Genetically elevated bilirubin and risk of ischaemic heart disease: three Mendelian randomization studies and a meta-analysis

Abstract

Background

Elevated plasma levels of bilirubin, an endogenous antioxidant, have been associated with reduced risk of ischaemic heart disease (IHD) and myocardial infarction (MI). Whether this is a causal relationship remains unclear.

Objective

We tested the hypothesis that elevated plasma bilirubin is causally related to decreased risk of IHD and MI.

Design

We used a Mendelian randomization approach and three independent studies from Copenhagen, Denmark. We measured bilirubin in 43 708 white individuals from the general population, and genotyped rs6742078 G>T in the uridine diphosphate glucuronosyltransferase 1A1 (UGT1A1) gene in 67 068 individuals, of whom 11 686 had IHD.

Results

Third versus first tertile of baseline bilirubin levels was associated with 134% increased bilirubin levels, with sex- and age-adjusted hazard ratios (HRs) of 0.86 [95% confidence interval (CI), 0.76–0.98; P = 0.02] for IHD and 0.81 (95% CI, 0.66–0.99; P = 0.04) for MI, but with corresponding multifactorially adjusted HRs of 0.93 (95% CI, 0.82–1.06; P = 0.29) and 0.90 (95% CI, 0.73–1.12; P = 0.35). UGT1A1 rs6742078 TT versus GG genotype was associated with 95% increased bilirubin levels (P < 0.001); TT versus GG genotype was associated with odds ratios (ORs) of 1.03 (95% CI, 0.96–1.11; P = 0.73) for IHD and 1.01 (95% CI, 0.92–1.12; P = 0.68) for MI. Finally, in a meta-analysis of the present three studies and eight previous studies including a total of 14 711 cases and 60 324 controls, the random effects OR for ischaemic cardiovascular disease for genotypes with approximately 100% increased bilirubin levels versus reference genotypes was 1.01 (95% CI, 0.88–1.16).

Conclusion

These data suggest that plasma bilirubin is not causally associated with risk of IHD.

Introduction

According to ‘the antioxidant hypothesis’, biological antioxidants have an important role in protecting tissues and organs against oxidative damage [1]. Elevated plasma levels of bilirubin, an endogenous antioxidant [2], have been associated with reduced risk of ischaemic heart disease (IHD) and myocardial infarction (MI) in previous mainly case–control and retrospective epidemiological studies [3, 4]. However, whether these associations reflect a true atheroprotective effect of bilirubin rather than confounding or reverse causation remains unknown [5, 6].

Mendelian randomization is an epidemiological approach based on the fact that individuals inherit genetic variants randomly from their parents [7]. Genetic variants with effect on plasma bilirubin are therefore ideal for avoiding confounding and reverse causation, limitations that are inherent to observational epidemiological studies [7]. Genetic variation in the uridine diphosphate glucuronosyltransferase 1A1 gene (UGT1A1) is a common cause of elevated plasma bilirubin [8] and is therefore useful for testing whether a lifelong elevated bilirubin level is a direct cause of reduced risk of IHD and MI using a Mendelian randomization approach.

We tested the hypothesis that elevated plasma bilirubin is causally related to decreased risk of IHD and MI, using a Mendelian randomization approach. Accordingly, we first tested whether elevated baseline plasma bilirubin levels predicted decreased risk of IHD and MI in the Copenhagen General Population Study (CGPS), a study of 46 538 white individuals from the Danish general population followed for up to 7.5 years. Secondly, we investigated whether the genetic variant UGT1A1 rs6742078, previously shown to be in strong linkage disequilibrium with the TA repeat polymorphism underlying Gilbert's syndrome [9], was associated with elevated plasma bilirubin in the CGPS. Thirdly, we investigated whether this variant was associated with reduced risk of IHD and MI in three studies: the CGPS; the Copenhagen City Heart Study (CCHS; a prospective study comprising 10 264 white individuals from the Danish general population); and the Copenhagen Ischemic Heart Disease Study (CIHDS; a case–control study of 5133 IHD cases and 5133 healthy controls). Finally, we conducted a meta-analysis, including these three studies as well as eight earlier studies, of the association between genetically elevated bilirubin levels and risk of ischaemic cardiovascular disease; a total of 14 711 cases and 60 324 controls were included in the meta-analysis.

Methods

Study cohorts

The present study included three cohorts of white individuals of Danish descent (CGPS, CCHS and CIHDS). These cohorts were defined so that no individuals appeared more than once in any of the three analysis groups. The studies were approved by institutional review boards and Danish ethics committees, and conducted according to the principles of the Declaration of Helsinki. Written informed consent was obtained from participants.

The Copenhagen General Population Study

The CGPS [10, 11] is an ongoing prospective study of the Danish general population initiated in 2003 and still recruiting. Individuals are randomly selected to represent the Danish general population aged 20 to >80 years. The aim is to include more than 100 000 participants, with a focus on all multifactorial diseases. At the time of genotyping for the present study, 51 671 individuals had been included, 5133 of whom served as controls in the CIHDS (see below); in addition, 2830 individuals with prevalent IHD at baseline were excluded from the prospective analyses. Thus, a total of 43 708 participants remained in the CGPS, including 1583 with incident IHD and 587 with incident MI. To achieve maximum statistical power, both prevalent and incident cases of IHD (n = 4413) and MI (n = 1815) in the CGPS were included in the genetic analysis of all three studies combined (CGPS + CCHS + CIHDS). For a detailed definition of IHD and MI diagnoses, see the Supporting Information.

The Copenhagen City Heart Study

The CCHS [10-12] is a prospective study of the Danish general population initiated in 1976–1978 with follow-up examinations in 1981–1984, 1991–1994 and 2001–2003. Individuals were randomly selected to represent the Danish general population aged 20 to >80 years. We included in the analyses 10 264 participants who gave blood for DNA analysis at the 1991–1994 and/or 2001–2003 examinations. After exclusion of those with prevalent IHD at baseline (n = 83), 2057 with incident IHD and 967 with MI were included in the prospective analyses. Both prevalent and incident cases of IHD (n = 2140) and MI (n = 1040) in the CCHS were included in the genetic analysis of all three studies combined (CGPS + CCHS + CIHDS). IHD and MI were diagnosed as in the CGPS.

The Copenhagen Ischemic Heart Disease Study

The CIHDS [10, 11] comprised 5133 patients from the greater Copenhagen area referred for coronary angiography to Copenhagen University Hospital during the period 1991–2009 and 5133 controls without IHD from the CGPS matched by age and gender. In addition to a diagnosis of IHD or MI (defined as in the CGPS and CCHS), these cases had stenosis/atherosclerosis on coronary angiography and/or a positive exercise electrocardiography test.

Meta-analysis

All studies investigating an association between genetically elevated plasma bilirubin owing to genetic variation in UGT1A1 and risk of ischaemic cardiovascular disease published up until March 2012 were considered in the meta-analysis. Studies reported in English were identified in PubMed, Embase and Web of Science databases using the search criteria (bilirubin or UGT1A1) and (IHD or MI), and references therein were reviewed. We included studies with available genotype frequencies in both cases and controls for variants in UGT1A1 with effect on plasma bilirubin levels. We included a total of 11 studies, including the CGPS, CCHS and CIHDS.

Laboratory analyses

Venous blood samples were drawn in the nonfasting state into standard gel tubes; serum was harvested, after centrifugation within 1 h of blood sampling. Serum samples were stored protected from light at 4 °C until measurement (within 1–16 h) and were expressed as plasma levels.

Nonfasting plasma levels of total cholesterol, LDL cholesterol, HDL cholesterol, triglycerides and gamma glutamyl transferase (GGT) were measured using routine assays (Konelab, Helsinki, Finland, and Boehringer Mannheim, Mannheim, Germany). To ensure accurate measurement at higher triglyceride levels, LDL cholesterol was calculated using the Friedewald equation, if the triglyceride level was <4 mmol L⁻¹ (352 mg dL⁻¹), instead of the <4.5 mmol L⁻¹ (400 mg dL⁻¹) used in the original Friedewald paper [13]. At triglyceride levels ≥4 mmol L⁻¹, LDL cholesterol was measured directly. High-sensitivity C-reactive protein (CRP) was measured by nephelometry (Dade Behring, Deerfield, IL, USA) or turbidimetry (Dako, Glostrup, Denmark).

Plasma bilirubin was measured daily as total bilirubin using the same standard Konelab assay according to the manufacturer's instructions (Thermo Fisher Scientific Inc., Waltham, MA, USA), at the Department of Clinical Biochemistry, Herlev University Hospital. The day-to-day coefficient of variation (CV) was <3.5% for bilirubin; CVs were <4% for all other measurements (except for CRP which was <6%).

Finally, all participants were asked the time of their last meal at the time of blood sampling. The times of last food intake were categorized as 0–1, 1–2, 2–3, 3–4, 4–5, 5–6, 6–7, 7–8 or >8 h ago.

Covariates

Body mass index (BMI), physical activity, hypertension, diabetes mellitus, smoking status, alcohol consumption and use of lipid-lowering drugs were included as covariates in the present study. Further details are given in the Supporting Information.

Genotyping

An ABI PRISM 7900HT Sequence Detection System (Applied Biosystems Inc., Foster City, CA, USA) and a TaqMan-based assay were used for genotyping the single-nucleotide polymorphism (SNP) rs6742078. This SNP is located in a noncoding region approximately 3000 base pairs upstream of UGT1A1 and has previously been shown in a genome-wide association study to be associated strongly with elevated plasma bilirubin levels [9].

Statistical analysis

We used STATA (StataCorp, College Station, TX, USA) statistical software for analysis. Risk of IHD and MI was examined prospectively in the CGPS and CCHS using left truncation (or delayed entry). Cox regression models were used to estimate hazard ratios (HRs) with age as the time scale, adjusted for sex and age, or multifactorially for age, sex, BMI, hypertension, diabetes, smoking, use of lipid-lowering drugs, plasma levels of HDL cholesterol, LDL cholesterol, triglycerides and CRP, or all of the above including time since the last meal. In the case–control CIHDS, and in all three studies combined (CGPS + CCHS + CIHDS), odds ratios (ORs) for IHD and MI (adjustment for age and sex alone or multifactorial adjustment) were calculated by logistic regression analysis. For test for trend, tertiles of bilirubin or genotypes were coded (1, 2 and 3).

In the meta-analysis, we used the metan command to estimate fixed and random effects ORs. We examined risk of any ischaemic end-point in a recessive model for high versus low plasma bilirubin, using UGT1A1*28 7/7 genotype versus 6/7 + 6/6 in the seven studies that genotyped this repeat polymorphism [6, 14-19]; rs887829 AA versus AG + GG in the one study that genotyped this promoter SNP [20], and rs6742078 TT versus GT + GG in the CGPS, CCHS and CIHDS. Carriership of UGT1A1*28 7/7 versus 6/6, rs887829 AA versus GG or rs6742078 TT versus GG was associated with similar increases (on average 100%) in plasma bilirubin levels in the included studies (see Kronenberg [5] for an overview). Between-study heterogeneity was assessed by Cochran's Q test and the I² statistic and was interpreted as low (I² ≤ 25%), moderate (I² > 25–<50%) or high (I² ≥ 50%) [21]. The random effects model was selected in the case of moderate or high heterogeneity between studies [21]. Further description of the statistical analyses is provided in the Supporting Information.

Results

Clinical characteristics

Characteristics of the participants in the three studies from Copenhagen, Denmark, are shown in Table 1. Genotyping UGT1A1 rs6742078 in the CGPS identified 20 548 (47%) GG homozygotes, 18 791 (43%) GT heterozygotes and 4369 (10%) TT homozygotes. The corresponding numbers in the CCHS were 4630 (46%) GG homozygotes, 4512 (44%) GT heterozygotes and 1039 (10%) TT homozygotes. Frequencies of UGT1A1 rs6742078 did not deviate from those predicted by the Hardy–Weinberg equilibrium (P = 0.45 and 0.21 in the CGPS and CCHS, respectively). Cardiovascular risk factors did not differ according to UGT1A1 rs6742078 genotype (Table S1).

Table 1. Characteristics of individuals in the three studies (all white and of Danish descent)

	Copenhagen General Population Study		Copenhagen City Heart Study		Copenhagen Ischemic Heart Disease Study
	No event	IHD	No event	IHD	Controls	IHD
IHD, ischaemic heart disease. Data are from study enrolment in 2003–2009 for the Copenhagen General Population Study, from the 1991–1994 or 2001–2003 examinations of the Copenhagen City Heart Study, and from study enrolment in 1991–2009 in the Copenhagen Ischemic Heart Disease Study. Values are median and (interquartile range) or number of participants and (percentage). Numbers of participants vary slightly according to availability of data. P-values were calculated using Mann–Whitney rank sum test for continuous variables and Pearson's chi-square test for categorical variables. ^aP < 0.001; ^bP < 0.05. ^cData on body mass index, smoking status, diabetes, hypertension, lipid levels and lipid-lowering therapy were only available in a fraction of IHD cases (n = 846–4274) in the Copenhagen Ischemic Heart Disease Study.
Total, n (%)	42 125 (91)	4413 (9)	8124 (79)	2140 (21)	5133 (50)	5133 (50)
Women, n (%)	25 313 (60)	1771 (40)^a	4707 (58)	1034 (48)^a	1507 (29)	1507 (29)
Age (years)	54 (45–64)	67 (59–75)^a	55 (41–67)	68 (60–74)^a	63 (56–71)	64 (56–71)
Body mass index (kg m^−2)	25 (23–28)	27 (24–30)^a	25 (22–27)	26 (24–29)^a	26 (24–29)	26 (24–28)^c
Current smokers, n (%)	9130 (22)	982 (22)	3772 (46)	1051 (49)^b	1103 (22)	958 (22)^c
Diabetes mellitus, n (%)	1246 (3)	492 (11)^a	176 (2)	139 (6)^a	259 (5)	519 (12)^a,^c
Physical activity, n (%)	20 088 (48)	1756 (40)^a	3076 (38)	622 (29)^a	2611 (51)	N.A.
Alcohol consumption, n (%)	17 592 (42)	2239 (51)^a	4675 (58)	1179 (55)^b	2909 (57)	N.A.
Hypertension, n (%)	22 397 (53)	3356 (76)^a	3751 (46)	1519 (71)^a	3385 (66)	1452 (71)^b,^c
Total cholesterol (mmol L^−1)	5.6 (4.9–6.3)	5.3 (4.5–6.1)^a	5.8 (5.0–6.7)	6.4 (5.6–7.3)^a	5.7 (5.0–6.4)	4.9 (4.1–5.9)^a,^a
LDL cholesterol (mmol L^−1)	3.2 (2.6–3.8)	2.9 (2.2–3.6)^a	3.5 (2.8–4.3)	4.0 (3.2–4.8)^a	3.3 (2.7–3.9)	2.8 (2.2–3.9)^a^.^c
HDL cholesterol (mmol L^−1)	1.6 (1.3–2.0)	1.5 (1.2–1.8)^a	1.5 (1.2–1.9)	1.4 (1.1–1.7)^a	1.2 (0.9–1.5)	1.2 (1.0–1.5)^a,^c
Triglycerides (mmol L^−1)	1.4 (1.1–2.1)	1.6 (1.1–2.4)^a	1.4 (1.0–2.1)	1.8 (1.3–2.5)^a	1.6 (1.1–2.3)	1.6 (1.1–2.3)^c
Lipid-lowering therapy, n (%)	2768 (7)	1712 (39)^a	41 (0.5)	64 (3)^a	439 (9)	1890 (57)^a,^c

Plasma bilirubin levels, IHD and MI

Mean bilirubin levels were increased by 44% (3.3 μmol L⁻¹) in the approximate second tertile and by 134% (10.1 μmol L⁻¹) in the approximate third tertile versus the approximate first tertile in the CGPS (Fig. 1, top). After adjustment for sex and age, the HRs for IHD in the CGPS were 0.92 [95% confidence interval (CI), 0.82–1.04] and 0.86 (95% CI, 0.76–0.98) for individuals in the approximate second and third tertile versus the approximate first tertile of plasma bilirubin, respectively (P for trend = 0.02; Fig. 2, top). The corresponding HRs for MI were 0.83 (95% CI, 0.68–1.02) and 0.81 (95% CI, 0.66–0.99; P = 0.04; Fig. 2, bottom). After multifactorial adjustment for sex, age, BMI, physical activity, hypertension, diabetes mellitus, smoking status, alcohol consumption, use of lipid-lowering drugs and levels of HDL cholesterol, LDL cholesterol, triglycerides and CRP, the HRs for IHD were 0.96 (95% CI, 0.84–1.09) and 0.93 (95% CI, 0.82–1.06) for individuals in the approximate second and third tertile versus the approximate first tertile of plasma bilirubin, respectively (P = 0.29; Fig. 2, top). The corresponding multifactorially adjusted HRs for MI were 0.91 (95% CI, 0.74–1.13) and 0.90 (95% CI, 0.73–1.12; P = 0.35; Fig. 2, bottom).

Figure 1. Mean bilirubin levels as a function of approximate bilirubin tertile (top) or UGT1A1 rs6742078 genotype (bottom) in the Copenhagen General Population Study. P-values are test for trend.

Figure 2. Risk of ischaemic heart disease and myocardial infarction by approximate tertiles of plasma bilirubin in the Copenhagen General Population Study. Multifactorial adjustment for age, sex, body mass index, hypertension, diabetes mellitus, physical activity, smoking, alcohol intake, use of lipid-lowering therapy, LDL cholesterol, HDL cholesterol, triglycerides and C-reactive protein. P-values are test for trend of hazard ratios.

To test the robustness of these observations, we performed a range of post hoc sensitivity analyses. Stratification of bilirubin levels into approximate quartiles, quintiles or octiles (Fig. S1), or exclusion of participants with extreme levels (top 1%, 5% or 10%) of bilirubin or GGT (i.e. suspected prevalent liver disease) or events occurring within 1 or 2 years from study entry (to address potential reverse causation), did not provide evidence of an association between elevated bilirubin and reduced risk of IHD or MI in the CGPS after multifactorial adjustment (Table S2).

Because prolonged fasting is a determinant of plasma bilirubin levels, we determined bilirubin levels as a function of time since the last meal in the CGPS (Fig. S2, left). Bilirubin levels adjusted for age and sex were increased by 6% (0.7 μmol L⁻¹) in individuals in whom blood was sampled more than 8 h after the last meal versus 0–1 h (P = 0.02). However, after multifactorial adjustment including time since last meal, HRs for IHD for individuals in the approximate second and third tertile versus the approximate first tertile of plasma bilirubin were 0.96 (95% CI, 0.84–1.09) and 0.93 (95% CI, 0.82–1.06), respectively, (P for trend = 0.27; data not shown). The corresponding HRs for MI were 0.92 (95% CI, 0.74–1.14) and 0.92 (95% CI, 0.74–1.14) (P for trend = 0.42; data not shown). Furthermore, restricting the analysis to individuals in the CGPS in whom blood was sampled more than 4 h after the last meal (n = 8222) as in previous studies [6] resulted in corresponding multifactorially adjusted HRs for IHD of 0.85 (95% CI, 0.65–1.13) and 0.84 (95% CI, 0.64–1.11) (P for trend = 0.21), and for MI of 1.30 (95% CI, 0.83–2.02) and 0.90 (95% CI, 0.56–1.45) (P for trend = 0.69).

UGT1A1 rs6742078 and plasma bilirubin levels

In accordance with previous findings [9], UGT1A1 rs6742078 was associated strongly with elevated plasma bilirubin levels in the CGPS (Fig. 1, bottom). UGT1A1 rs6742078 was associated with increases in mean plasma bilirubin of 16% (1.6 μmol L⁻¹) in GT heterozygotes and 95% (9.6 μmol L⁻¹) in TT homozygotes, compared with GG homozygotes (Fig. 1; P < 0.001). UGT1A1 rs6742078 genotype accounted for 18% of all variation in plasma bilirubin in the CGPS.

In individuals in whom blood was sampled more than 8 h after the last meal compared with 0–1 h, plasma bilirubin levels were increased by 7% (0.7 μmol L⁻¹) for GG homozygotes (P = 0.01), 7% (0.8 μmol L⁻¹) for GT heterozygotes (P = 0.05) and 4% (0.8 μmol L⁻¹) for TT homozygotes (P = 0.66) (Fig. S2, right).

Genetically elevated bilirubin levels and risk of IHD and MI

UGT1A1 rs6742078 genotype was not associated with risk of IHD or MI in any of the three studies, or in the studies combined (Fig. 3; P for trend 0.51–0.94). In the combined studies, TT versus GG genotype was associated with multifactorially adjusted ORs of 1.03 (95% CI, 0.96–1.11) for IHD and 1.01 (95% CI, 0.92–1.12) for MI. In addition, adjusting for time since last meal in the CGPS, or restricting the analyses to individuals in whom measurements were made more than 4 h after the last meal, did not change the lack of association between UGT1A1 genotype and risk of IHD or MI (P  for trend 0.11–0.94; data not shown).

Figure 3. Risk of ischaemic heart disease (IHD) and myocardial infarction (MI) as a function of UGT1A1 rs6742078 genotype. CGPS, Copenhagen General Population Study; CCHS, Copenhagen City Heart Study; CIHDS, Copenhagen Ischemic Heart Disease Study. *Owing to inclusion of participants with both incident and prevalent IHD and MI in the CGPS and the CCHS in the combined analysis, the number of participants is higher in the combined analysis than the sum of participants in the individual studies, in which only incident events were included. P-values are test for trend of hazard ratios. CI, confidence interval.

There were no interactions between genotype and any of the covariates presented in Table 1 on risk of IHD or MI (data not shown). The associations remained nonsignificant after stratification by age at study entry, sex, smoking status or use of lipid-lowering therapy in both the CGPS (Table S3) and CCHS (data not shown).

Meta-analysis

In a meta-analysis of 11 studies [6, 14-20] of risk of ischaemic cardiovascular disease (14 711 cases and 60 324 controls), including the present studies (CGPS, CCHS and CIHDS; 11 686 cases and 55 382 controls), homozygous carriers of bilirubin-increasing variants in UGT1A1 had a random effects OR for any ischaemic event of 1.01 (95% CI, 0.88–1.16; P = 0.86) versus noncarriers and heterozygotes combined (Fig. 4). The corresponding OR using a fixed effects model was 1.01 (95% CI, 0.95–1.08; P = 0.67). There was a high between-study heterogeneity (I² = 54%, P for heterogeneity 0.02), suggesting that the random effects model was most appropriate.

Figure 4. Meta-analysis of studies of genetically elevated bilirubin and risk of ischaemic cardiovascular disease. OR, odds ratio; ICVD, ischaemic cardiovascular disease; CAD, coronary artery disease; MI, myocardial infarction; PAD, peripheral arterial disease; CVD, cardiovascular disease; IHD, ischaemic heart disease; CI, confidence interval. *UGT1A1*28 7/7 genotype versus 6/7 + 6/6, UGT1A1 rs887829 AA versus AG + GG and UGT1A1 rs6742078 TT versus GT + GG in the Copenhagen General Population Study (CGPS), Copenhagen City Heart Study (CCHS) and Copenhagen Ischemic Heart Disease Study (CIHDS) were all associated with mean increases in plasma bilirubin of approximately 100%.

Discussion

The principal finding of this study is that a genetically elevated bilirubin level is not associated with decreased risk of IHD or MI, suggesting that raised plasma bilirubin is not causally associated with a decreased risk of ischaemic cardiovascular disease. With a total of 67 068 individuals, including 11 686 with IHD and 5749 with MI, our study is the largest to date to examine the association between genetically elevated bilirubin levels and risk of ischaemic cardiovascular disease. Results from previous studies, including from 61 to 776 cases, have been conflicting. Eight studies reported no significant association between genetically elevated bilirubin and risk of MI [6, 14, 16], peripheral arterial disease [19], ischaemic stroke [22] or coronary artery disease [18, 23], or severity of coronary artery disease [24], whereas a protective effect of genetically elevated bilirubin level was found in three studies on risk of coronary artery disease in Han Chinese [20], cardiovascular disease in the Framingham Offspring Study [17] and cardiovascular events and mortality in chronic haemodialysis patients [15]. It has been suggested that the inconsistent results could be owing to differences in study designs and/or characteristics of participants [5]. We did not observe an association between genetically elevated bilirubin level and risk of IHD or MI in three large, independent studies using three different designs (two prospective studies and a case–control study, as well as a combined study of all three cohorts), or in a meta-analysis of 11 studies. Moreover, stratification according to age at study entry did not reveal a protective effect in young individuals alone, a possibility that has been previously suggested [5, 17]. Further evidence for the lack of association with ischaemic cardiovascular disease reported in the present study is the fact that UGT1A1 rs6742078 (or other genetic variants near UGT1A1) has not been implicated in any of the many genome-wide association studies of ischaemic cardiovascular end-points, despite being represented on commercially available genotyping arrays [25].

A number of previous mainly cross-sectional and retrospective observational studies have demonstrated an inverse association between elevated plasma bilirubin and decreased risk of cardiovascular disease [3, 4]. However, the majority of prospective studies have shown various U-shaped associations (i.e. low or high bilirubin levels versus intermediate levels associated with increased risk of cardiovascular disease) [26-28], or no association after multifactorial adjustment [6, 29]. We found no association between baseline plasma bilirubin level and incident IHD or MI in the prospective CGPS after extensive multifactorial adjustment for potential confounders. By contrast, we observed a stepwise decreased risk of IHD and MI in individuals with elevated bilirubin when adjusting for age and sex alone. These discrepancies between results from retrospective and prospective studies, and between adjusted and unadjusted models are likely to reflect the effects of reverse causation (heart failure [30] and/or medication [31, 32] may influence bilirubin levels), or the influence of strong confounders, including smoking [33], obesity [34] and hypertension [35].

According to the antioxidant hypothesis, biological antioxidants protect cells and tissues against oxidative damage, and thus oral intake of antioxidants might play a role in the prevention of a range of human diseases [1]. Bilirubin is a potent antioxidant both in vitro and in vivo [2, 36]. We found that lifelong genetically, and hence unconfounded by socio-economic and/or environmental factors, elevated antioxidative bilirubin did not confer any protection against the development of IHD or MI in the general population, or in case–control studies. The data presented here therefore seem to suggest that the plasma level of bilirubin is not causally associated with the risk of ischaemic cardiovascular disease. However, to extend our findings, additional Mendelian randomization studies of genetically elevated bilirubin levels are warranted. Noncardiovascular end-points of interest include various forms of cancer and chronic obstructive pulmonary disease, conditions in which elevated plasma bilirubin may have a protective role [32, 37]. In addition, a Mendelian randomization study design could be utilized for other endogenous antioxidants. For instance, an association between genetic variation in SLC23A1 and plasma vitamin C levels was recently demonstrated [38], providing an ideal means for testing the effects of lifelong elevated antioxidative vitamin C levels on risk of human disease.

Some potential limitations to our study should be considered. UGT1A1 rs6742078 is strongly linked to the UGT1A1 promoter TA polymorphism underlying Gilbert's syndrome, characterized by approximately 70% reduction in UGT1A1 transcription in homozygotes compared with noncarriers, and hence lifelong moderately elevated plasma levels of bilirubin [8, 9]. In addition to bilirubin, UGT1A1 conjugates other endogenous and exogenous substances [32]. Reduced UGT1A1 activity (in carriers of rs6742078) might therefore have biologically relevant pleiotropic effects, including reduced conjugation of toxic substances, hormones or drugs, which in theory could obscure or counteract an atheroprotective effect mediated by elevated bilirubin levels. However, our findings were similar regardless of sex, smoking status or use of lipid-lowering therapy, suggesting that pleiotropy did not play a major role. Another potential limitation is that lifelong genetically elevated bilirubin might cause a downregulation of other endogenous antioxidants not measured in the current study, in effect neutralizing the putatively protective effects of raised bilirubin levels. However, Bulmer et al. [39] observed an increased resistance to serum oxidation in individuals with hyperbilirubinaemia owing to genetic variation in UGT1A1. In addition, erythrocyte levels of other antioxidants (superoxide dismutase, glutathione peroxidase and catalase) were not associated with UGT1A1 genotype, suggesting that genetically elevated bilirubin does not lead to a compensatory downregulation of other antioxidants [39].

Fasting is known to influence plasma levels of bilirubin, and this is most evident in individuals with Gilbert's syndrome [40]. However, adjusting risk estimates for time since the last meal, or restricting the analyses to individuals in whom blood was sampled more than 4 h after the last meal, did not affect the lack of association between tertiles of plasma bilirubin or UGT1A1 genotype and risk of IHD or MI.

In conclusion, elevated levels of bilirubin did not predict decreased risk of IHD or MI after multifactorial adjustment, and genetic variation leading to lifelong elevated plasma bilirubin levels was not associated with reduced risk of IHD or MI in the general population. These data suggest that plasma bilirubin is not causally associated with the risk of IHD.

Conflict of interest statement

No conflict of interest to declare.

Funding

This work was supported by the Danish Medical Research Council (10-083788), COST Action (BM0904), the Research Fund at Rigshospitalet, Copenhagen University Hospital, Chief Physician Johan Boserup and Lise Boserup's Fund, Ingeborg and Leo Dannin's Grant, Henry Hansen's and Wife's Grant, and a grant from the Order of Odd Fellows.

Acknowledgements

We are indebted to the staff and participants of the Copenhagen General Population Study, Copenhagen City Heart Study and the Copenhagen Ischemic Heart Disease Study for their important contributions.