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

## TL;DR

This study reviews and evaluates risk prediction models for hepatic encephalopathy after TIPS, finding strong discrimination but methodological flaws.

## Contribution

The paper provides a systematic review and meta-analysis of post-TIPS HE models, identifying key predictors and highlighting methodological limitations.

## Key findings

- 32 unique models were identified from 24 studies involving 5,197 patients.
- Discrimination was generally good with an AUC range of 0.64–1.00.
- Key predictors of HE included older age, diabetes, and higher Child-Pugh score.

## Abstract

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.

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.

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.

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.

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