We developed a simple preoperative model to predict the OS of patients undergoing LR for BCLC stage B HCC. BCLC B1 was defined as tumor number ≤3 and MELD score ≤9. In our analysis using BCLC A2 (defined as a single large HCC >5.0 cm and BCLC A) as the comparator, the results showed that OS was comparable between BCLC B1 and BCLC A2, which suggests that patients with BCLC B1 HCC could be ideal candidates for LR. Although OS was comparable between BCLC A2 and B1, the RFS of BCLC A2 was better than that of B1. The discrepancy between OS and RFS may be due to post-recurrence survival being determined by recurrence modality, performance status, liver function reserve at recurrence, and treatment for recurrence [20]. However, the present study lacked data on these aspects.
Orimo et al. enrolled patients undergoing LR for BCLC stage 0, A, or B HCC. They divided patients with BCLC stage A or B into A1 (single tumor <5 cm or 2–3 tumors ≤3 cm), A2 (single tumor 5–10 cm), A3 (single tumor >10 cm), B1 (2–3 tumors >3 cm), and B2 (≥4 tumors). They found that OS was significantly different between stages A1 and A2 (p=0.012), between A2 and A3 (p=0.001), and between B1 and B2 (p=0.005), but not between A3 and B1 (p=0.474). They concluded that LR should be considered for BCLC B HCC if the number of tumors is ≤3 because OS was not significantly different between this group and the group with single HCC >10 cm.6 However, we believe that the comparison of OS between A3 (single HCC >10 cm) and B1 (2–3 tumors >3 cm) is not appropriate. As tumor size increases, the risk of microvascular invasion (MVI) increases [21]. Therefore, the prognosis of patients undergoing LR for large HCC >10 cm is poor [22]. It is not surprising that the OS of patients with single HCC >10 cm was comparable to that of patients with BCLC B1. Furthermore, only 75 (10%) patients with single HCC >10 cm were in BCLC stage A. Therefore, this subgroup of patients with BCLC stage A seems to be an outlier. In contrast, we used the criterion single large HCC >5.0 cm according to the BCLC guideline recommendation and our case number was higher (n=294 [29.0% of BCLC stage A]).
Child–Pugh B cirrhosis was the only independent factor associated with OS in the BCLC B1 group in Orimo et al.’s study. They thus concluded that LR should be considered for BCLC B HCC if the number of tumors is ≤3 in patients without cirrhosis or those with Child–Pugh class A liver disease [6]. Similar to our study, the ideal surgical candidate in BCLC stage B had ≤3 tumors and good liver function reserve, i.e., MELD score ≤9, although Orimo et al. adopted a different definition of good liver function reserve.
A Japanese study enrolled 489 patients with BCLC stage B HCC and Child–Pugh class A liver disease; 170 patients underwent LR and 319 TACE. Among patients with a Child–Pugh score of 5 and ≤3 tumors, the OS of the LR group was significantly better than that of the TACE group (p=0.014), whereas among patients with a Child–Pugh score of 6 or ≥4 tumors, treatment modality was not associated with OS (p=0.088).7 The results of this study are consistent with those of our study; both suggest that BCLC stage B patients with ≤3 tumors and good liver function reserve are ideal candidates for LR.
Multivariate analysis showed that a tumor number >3 was associated with worse 5-year OS of BCLC stage B patients in the present study. MVI is a well-known prognostic factor associated with patients with HCC undergoing LR [23]. Wang et al. enrolled 323 patients undergoing LR for BCLC stage B HCC. The presence of >2 nodules was one of the independent predictors of MVI [24]. Wang et al.’s [24] results could explain our findings: as the tumor number increased, the risk of MVI increased, which resulted in poorer OS of patients undergoing LR for BCLC B HCC.
The Japanese Society of Hepatology (JSH) guidelines recommend considering LR for patients with multiple tumors and a tumor number ≤3, whereas non-surgical treatments for >3 tumors [25]. An interesting finding is that the majority of patients undergoing LR for multiple tumors have ≤3 nodules, a phenomenon noted both in Eastern and Western centers. An Italian study enrolled 144 patients with multiple tumors (including tumors within and beyond Milan criteria) undergoing LR, of whom 118 (81.9%) had ≤3 nodules [8]. In an international study enrolling 164 patients with BCLC stage B, mainly from Western countries, 130 (79.3%) patients had ≤3 nodules [9]. In a Japanese study of 334 patients with BCLC stage B, 250 (74.9%) patients had ≤3 nodules [6]. In a Chinese study of 424 patients with BCLC stage B, 307 (72.4%) had ≤3 nodules [10]. In the present study, of the 170 patients with BCLC stage B, 129 (74.6%) patients had ≤3 nodules.
A previous study established the tumor burden score (TBS; γ), calculated with the largest tumor size (α) and tumor number (β) using the Pythagorean theorem (α2 + β2 = γ2) [26]. TBS can be used to stratify the OS of patients with BCLC stage B undergoing LR [9,11,12]. However, TBS cutoff values are inconsistent between studies. The TBS cutoffs to stratify OS were <3.36, 3.36–13.74, and >13.74 in Tsilimigras et al [11]. Alternatively, they were <5.83, 5.83–11.18, and >11.18 [9]. Lima et al. used cutoffs of ≤8 vs >8 [12]. With lessons learned from studies of TBS in patients with BCLC B HCC undergoing LR [9,11,12], we believed that commonly used cutoff values of prognostic variables we found in the literature should be used in the present study because a cutoff value created by us that is suitable for our cohort may not be applicable to other cohorts, thus limiting the clinical utility of our cutoff values. Therefore, the cutoff AFP level of 400 ng/ml was based on Cancer of the Liver Italian Program (CLIP) staging [27]. Tumor number ≤3 vs >3 was chosen based on JSH guidelines [25]. Liver function reserve is a well-known prognostic factor of patients undergoing LR for HCC. In Citterio et al.’s study of 543 patients with HCC undergoing LR, 508 (93.6%) were Child–Pugh class A and 35 (6.4%) were Child–Pugh class B. Multivariate analysis showed that a MELD score of >9 (odds ratio=2.26; 95% CI=1.10–4.58; p=0.02) was independently associated with postoperative liver decompensation, whereas a Child–Pugh classification of B was not (p=0.59). Citterio et al.’s [28] study is cited by the European Association for the Study of the Liver guideline [2]. Decompensation of liver disease (variceal bleeding, ascites, or encephalopathy) reflects non-preserved liver function irrespective of the Child–Pugh score [29]. Therefore, we believed that compensated liver function could be stratified with additional granularity by using a MELD score of >9 for patients with HCC undergoing LR.
The strength of the present study is its simplicity. We divided patients with BCLC stage B into two groups based on two variables, i.e., tumor number and MELD score. We used BCLC A2, i.e., single large HCC >5.0 cm, as the comparator; the results showed that OS was comparable between BCLC B1 and BCLC A2, which suggests that patients with BCLC B1 HCC could be ideal candidates for LR. Although there was no external validation of our model, two Japanese studies [6,7] showed results similar to ours, further supporting the feasibility of clinical application of our model.
Our study has several limitations. It was a retrospective and monocentric study. We did not compare the outcomes of BCLC stage B patients undergoing LR with those of patients undergoing TACE. The study lacked data on intra-operative parameters, i.e., blood loss, blood transfusion, open vs laparoscopic surgery [30,31], and postoperative complications, which may be associated with these patients’ outcomes [32].