Editorial
January 2019 JAMA Oncol. 2019
"There remains a concern that the observed difference in HCC rates is due to residual confounding. This is a limitation of all observational studies, but Choi and colleagues5 used a robust analytic approach that included multiple subgroup analyses, the use of propensity scores and competing risks, as well as validation in a hospital-based cohort having data on disease severity and virologic responses. Remarkably, the association of ETV vs TDF and HCC risk was minimally changed by any of these adjustments. However, adherence to antiviral therapy and surveillance programs was not captured and may have influenced rates of HCC. Indeed, adherence of less than 90% to antiviral therapy in patients with CHB has been associated with a higher risk of HCC.10 Moreover, especially in countries where prior lamivudine exposure is prevalent, patients treated with ETV may be at increased risk of virologic breakthrough with time, and this, in turn, affects risk for HCC. Indeed, in the study by Choi and colleagues, there was a substantially higher rate of changing therapy in patients treated with ETV (12%) vs TDF (0.2%). Might reduced adherence and increasing rates of ETV resistance represent unmeasured confounders that contribute to the divergence in rates of HCC evident 2 years after initiation of therapy? Ideally, a randomized study design would help overcome these challenges and provide stronger evidence of the association between type of antiviral agent and HCC outcomes. As such a study seems unlikely, at least in the near term, we would advocate for additional observational cohorts with data on liver disease severity, virologic response, and adherence to surveillance to evaluate this important question."