Measurement of the High–Molecular Weight Form of Adiponectin in Plasma Is Useful for the Prediction of Insulin Resistance and Metabolic Syndrome

The subjects of this study were 298 patients admitted to Tokyo University Hospital for the treatment of diabetes or coronary angiography. The present study was conducted according to the principles outlined in the Declaration of Helsinki. Informed consent was obtained from each of the study subjects. The protocol of the study was approved by the ethics review committee of Tokyo University School of Medicine.

Height, weight, hip, waist, fasting plasma glucose, serum insulin, serum total cholesterol, triglyceride, HDL cholesterol, and LDL cholesterol levels were measured the morning after an overnight fast. We modified the diagnostic criteria of the National Cholesterol Education Program Adult Treatment Panel III for metabolic syndrome (26) by adopting a cutoff level for the waist circumference of 85 cm in men and 90 cm in women, in accordance with the recommendations for the Japanese population by the International Diabetes Federation (IDF) (27). The homeostasis model assessment of insulin resistance (HOMA-IR) (28) was determined in 171 subjects (110 men and 61 women); however, it could not be assessed in subjects being treated with insulin or in some of the subjects admitted for coronary angiography. We defined subjects with a HOMA-IR of >2.5 as having insulin resistance. This cutoff has been adopted in the Japanese guideline for the treatment of diabetes. We analyzed the relationship between total adiponectin levels or HMWR values and severity and extent of coronary atherosclerosis using the score by Gensini (29). Subjects with a history of percutaneous coronary intervention were excluded from this analysis. Because thiazolidinediones have been reported to exhibit direct antirestenotic effects in the vasculature (30) and have protective effects against coronary artery disease (31), after the prescription of thiazolidinediones was incorporated into the model in addition to the conventional risk factors for atherosclerosis, stepwise regression analysis was done to analyze the relationship between HMWR values and Gensini score. The measurement method for plasma HMW adiponectin level is described elsewhere (H.E., O.M., H.Y., K.H., T.Y., T.K., unpublished observations) and coefficients of variations of the assay were 5.3% for total adiponectin and 3.3% for HMW adiponectin level.

The values of the clinical parameters were expressed as means ± SD. All statistical analyses were performed using JMP for Windows software (version 4.0; SAS Institute, Cary, NC). The significance of differences in plasma total adiponectin levels, HMW adiponectin levels, and HMWR values was analyzed by ANOVA. ROC curves were plotted and compared using the Stata software (College Station, TX). P values <0.05 were considered to denote statistical significance.

As shown in online appendix Table 1 (available at http://care.diabetesjournals.org), there is no difference in the proportion of the subjects who took at least one of the drugs potentially affecting plasma adiponectin levels, such as thiazolidinediones, biguanide, ACE inhibitors, and angiotensin receptor blockers, between subjects with and without metabolic syndrome (68 [42.5%] vs. 54 [39.1%], P = 0.686).

Women had higher plasma total adiponectin levels (5.59 ± 0.31 vs. 4.55 ± 0.22 μg/ml, P = 0.0069), HMW adiponectin levels (2.19 ± 0.14 vs. 1.54 ± 0.11 μg/ml, P = 0.0003), and HMWR values (35.9 ± 1.1 vs. 29.9 ± 0.8%, P = 0.00001) than men. There was an inverse correlation between BMI and plasma total adiponectin levels (r = −0.29, P = 0.0001), HMW adiponectin levels (r = −0.27, P = 0.0001), and HMWR values (r = −0.13, P = 0.012). In multivariate analysis taking into account BMI, sex, and the interaction between sex and BMI, sex and BMI were independently correlated with plasma total adiponectin levels (P = 0.0001 and 0.0025, respectively), plasma HMW adiponectin levels (P = 0.0001 and 0.0001, respectively), and HMWR values (P = 0.0075 and 0.0001, respectively). However, since there was no interaction between sex and BMI for these three different measurements of plasma adiponectin (P = 0.29, 0.23, and 0.34, respectively), the inverse correlation between BMI and plasma total adiponectin level, plasma HMW adiponectin levels, or HMWR values were not affected by sex. Plasma total adiponectin level, plasma HMW adiponectin levels, and HMWR values did not vary with age in either sex (data not shown).

Insulin resistance is closely linked with metabolic syndrome and is often observed even in the early stage of metabolic syndrome (32). We investigated the correlation between insulin resistance and each of the three different measurements of plasma adiponectin. The total adiponectin level was inversely correlated with the HOMA-IR (P = 0.0280) in 171 men and women, consistent with the results of previous studies (11,12); the HMW adiponectin levels (P = 0.0035) and HMWR values (P = 0.0008) were also inversely correlated with HOMA-IR. In 110 men, there was a significant inverse correlation between HOMA-IR and plasma HMW adiponectin levels (P = 0.0083) and HMWR values (P = 0.0010), while there was only a tendency toward inverse correlation between HOMA-IR and plasma total adiponectin levels (P = 0.0613). In 61 women, tendencies were found for decreased plasma total adiponectin levels and for HMW adiponectin levels or HMWR values to be associated with increased HOMA-IR values. HMWR values were significantly and inversely correlated with HOMA-IR, even after adjusting for age, sex, and BMI (P = 0.0304) or adjusting for BMI alone (P = 0.0389) in 171 men and women, suggesting that correlation between HMWR values and HOMA-IR was independent of the association between BMI and HOMA-IR. In contrast, the statistically significant association between total adiponectin and HOMA-IR disappeared after adjusting for age, sex, and BMI (P = 0.227) or adjusting for BMI alone (P = 0.249) in 171 men and women. These results are again consistent with the superiority of HMWR values over total adiponectin level to predict the presence of insulin resistance. We then plotted ROC curves to compare the power of plasma total adiponectin level and HMWR value to predict the presence of insulin resistance (Fig. 1). The AUC of HMWR values was significantly larger than that of plasma total adiponectin levels (0.713 [95% CI 0.620–0.805] vs. 0.615 [0.522–0.708], P = 0.0160) (Fig. 1), suggesting that the HMWR value had better predictive power for the prediction of insulin resistance than plasma total adiponectin level. When a cutoff value of 35% was used, the HMWR value predicted the presence of insulin resistance with a sensitivity of 72% and specificity of 66%. On the other hand, at a cutoff level of 4.2 μg/ml, plasma total adiponectin diagnosed insulin resistance with a sensitivity of 56% and specificity of 63%. When stratified according to sex, there was a significant difference between the AUCs for plasma total adiponectin levels and HMWR values in men (0.713 [0.605–0.821], P = 0.048 vs. 0.624 [0.514–0.733], P = 0.048), while we could detect no such difference in women (0.794 vs. 0.665, P = 0.09).

We then investigated the association between the number of risk factors defining the prediction of metabolic syndrome (see research design and methods) in the subjects and plasma total adiponectin levels, HMW adiponectin levels, and HMWR values. The plasma total adiponectin levels (P = 0.0001), HMW adiponectin levels (P = 0.0001), and HMWR values (P = 0.0001) decreased as the number of risk factors present increased; this tendency was observed irrespective of sex (data not shown).

We performed an ROC analysis to quantify the power of the HMWR values to make a prediction of metabolic syndrome. ROC curves for plasma total adiponectin levels and HMWR values to discriminate between subjects with and without metabolic syndrome were plotted for total subjects (data not shown), men (Fig. 2A), and women (Fig. 2B). The areas under the curve for plasma total adiponectin level and HMWR values were then compared to determine whether the HMWR value had a better predictive power for the prediction of metabolic syndrome than plasma total adiponectin level. The AUC for the HMWR values was significantly larger than that for plasma total adiponectin levels in all subjects (0.780 [95% CI 0.736–0.824] vs. 0.692 [0.632–0.751], P = 0.039) (data not shown), in men (0.806 [0.747–0.865] vs. 0.730 [0.660–0.800], P = 0.0025) (Fig. 2A), and in women (0.743 [0.659–0.828] vs. 0.637 [0.532–0.742], P = 0.0458) (Fig. 2B). When a cutoff value of 32.0% for men and 40.2% for women was adopted to maximize the sensitivity (%) plus specificity (%), HMWR value predicted the presence of metabolic syndrome with a sensitivity of 83.2% and specificity of 62.0% in men and a sensitivity of 77.4% and specificity of 55.0% in women.

We have undertaken the subgroup analyses in both diabetic and nondiabetic subjects. In both groups, AUCs for the HMWR values to predict metabolic syndrome were larger than those for total adiponectin levels (type 2 diabetic group: 0.815 [95% CI 0.755–0.875] vs. 0.676 [0.616–0.736], P = 0.0013; nondiabetic group: 0.843 [0.743–0.943] vs. 0.751 [0.651–0.851], P = 0.00475).

We have also plotted ROC curves for total adiponectin levels and HMWR values to predict metabolic syndrome defined by newly announced IDF criteria (27). Indeed, HMWR values were more predictive of metabolic syndrome defined by IDF than total adiponectin levels in total subjects (0.733 [95% CI 0.693–0.773] vs. 0.680 [0.640–0.720], P = 0.0037).

The HMWR values varied substantially even among subjects with similar total adiponectin levels in plasma. The present ROC analysis suggested the possibility that the HMWR value may be useful for predicting the presence of metabolic syndrome among subjects with similar plasma levels of total adiponectin. We examined whether the HMWR values were different between subjects with and without metabolic syndrome after stratifying the subjects into quartiles of plasma total adiponectin levels. HMWR values were significantly lower in all the quartiles of total adiponectin in subjects with metabolic syndrome than in those without, suggesting that the HMWR value may be useful for the prediction of metabolic syndrome, irrespective of plasma total adiponectin level.

We analyzed the relationship between total adiponectin levels or HMWR values and severity and extent of coronary atherosclerosis using the score by Gensini (29). Gensini scores were significantly associated with age (P = 0.0022), prescription of thiazolidinediones (P = 0.0443), and HMWR values (P = 0.0437). Subjects with lower HMWR value had higher Gensini score, suggesting that HMWR value might be associated with the onset and development of coronary artery disease independently of age and prescription of thiazolidinediones. Total adiponectin level was not correlated with Gensini score before and after adjustment for the conventional risk factors for atherosclerosis and prescription of thiazolidinediones.

Prediction of metabolic syndrome, defined by the presence of a cluster of metabolic abnormalities, including impaired glucose metabolism, high BMI and abdominal fat distribution, dyslipidemia, and hypertension, is very important because of its association with the subsequent development of type 2 diabetes and cardiovascular disease (28). Because of the epidemic of obesity and a sedentary lifestyle worldwide, metabolic syndrome is becoming increasingly commonly recognized. According to the National Cholesterol Education Program (NCEP) definition, roughly one-fourth of middle-aged men and women in the U.S. have metabolic syndrome (33). Development of a method for convenient prediction of metabolic syndrome in daily clinical practice presents a major challenge for physicians and public health policy makers facing the epidemic of obesity and a sedentary lifestyle.

There is a mounting body of evidence to suggest that adiponectin is an insulin-sensitizing hormone and that the plasma level of this hormone is the best predictor of the subsequent development of type 2 diabetes among the various plasma biomarkers (34). Recently, however, it has been reported that adiponectin forms a wide range of multimers in plasma and that mutations in the adiponectin gene that inhibit the formation of HMW adiponectin are closely associated with the subsequent development of type 2 diabetes. Therefore, it was considered that HMW adiponectin value might be an attractive biomarker for the prediction of insulin resistance and metabolic syndrome. Indeed, the present study demonstrated that the AUC of HMWR values was significantly larger than that of plasma total adiponectin levels and that the sensitivity of the HMWR value for predicting the presence of metabolic syndrome reached 80%. Thus, this study is the first to demonstrate that HMWR value is more closely associated with insulin resistance and the presence of metabolic syndrome than plasma total adiponectin level.

The present study provided evidence of the usefulness of a newly developed method of measurement of plasma HMW adiponectin level as a convenient and sensitive biomarker for the prediction of insulin resistance and metabolic syndrome.

This study was supported by the Program for Promotion of Fundamental Studies in Health Sciences of the National Institute of Biomedical Innovation (to T.K.), a grant-in-aid for The 21st Century COE Program (to R.N.), and a grant-in-aid from Health and Labor Sciences Research Grants, Translational research (to T.K.).

We thank S. Miyama for technical assistance.