Prediction of esophageal varices in patients with cirrhosis

Goals: To identify predictors of esophageal varices (EV) using available clinical, laboratory, and diagnostic imaging variables. Background: Patients with cirrhosis frequently undergo screening endoscopy for varices so that prophylactic therapy and/or follow up can be planned. It is unclear how often patients should be screened endoscopically for varices, and there are few data on the relationship of varices to nonendoscopic variables. Study: Charts were reviewed for 247 consecutive patients with cirrhosis who underwent screening esophagogastroduodenoscopy for varices. Results: A total of 184 patients (68 women) were studied. Ninety-four patients (51%) had varices; of whom, 90 had only EV (small, n = 66; large, n = 24), 13 had EV and gastric varices, and 4 had isolated gastric varices. The distribution of EV according to the Child-Turcotte-Pugh class was as follows: A, 35%; B, 60%; and C, 69%, with roughly equal prevalence of large varices (29%, 24%, and 24%, respectively) in each class. Independent predictors of large varices were thrombocytopenia (p = 0.02) and splenomegaly (p = 0.04) seen using imaging. A platelet count of less than 68,000/mm(3) had the highest discriminative value for large EV with a sensitivity of 71% and a specificity of 73%. Splenomegaly had sensitivity and specificity of 75% and 58%, respectively. Using these two variables, we placed patients into one of four groups, with a risk for large varices ranging from 4% to 34%. Conclusions: The prevalence of EV in cirrhosis increases with the severity of liver disease, as expected. Thrombocytopenia and splenomegaly are independent predictors of large EV in cirrhosis. Further prospective studies might result in a discriminating algorithm to predict which patients with cirrhosis would benefit from early or regular endoscopy to detect clinically significant varices.