Wang ZB, Zhu B, Meng MM, Wu YF, Zhang Y, Li DZ, Tian H, Wang FC, Lv YF, Ye QX, Liu FQ. Nomogram for predicting survival after transjugular intrahepatic portosystemic shunt in portal hypertension patients with bleeding. World J Gastrointest Surg 2025; 17(4): 104884 [DOI: 10.4240/wjgs.v17.i4.104884]
Corresponding Author of This Article
Fu-Quan Liu, Chief Physician, Professor, Center for Minimally Invasive Treatment of Liver Diseases, Beijing Shijitan Hospital, Capital Medical University, No. 10 Tieyi Road, Yangfangdian, Haidian District, Beijing 100080, China. lfuquan@aliyun.com
Research Domain of This Article
Gastroenterology & Hepatology
Article-Type of This Article
Retrospective Cohort Study
Open-Access Policy of This Article
This article is an open-access article which was selected by an in-house editor and fully peer-reviewed by external reviewers. It is distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/
Zhi-Bin Wang, Ming-Ming Meng, Yi-Fan Wu, Yu Zhang, Yi-Fan Lv, Qiu-Xia Ye, Fu-Quan Liu, Center for Minimally Invasive Treatment of Liver Diseases, Beijing Shijitan Hospital, Capital Medical University, Beijing 100080, China
Bing Zhu, Dong-Ze Li, Hua Tian, Fu-Chuan Wang, Fu-Quan Liu, Center for Diagnosis and Treatment of Hepatic Vascular Diseases, Fifth Medical Center of Chinese PLA General Hospital, Beijing 100071, China
Author contributions: Wang ZB, Zhu B, Meng MM, and Wu YF collected the data; Zhang Y, Li DZ, and Tian H performed the data analysis; Wang FC, Lv YF, and Ye QX contributed to data interpretation; Wang ZB and Liu FQ drafted the manuscript; Liu FQ supervised the study and revised the manuscript; Wang ZB, Zhu B, Meng MM, Wu YF, Zhang Y, Li DZ, Tian H, Wang FC, Lv YF, Ye QX, and Liu FQ designed the research study; and all authors have read and approved the final version of the manuscript.
Supported by the “14th Five-Year” Talent Training Program of Beijing Shijitan Hospital, Capital Medical University, No. 2023 LJRCLFQ.
Institutional review board statement: This study was approved by the Medical Ethics Committee of the Fifth Medical Center of the General Hospital of the People’s Liberation Army, approval No. KY-2023-12-83-1.
Informed consent statement: Since this research project was conducted on the basis of patients’ routine treatment surgeries, without any additional procedures and without causing harm to the patients’ bodies, we adopted the waiver of informed consent in the clinical setting.
Conflict-of-interest statement: All the authors report no relevant conflicts of interest for this article.
STROBE statement: The authors have read the STROBE Statement-checklist of items, and the manuscript was prepared and revised according to the STROBE Statement-checklist of items.
Data sharing statement: The dataset and statistical code are available from the corresponding author, Fu-Quan Liu, at lfuquan@aliyun.com upon reasonable request. Consent for data sharing was not obtained, but the presented data are anonymized, and the risk of identification is low.
Open Access: This article is an open-access article that was selected by an in-house editor and fully peer-reviewed by external reviewers. It is distributed in accordance with the Creative Commons Attribution NonCommercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: https://creativecommons.org/Licenses/by-nc/4.0/
Corresponding author: Fu-Quan Liu, Chief Physician, Professor, Center for Minimally Invasive Treatment of Liver Diseases, Beijing Shijitan Hospital, Capital Medical University, No. 10 Tieyi Road, Yangfangdian, Haidian District, Beijing 100080, China. lfuquan@aliyun.com
Received: January 5, 2025 Revised: February 17, 2025 Accepted: March 10, 2025 Published online: April 27, 2025 Processing time: 83 Days and 16.2 Hours
Abstract
BACKGROUND
Portal hypertension (PHT) is a life-threatening complication of cirrhosis, often resulting in gastrointestinal bleeding that requires transjugular intrahepatic portosystemic shunt (TIPS). While TIPS effectively reduces portal pressure, predicting long-term survival remains challenging due to the multifactorial nature of patient outcomes. Accurate survival prediction tools are lacking, and existing models often omit critical factors such as portal vein diameter. This study aimed to develop and validate a nomogram incorporating key clinical and biochemical variables to predict 1-year and 2-year survival following TIPS in PHT patients. We hypothesized that this model would provide improved risk stratification and guide clinical decisions.
AIM
To develop and validate a nomogram for predicting 1-year and 2-year survival in PHT patients post-TIPS.
METHODS
This retrospective cohort study included 848 TIPS-treated PHT patients with gastrointestinal bleeding from two tertiary hospitals (2013-2021). Mortality was the primary endpoint. Predictive variables were selected using least absolute shrinkage and selection operator regression, and a nomogram was developed with Cox regression to predict 1-year and 2-year survival. Model performance was evaluated through receiver operating characteristic curves, calibration plots, and decision curve analysis.
RESULTS
The mean age of the included (848) patients was 53.00 years ± 12.51, where 69.58% were men. Results showed that portal vein diameter, serum creatinine, potassium, and alpha-fetoprotein were the independent predictors of post-TIPS survival. Besides, the model showed strong discriminatory ability (C-index, 0.816 in the training set; 0.827 in the validation set) and good calibration. The area under the curve for 1-year and 2-year survival in the training set were 0.890 [95% confidence interval (CI): 0.802-0.948] and 0.838 (95%CI: 0.803-0.869), respectively. The area under the curve for 1-year and 2-year survival in the validation set were 0.934 (95%CI: 0.815-0.987) and 0.864 (95%CI: 0.811-0.907), respectively.
CONCLUSION
The developed nomogram could reliably predict 1-year and 2-year survival in patients undergoing TIPS for PHT-induced gastrointestinal bleeding.
Core Tip: This study developed and validated a novel nomogram to predict 1-year and 2-year survival in patients undergoing transjugular intrahepatic portosystemic shunt for portal hypertension-induced gastrointestinal bleeding. Key prognostic factors, including portal vein diameter, serum creatinine, potassium, and alpha-fetoprotein, were identified using least absolute shrinkage and selection operator and Cox regression analyses. The model demonstrated excellent discrimination, calibration, and clinical utility across training and validation cohorts. This tool provides an innovative approach to risk stratification and personalized prognosis, offering valuable guidance for clinical decision-making and long-term management of patients receiving transjugular intrahepatic portosystemic shunt.
Citation: Wang ZB, Zhu B, Meng MM, Wu YF, Zhang Y, Li DZ, Tian H, Wang FC, Lv YF, Ye QX, Liu FQ. Nomogram for predicting survival after transjugular intrahepatic portosystemic shunt in portal hypertension patients with bleeding. World J Gastrointest Surg 2025; 17(4): 104884
Hepatocellular carcinoma (HCC) is the sixth most common cancer worldwide. Notably, the incidence of HCC is highest in East Asia and sub-Saharan Africa, primarily due to viral hepatitis, alcohol abuse, and obesity. The global burden of HCC may increase by over 50% by 2040, mainly due to non-alcoholic steatohepatitis[1-5]. Cirrhosis-related deaths accounted for 2.4% of all global deaths in 2017[6]. Severe portal hypertension (PHT) in decompensated cirrhosis increases the risk of variceal rupture, with mortality rates of 15%-30%[7-11]. Hepatic sinusoidal endothelial dysfunction, stellate cell activation, abnormal vascular responses, and microvascular thrombosis lead to systemic circulatory expansion and exacerbate complications of PHT[12-14].
Transjugular intrahepatic portosystemic shunt (TIPS) is widely used to manage PHT and its complications by creating a shunt between the portal and hepatic veins, bypassing the fibrotic liver region. A self-expanding metal stent maintains the shunt's patency, allowing direct flow of portal venous blood into the systemic circulation, thereby reducing the pressure gradient and restoring normal blood flow[15-19]. Microvascular thrombosis exacerbates sinusoidal obstruction and liver fibrosis, increasing the risk of complications after TIPS[14]. Long-term survival after TIPS depends on liver function, the presence of HCC, and the severity of PHT[20]. Early TIPS reduces rebleeding and improves survival in high-risk cirrhotic patients, indicating that regular monitoring of portal pressure and liver function, as well as timely interventions, are essential for successful PHT treatment[21,22].
Nomograms have become popular for predicting patient prognosis in recent years by integrating clinical and biochemical variables, providing an intuitive method to estimate survival probabilities[23,24]. However, many existing models, often based on retrospective data, may lack key variables like portal vein diameter and changes in portal pressure, which are critical for long-term outcomes[10,20]. Therefore, more comprehensive and precise nomograms are needed to predict the long-term survival of patients undergoing TIPS.
This study aimed to develop and validate a nomogram for predicting 1-year and 2-year survival in patients undergoing TIPS. Key prognostic variables were identified using least absolute shrinkage and selection operator (LASSO) regression, then further analyzed with a Cox regression model[23]. The model was validated using a moderate-sized patient cohort. Therefore, the study may provide accurate risk stratification and prognosis assessment options, optimizing treatment strategies and improving long-term survival[20,24].
MATERIALS AND METHODS
Sample size determination
A total of 848 patients were included in this study. The sample size was calculated to ensure 80% statistical power at a 0.05 significance level, based on expected effect sizes from prior studies[7,10,12]. To account for potential data loss due to incomplete medical records or patient follow-up, an additional 10% margin was incorporated into the calculation, ensuring the reliability and robustness of the results.
Ethical statement
This retrospective study was approved by the Ethics Committee of Beijing Shijitan Hospital, approval No. 2018 (1), and the Ethics Committee of the Fifth Medical Center of the General Hospital of the People’s Liberation Army, approval No. KY-2023-12-83-1. Since this study was based on patients’ routine treatment procedures without additional interventions or harm, informed consent was waived in the clinical setting. The study adhered to the Declaration of Helsinki and relevant ethical guidelines, ensuring patient privacy and compliance with scientific and ethical standards.
Patients
This study was conducted at the Liver Disease Minimally Invasive Treatment Center in Beijing Shijitan Hospital, Capital Medical University, and the Center for Diagnosis and Treatment of Hepatic Vascular Diseases in the Fifth Medical Center of Chinese PLA General Hospital. Only patients who underwent TIPS for PHT-related gastrointestinal bleeding between January 2013 and June 2021 and with follow-up completed by June 2023 were included in the study. Patients with recurrent esophageal variceal bleeding, refractory ascites, Budd-Chiari syndrome, or Child-Turcotte-Pugh B or C classification without severe hepatic encephalopathy underwent TIPS. Contraindications included severe liver failure and uncontrolled hepatic encephalopathy[22,25]. Finally, 107 of 955 screened patients were excluded. Exclusion criteria included: Patients with non-HCC malignancies (n = 17), missing key clinical data (n = 18), severe cardiac, renal dysfunction (n = 34), and those undergoing repeat TIPS (n = 26) or liver transplantation postoperatively (n = 12). The remaining 848 patients were randomly assigned to a training set (n = 593) and a validation set (n = 255).
Surgical procedure
The TIPS procedures were performed by experienced physicians under fluoroscopic guidance via a transjugular approach. First, a catheter was inserted into the inferior vena cava and hepatic vein, followed by venography to define the anatomical characteristics and measure pressures. The portal vein was then accessed by puncturing the hepatic segment of the inferior vena cava or the hepatic vein, and venography was performed to measure portal pressure. A stent was placed between the portal vein and the inferior vena cava after dilating the liver parenchymal tract. The portal pressure was measured, and the pressure gradient between the portal vein and inferior vena cava was calculated[12,26,27].
Data collection
The baseline demographic information, clinical parameters before and after surgery, imaging findings, and laboratory results were collected from the hospital's electronic medical records for quality control. Key variables included portal vein diameter, creatinine, potassium, and alpha-fetoprotein (AFP)[28]. Portal vein diameter was measured through imaging, while creatinine, potassium, and AFP levels were determined via blood tests. The measurements were taken within one week before TIPS procedure. Data collection and analysis were conducted by independent reviewers blinded to patient group assignments following standardized procedures to maintain accuracy and consistency, thus ensuring unbiased predictions and reducing potential bias.
Follow-up
The patients were followed for two years to track survival and postoperative complications. Follow-up data were collected through outpatient visits, telephone interviews, and electronic medical record system. The primary endpoint was all-cause mortality, defined as death from any cause during the follow-up period.
Statistical analysis
R software (version 4.4.1; https://www.R-project.org) and MedCalc software (version 22.001; https://www.medcalc.org) were used for statistical analyses. There were no missing data in this study. Quantitative data were presented as median (interquartile range), and comparisons between groups were performed using the independent sample t-test or Mann-Whitney U test, depending on data distribution. Categorical data were expressed as frequencies and percentages, and comparisons were made using the χ2 test or Fisher’s exact test. Variables, such as portal vein diameter, creatinine, potassium, and AFP, were log-transformed before analysis to improve normality and model robustness. Key survival-related variables were identified using LASSO regression. A nomogram predicting 1-year and 2-year survival probabilities was built using Cox regression analysis. Model performance was evaluated using receiver operating characteristic curves and area under the curve (AUC), with calibration assessed by the Hosmer-Lemeshow test. Net benefit at various thresholds was assessed using decision curve analysis to confirm the model’s applicability across clinical decision-making scenarios. Survival between high-risk and low-risk groups was compared using Kaplan-Meier survival analysis. Patients lost to follow-up were censored at their last recorded follow-up date in survival analysis. A two-sided P value < 0.05 was considered statistically significant.
RESULTS
Patient characteristics
A total of 848 patients with PHT-induced gastrointestinal bleeding who underwent TIPS were included in the study. The patients were randomly assigned to the training set (593 patients) and validation set (255 patients) in a 7:3 ratio. The mean follow-up duration was 20.85 ± 6.82 months. During follow-up, 59 patients (6.96%) were lost to follow-up due to uncontactable status or withdrawal from the study. By the end of the first year, 80 patients (9.43%) had died, and by the end of the second year, the total number of deaths had reached 144 (16.98%). In the training set, 101 patients died, while 43 deaths occurred in the validation set (survival rates: 82.97% and 83.14%, respectively; P value = 0.90, Figure 1). The primary causes of death were liver failure, HCC, hepatic encephalopathy, and rebleeding.
Figure 1 Flowchart of this study.
TIPS: Transjugular intrahepatic portosystemic shunt; HCC: Hepatocellular carcinoma.
Compared with survivors, non-survivors had significantly higher preoperative portal vein pressure, pressure gradient, and right hepatic vein wedge pressure, with reduced portal vein diameter. Furthermore, creatinine and AFP levels of non-survivors were elevated, while potassium levels were reduced (Table 1). In both the training and validation cohorts, non-survivors exhibited distinct baseline characteristics compared to survivors. In the training cohort, they had a higher pre-TIPS portal pressure gradient (P value = 0.031) and a smaller portal vein diameter (P value < 0.001), along with elevated serum creatinine (P value < 0.001), lower potassium (P value < 0.001), and significantly higher AFP levels (P value < 0.001). Similarly, in the validation cohort, non-survivors had higher pre-TIPS portal venous pressure (P value = 0.034), smaller portal vein diameter (P value < 0.001), increased serum creatinine (P value < 0.001), lower potassium (P value = 0.022), and decreased albumin levels (P value = 0.012). Additionally, AFP (P value < 0.001) and cancer antigen 125 (P value = 0.049) were significantly elevated in non-survivors (Tables 2 and 3). Notably, the baseline characteristics, including age, gender, Child-Turcotte-Pugh classification, and portal vein pressure, were not significantly different between the training and validation sets. Biochemical indices were also consistent in both sets (Table 4).
Table 1 Baseline demographic and clinical characteristics of patients stratified by survival status, n (%).
Table 4 Baseline demographic and clinical characteristics of the training and validation cohorts, n (%).
Characteristic
Overall (n = 848)
Training cohort (n = 593)
Validation cohort (n = 255)
P value
Age (year)
53.00 (45.00-61.00)
53.00 (46.00-61.00)
51.00 (42.00-62.00)
0.078
Sex
Male
590 (69.58)
406 (68.47)
184 (72.16)
0.284
Female
258 (30.42)
187 (31.53)
71 (27.84)
-
Pre-TIPS Child Pugh class
A
417 (49.17)
292 (49.24)
125 (49.02)
0.969
B
372 (43.87)
259 (43.68)
113 (44.31)
-
C
59 (6.96)
42 (7.08)
17 (6.67)
-
Etiology
Alcohol
97 (11.44)
61 (10.29)
36 (14.12)
0.183
Autoimmune liver disease
46 (5.42)
37 (6.24)
9 (3.53)
-
Cryptogenic
252 (29.72)
180 (30.35)
72 (28.24)
-
Drug-induced liver injury
18 (2.12)
15 (2.53)
3 (1.18)
-
Hepatitis B virus
391 (46.11)
267 (45.03)
124 (48.63)
-
Hepatitis C virus
44 (5.19)
33 (5.56)
11 (4.31)
-
Pre-TIPS portal venous pressure (mmHg)
33.04 (29.00-37.55)
33.80 (29.29-37.55)
33.04 (29.00-38.00)
0.632
Post-TIPS portal venous pressure (mmHg)
22.00 (18.00-26.00)
22.00 (18.00-26.00)
22.00 (17.27-26.00)
0.774
Pre-TIPS inferior vena cava pressure (mmHg)
8.00 (6.00-10.00)
8.00 (6.00-10.00)
8.00 (6.00-10.00)
0.998
Post-TIPS inferior vena cava pressure (mmHg)
9.00 (8.00-12.00)
9.00 (8.00-11.00)
10.00 (8.00-12.00)
0.212
Pre-TIPS portal pressure gradient (mmHg)
25.47 (21.03-29.79)
25.30 (21.00-30.00)
26.00 (21.66-29.00)
0.991
Post-TIPS porta pressure gradient (mmHg)
11.72 (8.00-15.78)
11.90 (8.00-15.91)
11.02 (7.00-15.53)
0.311
Portal vein diameter (mm)
12.00 (10.20-15.82)
12.00 (10.30-15.80)
12.70 (10.20-15.85)
0.683
Right hepatic vein wedge pressure (mmHg)
27.00 (21.00-32.00)
27.00 (21.00-32.00)
27.00 (23.00-31.77)
0.637
Right hepatic vein free pressure (mmHg)
10.75 (9.00-12.00)
10.75 (9.00-12.00)
10.75 (8.00-12.00)
0.706
Pre-TIPS right atrial pressure (mmHg)
6.13 (5.00-6.13)
6.13 (5.00-6.13)
6.13 (4.25-7.00)
0.911
Pre-TIPS Child-Pugh score
7.00 (6.00-8.00)
7.00 (5.00-8.00)
7.00 (6.00-8.00)
0.600
Pre-TIPS MELD score
59.96 (56.61-63.00)
60.10 (56.68-63.25)
59.59 (56.59-62.66)
0.376
Total bilirubin (μmol/L)
22.65 (15.90-34.80)
22.70 (16.00-35.20)
22.10 (15.90-34.30)
0.882
Direct bilirubin (μmol/L)
9.90 (6.70-15.93)
9.80 (6.50-16.30)
10.50 (6.95-15.30)
0.424
Indirect bilirubin (μmol/L)
12.80 (8.98-19.62)
13.20 (9.00-19.40)
12.20 (8.95-19.82)
0.371
Creatinine (μmol/L)
62.00 (52.00-74.00)
62.00 (52.00-74.00)
62.00 (51.50-73.00)
0.643
Alanine aminotransferase (U/L)
20.00 (14.00-29.00)
19.00 (14.00-27.69)
21.00 (13.50-29.00)
0.622
Aspartate aminotransferase (U/L)
28.00 (21.00-38.00)
28.00 (21.00-37.00)
27.00 (21.00-41.50)
0.898
Kalium (mmol/L)
3.94 (3.62-4.22)
3.94 (3.61-4.24)
3.93 (3.63-4.16)
0.585
Natrium (mmol/L)
140.00 (138.00-142.00)
140.00 (138.00-142.00)
140.00 (138.00-142.00)
0.983
Chlorine (mmol/L)
107.00 (104.00-109.00)
107.00 (104.00-109.00)
107.00 (104.00-109.00)
0.975
Alkaline phosphatase (U/L)
88.00 (66.00-123.00)
86.00 (66.00-121.00)
90.00 (67.00-136.50)
0.188
Glutamyl transferase (U/L)
38.00 (20.00-73.00)
37.00 (20.00-71.00)
41.00 (19.00-78.50)
0.533
Serum urea nitrogen (mmol/L)
4.91 (3.99-6.41)
4.97 (4.04-6.44)
4.78 (3.90-6.29)
0.326
Total protein (g/L)
66.10 (61.71-70.40)
66.03 (61.30-70.00)
66.47 (62.41-71.65)
0.121
Albumin (g/L)
35.73 (32.10-39.12)
35.80 (32.30-39.30)
35.50 (32.00-38.80)
0.297
International normalized ratio
1.27 (1.15-1.44)
1.27 (1.15-1.42)
1.30 (1.16-1.44)
0.280
Activated partial thromboplastin time (s)
33.40 (31.00-36.83)
33.50 (30.90-37.10)
33.20 (31.20-36.00)
0.554
Prothrombin time (s)
14.00 (12.70-15.72)
13.90 (12.70-15.50)
14.40 (12.80-16.00)
0.146
Prothrombin time activity (%)
66.00 (56.00-77.00)
67.00 (57.00-77.00)
65.00 (56.00-76.00)
0.334
Fibrinogen (g/L)
1.99 (1.57-2.50)
1.98 (1.56-2.48)
2.01 (1.62-2.54)
0.396
Blood ammonia (μmol/L)
42.64 (32.40-59.50)
42.30 (32.20-58.65)
43.40 (32.95-62.85)
0.360
White blood cell count (× 109/L)
2.83 (1.90-4.19)
2.88 (1.90-4.14)
2.72 (1.86-4.23)
0.698
Hemoglobin (× 1012/L)
86.50 (74.00-104.00)
87.00 (74.00-104.00)
86.00 (75.00-104.00)
0.385
Platelet count (× 109/L)
76.00 (50.00-121.00)
76.00 (50.00-116.00)
76.00 (52.00-123.50)
0.287
Alpha fetoprotein (ng/mL)
220.56 (2.85-423.31)
233.19 (2.88-413.82)
156.85 (2.66-428.81)
0.465
Carcinoembryonic antigen (ng/mL)
2.44 (1.62-3.61)
2.40 (1.63-3.54)
2.64 (1.60-3.75)
0.378
Carbohydrate antigen 125 (U/mL)
125.24 (36.20-233.29)
137.25 (40.10-242.69)
100.70 (33.20-204.23)
0.057
Cancer antigen 15-3 (U/mL)
10.30 (7.50-14.20)
10.30 (7.50-14.10)
10.40 (7.10-14.65)
0.604
Glycosylated chain antigen CA199 (U/mL)
10.94 (4.43-15.26)
11.47 (4.81-15.26)
9.26 (4.07-15.26)
0.179
Identification of predictive factors
LASSO regression was first applied to the training set using the glmnet package (α = 1, Nlambda = 100) to select key prognostic variables while minimizing overfitting. 10-fold cross-validation determined the optimal lambda (λ = 0.028), yielding four significant predictors: Portal vein diameter (β = -1.256), creatinine (β = 0.613), potassium (β = -2.316), and AFP (β = 0.261). These LASSO-identified variables were further subjected to univariate Cox regression, confirming their significant association with survival (P value < 0.05). Multivariate Cox analysis validated their independent prognostic value: Portal vein diameter [hazard ratio (HR) = 0.195, P value < 0.001] and potassium (HR = 0.030, P value < 0.001) were protective factors, whereas creatinine (HR = 1.947, P value = 0.019) and AFP (HR = 1.455, P value < 0.001) were associated with increased mortality risk (Figure 2A, Table 5).
Figure 2 Prognostic model predicting the 1-year and 2-year survival following transjugular intrahepatic portosystemic shunt.
A: Forest plot of multivariable Cox regression analysis showing hazard ratios with 95% confidence interval; B: Nomogram predicting the 1-year and 2-year survival probabilities; C: Nomogram with a patient case example, illustrating predicted mortality probabilities at 1 year and 2 years. HR: Hazard ratio; CI: Confidence interval; AFP: Alpha-fetoprotein.
Table 5 Univariate and multivariate cox regression analyses of factors associated with mortality.
In this study, the predictive model was constructed exclusively using the training set and subsequently validated in an independent validation set. The developed nomogram is shown in Figure 2B. The nomogram included potassium, creatinine, AFP, and portal vein diameter and estimated patient survival probability by summing the scores of these variables. Cox regression analysis indicated that potassium (β = -3.508, P value < 0.001) and portal vein diameter (β = -1.635,P value < 0.001) were protective factors, while AFP (β = 0.375, P value < 0.001) and creatinine (β = 0.667, P value = 0.019) were associated with worse prognosis. For instance, a post-TIPS patient with a portal vein diameter of 12 mm, creatinine of 50 μmol/L, potassium of 4.0 mmol/L, and AFP of 500 ng/mL had an estimated 1-year and 2-year mortality rates of 9.5% and 19.5%, respectively (Figure 2C).
Discrimination
The discrimination ability of the model was evaluated based on several metrics, such as the likelihood ratio χ² test, C-index, and AUC. The likelihood ratio χ² value was 146.17 (P value < 0.01), yielding a C-index of 0.816 [95% confidence interval (CI): 0.777-0.855] in the training set and 0.827 (95%CI: 0.771-0.883) in the validation set. The AUC for 1-year and 2-year survival in the training set were 0.890 (95%CI: 0.802-0.948) and 0.838 (95%CI: 0.803-0.869), respectively. The AUC for 1-year and 2-year survival in the validation set were 0.934 (95%CI: 0.815-0.987) and 0.864 (95%CI: 0.811-0.907), respectively. These results demonstrate strong discrimination ability at different time points (Figure 3).
Figure 3 Receiver operating characteristic curves indicating the area under the curve for the nomogram model.
A: Training set; B: Validation set. AUC: Area under the curve.
Calibration of the predictive model
The calibration of the nomogram model in the training and validation sets is shown in Figure 4. The calibration curves in both sets showed strong alignment between predicted and observed survival rates. Similarly, Hosmer-Lemeshow test results indicated good calibration in both sets, with P value of 1.
Figure 4 Calibration curves of the nomogram for predicting the 1-year and 2-year survival.
A and B: Calibration curves of the nomogram for predicting the 1-year and 2-year survival in the training set; C and D: Calibration curves of the nomogram for predicting the 1-year and 2-year survival in the validation set.
Clinical use
Decision curve analysis of the nomogram in both the training and validation sets, assessing net benefit across different risk thresholds, is shown in Figure 5. The optimal probability ranges for 1-year and 2-year survival prediction in the training set were 2%-49%, and 3%-99%, respectively: And 2%-56% and 3%-97%, respectively, in the validation set. The nomogram yielded the greatest net benefit within these ranges.
Figure 5 Decision curve analysis for the nomogram predicting the mortality.
A: Decision curve analysis for the nomogram predicting the mortality in the training set; B: Decision curve analysis for the nomogram predicting the mortality in the validation set. DCA: Decision curve analysis.
Risk stratification
The patients in the training and validation sets were stratified into high-risk and low-risk groups based on the optimal cut-off value of the risk score (1.625). Kaplan-Meier analysis showed that the high-risk groups had significantly lower survival rates than the low-risk groups (P value < 0.0001, Figure 6).
Figure 6 Kaplan-Meier survival curves comparing overall survival between high-risk and low-risk groups.
A: Kaplan-Meier survival curves comparing overall survival between high-risk and low-risk groups in the training cohort; B: Kaplan-Meier survival curves comparing overall survival between high-risk and low-risk groups in the validation cohort.
DISCUSSION
In this study, a nomogram was developed and validated for the prediction of 1-year and 2-year survival in patients with PHT-induced gastrointestinal bleeding who underwent TIPS. Portal vein diameter, creatinine, potassium, and AFP were identified as key factors influencing long-term survival. Besides, these variables demonstrated significant independent prognostic value. The nomogram effectively predicted postoperative survival, showing strong discrimination ability and good calibration in the training and validation sets. Kaplan-Meier analysis further confirmed that the survival was significantly different between the high-risk and low-risk groups. PHT management strategies influence the model application and long-term prognosis. Notably, TIPS should be conducted cautiously for patients with both HCC and PHT, balancing portal pressure reduction with the risk of liver function deterioration, underscoring the complexity of PHT management[8,13,16]. Although emerging therapies such as selective β2-adrenergic receptor agonists and multikinase inhibitors have shown promise in reducing PHT, this study did not include patients receiving these treatments, as they were not widely adopted during the study period. Future research incorporating these novel therapies into predictive models could further refine risk stratification and optimize clinical decision-making[13]. The developed nomogram aligns with American Association for the Study of Liver Diseases-recommended PHT management strategies; however, regional variations in cirrhosis burden may influence its applicability[12]. For instance, cirrhosis-related mortality is highest in sub-Saharan Africa due to widespread viral hepatitis, whereas in Western countries, increasing HCC incidence is primarily linked to alcoholic liver disease and non-alcoholic fatty liver disease[27,29,30].
In this study, portal vein diameter, creatinine, potassium, and AFP were identified as key survival predictors after TIPS, consistent with existing findings[9,10,28,31]. Portal vein diameter is crucial for assessing PHT severity and is closely linked to prognosis[9,31]. Elevated creatinine indicates kidney dysfunction and may predict outcomes after TIPS[10,28]. Low potassium levels reflect electrolyte and acid-base imbalances[28]. Although AFP is traditionally considered an HCC marker, elevated AFP was associated with poorer survival after TIPS. This suggests that AFP may not only indicate HCC but also reflect liver function and inflammatory responses, reinforcing its role as an inflammatory marker[32]. Furthermore, portal pressure gradient is a critical predictor of post-TIPS outcomes. In this study, AFP and serum creatinine were more relevant for predicting long-term survival. These differences may be attributed to variations in patient populations, timing of stent placement, and technical factors[20,24]. In this study, log transformation was applied to key variables to handle non-normal data and improve model robustness, thus improving predictive accuracy[10,28]. LASSO regression is highly effective in managing high-dimensional data and avoiding overfitting, making it widely used in biomedical research[23]. LASSO shrinks the coefficients of non-informative variables to zero via L1 regularization, retaining only the most predictive ones, which simplifies the model and enhances both interpretability and robustness[10,20,23]. In this study, LASSO identified portal vein diameter, creatinine, potassium, and AFP as key variables, all of which showed significant independent prognostic value in Cox regression, consistent with previous findings on post-TIPS outcomes[28,33]. The developed model showed high discrimination ability and good calibration in training and validation sets, further demonstrating the utility of combining LASSO and Cox regression for handling complex datasets[10,20,23].
Nonetheless, this study has some limitations. First, this is a retrospective study, and thus selection bias is inevitable. Additionally, the sample size may limit the generalizability of the model[20]. Besides, regional differences in cirrhosis and HCC risk factors may affect the model’s applicability. For instance, 50%-60% of HCC cases in Eastern Europe are alcohol-related, compared to only 6%-14% in the Middle East and North Africa[27]. Therefore, localized adjustments should be applied to ensure accuracy. Nonetheless, further global studies are crucial to optimizing cirrhosis and PHT management strategies, which may enhance the model’s applicability and clinical utility[29]. Although this study incorporated a broad range of cirrhosis-related factors into the model, key factors like genetic background and polymorphisms were not included, potentially affecting its predictive accuracy and external validity[23]. Therefore, future research should incorporate more population-specific characteristics and biomarkers to create more precise and personalized predictive models[33]. Future studies should also focus on several key areas to enhance the clinical applicability of the model. Prophylactic TIPS can significantly reduce rebleeding risk and improve long-term survival in high-risk patients, especially in patients with acute-on-chronic liver failure. Therefore, future work should clarify the indications for prophylactic TIPS and refine patient selection through individualized risk assessment to improve both procedure success and long-term survival[17,34]. One key limitation is that the follow-up period in this study was limited to two years, which may not fully capture long-term survival outcomes or late complications such as hepatic encephalopathy and shunt dysfunction. While our model provides robust survival predictions within this timeframe, future studies should explore its applicability beyond two years, incorporating long-term follow-up data to assess dynamic prognostic factors influencing survival. Additionally, exploring the impact of evolving therapies and personalized treatment approaches on long-term outcomes will be crucial for refining predictive models and improving clinical decision-making in TIPS patients[35].
CONCLUSION
Accurate postoperative survival prediction has become a key concern in clinical practice. In this study, a nomogram model was developed to predict survival in TIPS patients by combining LASSO and Cox regression models. While the model showed promising potential in clinical applications, its generalizability and accuracy should be further validated in diverse populations. Therefore, future research should consider incorporating more biomarkers and genetic factors to improve model prediction. In conclusion, developing more precise and personalized predictive tools is critical for improving long-term outcomes among TIPS patients.
Footnotes
Provenance and peer review: Unsolicited article; Externally peer reviewed.
Peer-review model: Single blind
Specialty type: Gastroenterology and hepatology
Country of origin: China
Peer-review report’s classification
Scientific Quality: Grade C, Grade C
Novelty: Grade C, Grade C
Creativity or Innovation: Grade C, Grade C
Scientific Significance: Grade B, Grade C
P-Reviewer: Hu MJ; Wang B S-Editor: Bai Y L-Editor: A P-Editor: Zhao YQ
Büttner L, Aigner A, Pick L, Brittinger J, Steib CJ, Böning G, Streitparth F. 25 years of experience with transjugular intrahepatic portosystemic shunt (TIPS): changes in patient selection and procedural aspects.Insights Imaging. 2022;13:73.
[RCA] [PubMed] [DOI] [Full Text] [Full Text (PDF)][Cited in This Article: ][Reference Citation Analysis (0)]
Wang G, Zhang F, Ojeda A, Shalaby S, Hernandez-Gea V, Garcia-Pagan JC. The evolution of the TIPS placement technique and its applications over four decades.Dig Liver Dis. 2024;56:1980-1988.
[RCA] [PubMed] [DOI] [Full Text][Cited in This Article: ][Reference Citation Analysis (0)]
Larrue H, D'Amico G, Olivas P, Lv Y, Bucsics T, Rudler M, Sauerbruch T, Hernandez-Gea V, Han G, Reiberger T, Thabut D, Vinel JP, Péron JM, García-Pagán JC, Bureau C. TIPS prevents further decompensation and improves survival in patients with cirrhosis and portal hypertension in an individual patient data meta-analysis.J Hepatol. 2023;79:692-703.
[RCA] [PubMed] [DOI] [Full Text][Cited in This Article: ][Cited by in Crossref: 4][Cited by in RCA: 35][Article Influence: 17.5][Reference Citation Analysis (1)]
Gong J, Xia Z, Zhou Z, Chen L, Wang X, Zhou F. Effectiveness and prognosis of covered stents with different diameters in transjugular intrahepatic portosystemic shunt: a meta-analysis.Eur J Gastroenterol Hepatol. 2024;36:229-237.
[RCA] [PubMed] [DOI] [Full Text][Cited in This Article: ][Reference Citation Analysis (0)]
Lv Y, Chen H, Luo B, Bai W, Li K, Wang Z, Xia D, Guo W, Wang Q, Li X, Yuan J, Cai H, Xia J, Yin Z, Fan D, Han G. Transjugular intrahepatic portosystemic shunt with or without gastro-oesophageal variceal embolisation for the prevention of variceal rebleeding: a randomised controlled trial.Lancet Gastroenterol Hepatol. 2022;7:736-746.
[RCA] [PubMed] [DOI] [Full Text][Cited in This Article: ][Cited by in Crossref: 11][Cited by in RCA: 30][Article Influence: 10.0][Reference Citation Analysis (0)]
Büttner L, Pick L, Jonczyk M, Fehrenbach U, Collettini F, Auer TA, Schnapauff D, De Bucourt M, Wieners G, Gebauer B, Aigner A, Böning G. Shunt dysfunction and mortality after transjugular intrahepatic portosystemic shunt (TIPS) in patients with portal hypertension.Insights Imaging. 2024;15:193.
[RCA] [PubMed] [DOI] [Full Text][Cited in This Article: ][Reference Citation Analysis (0)]
Fang K, Yang Q, Lin Y, Zheng L, Wang HL, Wu J. Global cirrhosis prevalence trends and attributable risk factors-an ecological study using data from 1990-2019.Liver Int. 2022;42:2791-2799.
[RCA] [PubMed] [DOI] [Full Text][Cited in This Article: ][Reference Citation Analysis (0)]
Zhou X, Liu M, Zheng Z, Cao X, Lin Y, Xu Y, Li Y, Wang HC, Sun Q. Nomogram predicts survival and surgical benefits for patients with breast cancer with initial bone metastasis: A population-based study.Cancer. 2024;130:1464-1475.
[RCA] [PubMed] [DOI] [Full Text][Cited in This Article: ][Reference Citation Analysis (0)]
Trebicka J, Gu W, Ibáñez-Samaniego L, Hernández-Gea V, Pitarch C, Garcia E, Procopet B, Giráldez Á, Amitrano L, Villanueva C, Thabut D, Silva-Junior G, Martinez J, Genescà J, Bureau C, Llop E, Laleman W, Palazon JM, Castellote J, Rodrigues S, Gluud L, Ferreira CN, Barcelo R, Cañete N, Rodríguez M, Ferlitsch A, Mundi JL, Gronbaek H, Hernández-Guerra M, Sassatelli R, Dell'Era A, Senzolo M, Abraldes JG, Romero-Gómez M, Zipprich A, Casas M, Masnou H, Primignani M, Weiss E, Catalina MV, Erasmus HP, Uschner FE, Schulz M, Brol MJ, Praktiknjo M, Chang J, Krag A, Nevens F, Calleja JL, Robic MA, Conejo I, Albillos A, Rudler M, Alvarado E, Guardascione MA, Tantau M, Bosch J, Torres F, Pavesi M, Garcia-Pagán JC, Jansen C, Bañares R; International Variceal Bleeding Observational Study Group and Baveno Cooperation. Rebleeding and mortality risk are increased by ACLF but reduced by pre-emptive TIPS.J Hepatol. 2020;73:1082-1091.
[RCA] [PubMed] [DOI] [Full Text][Cited in This Article: ][Cited by in Crossref: 72][Cited by in RCA: 124][Article Influence: 24.8][Reference Citation Analysis (0)]
