Decrease in Serum Adiponectin Level Due to Obesity and Visceral Fat Accumulation in Children


Department of Pediatrics, School of Medicine, University of Occupational and Environmental Health, Japan, 1-1 Iseigaoka, Yahatanishi-ku, Kitakyushu 807-8555, Japan. E-mail:


Objective: To determine whether serum adiponectin is decreased in obesity and is restored toward normal level after treatment in children.

Research Methods and Procedures: Subjects were 53 Japanese obese children, 33 boys and 20 girls (6 to 14 years old), and 30 age-matched nonobese controls for measuring adiponectin (16 boys and 14 girls). Blood was drawn after an overnight fast, and the obese children were subjected to anthropometric measurements including waist and hip circumferences and skinfold thicknesses. Paired samples were obtained from 21 obese children who underwent psychoeducational therapy. Visceral adipose tissue area was measured by computed tomography. Adiponectin was assayed by an enzyme-linked immunosorbent assay.

Results: The serum levels of alanine aminotransferase, uric acid, triglyceride, total cholesterol, low-density lipoprotein-cholesterol, total cholesterol/high-density lipoprotein-cholesterol, apo B, apo B/apo A1, and insulin in obese children were higher than the reference values. Serum adiponectin level was lower in the obese children than in the controls (6.4 ± 0.6 vs. 10.2 ± 0.8 mg/L, means ± SEM, p < 0.001). In 21 obese children whose percent overweight declined during therapy, the adiponectin level increased (p = 0.002). The adiponectin level was correlated inversely with visceral adipose tissue area in obese children (r = −0.531, p < 0.001). The inverse correlations of adiponectin with alanine aminotransferase, uric acid, and insulin were significant after being adjusted for percentage overweight, percentage body fat, or sex.

Discussion: Serum adiponectin level is decreased in obese children depending on the accumulation of visceral fat and is restored toward normal level by slimming.


Adipose tissue-specific plasma proteins, which are collectively named adipokines (1), mediate metabolic derangement associated with obesity independently of other risk factors for atherosclerosis. Adiponectin, an adipose-specific gene product, which has a matrix-like structure, is abundantly present in the bloodstream (2). Adiponectin, which is the same as gelatin-binding protein of 28 kDa (GBP28), adheres to injured vascular walls in vitro (3). Adiponectin was revealed to have anti-inflammatory effects on the cellular components of vascular wall (4). It inhibits endothelial nuclear factor κB signaling (5) and smooth muscle cell proliferation induced by heparin-binding epidermal growth factor-like growth factor and platelet-derived growth factor (6).

Recently an enzyme-linked immunosorbent assay for human adiponectin has been developed (7) and is used for measuring its serum level. Its plasma levels were significantly lower in adult human subjects with obesity (7), type 2 diabetes (8), or coronary artery disease (4). Therefore, it provides a biological link between obesity and obesity-related disorders such as atherosclerosis, against which it confers protection (9). Animal studies have revealed that adiponectin protects against insulin resistance, diabetes, and atherosclerosis (10, 11, 12).

Epidemiological evidence supports the theory that the relation between obesity and disease risk begins early in life (13). Adult studies (14) show that Asian people are more susceptible to metabolic derangement induced by mild obesity than white people are. Thus, mild obesity may have a significant impact on the health problems in Japanese children. Strong relevance of visceral fat accumulation to metabolic derangements, such as dyslipidemia and hyperinsulinemia, has been recognized in children, as well as in adults. We have previously demonstrated in Japanese children that visceral adipose tissue (VAT)1 area better correlates with metabolic indices than anthropometric surrogates do (15). To date, there has been no clinical report in children suggesting the negative effect of visceral fat accumulation on the blood level of adiponectin.

This study was designed to determine (1) whether serum adiponectin level is decreased, (2) whether it inversely correlates to visceral fat accumulation in obese children, and (3) whether it is restored toward normal level by the reduction of percentage overweight (POW).

Research Methods and Procedures


Fifty-three obese Japanese children, consisting of 33 boys and 20 girls, who visited the Clinic for Obese Children in either University of Yamanashi or the University of Occupational and Environmental Health, were consecutively enrolled in the study. According to the criteria for obesity in childhood adopted by the Ministry of Health, Labor and Welfare in Japan, a child was considered to be obese when the body weight exceeded 120% of the standard body weight, which is defined as the mean body weight corresponding to the height for that age obtained from national statistics for Japanese school children in 1990. The age of the subjects ranged from 6 to 14 years (10.3 ± 0.3 years; Table 1). They had no endocrine, metabolic, or kidney disease. Blood was drawn after an overnight fast, and the children were subjected to anthropometric measurements including height, body weight, waist circumference, hip circumference, and triceps and subscapular skinfold thicknesses. Paired samples were obtained from 21 obese children, 12 boys and 9 girls, who underwent psycho-educational therapy, which was described previously (16).

Table 1.  Anthropometric data and serum adiponectin level for control and obese children
 Boys (n = 16)Girls (n = 14)Boys (n = 33)Girls (n = 20)B vs. GC vs. OBInteraction
  1. B vs. G, boys vs. girls; C vs. OB, control vs. obese; ND, not determined.

  2. Data are expressed as means ± SEM. Statistical difference was evaluated by two factors factorial analysis of variance and unpaired Student's t test.

Age (years)10.8 ± 0.69.8 ± 0.810.6 ± 0.49.7 ± 0.50.0830.8040.952
Height (cm)136.9 ± 3.4135.6 ± 4.1146.3 ± 2.4135.9 ± 2.90.0170.0860.154
Body weight (kg)31.9 ± 2.132.1 ± 3.158.9 ± 2.746.5 ± 2.80.001<0.0010.040
POW−0.4 ± 1.5−0.6 ± 2.356.3 ± 2.547.7 ± 3.1<0.001<0.0010.146
Pondreal index (kg/m3)12.6 ± 0.2212.4 ± 0.3118.5 ± 0.3418.3 ± 0.420.097<0.0010.571
Waist circumference (cm)NDND87.8 ± 1.677.3 ± 1.7<0.001  
WHRNDND0.975 ± 0.0080.920 ± 0.012<0.001  
PBFNDND37.7 ± 1.237.0 ± 1.50.776  
Adiponectin (mg/L)9.8 ± 1.310.7 ± 1.25.8 ± 0.86.6 ± 0.70.241p < 0.0010.898

The age-matched control group for measuring adiponectin (mean age, 10.1 ± 0.6 years) consisted of 30 nonobese children, 16 boys and 14 girls (Table 1). Blood was drawn in the morning after an overnight fast. The clinical laboratory data from obese children were compared with the reference values for children in the University of Yamanashi. The reference values of the serum biochemical indices were obtained from fasting samples of 131 nonobese children as described previously (17). There were no significant sex-related differences in the clinical laboratory data among the children of the control group, as described previously (17).

The Human Study Committee of the University of Yamanashi and the University of Occupational and Environmental Health approved this study. Informed consent was obtained either from each subject or from his or her parents as appropriate.

Anthropometric Measurements

Anthropometric measurements were performed, as described previously (15, 18, 19), by the medical staff in both clinics. In brief, height was measured to the nearest 0.1 cm and body weight to the nearest 0.1 kg using a stadiometer. The waist circumference was measured at the level of the umbilicus and hip circumference at the level of maximum extension of the buttocks, to the nearest 0.1 cm. Skinfold thickness was measured to the nearest 0.1 cm using the skinfold calipers at triceps (halfway between the acromion and the olecranon) and subscapular (1 cm below the inferior angle of the scapula). The POW was calculated using a small programmed calculator (Pocket Growth Checker GEN-185; Sumitomo Pharmaceuticals Co., Osaka, Japan). The percentage body fat (PBF) based on the sum of triceps and subscapular skinfold thicknesses was obtained using the equation of Brozek et al. (20) after body density was calculated according to the formula of Nagamine and Suzuki (21).

Measurement of Abdominal Adipose Tissue Distribution

Subcutaneous adipose tissue (SAT) and VAT were measured by computed tomography (CT) running on 120 kVp, 200 mA, 2.0 s scan time and 10-mm slice thickness (15). A single-slice CT scan of the abdomen was performed at the level of the umbilicus and analyzed for cross-sectional area of adipose tissue. Adipose tissue area was measured in centimeters squared, assuming a density of −40 to −140 Hounsfield Units for adipose tissue, as described by Tokunaga et al. (22). The values of VAT and VAT/SAT ratio (V/S ratio) for Japanese male children were described previously (15).

Biochemical Measurements

Adiponectin in human serum was assayed by an enzyme-linked immunosorbent assay (Chugai Diagnostics Science Co., Ltd., Tokyo, Japan) as described previously (23). The sera were stored frozen at −80 °C until measurement. The sera were diluted by 441-fold before assay. A total of 100 μL of diluted sera were used for the assay. The intra- and interassay CVs were 4.8% to 4.9% and 3.3% to 6.8%, respectively. Serum total cholesterol (TC), triglyceride (TG), high-density lipoprotein-cholesterol (HDL-C), apolipoproteins (apo) A1 and B, alanine aminotransferase (ALT), uric acid, and insulin were measured in the clinical laboratories of both University Hospital of Yamanashi and University Hospital of Occupational and Environmental Health. Low-density lipoprotein-cholesterol (LDL-C) was calculated from the equation of Friedewald et al. (LDL-C = TC − HDL-C − TG/2.18) (24).


Data are presented as means ± SEM. Because the data for TG, ALT, and insulin were skewed, they were transformed logarithmically before performing a statistical analysis. The statistical significance between means was estimated by two factors factorial ANOVA for the two-way (i.e., sex and disease) model and an unpaired Student's t test for two groups. The paired samples were estimated by a paired Student's t test. Differences were considered statistically significant at p < 0.05. Pearson's correlation coefficients were calculated by least-squares linear regression analysis. Partial correlation was calculated according to the analysis of covariance. The statistical analyses were performed using SPSS version 8.01J (SPSS Inc., Chicago, IL).


Anthropometric data for control and obese children are shown in Table 1. Age was not different between sexes or between the obese and control children. The boys were taller than the girls because of the accelerated growth in the obese boys. Body weight and POW were significantly different between sexes because there were more obese boys than obese girls, whereas the sex-related difference in ponderal index was not significant. Body weight, POW, and ponderal index were, by definition, much higher in the obese than in the control children. Waist circumference and waist-hip ratio (WHR), which were determined only in the obese children, were larger in the boys than in the girls. PBF was not different between sexes.

Table 2 summarizes the biochemical data obtained from nonobese control children (i.e., reference values) and obese children. The serum levels of ALT, uric acid, TG, TC, and LDL-C in obese children were significantly higher than those in the control children. Although the levels of HDL-C and apo A1 in the obese children were similar to the controls, the ratio of TC/HDL-C was significantly increased in the obese children. The apo B and the apo B/apo A1 were markedly increased in the obese children. The serum insulin level was significantly increased in the obese children. The sex-related difference was not significant in the parameters listed in Table 2, except for ALT. The significant interaction between effects of sex and disease, observed in TC, LDL-C, TC/HDL-C, apo B, and apo B/apo A1, was because of the differential effect of gender in control and obese children; those parameters tended to be lower in control girls but higher in the obese girls compared with their male counterparts.

Table 2.  Blood biochemistry for reference and obese children
 Boys (n = 69)Girls (n = 62)Boys (n = 33)Girls (n = 20)B vs. GC vs. OBInteraction
  • *

    These parameters were subjected to statistical analysis after logarithmic transformation.

  • Data are means ± SEM. Statistics: two factors factorial ANOVA.

  • B vs. G, boys vs. girls; C vs. OB, control vs. obese.

ALT (U/L)*21 ± 118 ± 169 ± 1241 ± 130.020<0.0010.047
Uric acid (μM)295 ± 10293 ± 7365 ± 19329 ± 180.159<0.0010.197
TG (mM)*0.81 ± 0.030.83 ± 0.041.19 ± 0.191.13 ± 0.140.690<0.0010.681
TC (mM)4.30 ± 0.074.21 ± 0.094.63 ± 0.185.09 ± 0.270.877<0.0010.043
HDL-C (mM)1.36 ± 0.031.41 ± 0.041.45 ± 0.061.44 ± 0.100.5430.2120.585
LDL-C (mM)2.56 ± 0.062.42 ± 0.062.64 ± 0.153.13 ± 0.230.9500.0020.004
TC/HDL-C3.20 ± 0.053.06 ± 0.063.33 ± 0.163.82 ± 0.340.9470.0020.012
Apo A1 (μM)45.6 ± 0.945.5 ± 1.046.3 ± 1.244.7 ± 1.70.6220.9630.600
Apo B (μM)1.33 ± 0.031.28 ± 0.031.53 ± 0.081.73 ± 0.130.988<0.0010.028
Apo B/Apo A1 (×102)2.94 ± 0.072.89 ± 0.073.35 ± 0.174.01 ± 0.370.561<0.0010.008
Insulin (pM)*70.5 ± 3.673.5 ± 5.0124.9 ± 16.794.3 ± 13.70.241<0.0010.068

Serum adiponectin levels in the obese boys and obese girls are shown in Table 1. The levels were lower in the obese children than in the controls (6.4 ± 0.6 vs. 10.2 ± 0.8 mg/L, p < 0.001, when both sexes were combined). There was no sex-related difference in adiponectin level or interaction between effects of sex and disease (Table 1).

Abdominal adipose tissue distribution was measured in 30 obese boys and 17 obese girls. Table 3 lists the VAT and SAT areas in obese children. VAT area and the V/S ratio in the obese boys were significantly greater than those in the obese girls, whereas SAT area was not significantly different between sexes.

Table 3.  Visceral and subcutaneous adipose tissue areas in obese children
 BoysGirlsBoys vs. girls
  1. Visceral and subcutaneous adipose tissue areas were measured by a single CT slice at the level of umbilicus, as described in the text.

VAT (cm2)81.2 ± 5.856.7 ± 3.50.004
SAT (cm2)242.6 ± 12.6208.8 ± 9.80.080
V/S ratio0.346 ± 0.0240.269 ± 0.0170.032

Table 4 summarizes the relationships between anthropometric indices and visceral adipose tissue area in obese children. The waist circumference was correlated closely with POW, VAT, and SAT, and moderately with PBF, but not with V/S ratio. The WHR was correlated moderately with POW, PBF, VAT, and SAT, but not with V/S ratio. The ponderal index was correlated closely with POW and moderately with PBF and SAT, but not with VAT or V/S ratio.

Table 4.  Relationship between anthropometric indices and visceral adipose tissue area in obese children
  • *

    p < 0.05,

  • p < 0.01,

  • p < 0.001.

  • NS, not significant.

Waist circumference0.5050.4200.7120.757NS
Ponderal index0.9070.381NS0.405NS

The adiponectin level was inversely correlated with waist circumference (r = −0.559, p < 0.001, n = 53), VAT area (r = −0.531, p < 0.001, n = 47; Figure 1), SAT area (r = −0.404, n = 47, p = 0.005), and V/S ratio (r = −0.346, n = 47, p = 0.017). The partial correlation between VAT area and adiponectin was significant even after being adjusted for sex (−0.567, p < 0.001), age (−0.367, p = 0.012), POW (−0.520, p < 0.001), PBF (−0.492, p = 0.001), or height (−0.346, p = 0.019). Correlations of adiponectin with POW (r = −0.170, n = 53), ponderal index (r = −0.020, n = 53), or PBF (r = −0.190, n = 53) were not significant.

Figure 1.

Relationship between visceral adipose tissue area of the obese children (n = 47) and serum adiponectin level. Pearson's correlation coefficient is −0.531 (p < 0.001).

Table 5 lists the correlations between adiponectin level and other biochemical data for obese children. The adiponectin level was inversely correlated with ALT, uric acid, TG, TC/HDL-C, apo B/apo A1, and insulin, but not with TC, LDL-C, apo A1, or apo B. The correlation profile did not change greatly when adjusted for POW, PBF, or sex. The adiponectin level was correlated inversely with ALT, apo B/apo A1, and insulin, and positively with HDL-C and apo A1, when adjusted for age.

Table 5.  Relationships between adiponectin and other blood biochemistry data for obese children
  • *

    These parameters were subjected to statistical analysis after logarithmic transformation.

  • p < 0.05,

  • p < 0.01.

  • The number of observation was 47 for VAT and SAT and 53 for the correlation between adiponectin and blood biochemistry and the partial correlations adjusted for age, POW, PBF, and sex.

  • NS, not significant.

Uric acid−0.383NS−0.356−0.370−0.376NSNS
Apo B/Apo A1−0.285−0.279NSNS−0.317NSNS

The data for the paired samples of adiponectin and each respective POW are summarized in Figure 2. The POW declined in all children (from 56.5 ± 3.1% to 44.5 ± 2.9%, p < 0.001) during the psycho-educational therapy (duration: mean, 8 ± 2 months; range, 2 to 20 months), with a concomitant increase in adiponectin level in 18 children. (Adiponectin was decreased in one child and unchanged in two children.) The initial adiponectin level differed significantly from the second value (5.8 ± 0.7 vs. 8.0 ± 1.0 mg/L, p = 0.002). Figure 3 depicts the relationship between ΔPOW and the ratio of post-/pretreatment levels of adiponectin. They were correlated inversely (r = −0.571, p = 0.007). The change in body weight during treatment was not significant (from 50.9 ± 2.6 to 50.5 ± 2.5 kg, p = 0.677).

Figure 2.

Repeated measures of adiponectin in obese children. A significant reduction in POW was noted in 21 children who underwent psycho-educational therapy for various periods of time (2 to 20 months). Serum adiponectin level decreased in 18, remained unchanged in 2, and increased in 1 by the treatment.

Figure 3.

Effect of slimming on serum adiponectin level in obese children (n = 21). The ΔPOW is inversely correlated with the ratio of post-/pretreatment adiponectin level that is indicated as adp (post/pre).


There was no sex-related difference in serum adiponectin level in these children. In two previous large-scale adult studies, plasma adiponectin level was found to be lower in men than in women (25, 26). The studies in mice and cultured adipocytes suggest that androgens decrease plasma adiponectin (26). In contrast, Stefan et al. (27) reported the plasma adiponectin concentration in 5- and 10-year-old Pima Indian children. They found no sex-related difference, as was the case in the present study. Thus, sex-related difference in serum adiponectin level does not seem to be evident in the pediatric age group.

Serum adiponectin level was decreased in our obese children and was restored toward normal level by slimming. The decrease in POW was closely correlated to the restoration of adiponectin level relative to baseline value. It has previously been demonstrated in adults that plasma adiponectin level is decreased in obesity (7) and that weight reduction significantly elevates plasma adiponectin levels in diabetic and nondiabetic subjects (8, 9). The present change in adiponectin in response to change in obesity status in children is similar to that observed in adults. We previously demonstrated that serum level of cholesteryl-ester transfer protein, another adipokine, was increased in obese children, and that it declined in response to mild slimming (19). The present change in serum adiponectin level was a mirror image of what was observed by us previously in the cholesteryl-ester transfer protein level of obese children.

Adiponectin gene expression in adipose tissue paradoxically decreases despite the increase in tissue mass in obesity. This paradox is, at least partly, explained by the antagonism of tumor necrosis factor-α (TNF-α) to adiponectin (28) and vice versa. TNF-α, which is overexpressed in adipose tissue of the obese subjects, blocks insulin action through inhibition of insulin receptor substrate-1 and tyrosine kinase activity (29). TNF-α dose-dependently reduces the expression of adiponectin in 3T3-L1 adipocytes by suppressing its promoter activity (28). On the other hand, thiazolidinedione derivatives enhance the mRNA expression and secretion of adiponectin in a dose- and time-dependent manner and restore the inhibitory effect of TNF-α (28). Adiponectin enhances insulin sensitivity by blocking TNF-α signals (5).

Adult studies have revealed that waist circumference is a good anthropometric surrogate for VAT area. Men are characterized by a preferential accumulation of abdominal adipose tissue as revealed by an increased waist circumference and a greater VAT accumulation compared with women with the same amount of total body fat mass (30). The waist circumference and WHR were significantly greater in the boys than in the girls of our previous (18) and present series of obese children who were matched for age. The VAT area was greater in the obese boys than in the obese girls in this study, reflecting a larger WHR and waist circumference in the boys than in the girls.

The serum adiponectin level was inversely correlated with the VAT area in these obese children. Concerning the relation between abdominal obesity and serum adiponectin level, Yang et al. (9) reported that the change in plasma adiponectin levels was inversely correlated with changes in waist and hip circumferences in obese patients who received gastric partition surgery. Motoshima et al. (31) reported that cultured human omental adipocytes secreted more adiponectin than subcutaneous adipocytes and that the secretion was inversely correlated to BMI. They also found that secretion from the subcutaneous cells was unrelated to BMI. Adiponectin expression was lower in VAT of genetically obese than in VAT of lean rats, whereas no differences were observed when SAT of the same animals was compared (32). Weight loss resulted in an increase of adiponectin expression in VAT (32). Thus, VAT is the main source of adiponectin secretion, and certain metabolic derangement related to VAT accumulation is postulated to suppress adiponectin expression in VAT. The mechanism by which serum adiponectin level correlates inversely with VAT area needs to be elucidated in further studies.

Epidemiological studies in adults show that hypoadiponectinemia is linked to various components of the metabolic syndrome such as high blood pressure, insulin resistance, glucose intolerance, dyslipidemia, hyperuricemia, and low HDL-C (25, 33). In these obese children, serum adiponectin level was inversely correlated with biochemical parameters that showed obesity-related elevation in the serum levels (i.e., ALT, uric acid, TG, and insulin) even after being adjusted for POW, PBF, or sex. Correlations between adiponectin and most of such biochemical parameters were no longer significant when being adjusted for VAT or SAT, indicating that the relationship is highly dependent on abdominal fat accumulation. Adiponectin was also inversely correlated with atherogenic biochemical parameters (i.e., TC/HDL-C and apo B/apo A1) in these obese children. These results collectively suggest that hypoadiponectinemia represents the metabolic syndrome including insulin resistance, VAT accumulation, and atherogenic lipoprotein changes in children, as has been the case in several previous adult studies.

Animal models provide direct evidence that adiponectin protects against development of insulin resistance and subsequent atherosclerosis. Adiponectin-deficient mice showed insulin resistance with glucose intolerance, with severe neointimal thickening in mechanically injured arteries (34). Adiponectin stimulates glucose use and fatty-acid oxidation through phosphorylation and activation of the 5′-adenosine 3′, 5′-monophosphate-activated protein kinase, thereby directly regulating glucose metabolism and insulin sensitivity (10). When apolipoprotein E-deficient mice were treated with recombinant adenovirus expressing adiponectin, plasma adiponectin level was markedly elevated, and it suppressed the development of atherosclerosis in vivo (11). Globular domain adiponectin transgenic ob/ob mice showed amelioration of insulin resistance and β-cell degranulation as well as diabetes (12).

In conclusion, the negative association of adiponectin with insulin and atherogenic biochemical parameters, observed in this study, suggests that hypoadiponectinemia can be a component of metabolic derangement, leading to early development of vascular lesions caused by the metabolic syndrome in the pediatric age group.


This work is supported in part by Health Science Research Grants (Research on Children and Families) from Ministry of Health, Labor and Welfare, Japan, and also in part by Grant-in-Aid 12670735 from the Ministry of Education, Science and Culture of Japan. The authors thank Mitsuo Otsuka and Koichi Kawahara, Chugai Diagnostics Science Co., Ltd., Tokyo, Japan, for their valuable technical assistance.


  • 1

    Nonstandard abbreviations: VAT, visceral adipose tissue; POW, percentage overweight; PBF, percentage body fat; SAT, subcutaneous adipose tissue; CT, computed tomography; V/S ratio; VAT/SAT ratio; TC, total cholesterol; TG, triglyceride; HDL-C, high-density lipoprotein-cholesterol; apo, apolipoprotein; ALT, alanine aminotransferase; LDL-C, low-density lipoprotein-cholesterol; WHR, waist-to-hip ratio; TNF-α, tumor necrosis factor alpha.