Prospective validation of two models for ultrasonographic diagnosis of cirrhosis

Abstract

Objective: To perform a prospective validation and comparative analysis of two ultrasonographic diagnostic scores of cirrhosis in patients with silent liver disease.

Design: cross-sectional study, prospective and blind. ROC curves evaluated the diagnostic utility of: a) Bologna score (BS): assessment of liver surface nodularity and portal flow velocity, and b) Cadiz score (CS): assessment of liver echo structure, portal vein caliber and spleen area. Liver biopsy was considered the gold standard for the diagnosis of cirrhosis.

Patients: One hundred and thirteen patients, 76 men and 37 women, mean age 44 years old (range 18-73 years) referred for evaluation of chronic liver disease without clinical or biochemical evidence of advanced disease (absence of jaundice, ascites, encephalopathy, malnutrition or coagulopathy).

Results: Cirrhosis was diagnosed in 25 patients (22.1%). BS: sensitivity 84%, specificity 79.5%, area under the ROC curve 86.7%. CS: sensitivity 84%, specificity 89.8%, area under the ROC curve 92.4%. Portal vein was not displayed in 7 patients(6%) and portal flow velocity was not recorded in 13 (11.5%). These results agree with those obtained in the original articles developing both scores. There were no statistically significant differences between the two scores. Specificity reached 97% with joint use of both models, but sensitivity decreased to 72%.

Conclusions: Presence or absence of cirrhosis in patients with silent liver disease can be established by Doppler ultrasound with high diagnostic accuracy. The joint use of both scores has high diagnostic specificity. Both diagnostic models are highly re-producible.