Leptin and Adiponectin Concentrations Independently Predict Future Accumulation of Visceral Fat in Nondiabetic Japanese Americans
Abstract
Objective
Whether leptin and adiponectin are independently associated with regional body fat distribution was investigated in a prospective study of Japanese Americans.
Methods
Nondiabetic participants 39 to 79 years of age were followed for 5 years to assess change in body composition. Leptin and adiponectin concentrations were evaluated at baseline and by single-slice computed tomography measurements of intra-abdominal fat (IAF), abdominal subcutaneous fat (SCF), and thigh SCF cross-sectional areas at baseline and at 5 years.
Results
Ninety-six men and ninety-five women without diabetes had the following baseline mean (SD) values: age 45.7 (3.5) years and 46.4 (3.9) years, IAF 78.7 (38.6) cm2 and 62.1 (39.0) cm2, leptin concentration 4.5 (2.3) μg/L and 10.2 (5.2) μg/L, and adiponectin concentration 7.4 (3.2) μg/mL and 10.8 (4.7) μg/mL, respectively. Baseline leptin (β = 1.7722, P = 0.014) and adiponectin concentrations (β = −0.4162, P < 0.001) were significantly associated with IAF change over 5 years in multivariable models adjusting for age, sex, diabetes family history, weight change over 5 years, and baseline measurements of BMI, IAF, abdominal SCF, waist circumference, thigh fat, and homeostatic model assessment of insulin resistance.
Conclusions
In nondiabetic Japanese Americans, a higher concentration of leptin was associated with greater accumulation of IAF and a higher concentration of adiponectin with lesser accumulation of IAF over 5 years.
Study Importance
What is already known?
- ► Adipose tissue is recognized to be an important endocrine organ responsible for expression and secretion of a variety of biologically active polypeptides known as adipokines.
- ► Circulating leptin and adiponectin concentrations are strongly correlated.
- ► There have been no reports as to whether each might independently predict longitudinal changes of body fat distribution.
What does this study add?
- ► This longitudinal study provides novel evidence that each baseline adipokine measurement was independently associated with change in intra-abdominal fat, but in different directions.
Introduction
Adipose tissue is recognized to be an important endocrine organ responsible for expression and secretion of a variety of biologically active polypeptides known as adipokines ((1-3)). Since the discovery of leptin, numerous other adipokines have been identified ((4)). We have examined two major adipokines, leptin and adiponectin, with respect to longitudinal changes of body fat distribution over 5 years.
Leptin primarily regulates energy homeostasis by controlling satiety and body weight ((5)). The adipokine adiponectin plays an important role in regulating whole-body energy homeostasis by the regulation of appetite and satiety, body fat distribution, insulin sensitivity, insulin secretion, and energy expenditure ((6-9)). Adipokine dysfunction, though, can occur in association with obesity, as in the case of leptin, for which the ability to regulate energy expenditure is diminished, which has been described by the term “leptin resistance” ((3, 10, 11)). The production of adiponectin is reduced by higher levels of inflammatory adipokines associated with visceral fat accumulation ((6)).
The adipokines adiponectin and leptin have roles as mediators linking fat mass and adipose tissue function to metabolic diseases ((8, 12)). They not only have organ-specific effects but also interact with each other to regulate energy homeostasis. These interactions result from a coordinated change of transcriptional and post-translation mechanisms affecting adipocyte genes and protein expression ((6)).
In addition to the effects that adipokines have on each other and on organ systems, they are also associated with weight change and regional body fat distribution. In separate analyses, higher concentrations of circulating leptin and lower plasma adiponectin concentrations predicted future abdominal visceral fat accumulation in nondiabetic Japanese Americans ((13, 14)). Higher concentrations of circulating leptin were also associated with a greater increase in body weight in a prospective study conducted in this same population ((15)). An increase in visceral fat area among premenopausal women followed prospectively through menopause was positively correlated with change in leptin concentration and negatively with change in adiponectin concentration in univariate models ((16)).
Circulating leptin and adiponectin concentrations are strongly correlated ((17, 18)). Since the available research has only investigated the association between change in visceral adiposity and each of these adipokines individually, uncertainty remains about whether, for example, the association between leptin and change in visceral adiposity is due only to its correlation with adiponectin only and vice versa. There have been no reports as to whether each might independently predict longitudinal changes in body fat distribution in models with mutual adjustment for both of these adipokines. We investigated this over 5 years in a prospective study of Japanese Americans using computed tomography (CT) to evaluate regional body fat distribution.
Methods
Study population
The study population consisted of third-generation Japanese American men and women of 100% Japanese ancestry enrolled in the Japanese American Community Diabetes Study. A detailed description of the selection and recruitment of the study participants has been published previously ((19, 20)). For the current analysis, eligible participants did not have diabetes based on fasting plasma glucose < 126 mg/dL, had 2-hour plasma glucose < 200 mg/dL after a 75-g oral glucose tolerance test, and were not taking glucose-lowering medications. They also had to have completed baseline and 5-year follow-up examinations. A total of 191 nondiabetic individuals meeting these criteria were included in this analysis. The protocol was approved by the Human Subjects Review Committee of the University of Washington. Signed informed consent was obtained from all participants.
Clinical and laboratory examination
All evaluations were performed at the General Clinical Research Center, University of Washington Medical Center, Seattle, Washington. At baseline, a complete physical examination was performed, and personal medical history and lifestyle factors, including cigarette smoking, alcohol consumption, and physical activity, were evaluated using a standardized questionnaire. Smoking was classified into three groups (current smoker [smoker at the time of the examination], past smoker [smoker prior to the time of the examination but currently not a smoker], and never smoked). The Paffenbarger Physical Activity Index Questionnaire was used to determine physical activity level (usual kilocalories spent weekly) ((21)). Alcohol consumption was defined in grams per day ((21)). Family history of diabetes was considered positive if any first-degree relative had diabetes. Waist circumference was measured at the level of the umbilicus.
Blood samples were obtained following an overnight fast of 10 hours. Plasma glucose and insulin were measured as reported previously ((22)), glucose by the hexokinase method using an autoanalyzer (Department of Laboratory Medicine, University of Washington, Seattle, Washington) and insulin by radioimmunoassay (Immunoassay Core, Diabetes Research Center, University of Washington). To estimate insulin sensitivity, the homeostatic model assessment of insulin resistance index based on fasting glucose and insulin concentrations was calculated as (fasting serum insulin [µU/mL] × fasting serum glucose [mg/dL]) ÷ 405 ((23)). Leptin and adiponectin measurements were obtained from frozen plasma. The plasma concentration of leptin and total adiponectin were determined in duplicate using a commercially available radioimmunoassay kit (Linco Research) ((13, 14, 24)). Plasma from fasting morning blood samples (7:30-8:30 am) was stored at −70°C and thawed just before use for leptin measurement. Intra-assay and inter-assay coefficients of variation were 3.53% and 5.2% in the leptin assay and 6.21% and 9.25% in the adiponectin assay, respectively.
Regional body fat distribution was quantified by CT at baseline and at 5-year follow-up. Single 10-mm-slice CT scans were performed at the level of the umbilicus to measure cross-sectional fat areas (centimeters squared) of abdominal subcutaneous fat between skin and abdominal muscle (SCF) and intra-abdominal fat (IAF) within the transversalis fascia. Right-thigh fat area was measured midway between the greater trochanter and superior margin of the patella. CT scans were analyzed using density contour software (Standard General Electric 8800 computer software). Attenuation range for identification of fat was −250 to −50 Hounsfield units ((25)). Changes in IAF and weight were calculated by subtracting baseline results from results at 5-year follow-up.
Statistical analyses
Continuous variables are expressed as mean (SD) for normally distributed data as assessed by the Shapiro-Wilk test or median with interquartile range for non-normally distributed data. Categorical variables are expressed as count and percentage. The χ2 test was used to compare differences in frequencies for categorical data, as appropriate. We estimated unadjusted linear regression coefficients between change in IAF from baseline to 5 years (IAF at 5 years minus baseline IAF [Δ IAF]) as the continuous dependent variable and metabolic variables, anthropometric variables, or body composition measures on CT scan as the independent variables. Partial correlation analysis and multiple regression analysis were carried out to assess the independent relationships between Δ IAF in relation to leptin or adiponectin concentration while adjusting for covariates such as age, sex, BMI, regional fat depots, insulin sensitivity, family history of diabetes, alcohol consumption, physical activity, and smoking. Covariates chosen for inclusion in the multivariable models included factors that are potentially associated with both adipokine level and Δ IAF and as such might satisfy the criteria for confounding bias. Possible interactions between sex and adipokines (leptin or adiponectin) with regard to Δ IAF were tested by the insertion of first-order interaction terms into the regression model. Multivariable models were assessed for collinearity using the variance inflation factor, with a value greater than 4 suggesting its presence. To assess the normality of the residuals of the dependent variable in the linear regression models, we produced a kernel density plot of the residuals overlaying a normal probability plot and found no important deviations from normality. All statistical analyses were performed using Stata/MP version 15.1 (StataCorp LLC). A two-sided P < 0.05 was considered to indicate statistical significance.
Results
Baseline characteristics of the study participants
The clinical and laboratory characteristics of the participants are shown in Table 1. A total of 191 individuals without diabetes (96 men, 95 women) were included in this analysis with a mean age of 46.0 (3.7) years and a mean BMI of 24.8 (3.6) kg/m2 (Table 1). The mean concentrations of leptin and adiponectin were, as expected, lower in men than in women ((15, 26)). Although the mean IAF and waist circumference were greater in men than in women, the mean abdominal SCF and thigh SCF were lower in men than in women. Mean alcohol consumption, physical activity, and proportion currently smoking were higher in men than in women.
| Total (N = 191) | Male (N = 96) | Female (N = 95) | P | ||||
|---|---|---|---|---|---|---|---|
| Mean or median | SD or IQR | Mean or median | SD or IQR | Mean or median | SD or IQR | ||
| Age (y) | 44.9 | 43.3 to 48.5 | 44.8 | 43.4 to 46.5 | 45.5 | 43.3 to 49.7 | 0.151 |
| Family history of diabetes (N, %) | 60 (31.4%) | 27 (28.1%) | 33 (34.7%) | 0.352 | |||
| Weight (kg) | 66.0 | 12.3 | 73.3 | 9.8 | 58.6 | 10.0 | < 0.001 |
| BMI (kg/m2) | 24.1 | 22.3 to 26.9 | 25.5 | 23.5 to 27.0 | 23.2 | 21.4 to 26.5 | 0.003 |
| Fasting glucose (mg/dL) | 97.4 | 8.3 | 100.0 | 7.8 | 94.8 | 8.1 | < 0.001 |
| 2-hour glucose (mg/dL) | 135.3 | 27.8 | 131.8 | 26.1 | 138.8 | 29.1 | 0.077 |
| Fasting insulin | 12.0 | 9.0 to 18.0 | 13.0 | 10.0 to 18.5 | 12.0 | 8.0 to 17.0 | 0.114 |
| HOMA-IR | 2.9 | 2.2 to 4.3 | 3.2 | 2.4 to 4.5 | 2.7 | 1.8 to 4.1 | 0.064 |
| Adiponectin (µ g/mL) | 8.4 | 6.0 to 11.6 | 7.0 | 4.8 to 9.7 | 10.3 | 7.8 to 13.0 | < 0.001 |
| Leptin (pmol/L) | 6.0 | 3.6 to 10.1 | 4.2 | 2.8 to 5.9 | 9.6 | 6.9 to 13.3 | < 0.001 |
| Abdominal circumference (cm) | 84.4 | 9.8 | 87.8 | 7.4 | 80.9 | 10.7 | < 0.001 |
| Baseline IAF (cm2) | 65.7 | 36.7 to 99.3 | 75.7 | 52.2 to 106.7 | 50.2 | 30.9 to 84.1 | 0.004 |
| Baseline abdominal SCF (cm2) | 163.3 | 107.8 to 221.4 | 139.4 | 106.2 to 191.5 | 187.7 | 122.1 to 245.4 | < 0.001 |
| Baseline thigh SCF (cm2) | 63.7 | 44.5 to 88.2 | 45.2 | 34.6 to 60.4 | 85.7 | 68.8 to 107.0 | < 0.001 |
| Change in IAF (cm2) | 9.2 | −6.6 to 23.0 | 13.7 | −9.0 to 32.3 | 6.2 | −4.9 to 17.7 | 0.061 |
| Change in abdominal SCF (cm2) | 13.0 | −5.0 to 41.7 | 12.8 | −3.8 to 34.4 | 13.1 | −6.8 to 49.1 | 0.944 |
| Change in thigh SCF (cm2) | 1.1 | −10.7 to 13.5 | −0.59 | −8.5 to 6.6 | 2.5 | −21.5 to 22.8 | 0.455 |
| Alcohol consumption (g/d) | 0.0 | 0.0 to 6.32 | 1.9 | 0.0 to 12.6 | 0.0 | 0.0 to 1.6 | < 0.001 |
| Physical activity (kcal/wk) | 2,492 | 1,648 to 3,553 | 2,759 | 1,947 to 4,176 | 2,061 | 1,315 to 3,066 | 0.001 |
| Current smoking (N, %) | 33 (17.3%) | 18 (18.8%) | 15 (15.8%) | 0.030 | |||
- HOMA-IR, homeostatic model assessment of insulin resistance; IAF, intra-abdominal (visceral) fat; IQR, interquartile range; SCF, subcutaneous fat.
Relationship of baseline characteristics with Δ IAF
The univariate analysis of associations between baseline characteristics and Δ IAF showed significant positive associations with change in body weight from baseline to 5 years (Δ body weight) and negative associations with baseline IAF, baseline abdominal SCF, and baseline thigh SCF (Table 2). Age, family history of diabetes, BMI, fasting glucose, alcohol consumption, physical activity, and leptin and adiponectin concentrations were not significantly associated with Δ IAF in unadjusted analysis. Since adiponectin and leptin are known to vary by sex and have an association with each other, we explored the association between leptin and adiponectin concentrations while adjusting for sex and found a significant negative correlation (r = −0.2950, P < 0.001).
| Baseline characteristic | βa | P | 95% CI |
|---|---|---|---|
| Age | −0.7760 | 0.227 | −2.0402 to 0.4883 |
| Female | −8.9425 | 0.062 | −18.3454 to 0.4603 |
| Family history of diabetes | 3.7860 | 0.465 | −6.4207 to 13.9927 |
| Weight | −0.0472 | 0.807 | −0.4289 to 0.3346 |
| BMI | −1.1141 | 0.089 | −2.4017 to 0.1736 |
| Fasting glucose | −0.2577 | 0.363 | −0.8157 to 0.3003 |
| 2-hour glucose | −0.1263 | 0.149 | −0.2984 to 0.0458 |
| Fasting insulin | −0.1308 | 0.537 | −0.5486 to 0.2869 |
| HOMA-IR | −0.8104 | 0.335 | −2.4650 to 0.8442 |
| Adiponectin | −0.1653 | 0.766 | −0.0013 to 0.0009 |
| Leptin | −0.9635 | 0.052 | −1.9365 to 0.0095 |
| Δ Body weight | 4.9902 | < 0.001 | 4.0143 to 5.9662 |
| Waist circumference | −0.2896 | 0.231 | −0.7648 to 0.1856 |
| Baseline IAF | −0.2581 | < 0.001 | −0.3721 to −0.1442 |
| Baseline abdominal SCF | −0.0692 | 0.008 | −0.1205 to −0.0180 |
| Baseline thigh SCF | −0.1580 | 0.017 | −0.2873 to −0.0287 |
| Alcohol consumption | 0.0224 | 0.913 | −0.3791 to 0.4238 |
| Physical activity | −0.0009 | 0.464 | −0.0032 to 0.0014 |
- a Regression coefficient.
- HOMA-IR, homeostatic model assessment of insulin resistance; IAF, intra-abdominal (visceral) fat; SCF, subcutaneous fat.
Because there were no significant correlations between adipokines and Δ IAF in the univariate analysis, we next performed partial correlation analyses (Table 3), adjusting for each factor related to leptin and adiponectin concentration and examining whether such adjustments changed the association between either adipokine and Δ IAF. Each factor was individually included in the models with both adipokines as independent variables. This analysis showed that adiponectin was significantly negatively correlated with Δ IAF after adjustment for baseline IAF. With adjustment for family history of diabetes and smoking, there was a significant negative correlation between leptin concentration and Δ IAF. Δ body weight and baseline abdominal SCF had significant positive and negative correlations, respectively, with Δ IAF, but thigh SCF lost its significant correlation with Δ IAF in models that contained both adipokines as independent variables (Table 3).
| Covariate | Leptin | Adiponectin | ||||
|---|---|---|---|---|---|---|
| r | β | P | r | β | P | |
| None | −0.1529 | −0.9638 | 0.053 | −0.0242 | −0.1678 | 0.761 |
| Age | −0.1493 | −0.9401 | 0.060 | −0.0163 | −0.1135 | 0.838 |
| Female | −0.0875 | −0.7032 | 0.271 | 0.0053 | 0.0422 | 0.947 |
| Family history of diabetes | −0.1663 | −1.0796 | 0.036 | −0.0136 | −0.0951 | 0.864 |
| Weight | −0.1487 | −0.9437 | 0.061 | −0.0337 | −0.2658 | 0.673 |
| BMI | −0.0831 | −0.5995 | 0.296 | −0.0655 | −0.5042 | 0.410 |
| Fasting glucose | −0.1421 | −0.8975 | 0.073 | −0.0469 | −0.3366 | 0.556 |
| Postprandial glucose | −0.1184 | −0.7848 | 0.136 | −0.0343 | −0.2392 | 0.667 |
| HOMA-IR | −0.1341 | −0.8706 | 0.091 | −0.0420 | −0.3091 | 0.598 |
| Waist circumference | −0.1092 | −0.7408 | 0.171 | −0.0537 | −0.4255 | 0.502 |
| Δ Body weight | −0.1174 | −0.5864 | 0.139 | −0.0776 | −0.4285 | 0.330 |
| IAF | 0.0225 | 0.1422 | 0.778 | −0.2237 | −1.6870 | 0.004 |
| Abdominal SCF | 0.0181 | 0.1699 | 0.820 | −0.0794 | −0.5837 | 0.318 |
| Thigh SCF | −0.0294 | −0.2538 | 0.713 | −0.0152 | −0.1050 | 0.849 |
| Alcohol consumption | −0.1546 | −1.0022 | 0.051 | −0.0263 | −0.1834 | 0.741 |
| Physical activity | −0.1320 | −0.8437 | 0.096 | −0.0042 | −0.0295 | 0.958 |
| Smoking | −0.1586 | −1.0127 | 0.046 | 0.0000 | 0.0002 | 1.000 |
- HOMA-IR, homeostatic model assessment of insulin resistance; IAF, intra-abdominal (visceral) fat; SCF, subcutaneous fat.
We performed multivariate analyses to determine whether leptin and adiponectin independently predicted Δ IAF with adjustment for multiple covariates (Table 4). The first model (Model 1) included age, family history of diabetes, BMI, homeostatic model assessment of insulin resistance, IAF, abdominal SCF, thigh SCF, sex, Δ body weight, and leptin. Only the last three variables were significantly associated with Δ IAF. Δ body weight showed significant positive associations, while sex and baseline IAF showed a significant negative association with Δ IAF. Further adjustment of Model 1 for alcohol consumption (Model 2), alcohol consumption and physical activity (Model 3), and alcohol consumption, physical activity, and smoking (Model 4) yielded similar results for Δ body weight, baseline IAF, and sex. In these models, leptin remained significantly positively associated with Δ IAF. Adiponectin showed significant negative associations with Δ IAF in Models 3 and 4. No multivariable model exhibited collinearity as all variance inflation factors were less than 4.
| Baseline characteristic | Model 1 | Model 2 | Model 3 | Model 4 | ||||
|---|---|---|---|---|---|---|---|---|
| βa | P | βa | P | βa | P | βa | P | |
| Age | 0.0504 | 0.915 | 0.0564 | 0.905 | −0.0371 | 0.938 | −0.1666 | 0.733 |
| Female | −16.4671 | 0.031 | −16.0340 | 0.037 | −15.9550 | 0.037 | −16.5320 | 0.032 |
| Family history of diabetes | 6.6556 | 0.091 | 6.7131 | 0.090 | 6.6063 | 0.093 | 6.2054 | 0.118 |
| BMI | 1.7816 | 0.112 | 1.8065 | 0.109 | 2.2431 | 0.052 | 2.1091 | 0.070 |
| HOMA-IR | −0.3920 | 0.578 | −0.4297 | 0.546 | −0.6005 | 0.402 | −0.6401 | 0.373 |
| Leptin | 1.5254 | 0.030 | 1.5463 | 0.029 | 1.6090 | 0.023 | 1.7722 | 0.014 |
| Adiponectin | −0.9501 | 0.054 | −0.9573 | 0.053 | −1.0167 | 0.040 | −1.0690 | 0.033 |
| IAF | −0.4041 | < 0.001 | −0.4034 | < 0.001 | −0.4105 | < 0.001 | −0.4162 | < 0.001 |
| Abdominal SCF | −0.0462 | 0.278 | −0.0472 | 0.270 | −0.0602 | 0.165 | −0.0604 | 0.165 |
| Thigh SCF | −0.0086 | 0.915 | −0.0082 | 0.920 | −0.0204 | 0.801 | −0.0155 | 0.849 |
| Δ Body weight | 4.6479 | < 0.001 | 4.6701 | < 0.001 | 4.6152 | < 0.001 | 4.6225 | < 0.001 |
| Alcohol consumption | 0.707 | 0.0714 | 0.640 | 0.0774 | 0.610 | 0.0286 | 0.856 | |
| Physical activity | −0.0014 | 0.111 | −0.0014 | 0.115 | ||||
| Current smoking | 3.1922 | 0.532 | ||||||
| R 2 | 0.5442 | 0.5449 | 0.5527 | 0.5573 | ||||
- a Regression coefficient.
- HOMA-IR, homeostatic model assessment of insulin resistance; IAF, intra-abdominal (visceral) fat; SCF, subcutaneous fat.
We also conducted analyses through insertion of leptin × adiponectin interaction terms into the multivariate models in Table 4 to determine whether the association between leptin, adiponectin, and Δ IAF differed by the concentration of the other adipokine. No significant interaction was observed between those two adipokines in these multivariate models (coefficient 0.1009, P = 0.395). Furthermore, we examined whether interaction was present between sex and either adipokine by insertion of leptin × sex or adiponectin × sex terms into the multivariate models. None of the adipokine × sex interaction terms were significant. A sex-stratified version of Model 4 in Table 4 yielded the following values of the coefficients: leptin in men (1.82, P = 0.340) and women (2.20, P = 0.002) and adiponectin in men (−0.001, P = 0.364) and women (−0.001, P = 0.025). The results shown in Table 4 were similar when we repeated Models 1-4 after excluding the five participants (n = 4 men and 1 woman) who were taking lipid-lowering medications (data not shown).
Discussion
In the current prospective study of Japanese American, nondiabetic men and women, we found that leptin and adiponectin concentrations both independently predicted future accumulation of IAF over 5 years without any interaction between these adipokines or between these adipokines and sex. This association was independent of age, sex, insulin sensitivity, glycemia, body composition, smoking, and lifestyle factors potentially affecting leptin and adiponectin level such as alcohol consumption, smoking, and physical activity. There was greater accumulation of IAF in those with higher leptin or lower adiponectin concentrations. Female sex showed the greatest absolute magnitude and negative association with Δ IAF over 5 years, as would be expected because of the well-known greater average size of this fat depot in men than in women. No significant differences were seen in the associations between the adipokines and Δ IAF as shown by the nonsignificant sex × leptin and sex × adiponectin interaction terms. The values of the coefficients for leptin and adiponectin in multivariable adjusted models in predicting Δ IAF were similar in magnitude, as expected given the nonsignificant interaction with sex. The adipokine– Δ IAF association in the sex-stratified model remained significant in women but not in men, potentially owing to multiple reasons, including less variability in these associations in women, a reduction in the sample size of each model because of stratification, or insufficient power of the interaction analysis to detect a difference by gender.
Although leptin plays an important role in energy intake and expenditure in rodents, it is well documented that, except in certain genetic disorders, externally administrated leptin is ineffective in reducing body adiposity in humans, giving support for the existence of leptin resistance ((1, 27, 28)). Lending support to this concept is our finding of higher concentrations of leptin associated with greater accumulation of IAF. Higher leptin concentration has been associated with greater weight gain in humans, giving support for the concept of leptin resistance. It is not clear why such gain should favor the IAF depot, as our results demonstrate, in which an increase in IAF was seen after adjustment for overall weight gain. A potential explanation may be derived from the associations between exercise and leptin concentration, which falls in response to aerobic exercise ((29)). Also, it has been shown that visceral fat area decreased to a greater degree than SCF fat area in response to a lifestyle intervention including diet and exercise in the Diabetes Prevention Program ((30)). Although we adjusted for physical activity in these analyses, it is possible that lower leptin concentration reflects greater physical activity and, for this reason, may predict less visceral fat accumulation.
Higher adiponectin concentrations have been associated with lesser amounts of IAF, and our findings suggest that it may directly inhibit or be a marker for a factor that inhibits visceral fat accumulation ((5, 31-33)). The gender differences noted in adiponectin concentrations may be relevant to the negative association that we observed between concentration of this adipokine and IAF accumulation. Adiponectin concentration differs by gender, with these levels on average higher in women than in men ((34)). Women also, on average, have less IAF than men, suggesting a possible causal link between this higher adipokine concentration and lower IAF accumulation ((35)). Another possible explanation is that adipokine concentrations might reflect adipocyte dysfunction in subcutaneous white adipose tissue. Lower adiponectin concentration and impaired lipolysis owing to endoplasmic reticulum stress have been observed in mice fed a high-fat diet ((36)). This dysfunction may contribute to lesser SCF accumulation and greater deposition in ectopic sites such as the visceral depot. Therefore, low adiponectin level may serve as a marker for early dysfunction leading to future visceral fat accumulation.
Multiple studies have examined the relationship between plasma leptin or adiponectin concentrations and body composition ((26, 33, 37-41)). In several populations, including in adult women, Alaska Natives, nondiabetic participants, and offspring of type 2 diabetes, leptin has been shown to be positively associated with visceral fat and plasma adiponectin shown to be negatively associated with percent body fat, visceral fat, and leptin concentration ((15, 26)). Although these previous studies have demonstrated associations between adipokines and IAF, the temporal sequence of these associations could not be elucidated from these cross-sectional studies. Previous longitudinal analyses on the association between adipokines and change in body composition have examined only single adipokines and therefore could not ascertain whether, when more than one adipokine is studied, these have independent associations with a change in body composition ((13-16, 42)). We were able to confirm this for leptin and adiponectin in multivariate models, but the directions of the associations with Δ IAF were opposite: leptin concentration showed a positive association, while adiponectin concentration showed a negative association.
Interestingly, significant associations between these adipokine concentrations and Δ IAF only emerged after adjustment for confounding factors in multivariable regression analysis. In univariate analyses, neither leptin nor adiponectin concentrations were significantly related to Δ IAF. After adjustment for baseline IAF, a significant association emerged between adiponectin concentration and Δ IAF, with the magnitude of the negative association increased between adiponectin and Δ IAF (Tables 2-3). A common practice is to include only statistically significant variables in univariate analysis into multivariate analysis. However, some variables not significant in the univariate analysis may sometimes become significant in multivariate analysis owing to adjustment for other variables ((43)). Therefore, baseline IAF is an important mediator in this relationship between adipokines and Δ IAF.
The strengths of the present study include direct CT measurements of fat depots that provided an accurate assessment of fat areas at the regions of interest, which allowed us to explore the association between baseline leptin and adiponectin concentrations and future change in regional body fat distribution beyond what is possible using BMI or surface measurements of adiposity. In addition, the prospective design permitted assessment of temporal sequence not possible in cross-sectional research. To our knowledge, this is the first study that has examined the association between both leptin and adiponectin and accumulation of IAF over time. We confirmed individual associations of leptin and adiponectin with Δ IAF by simultaneously including both adipokines in regression models. Our results demonstrate that both adipokines are independent, individual markers for IAF accumulation. Weight change might also have accounted for Δ IAF, as any change in weight would affect both SCF and IAF depots, but we observed independent and opposite associations between Δ IAF and both adipokines even after adjustment for weight change.
However, our study also has some limitations. First, since this study focused entirely on Japanese Americans, these results may not be applicable to other ethnic groups. There are known differences in mean adiponectin concentration in Japanese compared with other ethnic groups. Adiponectin concentrations were lower in a waist-circumference-stratified analysis in 98 US Caucasians compared with 92 Japanese men aged 40 to 49 years without diabetes or coronary heart disease ((44)). A sample of Asian women of whom 53% were Japanese American had significantly lower concentrations of both adipokines in BMI-stratified analysis ((45)). Whether these differences extend to the association between Δ IAF and adipokine concentration will require further research in non-Japanese-ancestry populations. Second, the present findings should be qualified given the observational nature of our study design, which precludes conclusions about causality. In addition, as is true for all observational research, it is possible that unknown confounding factors may be responsible for the associations noted between leptin and adiponectin concentrations and adiposity although we adjusted for many covariates. Finally, because we measured leptin and adiponectin concentrations at baseline only, we were not able to assess whether changes in adiposity were associated with longitudinal changes in leptin and adiponectin concentration. Despite these limitations, to our knowledge our study is unique in that it is the first prospective study demonstrating that leptin and adiponectin independently predict future Δ IAF as measured by imaging.
In conclusion, this longitudinal study provides novel evidence that each baseline adipokine measurement was independently associated with Δ IAF but in different directions, with leptin concentration positively associated and adiponectin concentration negatively associated with visceral fat accumulation.
Acknowledgments
We are grateful to the King County Japanese American community for support and cooperation.
Funding agencies
This work was supported by National Institutes of Health grants DK-31170 and HL-49293, as well as by facilities and services provided by the Diabetes Research Center (DK-017047), the Nutrition Obesity Research Center (DK-035816), and the General Clinical Research Center (RR-000037) at the University of Washington. The funding entities had no role in the conduct of this study or interpretation of its results. VA Puget Sound provided support for the participation of EJB and SEK in this research.
Disclosure
EJB received compensation for serving on an Advisory Board for Bayer AG. The other authors declared no conflict of interest.
Author contributions
SOS, WYF, and EJB researched the data, wrote the manuscript, and provided approval for the final version. SJH, SEK, and DLL contributed to the discussion, reviewed/edited the manuscript, and provided approval for the final version. SOS and EJB are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
