Construction and Validation of a Novel Risk Index for Hepatocellular Carcinoma in Central European In-Patients

Abstract

Listen

Translate article

Introduction: Screening for hepatocellular carcinoma (HCC) is recommended for cirrhotic patients. Unfortunately, current screening tests have less than optimal sensitivity and specificity. Lack of resources precludes the use of more sensitive procedures such as magnetic resonance imaging in all patients. In this study, we develop and validate a screening index to discriminate patients with HCC from patients with cirrhosis but without HCC. Patients and methods: 409 consecutive in-patients with either liver cirrhosis or HCC were included in this study. Following univariate analysis to define differences between the patients with HCC and cirrhotic patients, stepwise logistic regression was performed to construct an index to differentiate between HCC patients and cirrhotics. This index was then validated using another unrelated cohort of 558 consecutive in-patients. 237 patients with HCC from the German national registry of HCC served as a further independent control group. Results: The screening index includes age, sex, ALT, the presence of ascites, portal vein thrombosis, markers of hepatitis B or C, and performance status. This index detected HCC patients (n=111) among the total validation cohort of 558 patients with a sensitivity of 96% (95%-confidence interval (CI): 91–99%). The negative predictive value was 98% (CI 94–99%) while the positive predictive value was 28%. Testing the score on yet another 237 HCC patients from the German National registry of HCC yielded a sensitivity of HCC detection of 94% (CI 89–96%). Similar results were obtained in the subgroup of AFP-negative patients. Conclusions: A simple score derived from seven readily available variables distinguishes between in-patients with HCC or cirrhosis. The index could be used to identify patients with a high likelihood for HCC who could then be further evaluated. Conversely, low risk individuals in whom costly procedures may not be indicated could be identified.