Urinary Adiponectin Is an Independent Predictor of Progression to End-Stage Renal Disease in Patients With Type 1 Diabetes and Diabetic Nephropathy

Diabetic nephropathy (DN) in type 1 diabetes is associated, already at the microalbuminuric stage, with a twofold higher mortality rate, which increases substantially toward end-stage renal disease (ESRD) (1). Prevention of DN is recommended and should be based on screening using biomarkers of progression, such as the albumin excretion rate (AER) or estimated glomerular filtration rate (eGFR). Although the molecular pathogenic mechanisms of DN are not completely understood, adiponectin (ADP) may also play a role in DN pathogenesis.

ADP is a small protein encoded by the adiponectin gene (ADIPOQ), which is primarily expressed in adipocytes with no apparent expression in the kidney (2). ADP is present in serum in three molecular isoforms: low molecular weight (LMW), medium molecular weight, and high molecular weight (HMW). ADP has a wide range of well-known protective effects against insulin resistance, vascular dysfunction, atherosclerosis, and inflammation (3). Animal studies suggested that ADP regulates albuminuria and podocyte function (4). In humans, serum ADP usually is inversely correlated with eGFR, whereas ADP is abundantly present in biopsy specimens of nondiabetic human kidneys (5,6). In addition, serum ADP is increased in patients with type 1 diabetes and may predict progression to ESRD (7,8). Being a protein molecule, ADP possibly passes the glomerular basement membrane and is excreted in the urine, thereby reflecting glomerular damage. Indeed, various ADP isoforms can be measured in the urine, where ADP is considered to be a marker of vascular damage (4,6). However, urinary ADP (uADP) has also been linked to renal tubular injury (9). Thus, uADP may reflect both glomerular and tubular damage in DN. In this context, studies regarding uADP level as a possible more comprehensive predictor of DN progression are warranted. Therefore, the aims of the present study were to 1) evaluate the predictive role of uADP for progression of DN in patients with type 1 diabetes, 2) investigate the added predictive benefit of uADP on top of AER or eGFR, and 3) examine the principal determinants of uADP.

Patients with type 1 diabetes included in this study were part of the Finnish Diabetic Nephropathy Study (FinnDiane) and were enrolled between January 1998 and December 2002. For this study, we also used a group of subjects without diabetes and without a family history of kidney disease or diabetes. Blood and urine samples were collected at study baseline and stored at −20°C until measured in 2008. Patients were followed for a median of 5.8 (4.4–6.9) years, and then clinical outcomes were evaluated. This study was performed with the approval of local ethics committees in accordance with the revised Declaration of Helsinki.

At baseline, we used a standardized questionnaire to assess patient clinical characteristics. The questionnaires were completed by the attending physician based on patient medical records. Blood pressure and anthropometric parameters were measured. Venous blood was collected for the assessment of common biochemical variables, which were measured using standard methods as previously described (10). During the follow-up period, all patients with type 1 diabetes were managed by their own practitioner together with his or her diabetes team. The detailed FinnDiane Study protocol has been described in detail elsewhere (11,12).

Renal status was defined based on the AER in at least two of three consecutive timed urine collections. On the basis of the AER, patients were divided into three categories: normal AER (<20 μg/min or <30 mg/24 h), microalbuminuria (≥20 but <200 μg/min or ≥30 but <300 mg/24 h), and macroalbuminuria (≥200 μg/min or ≥300 mg/24 h). ESRD was considered present if patients were undergoing dialysis or had received a kidney transplant. Patients with ESRD at baseline were excluded. Glomerular filtration rate was estimated using the CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) creatinine equation (13). Progression of DN was defined as the passage from one AER stage to the next or progression to ESRD for patients with initial macroalbuminuria.

We measured uADP in a single 24-h urine collection with an ALPCO Diagnostics kit (Salem, NH) for quantitative measurement of multimeric ADP, using a modified protocol for urine samples without protease pretreatment. The uADP levels were then normalized for urinary creatinine. Urinary kidney injury molecule 1 (KIM-1), serum ADP, and urinary liver-type fatty acid–binding protein (L-FABP) values were also available. Urinary KIM-1 was measured using a cobas Elecsys 411 immunoanalyzer with a DuoSet ELISA Development Kit from R&D Systems (Abingdon, Oxon, U.K.); serum ADP and urinary L-FABP measurement methods are described elsewhere (14,15).

Normally distributed variables are presented as mean ± SD. Nonnormally distributed variables are presented as median and interquartile range. Frequencies are given as percentages. Comparisons between groups were performed by independent samples t test for normally distributed variables and Mann-Whitney U test for nonparametric distributions. Categorical variables were compared between groups using χ² test.

We used Cox proportional hazards models to assess the ability of uADP to predict DN progression. The algorithm by which we constructed the basic models of DN progression in the cohort was previously published (15). First, we used simple Cox proportional hazards models to investigate uADP as a predictor for progression of nephropathy. After that, we adjusted the result with the basic models of progression for each stage to test the independence of the biomarker. Next, we included AER, KIM-1, L-FABP, or serum ADP in the models to see whether uADP predicts progression independent of these other markers. The Cox model fit was assessed by cumulative Cox-Snell residuals to (−log) Kaplan-Meier estimates. Fine and Gray regression analysis, which extends the Cox proportional hazards model to competing risk data by consideration of the subdistribution hazard, was also performed to take into account the competing event of death instead of progression to a higher stage (16,17). In the competing risk analysis, we included the same covariates as in the previous Cox regression analysis. The colinearity of the models was estimated by the variance inflation factor, tolerance, and R values. The cutoff values considered acceptable were <10 for variance inflation factor, >0.5 for tolerance, and <0.7 for R (18). The validity of the assumption for the prediction models was tested by checking the normal distribution of the residuals using the D’Agostino-Pearson test (19).

Receiver operating characteristic (ROC) curve analysis of the Cox models was performed and the areas under the curve (AUCs) compared using the method described by DeLong et al. (20) to assess the predictive abilities of uADP for progression of DN. First, we compared the ROC curves of uADP and AER alone. Next, we compared the ROC curve of the model formed by AER and uADP together with either AER or ADP alone. The improvement of prediction given by the addition of uADP to either AER alone or basic progression models plus AER was assessed by calculating the continuous net reclassification improvement (NRI) and the integrated discrimination improvement (IDI) obtained by 10-fold cross-validation using 1,000 bootstrap repetitions of the whole data set through the Stata incrisk module (21,22). The predictive performance of the Cox models was evaluated by the explained variation (R²) using 1,000 bootstrap repetitions of the whole data set through the Stata str2ph module (23). In addition, the mean risk increment and delta standardized total gain were calculated using the same Stata module.

For all tests, P < 0.05 was considered statistically significant. The data analysis was performed using MedCalc 12.1.3.0 (MedCalc Software BVBA, Mariakerke, Belgium) and Stata/MP2 version 13 (StataCorp LP, College Station, TX) software.

At baseline, 1,451 patients had normal AER, 319 had microalbuminuria, and 320 had macroalbuminuria. In addition, 111 subjects without diabetes were enrolled. The baseline characteristics of the cohort are shown in Table 1. Patients were followed for a median of 5.8 (4.4–6.9) years. During the follow-up, 101 patients progressed from normal AER to microalbuminuria, 42 progressed from micro- to macroalbuminuria, and 71 progressed from macroalbuminuria to ESRD. The differences in baseline clinical characteristics between the patients who progressed to a higher stage of DN and nonprogressors are detailed in Supplementary Table 1.

The uADP concentrations were higher in patients with diabetes and normal AER than in subjects without diabetes (0.56 vs. 0.34 μg/g, P < 0.0001). Furthermore, uADP increased with worsening DN (P < 0.0001) (Table 1 and Supplementary Fig. 1), and uADP was significantly higher in patients who progressed to a higher stage of DN compared with nonprogressors (Supplementary Table 1 and Supplementary Fig. 1).

Univariate Cox proportional hazards models showed that uADP predicted progression to a higher stage of DN in patients with normal AER (hazard ratio [HR]_normalAER 1.32 [95% CI 1.13–1.54]), microalbuminuria (HR_micro 1.38 [95% CI 1.09–1.74]), and macroalbuminuria (HR_macro 2.03 [95% CI 1.76–2.34]) at baseline. uADP was still a significant predictor of progression after adjustment for the basic progression models and sex (HR_normalAER 1.25 [95% CI 1.05–1.48], HR_micro 1.35 [95% CI 1.04–1.73], HR_macro 1.52 [95% CI 1.30–1.78]). However, after AER was added to the model, uADP independently predicted progression only to ESRD (HR_normalAER 0.99 [95% CI 0.82–1.20], HR_micro 1.02 [95% CI 0.76–1.37], HR_macro 1.37 [95% CI 1.07–1.75]) (Table 2). The results did not change in the competing risk analysis considering death as a competing event for progression at any stage (Supplementary Table 2).

Comparison of ROC AUCs for uADP and AER alone showed AER as a better predictor of progression to microalbuminuria (ΔAUCs_normalAER 0.180, P < 0.0001) and macroalbuminuria (ΔAUCs_micro 0.218, P < 0.0001) but not to ESRD (ΔAUCs_macro 0.057, P = 0.04), where uADP was superior. Comparison of the ROC curve of the model formed by AER together with uADP with AER alone showed an added predictive benefit regarding progression to ESRD in favor of the model formed by uADP together with AER (ΔAUCs_macro 0.056, P = 0.02). Finally, the ROC curve comparisons between uADP (AUC_ADPmacro 0.842) and eGFR alone (AUC_eGFRmacro 0.853) showed no significant differences (ΔAUCs_macro 0.010, P = 0.79), but when uADP was added to eGFR, there was again an added predictive benefit with respect to progression toward ESRD (ΔAUCs_macro 0.029, P = 0.03) compared with eGFR alone (Table 3 and Supplementary Fig. 2).

There was an improvement in the prediction of progression to ESRD by adding uADP to AER (NRI 0.794, P = 0.03; IDI 0.115, P < 0.0001) but not by adding uADP on top of the basic progression models plus AER (NRI 0.042, P > 0.05; IDI 0.015, P > 0.05). For all the other stages of DN, there was no improvement. Furthermore, in patients with baseline macroalbuminuria, addition of uADP improved the prediction of progression to ESRD compared with eGFR alone (NRI 0.637, P < 0.001; IDI 0.087, P < 0.0001), as did the addition of uADP to the basic progression models (NRI 0.674, P < 0.001; IDI 0.084, P < 0.0001) (Table 4).

Explained variability (R²) of the Cox models for progression to ESRD was better when uADP was added either to AER alone (0.651 vs. 0.472, P < 0.0001) or to the basic progression model plus AER (0.818 vs. 0.772, P = 0.007). In addition, the mean risk increment was 0.088 (95% CI 0.0291–0.172) when uADP was added to eGFR or 0.117 (95% CI 0.056–0.200) when uADP was added to AER (Supplementary Table 3).

The uADP predicted progression independently of serum ADP in patients with normal AER (HR 1.26 [95% CI 1.01–1.56], P = 0.04) but not in patients with microalbuminuria (HR 1.29 [95% CI 0.91–1.83], P = 0.15). However, for patients with baseline macroalbuminuria, both variables were independent predictors of progression to ESRD (HR 1.36 [95% CI 1.13–1.64], P = 0.001). uADP predicted progression to ESRD independently of urinary L-FABP (HR 1.28 [95% CI 1.14–1.70], P = 0.001) or urinary KIM-1 (HR 1.29 [95% CI 1.18–1.80], P = 0.02) (Table 2).

The common determinants of uADP levels at all baseline stages of DN were HbA_1c, AER, urinary L-FABP, and urinary KIM-1, whereas for patients with macroalbuminuria, age (β = −0.019, P = 0.02), BMI (β = −0.048, P = 0.02), serum ADP (β = 0.015, P = 0.03), eGFR (β = −0.018, P < 0.0001), and LDL cholesterol (β = −0.196, P = 0.03) also played a role (Supplementary Table 4).

We show that uADP is a strong and independent predictor of DN progression to ESRD in patients with type 1 diabetes. In addition, when uADP was used together with AER or eGFR, it improves the risk prediction of DN progression to ESRD. Finally, the main determinants of uADP levels were the markers of glomerular damage, tubular dysfunction, glycemic control, and serum ADP levels, but other factors may also be involved.

The main finding of this study was that uADP predicts progression to ESRD, both in Cox regression models and in the Fine and Gray competing risk analysis, independently of the progression model (with eGFR included), AER, and the tubular markers L-FABP and KIM-1, although all these variables are also associated with uADP levels at this stage. An increase in uADP in parallel with AER in patients with type 1 diabetes was previously observed only in a small study that could not explore the relationship further due to insufficient power (24). We show that uADP not only strongly predicted progression to ESRD but also is a better predictor than AER and adds a significant predictive benefit when used together with eGFR or AER. The important added value of uADP on top of either AER or eGFR is demonstrated for the calculated metrics (increment of AUC, continuous NRI, IDI, R² increase, and mean risk difference between models). To our knowledge, this study is the first to evaluate using a wide range of robust statistical methods the added value of uADP as a biomarker for the progression of DN in patients with type 1 diabetes. From a clinical point of view, these results are important because the risk of progression to ESRD in patients with type 1 diabetes is not easy to assess based on either AER or eGFR given each measure’s limitations (25–27). Assessment of uADP on top of AER or eGFR may therefore add a predictive benefit. A comparison with other biomarkers with respect to these new metrics is difficult because of the lack of comparable data. However, the added predictive benefit of uADP was comparable to another promising biomarker, the soluble tumor necrosis factor-α receptor 1 (28).

Another finding of the present study was that in patients with type 1 diabetes and normal AER or microalbuminuria, uADP predicted progression independently of the other clinical risk factors but not independently of AER. One possible interpretation of these results might be that uADP and AER share common pathophysiological mechanisms. Similarly to albumin, uADP is a circulating plasma protein filtered at the glomerular level (2,29). Moreover, uADP has been suggested to be a marker of vascular damage in type 2 diabetes (6,30,31)and has been previously related to the preservation of podocyte permeability and maintenance of glomerular barrier integrity (4). The current findings support a potential relationship between AER and uADP because in the present cohort, AER was a constant determinant of uADP levels across all stages of DN. This connection may be one explanation for the lack of independent predictive ability for uADP regarding AER because albuminuria is also an important predictor of progression at the earlier DN stages (32).

In addition to AER, the common determinants of uADP across all stages were HbA_1c, L-FABP, and KIM-1. In this regard, uADP could capture another important pathophysiological mechanism of DN on top of glomerular damage: tubular injury due to poor glycemic control. One possible explanation is that hyperglycemia may trigger not only glomerular damage but also tubular dysfunction because ADP may be lost first at the glomerular level, then later in intrarenal arteries/arterioles, and finally at all kidney levels (6). In addition, ADP has been linked to tubular damage and suggested to play a protective role in renal fibrosis (9,33,34). From these findings, we can assume that the initial ADP loss in urine possibly due to hyperglycemia is connected to glomerular dysfunction, whereas the later increase in uADP could be related to tubular dysfunction.

Finally, the study shows low BMI as a significant determinant of uADP levels only in patients with macroalbuminuria, which suggests that uADP also captures renal cachexia. This notion is supported by human studies showing that well-nourished patients undergoing hemodialysis had significantly lower serum ADP levels compared with malnourished patients, whereas in animal studies, high serum ADP levels were associated with increased energy expenditure and weight loss (35–37). Furthermore, ADP measured in the urine was mainly the LMW isomer and had an unexpected homology with tumor necrosis factor (or cachectin), whereas in type 1 diabetes, the serum ADP increase was based on HMW ADP—the main active protective molecule (6,38,39). These data suggest that the biggest increase in ADP in serum and urine observed at the macroalbuminuric stage could be a protective mechanism against cachexia (6).

This study has some limitations, which in our view do not critically influence the results or their interpretation. One limitation is that we could not provide an analysis by sex, although there were some differences in the uADP concentrations within each sex. However, to diminish the influence of sex on the Cox regression models, all final analyses were adjusted for sex. Another limitation may be that we had performed only a single measurement of ADP concentration in the serum and urine, but it has been suggested previously that plasma ADP levels show only minimal short- and long-term variability and are not markedly influenced by normal daily activity (27). To our knowledge, any major variability of the uADP has not been shown in previously published studies. Unfortunately, we were not able to measure the ADP isoforms, but previous studies have shown that the HMW ADP is the main uADP isoform, whereas the LMW isoform is also present in the urine of patients with diabetes (6,40).

In patients with type 1 diabetes and macroalbuminuria, uADP not only is a strong independent predictor for DN progression to ESRD but also adds significant predictive benefit when used together with either AER or eGFR. This may be due to uADP capturing recognized risk factors for DN progression such as glomerular damage, tubular dysfunction, and glycemic control as well as other factors important for DN progression like cachexia.

Acknowledgments. The authors thank laboratory technicians Maikki Parkkonen, Anna-Reetta Salonen, Anna Sandelin, Tuula Soppela, Jaana Tuomikangas (all at Folkhälsan Research Center), Renate Sedlmaier-Prasselsperger, and Kerstin Jaensch (Roche Diagnostics), for the excellent organization and measurements of urine samples on the Elecsys system. They also thank the physicians and nurses at each center participating in the recruitment of patients. (The complete list of physicians and nurses is presented in the Supplementary Data.)

Funding. The study was supported by grants from the Folkhälsan Research Foundation, the Wilhelm and Else Stockmann Foundation, the Liv och Hälsa Foundation, Novo Nordisk Foundation, Academy of Finland, and Finnish Medical Society (Finska Läkaresällskapet). N.M.P. was supported by the Nicolae Paulescu Fellowship from the Romanian Society of Diabetes Nutrition and Metabolic Diseases and by the European Association for the Study of Diabetes–Albert Renold Fellowship.

Duality of Interest. Analyses and assays for uADP, L-FABP, and KIM-1 were partly sponsored by Roche Diagnostics. M.S. is an advisory board member for Medtronic in Scandinavia and has received lecture fees from Eli Lilly, Medtronic Finland, Novartis, Novo Nordisk, Roche, Sanofi, and Merck Sharp & Dohme. P.-H.G. has received research grants from Eli Lilly and Roche and is an advisory board member for Boehringer Ingelheim, Novartis, Cebix, and Abbott. He also has received lecture fees from Boehringer Ingelheim, Eli Lilly, Genzyme, Novartis, Novo Nordisk, Sanofi, and Merck Sharp & Dohme. No other potential conflicts of interest relevant to this article were reported.

The sponsors were not involved in the conduct of the study.

Author Contributions. N.M.P. contributed to the study concept and design, data collection and follow-up, data research, statistical analyses, and writing of the manuscript. M.S. and C.F. contributed to the study concept and design; data collection, follow-up, and research; discussion; and review and editing of the manuscript. L.M.T., D.G., J.W., and N.T. contributed to the data collection, follow-up, and research; discussion; and review and editing of the manuscript. A.B. contributed to the study concept and design. P.M.H. contributed to the study concept and design, data research, discussion, and review and editing of the manuscript. P.-H.G. contributed to the study concept and design; data collection, follow-up, and research; statistical analyses; discussion; and writing, review, editing, and final approval of the manuscript. P.-H.G. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented orally at the 48th Annual Meeting of the European Association for the Study of Diabetes, Berlin, Germany, 1–5 October 2012.