The Nonlinear Relationship Between Total Bilirubin and Coronary Heart Disease: A Dose-Response Meta-Analysis

Background: Evidence suggests that the total bilirubin has a protective effect on coronary heart disease (CHD), but the dose-response relationship remains controversial, and there is no meta-analysis to assess the relationship. Methods: As of October 1, 2021, relevant literature was selected from four databases (PubMed, Web of Science, Cochrane Library, and Embase) by using a retrieval strategy. The dose-response curve between the total bilirubin and CHD was fitted by a restricted cubic spline. Stata 12.0 was used for statistical analysis. Results: A total of 170,209 (6,342 cases) participants from 7 prospective studies were analyzed in our meta-analysis. We calculated the pooled relative risks (RRs) and 95% CIs for the association between serum bilirubin level and risk of CHD using random-effects models. Compared with the first quantile, the bilirubin level in the third quantile had a protective effect on the risk of CHD (RR, 0.90; 95% CI, 0.82–0.99). The restricted cubic spline functions depicted a U-type curve relationship between bilirubin (3.42–49 μmol/L) and CHD (P linear < 0.001). When the bilirubin level was in the range of 3.42–13μmol/L, the protective effect of bilirubin on CHD was enhanced with increasing bilirubin levels. When the bilirubin level exceeded 13μmol/L, the protective effect of bilirubin weakened, and a dangerous effect gradually appeared with further increases in bilirubin levels. Conclusions: Compared with a low bilirubin level, a high bilirubin level has a protective effect on the risk of CHD, and there was a U-shaped dose-response relationship between them.


INTRODUCTION
Coronary heart disease (CHD) accounts for one-third to one-half of cardiovascular diseases and is the most common heart disease (1). According to WHO estimates, the number of deaths from cardiovascular disease has increased by more than 2 million since 2000 to nearly 9 million in 2019 (2). The increase in cardiovascular disease deaths was due to an increase in the prevalence of CHD (3). Bilirubin is formed from biliverdin through heme catabolism and biliverdin reductase activity (4). Traditionally, bilirubin is considered toxic and can cause irreversible damage to the nervous system (5). In the clinic, bilirubin concentration is used as a marker for the diagnosis of some liver diseases (6). However, there is evidence that bilirubin has antioxidant activity, and its antioxidant effect is stronger than that of bilirubin αtocopherol (7). It plays an important role in the occurrence and development of oxidative stress diseases. Recent studies have shown that bilirubin has a certain clinical value in the diagnosis and prognosis of stroke (8), diabetes (9), peripheral arterial disease (10), Parkinson's disease (11), and heart failure (12), so it may be used as a new marker for cardiovascular and cerebrovascular diseases in clinical practice.
At present, the relationship between bilirubin and the risk of CHD has attracted increasing attention from scholars. However, the conclusions of these studies are not consistent. Some studies reported that bilirubin was not associated with the occurrence of CHD (13)(14)(15); some studies believe that there is a negative linear relationship between them, and with the increase of bilirubin level, the risk of CHD decreases accordingly (16); interestingly, other studies believe that there is a nonlinear relationship between them, such as L-type (17) and U-type (18)(19)(20). Stender et al. (21) conducted a meta-analysis of the relationship between hereditary elevated bilirubin levels and the risk of CHD and found no evidence of the association between bilirubin and CHD. Subsequently, a prospective metaanalysis of the relationship between bilirubin and cardiovascular disease reported consistent results (22). Their study only considered whether bilirubin was related to CHD but did not evaluate the relationship between bilirubin exposure level and CHD.
Considering that bilirubin may have certain potential value in the occurrence and development of CHD and no consistent conclusion has been reached on the relationship between the two, it is necessary to explore bilirubin and CHD. Based on published epidemiological studies, we calculated the relative risk of CHD by dose-response meta-analysis, and a dose-response curve was drawn by restrictive cubic spline function to obtain more accurate and intuitive evidence of the dose-response relationship between them.

Search Strategy
An extensive search strategy was carried out in four databases (PubMed, Web of Science, Cochrane Library, and Embase) for studies published before October 1, 2021. The following search keywords were used: "bilirubin and CHD, " "bilirubin and cardiovascular diseases, " and "bilirubin and ischemia heart disease." Relevant MeSH terms were searched in combination to identify relevant studies. Refer to Table 1 for the detailed method used to search PubMed. The strategies for the other databases were similar but adapted where necessary. We also searched references to reviews with related or similar themes for additional research (21,22,(25)(26)(27).

Inclusion and Exclusion Criteria
According to the inclusion criteria, the possibly relevant articles that had been collected were independently assessed by two authors (CXL and YMS). The third author (XMW) resolved any disagreements by discussion or consultation. Articles were included if they fulfilled all of the following criteria: (1) the article reported the risk estimate and 95% CI of the relationship between bilirubin and CHD; (2) the design of the study was prospective; (3) Provided risk estimates at different bilirubin exposure levels, or provided corresponding data to calculate them; (4) Provided sample sizes at different bilirubin exposure levels, or provided corresponding data to calculate them; (5) the sample source of the article was human; (6) if multiple articles were published from the same/similar population, we included data from the study with the largest sample size and/or the latest or highestquality result. Reviews, case reports, lab experiments, duplicate publications, and articles without full text or available data were excluded.

Risk of Bias and Methodological Quality
The risk of bias and methodological quality of the included studies were independently assessed by two reviewers (CXL and WYW) using the Newcastle Ottawa scale (NOS), with an asterisk assigned to each article. A study can get up to one star for each numbered item in the selection and outcome categories and up to two stars for comparability. One asterisk equals one score, and the quality of the studies was graded as poor (fewer than four scores), fair (four to six scores), and good (seven or more scores) (28,29).

Data Extraction
Data extraction was managed in Microsoft Excel. The following information was independently extracted from the included articles by two authors (CXL and WYW): author, year of publication, country, study design, sample source, follow-up period, sample size (case/total), effect value, adjusted risk factors, baseline characteristics [age, body mass index (BMI), smokers, alcohol], diseases (hypertension, diabetes, hyperlipidemia, and liver disease), drug, and biochemical indicators of the subjects. For studies that presented several estimates adjusted for different numbers of potential confounders, the estimate that adjusted for the maximum number of potential confounders was selected for analysis.
In addition, we also extracted the following information for the dose-response meta-analysis: (1) the risk estimates and their corresponding 95% CIs in each group; (2) the median or mean of total bilirubin concentration in each group; and (3) the number of cases and total number in each group.

Statistical Analysis
In our meta-analysis, pooled risk ratios (RRs) and 95% CIs were used to analyze the association between serum bilirubin and the risk of CHD. The result of the pooled estimate was generated using the adjusted odds ratio (OR), relative risk (RR), hazard ratio (HR), and a baseline measure of total bilirubin exposure. In the original research, the effect value (OR/RR/HR) was reported with different distributions of bilirubin levels (such as per tertiles, quartiles, quintiles in bilirubin levels). To enable a consistent approach to the meta-analysis and enhance the interpretation of findings, ORs were converted into RRs by the formula, and HRs were regarded as approximate to RRs (30,31). At the same time, as reported in previous studies, the results were transformed into RR between the upper tertiles and lower tertiles (32). The chi-square-based Q-test was used to assess the heterogeneity among the individual studies (33). Heterogeneity was quantified based on I 2 , which ranged from 0 to 100% (I 2 = 0-25%, no heterogeneity; I 2 = 25-50%, low heterogeneity; I 2 = 50-75%, moderate heterogeneity; I 2 = 75-100%, high heterogeneity). Considering the underlying clinical and methodological heterogeneity (e.g., baseline characteristics of the patients, adjustment for confounders, and follow-up duration), a random-effects model was used as the primary approach to combine results across studies (34). The possible sources of heterogeneity were explored by subgroup analysis. Egger tests were used to evaluate the possibility of publication bias (35). The sensitivity analysis was performed by excluding one study at a time to evaluate whether the results could have been affected by a single study (36). A two-sided value of P < 0.05 in the statistical process was considered statistically significant (37).
We used the two-stage generalized least squares method to estimate the relative risk of the original research report and established a restrictive cubic spline curve with five nodes (5,25,50,75, and 95) to fit the dose-response relationship between bilirubin and CHD (38)(39)(40). To ensure unit consistency in all

Literature Search
A total of 37,826 potential studies were retrieved. A total of 20,931 articles were excluded according to automation tools. The remaining literature was imported into EndNote X9 (Clarivate Analytics) for further screening. According to the analysis of titles and abstracts, 16,862 articles were excluded and 33 potential articles were included. In addition, 67 articles were included from the reference list of previous reviews. After reading the full text and detailed screening, 93 articles were excluded for various reasons. Finally, 7 prospective studies (16,18,20,21,(41)(42)(43) were included in the meta-analysis. The PRISMA flowchart of literature retrieval and selection is presented in Figure 1. Table 2 shows the details of the included studies in our metaanalysis. All the effect values in the included studies were adjusted for the maximum number of potential confounders. All of the studies were based on community populations. The publication year was from 1996 to 2018. The follow-up period ranged from 5 to 21.9 years. The included studies involved three regions, including Europe (20,21,42), Asia (16,18) and America (41,43). Table 3 shows the baseline characteristics and biochemical indicators of the subjects in each study. The total sample size is 170,209 (age: 27.5-70 years). The percentage of men ranged from 45 to 100%. The percentages of smoking and alcohol were 67 and 70%, respectively. The percentages of hypertension patients, diabetic patients, hyperlipidemia patients, and liver disease were 60, 70, 80, and 22.95%, respectively. The biochemical indicators included AST, ALT, TC, TG, LDL, and HDL. The biochemical indicators of the subjects in all the included studies are shown in detail in Table 3. Table 4 shows the quality scores of the included studies based on the NOS assessment tool. All 7 articles were assessed as good.

Correlation Between Total Bilirubin and CHD
We extracted 9 sets of statistical data from 7 articles, and 1 study reported the relationship between bilirubin and CHD by sex (43); in another study, the relationship between bilirubin and CHD was reported by two different populations, which were divided according to the time of bilirubin measurement (before and after 16 o'clock) (20). There was low heterogeneity in 7 studies (I 2 = 49.2%, P = 0.046), and a random effect model was used for analysis. Compared with the first quantile, the bilirubin level in  the third quantile had a protective effect on the risk of CHD (RR, 0.90; 95% CI, 0.82-0.99; P = 0.027; Figure 2. Subgroup analysis was performed according to sample size (subject <15,000, ≥15,000), which was consistent with the results of the holistic analysis ( Table 5). Egger's test (P = 0.202) suggested no publication bias (Figure 3). For sensitivity analysis, the pooled RRs ranged from 0.89 (95% CI:0.81-1) to 0.92 (95% CI:0.83-1.02) ( Table 6).

Dose-Response Analysis for Bilirubin and CHD
We extracted 9 sets of data from 7 studies for dose-response meta-analysis (16,18,20,21,(41)(42)(43)   The dose-effect relationship between bilirubin (3.42-49 µmol/L) and CHD was nonlinear (plinear < 0.001) (Figure 4). When the bilirubin level was in the range of 3.42-13 µmol/L, the protective effect of bilirubin on CHD was enhanced. When  the bilirubin level exceeded 13 µmol/L, the protective effect of bilirubin weakened, and dangerous effects gradually appeared with increasing bilirubin levels.

DISCUSSION
This was the first study to assess the relationship between bilirubin and CHD based on a dose-response meta-analysis. Our results showed that compared with the first quantile, the bilirubin level in the third quantile has a protective effect on the risk of CHD; at the same time, dose-response meta-analysis showed that the protective effect was U-type. Previous studies have reported a nonlinear relationship between bilirubin and CHD, and our study is consistent with these findings. A cross-sectional study in Japan reported a nonlinear relationship between bilirubin and CHD (44). A prospective study of middle-aged people in Britain reported consistent results, and their results showed that the nonlinear relationship may be U-shaped (20). However, these conclusions are limited to male subjects. In another prospective study including Chinese elderly individuals (mean age was 62.7 years; 45.2% men), Lai et al. (18) reported a U-shaped relationship between them. Due to different research methods and subject populations, the conclusions of these studies may have some limitations, but they showed that the protective effect of bilirubin on CHD is not a simple linear relationship. The U-shaped relationship between bilirubin and disease means that when the bilirubin level is approximately 12-16 µmol/L (18,20,45,46), the protective effect of bilirubin on disease will no longer be enhanced. When the bilirubin level exceeds this range, the protective effect will weaken, and a dangerous effect will begin to appear. Other studies have reported an absence of any such association, and the inconsistency of this conclusion can be attributed to the bilirubin exposure level of subjects included in the study not being high enough to detect an effect of bilirubin on CHD (15)(16)(17). For example, the results of Kim et al. (15) and Suh et al. (16) showed that there was a negative linear relationship between bilirubin and CHD. In their study, the fourth quartile of bilirubin levels was only 9.5 and 10.77 µmol/L. Therefore, the linear relationship of negative correlation they found may only be part of the U-type relationship. A similar dose-response relationship was also found between bilirubin and other diseases, such as stroke and diabetic retinopathy (DR). Liu et al. (45) reported that the protective relationship between bilirubin and stroke is nonlinear, and the risk of first ischemic stroke was the lowest when the bilirubin level was 17.0 µmol/L. A nonlinear correlation relationship was also reported between bilirubin and diabetic retinopathy (DR). The protective effect of bilirubin level on DR was the strongest in the range of 12-13.8µmol/L (46).
The change in bilirubin level plays an important role in the occurrence and development of CHD. On one hand, bilirubin can eliminate reactive oxygen species produced in the process of oxidative stress, inhibit the aggregation of inflammatory cells, improve the bioavailability of Nitric Oxide (NO), and play a protective role in the heart (47,48). However, an increase in bilirubin exceeding the physiological range may indicate potential liver injury, and ALP, ALT, and GGT released by liver injury are related to an increased risk of CHD (49,50). On the other hand, we should pay attention to the fact that bilirubin is also cytotoxic. It has been confirmed in animal experiments that when the accumulation of bilirubin in cells reaches a certain threshold, oxidation will change into cytotoxicity (51). Therefore, the U-type relationship between bilirubin and CHD may be a combination of antioxidant, hepatotoxicity, or cytotoxicity.
Our study has some advantages. Firstly, we established a restrictive cubic spline to fit the dose-response relationship between bilirubin and CHD by pooling original research and obtaining a more intuitive curved shape. Secondly, our doseresponse meta-analysis included 7 prospective studies, a large sample size (170,209), and a wide range of bilirubin levels (3.42-49µmol/L), providing sufficient statistical validity for the dose-response relationship model. Our study also has some limitations. Firstly, in a longitudinal analysis, biochemical indicators including bilirubin were collected from baseline measurement, which are sensitive to changes during the followup. It may fail to (at least partly) show a robust link between bilirubin and CHD. Secondly, bilirubin levels are disturbed by many endogenous and environmental factors, we had no access to individual participants' data and potential confounders could not be ruled out. Under confounding factors, there may be no reliable connection between bilirubin and CHD. Gajdos et al. (52) and Mahaba et al. (53) found a modest link of bilirubin to CHD. However, after adjusting for risk factors, they were no longer statistically significant. Similarly, in a retrospective cohort of patients with HIV, bilirubin was associated with ischemic events. After the addition of related treatment to the Cox model, bilirubin was no longer significantly associated with ischemic events (54). Their results suggested that the correlation between bilirubin and CHD may be related to uncontrolled confounding factors. Considering the influence of potential factors, we prioritize OR/RR/HR adjusted by most factors when collecting and extracting effect values. It is hoped that the combined value can present objective results after controlling the bias factors as much as possible. In future research to explore the relationship between bilirubin and CHD, bilirubin-related factors should also be considered when adjusting the risk factors for cardiovascular disease.
CONCLUSION Compared with a low bilirubin level, a high bilirubin level has a protective effect on the risk of CHD, and there was a U-shaped dose-response relationship between them.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material, further inquiries can be directed to the corresponding author.