HIF-1α associated logistic regression model serves for predicting decompensation of hepatitis B cirrhosis

HIF-1α is relevant to inammation and brosis in hepatitis B virus (HBV)-related liver diseases. Thus, we designed a predictive model for decompensated cirrhosis. Peripheral plasma HIF-1α levels were measured in 52 subjects, including 20 patients with HBV-related-compensated-cirrhosis (HBV-CC), 20 patients with HBV-related-decompensated-cirrhosis (HBV-DC) that underwent transjugular intrahepatic portosystemic shunt (TIPS), and 12 healthy controls (HC). Portal plasma HIF-1α levels were detected in HBV-DC patients. The correlation between clinical data and HIF-1α levels was assessed, logistic regression and nomogram were used to develop prediction model. aminotransferase-to-Platelet Ratio Index ; FIB-4: Fibrosis index based on the 4 factors; HIF-1α: Hypoxia inducible factor 1α; HVPG: Hepatic venous pressure gradient. PT: Prothrombin time, (11.0~14.0s); ST: Spleen thickness, (3~4cm); PVD: Portal vein diameter, (0.6~1.0cm); Peripheral HIF-1α median: 341.78pg/ml. Platelets; PT: Prothrombin time; INR: International normalized ratio; ST: PVD: vein diameter; MELD:Model for end-stage liver disease; APRI: Aspartate aminotransferase-to-Platelet Ratio Index ; FIB-4: Fibrosis index based on the 4 factors.


Background
Hepatitis B virus (HBV) infection is a widespread chronic infection causing major public health problems worldwide. As many as 40% of men and 15% of women with perinatal HBV infection will reportedly die from liver cirrhosis or hepatocellular carcinoma [1][2][3]. It is widely acknowledged that 90% of patients with liver cirrhosis progress to the decompensated stage with concomitant portal hypertension (PHT), which is the common cause of death in this patient population [4]. PHT is caused by the obstruction of portal blood ow and/or increased circulating blood ow, which leads to increased portal pressure and complications, including gastroesophageal varices, ascites, spontaneous bacterial peritonitis (SBP), and Hepatorenal syndrome (HRS) [5]. The pathogenesis of decompensation in liver cirrhosis is intricate, and patients with early stages of cirrhosis are largely asymptomatic [6]. Currently, clinical detection of hepatic venous pressure gradient (HVPG) is the recognized standard for determining portal vein pressure [7], having an essential value in staging cirrhosis and predicting the occurrence of decompensated complications. Nonetheless, this invasive method for the dynamic monitoring of portal vein pressure has many limitations. Although berendoscopy, color ultrasound, liver stiffness measurement (LSM), computed tomography (CT), and magnetic resonance elastography (MRE) have certain application values for non-invasive diagnosis of liver cirrhosis and assessment of portal hypertension, their accuracy, speci city, and sensitivity are largely in uenced by the experience of the examiner. In addition, according to the incidence of complications such as ascites and upper gastrointestinal bleeding, the diagnosis of decompensated liver cirrhosis can be made clinically; however, at this stage, these patients have already progressed to advanced decompensated cirrhosis associated with increased mortality and poor prognosis [8]. Accordingly, early identi cation and dynamic monitoring of high-risk factors are essential for screening patients at risk of decompensated cirrhosis and providing optimal treatment.
Hypoxia-inducible factor (HIF) is a nuclear transcription factor that plays an active role in hypoxia, adjusts the expression of many functional genes at the transcriptional level, and participates in the maintenance of oxygen balance in cell tissues and the body. It is a heterodimer composed of α subunit (HIF-1α, HIF-2α, or HIF-3α) and HIF-1β subunit. The active HIF-1α subunit degrades rapidly under normal oxygen conditions and remains stable under hypoxic conditions, playing a critical role in regulating oxygen balance in the microenvironment [9][10][11]. Interestingly, the level of HIF-1α and HIF-2α has been documented to be upregulated in chronic hepatitis B, liver cirrhosis, and liver cancer [12][13][14], which may be attributed to the signi cant increase in oxygen consumption of liver cells, in ltration of in ammatory leukocytes, and tissue hypoxia caused by an imbalance in metabolic demand and supply [15]. Vascular dysfunction, thrombosis, or brosis can also lead to a reduced oxygen supply. Our previous researches found that HIF-1α was observably related to in ammation and brosis in HBV-associated liver disease and exhibited good diagnostic value in differentiating the compensatory and decompensated stage of cirrhosis [16]. Therefore, plasma HIF-1α levels and clinical parameters were measured in this study, and a scoring model for predicting decompensation of hepatitis B cirrhosis was established based on the identi ed independent risk factors.

Participants
A total of 40 patients diagnosed with HBV-CC and HBV-DC between August 2020 and June 2021 were enrolled. The criteria for selecting research subjects: (1) Chinese citizens of 18-65 years of age; (2) HBsAg positive or negative, anti-HBcAb-positive, and a clear history of chronic HBV infection (HBsAg positive of > 6 months); (3) ultrasound, CT, or other imaging or liver pathology showing signs of cirrhosis; (4) decreased albumin levels (< 35 g/L) and/or INR > 1.3, or prolonged PT (stopping thrombolytic or anticoagulant drugs for more than seven days, longer than the reference > 3s) and/or platelet count < 100×10 9 /L, and other causes were excluded; (5) hepatitis B cirrhosis was divided into compensated and decompensated stages according to the occurrence of ascites, esophageal and gastric variceal bleeding, hepatic encephalopathy, and other serious complications and/or liver dysfunction, based on the diagnostic criteria in the Chinese guidelines for the management of liver cirrhosis [17] and guidelines for the prevention and treatment of chronic hepatitis B [18]. The criteria for excluding study subjects: (1) patients with liver cancer or combined with other tumors; (2) patients with liver disease complicated by other etiologies (drug liver disease, alcoholic liver disease, and autoimmune liver disease) or superinfection with other hepatitis viruses; (3) patients with severe respiratory, digestive, circulatory, and nervous system diseases; (4) patients with diabetes and thyroid diseases; (5) patients with serious mental and psychological diseases. Twelve healthy subjects with negative serological markers for viral hepatitis and normal liver function were enrolled in the control group during the same period.
Experimental operation (1) The liver biochemical pro le was detected using an Olympus automatic biochemical analyzer (Olympus AU640, Tokyo, Japan). Coagulation assays were performed using a Sta-Compact automatic analyzer (STAGO, France).

Binary logistic regression model
Sixteen risk factors that may affect decompensation of cirrhosis were screened, including sex, age, aspartate transaminase (AST), alanine aminotransferase (ALT), total protein (TP), albumin (ALB), total bilirubin (TB), direct bilirubin (DB), total bile acid (TBA), red blood cell (RBC), hemoglobin (HGB), platelets (PLT), prothrombin time (PT), spleen thickness (ST), portal vein main diameter (PVD), and peripheral plasma HIF-1α concentration. With the occurrence of cirrhosis decompensation as the dependent variable (0= no cirrhosis decompensation, 1= cirrhosis decompensation), the 16 factors included in the univariate analysis were considered as independent variables. If the continuous variable had a linear relationship with the result, the continuous variable was included in the regression formula; otherwise, the continuous variable was converted into a dichotomous or sequential variable [19]. Continuous independent variables were grouped and assigned to the corresponding order (Supplementary Table 1). Binary logistic regression was used to analyze and identify the risk factors that signi cantly affected decompensation of cirrhosis. Finally, a scoring model using a nomogram was established to predict the occurrence of cirrhosis decompensation.
Statistical analyses SPSS 21 and GraphPad 8.0, were used for the statistical analyses. The research data conforming to a normal distribution were showed as mean ± standard deviation, and an independent sample t-test or analysis of variance was used. Data not normally distributed were described by median and range, and a rank-sum test was used. The relationship between continuous variables was examined using the Pearson correlation analysis. Risk factors associated with HBV-DC using binary logistic regression, the odds ratio (OR), and 95% con dence interval (CI) of each factor for the risk of HBV-DC were calculated using multivariate analyses. The WALD stepwise screening method was used for binary logistic regression analysis of independent factors in uencing cirrhosis decompensation [20]. R software (version 4.0.2) was used to construct the nomogram. Received operating characteristic (ROC) curves were compared using MedCalc (version 20.009) module ROC. Statistical signi cance was set at P < 0.05.

Results
Baseline Clinical data of enrolled subjects This research included 12 cases of HC, 20 cases of HBV-CC, and 20 cases of HBV-DC; all patients with HBV-DC presented with symptoms and signs of portal hypertension. No signi cant difference in sex between the HC, CC, and DC groups (P = 0.055), while the mean age of the CC group is markedly lower than that of the DC group (P = 0.007). Liver biochemical parameters, including TB, DB, and TBA, are signi cantly higher in the DC group than in the CC group (P < 0.001). Moreover, the liver brosis index, model for end-stage liver disease (MELD) score, and portal vein main diameter are signi cantly higher in the DC group than in the CC group (P < 0.001, P = 0.010, P = 0.049, respectively) ( Table 1).  Gender, Age, and Peripheral HIF-1α were analyzed by variance analysis while the remaining parameters were analyzed by t-test.

HIF-1α expression in hepatitis B cirrhosis patients and healthy controls
The mean plasma HIF-1α level is signi cantly higher in patients with hepatitis B cirrhosis (CC+DC group) than in healthy controls (t = 2.690, P = 0.0097) (Fig. 1A), while HIF-1α expression in the peripheral plasma of the DC group is signi cantly higher than that in the other groups (P < 0.001) (Fig. 1B). Even though HIF-1α levels in the peripheral plasma of the HBV-DC group are higher than those in the portal blood, the difference is not statistically signi cant (t = 0.5041, P = 0.62) (Fig. 1C). According to the Child-Pugh scoring criteria [17], patients in the compensated and decompensated stages of cirrhosis are divided into class A (n = 23), B (n = 12), and C (n = 5). HIF-1α levels in class A patients are markedly lower than those in class B and C patients (P < 0.0001) ( Fig. 2A). Peripheral plasma HIF-1α levels are signi cantly higher in cirrhotic patients with ascites than in those without ascites (t = 4.178, P = 0.0002). However, there is no signi cant difference in peripheral plasma HIF-1α levels between patients with and without GI bleeding (t = 0.0394, P = 0.9688). Plasma HIF-1α levels in patients with both serious complications are signi cantly higher than those in patients without complications (t = 3.409, P = 0.0023) (Fig. 2B). Interestingly, in the HBV-DC group, peripheral plasma HIF-1α levels are negatively correlated with preoperative HVPG (r = -0.2005, P = 0.3968) (Supplementary Fig. 1A). The expression level of HIF-1α in the peripheral blood of patients with hepatitis B cirrhosis is positively correlated with age (r = 0.446, P = 0.0039) ( Supplementary  Fig. 1B). No signi cant difference in HIF-1α expression is found between the sexes in patients with hepatitis B cirrhosis (P = 0.4872) (Supplementary Fig. 3A). Furthermore, peripheral plasma HIF-1α levels in the healthy control group are not markedly correlated with age (r = -0.333, P = 0.290) ( Supplementary  Fig. 1C), and gender strati cation also shows no difference (P = 0.5932) (Supplementary Fig. 3B).
The above ndings suggest that the peripheral plasma HIF-1α levels of hepatitis B cirrhosis patients are closely related to the biochemical liver parameters, bile acid metabolism, cirrhosis grade, and progression of decompensated cirrhosis (Fig. 3A,B).

Risk factors for HBV-related decompensated cirrhosis
Univariate and multivariate logistic regression were performed to determine the signi cant risk factors for HBV-DC. Univariate logistic regression show that TB, PLT, PT, spleen thickness, and plasma HIF-1α concentration are signi cant risk factors (P < 0.05) (  (HIF-1α). However, our small sample size did not meet the requirements of the event per variable (EPV); accordingly, the ndings of our study were not robust enough. However, to our knowledge, few studies have successfully designed models to predict decompensation in HBV-induced cirrhosis.

Nomogram development
We validated that HIF-1α could independently affect the progression to decompensated cirrhosis. Subsequently, these three factors are used to draw a nomogram to construct a scoring model for HBV-DC ( Fig. 4).

Application of the scoring model in evaluating HBV-DC
We compared the application value of plasma HIF-1α level, scoring model, APRI, and FIB-4 score to assess the probability of HBV-DC. ROC curve analysis revealed that the optimal cut-off value for the probability of HBV-DC is > 45%, area under the curve was 0.954 (P < 0.001), with 95% sensitivity and speci city. In predicting the progression of HBV-DC, the scoring model achieves better speci city and sensitivity than the plasma HIF-1α level, APRI, and FIB-4 scores (Fig. 5).

Comparison of ROC curve between scoring model and clinical non-invasive score
No noteworthy difference is observed between the scoring model, APRI, and FIB-4 scores in predicting HBV-DC. However, there is a signi cant difference between the scoring model and plasma HIF-1α level alone in predicting decompensated cirrhosis (P = 0.0356) (Supplementary Table 2).

Discussion
HIF-1α has been documented to maintain energy metabolism in myeloid cells under hypoxic conditions, promoting them toward in ammatory tissues [21] and play an anti-in ammatory and regulatory role in innate immunity. However, it is widely acknowledged that uctuations in transaminase levels are highly predominant in HBV-induced cirrhosis patients who take antiviral and hepatoprotective drugs to lower transaminase levels. HIF-1α may adjust the expression of many target genes encoding metabolic enzymes and improve cell metabolic adaptation to hypoxia [22], suggesting that HIF-1α is involved in the metabolic functioning of the liver, which was veri ed by our statistical biochemical results. In our study, HIF-1α levels were positively correlated with the APRI, FIB-4, and MELD scores. HIF-1α can regulate the production of pro-brotic mediators to promote the development of liver brosis directly [23]. It has been found that hypoxia induces a series of angiogenic factors such as VEGF and PDGF-β to promote angiogenesis in primary human macrophages. PDGF-β can also promote the proliferation of hepatic stellate cells (HSCs),and continue to differentiate into myo broblasts, which produce large amounts of collagen, leading to brosis and cirrhosis [24][25][26]. When cirrhosis continues to develop, it results in diffuse necrosis and regeneration of liver cells, destruction of the hepatic lobule structure, and formation of pseudolobules, whereas sinusoidal occlusion or perisinusal brosis causes intrahepatic vascular blockage and blood ow obstruction, consequently increasing portal pressure gradually [27]. In the development of portal hypertension, vasoconstriction causes tissue hypoxia, which leads to the phosphorylation of the NADPH oxidase subunit and promotes the oxidation-reducing coenzyme II (NADPH), which induces HIF-1α upregulation in tissues [28]. In the presence of HIF-1α, placental growth factor (PLGF) stimulates the growth and migration of endothelial cells and participates in the generation of pathological blood vessels, while liver sinusoidal endothelial cells (LSECs), HSCs, and Kupffer cells promote the generation and reconstruction of pathological blood vessels through direct or indirect pathways under hypoxic stimulation and injury [29]. In the process of ber repair, these pathological microvessels, originating from the branches of blood vessels in the liver, bypass block the sinusoidal blood supply area, further exacerbating tissue hypoxia [30]. There were no distinct differences in plasma HIF-1α levels between Child-Pugh class B and C patients, and preoperative HVPG were negatively correlated with HIF-1α levels. We speculate that HIF-1α may be a sensitive biomarker in the early stages of cirrhosis. However, the progression of cirrhosis is accompanied by collateral circulation formation and visceral and systemic vasodilation. High extrahepatic blood circulation may in uence the peripheral plasma HIF-1α levels.
Moreover, HIF-1α is a transcription factor activated under hypoxia and regulates many genes involved in cell responses to hypoxia and other tissue environmental signals [31]. Therefore, HIF-1α is inhibited under aerobic conditions, and any increase in HIF-1α expression is suggestive of hypoxia. Furthermore, liver cirrhosis patients often suffer from malnutrition, decreased liver synthesis, and hypersplenism, resulting in decreased hematopoiesis and increased hemolysis, thus substantiating our ndings.
Regarding the prediction model, Bureau et al. [32] found that low bilirubin levels (< 50 umol/L) and high platelet count (> 75×10 9 /L) predicted outcome in refractory ascites patients. In addition, the albuminbilirubin score (ALBI score) provided a better evaluation of the severity and long-term survival of HBVinduced cirrhosis patients than the MELD and MELD-NA scores [33]. Splenomegaly is a sensitive indicator of portal hypertension with poor speci city. However, combining splenomegaly with noninvasive diagnostic indicators such as liver hardness test and platelet count can be used to detect esophageal varicose veins in patients with hepatitis B cirrhosis [34]. Spleen size can be quanti ed by measuring the longitudinal, anteroposterior, and transverse diameters; however, few studies have used spleen thickness [35]. Interestingly, we found that spleen thickness was an independent factor that in uenced the progression of HBV-DC.
Furthermore, our scoring model exhibited better predictive ability, speci city, and sensitivity than other non-invasive, highly reliable metrics for liver brosis [36], including APRI [37] and FIB-4 [38]. Although there was no noteworthy difference during the non-inferiority test, APRI and FIB-4 scores were related to the degree of brosis in untreated HBV patients [39]. Indeed, uctuations in transaminase levels are often observed in patients with cirrhosis who receive clinical treatment. Progression of liver brosis can lead to hypersplenism and further thrombocytopenia [40]. However, the APRI and FIB-4 scores are relatively inaccurate metrics. FIB-4 has been reported to have greater accuracy in excluding signi cant hepatitis brosis than in diagnosing severe hepatitis B brosis [41]. Therefore, the scoring model constructed in this study can reduce the interference of the clinical use of enzyme-lowering drugs on the prediction of HBV-DC and has better application value.

Conclusions
Plasma HIF-1α levels were correlated with liver biochemical function, bile acid metabolism, liver cirrhosis stage, and decompensation progression. The logistic regression model constructed from total bilirubin, spleen thickness, and HIF-1α has a high predictive value for HBV-DC and thus offers signi cant bene ts such as dynamic non-invasive monitoring. Considering the small number of cases included in this singlecenter retrospective clinical study, prospective, large-sample, multi-center researches are needed to substantiate the application of our scoring model in predicting decompensation in patients with HBVinduced cirrhosis. Declarations increased in hepatitis B cirrhosis (CC+DC group) patients than that in healthy controls. (B) The mean plasma HIF-1α levels are signi cantly increased in HBV-DC patients than that in HBV-CC patients and healthy controls (P = 0.0004, P < 0.001). (C) Differences in plasma HIF-1α expression between peripheral and portal blood in patients with decompensated liver cirrhosis (t = 0.5041, P = 0.62).*p < 0.05, **p < 0.01. Comparison of the mean plasma HIF-1α level (mean ± SEM) in Hepatitis B liver cirrhosis with different severity: with ascites vs. without patients, with gastrointestinal bleeding vs. without patients, with ascites and gastrointestinal bleeding vs. without patients.*p < 0.05, **p < 0.01,***p 0.001.  Nomogram for predicting clinical outcomes in patients with hepatitis B cirrhosis.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.