Sensitivity, specificity and predictive values of noninvasive markers of esophageal varices in cirrhosis of liver

Abstract

Objective: To assess the sensitivity, specificity, positive predictive values and negative predictive values of non invasive markers of esophageal
varices in liver cirrhosis.
Methodology: This two year validation study was started on 1st January 2010 after approval of ethical review committee. 739 cases of cirrhosis of
liver underwent upper GI endoscopy after informed written consent. Radiological parameters (liver size, portal vein, splenomegaly and ascites)
and biochemical/hematological parameters (bilirubin, INR, albumin, platelet count) was documented in all case. Data regarding grades of varices
and non invasive parameters was analyzed using SPSS 19.
Results: Serum albumin = 2.8 g/dl was 90.9% specific for varices in cirrhosis. Platelet count < 50,000/?L had specificity and PPV of 98.9% and 96.7%
with p value < 0.001 and odds of 8.19 (95% CI: 5.74 11.69). Serum bilirubin = 3 mg/dl was 33.4% and 78.9% sensitive and specific for varices
and the PPV and NPV was 63.7% and 51.6% respectively. INR > 2.2 had high predictive value for varices with sensitivity, specificity, PPV and
NPV of 60.4%, 91.7%, 89.0% and 67.6% respectively (p < 0.001) with odds of 16.89 (95% CI: 10.97 25.99). Portal vein diameter > 15 mm,
ascites, liver span < 8 cms and spleen size > 15 cms was 64.0%, 12.6%, 77.4% and 69.2% specific for varices (p < 0.001).
Conclusion: Platelet count < 50,000/?L is highly specific and sensitive predictive of varices, as is INR > 2.2. Biochemical markers have more
significant predictive value than radiological. Platelet count and INR may be used as a predictor of varices in cirrhosis of liver.