Risk prediction models for hepatic encephalopathy following TIPS: a systematic review and meta-analysis

Abstract

Background and aims:

Numerous risk-prediction models for hepatic encephalopathy (HE) after TIPS have been proposed, but their quality, discrimination, and clinical utility remain uncertain. We aimed to identify key predictors of post-TIPS HE and to critically appraise existing models.

Methods:

We performed a systematic review and meta-analysis of observational studies (inception 15 July 2025) from Chinese (VIP, Wanfang, CNKI, and CBM) and international (Embase, PubMed, Web of Science, and Cochrane Library) databases. Model quality and bias were assessed via PROBAST. Pooled AUCs and predictor effect sizes were calculated using STATA 15.0 and MedCalc.

Results:

A total of 24 studies (5,197 patients) yielded 32 unique models; the incidence of HE ranged from 19.9 to 46.6%. Discrimination was generally good (AUC range, 0.64–1.00; 30 models >0.70; 22 > 0.80). PROBAST flagged high bias—especially in model analysis. A meta-analysis produced a summary AUC of 0.815 (95%CI, 0.780–0.849). Consistent predictors (p < 0.01) included older age, diabetes, higher Child-Pugh score/class, elevated ammonia, and increased portal-to-splenic vein diameter ratio.

Conclusion:

Existing post-TIPS HE models demonstrate strong discrimination but suffer methodological limitations and bias. Future studies should employ multicenter cohorts, harmonized definitions, rigorous analytics, and external validation to yield robust, clinically actionable tools.

Systematic review registration:

https://www.crd.york.ac.uk/PROSPERO/view/CRD42024597699, CRD42024597699.

1 Introduction

Transjugular intrahepatic portosystemic shunt (TIPS) is an image-guided intervention that creates a low-resistance channel between the portal and hepatic veins to lower portal pressure in patients with portal hypertension-related liver disease (1). Originally introduced in the 1980s, TIPS has become a mainstay in tertiary centers for controlling variceal hemorrhage and refractory ascites (2, 3). Beyond hemostasis and fluid management, growing data indicate that TIPS confers long-term benefits, such as improved survival, fewer readmissions, and enhanced nutritional status and quality of life in advanced cirrhosis (4–7).

However, these gains must be balanced against the risk of hepatic encephalopathy (HE), the most frequent neurologic complication post-TIPS, which affects 23–54.5% of patients within 6 months (8, 9). HE—from mild cognitive decline to coma—drives morbidity, prolongs hospitalization, and elevates mortality (10, 11). Despite advances in peri-procedural care, post-TIPS HE remains unpredictable and may offset the procedure’s overall benefit.

Researchers have thus developed multivariable models—combining clinical parameters, imaging markers, and biochemical biomarkers—to stratify HE risk after TIPS. However, their methodological rigor, discrimination, and applicability across heterogeneous cohorts remain unexamined. We therefore performed a systematic review of post-TIPS HE prediction tools, critically appraising model development, validation methods, and key performance indices. By comparing their strengths and limitations, we provide hepatologists and interventional radiologists practical, evidence-based recommendations for integrating these tools into patient selection and peri-procedural management.

2 Materials and methods

This review adhered to the PRISMA 2020 statement (12), and its protocol was prospectively registered in PROSPERO (CRD42024597699).

2.1 Search strategy

We searched Chinese (VIP, Wanfang, CNKI, and CBM) and English (Embase, PubMed, Web of Science, and Cochrane Library) databases from inception to 15 July 2025 using MeSH and free-text terms for TIPS (“transjugular intrahepatic portosystemic shunt,” “portasystemic shunt,” and “transjugular intrahepatic”), HE (“hepatic encephalopathy” and “ammoniac encephalopathy”), and risk models (“nomogram,” “prediction model*,” and “risk assessment*”). Database-specific search syntax was applied, and reference lists of included studies and reviews were manually searched. Full search strategies are provided in the Supplementary materials.

We applied the PICOTS framework according to Checklist for Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modeling Studies (CHARMS) (13) to guide objective formulation, search strategy, and study selection (14).

P: Adults (≥18 years) undergoing TIPS for portal-hypertension complications.

I: Published risk-prediction models for post-TIPS HE.

C: Not applicable (no head-to-head model comparisons).

O: Development of hepatic encephalopathy.

T: Predictions based on baseline demographics, clinical scores, and laboratory data.

S: Models intended to yield individualized HE risk estimates to inform clinical decision-making and preventive measures.

2.2 Include inclusion and exclusion criteria

The inclusion criteria were as follows:

(1) Adults (≥18 years) undergoing TIPS; (2) studies developing HE risk-prediction models; (3) prospective or retrospective study design; and (4) published in English or Chinese.

The exclusion criteria were as follows:

(1) Duplicate publications; (2) models derived from systematic reviews or meta-analyses; (3) studies without full-text access or with insufficient abstract information; (4) studies from which relevant data could not be extracted; (5) conference abstracts, theses, and other unpublished materials; and (6) studies that analyzed risk factors without constructing predictive models.

2.3 Literature screening and data extraction

Study selection followed PRISMA guidelines. After de-duplication in Zotero, titles and abstracts were screened against our inclusion/exclusion criteria. Two reviewers then independently reviewed full texts for eligibility. Data were extracted into a CHARMS-based template and managed in Excel. Any discrepancies were resolved by discussion or, if required, by a third reviewer.

2.4 Quality evaluation of literature

Risk of bias and applicability were assessed using PROBAST (15). Bias was judged across four domains (participants, predictors, outcomes, and analysis) via 20 signaling questions: studies with all domains rated “low” were deemed low-risk, whereas any “high” or “unclear” rating conferred high overall risk. Applicability was evaluated in three domains (participants, predictors, and outcomes), with any “high” concern indicating poor applicability. Two independent reviewers performed all assessments; disagreements were resolved by discussion or, if needed, third-party adjudication.

2.5 Data synthesis and statistical analysis

Meta-analyses were performed in Stata 15.0, with between-study heterogeneity evaluated by the Q test and I² statistic. A fixed-effects model was used when the p-value was >0.1 and I² was <50%; otherwise, a random-effects model was applied. Predictors exhibiting substantial heterogeneity underwent sensitivity analyses to assess robustness and explore heterogeneity sources. Statistical significance was defined as a p-value of <0.05. Model discrimination was quantified by pooling AUCs in MedCalc; when only 95% CIs were reported, standard errors were derived by dividing the CI width by 3.92 (16).

3 Results

3.1 Study selection

Figure 1 presents the PRISMA 2020 diagram, which outlines the detailed literature search methodology and outcomes.

Figure 1

The systematic search yielded 1,177 records; after removing 367 duplicates, 810 titles and abstracts were screened. Of these, 53 full texts were assessed for eligibility. Fourteen studies lacked predictive models, 16 were conference abstracts, and two were unavailable in full text. Ultimately, 24 studies satisfied all criteria and were included in the review.

3.2 Study characteristics

A total of 24 studies (16–39) were included (9 in Chinese, 15 in English), comprising 13 case–control, 7 retrospective cohort, and 1 prospective cohort design. Sample sizes ranged from 53 to 621 participants. Key characteristics are summarized in Table 1.

Among 24 studies, AUC ranged from 0.64 to 1.00 (Figure 2). Except for one study, all retained continuous predictors, with variable selection by stepwise regression (n = 1), univariate → multivariate analysis (n = 16), or LASSO (n = 4). Missing data were variably addressed—multiple imputation without case count (n = 1), exclusion after reporting missingness (n = 1), or complete-case analysis (n = 10). Most models employed logistic regression (n = 16), with four using hybrid machine-learning approaches (e.g., random forest, SVM, and CatBoost), one using random forest alone, two using Cox models, and one using a competing-risks Cox model. Calibration was evaluated by both Hosmer–Lemeshow and calibration plots (n = 5), calibration plots alone (n = 7), while ten studies did not report any calibration assessment.

Figure 2

3.3 Models validation

Regarding validation, three studies performed both internal and external validation, twelve conducted internal validation only, and one reported external validation alone. The remaining studies did not perform any form of validation. Key model characteristics are summarized in Table 2.

3.4 Results of quality assessment

PROBAST was used to evaluate bias and applicability across the 24 studies (Table 3). (1) Participants: Four case–control studies (18, 21, 22, 26) were judged high risk; nested case–control and cohort designs were low risk. (2) Predictors: Four studies (16, 23, 28, 36) were high risk because not all predictors were available at model use; in twelve studies (17–19, 21, 22, 25, 26, 30, 31, 33, 34, 37) predictor definitions or timing were insufficiently reported, yielding unclear risk. (3) Outcome: Thirteen studies (17–19, 21–26, 28, 30, 32, 33) had unclear risk because it was not specified whether outcome assessment was blinded to predictors or the interval between predictor and outcome measurement. (4) Analysis: Ten studies (16, 17, 20, 22–24, 27–29, 33, 36) used complete-case analysis; one (25) study used multiple imputation and the remaining study did not report methods for handling missing data or adjusting for overfitting, resulting in unclear or high risk.

4 Meta-analysis of validation models included in the review

Statistical pooling of model discrimination (AUC) was performed in MedCalc (Figure 3). This summary analysis primarily utilizes the AUC metric from the model development dataset. Given high heterogeneity (I² = 96.56%, p < 0.001), a random-effects model yielded a summary AUC of 0.815 (95%CI, 0.780–0.849), reflecting strong overall discrimination. Publication bias was negligible, as evidenced by a symmetrical funnel plot (Figure 4) and a non-significant Egger’s test (p = 0.308). Subgroup and meta-regression analyses stratified by modeling algorithm (logistic regression vs. machine learning), validation type (internal vs. external), events-per-variable (≥10 vs. <10), and missing-data strategy (exclusion vs. imputation) failed to explain the remaining heterogeneity (I² > 85%), suggesting unmeasured clinical or methodological factors.

Figure 3

Figure 4

Across studies, 71 candidate predictors emerged (Figure 5), with age, diabetes history, Child-Pugh score/class, blood ammonia, and portal–splenic vein diameter ratio—each examined in ≥2 cohorts—significantly associated with post-TIPS HE risk (Table 4). Sensitivity analyses confirmed the robustness of these associations despite heterogeneity.

Figure 5

5 Discussion

In 24 studies (35 predictive models, 5,197 patients), the reported incidence of HE after TIPS ranged from 19.9 to 46.6%. Since 2016, many models have shown encouraging discriminative performance (30 models with AUC > 0.70, 22 with AUC > 0.80), but substantial methodological heterogeneity undermines confidence in their clinical applicability. Most models were developed retrospectively and relied primarily on internal validation, with only five studies reporting external validation—features that increase the risk of selection and information bias and limit generalizability. Predictor selection frequently relied on univariable screening, a practice that can overlook important multivariable associations and fails to address multicollinearity or interaction effects (40). Handling of missing data was inconsistent: only one study used multiple imputation, ten excluded incomplete cases, and the majority did not report a clear strategy, further weakening the robustness and reproducibility of results.

Despite these limitations, three recurring predictor domains emerged: patient-specific factors (e.g., age and history of diabetes), preoperative clinical indicators (e.g., Child-Pugh score and blood ammonia level), and imaging-based parameters.

Advanced age is associated with an increased risk of HE, likely mediated through age-related gastrointestinal dysmotility, compositional shifts in the gut microbiota, and diminished extrahepatic ammonia detoxification capacity (41, 42). Physiologically, TIPS further augments systemic exposure to gut-derived ammonia by circumventing hepatic sinusoidal clearance (43, 44). Similarly, diabetes predisposes patients to HE through multiple mechanisms, such as insulin resistance-induced impairment of hepatic and skeletal muscle ammonia metabolism (45), and chronic hyperglycemia-driven upregulation of intestinal glutaminase activity and permeability, which together facilitate ammonia and endotoxin translocation (46, 47); post-TIPS diversion of portal flow further compromises first-pass hepatic extraction, thereby amplifying this risk (48, 49).

Higher pre-TIPS liver dysfunction scores reflect reduced synthetic function and intrinsic shunting, which TIPS may exacerbate (50). Although preprocedural ammonia level was prognostic in several reports, assay variability and lack of standardization limit its routine use (51); standardized protocols or point-of-care assays could improve its utility. Emerging biochemical markers (e.g., glutamine/glutamate ratios and phenylacetate) and dynamic flow metrics (Doppler or 4D-flow MRI) may enhance risk stratification beyond static measures and merit prospective evaluation.

Recent studies have increasingly indicated that malnutrition and sarcopenia are modifiable risk factors for hepatic encephalopathy (HE) after TIPS. Skeletal muscle plays a critical role in extrahepatic ammonia metabolism, and a reduction in muscle mass can markedly impair peripheral ammonia clearance, thereby increasing susceptibility to HE. In our dataset, several models included nutrition- or body composition-related variables—such as CT- or BIA-derived measures of muscle mass or fat distribution, prealbumin levels, or low-fat mass indices (reported in the variable lists as “sarcopenia,” “prealbumin,” or “visceral fat area index”). However, due to substantial heterogeneity in measurement methods, cutoff values, and reporting strategies, these variables did not demonstrate consistent stability in pooled analyses. In other words, while current evidence supports the biological plausibility and potential clinical relevance of nutrition-related indicators, methodological heterogeneity (such as assessment tools, thresholds, and follow-up time points) has hindered their consistent confirmation across prediction models. Therefore, we recommend that future model development prioritize internationally accepted definitions of sarcopenia (e.g., EWGSOP2 or AWGS) and report raw imaging-based measurements to facilitate cross-cohort comparisons and individual participant data (IPD) meta-analyses.

6 Conclusion

Although many existing models demonstrate promising discriminative performance, the substantial methodological heterogeneity across studies—particularly regarding modeling frameworks, predictor selection, and validation strategies—undermines the reliability and generalizability of current evidence. AUCs derived from various modeling frameworks are often highly heterogeneous and emphasize different aspects of predictive performance; therefore, simply pooling these AUCs into a single summary metric is methodologically questionable and may misrepresent a model’s clinical utility. To move the field forward, future studies should adopt prospective, multicenter designs with clearly pre-specified predictor sets, apply modern multivariable selection techniques (e.g., LASSO/ridge) (52, 53) and routine multiple imputation for missing data (54), and perform transparent reporting and external validation in line with TRIPOD and PROBAST (55). Notably, prognostic and diagnostic modeling frameworks should be explicitly separated, and pooled estimates (such as aggregated AUCs) should not be interpreted uncritically when they combine fundamentally different model types (e.g., logistic, Cox) since such pooling can be misleading.

6.1 Limitation

This systematic review has several limitations that should be carefully considered when interpreting the findings. First, approximately two-thirds of the included studies were derived from East Asian cohorts, predominantly from mainland China. This geographic concentration may limit the generalizability of the results to other regions, as differences in etiological profiles (e.g., proportions of viral versus alcohol-related liver disease), TIPS indications, and peri-procedural management strategies (such as perioperative antibiotic use, shunt design, and nursing practices) may influence both the incidence of HE and the external validity of identified predictors. Nevertheless, it is noteworthy that earlier cohort studies from Europe and North America have also reported several core predictors, suggesting that certain factors—such as age, Child-Pugh score, and blood ammonia levels—exhibit consistency across regions. However, these observations still require confirmation through prospective, multicenter, multinational studies. To enhance transparency, we have listed the geographic origin of each included study in the supplementary materials to allow readers to better assess the applicability of our findings.

Second, several potentially important predictors, such as a history of hepatic encephalopathy and nutritional status or sarcopenia, were inconsistently reported or variably defined across the original studies, resulting in a lack of consistent support in pooled model analyses. With respect to prior HE, some studies excluded patients with overt cognitive impairment or recurrent HE episodes, while others did not explicitly document mild or covert cognitive dysfunction at baseline. In addition, a history of HE often coexists with more advanced liver dysfunction and malnutrition, leading to strong collinearity with liver function scores and nutritional variables and increasing the likelihood of exclusion during multivariable model selection. Consequently, although earlier studies have suggested an association between prior HE or baseline cognitive impairment and post-TIPS HE risk, this variable did not emerge as a consistently significant predictor in our pooled analysis due to heterogeneity in definitions, sampling strategies, and statistical approaches. Future studies should incorporate systematic documentation of prior HE and standardized assessments of mild cognitive impairment to better delineate their independent predictive value.

Third, malnutrition and sarcopenia have gained increasing attention in recent years as biologically plausible mechanisms contributing to HE. Our list of candidate predictors included several nutrition- and body composition-related variables, such as sarcopenia, subcutaneous or visceral fat indices, and prealbumin levels. However, substantial variability in measurement methods (e.g., CT-based cross-sectional assessments, body mass index-based thresholds, or laboratory surrogate markers) and cutoff values across studies resulted in unstable findings in pooled analyses. Although existing evidence supports the role of muscle and nutritional depletion in increasing HE susceptibility—primarily through reduced peripheral ammonia clearance—methodological heterogeneity limits comparability. We therefore recommend that future models incorporate standardized sarcopenia assessments and nutritional scoring systems at the design stage and report detailed measurement methods and thresholds to facilitate subsequent synthesis and comparison.

Finally, methodological limitations should also be acknowledged. Most included studies were retrospective in nature, and only a minority performed external validation. In addition, there was no uniform definition of HE or standardized follow-up window (e.g., mixing HE incidence within 6- and 12-month periods), and approaches to handling missing data varied considerably, with complete-case analysis being the most common. These factors are likely to increase heterogeneity and may introduce bias in the estimation of predictive performance. To address this issue, we have provided a comprehensive summary table in the supplementary materials detailing the number of events, events per variable (EPV), missing data handling strategies, and external validation status for each model, enabling readers to better evaluate their robustness and potential clinical applicability.

References

• 1.

  Bosch J Berzigotti A Garcia-Pagan JC Abraldes JG . The management of portal hypertension: rational basis, available treatments and future options. J Hepatol. (2008) 48:S68–92. doi: 10.1016/j.jhep.2008.01.021,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 2.

  Lee EW Eghtesad B Garcia-Tsao G Haskal ZJ Hernandez-Gea V Jalaeian H et al . AASLD practice guidance on the use of TIPS, variceal embolization, and retrograde transvenous obliteration in the management of variceal hemorrhage. Hepatology. (2024) 79:224. doi: 10.1097/HEP.0000000000000530

  • CrossRef
  • Google Scholar

• 3.

  Garcia-Pagan JC Francoz C Montagnese S Senzolo M Mookerjee RP . Management of the major complications of cirrhosis: beyond guidelines. J Hepatol. (2021) 75:S135–46. doi: 10.1016/j.jhep.2021.01.027,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 4.

  Larrue H D'Amico G Olivas P Lv Y Bucsics T Rudler M et al . 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. doi: 10.1016/j.jhep.2023.04.028,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 5.

  Rössle M . TIPS: 25 years later. J Hepatol. (2013) 59:1081–93. doi: 10.1016/j.jhep.2013.06.014,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 6.

  Wang Z Jiang M-S Zhang H-L Weng NN Luo XF Li X et al . Is post-TIPS anticoagulation therapy necessary in patients with cirrhosis and portal vein thrombosis? A randomized controlled trial. Radiology. (2016) 279:943–51. doi: 10.1148/radiol.2015150369,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 7.

  Liu J Ma J Yang C Ye T Meng J Shi Q et al . Impact of TIPS on splenic volume and thrombocytopenia. AJR Am J Roentgenol. (2021) 216:698–703. doi: 10.2214/ajr.20.22958,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 8.

  Tripathi D Stanley AJ Hayes PC Travis S Armstrong MJ Tsochatzis EA et al . Transjugular intrahepatic portosystemic stent-shunt in the management of portal hypertension. Gut. (2020) 69:1173–92. doi: 10.1136/gutjnl-2019-320221,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 9.

  Gairing S Müller L Kloeckner R Galle P Labenz C . Review article: post-TIPSS hepatic encephalopathy-current knowledge and future perspectives. Aliment Pharmacol Ther. (2022) 55:1265–76. doi: 10.1111/apt.16825,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 10.

  Amodio P . Hepatic encephalopathy: diagnosis and management. Liver Int. (2018) 38:966–75. doi: 10.1111/liv.13752,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 11.

  Stepanova M Mishra A Venkatesan C Younossi ZM . In-hospital mortality and economic burden associated with hepatic encephalopathy in the United States from 2005 to 2009. Clin Gastroenterol Hepatol. (2012) 10:1034–1041.e1. doi: 10.1016/j.cgh.2012.05.016,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 12.

  The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. Int J Surg. (2021) 88:105906. doi: 10.1016/j.ijsu.2021.105906,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 13.

  Moons KGM de Groot JAH Bouwmeester W Vergouwe Y Mallett S Altman DG et al . Critical appraisal and data extraction for systematic reviews of prediction modelling studies: the CHARMS checklist. PLoS Med. (2014) 11:e1001744. doi: 10.1371/journal.pmed.1001744

  • CrossRef
  • Google Scholar

• 14.

  Debray TPA Damen JAAG Snell KIE Ensor J Hooft L Reitsma JB et al . A guide to systematic review and meta-analysis of prediction model performance (2017) 356:i6460. doi: 10.1136/bmj.i6460,

  • CrossRef
  • Google Scholar

• 15.

  Moons KGM Wolff RF Riley RD Whiting PF Westwood M Collins GS et al . PROBAST: a tool to assess risk of bias and applicability of prediction model studies: explanation and elaboration. Ann Intern Med. (2019) 170:W1–W33. doi: 10.7326/M18-1377,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 16.

  Yang Y Fu S Cao B Hao K Li Y Huang J et al . Prediction of overt hepatic encephalopathy after transjugular intrahepatic portosystemic shunt treatment: a cohort study. Hepatol Int. (2021) 15:730–40. doi: 10.1007/s12072-021-10188-5,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 17.

  Li Y Pang H He X . Development and validation of a nomogram for early hepatic encephalopathy after transjugular intrahepatic portosystemic shunt. J Pract Med. (2020) 36:963–8.

  • Google Scholar

• 18.

  Huang T Wang G Huang G Zhang C . Predicting clinical outcomes of cirrhotic patients after transjugular intrahepatic portosystemic shunt using machine learning algorithms. Chin J Digest Dis Imaging. (2022) 12:4–10.

  • Google Scholar

• 19.

  Li X Li J Sun W . Development and evaluation of a nomogram to predict overt hepatic encephalopathy in cirrhotic patients with esophageal and gastric variceal bleeding after transjugular intrahepatic portosystemic shunt. J Clin Hepatobiliary Dis. (2024) 40:1605–11.

  • Google Scholar

• 20.

  Liao Y Zhang L Wang J-T Yue ZD Fan ZH Wu YF et al . A novel nomogram predicting overt hepatic encephalopathy after transjugular intrahepatic portosystemic shunt in portal hypertension patients. Sci Rep. (2023) 13:15244. doi: 10.1038/s41598-023-42061-w,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 21.

  Li K Cheng Y Zhao R Jiang H Zhang L Tong Y et al . Prediction of mortality and overt hepatic encephalopathy undergoing transjugular intrahepatic portosystemic shunt: a retrospective cohort study. Abdom Radiol. (2024) 49:908–18. doi: 10.1007/s00261-023-04086-7

  • CrossRef
  • Google Scholar

• 22.

  Yang Y Liang X Yang S He X Huang M Shi W et al . Preoperative prediction of overt hepatic encephalopathy caused by transjugular intrahepatic portosystemic shunt. Eur J Radiol. (2022) 154:110384. doi: 10.1016/j.ejrad.2022.110384,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 23.

  İnce O Önder H Gençtürk M Golzarian J Young S . Machine learning insights: predicting hepatic encephalopathy after TIPS placement. Cardiovasc Intervent Radiol. (2023) 46:1715–25. doi: 10.1007/s00270-023-03593-w,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 24.

  Cheng S Hu G Jin Z Wang Z Xue H . Prediction of hepatic encephalopathy after transjugular intrahepatic portosystemic shunt based on CT radiomic features of visceral adipose tissue. Acad Radiol. (2024) 31:1849–61. doi: 10.1016/j.acra.2023.10.013,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 25.

  Yang C Zhu X Liu J Shi Q du H Chen Y et al . Development and validation of prognostic models to estimate the risk of overt hepatic encephalopathy after TIPS creation: a multicenter study. Clin Transl Gastroenterol. (2022) 13:e00461. doi: 10.14309/ctg.0000000000000461

  • CrossRef
  • Google Scholar

• 26.

  Cheng S Yu X Chen X Jin Z Xue H Wang Z et al . CT-based radiomics model for preoperative prediction of hepatic encephalopathy after transjugular intrahepatic portosystemic shunt. Br J Radiol. (2022) 95:20210792. doi: 10.1259/bjr.20210792

  • CrossRef
  • Google Scholar

• 27.

  Yin X Zhang F Guo H Peng C Zhang W Xiao J et al . A nomogram to predict the risk of hepatic encephalopathy after transjugular intrahepatic portosystemic shunt in cirrhotic patients. Sci Rep. (2020) 10:9381. doi: 10.1038/s41598-020-65227-2

  • CrossRef
  • Google Scholar

• 28.

  Tong H Gan C Wei B Wang ZD Li XD Qian SJ et al . Risk factors for overt hepatic encephalopathy after transjugular intrahepatic portosystemic shunt creation in patients with liver cirrhosis. J Dig Dis. (2021) 22:31–40. doi: 10.1111/1751-2980.12957,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 29.

  Chen X Huang M Yu X Chen J Xu C Jiang Y et al . Hepatic-associated vascular morphological assessment to predict overt hepatic encephalopathy before TIPS: a multicenter study. Hepatol Int. (2024) 18:1238–48. doi: 10.1007/s12072-024-10686-2,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 30.

  Yuan W Li L Han X Liu J Li B Wu J . Development and performance analysis of a clinical prediction model for hepatic encephalopathy after TIPS. Mod Dig Interv Ther. (2021) 26:874–7.

  • Google Scholar

• 31.

  Zhuo S Zhao W Hu J . Incidence and risk factors of overt hepatic encephalopathy after transjugular intrahepatic portosystemic shunt. J Clin Radiol. (2023) 42:1806–11.

  • Google Scholar

• 32.

  Yu X Xie P Cheng S Cao L Gu L Zhu T . Predictive value of preoperative peripheral blood inflammatory markers for hepatic encephalopathy after transjugular intrahepatic portosystemic shunt. J Interv Radiol. (2021) 30:937–40.

  • Google Scholar

• 33.

  Zhang W Gan J . Predictive value of age combined with albumin-bilirubin (ALBI) grade for overt hepatic encephalopathy after transjugular intrahepatic portosystemic shunt. J Pract Clin Med. (2024) 28:53–8.

  • Google Scholar

• 34.

  Wang L Qin J Yao X Qi Q Liu L Tang S . Development of a risk prediction model for overt hepatic encephalopathy in hepatitis B cirrhosis patients with portal hypertension undergoing TIPS. J Clin Hepatobiliary Dis. (2024) 40:1149–55.

  • Google Scholar

• 35.

  Nardelli S Gioia S Pasquale C Pentassuglio I Farcomeni A Merli M et al . Cognitive impairment predicts the occurrence of hepatic encephalopathy after transjugular intrahepatic portosystemic shunt. Am J Gastroenterol. (2016) 111:523–8. doi: 10.1038/ajg.2016.29,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 36.

  Wang C Teng Y Gao J Zhang Z Li Y . Low adipose tissue index as an indicator of hepatic encephalopathy in cirrhotic patients following transjugular intrahepatic portosystemic shunt. Abdominal Radiology. (2023) 48:1454–67. doi: 10.1007/s00261-023-03813-4,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 37.

  Sparacia G Colelli G Parla G . Brain magnetic resonance imaging radiomic signature and machine learning model prediction of hepatic encephalopathy in adult cirrhotic patients. Life. (2025) 15:346.

  • Google Scholar

• 38.

  Liu D-J Jia L-X Zeng F-X Zeng W-X Qin G-G Peng Q-F et al . Machine learning prediction of hepatic encephalopathy for long-term survival after transjugular intrahepatic portosystemic shunt in acute variceal bleeding. World J Gastroenterol. (2025) 31:100401. doi: 10.3748/wjg.v31.i4.100401,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 39.

  Chang P Xue P Guo C . LASSO-logistic regression analysis and nomogram development for predicting hepatic encephalopathy after TIPS. J Pract Clin Med. (2025) 29:19–25.

  • Google Scholar

• 40.

  Ziegler A . Clinical prediction models: a practical approach to development, validation, and updating. Biom J. (2020) 62:1122–3.

  • Google Scholar

• 41.

  Nardelli S Riggio O Marra F Gioia S Saltini D Bellafante D et al . Episodic overt hepatic encephalopathy after transjugular intrahepatic portosystemic shunt does not increase mortality in patients with cirrhosis. J Hepatol. (2024) 80:596–602. doi: 10.1016/j.jhep.2023.11.033,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 42.

  Bettinger D Sturm L Pfaff L Hahn F Kloeckner R Volkwein L et al . Refining prediction of survival after TIPS with the novel freiburg index of post-TIPS survival. J Hepatol. (2021) 74:1362–72. doi: 10.1016/j.jhep.2021.01.023,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 43.

  Adhikary S Esmeeta A Dey A Banerjee A Saha B Gopan P et al . Impacts of gut microbiota alteration on age-related chronic liver diseases. Dig Liver Dis. (2024) 56:112–22. doi: 10.1016/j.dld.2023.06.017,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 44.

  Smith EE Biessels GJ Gao V Gottesman RF Liesz A Parikh NS et al . Systemic determinants of brain health in ageing. Nat Rev Neurol. (2024) 20:647–59. doi: 10.1038/s41582-024-01016-z,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 45.

  Hafizi Abu Bakar M Kian Kai C Wan Hassan WN Sarmidi MR Yaakob H Zaman Huri H . Mitochondrial dysfunction as a central event for mechanisms underlying insulin resistance: the roles of long chain fatty acids. Diabetes Metab Res Rev. (2015) 31:453–75. doi: 10.1002/dmrr.2601,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 46.

  Thaiss CA Levy M Grosheva I Zheng D Soffer E Blacher E et al . Hyperglycemia drives intestinal barrier dysfunction and risk for enteric infection. Science. (2018) 359:1376–83. doi: 10.1126/science.aar3318,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 47.

  Giri B Dey S Das T Sarkar M Banerjee J Dash SK . Chronic hyperglycemia mediated physiological alteration and metabolic distortion leads to organ dysfunction, infection, cancer progression and other pathophysiological consequences: an update on glucose toxicity. Biomed Pharmacother. (2018) 107:306–28. doi: 10.1016/j.biopha.2018.07.157,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 48.

  Portincasa P Bonfrate L Wang DQ-H Frühbeck G Garruti G Di Ciaula A . Novel insights into the pathogenic impact of diabetes on the gastrointestinal tract. Eur J Clin Investig. (2022) 52:e13846. doi: 10.1111/eci.13846

  • CrossRef
  • Google Scholar

• 49.

  de Kort S Keszthelyi D Masclee A . M. Leaky gut and diabetes mellitus: what is the link?Obes Rev. (2011) 12:449–58. doi: 10.1111/j.1467-789X.2010.00845.x,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 50.

  Li Y Wu Y-T Wu H . Management of hepatic encephalopathy following transjugular intrahepatic portosystemic shunts: current strategies and future directions. World J Gastroenterol. (2025) 31:103512. doi: 10.3748/wjg.v31.i15.103512,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 51.

  Hashim IA Cuthbert JA . Elevated ammonia concentrations: potential for pre-analytical and analytical contributing factors. Clin Biochem. (2014) 47:233–6. doi: 10.1016/j.clinbiochem.2014.08.013,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 52.

  Ranalli MG Salvati N Petrella L Pantalone F . M-quantile regression shrinkage and selection via the lasso and elastic net to assess the effect of meteorology and traffic on air quality. Biom J. (2023) 65:2100355. doi: 10.1002/bimj.202100355,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 53.

  Zhang Q Yuan K-H Wang L . Asymptotic bias of normal-distribution-based maximum likelihood estimates of moderation effects with data missing at random. Br J Math Stat Psychol. (2019) 72:334–54. doi: 10.1111/bmsp.12151,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 54.

  Guide A Garbett S Feng X Mapes BM Cook J Sulieman L et al . Balancing efficacy and computational burden: weighted mean, multiple imputation, and inverse probability weighting methods for item non-response in reliable scales. J Am Med Inform Assoc. (2024) 31:2869–79. doi: 10.1093/jamia/ocae217,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 55.

  Moons KGM Altman DG Reitsma JB Ioannidis JP Macaskill P Steyerberg EW et al . Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): explanation and elaboration. Ann Intern Med. (2015) 162:W1–W73. doi: 10.7326/M14-0698,

  • Pubmed Abstract
  • CrossRef
  • Google Scholar