Transjugular intrahepatic portosystemic shunt for refractory ascites: a meta-analysis of individual patient data.

BACKGROUND & AIMS Several randomized controlled trials have compared a transjugular intrahepatic portosystemic shunt (TIPS) with large-volume paracentesis in cirrhotic patients with refractory ascites. Although all agree that TIPS reduces the recurrence rate of ascites, survival is controversial. The aim of this study was to compare the effects of TIPS and large-volume paracentesis in cirrhotic patients with refractory ascites by means of meta-analysis of individual patient data from 4 randomized controlled trials. METHODS The study population consisted of 305 patients: 149 allocated to TIPS and 156 to paracentesis. Cumulative probabilities of transplant-free survival and of hepatic encephalopathy (HE) were estimated by the Kaplan-Meier method and differences assessed by log-rank test. The total number of HE episodes per patient was also compared between TIPS and paracentesis. RESULTS Tense ascites recurred in 42% of patients allocated to TIPS and 89% allocated to paracentesis (P < .0001). Sixty-five patients in the TIPS group and 78 in the paracentesis group died. The actuarial probability of transplant-free survival was significantly better in the TIPS group (P = .035). Cox regression analysis performed in a subgroup of 235 patients (114 allocated to TIPS and 121 to paracentesis) showed that age, serum bilirubin level, plasma sodium level, and treatment allocation were independently associated with transplant-free survival. The average number of HE episodes was significantly higher in the TIPS group (1.13 +/- 1.93 vs 0.63 +/- 1.18; P = .006), although the cumulative probability of developing the first episode of HE was similar between the groups (P = .19). CONCLUSIONS The present meta-analysis of individual patient data provides further evidence to the previous meta-analyses of literature data showing that TIPS significantly improves transplant-free survival of cirrhotic patients with refractory ascites.