Analysis of the treatment effect on recurrent bleeding and death in patients with cirrhosis and esophageal varices : Multistage competing-risks model compared to conventional methods

Background/Aims: Multiple recurrences of bleeding with high mortality in cirrhosis with esophageal varices have been inadequately analyzed in previous trials. We propose analysis by the multistage competing-risks model, specifying the effect on overall mortality as an effect on mortality during bleeding, rate of cessation of bleeding, mortality rate without bleeding, and rate of rebleeding. Methods: The Copenhagen Esophageal Varices Project enrolled patients after first bleeding and randomized 94 to usual treatment and 93 to sclerotherapy as supplement. During 9–52 months of follow-up, rebleeding occurred in 49 and 42, and death in 68 and 60 patients, respectively. The proportional hazards regression model (Cox model) was used for reanalysis both by the multistage competing-risks and by conventional analysis for overall mortality and rate of first rebleeding. In the multistage model, time zero was at entry to any new disease stage, of which the first four were analyzed - two bleeding stages and two bleeding-free stages. Results: The conventional analysis showed a reduction of overall mortality rate in the sclerotherapy group of borderline significance, but no effect on rate of rebleeding. The multistage model indicated that sclerotherapy reduced the rate of rebleeding late in the disease course, and particularly after the first rebleeding. Rate of cessation of bleeding and mortality rates during bleeding and without bleeding were not affected by sclerotherapy. Conclusions: Conventional analysis may give misleading conclusions, which might be avoided by applying the multistage model. The effect of sclerotherapy on overall mortality may be ascribed entirely to the reduced rate of rebleeding.