Treating Very Early-stage HCC: Have We Found the Holy Grail?
We thank the authors for their comments on our article published in the Annals of Surgery.1 They claimed that the study setting was not strong enough to support the preference for surgical resection (SR) over radio frequency ablation (RFA) in patients with Barcelona Clinic Liver Cancer very early-stage hepatocellular carcinoma (HCC). Critical appraisals open the chances for further investigations, which are beneficial to the entire hepatology community. However, we would like to address the specific points raised by the author.
Cirrhosis, strictly speaking, is a histological diagnosis. Clinically, the diagnosis of cirrhosis can be obtained from a combination of clinical, laboratory, and imaging features. In fact, a variable proportion of patients with HCC do not have cirrhosis. Some other studies reporting outcomes in very early or early-stage HCC did not report the percentages of patients with cirrhosis.2,3 It is not clear whether the Child-Turcotte-Pugh (CTP) grade is adequate for the assessment of liver dysfunction in patients with HCC. However, the CTP classification is still widely used for stratification of staging systems given that the degree of liver dysfunction is one of the most important prognostic factors for HCC. Moreover, the CTP system is an integral part of the Barcelona Clinic Liver Cancer classification scheme. In our study, 181 patients were CTP class A5, whereas 56 patients were class A6. The recently proposed albumin-bilirubin (ALBI) grade may allow better discrimination of liver function for this group of patients.4 By this grading system, 133 patients in the study were ALBI grade 1, whereas 104 patients were classified as ALBI grade 2. There are no significant differences in distribution of CTP class A5/A6 and ALBI grade 1/2 between patients receiving SR or RFA, both in all patients and in patients selected in the propensity model.
The propensity score matching analysis is used to reduce bias in an observational study. In practice, the success of propensity score modeling is judged by whether balance on covariate values is achieved after matching.5 Because of this, investigators can be more liberal with the inclusion of the covariates, taking care that the number of covariates used must be much less than the number of participants in each treatment group.5 Indeed, many of the covariates were not independent to each other. For example, serum albumin and bilirubin level and blood platelet counts and prothrombin time are all measurements for liver dysfunction. To evaluate the success of propensity score matching, a common technique is to compare covariates before and after matching.5 The comparison of baseline demographics showed no inclination to either SR or RFA group after matching. The nonparametric tests were used instead of t statistics because no assumptions about the probability distributions of the variables were assessed. Even when the t tests were applied, there were no statistically significant differences between the 2 matched cohorts. Finally, we did not include serum albumin level in the univariate and multivariate analyses of survival. If albumin level was included, using the median value of 4.0 g/dL as the cutoff point, the hazard ratio for mortality in patients with serum albumin less than 4.0 g/dL would be 2.033 (P = 0.031). When serum albumin level was introduced into multivariate analysis, the remaining significant predictor of adverse outcome was still blood platelet count less than 100,000/μL, total tumor volume of 2 cm3 or more, and model for end-stage liver disease score of 10 or more. Albumin less than 4.0 g/dL was not a significant predictor for adverse outcome in other analyses shown in Tables 3 and 4. In univariate and multivariate analyses in the propensity model, RFA therapy was an independent predictor for mortality and tumor recurrence.