High-Molecular-Weight Adiponectin and the Risk of Type 2 Diabetes in the ARIC Study

Abstract

Context: Adiponectin, synthesized by adipocytes, has been shown to be a predictor of type 2 diabetes. Adiponectin circulates in plasma as three oligomeric isoforms. High-molecular-weight (HMW) adiponectin is thought to be the most biologically active form of adiponectin in terms of glucose homeostasis.

Objective: Our objective was to investigate whether HMW adiponectin is more strongly associated with incident diabetes than is total adiponectin.

Design: A nested case-cohort study was conducted in a population-based cohort of 9740 middle-aged, initially healthy, white and African-American participants of the Atherosclerosis Risk in Communities (ARIC) study followed for up to 9 yr. Plasma total and HMW adiponectin were measured by ELISA in 550 incident diabetes cases and 540 noncases.

Results: Overall hazard ratios (95% confidence intervals) for developing diabetes for those in the fourth (vs. the first) quartile of total adiponectin, HMW adiponectin, and the ratio of HMW to total were 0.40 (0.25–0.64), 0.38 (0.23–0.63), and 0.65 (0.42–0.99), respectively, after adjustment for age, sex, ethnicity, study center, parental history of diabetes, hypertension, body mass index, and waist-to-hip ratio and 0.52 (0.32–0.85), 0.51 (0.31–0.86), and 0.77 (0.50–1.20), respectively, after additional adjustment for inflammation score (a score composed of six inflammation markers) and fasting insulin. When further adjusting for baseline fasting glucose, the graded associations were attenuated substantially and lost their gradation.

Conclusions: In this community-based sample of U.S. adults, higher total and HMW adiponectin concentrations were similarly associated with a lower incidence of diabetes over 9 yr of follow-up.

In this community-based sample of U.S. adults, higher total and high-MW adiponectin concentrations are similarly associated with a lower incidence of diabetes.

Type 2 diabetes is a leading cause of morbidity and mortality in most developed countries, and there is substantial evidence that it is epidemic in many developing and newly industrialized nations. Obesity or overweight is a major modifiable risk factor for type 2 diabetes. Adipose tissue has important endocrine functions including secreting various hormones and cytokines (adipokines) (1,2). Adiponectin, a cytokine exclusively synthesized by adipocytes (3), has been shown to improve insulin sensitivity, increase rates of fatty acid oxidation, and reduce inflammation and vascular injury (1,2). Unlike leptin and many other adipokines, adiponectin’s circulating concentration decreases with increasing adiposity. Observational studies have consistently found that higher total adiponectin concentration is associated with a reduced risk for type 2 diabetes in men and women and in Caucasians, African-Americans, Asians, Samoans, American Indians, and Asian Indians (4,5,6,7,8,9,10,11,12,13). We previously reported that higher total adiponectin concentrations were associated with a lower incidence of diabetes, and this association was of similar magnitude in men and women and in whites and African-Americans in the Atherosclerosis Risk in Communities (ARIC) Study (4). These results suggest that total adiponectin, with its unique antiatherogenic, antiinflammatory, and insulin-sensitizing properties, serves as a nexus between accumulated adipose tissue and the propensity to develop type 2 diabetes (2).

Recently, it was demonstrated that adiponectin exists in plasma in oligomeric complexes, consisting of trimers (low molecular weight), hexamers (medium molecular weight), and large multimers of 12–18 subunits [high molecular weight (HMW)]) (14,15). Studies suggested that the HMW form that mediates insulin sensitization in peripheral tissues may be the most biologically active form of adiponectin in terms of glucose homeostasis (16,17). Only a few epidemiological studies have investigated the association of HMW adiponectin with risk of developing type 2 diabetes; some researchers found that HMW adiponectin and the ratio of HMW to total adiponectin were inversely associated with the risk (8,18,19). Moreover, compared with total adiponectin, HMW adiponectin was more strongly related to the development of type 2 diabetes (18). However, the association between HMW adiponectin and the risk of type 2 diabetes has not yet been investigated in African-Americans and white men.

The aim of the current study was to investigate whether the association between HMW adiponectin and incident diabetes is stronger than that for total adiponectin in African-American and white adults. Additionally, heterogeneity in associations across categories of smoking, inflammatory burden, and other possible effect modifiers was investigated.

Subjects and Methods

Study design

The study is a prospective, stratified, nested case-cohort study within the ARIC study. Between 1987 and 1989, ARIC examined population-based samples of 15,792 residents, white and African-American men and women aged 45–64 yr in four U.S. communities in North Carolina, Mississippi, Minnesota, and Maryland. Incident cases of diabetes were ascertained in three follow-up examinations conducted over 9 yr of follow-up. Human subject research review committees at the involved institutions approved the study, and all participants gave written consent.

A case-cohort design, previously used to investigate the role of total adiponectin, leptin, ferritin, and other plasma biomarkers in the development of diabetes in ARIC (4,20,21), was applied in this study. Before sampling, we excluded 2018 participants with prevalent diabetes, 95 members of ethnic groups other than African-American or white, 853 not returning to any follow-up visit, 26 with no valid diabetes determination at follow-up, six with restrictions on stored plasma use, 12 with missing baseline anthropometrics, 2514 participants in previous ARIC case-control and case-cohort studies involving cardiovascular disease for whom stored plasma was either previously exhausted or held in reserve, 212 for incomplete fasting (<8 h), and 316 with missing information for key covariates (hypertension, smoking, and inflammation markers). A final sample of 9740 individuals (71% of those in the full cohort without diabetes at baseline) was established after exclusions. Of these, 1105 participants (11.3% of the final sample) were ascertained with incident type 2 diabetes during the 9-yr follow-up.

From these eligible members of this baseline cohort, we selected and measured analytes on ethnicity-stratified (50% white, 50% African-American) random samples of both cases of incident diabetes and eligible members of the full cohort (1095 individuals in total). Some participants within the cohort random sample developed diabetes at the end of the follow-up and were part of the case random sample as well. Of those sampled, we excluded five for missing values of total or HMW adiponectin, leaving a total of 1090 subjects, including 550 diabetes cases and 540 noncases for analysis. The cohort random sample contained 4.4% of eligible white participants and 14.6% of eligible African-Americans. Among those with incident diabetes, 40% of whites and 75% of African-Americans were included in the study sample.

Data collection

Glucose was measured at baseline and at follow-up visits by a hexokinase method and fasting serum insulin by nonspecific RIA. Technicians measured waist girth at the umbilical level and hip circumference at the maximum hip girth to obtain the waist-to-hip ratio. We defined parental history of diabetes as a report of diabetes in either parent. The definitions and methods for other baseline measurements [height, weight, smoking status, systolic blood pressure, hypertension, triglycerides, high-density lipoprotein (HDL) cholesterol, insulin, and inflammation markers] have been previously reported (22,23). We created a score to indicate low-grade systemic inflammation ranging from 0–6, attributing one point for a value greater than the median of the cohort sample for each of the six inflammation markers: IL-6, C-reactive protein, orosomucoid, sialic acid, white blood cell count, and fibrinogen (22).

HMW and total adiponectin were measured at a central laboratory on plasma specimens frozen at baseline. These samples, stored for approximately 20 yr at −70 C, were thawed and maintained at 4 C until measured, which was no longer than 24 h later. We measured total and HMW adiponectin by a sandwich ELISA (American Laboratory Products Company, Salem, NH) with a sensitivity of 0.04 ng/ml (24). With pretreatment of proteinase-K, which selectively digested all adiponectin isoforms except HMW, HMW adiponectin and total adiponectin were measured at the same time. Based on blinded quality-control duplicates collected from 39 subjects, reliability coefficients for total adiponectin, HMW adiponectin, and ratio of HMW to total adiponectin were 0.93, 0.98, and 0.70, respectively.

Ascertainment of incident diabetes

Cases were defined on the basis of 1) a reported physician diagnosis, 2) reported use of antidiabetic medications during the past 2 wk, 3) a fasting (≥8 h) glucose value of at least 7.0 mmol/liter, or 4) a nonfasting glucose value of at least 11.1 mmol/liter. The date of diabetes incidence was estimated by linear interpolation using glucose values at the ascertaining visit and the previous one, as previously described (4,20,21,22,23).

Statistical analysis

Weighted analyses were performed on the basis of the case-cohort sampling design to permit statistical estimation and inference relevant to the entire cohort. Weights were defined as the inverse of the ethnicity-specific sampling fractions. Weighted Spearman correlations were applied to describe unadjusted associations between study variables. Total and HMW adiponectin were logarithmically transformed to approximate normality. Weighted analysis of covariance was used to compute adjusted means of total and HMW adiponectin and ratio of HMW to total adiponectin among subgroups of cohort random sample. Based upon their distribution among the cohort random sample, adiponectin measures were divided into quartiles or tertiles. Cox proportional hazards regression was applied to analyze the relation between plasma total adiponectin, HMW adiponectin, and ratio of HMW to total adiponectin and time to onset of type 2 diabetes. Wald tests of interaction terms were used to test heterogeneity in these associations by possible effect modifiers, including sex, ethnicity, body mass index (BMI), inflammation score, impaired fasting glucose, smoking, and drinking.

All statistical testing was performed using two-sided tests with the significance level of type I error (α) set at 0.05. Statistical analyses were performed using the SAS (SAS Institute Inc., Cary, NC) and SUDAAN (Research Triangle Institute, Research Triangle Park, NC) statistical software packages.

Results

The study sample consisted of 1090 participants. The cohort random sample comprised 634 individuals, and the incident diabetes random sample included 527 individuals. Within the cohort random sample, 94 participants developed diabetes during the follow-up time, and 71 of them were included in the incident diabetes random sample as well (thus, the cohort added 23 incident diabetes cases to those in the incident diabetes random sample). The median (interquartile range) of time to diabetes was 3.0 yr (1.7–5.9) for those who developed diabetes, and the median time to censoring was 8.9 yr (8.8–9.0) for those who did not. Among cases, 147 (27%) were white men, 142 (26%) were white women, 68 (12%) were African-American men, and 193 (35%) were African-American women. Among noncases, the corresponding numbers were 115 (21%), 196 (36%), 70 (13%), and 159 (29%), respectively.

Strong positive correlations were observed between total adiponectin, HMW adiponectin, and the ratio of HMW to total adiponectin in the cohort random sample (Table 1). Adiponectin measures were moderately correlated with inflammation markers (absolute values of the correlations were 0.11–0.28 for total adiponectin, 0.09–0.25 for HMW adiponectin, and 0.01–0.18 for the ratio), elements of the metabolic syndrome (0.26–0.43 for total adiponectin, 0.27–0.43 for HMW adiponectin, and 0.15–0.35 for the ratio), and fasting insulin (0.41 for total and HMW adiponectin and 0.30 for the ratio). The correlations were generally negative for all covariates, except for HDL cholesterol, and statistically significant for all variables with P < 0.001, except for the correlation between the ratio of HMW to total adiponectin and sialic acid (P = 0.53). Correlations of the ratio of total to HMW adiponectin were less strong with markers of inflammation than the corresponding correlations of total or HMW adiponectin.

Table 1.

Weighted Spearman correlation coefficients between adiponectin measures and markers of inflammation and metabolic syndrome in the cohort random sample: ARIC study, 1987–1989

Markers	Spearman correlation coefficienta
	Total adiponectin	HMW adiponectin	Ratio of HMW to total adiponectin
HMW adiponectin	0.95
Ratio of HMW to total adiponectin	0.56	0.78
Inflammation markers
Orosomucoid	−0.28	−0.24	−0.10
C-reactive protein	−0.26	−0.25	−0.13
IL-6	−0.19	−0.22	−0.18
White blood cell count	−0.17	−0.17	−0.11
Sialic acid	−0.11	−0.09	−0.01
Fibrinogen	−0.18	−0.18	−0.12
Metabolic syndrome
BMI	−0.30	−0.34	−0.31
Waist circumference	−0.35	−0.40	−0.35
Triglycerides	−0.34	−0.31	−0.15
HDL cholesterol	0.43	0.43	0.29
Systolic blood pressure	−0.26	−0.27	−0.20
Diastolic blood pressure	−0.29	−0.29	−0.20
Fasting glucose	−0.29	−0.30	−0.24
Fasting insulin	−0.41	−0.41	−0.30

The cohort random sample included 634 subjects. 

^aWeighted to reflect stratified sampling strategy. P < 0.001 for all correlations, except for the correlation between the ratio of HMW to total adiponectin and sialic acid (P = 0.53). 

Women had 52, 81, and 19% higher concentrations of total adiponectin, HMW adiponectin, and ratio of HMW to total adiponectin than men, respectively (P < 0.001) (Table 2). Whites had a 50% higher level of total adiponectin, 87% higher level of HMW adiponectin, and 19% higher level of ratio of HMW to total adiponectin than African-Americans (P < 0.001). Participants with lower BMI level had higher concentrations of total and HMW adiponectin and ratio of HMW to total adiponectin (P < 0.001 for difference between BMI categories). Those with a low inflammation score (three or fewer of six markers with above-median values) had a 25% higher mean value of total adiponectin (P < 0.001), 33% higher mean value of HMW adiponectin than those with higher inflammation scores (P < 0.001), and 5% higher mean value of the ratio of HMW to total than those with higher score (P = 0.01). Participants with parental history of diabetes had nonsignificantly lower level of total (P = 0.10) and HMW (P = 0.22) adiponectin, and similar level of the ratio of HMW to total (P = 0.70), compared with those without parental history of diabetes.

Table 2.

Adjusted weighted mean (95% CI) of adiponectin in subgroups of the cohort random sample: ARIC study, 1987–1989

Group	na	Total adiponectin (μg/ml)		HMW adiponectin (μg/ml)		Ratio of HMW to total adiponectin
		Meanb (95% CI)	P	Meanb (95% CI)	P	Mean (95% CI)	P
Overall	634	6.41 (6.17–6.67)		2.56 (2.42–2.70)		0.41 (0.41–0.42)
Age (yr)c
45–54	421	6.19 (5.88–6.51)	0.02	2.44 (2.29–2.61)	0.04	0.41 (0.40–0.42)	0.32
55–64	213	6.84 (6.42–7.28)		2.76 (2.52–3.03)		0.42 (0.40–0.44)
Sex
Women	417	7.45 (7.09–7.83)	<0.001	3.16 (2.95–3.39)	<0.001	0.44 (0.43–0.45)	<0.001
Men	217	4.91 (4.60–5.24)		1.75 (1.60–1.91)		0.37 (0.36–0.38)
Ethnicity
White	338	6.92 (6.60–7.25)	<0.001	2.88 (2.70–3.06)	<0.001	0.43 (0.42–0.44)	<0.001
African-American	296	4.62 (4.34–4.93)		1.54 (1.38–1.70)		0.36 (0.34–0.37)
BMI (kg/m2)
<25	198	7.20 (6.70–7.74)	<0.001	3.06 (2.78–3.36)	<0.001	0.44 (0.43–0.46)	<0.001
≥25 and <30	258	6.19 (5.86–6.54)		2.45 (2.27–2.65)		0.41 (0.40–0.42)
≥30	178	5.54 (5.11–6.00)		1.98 (1.77–2.22)		0.37 (0.35–0.39)
Parental history of diabetes
Yes	150	6.04 (5.59–6.53)	0.10	2.40 (2.14–2.69)	0.22	0.41 (0.39–0.43)	0.70
No	484	6.52 (6.23–6.82)		2.60 (2.45–2.76)		0.42 (0.41–0.43)
Hypertension
Yes	212	5.85 (5.34–6.41)	0.02	2.21 (1.94–2.52)	0.01	0.39 (0.37–0.41)	0.02
No	422	6.61 (6.32–6.90)		2.68 (2.52–2.84)		0.42 (0.41–0.43)
Smoking status
Never smoker	304	6.43 (6.10–6.78)	0.57	2.54 (2.36–2.73)	0.87	0.41 (0.40–0.43)	0.64
Former smoker	189	6.56 (6.09–7.07)		2.61 (2.35–2.90)		0.41 (0.39–0.43)
Current smoker	141	6.16 (5.62–6.75)		2.50 (2.22–2.83)		0.42 (0.40–0.44)
Drinking status
Never drinker	201	6.44 (5.92–7.00)	0.83	2.52 (2.24–2.83)	0.85	0.40 (0.38–0.42)	0.46
Former drinker	101	6.18 (5.40–7.07)		2.46 (2.06–2.93)		0.42 (0.39–0.45)
Current drinker	332	6.46 (6.15–6.78)		2.59 (2.42–2.77)		0.42 (0.41–0.43)
Systemic inflammation score
≤3	360	6.98 (6.66–7.32)	<0.001	2.85 (2.67–3.04)	<0.001	0.42 (0.41–0.44)	0.01
>3	274	5.57 (5.23–5.94)		2.14 (1.95–2.34)		0.40 (0.38–0.41)

Means were adjusted for age, sex, ethnicity, and center through weighted analysis of covariance accounting for ethnicity-stratified sampling. 

^aNumber of participants in each subgroup within cohort random sample. 

^bGeometric mean (95% CI). 

^cAge was divided into two groups (45–54 or 55–64) with same age range. 

In multivariate models adjusting initially for age, sex, ethnicity, center, hypertension, and parental history of diabetes, both total and HMW adiponectin levels were strongly and inversely associated with the risk of developing incident diabetes (Table 3). Hazard ratios (HR) comparing the highest with the lowest quartiles were 0.22 [95% confidence interval (CI) = 0.14–0.34] for total adiponectin and 0.20 (95% CI = 0.13–0.32) for HMW adiponectin. Associations for total and HMW adiponectin remained graded, protective, and significant with additional adjustment for BMI, waist-to-hip ratio, inflammation score, and baseline fasting insulin. Further adjustment for baseline fasting glucose substantially attenuated the association, especially for the HR comparing extreme quartiles, which were 0.82 (95% CI = 0.48–1.42) for total adiponectin and 0.75 (95% CI = 0.43–1.31) for HMW adiponectin. These results were not materially different when waist rather than waist-to-hip ratio was used as an adjustment variable.

Table 3.

Hazard ratios (95% confidence intervals) for developing type 2 diabetes, by quartiles of adiponectin concentrations: ARIC study, 1987–1998

Variable	Quartile of adiponectin concentration				P for trenda
	1 (low)	2	3	4 (high)
Total adiponectin (μg/ml)
Weighted median (range)	3.48 (0.57–4.39)	5.33 (4.40–6.26)	7.22 (6.27–8.38)	10.61 (8.39–25.42)
Model 1	1.00 (Referent)	0.59 (0.43–0.82)	0.32 (0.22–0.48)	0.22 (0.14–0.34)	<0.001
Model 2	1.00 (Referent)	0.72 (0.51–1.01)	0.44 (0.29–0.67)	0.40 (0.25–0.64)	<0.001
Model 3	1.00 (Referent)	0.71 (0.51–1.00)	0.45 (0.29–0.69)	0.46 (0.29–0.74)	<0.001
Model 4	1.00 (Referent)	0.72 (0.51–1.01)	0.49 (0.32–0.76)	0.52 (0.32–0.85)	<0.001
Model 5	1.00 (Referent)	0.77 (0.52–1.14)	0.52 (0.31–0.87)	0.82 (0.48–1.42)	0.07
HMW adiponectin (μg/ml)
Weighted median (range)	1.04 (0.03–1.60)	2.11 (1.61–2.53)	3.14 (2.53–3.80)	5.41 (3.82–20.15)
Model 1	1.00 (Referent)	0.58 (0.42–0.82)	0.33 (0.23–0.49)	0.20 (0.13–0.32)	<0.001
Model 2	1.00 (Referent)	0.67 (0.47–0.96)	0.52 (0.35–0.77)	0.38 (0.23–0.63)	<0.001
Model 3	1.00 (Referent)	0.69 (0.48–0.98)	0.54 (0.36–0.80)	0.45 (0.27–0.75)	<0.001
Model 4	1.00 (Referent)	0.70 (0.49–1.00)	0.59 (0.39–0.88)	0.51 (0.31–0.86)	0.003
Model 5	1.00 (Referent)	0.75 (0.50–1.12)	0.69 (0.41–1.15)	0.75 (0.43–1.31)	0.15
Ratio of HMW to total adiponectin
Weighted median (range)	0.29 (0.02–0.34)	0.37 (0.34–0.41)	0.44 (0.41–0.48)	0.53 (0.48–0.85)
Model 1	1.00 (Referent)	0.72 (0.52–0.99)	0.58 (0.40–0.83)	0.37 (0.25–0.55)	<0.001
Model 2	1.00 (Referent)	0.88 (0.62–1.25)	0.83 (0.56–1.21)	0.65 (0.42–0.99)	0.06
Model 3	1.00 (Referent)	0.85 (0.60–1.22)	0.81 (0.55–1.19)	0.66 (0.43–1.03)	0.07
Model 4	1.00 (Referent)	0.96 (0.67–1.37)	0.91 (0.62–1.34)	0.77 (0.50–1.20)	0.29
Model 5	1.00 (Referent)	1.06 (0.71–1.60)	1.00 (0.63–1.59)	1.13 (0.70–1.81)	0.77

All subjects (n = 1090) were included in the analysis. Model 1 was adjusted for age, sex, ethnicity, center, hypertension, and parental history of diabetes. Model 2 was additionally adjusted for BMI and waist-to-hip ratio. Model 3 was additionally adjusted for inflammation score (based on the number of six inflammation markers having an above-median value). Model 4 was additionally adjusted for fasting insulin. Model 5 was additionally adjusted for fasting glucose. 

^aP for trend is the P value of testing a linear trend in quartile number of total and HMW adiponectin and the ratio of HMW to total adiponectin. 

The ratio of HMW to total adiponectin also showed a statistically significantly inverse association with risk of developing diabetes, after initial adjustment (HR comparing extreme quartiles = 0.37; 95% CI = 0.25–0.55) (Table 3). This association was attenuated and became statistically nonsignificant after further sequential adjustment for BMI, waist-to hip ratio, inflammation score, fasting insulin, and fasting glucose. Including total adiponectin in the models substantially attenuated the associations for the ratio of HMW to total adiponectin and these associations were nonsignificant in all models (data not shown).

Restricted to cases who were defined by physician’s diagnosis and diabetic medication use, the direction and magnitude of association between total and HMW adiponectin and incident diabetes was similar to that seen in analysis of all cases, including undiagnosed cases; redefining the time to incident diabetes by using the time points when diabetes was ascertained at visit, almost identical associations were observed (data not shown).

For HMW adiponectin, the slightly larger protective association with incident diabetes for African-Americans [HR (third vs. first tertile) = 0.46; 95% CI = 0.24–0.86] than for whites [HR (third vs. first tertile) = 0.63; 95% CI = 0.38–1.04] was not statistically significant (P for interaction = 0.59). Associations between HMW adiponectin and diabetes were of similar strength in men and women, across categories of BMI and inflammation score, in those with or without baseline impaired fasting glucose, and by smoking and drinking status (P for interaction >0.05). Although the different assay technique and a different selection of covariates for the final model make this report somewhat different from our group’s previous one (4), total adiponectin here, as before, showed a notably greater protective association in those with a lesser (<3) than a greater (>3) inflammation score (HR = 0.25 and 95% CI = 0.14–0.46 vs. 0.70 and 0.39–1.25; P = 0.02). For the ratio of HMW to total adiponectin, no significant heterogeneity was seen except for the associations stratified by baseline impaired fasting glucose status; the protective association of the ratio with incident diabetes was absent for those who had impaired fasting glucose at baseline [P for interaction= 0.02; HR (third vs. first tertile) = 1.44; 95% CI = 0.75–2.76].

Discussion

In the current study, high baseline concentrations of both total and HMW adiponectin were found to be associated with a substantially reduced risk for incident diabetes over 9 yr of follow-up in a cohort of initially healthy middle-aged African-American and white men and women, representative of four U.S. communities. These associations were independent of age, sex, ethnicity, hypertension, parental history of diabetes, the obesity indexes of BMI and waist-to-hip ratio, systemic inflammation score, and fasting insulin. When further adjusting for fasting glucose, the protective associations lost their graded nature and became nonsignificant. Overall, this protective association was of similar strength for total and HMW adiponectin and was weaker for the ratio of HMW to total adiponectin than for total and HMW adiponectin.

Although HMW adiponectin is recognized as a predictor of risk of developing type 2 diabetes and even a stronger predictor than total adiponectin in certain populations (8,18), the association between HMW adiponectin and the risk of type 2 diabetes has not, to our knowledge, been investigated in African-Americans and white men. The strengths of our study also include its case cohort study design, long follow-up, large sample size, a cohort including both men and women, and comprehensive information on lifestyle and biochemical variables. Sensitivity analysis restricting to cases defined by physician’s diagnosis and diabetic medication use, which yielded a similar direction and magnitude of association between total and HMW adiponectin and incident diabetes, further confirm the robustness of the study findings. Several potential limitations of this study should also be acknowledged. First, as in other epidemiological studies, we adjusted for known confounding variables; however, residual confounding caused by imperfectly measured or unmeasured confounders may still exist in the current study, and we could not completely eliminate them. Second, our blood samples were collected at baseline and stored for about 20 yr; this long sample storage time may have induced a bias toward the null and perhaps made it more difficult to differentiate HMW vs. total adiponectin. Third, potential selection bias, owing either to a participant not returning for follow-up or not having samples available for measurement, could possibly have affected the results of the study. However, we have little a priori reason to believe that that there are significant differences in the magnitude of the association between total and HMW adiponectin and incident diabetes among those lost to follow-up or without available samples.

We did not find that the association between HMW adiponectin and risk of type 2 diabetes was stronger than that for total adiponectin, contrary to our initial hypothesis. Our hypothesis was based on in vitro and in vivo data suggesting that HMW adiponectin may be the major biologically active isoform of adiponectin in relation to glucose homeostasis (16,17,25). For example, Kobayashi and colleagues (1) found that only HMW adiponectin and not other isoforms suppressed apoptosis in cultured endothelial cells (26). In another study, individuals with a mutation in the adiponectin gene (G90S) exhibited lower levels of HMW adiponectin and diabetes. Epidemiological studies showed that HMW adiponectin correlated better with glucose tolerance than total serum adiponectin (27,28). Additionally, the ratio of HMW to total adiponectin appeared to be closely correlated with insulin sensitivity (14).

In contrast to our study, one prospective epidemiological study in Japanese-Americans found that HMW adiponectin, using the same assay technique here applied, was more strongly associated with progression to type 2 diabetes than total adiponectin; HR in the lowest vs. highest tertile were 1.79 (95% CI = 1.01–3.17) for total and 2.49 (95% CI = 1.34–4.63) for HMW adiponectin (18). In another prospective study conducted among white women, also using the same assay technique here applied, researchers found that the ratio of HMW to total adiponectin was related to the risk of developing diabetes independent of total adiponectin (odds ratio comparing the highest with the lowest quintiles = 0.45; 95% CI = 0.31–0.65), which suggested that the relative proportion of HMW plays an important role in diabetes pathogenesis (8). Reasons for variability in results across studies are unclear, but the relatively low reliability coefficient (0.70) for the ratio of HMW to total adiponectin could be an explanation for the weaker associations between the ratio and incident diabetes in our study. We did not observe a significant interaction by race in the association between HMW and risk of diabetes; however, little research on adiponectin isoforms has been conducted in African-Americans and other populations at high risk for diabetes. In one study, HMW adiponectin was highly correlated with multiple metabolic syndrome traits in European-Americans, but low-molecular-weight adiponectin and trimer forms were most highly correlated with metabolic traits in African-Americans (29).

In this study, further adjustment for baseline fasting glucose produced nonsignificant and nonmonotonic associations of both total and HMW adiponectin with incident diabetes. One potential underlying reason is that baseline fasting glucose, likely reflecting adiponectin’s protective effect, is on the causal pathway of the association between total and HMW adiponectin and incident diabetes; adjustment for fasting glucose will potentially bias the associations between total and HMW adiponectin with incident diabetes. Although the exact mode of action of adiponectin on glucose metabolism is not fully understood, total adiponectin or its proteolytic fragments has been shown to decrease the postprandial rise of plasma free fatty acids and improve postabsorptive insulin-mediated suppression of hepatic glucose output (30). Replenishment of the fat-derived hormone adiponectin can significantly reduce hyperglycemia and reverse insulin resistance in murine models of obesity and type 2 diabetes (31).

In conclusion, both high total and HMW adiponectin concentrations are strongly associated with lower risk for incident type 2 diabetes, in African-American and white men and women, independent of BMI, systemic inflammation and other diabetes risk factors. The strength of this protective association with incident type 2 diabetes was similar for total and HMW adiponectin. The ratio of HMW to total adiponectin is associated with the risk of developing type 2 diabetes before further adjustment for BMI and waist-to-hip ratio, suggesting that the inverse association between the relative proportion of HMW adiponectin and type 2 diabetes was largely explained by adiposity. Given the small number of prospective epidemiological studies to date and conflicting findings, additional research in other populations is needed to further characterize the relation of adiponectin isoforms and diabetes incidence. A better understanding of the pathophysiological mechanisms linking adiponectin and its isoforms to type 2 diabetes is essential for developing effective strategies to prevent and treat diabetes.