Serum Adiponectin and Renal Dysfunction in Men With Type 2 Diabetes

The Health Professionals’ Follow-Up Study (HPFS) was established in 1986 when 51,529 U.S. male health professionals, aged 40–75 years at study initiation, returned a mailed questionnaire providing information about diet, lifestyle factors, and medical history (25). Participants were mailed follow-up questionnaires every 2 years to update information. In 1993–1994, blood samples were collected and frozen (−130°C) from a subset of these participants (n = 18,159) as previously described (8). Demographic and clinical characteristics at baseline were similar between men who provided blood samples and those who did not.

Diabetes was first identified by self-report on a biennial questionnaire and confirmed by a Diabetes Supplemental Questionnaire in 2000; the validity of the Diabetes Supplemental Questionnaire in confirming diabetes has been demonstrated in the HPFS cohort (26). The HPFS diabetes blood cohort consists of 1,000 men with confirmed diabetes diagnosed before study entry or newly diagnosed up through June 1998 who provided a blood sample in 1993–1994. Exclusion criteria for the current analyses were 1) age of onset of diabetes ≤25 years of age (to attempt to restrict the study to type 2 diabetes) (n = 31), 2) reported date of diabetes diagnosis after the date of blood draw (n = 224), 3) participants who reported on the Diabetes Supplemental Questionnaire that they were on dialysis (n = 9) or had a kidney transplant (n = 1), 4) serum creatinine >5.0 mg/dl (n = 1), and 5) serum creatinine ≤0.5 mg/dl (felt to be physiologically implausible) (n = 1). After these exclusions, 733 men were available for analysis. This study was approved by the institutional review boards at the Harvard School of Public Health and the Brigham and Women’s Hospital.

Adiponectin was measured by competitive radioimmunoassay using a commercial reagent set from Linco Research (St. Louis, MO). A previous study (27) demonstrated that adiponectin measurements are highly stable and reproducible under transport conditions and in frozen whole-blood samples. The coefficient of variation was 3.4%. A1C was measured by turbidimetric immunoinhibition using hemolyzed whole blood or packed red cells with a coefficient of variation of 7.5%. Plasma creatinine was measured by a modified kinetic Jaffe reaction with a coefficient of variation of 22%.

Race and height were initially reported on the 1986 questionnaire. Other clinical and lifestyle variables (hypertension, weight, physical activity, cigarette smoking, and medication use) were derived from data from the 1994 questionnaire, which included data closest to the time of blood draw. BMI was calculated by weight in kilograms divided by the square of height in meters. A weekly metabolic equivalent (MET) score was calculated from physical activity questions. Cardiovascular disease (myocardial infarction, coronary artery bypass grafting, or angina) was confirmed by medical record review (28).

Renal dysfunction was defined as an eGFR <60 ml/min per 1.73 m² as calculated by the four-variable MDRD (Modification of Diet in Renal Disease) equation (eGFR [ml/min per 1.73 m²] = 186 × [PCr {mg/dl}]^−1.154 × [age]^−0.203 × [1.21 if subject is black]) (29). We also examined estimated creatinine clearance by the Cockcroft-Gault equation as the measure of kidney function (30). Of note, because the Cockcroft-Gault results were noticeably influenced by weight, as expected, we chose to present data only with eGFR by the MDRD equation.

The Wilcoxon signed-rank test was used for comparisons of continuous variables. Spearman correlation coefficients were calculated for pairs of continuous variables. Logistic regression was used to calculate odds ratios for adiponectin levels with eGFR <60 ml/min per 1.73 m² as the outcome. Multivariate models were adjusted for age (continuous, years), hypertension (yes/no), BMI (continuous), cigarette smoking status (never, past, or current), physical activity (quartiles, METs/week), duration of type 2 diabetes (quartiles, years), measured A1C (quartiles), and cardiovascular disease (yes/no). ACE and statin use did not change point estimates when included in the model and were removed. All analyses were performed with SAS software version 8.2 (SAS Institute, Cary, NC).

Characteristics of the HPFS diabetes cohort are presented in Table 1. Participants had a median age of 67 years, were mostly Caucasian, and had been carrying the diagnosis of diabetes for 9 years at the time of blood draw. The majority were overweight (BMI ≥25 kg/m²), almost half were hypertensive, one-quarter had cardiovascular disease, the median serum creatinine was 1.0 mg/dl, and median eGFR was 78 ml/min per 1.73 m².

By Spearman correlation, adiponectin was positively correlated with age (r = 0.19, P < 0.001) and inversely correlated with weight (r = −0.18, P < 0.001) and BMI (r = −0.24, P < 0.001) but not with eGFR (r = 0.01, P = 0.76). There was no association between adiponectin and serum creatinine (r = −0.04, P = 0.25) or adiponectin and eGFR when considering only those with eGFR <60 ml/min per 1.73 m² (r = −0.083, P = 0.44). Adiponectin was also significantly associated with serum LDL (r = 0.10, P = 0.008), triglycerides (r = −0.33, P < 0.001), HDL (r = 0.41, P < 0.001), and with A1C (r = 0.09, P = 0.007).

Using multivariate logistic regression, men in each of the upper three quartiles of adiponectin had a reduced odds for having eGFR <60 ml/min per 1.73 m² when compared with the lowest quartile (Table 2). Because the results suggested a threshold rather than a graded association, we combined the upper three quartiles of serum adiponectin and found that the odds ratio for kidney dysfunction in these individuals was 0.48 (95% CI 0.28–0.81). These results were unchanged after we adjusted for individual lipid markers by quartiles (odds ratio for those with adiponectin in the upper three quartiles remained significant and changed by <15% in all lipid analyses) (data not shown).

We found that serum adiponectin was inversely associated with presence of renal dysfunction in men with type 2 diabetes, the majority of whom had well-preserved eGFR (87% had eGFR >60 ml/min per 1.73 m²). There appeared to be a threshold in that those with an adiponectin level higher than the first quartile all had similar decreased odds ratios for renal dysfunction. Because of the reported associations between dyslipidemia and adiponectin, we also adjusted for these factors in our multivariate models and found that the relation between adiponectin and renal dysfunction remained independent and unchanged.

We adjusted for several potential confounders in the relation between adiponectin and eGFR, including age, obesity, and hypertension. Consistent with our findings that adiponectin and age were positively correlated, previous investigations have reported that plasma adiponectin is higher in elderly men and women aged >70 years when compared with younger individuals (31). The relation between adiponectin and age appears stronger in diabetic subjects (Spearman r = 0.44) than in nondiabetic subjects (Spearman r = 0.15) (32). Serum adiponectin is decreased in obesity (10,32,33), in the presence and absence of diabetes, and in hypertension (34). By adjusting for these covariates, we found that the inverse association between serum adiponectin quartiles and presence of renal dysfunction was modestly strengthened.

There is a growing literature on adiponectin and renal disease. Adiponectin is 2.5 times higher in hemodialysis patients (15.0 vs. 6.3 μg/ml, P < 0.0001) (16) and three times higher in pediatric peritoneal dialysis patients (35) when compared with healthy control subjects. In predialysis individuals, two cross-sectional studies (32,36) of adiponectin and renal function in type 2 diabetic subjects have reported a positive association between adiponectin and renal function, in contrast to our study. Although both studies appropriately adjusted for potential confounding by age and BMI, the study populations were very different from ours, making direct comparisons of the results difficult. The report by Looker et al. (32) included 1,069 Pima Indians in contrast to the mostly Caucasian men in our cohort. The study by Guebre-Egziabher et al. (36) consisted of only 48 patients with a mean inulin GFR of 53.5 ml/min per 1.73 m², which is much lower than the eGFR of our study cohort. We did not observe an association, however, between adiponectin and eGFR even when we restricted our analyses to those with eGFR <60 ml/min per 1.73 m². Another study analyzed 543 type 1 diabetic subjects and found an adjusted inverse association between adiponectin and creatinine clearance estimated by the Cockcroft-Gault equation of −0.33 ml/min difference in creatinine clearance for every 1 SD increase in adiponectin levels (37). Our study included only type 2 diabetic subjects, however, and associations between renal clearance and adiponectin may not be the same in insulin-resistant states compared with insulin-deficient ones such as type 1 diabetes. Similarly, although a study of 227 nondiabetic renal patients reported a significant inverse association between adiponectin and directly measured GFR (r = −0.25, P < 0.01) (7), the association between adiponectin and renal clearance could be different in type 2 diabetic subjects because adiponectin levels have been consistently reported to be inversely associated with insulin resistance and are therefore significantly lower in type 2 diabetes.

An alternative explanation for our contrasting findings is that this may be a chance finding, especially in light of the fact that we saw no statistically significant association between adiponectin and eGFR by Spearman correlation. It should be noted, however, that we did not find this result through multiple testing but instead entered into this project with an a priori hypothesis that adiponectin and renal clearance would be directly associated because of the existing literature on inverse association between adiponectin and vascular disease.

One potential explanation for the inverse association observed is that adiponectin reduces vascular disease. Supporting evidence for the role of adiponectin in decreasing vascular endothelial dysfunction is accruing. We previously have reported that adiponectin was positively correlated with HDL cholesterol (Spearman r = 0.42, P < 0.01) and negatively correlated with triglycerides (r = −0.38, P < 0.01), apoprotein B (r = −0.19, P < 0.01), C-reactive protein (r = −0.18, P < 0.01), and fibrinogen (r = −0.18, P < 0.01), which was independent of A1C and HDL in adjusted analyses in these diabetic men (8), suggesting that dyslipidemia and inflammation might be attenuated when serum adiponectin is higher. Similar inverse associations between adiponectin and lipids and inflammatory markers have been reported for diabetic female participants in the Nurses’ Health Study (9). Adjusting for quartiles of lipid biomarkers in our analyses, however, did not influence the results.

The potential benefits of increasing serum adiponectin levels on decreasing future risk for renal function decline remains to be determined. As a circulating anti-inflammatory molecule, higher serum adiponectin has been reported to be associated with a decreased risk for cardiovascular events in diverse populations including people with nondiabetic chronic kidney disease (7), diabetic subjects (15), and end-stage renal disease patients (16), and therefore, its potential role in vascular disease is intriguing. In particular, diminishing the inflammatory state of diabetes may be central in modifying risk for progressive endothelial dysfunction and renal disease. For example, adiponectin levels are increased by thiazolidinediones (38,39), medications that improve insulin sensitivity at the cellular level (22), so raising low adiponectin states is achievable if proven to be beneficial. Rosiglitazone therapy has been confirmed to increase adiponectin levels in African Americans with impaired glucose tolerance or diabetes (40), although direct evidence for the benefit of thiazolidinediones in human diabetic nephropathy is currently lacking.

Adiponectin levels can be increased through nonmedication factors as well. Low dietary glycemic load and high-fiber diets may contribute to higher adiponectin levels in diabetic subjects (15). Moderate alcohol intake also appears to increase adiponectin levels by 0.8 mg/ml (P = 0.01) for each additional drink per day in this cohort (41). Cigarette smoking is another modifiable risk factor reported to be associated with lower adiponectin (42). The effects of these lifestyle modifications and higher adiponectin levels on kidney function decline over time remain to be determined.

Limitations to our study include its cross-sectional design, which limits conclusions about mechanism or temporal relation. No information is available on albuminuria because urine has not been collected and stored in this cohort. Lastly, renal function was estimated from creatinine-based prediction equations. The relatively high coefficient of variation of plasma creatinine would presumably result in random misclassification and bias the results toward the null.

Our observation of the inverse association of adiponectin with moderate renal dysfunction in men with type 2 diabetes needs to be replicated in non-Caucasians and in women. How serum adiponectin is related to eGFR decline over time also warrants further study. Based on these initial data, we hypothesize that sustained higher serum adiponectin levels, independent of glucose and lipid control, would be associated with slower rates of eGFR loss in type 2 diabetes.

This work was supported by National Institutes of Health Grants K08 DK066246 (to J.L.), R01HL065582 (to F.B.H.), and P01CA055075 and a Norman S. Coplon Extramural Research Grant through Satellite Healthcare (to J.L.).

A portion of this work was included in a poster presentation at Renal Week 2005, 8–13 November 2005, Philadelphia, Pennsylvania.