Our data showed that radiographic TBS could be used to stratify the OS, both for the entire cohort and for BCLC stages A and B, of HCC patients undergoing LR. This information is useful for preoperative prognostic prediction. This study used the X-tile bioinformatics tool [27] to determine the optimal TBS cutoff values for OS (low TBS was defined as < 2.6, medium TBS as 2.6–7.9, and high TBS as > 7.9), which were different from those of Tsilimigras et al (low TBS was defined as < 3.36, medium as 3.36–13.74, and high as > 13.74) [9]. The discrepancies between the two studies may be due to different patient characteristics and, most importantly, the use of pathologically defined TBS by Tsilimigras et al. [9] in contrast to our use of radiographically defined TBS.
The current study showed that the TBS is higher in patients with HCC recurrence than in those without. Relatedly, patients with a higher TBS should have a higher risk of HCC recurrence, although we did not evaluate whether the TBS could predict HCC recurrence.
The BCLC staging system is a commonly used system for prognosis prediction in patients with HCC [3]. In the current study, we showed that the prognostic utility of the TBS is similar to that of the BCLC stage [3] in the prediction of the OS. Nevertheless, the TBS and BCLC stage [3] both play an important role in prognostic prediction, fulfilling separate but complementary needs, considering that the TBS could stratify the OS between the patients with BCLC stage A and B HCC.
Herein, we showed that as a single parameter, the radiographic TBS can be reliable and easy to calculate. The C-index of using the radiographic TBS to predict the 1-, 3-, and 5-year OS rates ranged between 0.61 and 0.78 in the derivation and validation cohorts, respectively. Among researchers who used the TBS and enrolled patients with HCC who underwent LR, Tsilimigras et al. showed that a preoperative model composed of the radiographic TBS, presence of cirrhosis, AFP level, and American Society of Anesthesiologists classification could stratify the OS of patients with BCLC stage B HCC; the C-index was 0.68 in the training cohort, 0.70 in the internal validation cohort, and 0.67 in the external validation cohort [10]. In another study, Tsilimigras et al. demonstrated that a model composed of the pathological TBS, AFP level, and presence of lymphovascular invasion could predict recurrence beyond the Milan criteria; the C-index of the model was 0.76 and 0.71 in the training and validation datasets, respectively [18]. Tsilimigras et al. also demonstrated that the serum AFP level and pathological TBS had a synergistic impact on the prognosis (the C-index was not mentioned in this study) [13]. Fukami et al. demonstrated that a model composed of the pathological TBS and Controlling Nutritional Status score could predict the OS (the C-index was also not mentioned in this study) [19]. Relatedly, the current study showed that the radiographic TBS alone is simple to use, could be used preoperatively, and has a predictive accuracy comparable to that of other complex models [10, 18].
We enrolled patients with BCLC stage B HCC in this study. The BCLC guidelines recommend transarterial chemo-embolization for BCLC stage B patients [3]. However, the selected patients with BCLC B (e.g., patients with multiple tumors located in one lobe of the liver or patients with the main tumor(s) located in one lobe of the liver and a small tumor located on the surface of the contralateral lobe of the liver) were eligible for LR [29]. We excluded patients with BCLC stage C HCC in the current study. Although several retrospective studies have enrolled patients with BCLC stage C HCC without extrahepatic spread but with intrahepatic portal invasion, their results showed that LR is superior to systemic therapies in terms of the yielded prognosis [30, 31]. The regimen of systemic therapies used in these studies included sorafenib [30], chemotherapy, and immunotherapy other than atezolizumab plus bevacizumab [31]. With the advent of systemic therapies, treatment with atezolizumab plus bevacizumab [32] is currently the standard of care for patients with BCLC stage C HCC [33]. The median OS rate was 19.2 months (95% CI = 17.0–23.7) with atezolizumab plus bevacizumab treatment in patients with systemic treatment-naive, unresectable HCC [34]. Thus, LR might not be currently favorable for patients with BCLC stage C HCC.
The discrepancy between radiographic and pathological tumor numbers was mainly in nodules determined to be < 1 cm either by radiographic or by pathological examination. Diagnosis of sub-centimeter HCC may be challenging [35]. Diagnostic performance in sub-centimeter HCC detection can be improved with gadoxetic acid-enhanced MRI [36]. However, MRI is not feasible for patients with inadequate breath-holding ability, claustrophobia, MRI contraindications, etc. Therefore, the American guidelines do not recommend MRI in preference to CT for diagnostic evaluation of patients with HCC [2].
The strength of this study is the use of the radiographic TBS rather than the pathological TBS. The radiographic TBS is useful for preoperative prognostic prediction. The study has several limitations: (1) it is a single institutional study, which may not be generalizable to other institutions; (2) there was no external validation; (3) the number of BCLC stage B cases in the validation cohort was small (n = 23), and, therefore, we did not analyze whether the TBS could be used to stratify the prognosis of the cohort in this stage; (4) the HCC registry data did not include pathological data, e.g., for vascular invasion, satellite tumors [37] and liver capsule involvement [12], which are well-known factors of poor prognosis of HCC patients; as well as information on comorbidities (e.g., diabetes), performance status, and details of liver function, such as platelet count and creatinine level, albumin–bilirubin grade [38], and intraoperative and postoperative parameters [39], which are associated with the prognosis of HCC patients [2, 3]. (5). In general, MRI is better at detecting HCC than CT, especially in terms of the tumor number. In the current study, we evaluated the radiographic TBS using CT or MRI; however, we did not clearly state whether each TBS was obtained using either CT or MRI. If the radiographic TBS is to be used to stratify the prognosis, the detection process should be standardized, preferably using MRI owing to its high sensitivity [40]. (6) We randomly selected 180 patients to determine the relationship between imaging and pathological findings in terms of tumor size and number. We did not evaluate all cases. However, there was no significant difference found in the background factors between the 180 randomly selected cases and the other cases, except for age and cirrhosis status.