A comparison of prognoses between surgical resection and radiofrequency ablation therapy for patients with hepatocellular carcinoma and esophagogastric varices

There has been insufficient investigation of the differences in long-term outcomes between surgical resection (SR) and radiofrequency ablation (RFA) among patients with hepatocellular carcinoma (HCC) and esophagogastric varices (EGV). We retrospectively enrolled 251 patients with treatment-naïve HCC and EGV who underwent SR or RFA as a first-line treatment. Prognostic factors were analyzed using a Cox proportional hazards model. A total of 68 patients underwent SR, and the remaining 183 patients received RFA. Patients who underwent SR were younger, had better liver functional reserves, and had larger tumors. After a median follow-up duration of 45.1 months, 151 patients died. The cumulative 5-year overall survival (OS) rate was significantly higher among patients who underwent SR than those treated with RFA (66.7% vs. 36.8%, p < 0.001). Multivariate analysis showed that age > 65 years, multiple tumors, RFA, albumin bilirubin grade > 1, and the occurrence of major peri-procedural morbidity were the independent risk factors that are predictive of poor OS. In conclusion, SR could be recommended as a first-line treatment modality for HCC patients with EGV if the patients are carefully selected and liver function is well preserved.


Results
Baseline clinical characteristics. The baseline demographic data of the patients examined are shown in Table 1. Patients who underwent SR were younger than those who received RFA. Both groups were predominantly male, but the ratio of males-to-females was higher in the SR group. Chronic HBV infection was more prevalent in the SR group than the RFA group (58.8% vs. 38.3%, p = 0.005). Liver functional reserves were better in the SR group, which had lower scores in the model for end-stage liver disease (MELD), more patients with Child-Pugh Grade A, more patients with albumin bilirubin (ALBI) grade 1, higher serum albumin levels, lower bilirubin levels, lower prothrombin time international normalized ratios (PT INRs), and higher platelet counts. Furthermore, tumor sizes were larger in the SR group than the RFA group (median size: 3.2 cm vs. 2.2 cm, p < 0.001).
There were 251 patients who had EGV that was confirmed by an esophagogastroduodenoscopy (EGD). Regarding EGV status, 209 of these patients had esophageal varices (EV) alone at the time of HCC diagnosis, while 2 patients had gastric varices (GV) alone. The remaining 40 patients had both EV and GV. Moreover, 131 (52.2%) patients received prophylaxis therapy for EV bleeding, including 21 patients with non-selective betablockers (NSBB), 75 patients with esophageal variceal ligation (EVL) therapy and 35 patients with NSBB and EVL combination therapy.
Compared to the SR group, the RFA group had more cases of high-risk EV (68.3% vs. 55.9%, p = 0.092), and more patients received prophylaxis therapy for EV bleeding (57.9% vs. 36.8%, p = 0.005). However, among those who had high-risk varices, 23 patients (60.5%) patients in the SR group received prophylaxis therapy, while 92 patients (73.6%) received it in the RFA group (p = 0.179).
The safety of SR and RFA in HCC patients with EGV. No patients in our cohort died during the operations, both the SR group and the RFA group. As shown in Tables 1 and 2, there were 40 patients (15.9%) who developed peri-procedural morbidity and 15 patients (6.0%) who had major morbidity. The SR group had more peri-procedural morbidity than the RFA group (35.3% vs. 8.7%, p < 0.001), but the rates of major morbidity were comparable between both groups (8.8% vs. 4.9%, p = 0.390). The 90-day mortality rates were 2.9% and 1.1% in the SR group and RFA group, respectively (p = 0.371).
Among the 68 patients in the SR group, 63 patients underwent conventional open liver resection (OLR), including 15 patients with major resection (defined as resection of three or more segments). Among the 5 patients who underwent laparoscopic liver resection (LLR), 2 patients had two segmentectomies, and the other 3 patients had one segmentectomy. The patients who underwent major resection had more major morbidity than those who received minor hepatectomy (

Discussion
This study shows that SR could provide acceptable long-term outcomes for patients with HCC and EGV. The 10-year cumulative OS rates were 57.4% and 15.9% for patients who underwent SR and RFA, respectively. The survival benefits of SR over RFA were confirmed by the multivariate analysis. Moreover, SR could provide a lower rate of recurrence and a higher RFS rate than RFA. This indicates that SR is not contraindicated for HCC patients with EGV. On the contrary, it could have a survival advantage over RFA if patients are carefully selected. The presence of EGV, is a surrogate for CSPH and has been validated as an independent factor for poor prognosis among patients with HCC 18,31,32 . Several studies show that the incidence of developing liver decompensation after SR is high among HCC patients with CSPH, which would increase the risk of mortality 24,25,33 (Table 5). Consequently, it has been suggested that SR be reserved for patients without CSPH in the current guidelines for the management of HCC, whereas RFA is recommended for HCC patients with CSPH 23,34 . www.nature.com/scientificreports/ Nevertheless, recent technical innovations in surgical techniques, anesthesia, critical care, and spatial understanding of the intra-hepatic anatomy of the liver have led to an increasing number of liver resections, fewer post-operative hepatic failures, and lower treatment-related mortality 26,35 . As shown in Table 5, several studies from Eastern and Western countries have validated that CSPH alone is not a contraindication for SR 33,36,37 These findings suggest that the indications for SR could be extended to HCC patients with CSPH or EGV if they have well-preserved liver function. www.nature.com/scientificreports/ For patients with early-stage HCC, RFA is relatively safe and has lower costs, less serious adverse effects, and less destruction of non-neoplastic tissue than SR 7 . Moreover, it could provide an acceptable long-term OS for certain patients [38][39][40] . It has been reported that 5-year accumulative OS rates over 60% could be achieved with RFA among HCC patients with early-stage HCC 12,[38][39][40] . In our previous study, the 5-year and 10-year OS rates after RFA were 63.1% and 48.7%, respectively 40 . Consequently, RFA is regarded as a curative treatment modality for HCC patients 7 .
However, the recurrence rate after RFA is still high. For example, the 10-year RFS rate after RFA was only 12.4% in our previous report 40 . This might be caused by the incomplete ablation of liver tumors due to insufficient ablation-needle technology, tissue cooling by the neighboring blood vessels (through a heat sink effect), large tumor masses, and the ablation of tumors in close proximity to heat-sensitive organs 41 .
In contrast, SR could have a higher chance of complete excision of not only tumor tissue but also the hepatic parenchyma around the tumor, which might have microvascular invasion and micro-metastases 13,42 . Therefore, it could result in better local tumor control than RFA. However, the risk of liver decompensation and mortality after the operation are concerns when performing SR for HCC patients with a poorer liver functional reserve. Several studies compared the outcomes between SR and RFA for patients with early-stage HCC [12][13][14][15][16]43 . Most of these studies demonstrated that SR could reduce the risk of recurrence and might provide superior OS to RFA, although some studies observed that the OS rates were comparable between SR and RFA 12 .
Nevertheless, the differences in prognosis between SR and RFA for HCC patients with CSPH have not yet been investigated sufficiently. Qiu et al. demonstrated that SR is safe and could provide a better OS and RFS than ablation therapy for patients with HBV-related HCC and CSPH 30 . However, the patients enrolled in that study were limited to those with HBV-related HCC, and the ablation therapies included both RFA (79 patients) and microwave ablation (57 patients).
In the current study, we enrolled HCC patients from all etiologies and compared the prognoses between patients who underwent SR and RFA. In the SR group, no patient expired during the surgeries, and only one patient died within one month due to post-operative liver failure. The patients who underwent SR had a higher rate of post-procedure morbidity than those in the RFA group, but the rates of major morbidity and 90-day mortality were not statistically different between the two groups. Notably, our cohort revealed that SR could offer better long-term prognoses than RFA among HCC patients with EGV in terms of OS and recurrence. This was further validated by the multivariate analysis. The results could provide robust evidence for performing SR as a front-line treatment modality for patients with HCC and EGV if they have well-preserved liver function.
The patients in our cohort who underwent RFA as the primary treatment modality were older and had a poorer liver functional reserve than those who underwent SR. Older patients might choose RFA because of the greater chance of comorbidities than younger patients. RFA features less invasiveness, a lower complication rate, and lower costs, as well as higher repeatability in the event of recurrence 44 . This finding is similar to those of a www.nature.com/scientificreports/ nationwide cohort from Japan 45 . However, the survival benefit of SR over RFA was persistent after adjusting for the confounding factors for prognosis in the multivariate analysis. There were some limitations to our study. First, although EGD was recommended to screen for EGV among patients with a new diagnosis of HCC, the completion rate was only 48.6% in this cohort (Fig. 3). Selection bias might be present because of the retrospective nature of the study. Second, measuring HVPG levels is the gold  www.nature.com/scientificreports/ standard for assessing the degree of portal hypertension. However, it is invasive, expensive and not feasible in most medical centers. We did not perform HVPG measurement in out cohort. Furthermore, the spleen diameter was not recorded uniformly due to the retrospective study design. Therefore, we used the presence of EGV diagnosed by EGD as a surrogate for CSPH, which is more practical in the daily practice. Third, decisions made for treatment were shared between the physicians and patients. This patient-tailored approach is based on the multidisciplinary evaluation of each case and includes any alternative treatment options. This might have caused the significant demographic difference between the two groups of patients. Fourth, LLR is a recent technical innovation in SR for the treatment of HCC. It could provide shorter hospitalization, less blood loss, less wound pain, and a lower rate of postoperative liver failure and ascites formation than conventional OLR for cirrhotic patients with HCC [46][47][48] . Regarding the long-term outcomes, HCC patients who underwent LLR had similar OS and RFS rates to those who received OLR [46][47][48][49] . However, we could not compare the treatment efficacy and outcomes between LLR and OLR because the majority of patients in our cohort underwent OLR and only 5 patients received LLR. Further prospective studies are warranted to elucidate this issue. Lastly, this study enrolled HCC patients over a relatively long span of time, so the diagnoses, assessment of HCC patients, and SR and RFA techniques might not have been the same between different time periods. Despite these limitations, this study provides robust evidence to reassure physicians that SR could serve as a first-line treatment option for HCC, in spite of evidence of CSPH and EGV. However, the patients should be selected carefully. Consequently, aggressive treatments beyond the current guidelines could be considered when clinically applicable to achieve the maximum survival benefit.

Conclusion
SR could be recommended as the first-line treatment modality for HCC patients with EGV if the patients are carefully selected and liver function is well preserved.

Methods
Patients. This study retrospectively enrolled 251 treatment-naïve HCC patients who underwent SR or RFA as the first treatment modality for HCC. All of the patients had EGV diagnosed by EGD at the time of HCC diagnosis at Taipei Veterans General Hospital (Fig. 3). EV was classified as follows: F1: small and straight varices; F2: moderately sized, tortuous varices; and F3: large, tumorous varices 50 . High-risk EV was defined by the F2 and F3 classifications or by F1 accompanied by red coloring 51 .
The diagnosis of HCC was established based on the criteria from the American Association for the Study of Liver Disease consensus 52 . Multidisciplinary expert meetings and an HCC registration system in Taipei Veterans General Hospital have been conducted since October 2007. The diagnosis and treatment strategy are discussed at weekly multidisciplinary meetings for all of the HCC patients at this hsopital 53 . The decision about the treatment modality is shared between the patient and the physician after discussing the risks, benefits, complications, and efficacies of the available treatments, as well as the recommendations from the multidisciplinary expert meetings.
In our center, the indications for SR are as follows: (1) Child-Pugh grade of liver function of A or B, with an indocyanine-green 15-min retention rate (ICG-15R) of ≤ 30%; (2) tumor involving no more than two Healey's segments without involvement of the main portal vein trunk; and (3) the absence of extra-hepatic tumor dissemination 17 . RFA was performed in cases of (1) a single tumor with size < 5 cm or 2-3 tumors < 3 cm; (2) no extra-hepatic metastasis or major vascular invasion; (3) Child-Pugh grade A or B; (4) platelet count > 50,000 / mm 3 ; (5) the absence of ascites; and (6) no other comorbid diseases that might complicate RFA 40 . Liver transplantation was indicated for patients with end-stage liver disease, HCC, or acute liver failure according to the   54 . For patients with HCC, the tumor size criterion for liver transplantation was based on the criteria of the University of California San Francisco (UCSF) 55 . The indications for SR and RFA were not the same, so we selected HCC patients using the Milan criteria for this study. Consequently, the inclusion criteria were as follows: (1) solitary tumor with size < 5 cm, in which anatomic resection could be achieved after through evaluation or 2-3 tumors < 3 cm; (2) no extra-hepatic metastasis or major vascular invasion; (3) grade A or B Child-Pugh classification of liver function; (4) platelet count > 50,000 /mm 3 ; (5) absence of ascites; and (6) no other comorbid diseases that might complicate SR or RFA (e.g., infection, arrhythmia, acute myocardial infarction, uncontrolled congestive heart failure, chronic obstructive pulmonary disease with acute exacerbation, recent stroke, etc.).
There were 2958 consecutive patients who were enrolled in the HCC registration system database from 2008 to 2017. As shown in Fig. 3, 1465 of these patients received an EGD, and 625 patients had EGV at the time of HCC diagnosis. Patients were excluded for having tumor size > 5 cm, tumor number > 3, ascites or major vascular invasion. After the exclusion, 51 patients were enrolled in the SR group, and 163 patients were enrolled in the RFA group. Moreover, we also retrospectively recruited 17 patients in the SR group and 20 patients in the RFA group who fulfilled the inclusion criteria between 2003 and 2017 before the establishment of the HCC registration system. Consequently, a total of 68 patients who underwent SR and 183 patients who received RFA were enrolled in the final analysis.
After SR or RFA, the peri-procedural morbidities were recorded and graded by the Clavien-Dindo classification 56 . Grade III-V complications were defined as major morbidities. Postoperative liver failure was defined according to the International Study Group of Liver Surgery (ISGLS) 57 . All of the patients were followed up regularly every 3 months after SR or RFA until their last visit to our hospital or death.
This study was conducted in accordance with the Declaration of Helsinki and current ethical guidelines. Approval was obtained from the Institutional Review Board (IRB) of Taipei Veterans General Hospital (VGHIRB No. 2018-07-029BC). Informed consent was obtained before the patients underwent SR or RFA.

Statistical analyses.
The primary endpoint of the study was OS, which was calculated from the date of HCC diagnosis until death, the last visit, or loss to follow-up. Pearson's chi-squared analysis or Fisher's exact test was used to compare categorical variables. The medians with IQRs were used to express continuous variables and compared using the Mann-Whitney U test.
Cumulative rates of OS and RFS were estimated by the Kaplan-Meier method, and the results were compared using a Cox proportional hazards model. Variables that had statistical significance (p < 0.05) or were proximate to it (p < 0.1) in the univariate analysis were included in a multivariate analysis, which was conducted using a forward stepwise Cox regression model. The ALBI scores were derived from serum albumin and bilirubin levels, so we used the ALBI grade but not the serum albumin and bilirubin levels in the multivariate analysis.
A two-tailed p < 0.05 was considered statistically significant. All statistical analyses were performed using IBM SPSS Statistics for Windows, version 21.0 (IBM Corp., Armonk, NY, USA).

Data availability
The datasets generated during and analyzed during the current study are available from the corresponding author on reasonable request.