Long-term survival prediction for transjugular intrahepatic portosystemic shunt in severe cirrhotic ascites: assessment of ten prognostic models.

OBJECTIVES Patients with severe cirrhotic ascites have poor prognosis, yet individual patient survival varies greatly. Therefore, suitable prognostic models can be helpful in clinical decision making. The aim of this study was to evaluate and compare the performance of 10 scores in predicting transplant-free survival (TFS) after transjugular intrahepatic portosystemic shunt (TIPS) in severe cirrhotic ascites. METHODS Two hundred eighty consecutive cirrhotic patients with severe ascites undergoing TIPS between March 2006 and December 2017 were retrospectively screened and included from nine tertiary Chinese centers, consisting of 123 patients with refractory ascites and 157 with recurrent ascites. Discriminatory ability of these models was further assessed in the whole cohort and subgroups. RESULTS TFS rates of all 280 patients were 75.4, 65.7, and 53.6% at 6-month, 1-year, and 2-year follow-up, respectively. Compared with other prognostic systems, the integrated model for end-stage liver disease (iMELD, incorporating serum sodium and age) showed optimal performance in predicting 6-month, 1-year, and 2-year TFS. Cutoffs were determined according to c-index and were used to stratify patients into three strata presenting significantly different TFS for short-term and long-term: iMELD < 32, ≥32 but <38 and ≥38 (log-rank P < 0.001). CONCLUSIONS The iMELD score proved to be the best prognostic model in predicting TFS in patients with severe cirrhotic ascites receiving TIPS. Meanwhile, the model could stratify patients in three strata to help guiding clinical practice.