Clinical question: How can the model for end-stage liver disease (MELD)-based model be updated and utilized to predict inpatient mortality rates of hospitalized cirrhotic patients with acute variceal bleeding (AVB)?
Background: AVB in cirrhosis continues to carry mortality rates as high as 20%. Risk prediction for individual patients is important to determine when a step-up in acuity of care is needed and to identify patients who would most benefit from preemptive treatments such as a transjugular intrahepatic portosystemic shunt. Many predictive models are available but are currently difficult to apply in the clinical setting.
Study design: Initial comparison data was collected via a prospective study from clinical records. Confirmation of updated MELD model occurred via cohort validation studies.
Setting: Prospective data collected from Hospital Clinic in Barcelona, Spain. Validation cohorts for new MELD model calibration completed in hospital settings in Canada and Spain.
Synopsis: Data was collected from 178 patients with cirrhosis and esophageal AVB receiving standard therapy from 2007-2010. Esophageal bleeding was confirmed endoscopically. The primary endpoint was six-week, bleeding-related mortality. Among all the subjects studied, the average six-week mortality rate was 16%. Models evaluated for validity included the Child-Pugh, the D’Amico and Augustin models, and the MELD score.
Each model was assessed via discrimination, calibration, and overall performance in mortality prediction. The MELD was identified as the best model in terms of discrimination and overall performance but was miscalibrated. The original validation cohort from the Hospital Clinic in Spain was utilized to update the MELD calibration via logistic regression. External validation was completed via cohort studies in Canada (N=240) and at Vall D’Hebron Hospital in Spain (N=221).
Using the updated model, the MELD score adds a predictive component in the setting of AVB that has not been available. MELD values of 19 and higher predict mortality >20%, whereas MELD values lower than 11 predict mortality of 5%.
Bottom line: Utilization of the updated MELD model may provide a more accurate method to identify patients in which more aggressive preemptive therapies are indicated using prognostic predictions of mortality.
Citation: Reverter E, Tandon P, Augustin S, et al. A MELD-based model to determine risk of mortality among patients with acute variceal bleeding. Gastroenterology. 2014;146(2):412-419.