Abstract
Background
Current evidence on associations between circulating bilirubin and colorectal cancer (CRC) risk is inconsistent.
Methods
In this prospective study, we investigated associations of pre-diagnostic circulating levels of total and indirect bilirubin with CRC risk in 78,467 Korean adults aged 40–78 years at recruitment, considering potential non-linearity and sex differences. Hazard ratios (HR) and 95% confidence intervals (CI) for associations with CRC risk were estimated with Cox proportional hazard regression.
Results
During a median 7.9-year follow-up, 539 incident CRC cases were recorded. In multivariable-adjusted models, higher levels of total bilirubin were associated with a 26% (CI: 42% to 7%) lower risk of CRC among men and women combined, comparing the highest with the lowest tertile (P-linear trend = 0.003). A U-shaped association was observed in men, with the lowest risk at approximately 0.8 mg/dL (=13.7 μmol/L) of total bilirubin (P for non-linearity = 0.01). Although the association was largely null in women, there was no evidence for effect modification by sex (P-interaction = 0.73). Associations between indirect bilirubin and CRC risk were similar.
Conclusions
Higher circulating levels of total and indirect bilirubin were inversely associated with the risk of CRC among Korean adults. The associations were strongly inverse and U-shaped among men.
Subject terms: Cancer epidemiology, Risk factors
Introduction
Colorectal cancer (CRC) was the third most diagnosed cancer (10% of all new cases) and the second leading cause of cancer death (9% of all cancer deaths) globally for both sexes in 2020 [1]. In South Korea, CRC was also the third most common cancer (11% of all new cases) and the third leading cause of cancer death (11% of all cancer deaths) in 2020 [2].
Chronic inflammation is one of the hallmark characteristics of cancer [3]. Inflammatory cells can release reactive oxygen species (ROS), which are mutagenic for nearby cells [3]. Inflammation has long been proposed to be an important risk factor for CRC, and CRC might thus be a candidate for prevention by anti-inflammatory and antioxidative agents [4]. A compelling body of evidence from experimental and clinical studies has demonstrated that bilirubin, the primary end-product of haem catabolism, exhibits substantial anti-inflammatory and antioxidative properties with potential cancer-preventive relevance [5–9].
Bilirubin concentrations are lower in women than in men because of the influence of sex hormones on bilirubin conjugation and circulating levels [10]. These physiological differences in bilirubin metabolism could translate into divergent associations with CRC risk. There is also a current debate to establish physiological ranges for bilirubin concentrations separately in men and women as well as by ethnicity [11].
Several epidemiological studies have examined associations between bilirubin and the risk of CRC [12–15]. However, the results were inconsistent and did not allow firm conclusions. Most studies were based on cross-sectional or retrospective case-control studies, considered only circulating levels of total bilirubin, or did not consider potential differences in associations among men and women [12–15]. In our previous work in European populations, we found that unconjugated bilirubin levels, which roughly correspond to indirect bilirubin, associated with the risk of CRC differed by sex, with positive associations among men and inverse associations among women [16, 17]. In the UK Biobank, we found an inverse non-linear association between total bilirubin levels and the risk of CRC, with similar associations among men and women [18]. A study among a Japanese population indicated an inverse non-linear association between total bilirubin levels and colon cancer risk with similar risk among men and women [19].
A genome-wide association study in Korea showed considerable ethnic differences in genetically predicted circulating bilirubin levels between Korean and European-derived populations [20]. This may translate into differences in CRC risk. However, associations between circulating bilirubin levels, especially indirect bilirubin levels, and CRC risk in the Korean population have not been well established. One prospective study among a Korean population showed inverse associations between total bilirubin and overall cancer risk and cancer-related death, and suggestive weak inverse associations with CRC risk (n = 238 CRC cases) without considering potential non-linearity [21].
This study aimed to investigate associations of pre-diagnostic total and indirect bilirubin levels with the risk of CRC taking into consideration potential non-linearity and sex differences in a large Korean population from the Korean Genome and Epidemiology Study – Health Examinee Study (KoGES-HEXA) Cohort Study.
Methods
Study design and population
We used data from the KoGES-HEXA study, a community-based prospective cohort study. The KoGES-HEXA study design has been described in detail elsewhere [22, 23]. Briefly, the KoGES was initiated in 2001 by the Korea National Institute of Health (KNIH) and the Korea Disease Control and Prevention Agency (KDCA) to establish a genome epidemiological cohort consortium consisting of six independent prospective cohort studies for the research community, providing a health database and biobank to investigate the genetic and environmental aetiology of common complex diseases (i.e. type 2 diabetes, hypertension, obesity, metabolic syndrome, osteoporosis, cardiovascular disease, and cancer) and causes of death in Koreans with long-term follow-up. The KoGES-HEXA study is the largest prospective cohort within the KoGES cohort consortium based on the national health examination program, which is provided by the National Health Insurance Service. Men and women aged ≥40 years between 2004 and 2013 were recruited at 39 hospitals in metropolitan areas and major cities.
Among all participants of the KoGES-HEXA study (n = 173,202), serum bilirubin data was available for the participants recruited after the year 2009 (n = 84,033). Of these, 226 participants had missing bilirubin data and 1359 participants did not consent to the linkage to the cancer registry and mortality data. After further exclusion of 3981 participants diagnosed with cancers before recruitment, a total of 78,467 participants remained for analysis (Fig. 1).
Fig. 1.
Selection of study population.
Bilirubin measurement
Bilirubin concentrations (total and direct bilirubin, mg/dL) were measured by colourimetry in serum derived from at least 12 h or overnight fasting venous blood samples collected at enrolment. The vanadate oxidase method (7600-210 Automatic Biochemical Analyzer, Hitachi Ltd, Japan) was used in 2009~2011 (n = 67,021, 85% of the total subjects) and the diazo method (Modular Analytics P800, Roche, Switzerland) in 2012~2013. Indirect bilirubin level was derived by subtracting direct bilirubin level from total bilirubin level. Indirect bilirubin is 80–85% of total bilirubin and this ratio is constant under a physiological state [16].
Outcomes
Incident first primary CRC cases were identified by data linkage with the Korea Central Cancer Registry and deaths by linkage to the Death Registry of Korea until 31 December 2018. CRCs were defined by the International Classification of Diseases (ICD-10) codes, C18-19 (colon) and C20 (rectum).
Covariates
Body mass index (BMI, kg/m2) was calculated from height (m) and weight (kg), which were measured by trained examiners [22]. Data on socio-demographic status, lifestyle, and medical treatment were collected by trained interviewers [22]. Dietary intake was collected using a semi-quantitative food frequency questionnaire (FFQ) developed and validated for the KoGES [22, 24, 25]. There were missing data in the covariates ranging from n = 41 (0.1%) for height to n = 868 (1.1%) for dietary intake (Table 1). Available biomarkers relevant for CRC development or bilirubin levels, including circulating levels of inflammatory (high-sensitivity C-reactive protein, hs-CRP) and liver function markers (aspartate transferase, AST, and gamma-glutamyl transferase, GGT), measured using enzymatic calorimetric methods with automatic analyzers (ADVIA 1650 and ADVIA 1800, Siemens, Tarrytown, NY, USA) [26], were used for sensitivity analyses to assess their impact as potential mediators or confounders. The missing values in these biomarker data varied between 0 to 24% (Table 1).
Table 1.
General characteristics of the study population across tertiles of total bilirubin levels
| Total | Men | Women | |||||||
|---|---|---|---|---|---|---|---|---|---|
| T1 (n = 19,356) | T2 (n = 27,135) | T3 (n = 31,976) | T1 (n = 7,919) | T2 (n = 8,312) | T3 (n = 11,533) | T1 (n = 15,144) | T2 (n = 10,287) | T3 (n = 25,272) | |
| Bilirubin levela, mg/dL | |||||||||
| Total bilirubin | 0.4 ± 0.1 | 0.6 ± 0.0 | 1.0 ± 0.3 | 0.5 ± 0.1 | 0.7 ± 0.0 | 1.1 ± 0.3 | 0.4 ± 0.1 | 0.6 ± 0.0 | 0.9 ± 0.2 |
| Indirect bilirubin | 0.3 ± 0.1 | 0.4 ± 0.1 | 0.7 ± 0.2 | 0.3 ± 0.1 | 0.5 ± 0.1 | 0.8 ± 0.2 | 0.3 ± 0.1 | 0.4 ± 0.0 | 0.6 ± 0.2 |
| Age at recruitmenta, yr | 53.0 ± 8.3 | 53.1 ± 8.1 | 52.6 ± 8.2 | 53.6 ± 8.7 | 53.6 ± 8.6 | 53.1 ± 8.5 | 52.9 ± 8.1 | 52.9 ± 8.0 | 52.2 ± 7.9 |
| BMIa, kg/m2 (missing% <0.1%) | 24.0 ± 3.1 | 23.9 ± 2.9 | 23.8 ± 2.9 | 24.5 ± 2.9 | 24.5 ± 2.7 | 24.4 ± 2.7 | 23.9 ± 3.1 | 23.7 ± 3.0 | 23.4 ± 2.9 |
| Heighta, cm (missing% <0.1%) | 159.0 ± 7.4 | 160.1 ± 7.8 | 163.0 ± 8.3 | 168.6 ± 5.8 | 169 ± 5.7 | 169.4 ± 5.8 | 156.4 ± 5.3 | 156.4 ± 5.3 | 156.9 ± 5.3 |
| PAa, mn/wk (missing% =0.6%) | 145 ± 231 | 165 ± 240 | 184 ± 254 | 173 ± 270 | 197 ± 275 | 203 ± 268 | 140 ± 218 | 152 ± 216 | 163 ± 236 |
| Educationb | |||||||||
| ≤Middle school | 6487 (33.5) | 8425 (31.1) | 8081 (25.3) | 1794 (22.7) | 1573 (18.9) | 2051 (17.8) | 5499 (36.3) | 3696 (35.9) | 8380 (33.2) |
| High school | 8328 (43.0) | 11,305 (41.7) | 13,162 (41.1) | 3368 (42.5) | 3292 (39.6) | 4398 (38.1) | 6472 (42.7) | 4306 (41.9) | 10,959 (43.4) |
| ≥College | 4393 (22.7) | 7164 (26.3) | 10,460 (32.7) | 2687 (33.9) | 3367 (40.5) | 5000 (43.4) | 3063 (20.2) | 2205 (21.4) | 5695 (22.5) |
| Missing | 148 (0.8) | 241 (0.9) | 273 (0.9) | 70 (0.9) | 80 (1.0) | 84 (0.7) | 110 (0.7) | 80 (0.8) | 238 (0.9) |
| Smoking habitsb | |||||||||
| Never | 15,150 (78.3) | 20,154 (74.3) | 20,165 (63.1) | 1624 (20.5) | 2112 (25.4) | 3288 (28.5) | 14,335 (94.7) | 9795 (95.2) | 24,315 (96.2) |
| Former | 1606 (8.3) | 3457 (12.7) | 7143 (22.3) | 2829 (35.7) | 3435 (41.3) | 5235 (45.4) | 240 (1.6) | 143 (1.4) | 324 (1.3) |
| Current | 2535 (13.1) | 3383 (12.5) | 4531 (14.2) | 3439 (43.4) | 2723 (32.8) | 2981 (25.9) | 519 (3.4) | 296 (2.9) | 491 (1.9) |
| Missing | 65 (0.3) | 141 (0.5) | 137 (0.4) | 27 (0.3) | 42 (0.5) | 29 (0.3) | 50 (0.3) | 53 (0.5) | 142 (0.6) |
| Alcohol drinking habitsb | |||||||||
| Never | 11,004 (56.9) | 13,943 (51.4) | 12,963 (40.5) | 1635 (20.7) | 1577 (19) | 1979 (17.2) | 10,098 (66.7) | 6734 (65.5) | 15,887 (62.9) |
| Former | 644 (3.3) | 949 (3.5) | 1182 (3.7) | 593 (7.5) | 537 (6.5) | 660 (5.7) | 296 (2.0) | 217 (2.1) | 472 (1.9) |
| Current | 7630 (39.4) | 12,079 (44.5) | 17,658 (55.2) | 5658 (71.5) | 6153 (74.0) | 8850 (76.7) | 4688 (31.0) | 3270 (31.8) | 8748 (34.6) |
| Missing | 78 (0.4) | 164 (0.6) | 173 (0.5) | 33 (0.4) | 45 (0.5) | 44 (0.4) | 62 (0.4) | 66 (0.6) | 165 (0.7) |
| Use of hormone therapyb,c | |||||||||
| Non-user | 6619 (74.0) | 4746 (75.5) | 11,098 (75.1) | ||||||
| User | 2256 (25.2) | 1474 (23.4) | 3495 (23.7) | ||||||
| Missing | 73 (0.8) | 71 (1.1) | 175 (1.2) | ||||||
| Dietary intakea (missing% =1.1%) | |||||||||
| Energy intake, kcal/d | 1720 ± 580 | 1740 ± 560 | 1770 ± 560 | 1870 ± 580 | 1850 ± 540 | 1850 ± 540 | 1680 ± 570 | 1690 ± 560 | 1700 ± 550 |
| Red meat intake, g/d | 40 ± 50 | 40 ± 50 | 50 ± 50 | 60 ± 60 | 50 ± 60 | 50 ± 50 | 40 ± 50 | 40 ± 50 | 40 ± 50 |
| Processed meat intake, g/d | 1 ± 4 | 1 ± 3 | 1 ± 4 | 4 ± 4 | 1 ± 3 | 1 ± 3 | 1 ± 3 | 1 ± 3 | 1 ± 3 |
| Vegetable and fruits intake, g/d | 460 ± 310 | 480 ± 310 | 470 ± 310 | 440 ± 280 | 440 ± 270 | 440 ± 290 | 470 ± 310 | 480 ± 310 | 500 ± 320 |
| Dietary fibre intake, g/d | 6 ± 3 | 6 ± 3 | 6 ± 3 | 6 ± 3 | 6 ± 3 | 6 ± 3 | 6 ± 3 | 6 ± 3 | 6 ± 3 |
| Dairy intake, g/d | 120 ± 140 | 120 ± 140 | 110 ± 140 | 100 ± 120 | 90 ± 120 | 100 ± 130 | 120 ± 140 | 130 ± 150 | 130 ± 140 |
| hs-CRPa, mg/L (missing% =23.8%) | 0.2 ± 0.4 | 0.1 ± 0.4 | 0.1 ± 0.4 | 0.2 ± 0.5 | 0.2 ± 0.5 | 0.1 ± 0.3 | 0.2 ± 0.4 | 0.1 ± 0.3 | 0.1 ± 0.4 |
| ASTa, units/L (missing% =0%) | 22.6 ± 10.3 | 23.0 ± 29.1 | 24.0 ± 16.4 | 24.6 ± 12.8 | 24.9 ± 14.3 | 25.7 ± 15.7 | 22.0 ± 9.4 | 22.2 ± 33.3 | 22.3 ± 16.9 |
| GGTa, units/L (missing% =0.1%) | 28.4 ± 33.5 | 30.1 ± 40.4 | 35.6 ± 56.8 | 47.8 ± 53.1 | 47.5 ± 61.3 | 49.2 ± 75.0 | 23.0 ± 22.7 | 22.8 ± 23.6 | 22.7 ± 24.5 |
BMI Body Mass Index, PA Physical Activity, hs-CRP high-sensitivity C-Reactive Protein, AST Aspartate Transferase, GGT Gamma-Glutamyl Transferase.
aContinuous variables are presented as mean ± standard deviation (SD).
bCategorical variables are presented as n (%).
cOnly in post-menopausal women.
Statistical analysis
Prior to statistical analysis, missing values in covariates were imputed using multiple imputations with chained equations (MICE algorithm in R). The imputed datasets were used for further analysis. Cox proportional hazard models were used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) of total bilirubin and indirect bilirubin in tertiles associated with the risk of CRC in men and women separately and combined. Crude models were a priori stratified by study centre, age at recruitment (1-year categories), and sex (for sex combined analysis) [Model 1]. Fully adjusted models were further adjusted for BMI (kg/m2), height (cm), smoking status (never, former, current), education (≤middle school, high school, ≥college), alcohol drinking habits (never, former, current), physical activity (min/week), total energy intake (kcal/day), red meat intake (g/day), processed meat intake (g/day), dietary fibre intake (g/day), fruit and vegetable intake (g/day), and dairy food intake (g/day), and postmenopausal hormone use (non-user and user in postmenopausal women and non-applicable in men and premenopausal women), including an indicator variable accounting for the type of bilirubin measurements (2009~2011/2012~2013) [Model 2 as the main model]. These variables were identified as confounders guided by a directed acyclic graph (DAG) (Supplementary Fig. S1). For all models, collinearity and proportional hazard assumptions were assessed by time-varying covariate analyses and met. Tests for linear trend were performed using a score variable with values from 1 to 3 included in the model, consistent with the tertile grouping. The same analyses were repeated for anatomical subsites (colon and rectum) of CRC in men and women combined. In addition, to investigate potential non-linear dose-response associations between total/indirect bilirubin levels and CRC risk, we used cubic spline models with three knots at the sex-specific 10th (reference), 50th, and 90th percentiles of bilirubin, combined with a likelihood ratio test for non-linearity comparing the spline models with models that only included bilirubin as a linear term.
Biologically plausible effect modifiers were investigated including sex, BMI categories (normal, overweight, and obese), and smoking status (never, former, and current) by including interaction terms (using likelihood ratio tests) between these factors and bilirubin levels. Analyses were stratified for these factors and strata-specific risk estimates (per 1-SD increment in total bilirubin levels) were presented.
The following sensitivity analyses were performed: (i) re-coding diagnostic cases within one year of recruitment to non-cases to assess potential reverse causality bias; (ii) additional adjustment for available biomarkers relevant for CRC development or bilirubin levels including circulating levels of inflammatory (hs-CRP) and liver function markers (AST and GGT) to assess their impact as potential mediators or confounders; (iii) additionally re-censoring diagnostic cases within one year of recruitment to non-cases in the study population with available inflammatory and liver function markers; and (iv) restricting analyses to participants enrolled in 2009~2011, where bilirubin levels were measured with the same method.
All statistical analyses were performed using R (R version 4.3.3) and the SAS statistical program (Statistical Analysis Software version 9.4 SAS Institute Inc, Cary, North Carolina, USA). The relevant statistical code will be available upon request to the corresponding authors.
Results
General characteristics of the study population by total bilirubin levels
A total of 78,467 Korean adults (27,764 men and 50,703 women) were included in this study with a mean follow-up of 7.9 years (interquartile range: 7.3–9 years), during which 539 CRC (269 in men and 270 in women) cases occurred. The age range (mean ± SD) at recruitment of the study population was 40–78 (52.9 ± 8.2) years. The general characteristics of the study population across tertiles of total bilirubin levels are shown in Table 1. Age and BMI at recruitment were similar across tertiles in both men and women. The proportion of current smokers was lower in men with higher levels of total bilirubin (26% in tertile 3 vs. 43% in tertile 1). Both men and women with higher bilirubin levels were slightly more physically active. The mean levels of total and indirect bilirubin, respectively, were slightly higher in men (0.8 and 0.6 mg/dL) compared to those in women (0.7 and 0.5 mg/dL).
Total and indirect bilirubin and colorectal cancer
After multivariable adjustment, higher levels of total bilirubin were associated with a 26% (CI: 42% to 7%) lower risk of CRC among men and women combined, comparing the highest with the lowest tertile (Fig. 2a and Supplementary Table S1). Although a formal test did not indicate effect modification by sex (P-interaction = 0.73, Supplementary Table S2), inverse associations with CRC risk were observed in men, but not in women (Fig. 2a). The inverse association in men was U-shaped, with the lowest CRC risk at approximately 0.8 mg/dL (=13.6 μmol/L) of total bilirubin (P for non-linearity = 0.01), but this pattern was not observed in women (Fig. 3a). The findings for indirect bilirubin virtually mirrored those of total bilirubin (Figs. 2b and 3b and Supplementary Table S1). After stratifying CRC by its anatomical subsites, associations of both total and indirect bilirubin with cancers of the colon and rectum were very similar and effect sizes mirrored these of overall CRC (Supplementary Table S3). This analysis was only performed in men and women combined.
Fig. 2. Associations of bilirubin levels with the risk of colorectal cancer in Korean men and women.
a Total bilirubin. b Indirect bilirubin. Hazard ratios with 95% confidence intervals (CI) from Cox proportional hazard models stratified by study centre, age at recruitment (1-year units), and sex (only for total participants) and adjusted for bilirubin measurement (2009~2011 and 2012~2013), education (middle school or less, high school, and college or higher), smoking status (never, former, current), BMI (continuous), height (continuous), alcohol intake (non-drinkers, ex-drinkers, and current drinkers), physical activity (continuous), total energy intake (continuous), red meat intake (continuous), processed meat intake (continuous), fibre intake (continuous), fruit and vegetable intake (continuous), dairy food intake (continuous), and use of hormone therapy (non-user and user in postmenopausal women and not-applicable in men and premenopausal women).
Fig. 3. Associations of bilirubin levels with the risk of colorectal cancer in Korean men and women allowing for non-linearity.
a Total bilirubin. b Indirect bilirubin. Hazard ratios (solid line) with 95% confidence intervals (shade) from Cox proportional hazard models stratified by study centre, age at recruitment (1-year units), and sex (only for total participants) and adjusted for bilirubin measurement (2009~2011 and 2012~2013), education (middle school or less, high school, and college or higher), smoking status (never, former, current), BMI (continuous), height (continuous), alcohol intake (non-drinkers, ex-drinkers, and current drinkers), physical activity (continuous), total energy intake (continuous), red meat intake (continuous), processed meat intake (continuous), fibre intake (continuous), fruit and vegetable intake (continuous), dairy food intake (continuous), and use of hormone therapy (non-user and user in postmenopausal women and not-applicable in men and premenopausal women). Non-linear associations were modelled with restricted cubic splines using three knots at the sex-specific 10th (reference), 50th, and 90th percentiles of bilirubin. The X-axis ranged from the sex-specific 1st~95th percentiles of bilirubin. Non-linearity was examined by a likelihood ratio test comparing the spline models with models that only included bilirubin as a linear term. To convert bilirubin levels from mg/dL to µmol/L, multiply mg/dL by 17.1.
Effect modification by BMI and smoking status
There was little evidence for effect modification of associations between bilirubin levels and CRC risk by categories of BMI or smoking status (all P-interaction >0.46) (Supplementary Table S2).
Sensitivity analyses
After re-coding cases within one year of recruitment to non-cases to address potential reverse causation, effect sizes were similar to the main analyses, but less precise due to the smaller number of cancer cases (Supplementary Table S4). We also evaluated potential confounding by subclinical liver disease by additionally adjusting for markers of inflammation (hs-CRP) and liver function (AST and GGT). In this analysis, comparing tertiles 3 vs. 1 of total bilirubin levels (men and women combined), it was associated with a 27% (CI: 42% to 7%) lower risk of CRC (Supplementary Table S5). Since reverse causation bias could be more pronounced in this model, we also repeated this sensitivity analysis after additionally re-coding CRC cases within one year of follow-up to non-cases and observed associations that were very similar to our main analysis (Supplementary Table S6). We also confirmed that the different bilirubin measurement methods used in 2009~2011 and 2012~2013 did not affect the results as shown in our sensitivity analysis in participants with bilirubin data assessed only in 2009~2011 (Supplementary Table S7).
Discussion
In this prospective cohort study among Korean adults, pre-diagnostic serum levels of total and indirect bilirubin were both inversely associated with CRC risk. In sex-stratified analyses, inverse U-shaped associations were observed in men, while associations were largely null in women. However, a formal test for effect modification by sex did not support differences in CRC risk associations between men and women. Potential explanations for this are that there are truly no differences between men and women in CRC risk related to bilirubin or more likely that the sample size in sex-stratified analyses was not sufficiently large to detect possible differences.
Bilirubin has been long thought to protect against oxidative stress-induced cancer initiation [14]. In its unconjugated form (approximated by levels of indirect bilirubin), bilirubin is normally present in the gut and can cross gut cell membranes [27]. It has also been shown to increase the expression of p53, a protein that controls cell apoptosis and autophagy in CRC tumour cells [28]. Taken together, the hypothesis that bilirubin plays a role in CRC development is compelling, however, studies in humans with CRC endpoints have so far been inconclusive [12–19, 21]. Our study adds observational evidence that bilirubin may be inversely associated with CRC risk and expands existing studies to Korean populations.
Our findings contrast the previous study among Korean adults (238 CRC cases) that reported only a weak inverse association between total bilirubin and CRC risk with a confidence interval including the null [21]. We speculate that the larger sample size (539 CRC cases) in our study and consideration of a non-linear association can explain this difference. Our findings are congruent with studies among adults (men and women combined) from Japan [19], the USA [12, 14], and the UK [18]. In contrast, largely null [13, 17] or suggestive positive associations [15] between bilirubin (either total or indirect/unconjugated bilirubin) and CRC risk were observed in other studies among European populations in sex-combined analyses [15, 17]. Some of these inconsistencies between studies are most likely attributable to differences in study design and/or sample size.
Recent studies from our research team in European populations [16, 17] showed sex differences in the associations. Both nested case-control studies [16, 17] within the large European cohorts, the European Prospective Investigation into Cancer and Nutrition (EPIC) and Cooperative Health Research in the Region Augsburg (KORA) observed positive associations among men and inverse associations among women. Additionally, this different relationship of bilirubin in CRC development by sex was supported by a Mendelian randomization (MR) study, where genetically predicted higher levels of total bilirubin using a UGT1A1 SNP (rs6431625) were associated with a 7% increase in CRC risk in men, while there was no association in women [16]. Women seem to be less susceptive to oxidative stress, because of antioxidant properties of oestrogen, differences in nicotinamide adenine dinucleotide phosphate (NADPH)-oxidase activity or other yet unknown mechanism(s) [29]. The purported anti-oxidative properties of bilirubin could therefore play a more important role in men than women, which might be one explanation for sex differences in associations between bilirubin and oxidative stress-induced diseases such as CRC. Furthermore, bilirubin concentrations are lower in women than in men because serum levels of both bilirubin and oestrogens are competitively regulated by UGT1A1 [30].
We found strong evidence that the association between bilirubin and CRC risk was non-linear among men, indicating a U-shaped relationship with the lowest risk in the middle range (approximately 0.8 mg/dL = 13.7 μmol/L) of the total bilirubin levels. We observed a similarly U-shaped association in our UK Biobank analysis [18]. A study among a Japanese population [19] also examined a non-linear association between serum bilirubin and colon cancer risk. In that study, only high bilirubin levels ≥1.2 mg/dL, as in Gilbert’s syndrome (GS) [31], were significantly associated with a decreased risk of colon cancer, but not with lower bilirubin levels <1.2 mg/dL. Further, they also examined that low bilirubin levels <0.6 mg/dL were associated with an increased risk of all-cause mortality, as shown in our study with an increased CRC risk <0.8 mg/dL, but only in men. These findings, therefore, suggest that potential anti-cancer activity of bilirubin is in the range of mild hyperbilirubinemia.
In the current study, we found that the effect size of indirect bilirubin (indicating unconjugated bilirubin) was almost the same as of the total bilirubin (HR tertile 3 vs. 1 for total bilirubin = 0.74; HR tertile 3 vs. 1 for indirect bilirubin = 0.74). A similar non-linear association was also observed for indirect bilirubin in men. This supports the stronger anti-cancer potential of the unconjugated form of bilirubin [32, 33].
Bilirubin is routinely measured in the clinical setting as part of standard liver function tests [34]. Our results support further studies on repurposing bilirubin additionally as a low-cost biomarker for CRC risk stratification. Accurate risk stratification is important clinically and for cost-effectiveness of screening programs. Given the high incidence of CRC and low participation rates in population-based screening, including in Korea – about 30% in 2017 [35], even a small improvement in risk prediction could have a meaningful impact and is worth exploring. Yet, clinical cut-off values for serum bilirubin concentrations that predict diseases beyond liver function remain currently intensively debated [36].
Our findings should be interpreted in light of the following limitations. This is an observational study and cannot, therefore, establish causality. However, the prospective design with a relatively long follow-up and the nationwide community-based sample provides greater confidence in our conclusions. The limited number of cancer events may partly explain why we did not detect effect modification by sex and the other potential effect modifiers investigated. Although we adjusted for several established risk factors of CRC, unobserved or residual confounding cannot be entirely ruled out. Another concern is reverse causation, which we addressed by censoring CRC events ascertained within one year without appreciable changes in results and by additional adjustment for inflammatory and liver function markers. The strengths of our study include the measurement of total and indirect bilirubin in pre-diagnostic serum of a nationwide community-based study population, which limits selection bias and strengthens the generalizability of our findings. It is specifically important to investigate the potential of the host factor bilirubin in preventing cancer risk in a population with increasing cancer incidence but a lower prevalence of GS. Future studies should establish or refute these associations, for example, by using MR approaches, and consider circulating bilirubin for CRC risk stratification in risk prediction models. Clinically, low bilirubin levels could be equally of concern as are abnormally high bilirubin levels, the latter mainly due to liver diseases.
Conclusions
In this prospective study among Korean men and women, higher circulating levels of total and indirect bilirubin were inversely associated with CRC risk. In sex-stratified analyses, associations were strongly inverse and U-shaped in men, while no associations were observed in women. Future studies that examine potential causality accounting for non-linearity and consider bilirubin for CRC risk stratification would be required.
Disclaimer
Where authors are identified as personnel of the International Agency for Research on Cancer/ World Health Organization, the authors alone are responsible for the views expressed in this article and they do not necessarily represent the decisions, policy or views of the International Agency for Research on Cancer/ World Health Organization.
Supplementary information
Acknowledgements
The authors appreciate all participants in the KoGES-HEXA Cohort Study.
Author contributions
All authors have contributed to the publication according to the ICMJE guidelines for authorship. The specific contributions of each author are as follows. Study conceptualization and funding acquisition: K-HW, AS, and HF; Study concept and design: AS and HF; Data acquisition: JL, DK, and AS; data analysis: JL; supervision and interpretation of data analysis: HN, AS, and HF; Drafting of the manuscript: HN and HF; Reviewing of the manuscript: JL, NSK, LP-N, BF, AS, and HF; All authors approved the final version for submission.
Funding
This work was supported by the Austrian Science Fund (FWF, grant nr. P 32303) and the National Research Foundation of Korea (NRF) grant funded by the Korean government (MSIT) (No. 2022R1A2C1004608).
Data availability
The datasets analysed during the current study are available after the approval of the review committee of the Korea National Institute of Health. The authors are not authorized to share the data other than the approved research group.
Competing interests
The authors declare no competing interests.
Ethics approval and consent to participate
This research was conducted according to the principles expressed in the Declaration of Helsinki. All study participants provided informed consent. Ethical approval was obtained from the IRBs of Seoul National University and collaborating centres of the KoGES groups, and an additional ethical approval specifically for this project was obtained from the IRB of Seoul National University College of Medicine/Hospital (Reference no.: E1810-006-974).
Consent for publication
The manuscript does not contain any individual person’s data. All reasonable measures have been taken to ensure the anonymity of study participants.
Footnotes
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
These authors contributed equally: Hwayoung Noh, Jeeyoo Lee.
Contributor Information
Aesun Shin, Email: shinaesun@snu.ac.kr.
Heinz Freisling, Email: FreislingH@iarc.who.int.
Supplementary information
The online version contains supplementary material available at 10.1038/s41416-024-02847-9.
References
- 1.IARC/WHO. The Global Cancer Observatory - Cancer Fact Sheets: Colorectal Cancer. Lyon: International Agency for Research on Cancer (IARC); 2020.
- 2.Kang MJ, Jung KW, Bang SH, Choi SH, Park EH, Yun EH, et al. Cancer Statistics in Korea: Incidence, Mortality, Survival, and Prevalence in 2020. Cancer Res Treat. 2023;55:385–99. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell. 2011;144:646–74. [DOI] [PubMed] [Google Scholar]
- 4.Lasry A, Zinger A, Ben-Neriah Y. Inflammatory networks underlying colorectal cancer. Nat Immunol. 2016;17:230–40. [DOI] [PubMed] [Google Scholar]
- 5.Wagner KH, Wallner M, Mölzer C, Gazzin S, Bulmer AC, Tiribelli C, et al. Looking to the horizon: the role of bilirubin in the development and prevention of age-related chronic diseases. Clin Sci. 2015;129:1–25. [DOI] [PubMed] [Google Scholar]
- 6.Sedlak TW, Saleh M, Higginson DS, Paul BD, Juluri KR, Snyder SH. Bilirubin and glutathione have complementary antioxidant and cytoprotective roles. Proc Natl Acad Sci USA. 2009;106:5171–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Rodrigues CMP, Solá S, Brito MA, Brites D, Moura JJG. Bilirubin directly disrupts membrane lipid polarity and fluidity, protein order, and redox status in rat mitochondria. J Hepatol. 2002;36:335–41. [DOI] [PubMed] [Google Scholar]
- 8.Hansen TW, Mathiesen SB, Walaas SI. Bilirubin has widespread inhibitory effects on protein phosphorylation. Pediatr Res. 1996;39:1072–7. [DOI] [PubMed] [Google Scholar]
- 9.Fevery J. Bilirubin in clinical practice: a review. Liver Int J Int Assoc Study Liver. 2008;28:592–605. [DOI] [PubMed] [Google Scholar]
- 10.Buckley DB, Klaassen CD. Mechanism of gender-divergent UDP-glucuronosyltransferase mRNA expression in mouse liver and kidney. Drug Metab Dispos Biol Fate Chem. 2009;37:834–40. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Vítek L, Tiribelli C. Bilirubin: The yellow hormone? J Hepatol. 2021;75:1485–90. [DOI] [PubMed] [Google Scholar]
- 12.Zucker SD, Horn PS, Sherman KE. Serum bilirubin levels in the U.S. population: gender effect and inverse correlation with colorectal cancer. Hepatol Balt Md. 2004;40:827–35. [DOI] [PubMed] [Google Scholar]
- 13.Ioannou GN, Liou IW, Weiss NS. Serum bilirubin and colorectal cancer risk: a population-based cohort study. Aliment Pharm Ther. 2006;23:1637–42. [DOI] [PubMed] [Google Scholar]
- 14.Jirásková A, Novotný J, Novotný L, Vodicka P, Pardini B, Naccarati A, et al. Association of serum bilirubin and promoter variations in HMOX1 and UGT1A1 genes with sporadic colorectal cancer. Int J Cancer. 2012;131:1549–55. [DOI] [PubMed] [Google Scholar]
- 15.Kühn T, Sookthai D, Graf ME, Schübel R, Freisling H, Johnson T, et al. Albumin, bilirubin, uric acid and cancer risk: results from a prospective population-based study. Br J Cancer. 2017;117:1572–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Seyed Khoei N, Jenab M, Murphy N, Banbury BL, Carreras-Torres R, Viallon V, et al. Circulating bilirubin levels and risk of colorectal cancer: serological and Mendelian randomization analyses. BMC Med. 2020;18:229. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Seyed Khoei N, Anton G, Peters A, Freisling H, Wagner KH. The Association between Serum Bilirubin Levels and Colorectal Cancer Risk: Results from the Prospective Cooperative Health Research in the Region of Augsburg (KORA) Study in Germany. Antioxidants. 2020;9:908. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Seyed Khoei N, Wagner KH, Carreras-Torres R, Gunter MJ, Murphy N, Freisling H. Associations between Prediagnostic Circulating Bilirubin Levels and Risk of Gastrointestinal Cancers in the UK Biobank. Cancers. 2021;13:2749. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Inoguchi T, Nohara Y, Nojiri C, Nakashima N. Association of serum bilirubin levels with risk of cancer development and total death. Sci Rep. 2021;11:13224. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Kang TW, Kim HJ, Ju H, Kim JH, Jeon YJ, Lee HC, et al. Genome-wide association of serum bilirubin levels in Korean population. Hum Mol Genet. 2010;19:3672–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Kim YR, Choi CK, Lee YH, Choi SW, Kim HY, Shin MH, et al. Association between Albumin, Total Bilirubin, and Uric Acid Serum Levels and the Risk of Cancer: A Prospective Study in a Korean Population. Yonsei Med J. 2021;62:792–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Kim Y, Han BG, Group the K. Cohort Profile: The Korean Genome and Epidemiology Study (KoGES) Consortium. Int J Epidemiol. 2017;46:e20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Health Examinees Study Group. The Health Examinees (HEXA) study: rationale, study design and baseline characteristics. Asian Pac J Cancer Prev APJCP. 2015;16:1591–7. [DOI] [PubMed] [Google Scholar]
- 24.Ahn Y, Kwon E, Shim JE, Park MK, Joo Y, Kimm K, et al. Validation and reproducibility of food frequency questionnaire for Korean genome epidemiologic study. Eur J Clin Nutr. 2007;61:1435–41. [DOI] [PubMed] [Google Scholar]
- 25.Kim J, Kim Y, Ahn YO, Paik HY, Ahn Y, Tokudome Y, et al. Development of a food frequency questionnaire in Koreans. Asia Pac J Clin Nutr. 2003;12:243–50. [PubMed] [Google Scholar]
- 26.Kityo A, Lee SA. Longitudinal changes in high sensitivity C-reactive protein associated with serum uric acid in the Korean Genome and Epidemiology Study. Sci Rep. 2024;14:374. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Zucker SD, Goessling W, Hoppin AG. Unconjugated bilirubin exhibits spontaneous diffusion through model lipid bilayers and native hepatocyte membranes. J Biol Chem. 1999;274:10852–62. [DOI] [PubMed] [Google Scholar]
- 28.Niknam M, Maleki MH, Khakshournia S, Rasouli M, Vakili O, Shafiee SM. Bilirubin, an endogenous antioxidant that affects p53 protein and its downstream apoptosis/autophagy-related genes in LS180 and SW480 cell culture models of colorectal cancer. Biochem Biophys Res Commun. 2023;672:161–7. [DOI] [PubMed] [Google Scholar]
- 29.Kander MC, Cui Y, Liu Z. Gender difference in oxidative stress: a new look at the mechanisms for cardiovascular diseases. J Cell Mol Med. 2017;21:1024–32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Cheng Z, Rios GR, King CD, Coffman BL, Green MD, Mojarrabi B, et al. Glucuronidation of catechol estrogens by expressed human UDP-glucuronosyltransferases (UGTs) 1A1, 1A3, and 2B7. Toxicol Sci J Soc Toxicol. 1998;45:52–7. [DOI] [PubMed] [Google Scholar]
- 31.Bosma PJ, Chowdhury JR, Bakker C, Gantla S, de Boer A, Oostra BA, et al. The genetic basis of the reduced expression of bilirubin UDP-glucuronosyltransferase 1 in Gilbert’s syndrome. N Engl J Med. 1995;333:1171–5. [DOI] [PubMed] [Google Scholar]
- 32.Keshavan P, Schwemberger SJ, Smith DLH, Babcock GF, Zucker SD. Unconjugated bilirubin induces apoptosis in colon cancer cells by triggering mitochondrial depolarization. Int J Cancer. 2004;112:433–45. [DOI] [PubMed] [Google Scholar]
- 33.Corich L, Aranda A, Carrassa L, Bellarosa C, Ostrow JD, Tiribelli C. The cytotoxic effect of unconjugated bilirubin in human neuroblastoma SH-SY5Y cells is modulated by the expression level of MRP1 but not MDR1. Biochem J. 2009;417:305–12. [DOI] [PubMed] [Google Scholar]
- 34.Guerra Ruiz AR, Crespo J, López Martínez RM, Iruzubieta P, Casals Mercadal G, Lalana Garcés M, et al. Measurement and clinical usefulness of bilirubin in liver disease. Adv Lab Med. 2021;2:352–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Park B, Her EY, Lee K, Nari F, Jun JK, Choi KS, et al. Overview of the National Cancer Screening Program for Colorectal Cancer in Korea over 14 Years (2004-2017). Cancer Res Treat. 2023;55:910–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Vítek L, Tiribelli C. Gilbert’s syndrome revisited. J Hepatol. 2023;79:1049–55. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets analysed during the current study are available after the approval of the review committee of the Korea National Institute of Health. The authors are not authorized to share the data other than the approved research group.