In adults, lower circulating levels of the adipocyte-derived hormone adiponectin are associated with obesity, type 2 diabetes, and cardiovascular disease risks. Its use as a risk marker in children is less clear. In 839 children aged 8 years from a representative birth cohort, circulating adiponectin levels were associated with body weight, BMI, waist circumference, and fasting and 30-min insulin levels, but the associations were opposite in boys, with positive associations, and girls, with inverse associations (P = 0.008–0.00001 for interaction with sex). Girls had overall higher adiponectin, higher total cholesterol, lower HDL cholesterol, and higher triglyceride levels than boys, even after adjustment for BMI. With increasing BMI, girls showed steeper declines in HDL cholesterol (P = 0.01 for interaction) and adiponectin levels (P = 0.0005 for interaction) and a steeper increase in triglyceride levels (P = 0.009 for interaction) compared with boys. In conclusion, plasma adiponectin is not a simple marker of central fat and insulin sensitivity in children. With increasing BMI, decreasing adiponectin levels in girls could contribute to their faster deterioration in lipid profiles in comparison with boys. Our data suggest a complex age- and sex-related regulation of adiponectin secretion or clearance.
Adiponectin is an adipokine with potent insulin-sensitizing properties in mice and humans (1,2). In adult populations, lower circulating adiponectin levels are associated with insulin resistance and lower HDL cholesterol levels and predict type 2 diabetes and cardiovascular disease risks (3,4). However, the strong inverse relationship observed between circulating adiponectin levels and measures of both body fat mass and insulin resistance is counterintuitive, considering that it derives exclusively from fat cells (5). In contrast, most other adipokines, such as leptin and visfatin, show positive relationships with adiposity (6,7).
The regulation of circulating adiponectin levels is complex. It is secreted from fat cells in one of three main forms: a low–molecular weight 67-kDa hexamer, a medium–molecular weight 136-kDa hexamer, and a high–molecular weight >300-kDa hexamer complex (8). Current assays detect all three molecular forms of circulating adiponectin but do not yet distinguish between each specific form. A number of hormones may influence adiponectin secretion (6), including variable effects of insulin on adiponectin mRNA expression and protein levels (9–11).
In some childhood populations, as in adults, lower adiponectin levels have been proposed as a marker of obesity and risk of developing type 2 diabetes (12,13). However, in the newborn, adiponectin levels are around twofold higher and show positive associations with birth weight and BMI (14–16). Furthermore, in newborns, adiponectin levels increase with higher leptin levels and gestational age, suggesting a positive relationship with the development of fetal adiposity (15,17). Adiponectin levels decrease during early childhood, and this is related to the rate of postnatal weight gain (18,19). A reversal in the direction of association (from positive to negative) between adiponectin and adiposity must therefore occur at some time during childhood (14). To assess the direction of associations between adiponectin levels and adiposity, insulin sensitivity, and blood lipid levels in children, we measured its circulating levels in a well-characterized representative birth cohort at age 8 years.
RESEARCH DESIGN AND METHODS
The Avon Longitudinal Study of Parents and Children (ALSPAC) is a prospective study of 14,541 pregnancies recruited from all pregnancies in three Bristol-based District Health Authorities with expected dates of delivery between April 1991 and December 1992 (http://www.alspac.bris.ac.uk). All children were measured at birth and at age 7 years; details of measurements have been previously described (20,21). The new data in this report relate to fasting adiponectin levels and lipid profiles at age 8 years.
Blood samples and anthropometry at age 8 years.
A total of 885 unselected full-term, singleton birth, 8-year-old ALSPAC children attended a substudy of fasting and 30-min postoral glucose (1.75 g/kg, max. 75 g) blood sampling and measurements of body weight, height, and waist circumference, as previously described (22). Insulin and glucose levels were measured on fasting and 30-min venous blood samples. We measured plasma adiponectin levels in fasting plasma samples at this age 8–year visit in 839 children (449 males). These children did not differ from other ALSPAC children with regard to size at birth or at age 7 years; compared with a current U.K. growth reference (23), they had a mean ± SD birth weight SD score of 0.01 ± 1.04 and a BMI SD score at 7 years of age of 0.17 ± 1.04. Only 16 children were nonwhite (1.9%) and had no differences in adiponectin, BMI, or insulin sensitivity compared with the other children.
Assays.
Plasma adiponectin was determined by an in-house time-resolved immunofluorometric assay as previously described (24). Briefly, the assay is based on commercially available antibodies and recombinant human adiponectin from R&D Systems (Abingdon, U.K.) and detects several molecular weight forms of adiponectin, including the three major molecular isoforms. All samples were analyzed in duplicate in a final dilution of 1:200. The detection limit (nonspecific binding +3 SD) was estimated to <1.5 μg/l, and assay standards ranged from 2 to 500 μg/l. Within-assay coefficients of variation (CVs) of standards and unknown samples averaged <5%. In between–assay CVs were estimated by repetitive analysis of a control sample diluted 1:2,500, 1:500, and 1:50. After 125 setups, in between–assay CVs averaged 10.3% at 0.36 μg/l (final dilution 1:2,500), 6.2% at 1.94 μg/l (final dilution 1:500), and 3.6% at 21.0 μg/l (final dilution 1:50). The recovery of exogenously added adiponectin to serum was 101 ± 1% (mean ± SEM based on 10 samples).
Triglyceride and total and HDL cholesterol levels were measured on a Dimension RXL system (Dade Behring) using reagents and calibrants supplied by the manufacturer; inter- and intra-assay CVs were <4%. Assays for insulin and glucose levels have been previously reported (22).
Calculations.
BMI was calculated as weight divided by the square of height in meters. Insulin sensitivity estimated from fasting insulin and glucose levels using the homeostasis model (HOMA-CIGMA Calculator Programme v2.00) (22) showed extremely high correlation with fasting insulin levels (R2 = 95.7%).
Statistics and ethics.
Insulin and triglyceride levels, body weight, BMI, and waist circumference data were log transformed to normal distributions to allow use of parametric analyses. Univariate correlations were tested by Pearson’s test, and correlation coefficients (R) are presented. ANCOVA (general linear models) was used to test interactions with sex on the relationship between adiponectin, lipid levels, and body size and other hormone levels. Analyses were performed using SPSS for Windows (SPSS, Chicago, IL). P values <0.05 were taken as significant. Ethical approval was obtained from the ALSPAC and the local research ethics committees. Signed consent was obtained from a parent or guardian, and verbal assent was obtained from the child.
RESULTS
Sex differences in metabolic variables.
Body size and metabolic variables at age 8 years are summarized by sex in Table 1. Girls had higher adiponectin, lower HDL cholesterol, higher total cholesterol, and higher triglyceride levels than boys (Table 1). These differences persisted even after adjustment for weight and height or BMI. Furthermore, compared with boys, girls, with increasing BMI, showed steeper declines in HDL cholesterol (P = 0.01 for interaction with sex) and adiponectin levels (P = 0.0005 for interaction) and a steeper rise in triglyceride levels (P = 0.009 for interaction) (Fig. 1). Differences in rates of change in adiponectin, HDL cholesterol, and triglycerides between boys and girls were also seen with increasing waist circumference (data not shown).
Fasting insulin levels were higher in girls than in boys (Table 1) but showed a similar rate of increase with BMI in both sexes (P = 0.4 for interaction with sex).
Sex-discordant associations with adiponectin.
Overall (in boys and girls combined), adiponectin levels at age 8 years were largely unrelated to current body size (Table 2). However, significant associations were seen in boys and girls separately and in opposite directions (P = 0.008–0.00009 for interaction with sex; Table 2). Similarly, adiponectin levels showed significant associations with fasting insulin and 30-min insulin levels in boys and girls separately and again in opposite directions (Table 2). Adiponectin levels were unrelated to fasting or 30-min glucose levels in either boys or girls (P = 0.1–0.8, not shown).
Adiponectin associations with HDL cholesterol.
Adiponectin levels were positively and independently associated with HDL cholesterol levels (correlation coefficient R = 0.11, P = 0.0008, adjusted for BMI, height, and fasting insulin levels). In contrast to the associations with current body size and insulin levels, when analyzed separately in each sex, the adiponectin associations with HDL cholesterol were positive in both boys (adjusted R = 0.10, P = 0.03) and girls (adjusted R = 0.12, P = 0.02) (P = 0.8 for interaction with sex).
DISCUSSION
In this cohort of representative 8-year-old U.K. children, plasma adiponectin levels declined with increasing BMI, waist circumference, and fasting and stimulated insulin levels in girls but not in boys. In contrast, adiponectin levels were positively related to body weight and insulin levels in boys.
In adults, circulating adiponectin levels show strong consistent inverse associations with central fat and insulin resistance (25–28) and predict future risk of type 2 diabetes and cardiovascular disease (29,30). Among children, similar inverse associations with BMI and insulin resistance have been reported (12,31,32). However, most of the children in the previous studies were more overweight or older than our population, and few studies explored associations in boys and girls separately (see online appendix supplementary data 1 [available at http://diabetes.diabetesjournals.org]). One study of 500 Taiwanese schoolchildren aged 6–18 years reported inverse associations between adiponectin and BMI or insulin levels only in girls and in boys older than age 15 years (33). However, that study was cross-sectional, and in the small number of younger boys (n = 35), the apparent positive association between adiponectin and fasting insulin levels was not significant (33). Together with the recent reports of high adiponectin levels and positive associations with body weight in newborns (14,34), our current data question the simple relationship between adiponectin levels and adiposity and insulin sensitivity among young children.
These findings suggest a further complexity in the hormonal regulation of adiponectin secretion, degradation, or clearance (6). Consistent effects of sex hormones on adiponectin are reported (35). The inhibitory effects of androgens and the stimulatory effects of estrogens on adiponectin secretion would explain the higher adiponectin levels in girls than boys. Increased adipocyte aromatase activity leading to higher estrogen levels could also possibly explain the higher adiponectin levels in obese boys but not the lower adiponectin levels in obese girls and obese adults.
Conflicting effects of insulin on adiponectin secretion are reported. In vitro studies on human visceral adipose tissue and mouse brown adipocytes report that insulin stimulates adiponectin gene expression (10,36). This could explain the positive association between both fasting and 30-min insulin levels with adiponectin in boys and also the higher adiponectin levels in adults with type 1 diabetes (37). However, in contrast to those studies, other in vivo studies show that adiponectin levels were suppressed by 20% in humans and by 50% in rats with a hyperinsulinemic- euglycemic clamp (11), and these findings are supported by in vitro studies of mouse 3T3-L1 adipocytes (9).
In vitro data suggest that adiponectin is produced mainly by visceral adipose tissue (38). Girls have a larger central fat mass than boys, despite similar waist circumference, and this could explain their relative insulin resistance in comparison with boys (39). During puberty, boys rapidly gain intra-abdominal fat (40); subsequently, adult men have similar or higher levels of intra-abdominal fat and insulin resistance than women (28,41). Differences in the timing of accumulation of intra-abdominal or intrahepatic fat could explain these age-related differences in insulin sensitivity, and also lipid levels, between males and females. Further analyses in our population suggest that adiponectin levels may indeed start to fall in those boys with the highest levels of central fat (see online appendix supplementary data 2). We hypothesize that the reversal in the direction of adiponectin associations with age (Fig. 2) might therefore reflect the accumulation of central fat, possibly due to the inhibitory actions of other adipokines on adiponectin levels (42).
Our cross-sectional and observational study design limits the ability to infer causal links. Further follow-up during puberty would confirm these findings and demonstrate the timing of the change in direction of adiponectin associations in boys. Furthermore, our adiponectin assay only detected total circulating adiponectin levels, and it is possible that the proportions of the main adiponectin complexes could differ between boys and girls; however, specific assays for use in epidemiological studies are not yet available.
In contrast to the sex-discordant associations between adiponectin and insulin resistance, adiponectin levels were positively associated with HDL cholesterol levels in both boys and girls. These associations were independent of BMI and fasting insulin levels and are consistent with findings in adults (3,28). Adiponectin activation of peroxisome proliferator–activated receptor-α (2) could explain a direct link between adiponectin and higher HDL cholesterol levels, and that pathway is independent of the adiponectin activation of AMP-activated protein kinase, which leads to insulin sensitization. With increasing BMI or waist circumference, girls showed a significantly steeper fall in HDL cholesterol levels than boys. It is possible that the decline in adiponectin levels with BMI, only seen in girls, could contribute to this sex difference.
In conclusion, while plasma adiponectin levels showed consistent associations with HDL cholesterol in boys and girls, it is not a simple marker of insulin sensitivity and adiposity in young children. The data need to be confirmed by a large longitudinal study; however, identification of putative inhibitors of adiponectin secretion that emerge during childhood could reveal potential targets to prevent adult metabolic disease.
Plasma adiponectin, HDL cholesterol, and triglyceride levels related to BMI in boys (□) and girls (▪) at age 8 years. A BMI of 16 kg/m2 is equivalent to the 50th centile, and a BMI of 20 kg/m2 is roughly equivalent to the 90th centile. Means (geometric means for triglycerides) + 95% CI are displayed.
Plasma adiponectin, HDL cholesterol, and triglyceride levels related to BMI in boys (□) and girls (▪) at age 8 years. A BMI of 16 kg/m2 is equivalent to the 50th centile, and a BMI of 20 kg/m2 is roughly equivalent to the 90th centile. Means (geometric means for triglycerides) + 95% CI are displayed.
![FIG. 2. Schematic diagram showing the changing relationships between adiponectin levels and body size at birth in both sexes (based on Kamoda et al. [34]), at age 8 years in boys (□) and girls (▪) (based on the current data), and in adults of both sexes (based on Arita et al. [5]).](/view-large/figure/4239940/zdb0050643190002.jpeg)
Schematic diagram showing the changing relationships between adiponectin levels and body size at birth in both sexes (based on Kamoda et al. [34]), at age 8 years in boys (□) and girls (▪) (based on the current data), and in adults of both sexes (based on Arita et al. [5]).
Schematic diagram showing the changing relationships between adiponectin levels and body size at birth in both sexes (based on Kamoda et al. [34]), at age 8 years in boys (□) and girls (▪) (based on the current data), and in adults of both sexes (based on Arita et al. [5]).
Hormone levels and body size at age 8 years
| Boys | Girls | P* | |
|---|---|---|---|
| Age 8 years visit (n) | 449 | 390 | |
| Age (years) | 8.16 (8.12–8.24) | 8.16 (8.12–8.24) | NS |
| Weight (kg) | 28 (25–31) | 28 (25–32) | NS |
| Height (cm) | 131 (127–134) | 130 (126–134) | 0.008 |
| BMI (kg/m2) | 16.1 (15.0–17.2) | 16.5 (15.3–18.1) | 0.001 |
| Waist circumference (cm) | 56 (54–60) | 56 (53–60) | NS |
| Adiponectin (mg/l) | 13.0 (11.1–16.0) | 14.0 (11.4–17.6) | 0.0007† |
| HDL cholesterol (mmol/l) | 1.56 (1.34–1.76) | 1.47 (1.25–1.69) | <0.0001† |
| Total cholesterol (mmol/l) | 3.8 (3.4–4.3) | 4.0 (3.5–4.4) | 0.01† |
| Triglycerides (mmol/l) | 0.67 (0.6–0.8) | 0.76 (0.6–0.9) | <0.0001† |
| Insulin sensitivity (HOMA) | 211 (122–301) | 166 (102–252) | 0.005† |
| Fasting insulin (mU/l) | 4.2 (2.9–7.1) | 5.3 (3.5–8.5) | 0.005† |
| 30-min insulin (mU)‡ | 37.7 (26.7–54.5) | 48.9 (34.4–70.2) | <0.0001† |
| Boys | Girls | P* | |
|---|---|---|---|
| Age 8 years visit (n) | 449 | 390 | |
| Age (years) | 8.16 (8.12–8.24) | 8.16 (8.12–8.24) | NS |
| Weight (kg) | 28 (25–31) | 28 (25–32) | NS |
| Height (cm) | 131 (127–134) | 130 (126–134) | 0.008 |
| BMI (kg/m2) | 16.1 (15.0–17.2) | 16.5 (15.3–18.1) | 0.001 |
| Waist circumference (cm) | 56 (54–60) | 56 (53–60) | NS |
| Adiponectin (mg/l) | 13.0 (11.1–16.0) | 14.0 (11.4–17.6) | 0.0007† |
| HDL cholesterol (mmol/l) | 1.56 (1.34–1.76) | 1.47 (1.25–1.69) | <0.0001† |
| Total cholesterol (mmol/l) | 3.8 (3.4–4.3) | 4.0 (3.5–4.4) | 0.01† |
| Triglycerides (mmol/l) | 0.67 (0.6–0.8) | 0.76 (0.6–0.9) | <0.0001† |
| Insulin sensitivity (HOMA) | 211 (122–301) | 166 (102–252) | 0.005† |
| Fasting insulin (mU/l) | 4.2 (2.9–7.1) | 5.3 (3.5–8.5) | 0.005† |
| 30-min insulin (mU)‡ | 37.7 (26.7–54.5) | 48.9 (34.4–70.2) | <0.0001† |
Data are median (interquartile range).
Sex difference;
adjusted for BMI at 8 years;
30-min post–oral glucose load. HOMA, homeostasis model assessment; NS, not significant (P > 0.05).
Discordant associations with adiponectin levels at age 8 years in boys and girls
| Determinant | Both sexes | Boys | Girls | P* |
|---|---|---|---|---|
| Weight | −0.03 | 0.09† | −0.18‡ | 0.00009 |
| Height | 0.00 | 0.07 | −0.11† | 0.008 |
| BMI | −0.04 | 0.08 | −0.16‡ | 0.0005 |
| Waist circumference | −0.07† | 0.06 | −0.19‡ | 0.0002 |
| Insulin sensitivity | 0.00 | 0.14‡ | −0.15‡ | 0.00003 |
| Fasting insulin | 0.00 | 0.12† | −0.18‡ | 0.00001 |
| 30-min insulin | 0.00 | 0.17‡ | −0.13† | 0.00004 |
| Determinant | Both sexes | Boys | Girls | P* |
|---|---|---|---|---|
| Weight | −0.03 | 0.09† | −0.18‡ | 0.00009 |
| Height | 0.00 | 0.07 | −0.11† | 0.008 |
| BMI | −0.04 | 0.08 | −0.16‡ | 0.0005 |
| Waist circumference | −0.07† | 0.06 | −0.19‡ | 0.0002 |
| Insulin sensitivity | 0.00 | 0.14‡ | −0.15‡ | 0.00003 |
| Fasting insulin | 0.00 | 0.12† | −0.18‡ | 0.00001 |
| 30-min insulin | 0.00 | 0.17‡ | −0.13† | 0.00004 |
Data are correlation coefficients (R).
For interaction with sex.
P < 0.05,
P < 0.005.
Additional information for this article can be found in an online appendix at http://diabetes.diabetesjournals.org.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Article Information
ALSPAC is supported by the Medical Research Council, the Wellcome Trust, the Department of Health, the Department of the Environment, the European Commission, and many others. D.B.D. is supported by the Wellcome Trust and the Juvenile Diabetes Research Foundation. A.F. is supported by the Danish Medical Research Council and the Danish Diabetes Association. J.F. is supported by the Danish Health Research Council.
We thank Hanne Petersen for skilled technical assistance. We are extremely grateful to all of the children and parents who took part in both studies and to the midwives for their cooperation and help in recruitment.
