Eradication of hepatitis C virus (HCV) reduces, but does not eliminate, the risk of hepatocellular cancer (HCC).1-3 Many risk scores have been developed in the last few years.4 We developed and validated a simple scoring system called the General Evaluation Score (GES) to stratify patients according to the risk of HCC among those with compensated advanced chronic liver disease (cACLD) from chronic HCV infection who achieved sustained virologic response (SVR) after therapy with a direct-acting antiviral (DAA) agent, based on pre-treatment data.5, 6 We further studied the dynamics of the GES, investigating and evaluating its use after completion of DAA treatment, and developed an algorithm using GES scores, pre- and post-treatment.7

We thus read with great interest the article from Tahata et al.8 They proposed a model based on four variables (baseline BMI, baseline FIB-4 index, albumin level at SVR, and alpha-foetoprotein [AFP] level at SVR) to construct a prediction model for HCC occurrence.

In response to the authors' recommendation to evaluate their model in different patient populations, we validated the model in 938 Egyptian patients with chronic hepatitis C genotype 4, and with cirrhosis (F4; 570 patients) or advanced fibrosis (F3; 368 patients), who had achieved SVR following DAAs. Duration of follow-up was 6–72 months after the end of treatment. HCC developed in 54 patients during the study period. Figure 1A shows cumulative HCC occurrence rates after SVR according to the model, with log rank P value of 0.010 and Harrell’s c statistic of 0.6600.

Cumulative HCC occurrence rate after SVR according to (A) published model, (B) modified model

The use of the six levels model was found difficult in our cohort. Therefore, we propose classification of the scale of model into three risk groups; low (0-1), intermediate (2-3) and high (4-5). The results are presented in Figure 1B. HCC incidence was 0.83/100 patient-years (py) (95% CI = 0.30-1.83) in the low-risk group, 2.57/100 py (95% CI = 1.63-3.86) in the intermediate-risk, and 4.85/100 py (95% CI = 3.29-6.92) in the high-risk group. Log-rank P value changed to 0.002, and Harrell’s c statistic is 0.6600.

Two additional points should be clarified. First, in Japan, where the model was developed, obesity rates are estimated to be 4.3%. However, in Egypt, it is estimated to be 32.0% according to the most recent data available from the World Health Organization (WHO) as of 26 March 2020.9 Differences in BMI among different populations may interfere with the accuracy of the HCC prediction model.

Second, dynamic models that incorporate changes in clinical variables and risk biomarkers over time are usually more predictive of HCC incidence than simple models.4 The inclusion of AFP at SVR into this model was a strength that was not clearly emphasised. This was recently supported by Kuwano et al who concluded that AFP at end of treatment is a useful predictor of HCC occurrence.10

Declaration of personal interests: None.