Differences between previous published version of the review and this review

The whole review protocol part was thoroughly updated following the guidelines in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011): text was added or rewritten for greater precision and understanding. Outcomes were rearranged so that they now are patient oriented. An attempt to collect evidence on harm was also made. The statistical approach was also revised. Based on updated domains for bias risk, we judged bias risk domains anew (Higgins 2011). In the present update, only one trial was assessed as having low risk of bias. Minor inaccuracies found in data previously extracted were now corrected. Number of patients who died, were liver transplanted, obtained hepatic encephalopathy or were diagnosed with bleeding varices are now added to the analyses on serious adverse events. New trial data were added to some of the secondary outcome measures. Due to addition of data to the outcome measures on ascites and histology, the result of the effect for ascites changed from significant into insignificant, and the result of the effect for histology changed from insignificant into significant. We performed subgroup analyses using data in the already included trials as well data found in follow‐up publications of the included trials. Results on all‐cause mortality and mortality or liver transplantation as well as the following secondary outcomes ‐ pruritus, serum bilirubin and serum alkaline phosphatases ‐ were analysed with trial sequential analysis. The random‐effects model meta‐regression showed that none of examined covariates (bias risk of the trials, disease severity of patients at entry, ursodeoxycholic acid dosage, and trial duration) were significantly associated with the estimated intervention effect on mortality or mortality or liver transplantation. Summary of Findings tables and grading of the evidence were also performed. We abandoned our Baysian analyses as they did not seem to add new information.