Reply : Understanding survival comparisons in non-randomized treatment comparisons for patients with early-stage HCC

We thank Dr. Khosla and colleagues for the reply¹ to our manuscript² and would like to address their comments.

First, the authors highlight the potential for differences in the stereotactic body radiation therapy (SBRT) technique among the included centers. We acknowledge site-level variability but would like to point out that we limited the study to specific tertiary centers highly experienced in SBRT. Technical details from each center are included in the supplementary materials. We considered the inclusion of several centers with different techniques as a strength since it improves generalizability, as variability can exist in the SBRT technique. Internationally, there are variable definitions for SBRT itself and no agreement on what defines ablative dose radiotherapy. There is a need for consensus definitions on ablative dose radiotherapy and more research on optimal radiotherapy treatment techniques.

Second, these authors point out that the increased risk of Child-Pugh score worsening ≥ 2 points among SBRT treatments compared with thermal ablation was statistically significant at 3 months but not 6 months. They suggest that the differences could have been attributable to the larger median lesion size for SBRT (2.4 cm) compared with ablation (2.1 cm), but differences remained statistically significant after adjustment for tumor size. Furthermore, the lower proportion of SBRT-treated patients with Child-Pugh increase at 6 months (11.1%) compared with 3 months (13.0%) could be in part attributable to the 13 SBRT-treated patients who experienced Child-Pugh increase at 3 months and died before the 6-month mark. While we were unable to ascertain the cause of death, it is possible that deaths related to liver toxicity in the SBRT group could have contributed to the lower proportion experiencing Child-Pugh increase at 6 months.

Lastly, Khosla et al raise concerns of residual confounding, including tumor biology and medical comorbidities. We agree with this point. Each institution clearly had biases in making their treatment decisions, and this certainly could impact outcomes. Ultimately, randomized controlled trials would be beneficial to assess the comparative effectiveness and toxicity of these treatments. There are challenges in completing such trials given that thermal ablation has been proposed as the ablative “treatment of choice” by many, including the recent American Association for the Study of Liver Diseases guidelines.³ Trials may be particularly challenging in populations with decompensated liver disease or comorbidities. Multicenter observational studies such as our study, while subject to biases, provide important information supporting the use of both thermal ablation and SBRT while we await prospective comparative effectiveness data.