Bilirubin and Progression of Nephropathy in Type 2 Diabetes: A Post Hoc Analysis of RENAAL With Independent Replication in IDNT

The incidence of type 2 diabetes and its complications are increasing worldwide. One of the major complications of type 2 diabetes is diabetic nephropathy (DN). Nephropathy develops in ∼20–40% of patients with diabetes and is the single leading cause of end-stage renal disease (ESRD) around the world (1).

Bilirubin is a product of heme catabolism and is known to be a potent endogenous antioxidant (2). As such, bilirubin has consistently been associated with protection against the development of cardiovascular disease (CVD) (3,4). A study in rodents suggested that bilirubin is also protective against progression of DN (5). This notion is supported by several cross-sectional studies in humans demonstrating that low levels of bilirubin are associated with DN (6–8).

To our knowledge, there are no prospective studies to date that investigated whether bilirubin levels are associated with progression of DN toward ESRD. Therefore, our primary objective was to prospectively investigate the association of bilirubin with progression of nephropathy in patients with type 2 diabetes. To this end, we performed a historical prospective study in the Reduction of Endpoints in NIDDM with the Angiotensin II Antagonist Losartan (RENAAL) trial (9,10). Subsequently, independent replication was sought in the Irbesartan Diabetic Nephropathy Trial (IDNT) (11,12).

In the RENAAL and IDNT studies, patients were treated with an angiotensin receptor blocker (ARB) (losartan in RENAAL and irbesartan in IDNT). Several studies have shown that ARBs reduce hemoglobin levels (13–15). Because bilirubin is a product of heme catabolism, the use of ARBs could consequently reduce bilirubin levels. Therefore, our secondary aim was to investigate the effect of ARB treatment on serum concentrations of hemoglobin and bilirubin.

The current study was conducted in patients with type 2 diabetes and nephropathy participating in the RENAAL and IDNT studies. The design, rationale, and study outcomes for these trials have been published elsewhere (9–12).

Both trials investigated the efficacy of an ARB (losartan in RENAAL and irbesartan in IDNT) on renal outcomes in patients with type 2 diabetes, nephropathy, and proteinuria. Inclusion criteria for both trials were similar, with minor differences in details. Patients with type 2 diabetes, hypertension, and nephropathy aged 30–70 years were eligible for inclusion in both trials. Serum creatinine levels ranged from 1.0 to 3.0 mg/dL. All subjects were required to have proteinuria defined as a urinary albumin-to-creatinine ratio (ACR) ≥300 mg/g or a 24-h urinary protein excretion >500 mg/day in RENAAL and ≥900 mg/day in IDNT. Major exclusion criteria for participation in both trials were type 1 diabetes, nondiabetic renal disease, and screening values of liver enzymes (alanine aminotransferase [ALT], aspartate aminotransferase [AST]) or total bilirubin >1.5 times the upper limit of normal (ULN). The inclusion and exclusion criteria of the RENAAL and IDNT studies are summarized in Supplementary Table 1 [adapted from Packham et al. (16)]. Subjects with missing data for baseline measurements of total bilirubin were excluded from the analyses (RENAAL n = 15 [1.0%], IDNT n = 8 [0.5%]).

Laboratory and physical assessment data were collected every 6 months during follow-up for subjects participating in RENAAL and IDNT and included blood pressure measurements, glycated hemoglobin (HbA_1c), lipid profile, hemoglobin, total bilirubin, serum albumin, ALT, AST, and serum creatinine. For both trials, all biochemical measurements were conducted in a central laboratory according to standardized conditions. Estimated glomerular filtration rate (eGFR) was calculated using the Modification of Diet in Renal Disease (MDRD) equation (17). The primary end point for the current study was the composite of a confirmed doubling of serum creatinine (DSCR) level or ESRD (defined as the need for long-term dialysis or renal transplantation). All end points were adjudicated by an independent committee using rigorous guidelines and definitions.

Statistical analyses were performed using SPSS version 18.0 for Windows (IBM Corporation, Chicago, IL) and Stata 11 (StataCorp LP, College Station, TX) software. Results are presented as mean ± SD for variables with a normal distribution and as median (interquartile range) for variables with a nonnormal distribution. Nominal data are presented as the total number of patients with percentages. A two-sided P < 0.05 was considered statistically significant.

To assess which baseline variables were associated with baseline total bilirubin, the study populations were subdivided into tertiles of baseline total bilirubin concentration, and patient characteristics were presented accordingly (Table 1). P values for trend across tertiles of baseline total bilirubin were assessed using linear regression analyses. Variables with a skewed distribution were log-transformed to fulfill criteria for linear regression analyses. Multivariable linear regression analyses were used to investigate which clinical parameters at baseline were independently associated with bilirubin at baseline (Table 2).

The course of clinical parameters over time are presented according to tertiles of baseline bilirubin levels (Supplementary Table 2). We investigated whether the change in total bilirubin concentration over time differed among tertiles in subjects with complete bilirubin measurements at baseline and 12, 24, and 36 months using one-way ANOVA (Supplementary Table 3).

To investigate the association of total bilirubin with progression of nephropathy, we used Cox proportional hazard regression analyses with time-varying covariates, which takes the change of clinical parameters over time into account (Table 3). Logarithmic transformation (base 2) of bilirubin levels was applied to present the hazard ratios (HRs) derived from Cox regression analyses per doubling of bilirubin levels. Multivariable analyses were conducted using a Cox regression model, including the potential confounding factors of age, sex, baseline eGFR, baseline log ACR, race, smoking at baseline, history of CVD at baseline, baseline BMI, total cholesterol, diastolic blood pressure, HbA_1c, treatment assignment, and log AST.

In sensitivity analyses, we repeated the Cox regression analyses in the subgroups receiving an ARB (i.e., losartan in RENAAL, irbesartan in IDNT) or placebo in both trials. In further sensitivity analyses, we investigated whether the change in bilirubin values during the course of the trials was associated with renal disease progression. To investigate the effect of treatment with an ARB on serum concentrations of hemoglobin and total bilirubin, we used the independent sample t test to compare these concentrations between treatment groups in both trials (Fig. 2).

In RENAAL, bilirubin concentrations were available for 1,498 (99.0%) patients. Mean baseline bilirubin level was 0.57 ± 0.19 mg/dL. Baseline patient characteristics according to tertiles of baseline bilirubin levels are presented in Table 1. Prevalence of male sex, age, history of CVD, hemoglobin, serum albumin, liver enzymes, and eGFR increased with increasing bilirubin levels, whereas the prevalence in use of diuretics, use of insulin, BMI, HbA_1c, cholesterol, triglycerides, and urinary ACR decreased with increasing bilirubin levels. Multivariable linear regression analyses showed that baseline bilirubin levels were independently associated with age, smoking, BMI, HbA_1c, hemoglobin, serum albumin, log AST, and total cholesterol (Table 2).

In IDNT, bilirubin concentrations were available for 1,707 (99.5%) patients. Mean baseline bilirubin level in IDNT was similar to that in RENAAL (0.54 ± 0.21 mg/dL). In general, associations and trends of bilirubin with baseline characteristics were similar to those observed in RENAAL. In multivariable linear regression analyses, all variables that were independently associated with bilirubin in RENAAL (except AST) were also independently associated with bilirubin in IDNT. Sex, race, diastolic blood pressure, duration of diabetes (≥5 years), and eGFR were also significantly associated with bilirubin levels in IDNT.

The course of clinical parameters over time is shown in Supplementary Table 2. In RENAAL, the change in total bilirubin concentration only differed among tertiles of baseline bilirubin at 12 months. After 24 and 36 months, there were no significant differences in change in bilirubin concentrations among tertiles of bilirubin (Supplementary Table 3). These results were confirmed in IDNT (Supplementary Table 3).

After a mean follow-up period of 3.4 years, 471 (31%) subjects had reached the renal end point of DSCR or ESRD in RENAAL. Univariable time-varying Cox regression analysis showed that total bilirubin was significantly associated with progression of nephropathy in RENAAL (HR 0.54 [95% CI 0.45–0.65], P < 0.001) (Table 3, model 1). These associations remained significant after adjustment for potential confounding factors, which were age, sex, race, eGFR, log ACR, BMI, smoking status, history of CVD, total cholesterol, diastolic blood pressure, HbA_1c, treatment, and log AST (0.67 [0.55–0.83], P < 0.001). The risk for the renal end point according to total bilirubin concentrations in the RENAAL trial is shown in Fig. 1A.

In IDNT, 381 (22%) patients reached the renal end point after a mean follow-up of 2.6 years. The results of time-varying Cox proportional hazard regression analyses were similar to those of RENAAL (HR 0.64 [95% CI 0.55–0.76], P < 0.001) for the final multivariable model (Table 3). The graph indicating the risk for the renal end point according to total bilirubin concentrations in IDNT is similar to that for RENAAL (Fig. 1B).

In sensitivity analyses, we investigated whether total bilirubin was associated with progression of nephropathy irrespective of ARB or placebo assignment. In RENAAL, total bilirubin was significantly and inversely associated with progression of nephropathy for subjects receiving losartan (HR 0.66 [95% CI 0.48–0.89], P = 0.008) and those receiving placebo (0.70 [0.52–0.94], P = 0.02). These results were confirmed in IDNT for subjects receiving irbesartan (0.59 [0.43–0.81], P = 0.001) and those receiving placebo (0.61 [0.46–0.80], P < 0.001).

In further sensitivity analyses, the change in total bilirubin during the course of the trial was not associated with the renal end point (HR 1.07 [95% CI 0.97–1.19], P = 0.2, per 0.1 mg/dL), whereas total bilirubin remained significantly associated with the renal end point in RENAAL (0.59 [0.44–0.79], P < 0.001). These results were confirmed in IDNT for change in total bilirubin (1.05 [0.97–1.13], P = 0.2) and for total bilirubin (0.62 [0.52–0.74], P < 0.001). The results remained essentially unchanged when stratified for treatment.

Because treatment with ARBs influence serum concentrations of hemoglobin (13) and could consequently affect bilirubin levels, we investigated the effect of treatment with an ARB (losartan in RENAAL and irbesartan in IDNT) on serum concentrations of hemoglobin and bilirubin. Hemoglobin and bilirubin concentrations over time in the RENAAL trial are shown in Fig. 2A and B. Hemoglobin levels slowly decreased over time in the placebo group, whereas an initial decrease followed by a stabilization of hemoglobin levels was seen in losartan-treated patients (Fig. 2A). After 48 months of treatment, hemoglobin levels were no longer significantly different between treatment groups (Fig. 2A).

Bilirubin levels were slightly, but not significantly, lower in the losartan group than in the placebo group (Fig. 2B). Despite the initial fall in hemoglobin levels, there was no initial fall in bilirubin levels in the losartan group. Bilirubin values decreased over time in both treatment groups, and no significant differences in bilirubin concentrations were observed between treatment groups after 12, 36, and 48 months.

Hemoglobin and bilirubin values of subjects using placebo and irbesartan in IDNT are shown in Fig. 2C and D. In general, the pattern of changes in these markers over time in IDNT was similar to the RENAAL trial. Although hemoglobin levels significantly decreased after initiation of treatment with irbesartan (Fig. 2C), as in RENAAL, no significant differences in bilirubin concentrations were observed between treatment groups (Fig. 2D).

In this historical prospective analysis of the RENAAL trial, we found an independent inverse association of bilirubin levels with progression of nephropathy in patients with type 2 diabetes. This finding was independently replicated in IDNT. Furthermore, we showed that treatment with the ARBs losartan or irbesartan did not result in a decrease in bilirubin concentrations, despite an initial decrease in hemoglobin levels.

One of the major pathophysiologic mechanisms that has been identified in the development and progression of DN is oxidative stress, described as increased levels of reactive oxygen species (18–20). Bilirubin is known to be a potent endogenous antioxidant (2), and a recent study in rodents found a protective effect of bilirubin against DN through inhibition of oxidative stress by downregulation of renal NADPH oxidase (5).

In humans, several cross-sectional studies have provided additional evidence for a protective effect of bilirubin on DN. Inoguchi et al. (6) showed a lower prevalence of vascular complications as well as reduced markers of oxidative stress and inflammation in patients with Gilbert syndrome (a congenital hyperbilirubinemia) and diabetes. Fukui et al. (8) reported a negative correlation of bilirubin with log ACR and a positive correlation with eGFR. In addition, it was shown that bilirubin levels were higher in patients without DN than in those with DN (7). However, these studies were cross-sectional in design, precluding investigation of the prospective association of bilirubin with renal impairment. To our knowledge, the current prospective study is the first to indicate an inverse association of bilirubin and progression of nephropathy toward ESRD in type 2 diabetes.

In animal models, antioxidants have been shown to be effective in treating DN (21,22). In combination with the current human data showing an independent association between bilirubin and renal outcome, we speculate that treatments intended to slightly raise bilirubin levels might have a beneficial effect on progression of nephropathy in patients with type 2 diabetes and low bilirubin levels.

A moderate increase in bilirubin levels could be attained through induction of heme oxygenase-1 (HO-1), the enzyme that catalyzes the rate-limiting step in heme degradation. HO-1 splits heme into carbon monoxide (CO) and biliverdin, which is subsequently reduced to bilirubin (2,23). The HO-1 system and heme degradation products CO, biliverdin, and bilirubin have repeatedly been shown to have renoprotective properties (2,23). Therefore, the renoprotective effects of bilirubin in this study are possibly, and at least partly, mediated by induction of HO-1 and by-products of heme degradation (i.e., CO, biliverdin). A study in rats showed that induction of HO-1 reduces renal oxidative stress and protects against diabetes-related renal injury (24). HO-1 is a highly inducible enzyme that can be induced by many drugs routinely used in clinical medicine (i.e., nonsteroidal anti-inflammatory drugs [NSAIDs], peroxisome proliferator–activated receptor α agonists) (4). However, given the adverse effects of NSAIDs, long-term use of NSAIDs is not recommended in patients with advanced renal function impairment. Natural HO-1 inducers include curcuma and polyphenols (i.e., resveratrol) (4,25). Partial inhibition of conjugation of bilirubin by uridine diphosphate-glucuronyltransferase, an enzyme encoded by the UGT1A1 gene, reduces bilirubin excretion and is known to result in mild increases in bilirubin concentrations (4,23,26). Pharmaceuticals capable of a partial inhibition of UGT1A1 are probenecid and atazanavir (4).

A number of studies have reported that the use of ACE inhibitors and ARBs decrease hemoglobin levels (13–15), which can be enhanced by the use of diuretics (14). Because bilirubin is a product of heme catabolism, changes in hemoglobin levels could subsequently influence bilirubin concentrations. Although the use of losartan and irbesartan slightly, but significantly, decreased hemoglobin levels compared with placebo, it did not affect bilirubin levels in either trial. Several studies reported that the use of ACE inhibitors and ARBs reduce erythropoietin and, consequently, hemoglobin levels by blocking the effects of angiotensin II on erythropoiesis (27,28). Because the enzymatic degradation of hemoglobin by HO-1 is known to be the rate-limiting step in the formation of bilirubin (23,29) and not in the synthesis of hemoglobin, it is conceivable that small changes in the synthesis of hemoglobin do not affect the formation and levels of bilirubin.

This study has several limitations. First, patients with liver enzymes (ALT, AST) and bilirubin levels >1.5 times the ULN were excluded from participation in both trials, which resulted in a relatively narrow range of bilirubin concentrations (i.e., 0.1–2.1 mg/dL, with a mean value of 0.57 mg/dL in RENAAL and 0.54 mg/dL in IDNT). In earlier cross-sectional studies on the association of bilirubin with DN, bilirubin levels were higher, with values of 1.4 (1.3–1.6) mg/dL in subjects with Gilbert syndrome in the study by Inoguchi et al. (6), and 0.71 ± 0.21 mg/dL in subjects with type 2 diabetes in the study by Fukui et al. (8). In a prospective study on development and progression of albuminuria in patients with type 2 diabetes by Mashitani et al. (30), mean bilirubin levels were 0.63 ± 0.28 mg/dL. Within the relatively small range of bilirubin levels in the current study, we could not identify a nonlinear component in the association of bilirubin with the renal end point. A larger range of bilirubin concentrations in future studies might allow for identification of a cutoff value of bilirubin above which the association with progression of renal function might flatten, which could help to identify an optimal target concentration for bilirubin in intervention trials. Second, in both RENAAL and IDNT, only total bilirubin was measured. Direct (conjugated) bilirubin was not measured separately because serum bilirubin comprises >95% of indirect (unconjugated) bilirubin (26), and subjects with total bilirubin levels >1.5 times the ULN were excluded from participation in both trials. Therefore, examining differences between unconjugated and conjugated bilirubin levels was not possible. Furthermore, given the observational nature of this study and the inability to focus on the HO-1 system and its by-products in more detail, it is impossible to draw a definitive conclusion about the causality of bilirubin and progression of DN. Mendelian randomization has been proposed as a method that enables estimation of causal relationships in observational studies (31,32). This method uses genotype to estimate causal relationships between a gene product and physiological outcomes (32). Because there is a strong relation between UGT1A1 (genotype) and bilirubin levels (phenotype) (32), Mendelian randomization can be used to establish a possible causal relation between bilirubin and DN. The strengths of this study are the large sample size, the large number of renal events, and the independent replication of the current findings in another large cohort of >1,700 subjects.

In conclusion, the results show an independent inverse association of bilirubin levels with progression of nephropathy in patients with type 2 diabetes, suggesting that measurement of bilirubin levels may identify patients at risk for renal disease progression. In addition, the study suggests a protective effect of bilirubin against progression of DN, thereby potentially implying its role as an antioxidant.

Acknowledgments. The authors thank all the RENAAL and IDNT investigators, support staff, and participating patients.

Funding. This research was performed within the framework of the Center for Translational Molecular Medicine (CTMM) (www.ctmm.nl) project PREDICCt (Prediction and Early Diagnosis of Diabetes and Diabetes-related Cardiovascular Complications) (grant 01C-104) and supported by the Dutch Heart Foundation, Dutch Diabetes Research Foundation, and Dutch Kidney Foundation. The work leading to this article has received funding from the Seventh Framework Programme of the European Community under grant agreement HEALTH-F2-2009-241544 (Syskid). I.J.R. and S.J.L.B. received support from The Netherlands Heart Foundation, Dutch Diabetes Research Foundation, and Dutch Kidney Foundation, together participating in the framework of the CTMM project PREDICCt. H.J.L.H. is supported by a Veni grant from The Netherlands Scientific Organisation.

Duality of Interest. The RENAAL trial was funded by Merck & Co. The IDNT trial was sponsored by the Bristol-Myers Squibb Institute for Medical Research and Sanofi-Synthelabo. M.E.C. and D.d.Z. have received financial support from Merck for their participation on the RENAAL Steering Committee. No other potential conflicts of interest relevant to this article were reported.

Author Contributions. I.J.R., S.J.L.B., and H.J.L.B. contributed to the data analysis and interpretation, writing of the manuscript, and approval of the final manuscript. P.E.D., G.N., M.E.C., J.B.L., and D.d.Z. contributed to the critical revision of the manuscript and approval of the final manuscript. I.J.R. and H.J.L.H. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form at the American Society of Nephrology Kidney Week 2012, San Diego, CA, 30 October–4 November 2012.