Response to developing a feasible classification model for surgical hepatocellular carcinoma: More questions than answers

Response to Professor Hou: Thank you very much for your valuable feedback.1 When this paper2 was accepted, we anticipated a range of opinions from various clinicians. In the field of pancreatic cancer, resectability classifications are now widely used, and preoperative chemotherapy is administered to patients with borderline resectable (BR) pancreatic cancer. In Japan, resectability classifications have also become a topic of interest in hepatocellular carcinoma (HCC), and the Japanese Society of Hepato-Biliary-Pancreatic Surgery has proposed an oncological resectability classification for HCC.3 Although we proposed a resectability classification for HCC, it was developed based on retrospective data from two institutions and does not have a high level of evidence. It was introduced merely as a classification model with the hope that larger datasets would eventually enable the development of a more robust classification with strong correlations with safety and prognosis. As you know, HCC has oncological characteristics that are distinct from those of pancreatic cancer. During surgery, it is essential to consider the liver reserve function and future liver remnant (FLR) to avoid postoperative liver failure. Therefore, we believe that any classification of HCC resectability must combine surgical factors that consider operative safety, with oncological factors that account for the risk of postoperative recurrence. In this study, we used modified albumin–bilirubin (mALBI) for liver reserve evaluation as a surgical factor. As you have pointed out, the indocyanine green (ICG) test has been widely used in Asia, particularly in Japan, to assess liver reserves. However, in Western countries, the ICG test is not common. The ICG test is not feasible for certain patients because factors such as portal blood flow and biliary obstruction can affect the results, and the ICG test cannot accurately assess liver function in patients with a splenorenal shunt or biliary obstruction.4 It is also not suitable for liver function evaluation in patients with ICG excretory defects. Additionally, the mALBI can be used to assess liver function in patients who are unable to undergo ICG testing. We have also mentioned this in the Discussion section of our paper.2 The ICG test requires a reagent, which adds to the cost, and blood sampling is necessary at multiple points postinjection (5, 10, and 15 min after injection), making it more complicated than mALBI. Therefore, many Japanese physicians now prefer the mALBI for liver reserve assessment, and its ease of use and cost-effectiveness make it an ideal classification model. In Japan, ICG-Krem, which considers liver reserve and FLR, is widely used to evaluate surgical safety.5 In response to your query, we examined whether ICG-Krem or FLR/mALBI was more effective in selecting patients in the BR group using data from the training and validation cohorts. We compared ICG-Krem and FLR/mALBI as predictors of posthepatectomy liver failure (PHLF) grade B/C incidence in a cohort of 310 patients (resectable [R], BR, and unresectable treated with operation groups) at Hyogo Medical University. The incidence of PHLF grade B/C was 15.5% (n = 48) for patients with ICG-Krem ≥0.05 and 57.1% (n = 4) for patients with ICG-Krem <0.05 (p = 0.003). Similarly, FLR/mALBI had an incidence of 10.0% (n = 26) for values ≥0.323 and 44.1% (n = 26) for values <0.323 (p < 0.001). Receiver operating characteristic (ROC) analysis showed no significant difference between ICG-Krem (area under the curve [AUC] = 0.800) and FLR/mALBI (AUC = 0.805, p = 0.869). Next, in the validation cohort (295 cases) at Kyoto University, ICG-Krem ≥0.05 was associated with a PHLF grade B/C incidence of 9.5% (n = 27), while ICG-Krem <0.05 showed an incidence of 40% (n = 4) (p = 0.002). The incidence of FLR/mALBI was 7.9% (n = 17) for values ≥0.323 and 17.7% (n = 14) for values <0.323 (p = 0.015). The ROC analysis revealed no significant differences between ICG-Krem (AUC = 0.696) and FLR/mALBI (AUC = 0.666; p = 0.518). Consequently, there was no statistically significant difference in the predictive accuracies of ICG-Krem and FLR/mALBI. However, because FLR/mALBI <0.323 identified a greater number of PHLF grade B/C cases in both cohorts, we considered FLR/mALBI <0.323 as a more effective marker. Our resectability classification uses both surgical and oncological factors. The BR group based on surgical factors included patients with a high risk of complications (e.g., PHLF), whereas the BR group based on oncological factors included patients with a high risk of recurrence despite upfront surgery. Tumor size was used as an oncological factor, as reports indicate that tumors ≥7 cm are prone to lung metastasis, even without macrovascular invasion (MVI).6 In our classification, cases with tumor sizes ≥10 cm are operable if FLR/mALBI is ≥0.323, although they fall into the oncological BR group due to a high risk of recurrence. In our study, the 1-year recurrence-free survival (RFS) rate in the R group was 81.4% in the training cohort and 79.6% in the validation cohort, indicating that curative resection was beneficial for this group. However, in the BR group, the 1-year RFS was 37.8% in the training cohort and 55.4% in the validation cohort, indicating a poorer prognosis. In Asia, surgeries are often performed for BR cases, yet we believe that, as in BR pancreatic cancer cases, preoperative treatment should be considered for BR-HCC cases. Such preoperative treatment may have improved outcomes in the BR group, warranting further investigation. Our shared goal was to improve the treatment outcomes for patients with HCC. However, we may not distinguish between (1) ensuring surgical safety and (2) improving prognosis by reducing recurrence rates. Some surgeons believed that the approach that prioritizes performing safe surgeries by focusing on surgical factors might lead to a decline in postoperative outcomes by excluding cases that could potentially be treated with curative surgery. In contrast, we first considered the tumor condition to determine the surgical approach. We then evaluated whether this approach presented a high risk of postoperative liver failure by analyzing surgical factors, thereby distinguishing between the surgical R and BR groups. Next, within the surgical R group, we identified the oncological BR group—those with a high risk of postoperative recurrence—by examining MVI and tumor size. We believe that it is essential to distinguish between (1) ensuring safe surgery (surgical factors) and (2) reducing postoperative recurrence rates to improve treatment outcomes (oncological factors) and to identify BR cases that are eligible for preoperative treatment. We believe that administering appropriate preoperative treatment to the BR group may enhance postoperative outcomes. Authors declare no conflict of interests for this article.