Direct, indirect and total bilirubin and risk of incident coronary heart disease in the Dongfeng-Tongji cohort

Abstract

Background: Total bilirubin (TBIL) is known to be inversely associated with coronary heart disease (CHD) risk, however, whether this association is dose-response remains inconsistent and it is unclear which subtype of bilirubin is responsible for the potential protective effect.

Methods: We included 12,097 participants who were free of CHD, stroke, cancer and potential liver, biliary and renal diseases at baseline from September 2008 to June 2010 and were followed-up until October 2013. Cox proportional hazards models were used to assess the hazard ratios (HR) and 95% confidence interval (95% CI) of bilirubin with incident CHD risk.

Results: The adjusted HRs for incident CHD increased with increasing direct bilirubin (DBIL) (p for trend = .013). Participants within the highest quintile of DBIL had 30% higher risk of incident CHD compared to those in the lowest quintile (95% CI: 1.07, 1.58). In contrast, compared with subjects in the lowest quintile of TBIL, those in the third quintile had the lowest of 24% risk for CHD incidence (95% CI: 0.63, 0.92), which showed a U-shaped association (p for quadratic trend = .040).

Conclusions: DBIL was associated with a dose-response increased risk for CHD incidence. However, a U-shaped association existed between TBIL, indirect bilirubin and incident CHD risk.

Key messages

• Direct bilirubin is independently associated with incident coronary heart disease (CHD) in a dose-response manner.

• A similarly consistent U-shaped association was found between total bilirubin, indirect bilirubin and incident CHD.

• The potential protective effect of total bilirubin within the normal range on incident CHD should be mainly attributed to mild-to moderate elevated levels of indirect bilirubin.

Keywords:

• Bilirubin
• cohort study
• incident coronary heart disease

Introduction

Serum circulating total bilirubin (TBIL), a bile pigment of heme catabolism, consists of two forms: indirect bilirubin (IBIL) and direct bilirubin (DBIL), the former of which was converted to the latter by the hepatic enzyme UDP-glucuronosyltransferase (UGT1A1) and accounts for up to 70–80% of TBIL [1–3]. Although once regarded as a metabolic waste useful for evaluating liver function, it has been found to have potent antioxidant and anti-inflammatory properties both in vitro and in vivo, which makes it possible to protect against coronary heart disease (CHD) through improving markers of oxidative stress and cellular dysfunction [4,5]. However, in epidemiological studies, the association of TBIL with CHD has been inconsistent [6–17]. Several prospective studies and a recent meta-analysis have generally indicated a linear inverse association between TBIL levels and incident CHD risk [9,15] and a few reported L-shaped [11] or even U-shaped pattern [6,7,10], but others showed modestly positive [8,16,17] or null associations [12–14]. Thus, it remains unclear about the consistency and shape of the association because that majority of the previous studies were conducted in populations with high cardiovascular disease (CVD) risk factors [11,18] or with coexisting morbidities [19,20]. In addition, previous epidemiology studies of CHD in relation to bilirubin were only focused on TBIL without separation of bilirubin types, no studies so far have examined the association of TBIL and its subtype (DBIL or IBIL) together with risk for CHD incidence. Given that TBIL may influence the risk of CHD, one might expect its effect to be mediated via DBIL or IBIL.

Therefore, the study was designed to fit the curve shape of association between TBIL and incident CHD risk and further determine which subtype of bilirubin is responsible for the observed association among a large-scale middle-aged and elderly Chinese population.

Materials and methods

Study population

The data was derived from Dongfeng-Tongji Cohort which was described in detail previously [21]. Briefly, a total of 27,009 retired workers at Dongfeng Motor Corporation (DMC) agreed and completed baseline questionnaire, medical examinations and provided fasting blood samples between September 2008 and June 2010. Simultaneously, all retired employees were covered by the DMC’s health-care service system with specific medical insurance card number and identification number, which made it easy to track health service use, disease prevalence , incidence and mortality. Therefore, the prevalence of major chronic diseases was confirmed though the medical insurance system and medical record reviews. At the follow-up investigation, participants repeated the questionnaire interview, physical examinations and blood collection as those during the baseline survey. Among the above participants, 25,978 individuals (96.2%) were followed-up for the first time from June 2013 to October 2013. In this study, we excluded subjects with CHD, stroke, cancer (n = 6712), missing data of serum bilirubin (n = 2043) and other covariates (n = 978), potential liver diseases group including chronic hepatitis, liver cirrhosis, fatty liver disease and liver cyst; individuals with increased liver enzymes levels including alanine aminotransferase, ALT ≥80 IU/L, aspartate aminotransferase, AST ≥80 IU/L or alkaline phosphatase, ALP ≥220 IU/L and abnormal urobilirubin and urobilinogen; n = 2257) renal diseases (nephritis and nephrolith; n = 534), biliary diseases (cholecystitis and gallstone; n = 1357) at baseline. Finally, the eligible sample size for analyses was 12,097 subjects in the present study (Figure 1). Compared to the excluded individuals, the included participants were younger and more likely to be current smokers or drinkers, had lower levels of three types of bilirubin, body mass index, liver enzymes and lower prevalence of hypertension, diabetes and hyperlipidemia (Supplemental Table S1). Overall, the analytical sample was relatively “healthy”. The study was approved by the Ethics and Human Subject committee of the School of Public Health, Tongji Medicine College and Dongfeng General Hospital, DMC. Written informed consents were received from all participants.

Figure 1. The flow chart of study population in the present study.

Assessment of bilirubin

Bilirubin and other biochemical indexes including lipids, glucose, hepatic function, and renal function were detected at the DMC-owned hospital’s laboratory with ARCHITECT Ci8200 automatic analyzer (ABBOTT Laboratories, Abbott Park, IL).

Definition of covariates

Trained interviewers used a structured questionnaires to collect baseline information, including age, sex and education levels (primary school or below, middle school, high school or above), lifestyle such as smoking status (nonsmoker, ex-smoker, current smoker; current smoker was defined as smoking at least one cigarette per day for more than half a year), alcohol consumption status (nondrinker, former drinker, current drinker; current drinker was defined as drinking at least one time per week for more than half a year) and physical activity (yes/no; yes means one who regularly exercised more than 20 minutes per day in the last six months), diet, occupational history, environmental exposure, family and medical histories (diagnosed medial conditions, use of health services and use of medicines for the most recent two weeks). Body mass index (BMI) was calculated as mass (kg) divided by the square of height (m²). Hypertension was determined as the presence of one of the following conditions: self-reported physician diagnosis of hypertension or blood pressure ≥140/90 mm Hg or current use of antihypertensive medication [22,23]. Participants with self-reported physician diagnosis of diabetes, fasting glucose level    ≥ 7.0 mmol/L or taking oral hypoglycemic medication or insulin were defined as diabetes [22,23]. Subjects with total cholesterol >5.72 mmol/L or triglycerides >1.70 mmol/L at medical examination or a previous self-reported physician diagnosis of hyperlipidemia or taking lipid-lowering medication were considered as hyperlipidemia [22–24].

Ascertainment of incident CHD

All retired employees were covered by DMC’s health-care service system and each person had a unique medical insurance card number and ID which was used to retrieve disease incidence and mortality. The diagnosis of CHD was identified through medical insurance system and medical records and death certificates according to well-accepted international standards. The incident CHD was defined as the first hospital admission with an occurrence of an angina pectoris (ICD-10 code I20), acute myocardial infarction (AMI, I21), subsequent myocardial infarction (I22), other forms of acute (I24) or chronic (I25) heart disease, percutaneous transluminal coronary angioplasty or coronary artery bypass graft and cardiac arrest (I46) or death with CHD (I20–I25) as the underlying cause [23].

Statistical analyses

Baseline characteristics of participants were compared across quintiles of DBIL, IBIL and TBIL. Continuous variables were presented as mean ± SD or median (interquartile range) and categorical variables as numbers (percentages). Trends of baseline characteristics across the quintiles of DBIL levels were tested by linear or logistic regression using the median value of DBIL for each quintile as an ordinal variable after adjusting for age and sex. Cox proportional hazard regression models were used to estimate the relationship of three types of bilirubin levels (separately for DBIL, IBIL and TBIL) with risk of incident CHD after adjusting for potential confounders. Model 1 was adjusted for age, sex, BMI, smoking status, drinking status, physical activity, education levels, hypertension, diabetes, hyperlipidemia and family history of CHD. Model 2 was additionally adjusted for ALP, ALT and AST. To assess the shape of the relationship of bilirubin levels with CHD incidence, both linear trend and nonlinear trend were examined, respectively. The test for linear trend was performed using the median bilirubin (TBIL, IBIL and DBIL, respectively) concentration for each quintile as a continuous variable in the multivariate model. We also entered bilirubin levels (TBIL, IBIL and DBIL, respectively) both as a linear and quadratic term in its continuous form in the final model to examine the potential non-linearity associations. Furthermore, the curve shape relationships of bilirubin levels (TBIL, IBIL and DBIL, respectively) with risk for CHD incidence were further examined and plotted by restricted cubic spline Cox regression utilizing four knots placed at the 5th, 35th, 65th and 95th percentiles of three types of bilirubin levels respectively [11,22,23], with 2.0 μmol/L (median value of the first quintile of DBIL) as the reference group universally. Considering that DBIL was dose-responsive related to increased risk of CHD incidence, we further applied Cox regression to test DBIL as continuous variable (after log_e-transformation) in unadjusted and multivariable adjusted analysis. In addition, stratified analyses for the dose-response association between DBIL and risk of CHD incidence were performed according to major baseline characteristics. Potential interactions were tested by adding the products of those covariates with DBIL levels in total study population, respectively. All statistical analyses were carried out using SAS version 9.3 (SAS institute Inc., Cary, NC). A 2-sided p value <.05 was considered to be statistically significant.

Results

Characteristics of study population

Overall, the mean age of participants at baseline was 62.7 years and 45.2% were males. The median (interquartile range) overall were 3.7 (2.8, 4.8) μmol/L for DBIL, 9.5 (7.0, 12.2) μmol/L for IBIL and 13.0 (10.4, 16.5) μmol/L for TBIL, respectively. Baseline characteristics of participants were presented in Table 1. Participants with higher serum DBIL levels were older and more likely to be men, low educational, physical active, current smokers, current drinkers and present with hypertension. In addition, they also had elevated liver enzymes levels of ALT, AST and ALP. In contrast, the prevalence of hyperlipidemia and family history of CHD decreased as DBIL levels increased (all p for trend < .05). The baseline characteristics according to IBIL and TBIL were shown in Supplemental Table S2 and S3. Similar baseline characteristic distribution was found for the three types of bilirubins.

Table 1. Baseline characteristics of participants across quintiles of direct bilirubin levels.^a

	Quintiles of direct bilirubin level (μmol/L)
Characteristics	Q1	Q2	Q3	Q4	Q5	p^b trend
n	2423	2480	2494	2439	2261
Total bilirubin	11.3 (8.1, 14.0)	10.2 (8.9, 11.9)	12.2 (10.7, 14.0)	15.2 (13.2, 17.2)	19.5 (17.0, 23.6)	–
Indirect bilirubin	9.6 (6.0, 12.0)	7.3 (6.0, 8.8)	8.6 (7.0, 10.2)	10.8 (8.7, 12.7)	13.5 (11.0, 17.2)	–
Direct bilirubin	2.1 (1.5, 2.3)	3.0 (2.8, 3.1)	3.7 (3.5, 3.9)	4.5 (4.3, 4.9)	6.0 (5.5, 7.0)	–
Age (year)	61.8 (7.8)	61.6 (7.7)	62.9 (7.9)	63.2 (7.8)	64.3 (7.6)	.005
Male, n (%)	772 (31.9)	770 (31.1)	1097 (44.0)	1285 (52.7)	1540 (68.1)	<.0001
Body mass index (kg/m^2)	24.3 (3.3)	23.9 (3.2)	24.0 (3.1)	24.0 (3.1)	24.0 (3.1)	.075
High school or above, n (%)	839 (34.6)	904 (36.5)	804 (32.2)	793 (32.5)	644 (28.5)	<.0001
Physical activity, n (%)	2145 (88.5)	2178 (87.8)	2227 (89.3)	2213 (90.7)	2063 (91.2)	.002
Current smokers, n (%)	356 (14.7)	382 (15.4)	472 (18.9)	552 (22.6)	582 (25.7)	.013
Current drinkers, n (%)	427 (17.6)	403 (16.3)	554 (22.2)	628 (25.8)	815 (36.1)	<.0001
Family history of CHD, n (%)	124 (5.1)	150 (6.1)	122 (4.9)	107 (4.4)	70 (3.1)	.030
Hypertension, n (%)	1157 (47.8)	1145 (46.2)	1141 (45.8)	1213 (49.7)	1201 (53.1)	.026
Diabetes, n (%)	360 (14.9)	364 (14.7)	397 (15.9)	391 (16.0)	374 (16.5)	.644
Hyperlipidemia, n (%)	1301 (53.7)	1366 (55.1)	1275 (51.1)	1131 (46.4)	929 (41.1)	<.0001
ALT (U/L)	19.0 (15.0, 24.0)	20.0 (15.0, 26.0)	20.0 (15.0, 26.0)	20.0 (15.0, 27.0)	21.0 (16.0, 28.0)	<.0001
AST (U/L)	22.0 (18.0, 26.0)	22.0 (19.0, 27.0)	23.0 (19.0, 27.0)	23.0 (19.0, 28.0)	24.0 (20.0, 29.0)	<.0001
ALP (U/L)	79.0 (68.0, 100.0)	88.0 (73.0, 107.0)	88.0 (74.0, 107.0)	88.0 (73.0, 104.0)	87.0 (72.0, 103.0)	<.0001

CHD: coronary heart disease; ALT: alanine aminotransferase; AST: aspartate aminotransferase; ALP: alkaline phosphatase.

^a Data were means (SD), median (P₂₅, P₇₅) for continuous variables or numbers (%) for category variables, respectively.

^b p values were calculated after adjustment for age and sex except for itself.

Associations between bilirubin levels and incident CHD risk

As shown in Table 2, the increasing DBIL levels were independently associated with elevated risk of CHD incidence in a dose-response manner. Compared with the first quintile of DBIL, the fully adjusted HRs (95% CI) for incident CHD from the second to the fifth quintile of DBIL were 1.21 (1.01, 1.47), 1.09 (0.90, 1.32), 1.26 (1.05, 1.53) and 1.30 (1.07, 1.58), respectively (p for tend = .013). Simultaneously, this dose-response relationship was further demonstrated with restricted cubic splines (p for linear trend = .002; Figure 2(a)). The fully adjusted HR per 1 μmol/L change in log_e DBIL was 1.23 (95% CI: 1.09, 1.39). However, a U-shaped relationship was observed between TBIL, IBIL and incident CHD risk. Compared with participants in the first quintile of TBIL, those in the third quintile had the lowest risk for CHD incidence (HR: 0.76; 95% CI: 0.63, 0.92; p for quadratic trend = .040). Similar results were observed for IBIL. Additionally, the fitted restricted cubic splines further showed the apparent U-shaped relationships of IBIL, TBIL with risk for CHD incidence (Figure 2(b,c), respectively).

Figure 2. Multivariable adjusted restricted cubic splines for the relationship between DBIL (a), IBIL (b), TBIL (c) levels and risk for CHD incidence in a Cox proportional hazard model. The model was adjusted for age, sex, body mass index, smoking, drinking, physical activity, education levels and family history of CHD, hypertension, hyperlipidemia, diabetes, ALT, AST and ALP. The solid lines and dashed lines indicates the adjusted hazard ratios (HR) and 95% confidence interval (CI), respectively. DBIL: direct bilirubin; IBIL: indirect bilirubin; TBIL: total bilirubin; ALT: alanine aminotransferase; AST: aspartate aminotransferase; ALP: alkaline phosphatase; CHD: coronary heart disease.

Table 2. Hazard ratios (95% CI) for CHD incidence according to three types of serum bilirubin levels.

		Hazard ratios (95% CI)
Bilirubin types	Incidents/Participants	Unadjusted	Model 1^a	Model 2^b
Direct bilirubin (range; median, μmol/L)
Quartile 1 (0–2.6; 2.1 μmol/L)	198/2423	1.00 (Ref.)	1.00 (Ref.)	1.00 (ref.)
Quartile 2 (2.6–3.3; 3.0 μmol/L)	253/2480	1.23 (1.02, 1.48)	1.24 (1.03, 1.50)	1.21 (1.01, 1.47)
Quartile 3 (3.3–4.1; 3.7 μmol/L)	241/2494	1.18 (0.98, 1.43)	1.12 (0.92, 1.35)	1.09 (0.90, 1.32)
Quartile 4 (4.1–5.1; 4.5 μmol/L)	260/2439	1.34 (1.12, 1.61)	1.30 (1.08, 1.57)	1.26 (1.05, 1.53)
Quartile 5 (5.1–24.8; 6.0 μmol/L)	252/2261	1.44 (1.19, 1.73)	1.34 (1.10, 1.62)	1.30 (1.07, 1.58)
p for trend	–	.0001	.005	.013
Natural loge-transformed continuous  Direct bilirubin, per 1μmol/L	–	1.31 (1.16, 1.47)	1.26 (1.11, 1.42)	1.23 (1.09, 1.39)
Indirect bilirubin (range; median, μmol/L)
Quartile 1 (1.5–6.6; 5.3 μmol/L)	252/2440	1.00 (Ref.)	1.00 (Ref.)	1.00 (ref.)
Quartile 2 (6.6–8.5; 7.5 μmol/L)	243/2341	0.98 (0.82, 1.17)	0.93 (0.78, 1.11)	0.93 (0.78, 1.11)
Quartile 3 (8.5–10.5; 9.5 μmol/L)	231/2453	0.92 (0.77, 1.10)	0.88 (0.74, 1.06)	0.89 (0.74, 1.07)
Quartile 4 (10.5–13.1; 11.6 μmol/L)	232/2494	0.88 (0.73, 1.05)	0.82 (0.69, 0.98)	0.83 (0.69, 0.99)
Quartile 5 (13.1–50.8; 15.3 μmol/L)	246/2369	1.01 (0.84, 1.20)	0.94 (0.79, 1.13)	0.94 (0.79, 1.13)
p for quadratic trend	–	.065	.014	.021
Total bilirubin (range; median, μmol/L)
Quartile 1 (2.5–9.9; 8.2 μmol/L)	258/2412	1.00 (ref.)	1.00 (ref.)	1.00 (ref.)
Quartile 2 (9.9–12.0; 11 μmol/L)	239/2534	0.90 (0.75, 1.07)	0.89 (0.74, 1.06)	0.89 (0.74, 1.06)
Quartile 3 (12.0–14.1; 13.0 μmol/L)	196/2315	0.82 (0.68, 0.99)	0.76 (0.63, 0.91)	0.76 (0.63, 0.92)
Quartile 4 (14.1–17.5; 15.7 μmol/L)	238/2422	0.93 (0.78, 1.11)	0.89 (0.75, 1.07)	0.90 (0.75, 1.07)
Quartile 5 (17.5–59.7; 20.0 μmol/L)	273/2414	1.08 (0.91, 1.28)	1.01 (0.85, 1.20)	1.01 (0.84 1.20)
p for quadratic trend	–	.013	.026	.040

CHD: coronary heart disease; ALT: alanine aminotransferase; AST: aspartate aminotransferase; ALP: alkaline phosphatase.

^a Adjusted for age, sex, body mass index, education levels, smoking, drinking, physical activity, history of CHD, hypertension, hyperlipidemia and diabetes.

^b Adjusted for model 1 plus liver enzymes including ALT, AST and ALP.

Stratified analyses for association between DBIL and risk of CHD incidence

The significant dose-response relationship between DBIL and risk for CHD incidence was more pronounced in female, individuals younger than 65 years of age, non-current smokers and drinkers, subjects without hypertension, diabetes and hyperlipidemia. However, the significant interaction was only found between DBIL levels and hyperlipidemia (Table 3).

Table 3 Associations of CHD incidence with serum direct bilirubin levels stratified by baseline characteristics^a.

Stratified covariates	Q1	Q2	Q3	Q4	Q5	p trend	p for interaction
Age							.896
<65						.042
Incidents / participants	84/1453	118/1667	104/1354	120/1562	119/1562
HR (95% CI)	1.00 (ref.)	1.28 (0.96, 1.70)	1.36 (1.01, 1.83)	1.36 (1.02, 1.82)	1.37 (1.02, 1.84)
≥ 65						.054
Incidents / participants	116/897	142/939	127/889	140/869	134/905
HR (95% CI)	1.00 (ref.)	1.08 (0.85, 1.39)	1.12 (0.87, 1.45)	1.32 (1.03, 1.70)	1.25 (0.97, 1.62)
Sex							.941
Male						.245
Incidents / participants	110/1067	140/1181	127/1035	132/1106	125/1075
HR (95% CI)	1.00 (ref.)	1.02 (0.79, 1.32)	1.18 (0.91, 1.53)	1.17 (0.91, 1.51)	1.13 (0.87, 1.47)
Female						.017
Incidents / participants	87/1283	124/1476	117/1156	120/1428	122/1290
HR (95% CI)	1.00 (ref.)	1.26 (0.95, 1.66)	1.40 (1.06, 1.86)	1.25 (0.94, 1.65)	1.48 (1.12, 1.96)
Smoking status							.316
Non-current smokers						.005
Incidents / participants	150/2057	174/1885	174/1757	212/1972	212/2082
HR (95% CI)	1.00 (ref.)	1.30 (1.04, 1.62)	1.22 (0.98, 1.53)	1.41 (1.14, 1.75)	1.39 (1.12, 1.73)
Current smokers						.956
Incidents / participants	62/469	51/444	58/514	51/453	60/464
HR (95% CI)	1.00 (ref.)	0.82 (0.56, 1.19)	0.85 (0.59, 1.22)	0.89 (0.61, 1.31)	0.93 (0.64, 1.34)
Drinking status							.905
Non-current drinkers						.002
Incidents / participants	140/1771	173/1833	197/1932	213/1894	201/1840
HR (95% CI)	1.00 (ref.)	1.26 (1.01, 1.58)	1.21 (0.97, 1.51)	1.42 (1.14, 1.76)	1.42 (1.13, 1.77)
Current drinkers						.872
Incidents / participants	47/554	62/569	49/584	57/544	65/576
HR (95% CI)	1.00 (ref.)	1.06 (0.72, 1.57)	0.89 (0.60, 1.33)	1.05 (0.70, 1.55)	1.03 (0.70, 1.52)
Hypertension							.894
No						.045
Incidents / participants	75/1262	73/1205	82/1164	96 / 1303	115 / 1306
HR (95% CI)	1.00 (ref.)	0.94 (0.68, 1.30)	1.03 (0.75, 1.42)	1.11 (0.98, 1.51)	1.26 (1.03, 1.71)
Yes						.129
Incidents / participants	122/1157	170/1156	165/1254	151 / 1091	155 / 1199
HR (95% CI)	1.00 (ref.)	1.38 (1.09, 1.75)	1.19 (0.94, 1.51)	1.32 (1.03, 1.69)	1.30 (1.01, 1.66)
Hyperlipidemia							.004
No						.005
Incidents / participants	54/1120	74/1052	100/1153	119 / 1285	140 / 1485
HR (95% CI)	1.00 (ref.)	1.33 (0.94, 1.90)	1.23 (0.88, 1.73)	1.46 (1.05, 2.03)	1.59 (1.15, 2.20)
Yes						.379
Incidents / participants	143/1299	169/1309	147/1265	128 / 1109	130 / 1020
HR (95% CI)	1.00 (ref.)	1.16 (0.93, 1.46)	1.08 (0.85, 1.36)	1.13 (0.88, 1.44)	1.15 (0.90, 1.48)
Diabetes							.843
No						.019
Incidents / participants	149/2051	185/2128	197/2097	202 / 2048	189 / 1887
HR (95% CI)	1.00 (ref.)	1.18 (0.95, 1.47)	1.16 (0.94, 1.44)	1.28 (1.03, 1.60)	1.31 (1.05, 1.64)
Yes						.263
Incidents / participants	47/360	76/413	38/348	58 / 391	63 / 374
HR (95% CI)	1.00 (ref.)	1.34 (0.92, 1.94)	0.95 (0.61, 1.47)	1.25 (0.84, 1.86)	1.34 (0.90, 1.99)

CHD: coronary heart disease; ALT: alanine aminotransferase; AST: aspartate aminotransferase; ALP: alkaline phosphatase.

^a Adjusted covariates for age, sex, body mass index, education levels, smoking, drinking, physical activity, history of CHD, hypertension, hyperlipidemia, diabetes, liver enzymes including ALT, AST, and ALP except for the stratified covariate itself.

Discussion

In the present study, we unexpectedly found that DBIL is independently associated with a linear dose-response increased risk for CHD incidence in middle-aged and old Chinese. Whereas the results showed a U-shaped association between TBIL, IBIL levels and risk of CHD incidence, which might indicate that only mild-to-moderate elevated TBIL levels within physiological concentration range contribute to decreased risk for CHD incidence.

It is worth noting that, contrary to our expectations, DBIL levels were associated with an increased but not decreased risk of incident CHD in the current study. To the best of our knowledge, the present study is the first to investigate the association between DBIL levels and risk for CHD incidence in a prospective study, we could not directly compare our results with previous studies because there is no data showing the effect of DBIL on CHD incidence . Results from two independent cohorts including Dongfeng-Tongji cohort and the Singapore Chinese Health Study found elevated DBIL were associated with an increased risk of incident type 2 diabetes, which is consistent with our finding [25]. Evidence suggested that diabetes and CVD may share an underlying cause(s), a theory known as the “common soil” hypothesis [26–28]. In fact, the vast majority of diabetic patients die of cardiovascular complications [27]. Thus, our observation is partly supported by the two independent cohorts study. Similarly, other two prospective study showed that higher DBIL was associated with greater stroke severity among ischemic stroke patients [29] and mortality in patients with idiopathic pulmonary arterial hypertension [30]. All the above studies indicated that DBIL was more likely a risk factor than its potential antioxidant property in oxidative stress-mediated diseases. On the contrary, Sung et al. found an inverse relationship between conjugated bilirubin and coronary artery calcium score (CACS), but the cross-sectional study could not infer a causal relationship and CACS was only an indicator of atherosclerosis [31]. Thus, the contrast results might be due to study design and different end points. In particular, subgroup analyses showed that the dose-response association was more evident in females and relatively healthy individuals such as participants younger than 65 of age, non-current smokers or drinkers, subjects without hypertension, diabetes and hyperlipidemia. Although the potential pathophysiological mechanisms were unclear, it is possible that the presence of above CHD risk factors might cover up the effect of DBIL on risk of CHD.

The biological mechanisms linking elevated DBIL and higher CHD risk are still unclear. The serum DBIL in human usually only accounted for 20–30% of TBIL [1–3]. Compared with lipophilic IBIL, DBIL is soluble in serum and only weakly bound to albumin and thus more easily available in active form. Meanwhile, water-soluble DBIL might be difficult to enter into the vascular intima within the atherosclerotic plaque to play an antioxidant role [1]. As reported by Franchini et al. given that markedly elevated bilirubin levels may exert toxic effects itself in certain situation, DBIL was more likely to be a kind of toxic waste than its potential antioxidant property. In fact, higher DBIL may indicate hepatocellular injury if TBIL is within normal range [32], thus the positive association of DBIL levels with incident CHD might reflect the relationship between hepatic-biliary dysfunction and CHD risk.

On the other hand, the U-shaped associations between TBIL, IBIL levels and incident CHD was found in the present study, which was consistent with some previous studies. Lars et al. firstly reported a U-shaped relationship between serum TBIL and risk of ischemic heart disease in middle-aged British men [6]. Another nested case-control study confirmed the U-shaped association between TBIL and CHD risk once again [10]. More recently, the biggest prospective study conducted by Horsfall et al. actually showed a U-shaped appearance between TBIL and CHD incidence through the restricted cubic spline graph, even though they reported an L-shaped association between TBIL and CVD incidence [11]. Regardless of what the relationship was L-patterned or U-shaped, all of these prospective studies implied that the inverse association was not that friendly linear but had a true nadir, that is to say, the risk for CHD or CVD incidence would no longer obviously decrease when median TBIL levels reaches up to approximately 10 to 15 μmol/L [6,10,11]. However, a recent prospective study showed a log-linear inverse association between TBIL levels and risk for CVD incidence in the Netherlanders with a mean age of 48 years [15]. The median TBIL level in that study was 7 μmol/L and the range of TBIL actually fluctuated within 14 μmol/L (within the normal reference) among most participants; whereas in our study, the participants had a mean age of 62.7 years and the median TBIL level was up to 13.0 μmol/L (ranging from 2.5 to 59.7 μmol/L), which was similar to those reported for the population from the Framingham offspring cohort [16] and the United kingdom primary care database [11]. In addition, an earlier meta-analysis reported an inverse dose-response relationship between TBIL and atherosclerosis disease including CHD, subclinical disease (subclinical carotid plaques and coronary stenosis) and peripheral artery disease [9]. However, the study was only limited to men and included prospective, cross-sectional and nested case-control studies. In fact, the metal-analysis took U-shaped and linear study into account but only claimed a dose-response relation. The distribution difference of TBIL levels in different study might be mainly due to inherent differences in population characteristics such as age [33] and race [34,35], health status [18,20] and sample size. Thus, different TBIL levels in different population might contribute to the inconsistent findings. Anyway, it could be inferred that the log-linear association was mainly established within the normal range of 14 μmol/L, which was similar to our finding that the risk for CHD incidence decreased gradually with increasing TBIL levels when TBIL was below 12 μmol/L. In fact, this inverse association was just first half part of the whole U-shaped association in our study. Meanwhile, it should be noted that elevated TBIL levels beyond the normal range might indicate potential hepatocellular injury such as hepatic or obstructive jaundice followed by increased serum transaminases and alkaline phosphatase [1], while elevated liver enzymes were associated with increased risk of CHD incidence [36,37]. Consequently, the U-shaped association might be the result of a combined effect of potential antioxidant effect and hepatocellular toxicity. The overall evidence from these studies indicated that it was actually a U-shaped association between TBIL levels and risk for CHD incidence in the general population. Considering that TBIL is mainly composed of IBIL and there is a consistently similar U-shaped relationship of IBIL levels with incident CHD risk, we might speculate that the potential anti-atherosclerosis effect of TBIL found in previous studies should be mainly attributed to mild-to moderate increased IBIL levels.

There were several limitations in our study. Firstly, serum bilirubin at baseline was measured only once, so we were unable to account for within-individual variability in a period of time and the bilirubin levels might not reflect a long-term exposure status. Secondly, the current study was conducted in middle-aged and older Chinese and the participants with hypertension, diabetes and hyperlipidemia at baseline were higher than general population, and therefore the findings might not be generalized to populations of other ages, different health conditions or other ethnicities. Thirdly, this was the first prospective study which found that DBIL was independently related to increased risk for CHD incidence, which might be only hypothesis generating and should be confirmed or refuted by future larger, more rigorously designed prospective studies. Finally, the follow-up period of this study was comparatively shorter than previous studies.

Conclusions

In summary, DBIL was independently associated with a dose-response increased risk for CHD incidence. However, a U-shaped association existed between TBIL, IBIL and risk of CHD incidence, which implied that only mildly-moderately elevated TBIL levels within the normal range might be associated with decreased risk of incident CHD. Further studies are warranted to investigate the role of IBIL rather TBIL as a potential protective agent for CVD diseases.

Acknowledgements

The authors would like to acknowledge all researchers for participating in the present Dongfeng-Tongji Cohort study as well as volunteers for assisting in collecting the sample, questionnaire data and clinic data.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This work was supported by the National Natural Science Foundation of China (81373093 and 81673139) and the National Key Program of Research and Development of China (2016YFC0900800).