Predictive value of ultrasonography in the screening of non-ascitic cirrhotic patients with large varices.

To evaluate the accuracy of ultrasonography in the screening of cirrhotic patients with large varices, the receiver operator characteristic (ROC) curves of the portal, mesenteric, and splenic veins of 215 consecutive non-ascitic patients were plotted in order to select the most indicative optimal threshold. The contribution of respiratory variations of splenic and mesenteric veins was also evaluated. Taking 13 mm as the limit for the normal diameter of the portal vein, 10 and 9 mm for the upper limits of the mesenteric and splenic veins, respectively, and respiratory variations being absent, the positive rates were 91, 88, and 92 per cent and the false-positive rates were 44, 38, and 42 per cent, respectively. The predictive values of a positive test were 0.34, 0.33, and 0.34, respectively, and the predictive values of a negative test were 0.96 for the portal and mesenteric veins and 0.97 for the splenic vein. Because the portal vein is always visible, the portal diameter may be employed in the screening through the presence of large varices in association with the assessment of respiratory variation in the splenic or mesenteric diameter. A cut-off point of 13 mm and the absence of respiratory variations precludes the need for endoscopy in 47 per cent of patients.