Background: All patients with liver cirrhosis are recommended for evaluation of oesophagogastric varices (EGV) regularly. This prospective study was designed to develop a predictive model for EGV in cirrhotic patients.
Methods: Ninety-two patients were recruited. From all patients studied, the size of palpable spleen, liver chemistry value, platelet count, prothrombin time, diameter of main portal vein and splenic length as assessed by ultrasonography were determined. Upper endoscopy was performed. Oesophageal varices (EV) and gastric varices (GV) were graded (EV, grade 1-4; GV, grade 1-3). In the predictive model, the EGV was classified into two grades (low, grade 1-2 EV or grade 1 GV; high, grade 3-4 EV or grade 2-3 GV).
Results: There were 53 patients with EGV and 39 patients without EGV as determined by endoscopy. Patients with EGV had a significantly higher degree of ascites and hepatic encephalopathy, lower platelet count and longer splenic length than those without EGV. Low platelet count and presence of ascites were the significant independent predictors for high-grade EGV (concordance rate 0.83). The optimal critical value for the platelet count was 150 x 10(9)/L. Of patients without thrombocytopenia and ascites, 37% had low-grade EGV but none had high-grade EGV, whereas 38 and 35% of patients with thrombocytopenia or ascites had low and high-grade EGV, respectively. Therefore, this predictive model for high-grade varices had a positive and negative predictive value of 35 and 100%, respectively.
Conclusion: Endoscopic screening for EGV was not necessary until thrombocytopenia or ascites occurred.