Predictive value of different bilirubin subtypes for clinical outcomes in patients with acute ischemic stroke receiving thrombolysis therapy

Abstract Aims To explore the association of total bilirubin (TBIL), direct bilirubin (DBIL), and indirect bilirubin (IBIL) levels with, as well as the incremental predictive value of different bilirubin subtypes for, poor outcomes in acute ischemic stroke patients after thrombolysis. Methods We analyzed 588 individuals out of 718 AIS participants, and all patients were followed up at 3 months after thrombolysis. The primary outcome was 3‐month death and major disability (modified Rankin Scale (mRS) score of 3–6). The secondary outcomes were 3‐month mortality (mRS score of 6), moderate‐severe cerebral edema, and symptomatic intracranial hemorrhage (sICH), respectively. Results Elevated DBIL pre‐thrombolysis was associated with an increased risk of primary outcome (OR 3.228; 95% CI 1.595–6.535; p for trend = 0.014) after fully adjustment. Elevated TBIL pre‐thrombolysis showed the similar results (OR 2.185; 95% CI 1.111–4.298; p for trend = 0.047), while IBIL pre‐thrombolysis was not significantly associated with primary outcome (OR 1.895; 95% CI 0.974–3.687; p for trend = 0.090). Multivariable‐adjusted spline regression model showed a positive linear dose‐response relationship between DBIL pre‐thrombolysis and risk of primary outcome (p for linearity = 0.004). Adding DBIL pre‐thrombolysis into conventional model had greater incremental predictive value for primary outcome, with net reclassification improvement (NRI) 95% CI = 0.275 (0.084–0.466) and integrated discrimination improvement (IDI) 95% CI = 0.011 (0.001–0.024). Increased DBIL post‐thrombolysis had an association with primary outcome (OR 2.416; 95%CI 1.184–4.930; p for trend = 0.039), and it also elevated the incremental predictive value for primary outcome, with NRI (95% CI) = 0.259 (0.066–0.453) and IDI (95% CI) = 0.025 (0.008–0.043). Conclusion Increased DBIL pre‐thrombolysis had a stronger association with, as well as greater incremental predictive value for, poor outcomes than TBIL and IBIL did in AIS patients after thrombolysis, which should be understood in the context of retrospective design. The effect of DBIL on targeted populations should be investigated in further researches.


| INTRODUC TI ON
Acute ischemic stroke (AIS) has a high disability and mortality rate, which brings a huge economic burden to the society and the family. 1,2 To date, thrombolytic therapy with recombinant tissue plasminogen activator (rt-PA) and endovascular thrombectomy (EVT) are still the frontline treatment strategies for acute ischemic stroke within the time window. 3 However, the overall effectiveness of these treatments has been reported to be limited, with only 30%-50% of patients achieving good long-term outcomes. 4,5 Many patients receiving reperfusion therapy are at high risk of suffering from certain complications such as cerebral hemorrhage transformation and cerebral edema, who could not achieve good clinical outcome after discharge. To date, usefulness of a biomarker is limited to identify patients at high risk of getting worse clinical outcomes. 6,7 Therefore, alternative markers that have the potential to identify targeted patients pre-thrombolysis to escalate preventive therapy are thus needed.
Bilirubin, a potent endogenous antioxidant, is produced in the heme catabolic pathway, with liver being the primary organ responsible for metabolism and excretion of bilirubin. 8 Clinically, bilirubin levels are reported as total bilirubin (TBIL) and direct bilirubin (DBIL), and TBIL is the sum of DBIL and indirect bilirubin (IBIL). 9 Previous studies reported that bilirubin exhibits both neurotoxic and neuroprotective effects after ischemic stroke, without reaching a consensus for the prognosis of ischemic stroke. 10 However, these studies almost merely concentrated on one of the subtypes of bilirubin, without distinguishing the difference among them. To the best of our knowledge, studies on the predictive value of TBIL, DBIL, and IBIL for the clinical outcomes in patients diagnosed with AIS receiving intravenous thrombolysis are still lacking.
In this study, we analyzed the association of three subtypes of bilirubin pre-thrombolysis with clinical outcomes as well as compared the performance of them as an indicator of worse outcomes among AIS patients receiving intravenous thrombolysis after ischemic stroke to elucidate that DBIL pre-thrombolysis level has the potential to identify patients who are likely to be at increased risk of poor outcomes after intravenous thrombolysis to escalate preventive therapy. City. We enrolled all AIS patients ≥18 years old who received intravenous thrombolysis therapy. Patients who had a diagnosis of (1) chronic hepatitis; (2) increased liver enzymes whose ALT or AST >twofold upper limit of normal range; (3) nephritis; (4) nephrolith; (5) cholecystitis; (6) gallstone as well as (7) who were lost to follow-up were excluded (n = 130). All available hospitalization data, including medical history, clinical examination, laboratory examination, retrospective design. The effect of DBIL on targeted populations should be investigated in further researches.

K E Y W O R D S
bilirubin subtype, ischemic stroke, metabolism, neurotoxicity, predictive value, thrombolysis F I G U R E 1 The flowchart of study population in this study diagnostic examination, imaging examination and discharge diagnosis, were used for the diagnosis of the above diseases. A total of 588 participants were involved in final analysis ( Figure 1). been obtained from all participants in the study.
All the patients treated with thrombolytic treatment were in line with the written institutional guidelines. The time window for thrombolysis is extended and limited up to 9 h guided by perfusion imaging. 11,12 Intravenous rt-PA injection (administered at a standard dose of 0.9 mg/kg body weight) was given according to the procedure recommended by the European Stroke Organization (ESO) 13 : 10% of the total dose being given as the first dose and the remaining dose being given within the next hour. Continuous monitoring and evaluation were conducted during thrombolysis procedure. After thrombolysis was completed, patients were transferred to the neurology intensive care unit (NICU) for intensive nursing.
Peripheral venous blood samples from patients were collected both pre-thrombolysis upon admission and within 1-3 days postthrombolysis. Samples with hemolysis were discarded. The demographic characteristics, clinical features, and medical history of all enrolled patients were collected. Stroke severity was assessed both upon admission pre-thrombolysis and post-thrombolysis by trained neurologists using National Institutes of Health Stroke Scale (NIHSS).
All routine laboratory examination results were obtained at emergency department (pre-thrombolysis) and inpatient department (post-thrombolysis). Hypertension is identified as one of the following conditions: blood pressure ≥140/90 mmHg or self-reported physician-diagnosed hypertension or current use of antihypertensive medication. 14 According to Chinese guidelines on the prevention and treatment of hyperlipidemia, hyperlipidemia is considered as the abnormity of lipids in the blood (total cholesterol >6.22 mmol/L or triglyceride >2.26 mmol/L or low-density lipoprotein cholesterol >4.14 mmol/L) or self-reported history of physician diagnosis of hyperlipidemia. 15 Patients with fasting glucose level >7.0 mmol/L or selfreported physician-diagnosed of diabetes or taking oral hypoglycemic drugs or insulin were defined as diabetes mellitus. 14 Participants were followed up by modified Rankin Scale (mRS) score at 3 months by trained neurologists who were not aware of the treatment allocation. The primary outcome was defined as 3month death and major disability (mRS score of 3-6), and secondary outcomes were defined as (1) 3-month mortality (mRS score

| Statistical analysis
To analyze the association of three subtypes of bilirubin with clinical outcomes, the participants were divided into 4 groups according to quartiles of TBIL, IBIL and DBIL, respectively. Kolmogorov-Smirnov (K-S) test of normality was used to assess data distribution.
Continuous variables with normal distributions were represented as mean ± standard difference (SD), while other variables were represented as median (interquartile range). Categorical variables were represented as numbers (percentages). The quartile differences of the baseline characteristics of each bilirubin subtype were tested with chi-square tests for categorical characteristics and analysis of variance (ANOVA) for continuous characteristics. Binary logistic regression models were used to estimate the relationship of three subtypes of bilirubin with primary and secondary outcomes. Odds ratios (ORs) and 95% confidence intervals (CIs) of primary and secondary clinical outcomes for higher quartiles compared with the lowest quartile and for each SD increase of log-transformed of three subtypes bilirubin were calculated. We constructed two models with progressive adjustment: model 1 was adjusted for age, sex, onset-time to treatment (OTT), admission glucose level, admission ALT, admission AST, cigarette smoking, alcohol drinking, history of stroke, cerebral hemorrhage, hypertension, diabetes mellitus and hyperlipidemia; model 2 was additionally adjusted for admission NIHSS score. We tested the linear trends across the quartiles of three subtypes of bilirubin by including the quartiles in the models as continuous variable. Restricted cubic spline (RCS) model with knots at the 5th, 35th, 65th, and 95th percentiles 17 was used to characterize the shape of the association of bilirubin level with primary outcome. Receiver operating characteristic (ROC) curve which is equivalent to the C-statistic was constructed to estimate the discriminative power of three subtypes bilirubin for primary and secondary outcomes as well as compare the discriminative power of DBIL both pre-and post-thrombolysis for primary outcome. The predictive power of each bilirubin pre-thrombolysis when added to conventional model (CM) was assessed by net reclassification improvement (NRI) and integrative discriminative improvement (IDI). 18 Additionally, subgroup analysis was performed to assess the potential modified effect of 12 interesting factors on the association between DBIL pre-thrombolysis and primary outcome. A two-sided p value <0.05 was considered to be statistically significant.
Statistical analyses were carried out using R software (

| Characteristics of study population
Overall, the mean age of participants at baseline was 64.7 years, and 66.4% of them were men. The median (interquartile range) overall were 3.6 (2.7-5.1) μmol/L for DBIL, 10.4 (7.8-14.2) μmol/L for TBIL, 6.8 (4.8-9.5) μmol/L for IBIL, respectively. Baseline characteristics of participants by DBIL level quartiles are shown in Table 1. Participants with higher DBIL were more likely to be older; to have higher admission NIHSS score; to have longer OTT; to have higher liver enzyme level of AST; and to have higher prevalence of 3-month mRS score of 3-6, 3-month mRS score of 6, moderate-severe cerebral edema as well as sICH; In contrast, the prevalence of hyperlipidemia decreased as DBIL level increased. And the similar characteristics of the study population by TBIL and IBIL levels are shown in Tables S1 and S2.

| Association of different bilirubin subtypes prethrombolysis with primary outcome
As shown in Table 2

| Association of different bilirubin subtypes prethrombolysis with secondary outcomes
As shown in Table 3, compared with first quartile of DBIL, the fully adjusted OR of 3-month mortality from the second to the fourth quartile in model 2 were 3.002 (0.748-11.502), 4.499    Table S3). In addition, none of DBIL, TBIL, and IBIL was in association with sICH (shown in Table S4).

| Performance of different bilirubin subtypes pre-thrombolysis as a biomarker for different clinical outcome
In the ROC analysis shown in Figure 3 and TA B L E 2 Odds ratios and 95% CI of primary outcome for quartiles of each serum bilirubin pre-thrombolysis risk factors in fully adjusted model 2, respectively. As shown in Table 4,  (Tables S6 and S7).

| Subgroup analysis for association between DBIL and primary outcome
We performed stratified analysis for the dose-response association of DBIL pre-thrombolysis with primary clinical outcome according to prespecified factors. The significant interaction was only found between DBIL and admission NIHSS score ≤10 (p for interaction <0.001, Figure 4).

| Association and Performance of DBIL Postthrombolysis for Primary Outcome
Firstly, we compared the concentration of DBIL pre-and postthrombolysis, showing that concentration of DBIL post-thrombolysis was higher than that of pre-thrombolysis (p < 0.001, Figure S1).  (Table S8). Furthermore, we explored the incremental predictive value of DBIL for primary outcome as well. As shown in Table S9, the addition of DBIL to the CM allowed a significant incremental prediction of risk with NRI (95% CI) = 0.259 (0.066-0.453) and IDI (95% CI) = 0.025 (0.008-0.043).
However, the overall discriminative power comparison of DBIL preand post-thrombolysis for primary outcome applied by ROC curves showed that the discriminative power difference was not significant (p = 0.908, Figure S2).

TA B L E 3
Odds ratios and 95% CI of 3-month mortality for quartiles of each serum bilirubin pre-thrombolysis F I G U R E 3 ROC analysis of three subtypes bilirubin for 3-month death and disability (A), 3-month mortality (B), and moderate-severe cerebral edema (C) showing a stronger association in individuals with admission NIHSS score ≤10 (mild-moderate stroke), while the underlying mechanism for this observation was unclear which needs further exploration.
What's more, DBIL post-thrombolysis was in association with, as well as provided excellent predictive value for, 3-month death and major disability. While the difference of discriminative accuracy for DBIL between pre-and post-thrombolysis was not significant, which meant that both DBIL pre-and post-thrombolysis levels were equivalent in predicting the primary outcome. Importantly, DBIL measured pre-thrombolysis would be much more clinically relevant, which may aid in predicting the risk of 3-month death and major disability. And DBIL post-thrombolysis may help clinicians decide which patients should be monitored more closely after thrombolysis.
As a systemic disease, ischemic stroke causes damage to other remote organs in the body, altering their signaling and metabolisms, including the liver. 26 A large accumulation of data has demonstrated that leukocytes infiltration could also occur inside liver after brain and is bound tightly to albumin to be transferred to the liver for the production of the conjugated form (i.e. direct bilirubin). 9 Although the exact mechanism of bilirubin uptake into hepatocyte is not that clear, it appears that circulating bilirubin dissociates from albumin before entering hepatocytes by organic anion transporters (OATP) family, particularly OATP1B1 and OATP1B3, which are present in the lipid bilayer of liver cell membranes. 33 Once in the hepatocytes of the liver, glucuronic acid is added to the unconjugated bilirubin by UDP-glucuronosyltransferase (UGT1A1), forming the conjugated bilirubin. 8 On the one hand, the newly conjugated bilirubin, secreted by the multidrug-resistant protein MRP2 (ABCC2), enters into the bile through canalicular membrane. On the other hand, MRP3 (ABCC3) deposited the conjugated bilirubin back into the blood. 34 The uptake of bilirubin into brain at the blood-brain barrier (BBB) may occur via one of organic anion transporters OATP1, and the efflux of bilirubin from the brain may be mediated by ATP-binding cassette subfamily B member 1 (ABCB1; also referred to as MDR1 Pglycoprotein), 35 and the expression of OATP1 and MDR1 at the BBB was reported to be upregulated after brain ischemia ( Figure S4). [35][36][37] Previously, most studies have investigated the mechanisms as Akt-protein kinase B system. 45 Furthermore, not only caspase-3 but also caspase-8 and caspase-9 could be activated by bilirubin, leading to neuronal apoptosis and necrosis, 46,47 and N-methyl-Daspartate (NMDA)-induced glutamate release and excitotoxic cell death were observed to be increased at higher concentration of bilirubin. 48,49 Bilirubin can also directly induce glial death, 50 -53 and it seems that neurons are more susceptible to bilirubin toxicity than astrocytes. 54 Actually, the integrity of BBB is broken after stroke, resulting in the increased entry of bilirubin from peripheral circulation into brain to exert neurotoxic effect. Additionally, as all types of cell could maintain certain concentration of bilirubin both at physiological and pathological condition, 55 it is therefore reasonable to speculate that brain cells could release intracellular bilirubin into bloodstream after brain ischemia reperfusion because of cell damage, together with peripheral-derived bilirubin, resulting in hyperbilirubinemia. In a word, hyperbilirubinemia after brain ischemia reperfusion probably acts as a cell injury marker Future studies are thus required to elucidate the specific differences of different bilirubin subtypes with respect to their molecular mechanisms of action.
F I G U R E 4 Subgroup analyses of the association between DBIL prethrombolysis and primary outcome.
Interactions between DBIL and interesting factors on the primary outcome were tested by the likelihood ratio test with adjustment for the same variables in model 2. Odds ratio and 95% CIs were shown by forest plot There were several limitations in our study. Firstly, the current study is a retrospective study with selection bias conducted in middle-aged and elderly Chinese population, and further research on populations of different ethnic and age is needed to confirm our findings. Secondly, our sample size is not large enough which may have an influence on the results. Extensive large sample studies are needed to explore the underlying mechanisms of bilirubin and poor clinical outcomes in future. Thirdly, HO-1 plays an important role in the effects of bilirubin generation, but we did not measure the levels of HO-1 as well as markers of inflammation and oxidative stress in the population. Last but not the least, three subtypes of bilirubin were only assessed at two time points and the follow-up period of this study was relatively short.

| CON CLUS IONS
We found that increased DBIL pre-thrombolysis had a stronger association with, as well as significantly improved the risk prediction of, poor clinical outcomes of 3-month death and major disability, And we indeed appreciate the data generously provided by Central Hospital of Hefeng County, the Frist People's Hospital of Yichang City.

CO N FLI C T O F I NTE R E S T
The authors declare no conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.