Relationship between serum total bilirubin levels and mortality in uremia patients undergoing long-term hemodialysis: A nationwide cohort study

Atherosclerosis 265 (2017) 155e161

Contents lists available at ScienceDirect

Atherosclerosis journal homepage: www.elsevier.com/locate/atherosclerosis

Relationship between serum total bilirubin levels and mortality in uremia patients undergoing long-term hemodialysis: A nationwide cohort study Hui-Hsien Su a, 1, Chia-Man Kao b, 1, Yi-Chun Lin e, f, Yen-Chung Lin b, c, d, **, 2, Chih-Chin Kao c, Hsi-Hsien Chen c, Chih-Cheng Hsu g, Kuan-Chou Chen b, Chiung-Chi Peng b, Mai-Szu Wu b, c, d, *, 2 a

Division of Gastroenterology, Department of Internal Medicine, Taipei Medical University Hospital, 252, Wu-xing St., Taipei, Taiwan, ROC Graduate Institute of Clinical Medicine, College of Medicine, Taipei Medical University, 250, Wu-xing St., Taipei, Taiwan, ROC c Division of Nephrology, Department of Internal Medicine, Taipei Medical University Hospital, 252, Wu-xing St., Taipei, Taiwan, ROC d Department of Internal Medicine, School of Medicine, College of Medicine, Taipei Medical University, 252, Wu-xing St., Taipei, Taiwan, ROC e Division of Endocrinology & Metabolism, Department of Medicine, Taipei Veterans General Hospital, 201, Sec.2, Shi-Pai Rd., Taipei, Taiwan, ROC f Faculty of Medicine, National Yang-Ming University, 155, Sec.2, Linong St., Taipei, Taiwan, ROC g Institute of Population Health Sciences, National Health Research Institutes, 35 Keyan Rd, Zhunan, Miaoli county, Taiwan, ROC b

a r t i c l e i n f o

a b s t r a c t

Article history: Received 24 April 2017 Received in revised form 11 June 2017 Accepted 1 September 2017 Available online 4 September 2017

Background and aims: Previous studies show that serum bilirubin has potent antioxidant effect and is associated with protection from kidney damage and reduce cardiovascular events. The aim of this study was to examine the association of serum total bilirubin level and mortality in uremia patients who underwent hemodialysis. Methods: This is a nationwide retrospective long-term cohort study. Patients were registered in the Taiwan Renal Registry Data System (TWRDS) from 2005 to 2012. A total of 115,535 hemodialysis patients were surveyed and those with valid baseline total bilirubin (TB) data were enrolled. All-cause mortality was the primary outcome. Results: A total of 47,650 hemodialysis patients followed for 27.6 ± 12 months, were divided into 3 groups according to different baseline serum total bilirubin levels (0.1e0.3, 0.3e0.7, 0.7e1.2 mg/dL). Mean age was 61.4 ± 13.6 years, 50% were male, 13% were hepatitis B carriers, and 20% were hepatitis C carriers. Primary outcome was the 3-year mortality. The TB level 0.7e1.2 mg/dL group had high mortality, statistically significant hazard ratio of mortality was 1.14 (crude HR, 95% 1.07e1.20, p < 0.01), and adjusted HR was 1.18 (model 1, 95% CI 1.11e1.25), 1.21 (model 2, 95% CI 1.14e1.29, p < 0.01), 1.44 (model 3, 95% CI 1.06e1.96, p < 0.01), respectively. Sensitivity test showed that after excluding 14,899 patients with hepatitis B or C, or abnormal liver function, the highest level of TB associated with higher significant mortality was still robust. Conclusions: In our study, high TB level is associated with mortality in uremia patients undergoing longterm hemodialysis, but further studies of the different effects of unconjugated or conjugated bilirubin on hemodialysis patients are needed. © 2017 Elsevier B.V. All rights reserved.

Keywords: Atherosclerosis Bilirubin End stage renal disease Hemodialysis Unconjugated

1. Introduction * Corresponding author. Present address: 110 No. 252, Wu-xing St., Taipei Medical University Hospital, Taiwan, ROC. ** Corresponding author. Present address: 110 No. 252, Wu-xing St., Taipei Medical University Hospital, Taiwan, ROC. E-mail addresses: [email protected] (Y.-C. Lin), [email protected] (M.-S. Wu). 1 These authors contributed equally to this work as first authors. 2 These authors contributed equally to this work as corresponding authors. http://dx.doi.org/10.1016/j.atherosclerosis.2017.09.001 0021-9150/© 2017 Elsevier B.V. All rights reserved.

Cardiovascular disease-related mortality risk increases about 30-fold in end-stage renal disease patients receiving hemodialysis. Lipid oxidation or oxidative stress is important in the pathogenesis of atherosclerosis [1], which may be attributed to bilirubin's antioxidant properties [2]. Thus, a low serum total bilirubin (TB) concentration is associated with accelerated progression of chronic

156

H.-H. Su et al. / Atherosclerosis 265 (2017) 155e161

kidney disease (CKD), is an independent predictor of CKD progression [3], and is associated with a high incidence of cardiovascular events [4]. In patients with uremia, unlike other anti-oxidative agent-uric acids [5], serum TB level is stable and not affected by hemodialysis therapy itself. There is only one study in which a graded, reverse association was observed between serum TB and adverse outcomes among chronic hemodialysis patients [6]. However, it was a small, single-center, observational study. A larger cohort study with 5900 Korean men showed that higher serum direct bilirubin was associated with a lower risk of new-onset non-alcoholic fatty liver disease. However, body mass index also showed an inverse trend with TB. There is a counterintuitive phenomenon present in hemodialysis patients: obese patients may have a better prognosis than non-obese patients [7]. Therefore, how TB affects clinical outcomes in dialysis patients remains controversial due to a paucity of data. The aim of this study was to investigate the relationship between TB and mortality in patients undergoing long-term HD using the nationwide TWRDS database. TWRDS is a registered online system for the dialysis center to upload their annual biochemistries for reimbursement from the National Health Insurance bureau in Taiwan. The system was operated by specialized nurses, and repeated records of bilirubin, taken every 3 months in each hemodialysis (HD) patient, were used to predict its association with mortality. 2. Materials and methods This study was approved by the ethics committee of the Taipei Medical University Institutional Review Board (No. N201610006) and was carried out in accordance with the Declaration of Helsinki of 1975, as revised in 2013. The Waiver of the Requirements for Obtaining Informed Consent was approved by the Taipei Medical University Institutional Review Board since the study met all applicable regulations. 2.1. Taiwan Renal Registry Data System The TWRDS was initially established in 1987 for the accreditation of dialysis therapy. All dialysis units in Taiwan are obligated to upload relevant data to the website constructed by TWRDS in 2005. This provided only a very small portion of incentives (5%) for dialysis centers and did not correlate with the NHI's medical reimbursement for overall expenditures from the organization. A registered nurse from each dialysis unit submits a quarterly report. The data in the TWRDS form a solid foundation for continual dialysis quality control at the national level. 2.2. Patient enrollment We included in the analysis patients registered with the TWRDS from 2005 to 2012 (n ¼ 115,565). Patients who had received HD or peritoneal dialysis (PD) for >3 months were distributed to either the HD group or the PD group, respectively. There were 4661 patients who changed their dialysis modality, 9232 who used PD (categorical or continuous), 998 patients with extremely old or young age (>90 years or <20 years), 42,887 patients without a TB record (with the same sex distribution and a similar average age as the study population), and 2978 patients with higher (>1.2 mg/dL) or lower (<0.1 mg/dL) TB. These patients were excluded from the analysis and, finally, 47,650 HD patients with valid baseline TB data (average TB level in the first year) were followed for 3 years (Fig. 1).

2.3. Statistical analysis All variables were analyzed with the Chi-squared test (categorical variables), analysis of variance (continuous variables normally distributed), or the ManneWhitney U test (continuous variables with non-parametric distributions). After excluding 2978 patients with a TB value beyond the reference level (0.1e1.2 mg/dL), patients were divided into groups based on a clinically acceptable level of 0.3 and 0.7 mg/dL, and quartiles of TB levels were also used for grouping. Kaplan-Meier curves were used to determine the unadjusted survival curves, and a Cox regression analysis was used to test the association between predictors and primary outcomes, with adjustment for multiple confounders. A sensitivity test was performed by excluding from the population hepatitis B virus (HBV) and hepatitis C virus (HCV) carriers, and patients with liver function abnormalities. A two-tailed p-value <0.05 was used to indicate statistical significance. 3. Results 3.1. Demographic characteristics of the overall 47,650 uremia patients who underwent hemodialysis Table 1 shows the baseline characteristics of the population, stratified by 3 different TB levels (0.1e0.3, 0.3e0.7, and 0.7e1.2 mg/ dL). The average TB levels in these groups were 0.2 ± 0.1, 0.5 ± 0.1, and 0.8 ± 0.1 mg/dL, respectively; the mean age was 61.4 ± 13.6 years, 50% were men, 13% carried HBV, and 20% carried HCV. The groups had similar biochemistries, including with respect to uric acid, ferritin, cholesterol, triglycerides, albumin, hematocrit, and calcium or phosphate levels, except for total bilirubin and liver function (ALT). During the follow-up period, only 908 (1.9%) patients received kidney transplantation, the comparisons of basic characteristics between those who had kidney transplantation or not are listed in Supplementary Data 1, there were no significant differences in TB levels. The issue of selection bias could be avoided. 3.2. Unadjusted and adjusted survival analysis for total bilirubin as predictors (categorical or continuous) Fig. 2 shows a Kaplan-Meier curve and reveals that those with TB levels <0.3 mg/dL had the best unadjusted survival rate, followed by those with TB levels between 0.3 and 0.7 and > 0.7 mg/dL. The average follow-up duration was 27.6 ± 12 months, similar across the different TB groups. Table 2 shows more detailed information for the groups with TB levels ranging from 0.3 to 1.2 mg/dL with an interval of 0.1 mg/dL. The group with the highest level of TB (between 0.7 and 1.2 mg/dL) had a statistically significant hazard ratio (HR) for mortality of 1.14 (crude HR; 95% confidence interval (CI), 1.07e1.20, p < 0.01), and adjusted HRs of 1.18 (model 1; 95% CI, 1.11e1.25), 1.21 (model 2; 95% CI, 1.14e1.29), and 1.44 (model 3; 95% CI, 1.06e1.96). For continuous predictors, TB showed similar trends: crude HR 1.03 (95% CI 1.03e1.04) and adjusted HRs 1.03 (95% CI, 1.03e1.04) in model 1, 1.04 (95% CI, 1.03e1.05) in model 2, and 1.05 (95% CI, 1.01e1.09) in model 3, for each 0.1-mg/dL increase in TB. The sensitivity test was also performed by including those with TB > 1.2 mg/dL, and <0.1 mg/dL (total N ¼ 50,601), the results that higher TB was associated with high morality remained unaffected (Supplemental Data 2). 3.3. Survival analysis stratified by quartile of total bilirubin levels Table 3 shows the crude and adjusted HRs for mortality for HD patients according to quartile (Q1 to Q4). It reveals that the Q4 group with the highest average TB (mean TB, 0.93 ± 0.78 mg/dL)

H.-H. Su et al. / Atherosclerosis 265 (2017) 155e161

157

Fig. 1. Flow chart of the TWRDS population.

Table 1 Baseline characteristics of HD patients followed up for 3 years, by groups according to total bilirubin levels.

Number Age (years) Male (%) HTN (%) CHF (%) LVH (%) DM (%) CVA (%) CAD (%) HBV (%) HCV (%) HD duration (years) [25th-75th](median) Laboratory data Total bilirubin (mg/dL) WBC (x1000/mL) BUN (mg/dL) Creatinine (mg/dL) Residual GFR (ml/min) ALT(GPT) (IU/L) Uric acid (mg/dL) Ferritin (ng/mL) Cholesterol (mg/dL) Triglyceride (mg/dL) Albumin (g/dL) Glucose (mg/dL) Hct (%) Ca (mg/dL) P (mg/dL) Ca*P product

Total

Total bilirubin (mg/dL) 0.1e0.3

0.3e0.7

0.7e1.2

p

47,650 61.4 ± 13.6 23,964(50%) 21,766(46%) 6648(14%) 6772(14%) 24,325 (51%) 3749(8%) 6347(13%) 6068(13%) 9701(20%) 0.89e3.45 (1.11)

10,301 61.0 ± 13.8 5034(49%) 4548(44%) 1194(12%) 1092(11%) 5621(55%) 734(7%) 1272(12%) 1158(11%) 1704(17%) 0.86e1.93 (1.02)

31,035 61.8 ± 13.5 15,359(49%) 14,443(47%) 4495(14%) 4520(15%) 5991(53%) 2509(8%) 4157(13%) 3965(13%) 6301(20%) 0.90e3.53 (1.11)

6314 60.4 ± 13.7 3571(57%) 2775(44%) 959(15%) 1160(18%) 2792 (44%) 506(8%) 918(15%) 945(15%) 1695(27%) 0.96e5.37 (1.72)

e <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001 0.0067 0.0002 <0,0001 <0.0001 <0.0001

0.5 ± 0.2 7.0 ± 3.2 71.1 ± 16.3 9.6 ± 2.5 6.2 ± 2.2 22.6 ± 19.9 7.2 ± 1.3 412.2 ± 220.6 172.2 ± 36.9 164.3 ± 107.8 3.8 ± 0.4 148.4 ± 67.6 30.7 ± 3.4 9.2 ± 0.8 4.9 ± 1.2 44.9 ± 12.4

0.2 ± 0.1 7.3 ± 5.5 72.5 ± 16.4 9.5 ± 2.6 6.3 ± 2.3 20.7 ± 16.2 7.4 ± 1.3 405.1 ± 213.7 173.2 ± 36.8 183.1 ± 120.7 3.8 ± 0.4 157.7 ± 73.0 30.4 ± 3.2 9.1 ± 0.8 4.9 ± 1.3 44.4 ± 12.5

0.5 ± 0.1 7.0 ± 2.2 70.8 ± 16.2 9.5 ± 2.5 6.2 ± 2.2 22.4 ± 20.6 7.2 ± 1.3 413.2 ± 220.1 172.4 ± 36.6 160.8 ± 103.0 3.8 ± 0.4 147.4 ± 66.6 30.7 ± 3.3 9.2 ± 0.8 4.8 ± 1.2 44.7 ± 12.3

0.8 ± 0.1 6.7 ± 2.2 70.2 ± 16.1 9.9 ± 2.6 6.3 ± 2.3 26.1 ± 21.2 7.1 ± 1.3 419.3 ± 234.0 169.7 ± 38.4 150.8 ± 105.1 3.8 ± 0.4 138.3 ± 61.6 31.2 ± 3.8 9.3 ± 0.8 5.0 ± 1.3 46.4 ± 12.7

<0.0001 <0.0001 <0.0001 <0.0001 0.2244 <0.0001 <0.0001 0.0003 <0.0001 <0.0001 0.0563 <0.0001 <0.0001 <0.0001 <0.0001 <0.0001

DM, diabetes; HTN, hypertension; CAD, coronary artery disease; CHF, congestive heart failure; LVH, left ventricular hypertrophy; CVA, cerebral vascular accident; HBV, hepatitis B; HCV, hepatitis C, ALT, alanine aminotransferase; Ca, calcium; P, phosphate; WBC, white blood cell, BUN, blood urea nitrogen, Cr, creatinine, GFR, glomerular filtration rate.

158

H.-H. Su et al. / Atherosclerosis 265 (2017) 155e161

Fig. 2. Kaplan-Meier curve of unadjusted survival in the whole population, categorized by different total bilirubin groups (<0.3 mg/dL, dotted line; 0.3e0.7 mg/dL, dashed line; >0.7 mg/dL, solid line).

Table 2 Models of multivariate Cox regression of mortality on hemodialysis patients by categorical total bilirubin levels (TB) (N ¼ 47,650). TB (mg/dL)

N

Crude hazard ratio (HR)

HR in model 1

HR in model 2

HR in model 3

0.1e0.3 0.3e0.4 0.4e0.5 0.5e0.6 0.6e0.7 0.7e1.2 Every 0.1 increase

10,301 10,598 7666 7046 5725 6314 47,650

0.95 0.97 1 1.04 0.97 1.14 1.03

0.98 0.96 1 1.04 0.99 1.18 1.03

0.97 0.99 1 1.05 1.01 1.21 1.04

1.04 0.98 1 1.23 1.22 1.44 1.05

(0.89e1.00) (0.92e1.02) (0.99e1.11) (0.92e1.04) (1.07e1.20)** (1.03e1.04)**

(0.93e1.04) (0.91e1.02) (0.98e1.10) (0.93e1.05) (1.11e1.25)** (1.03e1.04)**

(0.92e1.03) (0.94e1.05) (0.99e1.10) (0.95e1.08) (1.14e1.29)** (1.03e1.05)**

(0.82e1.31) (0.77e1.25) (0.93e1.63) (0.89e1.66) (1.06e1.96)** (1.01e1.09)**

**p < 0.01. Model 1: adjust for age, sex. Mode1 2: adjust for model 1 plus DM, HTN, CAD, CHF, LVH, CVA, HBV, HCV, HD duration. Model 3: adjust for mode 2 plus ALT, ferritin, uric acid, albumin, Hb, Ca, P, Kt/V, WBC, BUN, Cr, residual GFR. DM, diabetes; HTN, hypertension; CAD, coronary artery disease; CHF, congestive heart failure; LVH, left ventricular hypertrophy; CVA, cerebral vascular accident; HBV, hepatitis B; HCV, hepatitis C, ALT, alanine aminotransferase; Ca, calcium; P, phosphate; Kt/V dialytic dose; WBC, white blood cell, BUN, blood urea nitrogen, Cr, creatinine, GFR, glomerular filtration rate.

had the highest crude HR (1.12; 95% CI, 1.06e1.17, p < 0.01), adjusted HR in model 1 (1.11; 95% CI, 1.06e1.17), adjusted HR in model 2 (1.14; 95% CI, 1.09e1.20, p < 0.01), and adjusted HR in model 3 (1.24; 95% CI, 1.00e1.55. Furthermore, Fig. 3 shows the HR of each TB category and reveals that the highest TB category (between 0.7 and 1.2 mg/dL) had the highest mortality.

with abnormal liver function, defined by an alanine aminotransferase level >40 mg/dL. The finding that the highest level of TB was significantly associated with higher mortality remained evident. The crude HR in those with a TB level between 0.7 and 1.2 mg/dL was 1.16 (95% CI, 1.08e1.25), their adjusted HR in model 1 was 1.19 (95% CI, 1.10e1.28), in model 2 1.22 (95% CI, 1.13e1.31), and in model 3 1.17 (95% CI, 0.79e1.74). The group with the highest TB had statistically significantly higher mortality.

3.4. Sensitivity test excluding those with hepatitis B, hepatitis C, or abnormal liver function 4. Discussion Table 4 shows results of the sensitivity study on the 32,751 patients after excluding 14,899 patients with hepatitis B or C or

As a catabolic end-product of heme, bilirubin has potent

H.-H. Su et al. / Atherosclerosis 265 (2017) 155e161

159

Table 3 Models of multivariate Cox regression of mortality on hemodialysis patients, stratified by quarters of total bilirubin (TB) levels (N ¼ 47,650). Quarters of TB (mg/dL) Q1 Q2 Q3 Q4

(0.17 (0.35 (0.52 (0.93

± ± ± ±

0.09) 0.04) 0.05) 0.78)

Number

Crude hazard ratio (HR)

HR in model 1

HR in model 2

HR in model 3

12,769 11,403 11,602 11,876

1 1.01 (0.96e1.06) 1.10 (1.05e1.16)** 1.12 (1.06e1.17)**

1 0.98 (0.94e1.03) 1.07 (1.02e1.12)** 1.11 (1.06e1.17)**

1 1.00 (0.95e1.05) 1.07 (1.02e1.13)** 1.14 (1.03e1.20)**

1 0.87 (0.72e1.05) 1.07 (0.87e1.32) 1.24 (1.00e1.55)**

**p < 0.01. Adjust for model 1: age, sex. Adjust for model 2: age, sex, DM, HTN, CAD, CHF, LVH, CVA, HBV, HCV infection, HD duration. Adjust for model 3: model 1 plus ALT, ferritin, uric acid, albumin, Hb, Ca, P, Kt/V, WBC, BUN, Cr, residual GFR. DM, diabetes; HTN, hypertension; CAD, coronary artery disease; CHF, congestive heart failure; LVH, left ventricular hypertrophy; CVA, cerebral vascular accident; HBV, hepatitis B; HCV, hepatitis C, ALT, alanine aminotransferase; Ca, calcium; P, phosphate; Kt/V dialytic dose, WBC, white blood cell, BUN, blood urea nitrogen, Cr, creatinine, GFR, glomerular filtration rate.

Fig. 3. Crude and adjusted models in evaluating hazard ratio of mortality in different groups of total bilirubin levels.

Table 4 Models of multivariate Cox regression of mortality on hemodialysis patients by categorical total bilirubin (TB) levels after excluding chronic hepatitis carriers or abnormal liver function (N ¼ 32,751). TB (mg/dL)

Number

Crude HR

Adjusted HR (Model 1)

Adjusted HR (Model 2)

Adjusted HR (Model 3)

0.1e0.3

7574

0.3e0.4

7505

0.4e0.5 0.5e0.6

5313 4822

0.6e0.7

3727

0.7e1.2

3810

Every 0.1 mg/dL increase of TB

32,751

0.96 (0.90e1.03) 1.00 (0.93e1.07) 1 1.08 (1.01e1.16)* 1.00 (0.93e1.08) 1.16 (1.08e1.25)** 1.03 (1.02e1.04)**

1.00 (0.94e1.07) 0.99 (0.93e1.06) 1 1.07 (1.00e1.15) 1.01 (0.93e1.09) 1.19 (1.10e1.28)** 1.03 (1.02e1.04)**

0.99 (0.92e1.06) 1.02 (0.95e1.09) 1 1.07 (0.99e1.14) 1.02 (0.94e1.10) 1.22 (1.13e1.31)** 1.03 (1.02e1.04)**

1.03 (0.79e1.36) 1.08 (0.82e1.43) 1 1.15 (0.82e1.61) 1.27 (0.88e1.84) 1.17 (0.79e1.74) 1.02 (0.97e1.07)

*p < 0.05; **p < 0.01. Model 1: adjust for age, sex. Mode1 2: adjust for model 1 plus DM, HTN, CAD, CHF, LVH, CVA, HD duration. Model 3: adjust for mode 2 plus ALT, ferritin, uric acid, albumin, Hb, Ca, P, Kt/V, WBC, BUN, Cr, residual GFR. DM, diabetes; HTN, hypertension; CAD, coronary artery disease; CHF, congestive heart failure; LVH, left ventricular hypertrophy; CVA, cerebral vascular accident; HBV, hepatitis B; HCV, hepatitis C, ALT, alanine aminotransferase; Ca, calcium; P, phosphate; Kt/V dialytic dose, WBC, white blood cell, BUN, blood urea nitrogen, Cr, creatinine, GFR, glomerular filtration rate.

160

H.-H. Su et al. / Atherosclerosis 265 (2017) 155e161

antioxidant and cytoprotective properties [8,9]. With respect to vascular damage, TB is a promising biomarker associated with an increasing risk of amputation in patients with type 2 diabetes mellitus [10]. and increased atherosclerosis [11]. However, with respect to the metabolic syndrome, there was a negative association between TB level and metabolic syndrome in a meta-analysis study of type 2 diabetes patients [12], but a positive association in a large population cohort study [13]. In HD patients, low body mass index is associated with bad outcomes, a phenomenon called “reverse epidemiology”. In our nationwide study of HD patients, excluding patients with liver disease or extreme bilirubin levels, 47,650 HD patients were followed for 8 years. We found that a higher TB level was associated with high mortality among longterm HD patients, in contrast to the results of a small population study on HD patients by Chen et al. [6], which was a single-center small cohort without detailed information about serum bilirubin, such as whether it was conjugated bilirubin or unconjugated bilirubin. A reason for this discrepancy may be that higher bilirubin is correlated with a low body mass index, which is strongly correlated with uremia patients' mortality. Many cross-sectional studies showed positive associations between serum TB concentration and the estimated glomerular filtration rate (eGFR) [14e16], with the exception of two studies, in that high bilirubin concentrations were significantly associated with lower eGFR [17,18]. HD patients had common presentations of reverse epidemiology, including hypertension and obesity. As shown in the basic characteristics in Table 1, there was a trend for higher TB being associated with lower triglycerides and cholesterol levels. From the famous 4D study evaluating statin and cardiovascular outcome, decreasing the low-density lipoprotein level in HD patients did not result in better cardiovascular outcomes, probably because of malnutrition [19]. Additionally, the SHARP study also confirmed that uremia patients undergoing HD did not receive benefits from lipid-lowering agents [20]. In another study from our team, high cholesterol or triglyceride levels were not associated with mortality in HD patients who had already experienced myocardial infarction or cardiovascular disease [21]. HD itself is associated with increased oxidative stress, which can be relieved by vitamin E-coated hollow fiber membranes during 4h dialysis sessions [22]. HD removes water-soluble circulating antioxidants, including uric acid and ascorbate, but not hydrophobic substances, including unconjugated bilirubin, which is plasma albumin-bound [23]. In addition, uric acid, but not bilirubin, was significantly associated with new-onset coronary artery disease [24]. In our 3-year observational study, we tried to analyze the relationship of TB with mortality on dialysis patients. Unlike uric acid, which may change after each HD session [23], unconjugated bilirubin remains the same before and after HD, so it cannot indicate oxidative stress. Thus, in a special group undergoing HD, we found that bilirubin showed no protective role. The strength of our study is that it is a nationwide population study in a long-term cohort study focusing on TB levels and mortality in uremia patients undergoing HD. All TB levels were measured in a fasted state. With respect to study limitations, first, we did not have unconjugated bilirubin data, which is important because a recent review article described that only unconjugated bilirubin, which is lipid-soluble, had strong anti-oxidant effects [25]. Second, we only had all-cause mortality but we did not know the exact cause of death of the uremia patients; however, in clinical experience, most dialysis patients die from cardiovascular events. Third, we did not have records of C-reactive protein in our population to evaluate patients' inflammation status which might be associated with TB. In conclusion, we found that a high TB level was correlated with mortality in HD patients without liver disease or abnormal liver

function. A mildly elevated TB level was not associated with a protective role in uremia patients undergoing HD. Further studies are warranted to investigate the different effects of unconjugated or conjugated bilirubin on HD patients. Conflict of interest The authors declared they do not have anything to disclose regarding conflict of interest with respect to this manuscript. Financial support This research received grant from Taipei Medical University 106TMU-TMUH-23. Author contributions Y-CL (Yen-Chung Lin) and M-SW conceived and designed the experiments. C-CK and Y-CL (Yi-Chun Lin) performed the experiments. C-CH and H-HC analysed the data. K-CC and C-CP interpreted the results. H-CS and C-MK wrote the paper. Acknowledgements We would like to thank Mr. Kuang-wen Wang for analytical assistance. Appendix A. Supplementary data Supplementary data related to this article can be found at http:// dx.doi.org/10.1016/j.atherosclerosis.2017.09.001. References [1] M.F. Linton, P.G. Yancey, S.S. Davies, W.G.J. Jerome, E.F. Linton, et al., The role of lipids and lipoproteins in atherosclerosis, in: L.J. De Groot, G. Chrousos, K. Dungan, et al. (Eds.), Endotext, 2000. South Dartmouth (MA). [2] M.F. McCarty, Serum bilirubin may serve as a marker for increased heme oxygenase activity and inducibility in tissuesea rationale for the versatile health protection associated with elevated plasma bilirubin, Med. Hypotheses 81 (2013) 607e610. [3] M. Tanaka, M. Fukui, H. Okada, T. Senmaru, M. Asano, et al., Low serum bilirubin concentration is a predictor of chronic kidney disease, Atherosclerosis 234 (2014) 421e425. [4] M. Fukui, M. Tanaka, M. Yamazaki, G. Hasegawa, M. Nishimura, et al., Low serum bilirubin concentration in haemodialysis patients with Type 2 diabetes, Diabet. Med. 28 (2011) 96e99. [5] A. Beciragic, H. Resic, N. Prohic, J. Karamehic, A. Smajlovic, et al., Correlation between C-reactive protein and non-enzymatic antioxidants (albumin, ferritin, uric acid and bilirubin) in hemodialysis patients, Mater Sociomed. 27 (2015) 87e90. [6] Y.H. Chen, S.C. Hung, D.C. Tarng, Serum bilirubin links UGT1A1*28 polymorphism and predicts long-term cardiovascular events and mortality in chronic hemodialysis patients, Clin. J. Am. Soc. Nephrol. 6 (2011) 567e574. [7] J. Park, S.F. Ahmadi, E. Streja, M.Z. Molnar, K.M. Flegal, et al., Obesity paradox in end-stage kidney disease patients, Prog. Cardiovasc Dis. 56 (2014) 415e425. [8] R. Stocker, Y. Yamamoto, A.F. McDonagh, A.N. Glazer, B.N. Ames, Bilirubin is an antioxidant of possible physiological importance, Science 235 (1987) 1043e1046. [9] D.E. Baranano, M. Rao, C.D. Ferris, S.H. Snyder, Biliverdin reductase: a major physiologic cytoprotectant, Proc. Natl. Acad. Sci. U. S. A. 99 (2002) 16093e16098. [10] K.H. Chan, R.L. O'Connell, D.R. Sullivan, L.S. Hoffmann, K. Rajamani, et al., Plasma total bilirubin levels predict amputation events in type 2 diabetes mellitus: the Fenofibrate Intervention and Event Lowering in Diabetes (FIELD) study, Diabetologia 56 (2013) 724e736. [11] M.E. Vogel, G. Idelman, E.S. Konaniah, S.D. Zucker, Bilirubin prevents atherosclerotic lesion formation in low-density lipoprotein receptor-deficient mice by inhibiting endothelial VCAM-1 and ICAM-1 signaling, J. Am. Heart Assoc. (2017) 6. [12] M.J. Lee, C.H. Jung, Y.M. Kang, J.Y. Hwang, J.E. Jang, et al., Serum bilirubin as a predictor of incident metabolic syndrome: a 4-year retrospective longitudinal study of 6205 initially healthy Korean men, Diabetes Metab. 40 (2014)

H.-H. Su et al. / Atherosclerosis 265 (2017) 155e161 305e309. [13] Y.B. Lee, S.E. Lee, J.E. Jun, J.H. Jee, J.C. Bae, et al., Change in serum bilirubin level as a predictor of incident metabolic syndrome, PLoS One 11 (2016) e0168253. [14] S.S. Han, K.Y. Na, D.W. Chae, Y.S. Kim, S. Kim, et al., High serum bilirubin is associated with the reduced risk of diabetes mellitus and diabetic nephropathy, Tohoku J. Exp. Med. 221 (2010) 133e140. [15] M. Fukui, M. Tanaka, E. Shiraishi, I. Harusato, H. Hosoda, et al., Relationship between serum bilirubin and albuminuria in patients with type 2 diabetes, Kidney Int. 74 (2008) 1197e1201. [16] H.S. Shin, Y.S. Jung, H. Rim, Relationship of serum bilirubin concentration to kidney function and 24-hour urine protein in Korean adults, BMC Nephrol. 12 (2011) 29. [17] G. Targher, G. Zoppini, G. Cesare Guidi, G. Lippi, Relationship between serum bilirubin and kidney function in non-diabetic and diabetic individuals, Kidney Int. 75 (2009) 863. [18] G. Targher, C. Bosworth, J. Kendrick, G. Smits, G. Lippi, et al., Relationship of serum bilirubin concentrations to kidney function and albuminuria in the United States adult population. Findings from the National Health and Nutrition Examination Survey 2001-2006, Clin. Chem. Lab. Med. 47 (2009) 1055e1062. [19] C. Wanner, V. Krane, W. Marz, M. Olschewski, J.F. Mann, et al., Atorvastatin in

[20]

[21]

[22]

[23] [24]

[25]

161

patients with type 2 diabetes mellitus undergoing hemodialysis, N. Engl. J. Med. 353 (2005) 238e248. Sharp Collaborative, G, Study of Heart and Renal Protection (SHARP): randomized trial to assess the effects of lowering low-density lipoprotein cholesterol among 9,438 patients with chronic kidney disease, Am. Heart J. 160 (2010) 785e794 e710. Y.C. Lin, Y.C. Lin, H.H. Chen, T.W. Chen, C.C. Hsu, et al., Different effect of hypercholesterolemia on mortality in hemodialysis patients based on coronary artery disease or myocardial infarction, Lipids Health Dis. 15 (2016) 211. G. D'Arrigo, R. Baggetta, G. Tripepi, F. Galli, D. Bolignano, Effects of vitamin ecoated versus conventional membranes in chronic hemodialysis patients: a systematic review and meta-analysis, Blood Purif. 43 (2017) 101e122. J.S. Coombes, R.G. Fassett, Antioxidant therapy in hemodialysis patients: a systematic review, Kidney Int. 81 (2012) 233e246. B. Bagheri, M. Zargari, F. Meshkini, K. Dinarvand, V. Mokhberi, et al., Uric acid and coronary artery disease, two sides of a single coin: a determinant of antioxidant system or a factor in metabolic syndrome, J. Clin. Diagn Res. 10 (2016) OC27e31. S. Gazzin, L. Vitek, J. Watchko, S.M. Shapiro, C. Tiribelli, A novel perspective on the biology of bilirubin in health and disease, Trends Mol. Med. 22 (2016) 758e768.