Sonographic signs in portal hypertension: a multivariate analysis.

A prospective case controlled study was conducted to evaluate the value of real time ultrasound in establishing the diagnosis and the etiology of portal hypertension. Patients attending the outpatient department of a tertiary care centre were included. There were 324 consecutive patients with portal hypertension due to cirrhosis (n = 229), non cirrhotic portal fibrosis (NCPF: n = 64) and extrahepatic portal venous obstruction (EHPVO: n = 31). During this period, 146 patients with dyspepsia, 35 with splenomegaly and 32 with ascites due to varied causes served as negative and positive controls. Real time ultrasonography using a 3.5 MHz linear array scanner was performed in a fasting state in all subjects. Portal and splenic vein diameter greater than 10 mm, splenomegaly, hepatic and splenic hilar collaterals were suggestive of portal hypertension. Non visualization of the portal vein which was replaced by a cavernoma had a diagnostic accuracy of 98% in EHPVO. Splenic infarcts and absence of ascites were features of non cirrhotic portal hypertension (NCPF and EHPVO). Sonography had an overall diagnostic accuracy of 80%. A stepwise logistic regression with multivariate analysis using discriminate function showed that collaterals at the hepatic and splenic hilum, hepatomegaly, ascites and splenic infarcts were independent markers to differentiate cirrhotic from noncirrhotic causes of portal hypertension. The discriminate equation generated had a mismatch of 9.8%. Correlations between the sonographic signs demonstrated that the variceal grade correlated positively with the presence of splenic hilar collaterals and the liver size inversely correlated with presence of ascites. It was concluded that real time ultrasonography is an accurate method to establish the presence and etiology of portal hypertension.