Predicting death in patients with end-stage liver disease: a new model for assessing disease severity

Abstract

Objective. To define possibilities of clinical application of the new original method for estimating failure (death) probability in patients on a liver transplant waiting list.

Material and Methods. The study included 350 patients who had been on a liver transplant waiting list for 5 years. Using the comparison of Mann-Whitney test results and evaluation of sensitivity and specificity (ROC curves) it was established that values of age, MELD-Na score, leukocyte level, nature of liver failure and presence of portal vein thrombosis had statistically significant differences between the dead and survived patients (p < 0.05). By means of binary logistic regression, the model assessing the risk of death taking into consideration indicators mentioned above has been obtained. The new index of death probability of a patient on a liver transplant waiting list within one year has been created.

Results. Quality evaluation of the created model and the index derived from it showed that the new index had a stronger ability to estimate somatic status severity in a patient with cirrhosis and allowed to make more precise prognosis of adverse outcome risk for not less than 12 months as compared to the standard MELD-Na score. The predicted risk of death coincided with actual mortality of patients in 83% of cases. The EPV criterion was 17.4 what exceeded the minimum admissible threshold of the criterion (10) for small samples and allowed to use the obtained index.

Conclusion. The original method allows increasing the accuracy of assessment of failure (death) development in a patient with cirrhosis for one year at any time of its application in the non-invasive way, using the data of the examination standard. In addition, the method helps setting priority in liver transplantation.