Measurement and Correlation of Indices of Insulin Resistance in Patients on Peritoneal Dialysis

Abstract

♦ Background:

Insulin resistance (IR) is common in maintenance dialysis patients and is associated with excess mortality. Hyperinsulinemic euglycemic glucose clamp (HEGC) is the gold standard for measuring IR. There are limited studies using HEGC for comparison to other indirect indices of IR in peritoneal dialysis (PD) patients, nor have there been direct comparisons between patients receiving PD and those on maintenance hemodialysis (MHD) with regard to severity of IR, methods of measurement, or factors associated with the development of IR.

♦ Methods:

This is a cross-sectional, single-center study performed in 10 prevalent PD patients of median age 48 years (range 41 – 54); 50% were female and 60% were African American. Insulin resistance was assessed by HEGC (glucose disposal rate [GDR]), homeostatic model assessment of IR (HOMA-IR), HOMA-IR corrected by adiponectin (HOMA-AD), leptin adiponectin ratio (LAR), quantitative insulin sensitivity check index (QUICKI), McAuley's index, and oral glucose tolerance test (OGTT) at each time point for a total of 18 studies. Retrospective analysis compared this cohort to 12 hemodialysis patients who had previously undergone similar testing.

♦ Results:

The median GDR was 6.4 mg/kg/min (interquartile range [IQR] 6.0, 7.8) in the PD cohort compared with the MHD group, which was 5.7 mg/kg/min (IQR 4.3, 6.6). For both the PD and MHD cohorts, the best predictors of GDR by HEGC after adjusting for age, gender, and body mass index (BMI), were HOMA-AD (PD: r = −0.69, p = 0.01; MHD: r = −0.78, p = 0.03) and LAR (PD: r = −0.68, p < 0.001; MHD: r = −0.65, p = 0.04). In both groups, HOMA-IR and QUICKI failed to have strong predictive value. Eight of 10 PD patients had at least 1 abnormal OGTT, demonstrating impaired glucose tolerance.

♦ Conclusions:

Insulin resistance is highly prevalent in PD patients. The adipokine based formulas, HOMA-AD and LAR, correlated well in both the PD and MHD populations in predicting GDR by HEGC, outperforming HOMA-IR. The use of these novel markers could be considered for large-scale, epidemiological outcome studies.

Cardiovascular disease is the leading cause of death in patients with end-stage renal disease (ESRD) and those undergoing chronic dialysis treatments have a life expectancy of 25 – 30 years less than age-, race-, and sex-matched controls (1). Insulin resistance (IR) is a significant risk factor for the development of cardiovascular events and occurs at an increased frequency in peritoneal dialysis (PD) patients (2,3). This cluster of metabolic derangements includes a disturbance of glucose and insulin metabolism (4). Peritoneal dialysis patients are unique in that they experience absorption of high concentrations of glucose through the peritoneum, which is thought to lead to an insulin resistant state. Insulin resistance, the reciprocal of insulin sensitivity, is an independent predictor for the development of cardiovascular disease and has been previously shown to be highly prevalent at 47% and 21%, in maintenance PD and hemodialysis (MHD) patients, respectively (2). Measuring the presence of IR would identify ESRD patients who have a high risk of cardiovascular disease, and this could provide a target for intervention through diet, exercise, and pharmacotherapy (5).

The gold standard for quantifying IR is the hyperinsulinemic euglycemic clamp (HEGC), which is considered cumbersome and not readily usable in the clinical setting or in large-scale studies. In the existing literature, IR is more commonly estimated by homeostatic model assessment (HOMA-IR), which has been shown to be an independent predictor of cardiovascular mortality in non-diabetic, maintenance dialysis patients. However, the use of HOMA-IR in PD patients is flawed, as most, if not all, studies report this value after a dialysate dwell, or in a state of continuous glucose absorption, which violates the basal condition. Other more readily available indices of IR include the quantitative insulin sensitivity check index (QUICKI), McAuley's index (a triglyceride based method) (6), the oral glucose tolerance test (OGTT), and adipokine-based indices: HOMA adjusted by adiponectin (HOMA-AD) and leptin-adiponectin ratio (LAR), which have not been studied in PD in depth.

There are very limited studies using the gold standard HEGC to quantify the presence and extent of IR in PD patients and its comparison to readily available, practical methods. In this pilot study, we aimed to examine the extent of peripheral IR in PD patients using HEGC, to compare novel and commonly used indices to estimate IR in a true fasting state, and to evaluate the differences that may exist between PD and MHD patients. This is relevant because metabolic syndrome definitions may have limited applicability in PD due to the high prevalence of its different components and a large waist circumference (7). We hypothesized that IR will be both common and more prevalent in the PD population than in the MHD population primarily due to the continuous glucose exposure and absorption during PD. We also hypothesized that the adipokine-based indices will have a higher correlation with the glucose disposal rate (GDR) as determined by HEGC than other practical methods such as HOMA-IR.

Materials and Methods

This study was part of a randomized trial involving patients undergoing PD randomized to continue their typical prescription or receive a single long dwell daily of icodextrin over a period of 3 months (NCT01119196). The measurement of insulin sensitivity by all methods described below occurred at baseline (week 0), week 6, and week 12. These data were compared to data collected by Hung et al. (8) in MHD patients.

Subjects

Ten PD patients were recruited from the Vanderbilt University Medical Center (VUMC) Outpatient Dialysis unit between September 2010 and October 2014. Inclusion criteria included age older than 18, on PD for at least 3 months, weekly Kt/V > 1.7, and on a stable PD regimen containing glucose lactate-buffered solution. Patients were excluded if they had an infection within the previous 30 days, had chronic inflammatory conditions, used anti-inflammatory medication other than a low dose (< 10 mg daily) of prednisone, or had severe hypokalemia (level < 3.0 mEq/L), hypercalcemia (> 11.0 mEq/L), or insulin-dependent diabetes mellitus. Type 2 diabetics who were noninsulin dependent were allowed in the study unless they were treated by an insulin-sensitizing agent. Following the baseline assessment, 5 patients were randomized to a nightly dwell of icodextrin treatment. The remaining patients continued their routine prescription. A total of 18 studies were completed. The measurements were performed 3 times on each patient, except for 5 patients who only completed 1 study—1 patient due to intolerance to icodextrin and withdrawal from the parent study, another due to an inability to establish IV access, and 3 patients who provided informed consent for the baseline metabolic study. Two patients completed only 2 studies; the second study was aborted due to hypokalemia for 1 patient and the other had inability to establish IV access. This study was approved by Vanderbilt University Medical Center Institutional Review Board, and informed consent was obtained from all patients.

Hyperinsulinemic-Euglycemic Glucose Clamp Study (HEGC)

All studies were performed at the General Clinical Research Center at VUMC. On day 1, the PD subjects were admitted the night prior to the insulin clamp study. They received a standardized meal (35 kcal/kg body weight containing 50% carbohydrates, 32% fat, and 18% protein), then performed 3 manual exchanges, 2 hours each, of PD. Serum potassium levels were measured on arrival to the research center and potassium supplementation was provided within the dialysate for levels less than 5 mEq/L. All dialysate was drained at midnight and subjects remained dry and fasting until the completion of the insulin clamp study the following day (9). On the morning of the clamp study, a fasting blood glucose measurement was obtained as well as baseline values of inflammatory and hormonal markers. Two peripheral IVs were established, 1 on each upper extremity. One was used for infusion of insulin, dextrose, normal saline, and, if required, potassium chloride. The other was used for blood sampling. An infusion of insulin at a concentration of 2 mU/kg/min was started to suppress endogenous insulin production. Serum potassium levels were monitored every 15 minutes and supplementation provided if less than 4.5 mEq/L. Blood glucose was checked every 5 minutes to achieve normoglycemia (95 – 105 mg/dL) using variable infusion of 20% dextrose. Once steady state was reached and confirmed at 90 minutes, the average value of the glucose infusion rate was calculated over the last 30 minutes (M value). This was normalized to total body weight to estimate the glucose-disposal rate (GDR, mg/kg per minute). Glucose-disposal rates higher than 7.5 mg/kg per minute were considered to be an insulin-sensitive state, values between 4.0 and 7.5 mg/kg per minute were consistent with impaired glucose tolerance, and values lower than 4.0 mg/kg per minute were considered insulin-resistant (10).

After completion of the insulin clamp study, patients were provided with a meal of predetermined content. Again, they performed 4 rapid PD exchanges every 2 hours with the dialysate completely drained at midnight to begin another 8-hour fasting period. The rationale for the rapid exchanges was to achieve appropriate solute clearance during the 2-day dialysis-free state for estimation of true basal state IR.

Oral Glucose Tolerance Test (OGTT)

On day 2, fasting glucose and insulin levels were obtained per routine OGTT protocol. Each subject then received a 75-g glucose load followed by a blood glucose level obtained at 2 hours. An OGTT showing a 2-hour glucose of 120 – 199 mg/dL is consistent with impaired glucose tolerance; a level > 200 mg/dL is considered diagnostic of diabetes mellitus according to the American Diabetes Association (11) .

Derived Insulin Resistance Indices

In addition to HEGC and OGTT, homeostasis model assessment (HOMA-IR; insulin (uU/mL) × glucose (mg/dL)/405) (12), quantitative insulin sensitivity check index (QUICKI; 1/log glucose [mg/dL] + log insulin [uU/mL]) (13); McAuleys index exp (2.63 – 0.28 ln insulin [uU/mL] – 0.31 ln tri-glycerides [mM/mL]) (6), homeostasis model assessment corrected by adiponectin (HOMA-AD; insulin [uU/mL] × glucose [mg/dL]/405 × adiponectin [mg/mL]) (14), and leptin adiponectin ratio (LAR; leptin [ng/mL]/adiponectin [mg/mL]) were calculated (15). In the general population, previously determined reference values for IR are HOMA-IR ≥ 2.6, QUICKI ≤ 0.33, McAuley's index ≤ 5.8 (16,17).

Blood Samples

All blood samples were collected in similar fashion and technique to that described in Hung et al. 2010 (8). Glucose concentrations were measured by using the glucose oxidase method and insulin by double-antibody RIA, (Glucose Analyzer 2; Beckman Coulter, Brea, CA, and DA RIA; Millipore, St. Charles, MO, USA, respectively). C-reactive protein levels were measured using the UniCel Dxi Immunoassay System (Beckman Coulter, Brea, CA, USA). Adiponectin and leptin were measured using the MILLIPEX MAP Human Serum Adiponectin Panel A kit (Millipore, Billerica, MA, USA). All of the other measurements were performed using standard, certified, routine laboratory methods.

Body Composition by Dual-Energy X-Ray Absorptiometry (DEXA)

The assessment of body composition was performed using a Lunar iDEXA machine, encore 2007, v.11.40.004 and GE Lunar Body Composition Software (General Electric, Madison, WI, USA). Dual-energy X-ray absorptiometry was performed while the patient's abdomen remained empty of dialysate. The procedure was performed in similar fashion to that described previously by Hung et al. 2010 (8). Measurements obtained and reported here include percent body fat (%), truncal fat mass (%), trunk total mass, lean tissue, and fat (kg).

Power/Sample Size Calculation

Many of the primary outcomes of this study are descriptive and formal power calculations not applicable. To determine the degree of IR in patients receiving PD compared with those on MHD, our preliminary data showed that mean (± standard deviation [SD]) HOMA-IR was 2.66 ± 2.61 in hemodialysis patients. In PD patients, published data indicate a range of 4.6 – 8.2. A total of 18 measurements would have 90% power to detect a difference of 3.0 in HOMA-IR in MHD vs PD patients with 2-sided significance level of 5%. To compare the insulin sensitivity measured by the gold-standard HEGC with the more readily available methods including HOMA-IR, based on our preliminary data in MHD patients showing a correlation coefficient of 0.53, 18 measurements would have provided a minimum detectable Pearson correlation coefficient of at least 0.50 between glucose disposal rate and HOMA-IR, achieving a minimum of 80% power with 2-sided significance level 5%.

Statistical Analysis

Insulin resistance was quantified by GDR as determined by the HEGC. Data are presented as mean ± SD or as median with interquartile range (IQR) depending on their distribution. Differences in median values between PD and MHD groups were analyzed with the Mann-Whitney U test. Chi-square test was used for comparison of categorical data. Analysis of covariance (ANCOVA) was used to compare percent change of GDR from baseline to 12 weeks in the PD group. Linear mixed model was used to account for repeated measures per subject, and all models were adjusted for age, gender, and body mass index (BMI). Statistical tests include correlations between GDR and secondary indices separately in the PD and MHD groups, associations between selected covariates and IR in the PD and MHD groups, and the variation in IR correlations between the PD and MHD groups. Each of the proposed determinants of IR was selected a priori on the basis of previous literature. All insulin-resistance measures were natural log transformed for homoscedasticity. We used bootstrap validation to assess model overfitting, which was very minimal. Factors associated with IR were analyzed using 2 methods, proportional odds and linear mixed model, to assess for potential differences due to normality. In many cases, p values were significant in both models, and among models where both p values are statistically significant, the direction of the effect was the same. Factors associated with insulin resistance measured by HEGC were reported from the linear mixed model for ease of readership. The sensitivity analysis was performed for the Spearman Correlation between GDR and other IR indices and mixed model analysis after excluding the patients who met the criteria for diabetes mellitus. Statistical analysis was performed using SPSS, version 22 for Mac (IBM Corp, Armonk, NY, USA).

Results

Baseline characteristics of the study subjects are shown in Table 1. In the PD cohort, the mean age was 48 years (range 41 – 54), 50% were female, and 60% were African American. The median BMI was 30.1 kg/m² (IQR 26.4, 34.2), which is less than that previously reported in the MHD cohort (median 34.6 kg/m², IQR 26.9, 37.9). The median amount of time on dialysis was 17 months (IQR 9, 50), compared with 46 months (IQR 33, 94) in the MHD group. All patients were using continuous ambulatory PD. Peritoneal dialysis solutions that the patients were using were as follows: 3 patients were performing 3 exchanges for 4-hour dwell with D2.5 and 1 exchange with D1.5; 5 patients were performing 2 exchanges with D1.5 and 2 exchanges with D2.5; 1 patient was performing 3 exchanges with D2.27 and 1 exchange with D1.36, and 1 patient was performing 3 exchanges with D2.5 and 2 exchanges with D1.5. Of those recruited, 5 patients switched to icodextrin and 5 continued to use original glucose-based dialysis solution.

TABLE 1.

Baseline Characteristics of the Study Subjects

Insulin Resistance

A total of 18 studies were completed in 10 PD patients. Table 1 depicts the baseline values for the insulin-resistance indices in PD patients in comparison with previously obtained data in MHD patients. The median GDR was 6.4 mg/kg/min (IQR 6.0, 7.8) in the PD cohort compared with the MHD group, which was slightly more resistant at 5.7 mg/kg/min (IQR 4.3, 6.6). The median values of GDR for each time point for each subject are summarized in Figure 1. HOMA-IR in the PD group demonstrated insulin sensitivity with median 2.5 (IQR 2.0, 4.5) vs the MHD group, which was resistant with a median value 3.3 (IQR 2.4, 3.8). Likewise, the PD group was insulin-sensitive relative to the MHD group for QUICKI, median 0.33 (IQR 0.30, 0.34) vs 0.32 (IQR 0.31, 0.33), McAuley's Index, median 7.2 (IQR 6.7, 8.2) vs 5.9 (IQR 5.3, 7.4), and lastly, HOMA-AD 58 (IQR 48, 122) vs 67 (IQR 43, 119), respectively. Conversely, LAR in the PD cohort reflected a more insulin-resistant state, median 2.32 (IQR 1.04, 4.39) vs the MHD group, median 1.36 (IQR 0.67, 4.52).

Figure 1 —

Glucose disposal rate by HEGC at each study time point. The measurements were performed once in 5 patients, twice in 2 patients, and 3 times in 3 patients. HEGC = hyperinsulinemic euglycemic clamp.

Prior to study start, 1 patient (10%) of the PD cohort carried a prior diagnosis of diabetes mellitus and required an oral antihyperglycemic agent. According to the American Diabetes Association guidelines, 1 patient met the criteria for impaired fasting glucose with at least 2 values of 100 – 125 mg/dL, 3 other participants also had at least 1 value within this range. On the 2-hour OGTT, all 9 patients not previously known to have diabetes met criteria for impaired glucose tolerance with levels between 140 and 199 mg/dL. The patient with known diabetes mellitus type 2 had a 2-hour glucose level greater than 200 mg/dL (Table 2).

TABLE 2.

Oral Glucose Tolerance Test of PD Patients

Correlations Among Different Indices of Insulin Resistance

Correlation coefficients of the different indices of IR and HEGC are shown in Table 3 and Figure 2 for both groups. For the PD group, the best predictors of GDR by HEGC after adjusting for age, gender, and BMI were HOMA-AD (r = −0.69) and LAR (r = −0.68) with p values of 0.01 and < 0.01, respectively. Also, there were weak but significant correlations between GDR derived by HEGC with HOMA-IR (r = −0.35) and QUICKI (r = 0.35) (p values of 0.01 and 0.02, respectively). McAuley's was insignificant as a predictor of GDR by HEGC (r = 0.25, p = 0.22).

TABLE 3.

Correlations of the Different IR Indices with Glucose Disposal Rate by HEGC in Chronic Dialysis Patients

Figure 2 —

The data represent the best-fit regression line. The p values for the association between the different indices and GDR as determined by HEGC are derived from a linear mixed model. The correlation coefficient is derived by Spearman correlation. GDR = glucose disposal rate; HEGC = hyperinsulinemic euglycemic clamp; PD = peritoneal dialysis; HD = hemodialysis; LAR = leptin to adiponectin ratio; HOMA-AD = homeostatic model assessment-adiponectin; HOMA-IR = homeostatic model assessment insulin resistance; rs = spearman correlation coefficient.

Determinants of Insulin Resistance by HEGC

In the PD group, in the adjusted analysis for age, gender, and BMI, factors that were associated with IR included adiponectin (p < 0.001) and triglycerides (p < 0.001). In the MHD group, there were significant associations between GDR and adiponectin (p = 0.006) and leptin (p = 0.003). In the MHD cohort, there was also a significant association between GDR and LDL cholesterol levels (p = 0.008) and dialysis vintage (p = 0.03) and near significant association with total fat mass percentage (0.09) that were not present for patients undergoing PD (Table 4).

TABLE 4.

Factors Associated with IR Measured by HEGC

Sensitivity Analysis

Spearman correlation analysis between GDR and other IR indices and the mixed model analysis were repeated after excluding 1 patient from PD and 5 patients from MHD groups who met criteria for diabetes mellitus (Table 5, Figure 3). For correlations among different indices of IR, results were similar to those from the main analysis for the PD group. The best predictors of GDR by HEGC continued to be HOMA-AD (r = −0.58) and LAR (r = −0.42) after adjusting for age, gender, and BMI, with p values of < 0.001 and < 0.001, respectively. Also, there were weak but significant correlations between GDR derived by HEGC with HOMA-IR (r = −0.38) and QUICKI (r = 0.38) (p values of < 0.001 and < 0.001, respectively). McAuley's index was a weak but significant predictor of GDR by HEGC (r = 0.22; p = 0.01). In the MHD group, there were negative correlations between GDR derived by HEGC and LAR (r = −0.64) and HOMA-AD (r = −0.86).

TABLE 5.

Spearman Coefficients of the Different IR Indices with Glucose Disposal Rate by HEGC after Excluding Diabetic Patients

Figure 3 —

The association between GDR derived by HEGC and other insulin-resistance indices after excluding diabetic patients. GDR = glucose disposal rate; HEGC = hyperinsulinemic euglycemic clamp; PD = peritoneal dialysis; HD = hemodialysis; LAR = leptin to adiponectin ratio; HOMA-AD = homeostatic model assessment-adiponectin; HOMA-IR = homeostatic model assessment insulin resistance; rs = spearman correlation coefficient.

Serum adiponectin and triglyceride levels were statistically significant determinants of IR by HEGC in PD patients, similar to the primary analysis (p < 0.001 for both). In addition, the association between leptin and GDR by HEGC became significant with the exclusion of the patient with DM (p = 0.02). In the HD group, serum triglyceride (p = 0.04), serum LDL cholesterol (p < 0.001), and dialysis vintage (p = 0.004) were significantly associated with GDR by HEGC for the MHD group (Table 6).

TABLE 6.

Factors Associated with IR Measured by HEGC after Excluding Diabetic Patients

Discussion

In this pilot study, we found that IR is highly prevalent in PD patients as determined by HEGC. The adipokine-based formulas, HOMA-AD and LAR, performed well in PD patients in predicting GDR by HEGC, outperforming HOMA-IR, which is used most abundantly in large-scale studies. Our data suggest that routine use of HOMA-IR in epidemiological and interventional studies may be misleading in PD patients and may need to be replaced by more metabolically relevant measures such as ones adjusted by adipokines.

Insulin resistance or glucose intolerance was present in 82% of our total population (80% of PD, 83% of MHD patients) confirming what has been published previously (4,18). We also a priori hypothesized that the PD patients would show a greater degree of IR due to constant glucose exposure and absorption compared with MHD patients. However, our results were not concordant with this hypothesis. While this finding could be due to the differences in body habitus with the PD patients having slightly lower median BMI 30.1 (IQR 26.4, 34.2) vs the median BMI in the MHD group 34.6 (IQR 26.9, 37.9), we observed that fat mass did not predict IR in PD patients and was not a confounder in our multivariate analysis. Nevertheless, these observations should be interpreted with caution. Peritoneal dialysis patients had less trunkal obesity than MHD patients in our study, and the lack of a wide range of BMI or waist circumference in PD patients might have confounded the results. Overall, these data indicate that abnormalities in carbohydrate metabolism and insulin secretion are not only a consequence of loss of kidney function or body habitus and would require a multi-faceted approach for prevention or treatment.

A major aim of our study was to define to most suitable practical marker for assessing the extent of IR in PD patients. Homeostatic model assessment of IR is designed to measure fasting IR, the basal state of insulin sensitivity (12). In this study, we ensured a fasting state by draining the abdomen 8 hours prior to testing, unlike other published data. The PD and MHD groups appeared to have the same signal but different strength with respect to weak correlation between GDR and HOMA-IR. Perhaps differences in renal clearance of insulin may account for the difference in relationship; however, follow-up analysis did not show a significant correlation between residual renal function and presence of IR as determined by GDR. Even so, the adipokine-based formulas had a better correlation with GDR than did HOMA-IR. This is consistent with a previous report in the MHD population (8).

Adipokine-based formulas may have increased validity in the ESRD patient population because of reduced renal excretion, both in the PD and MHD populations (8,19). The presence of obesity in the PD population has also been theorized to lead to the development of IR, which is proposed to increase the cardiovascular risk profile. Truncal fat is the source of adipokines, adiponectin, and leptin. Here we show agreement between truncal fat mass and adipokine-based indices of IR and GDR. Perhaps it is not only the continuous absorption of glucose that creates a resistant state, but the development of obesity, specifically truncal obesity, that contributes to the development of IR in PD patients. Accordingly, adipokine-based measures of IR might be the most suitable markers to examine when evaluating the adverse clinical effects of IR in PD patients, as suggested in MHD patients (8). In accordance with these findings, a recent study shows that adipokine-based indices predict mortality in the PD population (20).

There are several strengths to this study. We used the gold standard method to detect and measure IR in both MHD and PD populations, the latter of which has limited published data available. This method has been shown to be highly reproducible and allowed for comparisons with novel markers and indices. While the use of HOMA-IR in PD patients is highly prevalent in the literature, after ensuring a fasting state, our results indicated a relatively poor correlation with GDR by HEGC. This is most likely related to the fact that HOMA-IR is developed based on a basal fasting state, which is commonly violated in PD patients due to continuous absorption of glucose in the peritoneum.

There are several limitations to this study, the small sample size being the most significant. The study was intensive, required inpatient stay, and, as a part of the prospective study, consent to change dialysis prescription based on randomization. While the total number of patients was small, we were able to test our hypotheses by using multiple studies within a patient, using appropriate statistical adjustments. As reported in the article by Hung et al. (8), the BMI for the MHD group was substantially higher than the reported average, 58% vs 35% with a BMI > 30. For the PD population, the BMI of our patients was also higher than what has previously been reported. While this may limit the generalizability in terms of prevalence and severity of IR, as BMI has changed over time in to the general PD and MHD populations, the trend toward a higher BMI in incident maintenance dialysis patients worldwide makes our results applicable.

In conclusion, IR is common in PD patients. The adipokine-based indices, HOMA-AD and LAR, were highly correlated with GDR by HEGC whereas the commonly used HOMA-IR did not perform as well when compared with the GDR in this study. The use of HOMA-AD or LAR in large epidemiologic studies involving maintenance dialysis patients may better predict higher cardiovascular risk or document risk reduction after intervention to reduce IR.

Disclosures

This study was supported in part by grants from Baxter Pharmaceutical, Clinical Translational Science Award UL1 TR000445 from the National Center for Research Resources, K24 DK062849 from the National Institute of Diabetes and Digestive and Kidney Diseases, Vanderbilt Diabetes Research and Training Center Grant P30 DK020593, Vanderbilt Center in Molecular Toxicology Grant P30 ES000267, and Vanderbilt O'Brien Mouse Kidney Center Grant P30 DK079341. AH's work was supported by a Career Development Award (2-031-09S) from the Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development Clinical Sciences Research. SMD was partly sponsored by an International Society of Nephrology/Turkish Society of Nephrology fellowship grant.