Cirrhotic patients with clinically significant portal hypertension (CSPH) often develop refractory ascites. Transjugular intrahepatic portosystemic shunt (TIPS) is a well-established treatment for CSPH, but it carries a substantial risk of overt hepatic encephalopathy (OHE). In this issue, Dr. Xuefeng Luo analyzes a recent study published in Hepatology, which applies machine learning techniques to determine the optimal reduction in post-TIPS portal pressure gradient (PPG) to effectively control ascites while minimizing the risk of OHE.Background: The Clinical Dilemma of TIPS – A Double-Edged Sword in Cirrhosis Treatment
Cirrhosis is a major global health burden, causing approximately 170,000 deaths annually in Europe alone (Pimpin L, et al. J Hepatol. 2018;69(3):718-735). CSPH is a key driver of complications such as ascites and variceal bleeding, signaling the transition from compensated to decompensated cirrhosis (D’Amico G, et al. J Hepatol. 2018;68(3):563-576).
TIPS is an effective intervention to relieve CSPH, reduce ascites, and lower bleeding risk. However, a significant limitation of the procedure is the high incidence of post-TIPS OHE, which restricts its widespread clinical application (Ehrenbauer AF, et al. JHEP Rep. 2023;5(9):100829).
A major clinical challenge is determining the optimal reduction in PPG—lowering it enough to control ascites while avoiding excessive shunting that increases the risk of OHE. This issue is central to the Baveno VII consensus agenda (de Franchis R, et al. J Hepatol. 2022;76(4):959-974). Since the relationship between PPG reduction and OHE risk is complex and nonlinear, traditional statistical methods are insufficient to establish precise thresholds. Large-scale data analysis using advanced computational approaches is needed to identify the “optimal balance point” for PPG reduction.
A study published in the January 2025 issue of Hepatology, titled Optimal Portal Pressure Reduction Strategy After TIPS for Ascites Control and Minimization of Hepatic Encephalopathy Risk: A Multicenter Study, is the first to leverage machine learning models and large multicenter datasets to determine the best post-TIPS PPG reduction strategy. This study provides high-level evidence to guide clinical decision-making in balancing ascites control with OHE prevention.
Study Design: Multicenter Data and Machine Learning-Driven Insights
This retrospective multicenter cohort study analyzed data from three European medical centers—Hannover, Vienna, and Hamburg—collected between 2000 and 2023. The study focused on cirrhotic patients who underwent TIPS using covered stents (PTFE) for refractory ascites or hepatic hydrothorax. Patients with hepatocellular carcinoma, vascular liver disease, or incomplete data were excluded, resulting in a final cohort of 729 patients. The dataset was split into a modeling group (n=438) and a validation group (n=291) in a 60:40 ratio to assess generalizability.
The primary variable was PPG, calculated as the pressure difference between the portal vein and the inferior vena cava before and after TIPS. The primary outcomes were hepatic decompensation within one year (defined as requiring paracentesis, long-term drainage, or low-flow devices) and OHE (West Haven grade 2–4). Competing events included death and liver transplantation.
A combination of machine learning and traditional statistical methods was used. The study first applied a random survival forest model for competing risks (RSF-CR), which accounts for multiple competing events (HDA, OHE, death, or transplantation) and identifies key predictive variables. Partial dependence plots (PDPs) were then used to visualize the independent effect of PPG reduction on HDA and OHE risk, helping to identify critical threshold points. These findings were further validated using maximally selected rank statistics and Fine & Gray competing risk analysis.
Key Findings: The Optimal PPG Reduction Range is 60%–80%
1. Baseline Characteristics of Patients
The median MELD scores for the total cohort (n=729), modeling cohort (n=438), and validation cohort (n=291) were 13 (IQR 10–16), 12 (IQR 10–16), and 13 (IQR 10–17), respectively. The Freiburg post-TIPS survival index (FIPS) had a median value of 0.00 (IQR -0.54–0.48) for the total cohort, 0.05 (-0.51–0.49) for the modeling cohort, and -0.07 (-0.59–0.47) for the validation cohort.
Alcohol-related liver disease was the predominant cause of cirrhosis in all groups (total cohort: 64.6%; modeling cohort: 66%; validation cohort: 62.5%). Pre-TIPS OHE was present in 23% of the total cohort, 21% of the modeling cohort, and 24.4% of the validation cohort.
2. Modeling Cohort Results
PDP analysis revealed that HDA risk significantly decreased when PPG reduction exceeded 60%, but OHE risk began rising when PPG reduction surpassed 80%. Fine & Gray analysis confirmed that the optimal PPG reduction range was 60%–80%.
Within this range:
- HDA risk decreased significantly (P=0.027, subdistribution hazard ratio [sHR]=0.70, 95% CI: 0.52–0.96).
- OHE risk remained stable, with no significant increase (P=0.62, sHR=0.92, 95% CI: 0.67–1.27).
3. Validation Cohort Results
- HDA risk decreased by 34% (P=0.028, sHR=0.66, 95% CI: 0.46–0.96).
- OHE risk remained unchanged (P=0.60, sHR=0.89, 95% CI: 0.61–1.32), confirming model robustness.
4. Subgroup and Sensitivity Analyses
- PPG reduction >80% was linked to increased OHE risk (P=0.036, sHR=1.58, 95% CI: 1.04–2.41).
- PPG reduction <60% was associated with higher HDA risk (P=0.048, sHR=1.38, 95% CI: 1.003–1.90).
Clinical Implications: Moving from a One-Size-Fits-All Approach to Individualized TIPS Management
- Procedure Optimization: Continuous intraoperative PPG monitoring is crucial to maintain reduction within 60%–80%. In high-risk patients (e.g., those with HanDeCT-positive findings), avoiding excessive PPG reduction (>80%) is key to preventing cardiac overload and ascites recurrence
- Use of Controlled-Expansion Stents: Gradual expansion stents allow for staged reductions (initially targeting 60%) with later adjustments if needed.
- Prioritizing OHE Prevention: While reducing HDA risk is important, preventing OHE has a greater impact on patient quality of life. In patients with poor liver reserve or prior OHE history, a conservative target closer to 60% is advisable.
Conclusion: A Milestone in Precision TIPS Therapy
This study establishes the 60%–80% PPG reduction range as the optimal target, advancing the field of interventional hepatology from empirical decision-making to data-driven precision medicine. With further development of real-time monitoring and AI-driven models, individualized TIPS therapy will continue to improve patient outcomes and quality of life.
