A sustained virologic response (SVR) after direct-acting anti-viral treatment (DAA) significantly reduces, but does not eliminate, the occurrence of hepatocellular carcinoma (HCC).1-3 Therefore, current guidelines recommend indefinite HCC surveillance in patients with cirrhosis, and possibly, advanced fibrosis, following DAA-mediated SVR.4-6 Risk prediction scores that more accurately stratify and inform future HCC risk post-SVR are sorely needed.

The study by Tahata et al7 evaluated the risk of HCC in a multicentre prospective cohort of 2209 Japanese patients after a DAA-mediated SVR. Median follow-up was 40 months, the median age was 68-69 years, only 12%-14% had cirrhosis, and 41% had a FIB-4 ≥3.25. Exclusions included Child-Pugh B/C cirrhosis, HCC <6 months post-SVR, and HBV/HIV co-infection. Cumulative HCC occurrence rates at 2, 3 and 4 years were 1.2%-2.1%, 2.5%-3.6% and 4.4%-4.5%, respectively. A BMI ≥25.0 kg/m2, baseline FIB-4 ≥3.25, albumin <4.0 g/dl at SVR and alpha-fetoprotein ≥5.0 ng/ml at SVR were independently associated with HCC occurrence, and were used to generate a prediction model. The optimal cutoff score of ≥2 (range 0-5) was associated with significantly higher HCC occurrence rates, with a sensitivity of 82%-85% and negative predictive value >94%, suggesting an efficient surveillance strategy for HCC post-DAA-mediated SVR.

Can this model be widely used to stratify future HCC risk and HCC surveillance? Several considerations may reduce more widespread applicability of this policy. HCC screening in Japan is performed in all patients with HCV infection, irrespective of fibrosis stage, contrary to international guidelines.8 This cohort was also older, conferring a higher HCC risk. The proportion with cirrhosis, defined by “compatible histology or imaging findings” was probably underestimated, affecting interpretation of results. A FIB-4 ≥3.25 was a major determinant of the prediction score. Although conventionally suggestive of advanced fibrosis, FIB-4 performs less well in older patients as age is an element of the FIB-4 score.9 Other non-invasive methods to determine fibrosis stage may be useful to refine the model, and define which patients truly require HCC surveillance. The low positive predictive value of this prediction model (4%-21%) further limits its utility but, as with other predictive models, may be useful to exclude patients not requiring HCC surveillance. Pre-test probabilities will alter the performance of this model.

Incident HCCs occurring 6-12 months post-SVR were included; such patients had been excluded in other post-SVR studies, as it is thought likely the HCC was present before DAA therapy. An analysis excluding these patients may further refine the model. Co-factors such as alcohol consumption and metabolic-associated fatty liver disease (MAFLD) were not assessed, but baseline BMI was independently associated with HCC, suggesting that MAFLD may be a significant confounder. Finally, longer term follow-up is required to determine how long the risk of HCC persists post-SVR, to crucially inform duration of HCC screening.

On balance, we believe the model requires validation and cost-effectiveness analyses in other populations with HCV infection; caution is required before applying this prediction model beyond the Japanese setting. Further studies are needed to define optimal prediction models to better inform and stratify future HCC risk among persons achieving DAA-mediated SVR.

Declaration of personal interests: Jacinta A. Holmes has as a speaker for Gilead and AbbVie, and an advisory board member for CSL. Jacinta A. Holmes and Sarah Romero are employees of St Vincent's Hospital, Melbourne, Australia.