TO THE EDITOR: We read with great interest the report by Riaz et al, in which they demonstrated that the response of serum alphafetoprotein (AFP) level after locoregional therapy in patients with hepatocellular carcinoma (HCC) was a useful surrogate marker to determinate tumor response and survival. In this study, AFP response was seen in 81 (64.8%) of 125 patients treated with chemoembolization and radioembolization. Also, the authors reported that AFP response was a prognostic factor for overall survival in univariate and multivariate analyses. It is still unclear whether serial AFP measurement or AFP/tumor volume for monitoring treatment response in patients with HCC should be used. Some studies have shown that serum AFP level might be used as a surrogate marker for response to systemic chemotherapy. A significant correlation between serum AFP levels and both size and number of HCC lesions has been reported. AFP/tumor volume may be a better prognostic indicator for patients with HCC than serum AFP value alone. Also, significant correlation between higher AFP levels and higher stages has been reported. Because the results of a multivariate analysis give important messages to readers, the analysis should be carefully used. For accurately interpreted results of a study, the sample size should be sufficient to perform multivariate analysis. Some authors have recommended that a minimum of 20 patients per predictor should be used for an accurately interpreted multivariate analysis. Also, each parameter considered to be a prognostic factor must be evaluated before univariate or multivariate analysis is performed, because one factor may interact with the others. We think that the statistical assessment is the most important limitation of the study by Riaz et al. The study included a total of 125 patients and six prognostic factors, such as stage, performance status, Child Pugh class, WHO response, European Association for the Study of the Liver response, and AFP response, which were evaluated for multivariate analysis. We think that the sample size of the study was quite small for multivariate analysis, and the interaction probability among prognostic factors analyzed in the study was quite high. We believe that this problematic issue may have affected the results presented in the report.