Low Serum Bilirubin Level as an Independent Predictor of Stroke Incidence
A Prospective Study in Korean Men and Women
Background and Purpose— Bilirubin is not only a waste end-product but also an antioxidant. Bilirubin is known to be associated with decrease in cardiovascular risk in men, but its relationship to stroke was not clearly understood.
Methods— Serum bilirubin concentrations were measured in 78 724 health examinees (41 054 men, aged 30–89 years) from 1994 to 2001. The subjects with potential hepatobiliary diseases or Gilbert syndrome were excluded from analysis. Stroke incidence outcome was collected from hospital records of admission attributable to stroke from 1994 to 2007.
Results— Serum bilirubin measurements were divided into 4 levels: 0 to 10.2, 10.3 to 15.3, 15.4 to 22.1, and 22.2 to 34.2 μmol/L. The number of stroke cases was 1137 in men and 827 in women. In Cox proportional hazard models, participants with a higher level of bilirubin showed lower hazard ratios in men with ischemic stroke after adjustment for multiple confounding factors compared to the lowest level of bilirubin (hazard ratio [HR], 0.72; 95% CI, 0.58–0.90 in level 3; HR, 0.66; 95% CI, 0.49–0.89 in level 4; P for trend=0.016). The risk of all stroke types also decreased as bilirubin levels increased (HR, 0.81; 95% CI, 0.68–0.97 in level 3; HR, 0.74; 95% CI, 0.58–0.94 in level 4; P for trend=0.0071). However, these associations were not seen in hemorrhagic stroke or in women.
Conclusions— These findings suggest that serum bilirubin might have some protective function against stroke risk in men.
It has been suggested that bilirubin is not only a waste end-product but also an antioxidant3–5⇓⇓ that may protect against diseases associated with oxidative stress.6 In several prospective studies, an inverse relationship has been reported between bilirubin and the following diseases: cardiovascular disease (CVD),7,8⇓ coronary heart disease,9 myocardial infarction,10 ischemic heart disease,11 and all-cause and cancer mortality in men,12 although some of them failed to reach statistical significance in CVD.12 Cross-sectional studies reported similar results with coronary artery disease,13,14⇓ peripheral vascular disease,15,16⇓ carotid intimal-medial thickness,17,18⇓ and stroke.19 This inverse relationship of bilirubin to CVD was confirmed by meta-analysis,20 and bilirubin has been discussed as a therapeutic target for CVD.5,6⇓
In a recent cross-sectional study, it was shown that total bilirubin and stroke prevalence have an inverse association in the representative national data in which higher bilirubin level was not only associated with reduced stroke prevalence but also associated with favorable stroke outcomes.19 However, the relationship between bilirubin and stroke has not been fully understood because of limited reports on this issue. Information such as subtypes of stroke or various ethnic groups is also limited.
Therefore, we analyzed the association between serum bilirubin levels and incidences of ischemic, hemorrhagic, and all stroke types in a large prospective cohort study in Korea.
Materials and Methods
The initial study population consisted of 122 769 participants who underwent health examinations at 2 health examination centers in Seoul from 1994 to 2001. Among the initial subjects, subjects with missing data in the major variables (n=15 503), hemoglobin level <10 g/dL or >20 g/dL (n=1318), past history of CVD, cerebrovascular accident, cancer, history of kidney disease or serum creatinine >2.0 mg/dL, or past history of admission (n=10 928) were excluded. Among the 95 220 (53 217 men) participants, potential Gilbert syndrome group (total bilirubin >34.2 μmol/L [2.0 mg/dL], aspartate aminotransferase <80 IU/L, alanine transaminase <80 IU/L, gamma glutamyl transpeptidase [GGT] <80 IU/L, and no self-reported history of hepatobiliary disease; n=662) and potential hepatobiliary disease group (total bilirubin >34.2 μmol/L or aspartate aminotransferase ≥80 IU/L or aspartate aminotransferase ≥80 IU/L or serum albumin <3.5 g/dL or positive self-reported history of hepatobiliary disease; n=15 834) were also excluded to avoid confounding factors. The definition of the potential Gilbert syndrome group was modified from previous works.21 The potential hepatobiliary disease group was defined according to the US NHANES study.19 Finally, 78 724 health examinees (41 054 men, aged 30–89 years) were selected for the analyses. Consent from each examinee was not specially obtained because the data were collected from routine health examinations. The study was approved by the Institutional review board of human research of Yonsei University.
Baseline information was obtained from the health examinations from 1994 to 2001. Each participant completed systemized questionnaires that include smoking habit (never smoker, ex-smoker, or current smoker), alcohol habit (nondrinker or drinker of any amount of alcohol), regular exercise (yes or no), and other characteristics, including history and any medication. The participants’ weights, heights, and blood pressures were measured in a standardized manner. The BMI was calculated as weight in kilograms divided by height in meters squared. Systolic and diastolic blood pressure were measured in a seated position with a mercury sphygmomanometer or automatic manometer. Fresh fasting serum specimens were analyzed for laboratory tests. Fasting blood glucose, serum cholesterol, liver function tests, and serum bilirubin concentrations were measured by automated biochemical profiling (Hitachi-7600 analyzer, Hitachi Ltd). Bilirubin concentrations were collected in milligram per deciliter units. Serum creatinine was measured by a kinetic rate Jaffe method.
Measurement of Outcome
The outcome variables were incidence of stroke and its subtypes (ischemic, hemorrhagic, and all stroke types) as recorded in hospital admission discharge records from 1994 to 2007 (median follow-up duration, 9.4 years). We ascertained these outcomes from health insurance claim data from the National Health Insurance Corporation and checked fatal cases from the national death certification data.22,23⇓ Ischemic, hemorrhagic, and all stroke types were defined according to the International Classification of Diseases 10th Revision codes: ischemic stroke, I63-I639; hemorrhagic stroke, I60-I629; and all stroke types, I60-I699. For those individuals with >1 event during the follow-up period, we considered only the first event in our statistical analyses. Morbidity was recorded according to the International Classification of Diseases 10th Revision codes in a standardized manner.22–24⇓⇓ The accuracy rate of the International Classification of Diseases codes for cerebrovascular diseases in medical insurance claims for men in Korea was reported as 83.0% in 2000.25
We classified the concentrations of serum bilirubin into 4 levels: 0 to 10.2, 10.3 to 15.3, 15.4 to 22.1, and 22.2 to 34.2 μmol/L (to convert bilirubin from milligrams per deciliter to micromoles per liter, multiply by 17.1). Type 2 diabetes was defined as participants with self-reported diagnosis of type 2 diabetes or participants with fasting blood glucose levels ≥126 mg/dL. Hypertension was defined as a systolic blood pressure of at least 140 mm Hg, a diastolic blood pressure of at least 90 mm Hg, or a self-reported diagnosis of hypertension.
To examine the differences of baseline characteristics of the 4 levels of bilirubin, age-adjusted ANOVA and χ2 test were used. The incidence rates per 100 000 person-year were calculated. Partial Pearson correlation coefficients were acquired to determine the relationship between log-transformed bilirubin and other variables with adjustment for age. Log-transformed bilirubin concentrations were used to achieve normal distribution. Cox proportional hazards models were used to estimate the risk of stroke incidence according to serum bilirubin levels. Crude (model I), age-adjusted (model II), and multivariable-adjusted (model III) analyses were performed with consideration for multicollinearity. We adopted the multivariable model used in the Framingham offspring study,7 in addition to the GGT, a new risk factor for cardiovascular disease mortality.26 Finally, the hazard ratios (HR) in model III were adjusted for age, smoking (nonsmoker, ex-smoker, current smoker), alcohol (yes or no), exercise (yes or no), alanine transaminase, total cholesterol, type 2 diabetes, hypertension, and GGT. The HR and 95% CI were calculated. All analyses were conducted using SAS statistical software version 9.1 (SAS institute Inc) and performed separately for men and women because frequency rates in the 4 bilirubin levels were different between men and women (P<0.0001 by χ2 test). All statistical tests were 2-sided, and statistical significance was determined as P<0.05.
Mean±SD of bilirubin concentration was 15.1±5.6 μmol/L in men (n=41 054) and 11.6±4.7 μmol/L in women (n=37 670; P<0.0001). During a 14-year follow-up, a total of 1964 men and women were hospitalized for stroke. Among all the stroke patients, the proportion of patients having ischemic stroke were 63.1% for men and 57.1% for women, respectively.
The mean age of study participants decreased as levels of serum bilirubin in both men and women increased (P<0.0001). There was not an even distribution of general characteristics among the 4 bilirubin levels which included age, BMI, alanine transaminase, total cholesterol, and smoking habit. GGT and hypertension in women, alcohol, exercise, and type 2 diabetes in men were also associated with bilirubin levels (Table 1). In age-adjusted partial correlation analysis, aspartate aminotransferase, systolic blood pressure, fasting blood glucose, triglyceride, high-density lipoprotein, serum creatinine, and hemoglobin were also correlated with log-transformed bilirubin concentrations (Table 2).
There was strong multicollinearity among aspartate aminotransferase, alanine transaminase, and GGT, and also between total cholesterol and triglyceride. When we analyzed with several alternate models with or without systolic blood pressure, fasting blood glucose, GGT, BMI, creatinine, hemoglobin, triglyceride, high-density lipoprotein, and cholesterol/high-density lipoprotein ratio,19 there was no difference in the final results showing the association between bilirubin levels and stroke incidence (data not shown).
In Cox proportional hazard models, participants with higher levels of bilirubin showed lower HR in men with ischemic stroke before and after adjustment for age (model II) or multiple (model III) confounding factors compared with the lowest level of bilirubin (HR, 0.72; 95% CI, 0.58–0.90 in level 3; HR, 0.66; 95% CI, 0.49–0.89 in level 4; P for trend=0.016). The risk for all stroke types also decreased as bilirubin levels increased (HR, 0.81; 95% CI, 0.68–0.97 in level 3; HR, 0.74; 95% CI, 0.58–0.94 in level 4; P for trend=0.0071). However, these associations were not seen in hemorrhagic stroke or in women (Tables 3 and 4⇓).
We also did the Cox proportional hazard analysis with bilirubin as a continuous variable. For ischemic stroke, a 1 μmol/L increment in bilirubin concentration was associated with a 2% reduction of HR (HR, 0.98; 95% CI. 0.97–0.99; P=0.0138) and a 1-log bilirubin (μmol/L) increment was associated with a 23% reduction in HR (HR, 0.77; 95% CI. 0.63–0.94) in a multivariate-adjusted model for men. For all stroke types, there was a 1% reduction of HR (HR, 0.99; 95% CI, 0.98–1.00; P=0.0382) and a 15% reduction of HR (HR, 0.85; 95% CI, 0.72–0.99; P=0.0358), respectively (data not shown). Similar results were observed when we divided the participants into quartiles (please see Table I, available online at http://stroke.ahajournals.org). We grouped the bilirubin concentrations into 5 levels using the design used in the Framingham offspring study.7 The multivariable adjusted HR, which compared ischemic stroke in the highest quintile to that in the lowest quintile, showed a significant decrease in men (HR, 0.71; 95% CI, 0.56–0.90), but the HR for all stroke types was not significant (please see Table II).
In men, we saw a protective trend of bilirubin against a risk of hemorrhagic stroke in an unadjusted model for the trend test (P for trend=0.0151), but it disappeared when we adjusted for age and other confounding variables (Table 3).
In women, we found significant associations between bilirubin and stroke in an unadjusted model for ischemic stroke (HR, 0.71; 95% CI, 0.54–0.94 in level 3; P for trend=0.0046) and all stroke types (HR, 0.72; 95% CI, 0.58–0.89 in level 3; P for trend=0.0024). In adjusted models, there was no remaining significant relationship between bilirubin and the various stroke types in women (Table 4).
Even after adjusting for age and other well-known risk factors, the serum bilirubin level still was independently associated with the incidence of ischemic and all stroke types dose-dependently in men, but not in women. The result of our study on stroke coincides with previous reports that bilirubin is an independent inverse predictor for CVD.
In previous studies, a U-shape correlation between bilirubin and cardiovascular risk was described9,11⇓ or suggested.7,19⇓ In our report, we divided the bilirubin concentrations into 4 levels by a concentration range in contrast to previous studies to provide information on these concentrations. However, we performed alternate grouping to check this issue. Our analysis using the same grouping as the Framingham offspring study7 showed decreased risk of ischemic stroke in higher bilirubin levels in men. However, there was no significant association between bilirubin and the risk of all stroke types in men. This result implies the possibility that the U-shape was attributable to the lack of analysis for the subtypes of CVD.
In our study, we presented each HR for ischemic, hemorrhagic, and all stroke types rather than total CVD, and found some differences between the results from ischemic stroke and the results from hemorrhagic stroke. A stronger association of bilirubin with the risk of ischemic stroke than with the risk of hemorrhagic stroke may support the antiatherogenic property of bilirubin suggested in previous reports.19
Bilirubin has been shown to be a natural antioxidant.3,4⇓ As an antioxidant, bilirubin demonstrated antiatherogenic function through inhibition of low-density lipoprotein oxidation27,28⇓ or through inhibition of vascular endothelial activation, which may mediate the antiatherogenic properties of heme oxygenase-1.29 The antiatherogenic potential of bilirubin was also shown in human studies. Carotid intima-media thickness, a predictor for atherosclerosis, increased in healthy subjects with low bilirubin levels not only in men17 but also in women.18
However, in longitudinal cohort studies, the association between bilirubin and mortality12 or between bilirubin and cardiovascular disease7 did not reach statistical significance in women. The Framingham Offspring Cohort reported inverse associations between bilirubin and cardiovascular diseases when they analyzed CVD not specific for coronary heart disease or stroke in men but not in women among 4276 participants during a 22-year follow-up. The multivariate-adjusted HR for risk of CVD for the higher bilirubin group (14.53–17.94 μmol/L) compared to the lowest group was 0.59 (95% CI, 0.39–0.89) in men,7 which is comparable with our findings (HR, 0.72; 95% CI, 0.58–0.90) of 15.4 to 22.1 μmol/L for ischemic stroke. These differences in men and women may be attributable to lifestyle factors, including tobacco, alcohol, or diet. Postmenopausal women and men accumulate higher levels of stored iron than young women,30 and high dietary intake of heme iron has been known to be associated with increased coronary heart disease risk.31 We cannot exclude a potential effect of iron load to the heme oxygenase-1 and bilirubin pathway.29
Defining hyperbilirubinemic subgroup and excluding potential liver disease patients was pointed as a key step to clarify the inverse associations between bilirubin and stroke in discussions from other published works.32 We modified the exclusion criteria of the NHANES study: a report on active liver diseases, 2-times higher than normal value of alanine transaminase, bilirubin >34.2 μmol/L, or albumin <3.5 mg/dL, and adjustment for alcohol intake.19 We also excluded the potential Gilbert syndrome group because it has been known that they had lower prevalence and incidence of coronary heart diseases and had higher high-density lipoprotein and antioxidant levels.6 However, we could not find any difference in the risk of stroke among the potential Gilbert syndrome group, the potential hepatobiliary disease group, and the main analysis group in our additional analysis. When we calculated HR without excluding them (n=95 220), the results did not differ from that reported.
This study has several limitations. The representation of our participants is limited and the variables of stroke prognosis or dietary intake are lacking. We also cannot rule out selection bias in our volunteers who may have had healthier lifestyles. Bilirubin concentration was measured only once, but we thought the possibility of confounders from biliary disease might be minimal because we excluded subjects with abnormal liver enzymes and with histories of hepatobiliary diseases. The accuracy rate for cerebrovascular diseases in medical insurance claims for men was reported previously,25 but there were limited data for women.
It was reported that among the stoke cases, ≈15% of the cases were hemorrhagic stroke, 70% of the cases were cerebral infarction, and 15% of the cases were transient ischemic attack (TIA).33 We did not include TIA in the all stroke types category because TIA patients are likely to be missed from the admission record because of the mild, transient symptom.
Our study has a few strengths. First, we analyzed a large population with men and women. Second, we excluded subjects with potential hepatobiliary diseases or Gilbert syndrome. Third, the characteristics of participants were homogenous because they were all urban middle class. Finally, we clarified the specific risk for the subgroups of stroke, including ischemic and hemorrhagic stroke.
Bilirubin assay is available in many laboratories and is not expensive. The results in this large cohort study support the potential role of bilirubin to predict stroke. Our findings suggest that serum bilirubin might have some protective function against stroke in men. Further studies are needed to confirm the association of bilirubin with stroke. A larger sample size or longer follow-up time might have detected significant findings in the risk of hemorrhagic stroke, as well as in the risk of any type of stroke in women. Clarification of the protective bilirubin cut-off level and intervention trials may answer the question of the potential for bilirubin as a therapeutic target for CVD.
The authors are indebted to the staff of the Korean National Health Insurance Corporation, particularly Mungyu Park, the deputy director.
Sources of Funding
This study was supported by a grant of the Seoul R&BD Program, Republic of Korea (10526). Also, this research was supported by a grant from the Korea Institute of Oriental Medicine (KIOM, K09200). The funding sources had no role in the design and conduct of the study.
- Received June 15, 2009.
- Revision received July 10, 2009.
- Accepted July 20, 2009.
- ↵Yusuf S, Reddy S, Ounpuu S, Anand S. Global burden of cardiovascular diseases. Part I: general considerations, the epidemiologic transition, risk factors, and impact of urbanization. Circulation. 2001; 104: 2746–2753.
- ↵Statistical office. Republic of Korea. 2006.
- ↵Stocker R, Yamamoto Y, McDonagh AF, Glazer AN, Ames BN. Bilirubin is an antioxidant of possible physiological importance. Science. 1987; 235: 1043–1046.
- ↵Ollinger R, Bilban M, Erat A, Froio A, McDaid J, Tyagi S, Csizmadia E, Graça-Souza AV, Liloia A, Soares MP, Otterbein LE, Usheva A, Yamashita K, Bach FH. Bilirubin: a natural inhibitor of vascular smooth muscle cell proliferation. Circulation. 2005; 112: 1030–1039.
- ↵Ganotakis ES, Vrentzos GE, Gazi IF, Papadakis JA, Jagroop A, Paraskevas KI, Nair DR, Mikhailidis DP. Fibrinogen, lipoprotein (a), albumin and bilirubin (F-L-A-B) levels and cardiovascular risk calculated using the Framingham equation. In Vivo. 2007; 21: 685–694.
- ↵Troughton JA, Woodside JV, Young IS, Arveiler D, Amouyel P, Ferrières J, Ducimetière P, Patterson CC, Kee F, Yarnell JW, Evans A; PRIME Study Group. Bilirubin and coronary heart disease risk in the Prospective Epidemiological Study of Myocardial Infarction (PRIME). Eur J Cardiovasc Prev Rehabil. 2007; 14: 79–84.
- ↵Breimer LH, Wannamethee G, Ebrahim S, Shaper AG. Serum bilirubin and risk of ischemic heart disease in middle-aged British men. Clin Chem. 1995; 41: 1504–1508.
- ↵Hopkins PN, Wu LL, Hunt SC, James BC, Vincent GM, Williams RR. Higher serum bilirubin is associated with decreased risk for early familial coronary artery disease. Arterioscler Thromb Vasc Biol. 1996; 16: 250–255.
- ↵Schwertner HA, Jackson WG, Tolan G. Association of low serum concentration of bilirubin with increased risk of coronary artery disease. Clin Chem. 1994; 40: 18–23.
- ↵Perlstein TS, Pande RL, Beckman JA, Creager MA. Serum total bilirubin level and prevalent lower-extremity peripheral arterial disease: National Health and Nutrition Examination Survey (NHANES) 1999 to 2004. Arterioscler Thromb Vasc Biol. 2008; 28: 166–172.
- ↵Erdogan D, Gullu H, Yildirim E, Tok D, Kirbas I, Ciftci O, Baycan ST, Muderrisoglu H. Low serum bilirubin levels are independently and inversely related to impaired flow-mediated vasodilation and increased carotid intima-media thickness in both men and women. Atherosclerosis. 2006; 184: 431–437.
- ↵Novotný L, Vítek L. Inverse relationship between serum bilirubin and atherosclerosis in men: a meta-analysis of published studies. Exp Biol Med (Maywood). 2003; 228: 568–571.
- ↵Kim HC, Kang DR, Nam CM, Hur NW, Shim JS, Jee SH, Suh I. Elevated serum aminotransferase level as a predictor of intracerebral hemorrhage: Korea medical insurance corporation study. Stroke. 2005; 36: 1642–1647.
- ↵International classification of diseases and related health problems: manual of the International Statistical Classification of Diseases, Injuries, and Causes of Death, 10th ed. Geneva: World Health Organization; 1992.
- ↵Park JK, Kim KS, Kim CB, Lee TY, Lee KS, Lee DH, Lee S, Jee SH, Suh I, Koh KW, Ryu SY, Park KH, Park W, Wang S, Lee H, Chae Y, Hong H, Suh JS. The accuracy of ICD codes for cerebrovascular diseases in medical insurance claims. Korean J Prev Med. 2000; 33: 76–82.
- ↵Ruttmann E, Brant LJ, Concin H, Diem G, Rapp K, Ulmer H. Vorarlberg Health Monitoring and Promotion Program Study Group. Gamma-glutamyltransferase as a risk factor for cardiovascular disease mortality: an epidemiological investigation in a cohort of 163,944 Austrian adults. Circulation. 2005; 112: 2130–2137.
- ↵Neuzil J, Stocker R. Free and albumin-bound bilirubin are efficient co-antioxidants for alpha-tocopherol, inhibiting plasma and low density lipoprotein lipid peroxidation. J Biol Chem. 1994; 269: 16712–16719.
- ↵Kawamura K, Ishikawa K, Wada Y, Kimura S, Matsumoto H, Kohro T, Itabe H, Kodama T, Maruyama Y. Bilirubin from heme oxygenase-1 attenuates vascular endothelial activation and dysfunction. Arterioscler Thromb Vasc Biol. 2005; 25: 155–160.
- ↵Van der ADL, Peeters PH, Grobbee DE, Marx JJ, van der Schouw YT. Dietary haem iron and coronary heart disease in women. Eur Heart J. 2005; 26: 257–262.
- ↵Mohr JP, Albers GW, Amarenco P, Babikian VL, Biller J, Brey RL, Coull B, Easton JD, Gomez CR, Helgason CM, Kase CS, Pullicino PM, Turpie AGG. Etiology of stroke. Stroke. 1997; 28: 1501–1506.