Optimizing the time-frame for the definition of bleeding-related death after acute variceal bleeding in cirrhosis

Objective: To identify the best time-frame for defining bleeding-related death after variceal bleeding in patients with cirrhosis. Design: Prospective long-term evaluation of a cohort of 155 patients admitted with variceal bleeding. Setting: Eight medical departments in seven hospitals in north-eastern Italy. Methods: Non-linear regression analysis of a hazard curve for death, and Cox's multiple regression analyses using different zero-time points. Results: Cumulative hazard plots gave two slopes, the first corresponding to the risk of death from acute bleeding, the second a baseline risk of death. The first 30 days were outside the confidence limits of the regression curve for the baseline risk of death. Using Cox's regression analysis, the significant predictors of overall mortality risk were balanced between factors related to severity of bleeding and those related to severity of liver disease. If only deaths occurring after 30 days were considered, only predictors related to the severity of liver disease were found to be of importance. Conclusion: Thirty days after bleeding is considered to be a reasonable time-frame for the definition of bleeding-related death in patients with cirrhosis and variceal bleeding.