Logistic regression model of non-invasively predicting risk for variceal bleeding

Objective To create a logistic regression model of non-invasively predicting risk for variceal bleeding,based on Doppler ultrasound and clinical parameters.Methods 60 patients with hepatic cirrhosis were enrolled.Doppler parameters and clinical data were used for binary logistic analysis,and multivariate logistic regression model was developed.14 patients were followed up for 1 year.Results Parameters of Q pv ,Q spv ,Q smv adn Q spv /Q pv were positively correlated with risk of bleeding,while negatively correlated with mean platelet volume(MPV).Logistic regression model with Q spv ,Q smv and MPV had a sensitivity of 83.8%,specificity of 86.2% and a positive predicting value of 85.7%,negative predicting value of 85.7% in predicting risk of bleeding.Conclusion It is feasible to use logistic regression model for predicting risk of variceal bleeding in practice.