Bilirubin, the Gold Within

Construction of machine learning diagnostic models for cardiovascular pan-disease based on blood routine and biochemical detection data

BACKGROUND

Cardiovascular disease, also known as circulation system disease, remains the leading cause of morbidity and mortality worldwide. Traditional methods for diagnosing cardiovascular disease are often expensive and time-consuming. So the purpose of this study is to construct machine learning models for the diagnosis of cardiovascular diseases using easily accessible blood routine and biochemical detection data and explore the unique hematologic features of cardiovascular diseases, including some metabolic indicators.

METHODS

After the data preprocessing, 25,794 healthy people and 32,822 circulation system disease patients with the blood routine and biochemical detection data were utilized for our study. We selected logistic regression, random forest, support vector machine, eXtreme Gradient Boosting (XGBoost), and deep neural network to construct models. Finally, the SHAP algorithm was used to interpret models.

RESULTS

The circulation system disease prediction model constructed by XGBoost possessed the best performance (AUC: 0.9921 (0.9911-0.9930); Acc: 0.9618 (0.9588-0.9645); Sn: 0.9690 (0.9655-0.9723); Sp: 0.9526 (0.9477-0.9572); PPV: 0.9631 (0.9592-0.9668); NPV: 0.9600 (0.9556-0.9644); MCC: 0.9224 (0.9165-0.9279); F1 score: 0.9661 (0.9634-0.9686)). Most models of distinguishing various circulation system diseases also had good performance, the model performance of distinguishing dilated cardiomyopathy from other circulation system diseases was the best (AUC: 0.9267 (0.8663-0.9752)). The model interpretation by the SHAP algorithm indicated features from biochemical detection made major contributions to predicting circulation system disease, such as potassium (K), total protein (TP), albumin (ALB), and indirect bilirubin (NBIL). But for models of distinguishing various circulation system diseases, we found that red blood cell count (RBC), K, direct bilirubin (DBIL), and glucose (GLU) were the top 4 features subdividing various circulation system diseases.

CONCLUSIONS

The present study constructed multiple models using 50 features from the blood routine and biochemical detection data for the diagnosis of various circulation system diseases. At the same time, the unique hematologic features of various circulation system diseases, including some metabolic-related indicators, were also explored. This cost-effective work will benefit more people and help diagnose and prevent circulation system diseases.

Clinical characteristics and predictors of pulmonary hypertension in chronic obstructive pulmonary disease at different altitudes

BACKGROUND

Pulmonary hypertension (PH) is a common complication in patients with chronic obstructive pulmonary disease (COPD) and is closely associated with poor prognosis. However, studies on the predictors of PH in COPD patients are limited, especially in populations living at high altitude (HA).

OBJECTIVES

To investigate the differences in the clinical characteristics and predictors of patients with COPD/COPD and PH (COPD-PH) from low altitude (LA, 600 m) and HA (2200 m).

METHODS

We performed a cross-sectional survey of 228 COPD patients of Han nationality admitted to the respiratory department of Qinghai People's Hospital (N = 113) and West China Hospital of Sichuan University (N = 115) between March 2019 and June 2021. PH was defined as a pulmonary arterial systolic pressure (PASP) > 36 mmHg measured using transthoracic echocardiography (TTE).

RESULTS

The proportion of PH in COPD patients living at HA was higher than that in patients living at LA (60.2% vs. 31.3%). COPD-PH patients from HA showed significantly different in baseline characteristics, laboratory tests and pulmonary function test. Multivariate logistic regression analysis indicated that the predictors of PH in COPD patients were different between the HA and LA groups.

CONCLUSIONS

The COPD patients living at HA had a higher proportion of PH than those living at LA. At LA, increased B-type natriuretic peptide (BNP) and direct bilirubin (DB) were predictors for PH in COPD patients. However, at HA, increased DB was a predictor of PH in COPD patients.

The role of tolvaptan in pulmonary hypertension: A retrospective study

Pulmonary hypertension (PH) is a severe form of pulmonary vascular disease that can lead to right heart failure (RHF). Nearly 2-thirds of patients with PH die within 5 years. Studies suggest that a new diuretic medication, called tolvaptan (TLV), can be used to treat PH. However, there is still insufficient evidence to confirm its effectiveness. Therefore, we investigated the role of TLV in patients with PH. This retrospective study included 73 patients with PH hospitalized in Shanghai Pulmonary Hospital between November 2019 and March 2022. All patients received 7.5 to 15.0 mg of TLV for 3 to 21 days starting at admission, in addition to targeted drugs and traditional diuretic therapy. The outcomes included the blood pressure, urine and water intake volumes, electrolyte concentrations, and renal, liver, and cardiac function indexes before and after TLV treatment. In addition, we assessed the clinical symptoms and adverse reactions during the treatment. After TLV treatment, the water intake and urine volumes significantly increased, and body weight, diastolic blood pressure (DBP) and mean arterial pressure significantly decreased. Total bilirubin, direct bilirubin, N-terminal pro-brain natriuretic peptide, and serum uric acid (UA) levels after TLV treatment were significantly lower than before treatment. After TLV treatment, dyspnea significantly improved in 71 of 73 patients, and lower limb edema disappeared in 42 of 53 patients. No obvious adverse reactions occurred during the TLV treatment period. These results suggest that adding TLV to targeted drug and traditional diuretic therapies is effective for patients with PH. However, more data are required to support these findings.

Dose-Response Association Between Bilirubin and Cardiovascular Disease: A Systematic Review and Meta-analysis

The association between bilirubin (BIL) and cardiovascular disease (CVD) remains controversial. We performed a meta-analysis of prospective studies to evaluate this association in the general population. We searched PubMed, EMBASE, Web of Science, Cochrane, and Scopus databases through to September 2021. The Newcastle-Ottawa Quality Assessment Scale was used to assess study quality. The pooled effect estimate was calculated by the fixed-effect model or random-effect model. We included 12 prospective studies (368 567 participants). The pooled risk ratio of CVD for the lowest vs highest groups of BIL levels was .75 (95% CI: .58-.97) with high heterogeneity (I² = 87.5%, P < .001). Similar associations were observed for coronary heart disease and stroke. We further performed a "dose-response" meta-analysis, and a significant U-shaped relationship between circulating (most values were serum bilirubin, but a few were plasma bilirubin) BIL and CVD (P < .01) was observed. The lowest risk of CVD events was observed in participants with a BIL of 17-20 µmol/L in serum and/or plasma. In conclusion, there was a U-shaped dose-response relationship between BIL and CVD, especially for men. Further studies are needed to confirm our findings and identify the mechanisms involved as well as any prognostic or therapeutic potential.

Characteristics and Risk Factors of Pulmonary Hypertension in Patients With Hyperthyroidism

OBJECTIVE

This study aimed to comprehensively assess the characteristics and risk factors of hyperthyroidism with pulmonary hypertension (PH).

METHODS

This was a retrospective cross-sectional analysis of 315 consecutive patients with hyperthyroidism admitted to the endocrinology department of Tongji Hospital from February 2016 to December 2017. PH was defined as a pulmonary arterial systolic pressure above 35 mm Hg measured by echocardiography.

RESULTS

Among the 315 patients, 208 were females, the median age was 42 (30-51) years, and the median disease duration was 12 (3-48) months. Thirty-five percent (111/315) of patients were identified with PH. Patients with hyperthyroidism and PH showed significantly higher serum concentrations of free thyroxine (FT4), free triiodothyronine, thyroid receptor antibodies, total bilirubin (TB), direct and indirect bilirubin, lower serum levels of hemoglobin and creatinine, and more severe cardiac load (P < .05 for each) compared with patients without PH. Levels of serum FT4, free triiodothyronine, thyroid receptor antibodies, and thyroid peroxidase antibody were different among groups of patients with different levels of pulmonary arterial systolic pressure (P < .05 for each). Multivariate logistic regression analysis indicated that serum FT4 (odds ratio, 1.02; 95% CI, 1.01-1.04; P = .004) and TB (OR, 1.03; 95% CI, 1.00-1.06; P = .030) were independent risk factors for PH in patients with hyperthyroidism.

CONCLUSION

Elevated serum FT4 and TB levels may be independent risk factors for PH in patients with hyperthyroidism and valuable indicators for the identification and treatment of patients with PH and hyperthyroidism.

Exploring the causal pathway from bilirubin to CVD and diabetes in the UK biobank cohort study: Observational findings and Mendelian randomization studies

BACKGROUND AND AIMS

Some studies reported that mildly elevated serum bilirubin levels were associated with decreased risk of cardiovascular disease (CVD) and diabetes. Whether these are causal relationships remains unclear. This study aims to examine the causal effects of bilirubin on CVD, diabetes and their subtypes.

METHODS

The data we used in this study includes individual data from the UK Biobank cohort with 331,002 white British participants, and summary data from published genome wide associations studies (GWAS) findings. We used individual data to perform logistic regression for the observational study and two-stage least squares method for the Mendelian randomization (MR) study. We also performed several traditional MR methods and MR-TRYX by summary data.

RESULTS

The observational study supported the association relationships between bilirubin and CVD and diabetes and their subtypes. Results of MR showed strong evidence for negative causal associations of log_e total bilirubin with CVD [OR 0.92, 95%CI 0.88-0.95, p-value 2.15 × 10^-6], coronary heart disease [OR 0.90, 95%CI 0.85-0.96, p-value 1.54 × 10^-3] and hypertensive diseases [OR 0.91, 95%CI 0.88-0.95, p-value 5.89 × 10^-6], but no evidence for diabetes [OR 0.94, 95%CI 0.86-1.02, p-value 0.14] and its subtypes. We also obtained similar results for direct bilirubin. We found that blood pressure, cholesterol, C-reactive protein, alcohol and white blood cell count played important roles in the causal pathway from bilirubin to CVD. Two sample MR and sensitivity analyses showed consistent results with one sample MR.

CONCLUSIONS

Genetically determined bilirubin was negatively associated with the risk of CVD but had no evident causal association with diabetes in the UK Biobank cohort of white British.

Circulating total bilirubin and risk of non-alcoholic fatty liver disease in the PREVEND study: observational findings and a Mendelian randomization study

The relationship between circulating total bilirubin and incident non-alcoholic fatty liver disease (NAFLD) is uncertain. We aimed to assess the association of total bilirubin with the risk of new-onset NAFLD and investigate any causal relevance to the association using a Mendelian randomization (MR) study. Plasma total bilirubin levels were measured at baseline in the PREVEND prospective study of 3824 participants (aged 28–75 years) without pre-existing cardiovascular disease or NAFLD. Incident NAFLD was estimated using the biomarker-based algorithms, fatty liver index (FLI) and hepatic steatosis index (HSI). Odds ratios (ORs) (95% confidence intervals) for NAFLD were assessed. The genetic variant rs6742078 located in the UDP-glucuronosyltransferase (UGT1A1) locus was used as an instrumental variable. Participants were followed up for a mean duration of 4.2 years. The multivariable adjusted OR (95% CIs) for NAFLD as estimated by FLI (434 cases) was 0.82 (0.73–0.92; p = 0.001) per 1 standard deviation (SD) change in log_e total bilirubin. The corresponding adjusted OR (95% CIs) for NAFLD as estimated by HSI (452 cases) was 0.87 (0.78–0.97; p = 0.012). The rs6742078 variant explained 20% of bilirubin variation. The ORs (95% CIs) for a 1 SD genetically elevated total bilirubin level was 0.98 (0.69–1.38; p = 0.900) for FLI and 1.14 (0.81–1.59; p = 0.451) for HSI. Elevated levels of total bilirubin were not causally associated with decreased risk of NAFLD based on MR analysis. The observational association may be driven by biases such as unmeasured confounding and/or reverse causation. However, due to low statistical power, larger-scale investigations are necessary to draw definitive conclusions.

Circulating Total Bilirubin and Future Risk of Hypertension in the General Population: The Prevention of Renal and Vascular End-Stage Disease (PREVEND) Prospective Study and a Mendelian Randomization Approach

BACKGROUND

Circulating total bilirubin is known to be inversely and independently associated with future risk of cardiovascular disease. However, the relationship of circulating total bilirubin with incident hypertension is uncertain. We aimed to assess the association of total bilirubin with future hypertension risk and supplemented this with a Mendelian randomization approach to investigate any causal relevance to the association.

METHODS AND RESULTS

Plasma total bilirubin levels were measured at baseline in the PREVEND (Prevention of Renal and Vascular End-Stage Disease) prospective study of 3989 men and women without hypertension. Hazard ratios (95% confidence intervals) of total bilirubin with incident hypertension were assessed. New-onset hypertension was recorded in 1206 participants during a median follow-up of 10.7 years. Baseline total bilirubin was approximately log-linearly associated with hypertension risk. Age- and sex-adjusted hazard ratio for hypertension per 1-SD increase in loge total bilirubin was 0.86 (0.81-0.92; P<0.001), which was attenuated to 0.94 (0.88-0.99; P=0.040) after further adjustment for established risk factors and other potential confounders. The association was marginally significant on further adjustment for high-sensitivity C-reactive protein (0.94; 0.88-1.00; P=0.067). A genetic variant at the UGT1A1*28 locus consistently shown to be strongly associated with circulating bilirubin levels-rs6742078-was not significantly associated with blood pressure or hypertension (P>0.05 for all), arguing against a strong causal association of circulating bilirubin with blood pressure.

CONCLUSIONS

The weak and inverse association of circulating total bilirubin with future hypertension risk may be driven by biases such as unmeasured confounding and/or reverse causation. Further evaluation is warranted.

Lineage-Specific Changes in Biomarkers in Great Apes and Humans

Although human biomedical and physiological information is readily available, such information for great apes is limited. We analyzed clinical chemical biomarkers in serum samples from 277 wild- and captive-born great apes and from 312 healthy human volunteers as well as from 20 rhesus macaques. For each individual, we determined a maximum of 33 markers of heart, liver, kidney, thyroid and pancreas function, hemoglobin and lipid metabolism and one marker of inflammation. We identified biomarkers that show differences between humans and the great apes in their average level or activity. Using the rhesus macaques as an outgroup, we identified human-specific differences in the levels of bilirubin, cholinesterase and lactate dehydrogenase, and bonobo-specific differences in the level of apolipoprotein A-I. For the remaining twenty-nine biomarkers there was no evidence for lineage-specific differences. In fact, we find that many biomarkers show differences between individuals of the same species in different environments. Of the four lineage-specific biomarkers, only bilirubin showed no differences between wild- and captive-born great apes. We show that the major factor explaining the human-specific difference in bilirubin levels may be genetic. There are human-specific changes in the sequence of the promoter and the protein-coding sequence of uridine diphosphoglucuronosyltransferase 1 (UGT1A1), the enzyme that transforms bilirubin and toxic plant compounds into water-soluble, excretable metabolites. Experimental evidence that UGT1A1 is down-regulated in the human liver suggests that changes in the promoter may be responsible for the human-specific increase in bilirubin. We speculate that since cooking reduces toxic plant compounds, consumption of cooked foods, which is specific to humans, may have resulted in relaxed constraint on UGT1A1 which has in turn led to higher serum levels of bilirubin in humans.

Circulating Markers of Liver Function and Cardiovascular Disease Risk

Measurement of serum concentrations of various liver enzymes and other nonenzymatic proteins and metabolites of heme metabolism (eg, bilirubin) is often undertaken in clinical practice. Measurement of these liver function tests is simple, quick, and relatively inexpensive. However, interpreting the liver function test results in patients without evidence of liver disease is often challenging. Concentrations of some of liver enzymes, such as γ-glutamyltransferase or alkaline phosphatase, and concentrations of liver-derived metabolites, such as bilirubin, may be influenced by metabolic processes beyond the liver, sometimes making interpretation of the test results difficult. This scenario frequently occurs both in individuals at risk of cardiovascular disease and in patients with known cardiovascular disease, often resulting in the clinicians ignoring the test results. In this brief review, we discuss the evidence for associations between key serum liver function tests and cardiovascular disease risk and where associations are robust; we provide an interpretation for possible mechanistic links between the liver function test and cardiovascular disease.

Bilirubin as a potential causal factor in type 2 diabetes risk: a Mendelian randomization study

Circulating bilirubin, a natural antioxidant, is associated with decreased risk of type 2 diabetes (T2D), but the nature of the relationship remains unknown. We performed Mendelian randomization in a prospective cohort of 3,381 participants free of diabetes at baseline (age 28-75 years; women 52.6%). We used rs6742078 located in the uridine diphosphate-glucuronosyltransferase locus as an instrumental variable (IV) to study a potential causal effect of serum total bilirubin level on T2D risk. T2D developed in a total of 210 participants (6.2%) during a median follow-up period of 7.8 years. In adjusted analyses, rs6742078, which explained 19.5% of bilirubin variation, was strongly associated with total bilirubin (a 0.68-SD increase in bilirubin levels per T allele; P < 1 × 10(-122)) and was also associated with T2D risk (odds ratio [OR] 0.69 [95% CI 0.54-0.90]; P = 0.006). Per 1-SD increase in log-transformed bilirubin levels, we observed a 25% (OR 0.75 [95% CI 0.62-0.92]; P = 0.004) lower risk of T2D. In Mendelian randomization analysis, the causal risk reduction for T2D was estimated to be 42% (causal OR for IV estimation per 1-SD increase in log-transformed bilirubin 0.58 [95% CI 0.39-0.84]; P = 0.005), which was comparable to the observational estimate (Durbin-Wu-Hausman χ(2) test, P for difference = 0.19). These novel results provide evidence that an elevated bilirubin level is causally associated with the risk of T2D and support its role as a protective determinant.

Circulating Total Bilirubin and Risk of Incident Cardiovascular Disease in the General Population

Objective—

To assess the association of circulating total bilirubin and cardiovascular disease (CVD) risk in a new prospective study and to determine whether adding information on total bilirubin values to established cardiovascular risk factors is associated with improvement in prediction of CVD risk.

Approach and Results—

Circulating total bilirubin levels were measured at baseline in the PREVEND (Prevention of Renal and Vascular End-stage Disease) prospective study of 7222 participants and 773 incident CVD events. Total bilirubin was log-linearly associated with CVD risk. Age- and sex-adjusted hazard ratio (95% confidence interval) for CVD per 1-SD increase in log e total bilirubin was 0.82 (0.76 to 0.88; P <0.001), which was minimally attenuated to 0.89 (0.82 to 0.96; P =0.003) after further adjustment for established risk factors. In a meta-analysis of 12 population-based prospective studies involving 9378 incident CVD cases, the pooled multivariate-adjusted relative risk (95% confidence interval) for CVD was 0.93 (0.90 to 0.97; P <0.001) per 1-SD increase in total bilirubin levels. The corresponding pooled risks for coronary heart disease and stroke were 0.95 (0.92 to 0.99; P =0.018) and 0.93 (0.88 to 0.98; P =0.006), respectively. Addition of information on total bilirubin to a CVD risk prediction model containing established risk factors was associated with a C-index change of 0.0013 (−0.0004 to 0.0029; P =0.13).

Conclusions—

There is a log-linear inverse association between circulating total bilirubin level and CVD risk, which is independent of established risk factors. Nonetheless, inclusion of total bilirubin in the standard established risk factors panel provides no significant improvement in CVD risk prediction.

Relation of conjugated bilirubin concentrations to the presence of coronary artery calcium

Whether conjugated bilirubin concentration, resulting from hepatic UDP-glucuronosyltransferase 1 A1 activity, is associated with cardiovascular disease is unknown. Our aim was to investigate the relation between serum conjugated bilirubin concentrations and coronary artery calcium score (CACS) as a measure of preclinical atherosclerosis. Data were analyzed from an occupational cohort of 14,583 subjects who underwent a cardiac computed tomographic estimation of CACS and measurements of risk factors. Logistic regression was used to describe associations between bilirubin concentrations and CACS. The proportion of subjects with a CACS >0 (total: men = 1,351, women = 111) decreased across increasing conjugated bilirubin quartiles (men p <0.001, women p = 0.005). After adjustment for age, gender, waist, systolic blood pressure, smoking, exercise, alcohol, homeostatic model assessment of insulin resistance, glucose, triglyceride, high- and low-density lipoprotein cholesterols, high-sensitivity C-reactive protein, lipoprotein (a), ferritin, fatty liver, cerebrovascular accident, coronary artery disease, hypertension, and diabetes, there was an inverse independent association between conjugated bilirubin and CACS >0 (odds ratio 0.76, 95% confidence interval 0.64 to 0.92, p = 0.004, for top vs bottom quartile). After further adjustment for estimated glomerular filtration rate, the odds ratio was 0.83 (95% confidence interval 0.69 to 0.99), p = 0.04. Exclusion of subjects with a high bilirubin (total >1.3 mg/dl) did not attenuate the association. In conclusion, there was a strong inverse and independent relation between conjugated bilirubin and CACS. The mechanism behind this association is not clear and may not be causal, but the effects of glucuronidation on cardiovascular disease risk should be tested.