Background and Aims
The modified PAGE‐B (mPAGE‐B) score comprising age, gender, platelet count and albumin was recently proposed to predict hepatocellular carcinoma (HCC) risk among chronic hepatitis B (CHB) patients undergoing antivirals. Here, in the independent cohort, we externally validated the predictive performance of the mPAGE‐B score and compared it with those of conventional HCC prediction models.
Methods
We consecutively recruited CHB patients treated with lamivudine, entecavir or tenofovir as the first‐line antiviral regimen. Patients with decompensated cirrhosis or HCC at baseline were excluded. Predictive performances of the mPAGE‐B score and other models were assessed with comparison.
Results
Among 1330 patients, 9.6% developed HCC during follow‐up. The mPAGE‐B score provided the highest Harrell's c‐index (0.769), followed by the GAG‐HCC (0.751), PAGE (0.744), REACH‐B (0.686) and CU‐HCC (0.618) scores. The mPAGE‐B score showed the similar performance to the PAGE‐B and GAG‐HCC scores and the better performance than the REACH‐B and CU‐HCC scores. Cumulative HCC probabilities at 5‐ and 7‐years were 0.0% and 0.0% in low‐risk group (mPAGE‐B score ≤ 8), 6.1% and 10.8% in intermediate‐risk group (mPAGE‐B score 9‐12) and 18.7% and 26.7% in high‐risk group (mPAGE‐B score ≥ 13) respectively (both P < 0.001 between adjacent two groups). C‐indices of the mPAGE‐B score were 0.785 and 0.724 among subgroups treated with entecavir or tenofovir (n = 1011) and with lamivudine (n = 319), respectively, which are overall similar to those of the PAGE‐B score.
Conclusion
The mPAGE‐B score showed acceptable predictive performances. Compared to the PAGE‐B score, addition of albumin as a constituent provided the marginal benefit.