Limitations and enhancement opportunities for variceal rebleeding prediction model in patients with cirrhosis

Abstract

A multicenter study recently published introduced a novel prognostic model for predicting esophagogastric variceal rebleeding after endoscopic treatment in patients with cirrhosis. The model incorporated six readily available clinical variables—albumin level, aspartate aminotransferase level, white blood cell count, ascites, portal vein thrombosis, and bleeding signs—and demonstrated promising predictive performance. However, limitations, including the retrospective design and exclusion of patients with hepatocellular carcinoma, may affect the generalizability of the model. Additionally, further improvement is needed in the model’s discrimination between intermediate- and high-risk groups in external. Prospective validation and inclusion of additional variables are recommended to enhance predictive accuracy across diverse clinical scenarios.

Key Words: Prognostic model; Liver cirrhosis; Variceal rebleeding; Risk stratification; Endoscopic treatment; Portal hypertension; Clinical prediction

Core Tip: A novel prognostic model for predicting esophagogastric variceal rebleeding after endoscopic treatment in patients with cirrhosis demonstrated excellent predictive performance using six clinical variables. Despite its promise, limitations, including the retrospective design and hepatocellular carcinoma patient exclusion, necessitate prospective validation and additional variable incorporation to enhance clinical utility across risk categories.

TO THE EDITOR

We read with great interest the article by Zhan et al[1] published in the World Journal of Gastroenterology, which introduces the rebleeding event prediction following endoscopic treatment (REPET) model for predicting esophagogastric variceal rebleeding in patients with cirrhosis. Although we recommend this potentially valuable prognostic tool in the evolving landscape of portal hypertension management[2,3], several important aspects warrant further discussion.

First, the model’s performance in terms of risk stratification requires careful consideration. Although the overall C-index (0.775–0.862) was promising, the discrimination between intermediate- and high-risk groups in external validation was suboptimal. This limitation can substantially affect patient care through several critical pathways. For instance, misclassification between these risk groups may lead to inappropriate management decisions, such as delayed transjugular intrahepatic portosystemic shunt (TIPS) placement in high-risk patients erroneously categorized as having an intermediate risk or, conversely, premature invasive interventions in intermediate-risk patients misclassified as having a high risk. Furthermore, this imprecise stratification can result in suboptimal endoscopic surveillance intervals, potentially missing early rebleeding in undertriaged patients or subjecting others to unnecessary frequent endoscopies. Recent guidelines emphasize risk-stratified management approaches, treatment intensity tailoring, and follow-up protocols based on predicted rebleeding risk[2,4]. For example, high-risk patients may benefit from earlier consideration of TIPS or more frequent endoscopic surveillance, whereas intermediate-risk patients can be managed using standard endoscopic follow-up protocols. This underscores the critical importance of accurate risk discrimination, as highlighted by recent studies demonstrating the impact of precise risk stratification on patient outcomes and resource utilization[5,6].

Second, the exclusion of patients with hepatocellular carcinoma (HCC) represents a significant limitation. Patients with HCC exhibit unique bleeding risk profiles owing to elevated portal pressure, tumor invasion, and coagulopathy[7], which frequently exacerbate bleeding risks independent of cirrhosis severity. Given that HCC commonly coexists with cirrhosis and can independently influence bleeding risk[8,9], this exclusion may restrict the model’s generalizability to real-world clinical settings, where comorbid conditions often complicate patient management.

Third, although the retrospective design allows for efficient model development, it introduces inherent bias. Prospective validation studies are crucial for establishing the clinical utility of prediction models for portal hypertension management[10]. Contemporary prognostic models, such as those for preemptive TIPS placement[11], have demonstrated the value of prospective validation in diverse patient populations. To enhance the clinical applicability of the model, future research should consider incorporating additional variables, such as portal pressure gradient measurements, model for end-stage liver disease scores, Child–Pugh scores, creatinine levels, endoscopic findings (particularly variceal size), and novel biomarkers. Furthermore, conducting substudies or sensitivity analyses to evaluate the impact of HCC on the model’s predictive capabilities will strengthen its applicability across diverse clinical scenarios.

Despite these limitations, the REPET model represents a valuable advancement in risk stratification for variceal rebleeding and aligns with the goals outlined in the Baveno VII consensus[3]. We encourage the authors to conduct prospective validation studies to address these concerns, thereby strengthening the clinical applicability and utility of the model for personalized patient care.

CONCLUSION

Although the REPET model represents a promising advancement in predicting the risk of variceal rebleeding in patients with cirrhosis, addressing its limitations through prospective validation and incorporating additional variables is essential. Future research should include diverse patient populations, including those with HCC, to enhance the clinical applicability and accuracy of the model across all risk categories.