syndromes or diseases known to affect body composition or fat distribution (e.g., Cushing sydrome, Down syndrome, or type 1 diabetes), or diagnosed with any major illness since birth. Ethnicity was determined by self-report. This study was approved by the Institutional Review Board at the University of Alabama at Birmingham (UAB), and parents provided informed consent before testing commenced.

Children were admitted to the General Clinical Research Center (GCRC) in the late afternoon for an overnight visit. Upon the children’s arrival, anthropometric measurements were obtained. A computed tomography (CT) scan was conducted in the Department of Radiology at UAB at ∼1700 h. The children were served dinner and an evening snack, with all food consumed before 2000 h. All children were fed a fixed meal consisting of 55% carbohydrate, 15% protein, and 30% fat. Consumption of only water and noncaloric noncaffeinated beverages was permitted between 2000 h and testing the following morning. Two weeks after testing at the GCRC, children returned to the Department of Nutrition Sciences at UAB for body composition analysis by dual-energy X-ray absorptiometry (DEXA).

Tanner stage was determined by a pediatrician and was based on breast stage and pubic hair development in girls (17) and genitalia development in boys (18).

S_I, the acute insulin response (AIR), and the disposition index (DI) (product of S_I and AIR) were determined using a frequently sampled intravenous glucose tolerance test (FSIGTT) in the early morning after an overnight fast, as previously reported (15). Briefly, fasting blood samples were drawn, glucose was administered intravenously at time zero (25% dextrose; 11.4 g/m²), tolbutamide (125 mg/m²) was injected intravenously at 20 min postglucose, and 18 blood samples were collected over 3 h. Sera were analyzed for glucose (Ektachem DT II System; Johnson & Johnson Clinical Diagnostics) and insulin (radioimmunoassay [RIA]; Diagnostic Products, Los Angeles, CA), and values were entered into the MINMOD computer program (version 3.0, Richard N. Bergman) for determination of S_I, AIR, and DI, which is the product of S_I and AIR and serves as an index of β-cell function (19).

Whole-body composition (fat mass and fat-free mass) was measured by DEXA using a Lunar DPX-L densitometer (LUNAR Radiation, Madison, WI) as previously described (20). Subcutaneous abdominal fat and visceral fat were measured by CT scanning with a HiLight/Advantage Scanner (General Electric, Milwaukee, WI), as previously described (13).

Three samples of blood were collected over 40 min after an overnight fast, and sera were separated, pooled, stored at −85°C, and assayed for various hormones. Estradiol, testosterone, androstendione, and follicle-stimulating hormone (FSH) were measured as indexes of reproductive maturation. IGF-I was measured as an index of growth hormone action. Leptin was measured because it is closely related to body fat levels in children (21) and has been hypothesized to play a role in pubertal development (22). Cortisol was measured because this hormone is negatively associated with S_I (23). The details for each of the hormone assays are as follows:

• Estradiol (Diagnostic Products; double-antibody RIA): intra-assay coefficient of variation (CV) 3.6%; interassay CV 5.2%; and assay sensitivity 15.42 pmol/l.

• Total testosterone (Diagnostic Products; solid-phase RIA): intra-assay CV 2.7%; interassay CV 8.6%; and assay sensitivity 0.41 nmol/l.

• Androstendione (Diagnostic Systems Laboratories, Webster, TX; solid-phase RIA): intra-assay CV 11%; interassay CV 6.1%; and assay sensitivity 0.03 ng/ml.

• FSH (Diagnostic Products; immunoradiometric assay): intra-assay CV 4.47%; interassay CV 4.60%; and assay sensitivity ∼0.06 mIU/ml.

• Total IGF-I (Diagnostic Systems Laboratories; immunoradiometric assay): intra-assay CV 3.7%; interassay CV 7.3%; and assay sensitivity ∼2.06 ng/ml.

• Leptin (Linco Research, St. Charles, MO; double-antibody RIA): intra-assay CV 3.1%; interassay CV 6%; and assay sensitivity 0.4 ng/ml.

• Cortisol (Diagnostic Products; solid-phase RIA): intra-assay CV 5.2%; interassay CV 6.8%; and assay sensitivity 0.39 μg/dl.

Variables that were not normally distributed were log transformed before analysis. These variables included fasting insulin, S_I, AIR, and hormone levels. For ease of interpretation, data are presented in the measured untransformed scale. Simple, unadjusted within-subject changes were assessed by paired t test. Repeated-measures analysis of variance was used to examine the influence of time on changes on each of the dependent variables (e.g., S_I) after controlling for the between-subject effects of sex, ethnicity, and puberty transition group (i.e., remaining at Tanner stage I or progressing to Tanner stage III) as well as the covariates of age and body fat at baseline. Interaction effects were examined to determine whether the pattern of time-related changes was similar across puberty transition groups, sex, and ethnicity. Pearson’s correlation coefficients were used to examine the association between changes in the main outcome variables (e.g., change in S_I) and the hypothesized factors (e.g., change in body fat mass and change in hormone level). All analyses were conducted using SPSS version 9.0, and data are presented as the means ± SD.

Data from 60 children were used in this analysis. The group was evenly split between the sexes (33 male and 27 female subjects) and ethnic groups (32 Caucasians and 28 African-Americans). All children were at Tanner stage I at their original visit and had a mean age of 9.2 ± 1.3 years. After a mean follow-up period of 2.0 ± 0.6 years, 31 of these children had progressed to Tanner stage III or IV, whereas 29 remained at Tanner stage I. The children who progressed to Tanner stage III or IV were older, with greater body fat than those remaining at Tanner stage I (Table 1).

The children who progressed from Tanner stage I to Tanner stage III or IV were aged 10.0 ± 1.1 years at baseline and 12.3 ± 1.1 at follow-up (Table 1). As expected, both fat mass and fat-free mass were significantly higher by Tanner stage III, with fat-free mass increasing by 37% and fat mass increasing by 44%. Fasting glucose and fasting insulin were significantly higher at Tanner stage III or IV compared with Tanner stage I (Table 1). S_I fell significantly by 32%, AIR increased significantly by 30%, DI fell significantly by 30% (Table 1), and glucose effectiveness remained constant (data not shown).

To separate the effects of changing Tanner stage from age-related changes, we compared the changes noted above with a group of children who remained at Tanner stage I over a similar follow-up period (Table 1). These children were younger, with a mean age of 8.4 ± 1.1 years at baseline and 10.0 ± 1.1 years at follow-up. As expected, both fat and fat-free mass also increased, but at a lower rate (14% increase in fat-free mass and 34% increase in fat mass). Fasting glucose and fasting insulin remained unchanged at follow-up, S_I increased slightly but significantly (by 15%), and AIR, DI, and glucose effectiveness did not change significantly over time.

To fully maximize the power of this study and examine the data collectively, we conducted a repeated-measures analysis of variance to examine the within-subject effects and compare the change over time in those children progressing to Tanner stage III or IV, versus those children remaining at Tanner stage I. As shown in Table 1, there was a significant time*group interaction effect for fasting insulin, S_I, and DI, indicating that the pattern of change in these variables was significantly different in the two groups of children. There were no significant interaction effects between time and sex or ethnicity in these models, indicating that the degree of change was similar across sex and ethnicity. In addition, these effects remained significant after controlling for age, fat-free mass, and body fat content. The group mean values for S_I across time in the sex and ethnic groups are shown in Fig. 1. Since the two groups of children varied in age, we examined changes in subgroups matched for age at baseline and at follow-up. Similar findings were observed in these age-matched subgroups. In fact, in this analysis the fall in S_I in those progressing to Tanner stage III or IV was slightly higher (39% fall), with a 53% increase in AIR, and no significant changes in either variable in the children remaining at Tanner stage I (data not shown).

To examine the effect of body fat on pubertal changes in S_I, we examined pubertal changes in children who fell in the lower, middle, and upper tertiles for body fat percentage at baseline among those who progressed to Tanner stage III (Table 2). The three tertiles were similar for age (9.8 ± 1.4, 10.0 ± 1.1, and 10.1 ± 0.8 years, respectively), but, varied greatly for percentage of body fat (17.8 ± 6.7, 28.7 ± 3.3, and 41.3 ± 5.0% fat at baseline, respectively). In repeated-measures analysis of variance, there was a significant within-subject effect of time (P = 0.001) on S_I that was consistent across the fat groups (P value for time*fat-group interaction = 0.38), as well as a significant between-subject effect of fat (P < 0.001). The reduction in S_I over time was 33% in the low-fat group, 30% in the medium-fat group, and 36% in the high-fat group. Results were similar when examining pubertal changes across fat groups for AIR, DI, and fasting insulin and glucose, as summarized in Table 2.

Changes in various hormone concentrations over time in those children progressing from Tanner stage I to Tanner stage III or IV versus those remaining at Tanner stage I are shown in Table 3. In those children remaining at Tanner stage I, there were no significant changes in estradiol, testosterone, cortisol, or leptin, but there were significant increases in FSH (70% increase) and a small but significant increase in IGF-I (13% increase). In those children progressing to Tanner stage III or IV, there were no significant changes in cortisol or leptin, but there were large increases in estradiol and testosterone (both undetectable at baseline), a doubling of androstendione, a threefold increase in FSH, and a 1.7-fold increase in IGF-I. Among those children who progressed from Tanner stage I to Tanner stage III or IV, change in S_I was not significantly associated with change in fat mass, change in subcutaneous or abdominal adipose tissue, or any of the hormone levels, in the group as a whole or when analyzed in sex and ethnic subgroups. Change in the AIR was not associated with change in any other variable except the change in androstendione (r = 0.39; P = 0.04; n = 29). This relationship was evident but not significant when examined separately in boys (r = 0.44; P = 0.11; n = 14) or girls (r = 0.43; P = 0.11; n = 15).

The phenomenon of pubertal insulin resistance has been well described in numerous studies (1,2,3,4,5,6,7), but there have been no longitudinal studies that have examined this change in detail. It is therefore unclear whether pubertal insulin resistance is related to compensatory changes in insulin secretion and is associated with pubertal changes in body fat and/or hormone levels. We used a longitudinal study design to compare within-subject changes over time, and our main findings were that pubertal transition from Tanner stage I to Tanner stage III or IV was associated with a 32% reduction in S_I and a disproportionately low increase in the AIR. These changes were consistent across sex, ethnicity, and obesity and were not apparent in growing children who did not experience a change in Tanner stage. The disproportionately low increase in the AIR observed in response to the decline in S_I suggests a conservation in β-cell function or an inadequate β-cell response. Finally, the fall in S_I was not associated with changes in body fat, IGF-I, or sex steroid hormone levels.

The present findings regarding pubertal changes in S_I generally agree with those of previous cross-sectional observations. The largest cross-sectional study was reported in data from 357 children (10–14 years of age) in whom insulin action was assessed by the euglycemic clamp technique (3). In this study, S_I began to fall by Tanner stage II, reached a nadir at Tanner stage III, and then recovered to near prepubertal levels by Tanner stage V (3).

We did not have substantial numbers of children within each of the four sex by ethnic subgroups to thoroughly examine the influence of sex and ethnicity on the magnitude of changes in S_I or AIR. In the repeated-measures analysis of variance, we did not detect any significant interactions between sex and time or ethnicity and time, other than the interaction of time*group*ethnicity (Table 1). This finding indicates that the fall in S_I may have been stronger in blacks than in whites. This was also evident in the simple paired t test analysis, in which the fall in S_I in children progressing to Tanner stage III was significant in African-Americans (4.0 ± 2.4 to 2.6 ± 1.7; P = 0.001; n = 18) but not in Caucasians (5.0 ± 3.7 to 3.4 ± 1.6; P = 0.16; n = 13). However, the lack of effect in Caucasians may be due to the smaller sample (n = 13) and greater variability in S_I.

The pubertal fall in S_I was accompanied by an increase in the AIR. However, based on the hyperbolic relationship between S_I and AIR (24), the increase in AIR (30% overall) was lower than expected to match the overall 32% reduction in S_I. This is reflected in the overall 30% fall in DI. According to the hyperbolic relationship, AIR would have to be increased by 50% to fully compensate for the fall in S_I. Such an increase would maintain a constant DI and reflect a healthy response at the level of the β-cell. The accompanying rise in AIR is consistent with previous findings (7), but the finding of a lower-than-expected increase is a new observation. This finding potentially indicates either that there is a conservation of β-cell activity or that the β-cell is failing.

Substantial indirect evidence suggests that growth hormone may play a key role in pubertal insulin resistance. This hypothesis is attractive because growth hormone and IGF-I are transiently higher in mid-puberty, mirroring the changes in S_I, and S_I is correlated with growth hormone and IGF-I (5). In addition, growth hormone–deficient children have increased S_I (25), and growth hormone has strong effects on β-cells (26). Also, the increase in growth hormone during puberty may contribute to insulin resistance via its effect on increasing lipolysis and free fatty acid concentration.

However, not all studies have found supporting evidence for the growth hormone theory. One study that performed detailed measures of growth hormone (9) found no significant difference across pubertal groups in overnight growth hormone secretion, peripheral growth hormone responsiveness (as indicated by growth hormone–binding protein), or growth hormone action (as indicated by circulating IGF-I). However, it may not be appropriate to base conclusions regarding the influence of growth hormone on pubertal S_I on these results, because no pubertal change in S_I was observed in this study. In our study, although there was a large increase in circulating IGF-I during puberty, this was not correlated with the fall in S_I. Since we only had a fasting plasma sample, we were unable to thoroughly evaluate the influence of growth hormone beyond an examination of IGF-I as a surrogate marker of growth hormone action. Clearly, there is a strong need to examine more carefully and thoroughly the relationship between changes in growth hormone action and insulin action and secretion during puberty.

Although changes in sex hormones may contribute to pubertal insulin resistance, this theory has not been previously examined in detail. In general, changes in sex steroids are not thought to drive changes in pubertal insulin resistance because sex steroids increase in early puberty but remain high, whereas S_I returns to normal levels by the end of puberty. Travers et al. (6) compared S_I in 50 Tanner stage II boys and girls with 47 Tanner stage III boys and girls. The effect of Tanner stage on S_I was significant only in girls, and neither testosterone nor estradiol was correlated with S_I. However, no other hormones were examined. In addition, this study is limited because it compared only Tanner stage II with Tanner stage III, and significant changes may have already occurred by Tanner stage II. In another study, 4 months of testosterone administration to adolescents with delayed puberty led to an increase in fat-free mass (through reductions in protein breakdown and protein oxidation), an increase in circulating testosterone (23 ± 4 to 422 ± 45 ng/dl), IGF-I (210 ± 28 to 505 ± 40 ng/ml), and mean nocturnal growth hormone (2.5 ± 0.5 to 6.0 ± 0.8 ng/ml), but had no effect on S_I (11). The lack of effect of testosterone administration on S_I in adolescents is consistent with other studies in adults (27).

The longitudinal data presented in this study are consistent with previous studies in that we did not find any correlation between the change in S_I and changes in circulating hormone levels. The only significant relationship we observed was between the change in AIR and the change in androstendione (the relationship did not reach statistical significance when analyzed separately in boys and girls, although the trend remained the same in both sexes). Androstendione is the biosynthetic precursor of testosterone and estradiol, and it is the predominant adrenal androgen in mid-puberty in both boys and girls. This hormone has been hypothesized to contribute to bone and muscle development during growth (28), but it may not be effective in adults since exogenous administration to young men had no anabolic effects on muscle protein metabolism (29). Circulating levels of androstendione gradually increase during pubertal development, with higher levels in girls than boys; the increase in androstendione in the first half of puberty is thought to derive predominantly from the adrenal glands, with a shift to production from the gonads by the end of puberty (30). Further studies in larger samples and with more detailed hormone measures are necessary to more fully examine the role of sex hormones on insulin resistance and insulin secretion during puberty. On the basis of the current data and power analysis, we estimated that sample sizes of 46 children per subgroup would be needed to detect a significant correlation of 0.4 (similar to what we observed for the correlation between change in S_I and change in androstendione) with a power of 0.8. In addition, more extended longitudinal observations in the current cohort should also be useful to reveal which factors are associated with the recovery of insulin resistance by the end of puberty.

Our results have several implications for the prevention of type 2 diabetes during puberty. Since insulin resistance is likely to play an integral role in healthy somatic growth, we do not think that it is prudent to attempt to prevent the pubertal decline in S_I. When a more detailed mechanism is identified, it may be important to identify interventions that ensure that S_I recovers during later puberty and that β-cell function remains adequate to support this period of insulin resistance. More importantly, dietary and physical activity interventions should be explored for decreasing body fat, increasing S_I, and sustaining β-cell function before and during pubertal development, especially in those subjects with very low levels of S_I. In the present study, early pubertal development was evident in fatter, age-matched children (Table 2). Early puberty is also a known risk factor for breast cancer (31) and could be hypothesized to contribute to risk for type 2 diabetes if the pancreas is unable to compensate for the pubertal decline in S_I. Thus, early puberty and a longer period of maturation may be additional risk factors for the development of type 2 diabetes during puberty, and further studies are needed to examine these issues.

In summary, the pubertal transition from Tanner stage I to Tanner stage III or IV was associated with a 32% reduction in S_I and a lower-than-expected increase in AIR. These changes were consistent across sex, ethnicity, and obesity subgroups and were not observed in growing children remaining at Tanner stage I. The lower-than-expected rise in AIR and the significant fall in DI suggest a conservation in β-cell function or an inadequate β-cell response to the fall in S_I. The fall in S_I was not associated with changes in body fat, visceral fat, IGF-I, testosterone, or estradiol, but the increase in AIR was positively correlated with the pubertal increase in androstendione concentration.

This study was supported by the National Institute of Child Health and Development (to M.I.G.: grants R29 HD 32668 and R01 HD/HL 33064), the National Institute of Aging (to B.A.G.: grant K01AG00740), and by GCRC Grant M01-RR-00032.

We are extremely grateful to Tena Hilario, who coordinated this project; the staff of the GCRC; and the children and their families who participated in this study.