LONG-TERM PREDICTION MODEL FOR HEPATOCELLULAR CARCINOMA IN PATIENTS WITH CHRONIC HEPATITIS B RECEIVING ANTIVIRAL THERAPY

Abstract

Background: Predicting the development of hepatocellular carcinoma (HCC) is a key clinical issue in patients with chronic hepatitis B (CHB). After the introduction of nucelos(t)ide analogues, the role of viral factors in the prediction of HCC decreased, suggesting that incorporation of the other factors such as liver stiffness is increasingly important. Moreover, long-term performance of the current HCC prediction models is still limited. The aim of this study was to develop an accurate, precise, and simple HCC risk score for up to 10 years.

Methods: CHB patients treated with entecavir or tenofovir disoproxil fumarate as a first antiviral therapy were retrospectively recruited. Exclusion criteria were 1) history of HCC, 2) lack of data from essential examinations at baseline, or 3) missing regular HCC surveillance examinations. Furthermore, patients without baseline liver stiffness data were excluded. Patients were randomized into derivation and validation cohorts at a 2:1 ratio. Variables proven to be independent risk factors for HCC in the derivation cohort were used to develop the prediction model.

Results: A total of 1,895 patients were randomized into the derivation (n=1,239) and validation (n=656) cohorts. In the multivariable analysis, the independent predictors for HCC were age [adjusted hazard ratio (aHR) = 1.771 at 40–49 years; aHR = 3.682 at ≥50 years; p <0.001], cirrhosis at baseline (aHR = 2.974, p <0.001), consumption of alcohol (aHR = 1.954, p = 0.003), stiffness of the liver (aHR = 1.314 at 9.7 ≤LS <14.9 kPa, aHR = 2.340 at LS ≥14.9 kPa, p = 0.001), and serum ALT <80 U/L (aHR = 1.633, p = 0.002). The ACCESS-HCC prediction model was developed using these five variables. The areas under the receiver operating characteristics curves were 0.798, 0.762, and 0.883 for HCC risk at 3, 5, and 10 years, respectively, which were higher than those of the other prediction models such as PAGE-B, mPAGE-B, REAL-B, and CAMD scores in this cohort. The scores were categorized according to significantly different HCC incidences: 0–4, low; 5–8, intermediate; and 9–14, high-risk. The performance of this model was validated in an independent cohort.

Conclusion: The ACCESS-HCC model shows improved long-term prediction and provides three distinct risk categories for HCC development in CHB patients receiving antiviral therapy.