Management of hepatocellular carcinoma: an overview of major findings from meta-analyses

This paper aims to systematically review the major findings from meta-analyses comparing different treatment options for hepatocellular carcinoma (HCC). A total of 153 relevant papers were searched via the PubMed, EMBASE, and Cochrane library databases. They were classified according to the mainstay treatment modalities (i.e., liver transplantation, surgical resection, radiofrequency ablation, transarterial embolization or chemoembolization, sorafenib, and others). The primary outcome data, such as overall survival, diseases-free survival or recurrence-free survival, progression-free survival, and safety, were summarized. The recommendations and uncertainties regarding the treatment of HCC were also proposed.


INTRODUCTION
Hepatocellular carcinoma (HCC) is one of the most common causes of cancer-related death [1][2][3]. Currently, the most widely accepted therapeutic algorithm is derived from BCLC staging system [4][5], in which the mainstay treatment options for HCC include liver transplantation (LT), surgical resection, radiofrequency ablation (RFA), percutaneous ethanol injection (PEI), transarterial embolization (TAE) or chemoembolization (TACE), and sorafenib. Several novel therapeutic modalities have been also explored, such as percutaneous acetic acid injection (PAI), three-dimensional conformal radiation therapy (3D-CRT), argon-helium cryotherapy system (AHCS), traditional Chinese medicine (TCMs), cytokine-induced killer (CIK) cell therapy, and portal vein embolization (PVE), etc. It remains unclear about whether or such novel therapeutic modalities could be applied to the clinical practice. Meta-analysis can provide the highest level of evidence for our clinical decisions by combining all scattered data [6][7]. Herein, we systematically reviewed the major findings from all meta-analyses regarding the treatment of HCC and attempted to propose the evidence-based recommendations and uncertainties.

RESULTS
Overall, 2039 papers were identified. Among them, 153 meta-analysis papers were finally included  ( Figure 1). The number of relevant papers was gradually increased over years (Supplementary Figure S1). The characteristics of these included papers were shown in Table 1. Their major findings were summarized according to the treatment modalities (Tables 2-5

LT Living donor LT (LDLT) versus deceased donor LT (DDLT)
Three meta-analyses compared the outcomes of LDLT versus DDLT [8,41,67]. All of them demonstrated that the OS was statistically similar between the two groups [8,41,67]. Two of them showed that the 1-, 3-, and 5-year DFS were statistically similar between the two groups [8,67], but another one favored DDLT in term of DFS [41]. One of them found

Primary versus salvage LT
Two meta-analyses compared the outcomes of primary versus salvage LT [59,157]. Both of them demonstrated that the OS and 1-and 3-year DFS were statistically similar between the two groups [59,157]. One of them favored primary LT in term of 5-year DFS [157]; by comparison, another one showed that the 5-year DFS was statistically similar between the two groups [59]. In addition, salvage LT had significantly longer operative time, increased intra-operative blood loss, and larger number of transfused units of packed red blood cells than primary LT [157]. But the length of hospital and ICU stay was statistically similar between the two groups [157].
Only non-RCT studies, rather than RCTs, were included in the two meta-analyses.
The meta-analysis by Zhu had a larger number of included studies than that by Li (14 versus 11) (Supplementary Table S10). Notably, there was an overlap of included studies between them. All studies which were included in the meta-analysis by Li were also included in the meta-analysis by Zhu. Given its superiority in the quantity of non-RCT studies, the results of the meta-analysis by Zhu might be more reliable. In details, salvage LT achieves the same short-and long-term survival as primary LT. However, primary LT was significantly superior to salvage LT in terms of operative time, blood loss, and blood transfusion.

Sirolimus-based immunosuppression after LT
Two meta-analyses compared the outcomes of sirolimus-based immunosuppression versus no sirolimus after LT [66,83]. Both of them favored the use of sirolimus after LT in terms of OS, DFS/RFS, and recurrence [66,83].
Only non-RCT studies, rather than RCTs, were included in the two meta-analyses.
Both of them had a similar number of included studies (5 versus 5) (Supplementary Table S11). But not all included studies were the same between them.
The results were completely consistent between the two meta-analyses. In details, the use of sirolimus after LT should be favored.

LT versus surgical resection
Seven meta-analyses compared the outcomes of LT versus surgical resection [25,46,96,98,129,131,146]. There were 4, 4, 6, and 1 meta-analyses to compare the 1-, 3-, 5-, and 10-year survival, respectively. As for the 1-year survival, three of them demonstrated that the survival was statistically similar between the two groups [129,131,146], but another one favored surgical resection [98]. As for the 3-year survival, two of them found that the survival was statistically similar between the two groups [46,131], but another two favored LT [129,146]. As for the 5-year survival, two of them showed that the survival was statistically similar between the two groups [96,98], but another four favored LT [25,129,131,146]. As for the 10-year survival, the only one meta-analysis favored LT [98]. There were 3, 3, 4, and 1 meta-analyses to compare the 1-, 3-, 5-, and 10-year DFS, respectively. As for the 1-year DFS, two of them favored LT [131,146], but another one found that the 1-year DFS was statistically similar between the two groups [98]. As for the 3-year DFS, all of them favored LT [46,131,146]. As for the 5-year DFS, all of them favored LT [98,129,131,146]. As for the 10-year DFS, the only one meta-analysis favored LT [98]. Two meta-analyses compared the recurrence. Both of them favored LT in term of recurrence [129,146].
Only non-RCT studies, rather than RCTs, were included in these meta-analyses.
The meta-analysis by Zhang had the largest number of included studies (n = 62) (Supplementary Table S12). By comparison, the number of included studies was less than 20 in 6 other meta-analyses.
Given its superiority in the quantity of non-RCT studies, the results of the meta-analysis by Zhang might be more reliable. In details, LT provides a significantly better survival and a lower recurrence.

Surgical resection Surgical resection margin 1 cm versus 2 cm
Only one meta-analysis compared the outcomes of hepatectomy with a margin aiming at 2 cm versus those with a margin aiming at 1 cm [109]. Regardless of study design, the 1-year survival was statistically similar between the two groups [109]. In the subgroup analysis of randomized studies, the 3-and 5-year survival and DFS were better in patients undergoing hepatectomy with a margin aiming at 2 cm than in those with a margin aiming at 1 cm [109]. Contrarily, in the subgroup analysis of non-randomized studies, the 3-and 5-year survival and DFS were statistically similar between the two groups [109].
One RCT and 4 non-RCT studies were included in this meta-analysis.

Laparoscopic versus open resection
Nine meta-analyses compared the outcomes of laparoscopic versus open resection [30, 62, 94-95, 111, 126, 135, 137, 156]. All of them demonstrated that the OS and DFS/RFS at any time points were statistically similar between the two groups [30, 62, 94-95, 111, 126, 135, 137, 156]. Two of them also found that the recurrence was statistically similar between the two groups [62,126]. Eight of them demonstrated statistically similar operative time between the two groups [30,62,95,111,126,135,137,156], but one demonstrated significantly longer operative time in laparoscopic resection group [94]. All of them demonstrated that blood loss or intraoperative bleeding was significantly less in laparoscopic resection group [30, 62, 94-95, 111, 126, 135, 137, 156]. Among the 7 meta-analyses evaluating the blood transfusion, 6 demonstrated significantly less blood transfusion in laparoscopic resection group [30,62,111,126,137,156], and one demonstrated statistically similar blood transfusion between the two groups [94]. Among the 6 meta-analyses evaluating the overall complications, 5 demonstrated significantly less complications in laparoscopic resection group [30,62,94,111,135], and one demonstrated statistically similar complications between the two groups [126]. Among the 8 meta-analyses evaluating the hospital length, all demonstrated significantly shorter hospital study in laparoscopic resection group [30,62,94,111,126,135,137,156].
Only non-RCT studies, rather than RCTs, were included in these meta-analyses.
The meta-analyses by Park, Xiong, and Yin had the largest number of included studies (n = 15) followed by the meta-analyses by Yao (n = 13), Zhou (n = 10), Li (n = 10), Fancellu (n = 9), Pang (n = 7), and Twaij (n = 4) (Supplementary Table S13). The included studies were completely same between the two meta-analyses by Xiong and Yin. However, the studies included in the metaanalysis by Park were different from those included in the meta-analyses by Xiong and Yin. Given its superiority in the quantity of non-RCT studies, the results of the meta-analyses by Park, Xiong, and Yin might be more reliable. In details, they suggested that the operative time was statistically similar between the two groups and that laparoscopic resection was superior to open resection in terms of blood loss, blood transfusion, complications, and hospital stay.

Anatomic resection versus non-anatomic resection
Six meta-analyses compared the outcomes of anatomic versus non-anatomic resection [15,23,60,110,136,152]. Four of them demonstrated that the OS was statistically similar between the two groups [15,60,110,136], but another two favored anatomic resection in term of 5-year survival [23,152]. One of them found that the DFS was statistically similar between the two groups [60], but another four favored anatomic resection in term of DFS [15,23,136,152]. Two of them showed that the recurrence was statistically similar between the two groups [15,110], but another two favored anatomic resection in term of local intrahepatic recurrence [136,152]. Post-operative complications were statistically similar between the two groups [23,110,136,152].
Only non-RCT studies, rather than RCTs, were included in these meta-analyses.
Given its superiority in the quantity of non-RCT studies, the results of the meta-analysis by Zhang might be more reliable. In details, anatomic resection was superior to non-anatomic resection in terms of OS and DFS.  Salvage LT achieves the same short-and long-term outcomes as primary LT.

Surgical resection + I 131 lipiodol versus surgical resection alone
Two meta-analyses compared the outcomes of surgical resection in combination with I 131 lipiodol versus surgical resection alone [36,40]. Both of them favored the combination therapy in terms of OS, DFS, and recurrence [36,40].
Only non-RCT studies, rather than RCTs, were included in these meta-analyses.
The meta-analysis by Gong had a larger number of included studies than that by Furtado (10 versus 5) (Supplementary Table S15). Notably, there was an overlap of included studies between them. All studies which were included in the meta-analysis by Furtado were also included in the meta-analysis by Gong. The results were completely consistent between the two meta-analyses. In details, surgical resection in combination with I 131 lipiodol should be favored.

Pre-operative TACE
Four meta-analyses compared the outcomes of surgical resection in combination with pre-operative TACE versus surgical resection alone [17,116,138,153]. All of them found that the OS, DFS, and recurrence were statistically similar between the two groups [17,116,138,153].
RCT studies were included in the meta-analyses by Cheng (n = 4), Wang (n = 3), and Zhou (n = 4), but not in the meta-analysis by Yu (n = 0). The meta-analysis by Zhou had the largest number of included studies (n = 21) followed by the metaanalyses by Yu (n = 7), Cheng (n = 4), and Wang (n = 3) (Supplementary Table S16). All studies which were included in the two meta-analyses by Cheng and Wang were also included in the meta-analysis by Zhou.
The results were completely consistent between the two meta-analyses. In details, pre-operative TACE did not improve the OS or DFS.

Post-operative TACE
Two meta-analyses compared the outcomes of surgical resection in combination with post-operative TACE versus surgical resection alone [17,149]. Both of them favored post-operative TACE in terms of OS, DFS, and recurrence [17,149].
Only RCT studies were included in the two metaanalyses.
Although the number of included studies was the same between the two meta-analysis by Cheng and Zhong (n = 6) (Supplementary Table S17), not all included studies were the same between them.
The results were completely consistent between the two meta-analyses. In details, post-operative TACE should be favored.

Surgical resection + adjuvant chemotherapy versus surgical resection alone
Five meta-analyses compared the outcomes of surgical resection in combination with adjuvant chemotherapy versus surgical resection alone [78,92,112,[147][148]. LT provides increased survival and lower recurrence rates than resection. Survival was similar between resection aiming at 2 cm and 1 cm.

Oral systemic chemotherapy
Oral systemic chemotherapy was evaluated in two meta-analyses [78,148]. The OS, RFS, and recurrence were statistically similar between patients with and without chemotherapy [78,148].
RCT studies were included in the meta-analyses by Zhong (n = 3) and Mathurin (n = 1).
The meta-analysis by Zhong had a larger number of included studies than that by Mathurin (3 versus 2) (Supplementary Table S18). Not all included studies were the same between them.
The results were completely consistent between the two meta-analyses. In details, the adjunctive use of oral systemic chemotherapy should not be favored in patients undergoing surgical resection.

Transarterial chemotherapy
Transarterial chemotherapy was evaluated in one meta-analysis [78]. As for the pre-operative transarterial chemotherapy, the overall analysis of both RCTs and non-RCTs demonstrated that chemotherapy improved the 2-year survival, but not the 1-or 3-year survival. The subgroup analysis of RCTs showed that the 1-, 2-, and 3-year recurrence were statistically similar between the two groups. As for the post-operative transarterial chemotherapy, the overall analysis of both RCTs and non-RCTs demonstrated that chemotherapy improved the 1-, 2-, and 3-year survival. The subgroup analysis of RCTs showed that chemotherapy improved the 2-and 3-year survival, but not the 1-year survival.
Approaches of chemotherapy were mixed in three meta-analyses [92,112,147]. The statistical results were largely inconsistent among them. One of them favored the chemotherapy in term of OS [147]; one showed that the OS was statistically similar between the two groups [112]; one demonstrated that the OS was decreased by chemotherapy [92].
The meta-analysis by Zheng had a larger number of included studies than those by Wang and Ono (48 versus 8 and 3) (Supplementary Table S19).
Given its superiority in the quantity of RCT studies, the results of the meta-analysis by Zheng might be more reliable.

Surgical resection + immunotherapy versus surgical resection alone
Four meta-analyses compared the outcomes of surgical resection in combination with immunotherapy versus surgical resection alone [32,75,112,124]. All of them demonstrated that the OS was statistically similar between the two groups [32,75,112,124]. One of them favored the combination therapy in term of RFS. One of them favored the combination therapy in terms of 1-and 3-year recurrence [124]; one favored the combination therapy in term of 1-year recurrence, but not 3-year recurrence [75]; one showed that the recurrence was statistically similar between the two groups [32].
The meta-analysis by Xie had the largest number of included studies followed by those by Ma, Wang, and Flores (6 versus 4, 3, and 2) (Supplementary Table S20). Notably, there was an overlap of included studies among them.
Given its superiority in the quantity of RCT studies, the results of the meta-analysis by Xie might be more reliable. In details, the adjunctive use of immunotherapy might not be favored in patients undergoing surgical resection.

Surgical resection + PVE versus surgical resection alone
One meta-analysis compared the outcomes of surgical resection in combination with PVE versus surgical resection alone [145]. The 1-, 3-, and 5-year survival and intrahepatic and distant recurrence were statistically similar between the two groups [145].
Only non-RCT studies were included in the metaanalysis by Zhao.

RFA versus surgical resection
Eighteen meta-analyses compared the outcomes of RFA versus surgical resection [11,16,24,27,31,33,47,61,69,72,89,97,106,117,119,127,151,153]. As for the OS, seven of them favored surgical resection [27,31,47,61,97,106,127]; four demonstrated that the OS was statistically similar between the two groups [11,16,119,151]; four showed that the 1-year survival was statistically similar between the two groups, but the 5-year survival was better in surgical resection group [33,72,89,117]; one found that the 1-and 5-year survival were statistically similar between the two groups, but the 3-year survival was better in surgical resection group [155]; one reported that surgical resection had better OS than RFA in the subgroup analyses of a single nodule 3-5 cm and ≤ 3 cm, but the OS was statistically similar between the two groups in the subgroup analyses of a single nodule < 2 cm and 2-3 nodules < 3 cm [24].
As for the DFS, nine of them favored surgical resection in terms of DFS/RFS at any time points [11,27,31,61,72,89,97,106,155]; three showed that the 1-year DFS was statistically similar between the two groups, but the 3-and/or 5-year DFS were better in surgical resection group than in RFA group [16,33,151]; one reported that surgical resection had better DFS than RFA in the subgroup analyses of a single nodule 3-5 cm and ≤ 3 cm, but the DFS was statistically similar between the two groups in the subgroup analyses of a single nodule < 2 cm and 2-3 nodules < 3 cm [24].
As for the recurrence, three of them favored surgical resection [11,47,61]; two favored RFA [127,151]; one found that the recurrence was statistically similar between the two groups [24]; three showed that the 1-year recurrence was statistically similar between the two groups, but the 3-year recurrence was less in surgical resection group than in RFA group [31,33,117]; one reported that the 1-and 3-year recurrence were statistically similar between the two groups, but the recurrence at the end of follow-up was less in RFA group than in surgical resection group [69]; one demonstrated that the distant intrahepatic recurrence was statistically similar between the two groups, but the local intrahepatic recurrence was less in surgical resection group than in RFA group [153]; one favored surgical resection in term of recurrence at previous sites, but favored RFA in term of recurrence at new sites [72].
According to the description of each meta-analysis, RCT studies were included the meta-analyses by Liu  Table S21).
Given the superiority in the quantity of RCT studies, the meta-analyses by Feng, Qi, Wang, and Weis might be more reliable. In details, surgical resection should be superior to RFA for the improvement of OS.

PEI versus surgical resection
Two meta-analyses compared the outcomes of PEI versus surgical resection [46,100]. Both of them demonstrated that OS and RFS were statistically similar between the two groups [46,100].
Only one RCT study was included in the metaanalysis by Schoppmeyer.
The meta-analysis by Hoshida had a larger number of included studies than that by Schoppmeyer (5 versus 1) (Supplementary Table S22). However, no included studies were overlapped between them.
The results were completely consistent between the two meta-analyses. In details, PEI was similar to surgical resection in terms of OS and RFS.

Non-surgical-resection ablation versus surgical resection
One meta-analysis compared the outcomes of nonsurgical-resection ablation versus surgical resection [26]. The 1-and 3-year survival and DFS were statistically similar between the two groups [26].

RFA versus PEI or PAI
Eight meta-analyses compared the outcomes of RFA versus PEI or PAI [9,20,26,38,93,102,119,130]. All of them favored RFA over PEI in terms of OS, DFS, and/or recurrence [9,20,26,38,93,102,119,130]. Additionally, one of them found that the OS, local recurrence, de novo tumor, and adverse event were statistically similar between RFA and PAI groups [38].
The results regarding the comparison between RFA v.s. PEI were completely consistent among metaanalyses. In details, RFA should be superior to PEI for the improvement of OS and DFS.

RFA versus cryosurgery ablation
One meta-analysis compared the outcomes of RFA versus cryosurgery ablation [51]. Although the OS was statistically similar between the two groups, RFA had less recurrence and complications than cryosurgery ablation [51].

RFA versus other therapeutic methods
One meta-analysis compared the outcomes of RFA versus any other therapeutic methods [50]. RFA was superior to other treatment methods for early HCC in terms of local recurrence and 3-year survival [50]. However, no subgroup analysis was performed according to the different treatment modalities.
Additionally, one meta-analysis compared the outcomes of RFA versus microwave or laser ablation [119]. However, only one trial was identified for each comparison.

PEI versus PAI
Two meta-analyses compared the outcomes of PEI versus PAI [38,100]. Both of them showed that the OS, RFS, and recurrence were statistically similar between the two groups [38,100].
Both of them had a similar number of included studies (n = 2) (Supplementary Table S24). However, not all of the included studies were identical.
The results were completely consistent between the two meta-analyses. In details, PEI was similar to PAI in terms of OS and RFS. www.impactjournals.com/oncotarget
Three meta-analyses compared the RFS of RFA in combination with TACE versus RFA or TACE alone [71,[87][88]. As for the 1-year RFS, one meta-analysis favored the combination therapy [71], but another two showed that the 1-year RFS was statistically similar between the two groups [87][88]. By comparison, all of them favored the combination therapy in term of 3-year RFS [71,[87][88].
RCT studies were included in the meta-analyses by Zhao  Table S25). Notably, all of the 19 included studies were completely identical among the three meta-analyses by Jiang, Kong, and Yan. Given the superiority in the number of RCTs, the meta-analyses by Zhao, Jiang, Kong, Ni, and Yan should be more reliable. In details, RFA in combination with TACE should be favored in term of OS.

PEI + TACE versus mono-therapy
Three meta-analyses compared the outcomes of PEI in combination with TACE versus PEI or TACE alone [68,[114][115]. Two of them favored the combination therapy in term of OS [114][115]. Another one metaanalysis was performed according to the study design. In the subgroup analysis of RCTs, the combination therapy significantly improved the 3-year survival, rather than 1-year survival. By contrast, in the subgroup analysis of observational studies, the combination therapy significantly improved the 1-year survival, rather than 3-year survival [68].
RCT studies were included in all of the 3 metaanalyses by Wang Table S26). However, not all studies included by Wang W and Liao were included by Wang N.
Given the superiority in the number of RCTs, the meta-analysis by Wang N (Med Oncol, 2011) might be more reliable. In details, PEI in combination with TACE should be favored in term of OS.

Any ablation therapy + TACE versus mono-therapy
Two meta-analyses compared the outcomes of unclassified ablation therapies in combination with TACE versus mono-therapy [42,115]. Both of them favored the combination therapy in terms of OS, recurrence, and tumor response [42,115].
RCT studies were included in the meta-analyses by Wang (n = 10) and Gu (n = 7).
The meta-analysis by Gu had a larger number of included studies than that by Wang (18 versus 10) (Supplementary Table S27). However, not all studies included by Wang were included by Gu. The results were completely consistent between the two meta-analyses. In details, TACE in combination with ablation therapy was favored.
The meta-analysis by Marelli had the largest number of included studies (n = 9) followed by those by Oliveri  Table S28). However, not all included studies were completely overlapped among them.
Given the superiority in the number of RCTs, the meta-analysis by Marelli might be more reliable. In details, TACE/TAE should be favored.

TACE versus TAE
Three meta-analyses compared the outcomes of TACE versus TAE [12,77,125]. All of them showed that the OS was statistically similar between the two groups [12,77,125].
The meta-analysis by Xie had a larger number of included studies than those by Marelli and Camma (5 versus 3 and 2) (Supplementary Table S29). However, not all included studies were completely overlapped among them.
The results were completely consistent among them. In details, TACE was similar to TAE in term of OS.

Drug-eluting bead (DEB)-TACE versus conventional TACE (cTACE)
Three meta-analyses compared the outcomes of DEB-TACE versus cTACE [37,44,48]. One of them evaluated the OS [48]. DEB-TACE was significantly better than cTACE in terms of 1-and 2-year survival. But the 6-month and 3-year survival were statistically similar between the two groups.
Two of them demonstrated that tumor response or disease control rate was statistically similar between them. Another one meta-analysis demonstrated that tumor response rate was significantly higher in DEB-TACE group than in cTACE group.
Two of them evaluated the complications [37,44]. The incidence of complications was statistically similar between the two groups.
RCT studies were included in the meta-analyses by Han (n = 3) and Huang (n = 2), but not in the metaanalysis by Gao. The meta-analysis by Huang had a larger number of included studies than those by Han and Gao (7 versus 5 and 2) (Supplementary Table S30). However, not all included studies were completely overlapped among them.
Given the superiority in the number of RCTs, the meta-analysis by Han might be more reliable. In details, DEB-TACE was similar to cTACE in the term of tumor response.

TACE versus microsphere embolization
One meta-analysis compared the outcomes of TACE versus microsphere embolization [123]. Microsphere embolization was superior to TACE in terms of OS, TTP, and tumor response [123]. In the subgroup analyses, the benefit was statistically significant in patients undergoing 32 P glass microspheres, but not in those undergoing 90 Y microspheres.
RCT studies were included in the meta-analysis by Xie (n = 7).

TACE + sorafenib versus TACE
Four meta-analyses compared the outcomes of TACE in combination with sorafenib versus TACE alone [35,70,134,140]. Three of them favored the combination therapy in term of OS [35,134,140], but another one found that the OS was statistically similar between the two groups [70]. The survival benefit of the combination therapy was statistically significant in the subgroup analysis of retrospective studies, but not in that of RCTs [134].
RCT studies were included in the meta-analyses by Liu (n = 3), Yang (n = 3), and Zhang (n = 2), but not in the meta-analysis by Fu. The meta-analysis by Fu had the largest number of included studies (n = 9), followed by those by Liu (n = 7), Yang (n = 6), and Zhang (n = 6) (Supplementary Table  S31). However, not all included studies were completely overlapped among the 4 meta-analyses.
Given the superiority in the number of RCTs, the meta-analysis by Liu and Yang might be more reliable. In details, TACE plus sorafenib was not favored in term of OS.

TACE + high-intensity focused ultrasound (HIFU) versus TACE
Two meta-analyses compared the outcomes of TACE in combination with HIFU versus TACE alone [13,68]. One of them demonstrated that both OS and tumor response were improved by the combination therapy [13]. Another one meta-analysis was performed according to the study design [68]. In the subgroup analysis of observational studies, both 1-and 3-year survival were significantly improved by the combination therapy [68]. By comparison, in the subgroup analysis of RCTs, only 1-year survival, rather than 3-year survival, was significantly improved by the combination therapy [68].
RCT study was included in the meta-analysis by Liao (n = 1), but not in the meta-analysis by Cao. The meta-analysis by Cao had a larger number of included studies than that by Liao (9 versus 5) (Supplementary Table S32). All studies which were included in the meta-analysis by Liao were also included in the meta-analysis by Cao. Given the superiority in the number of RCTs, the meta-analysis by Liao might be more reliable. In details, TACE plus HIFU should be favored in term of 1-year survival, but not 3-year survival.

TACE + thermotherapy versus TACE
Only one meta-analysis compared the outcomes of TACE in combination with thermotherapy versus TACE alone [64]. Both 1-and 2-year survival were significantly improved by the combination therapy, but the 0.5-, 1.5-, and 3-year survival were statistically similar between the two groups [64]. Additionally, the overall effective rate and quality of life were improved by the combination therapy [64].

TACE + AHCS versus TACE or AHCS
Only one meta-analysis compared the outcomes of TACE in combination with AHCS versus TACE or AHCS alone [65]. Compared with TACE alone, the combination therapy had significantly better 0.5-, 1-, 1.5-, 2-, and 2.5year survival, but the 3-year survival was statistically similar between the two groups [65]. Compared with AHCS alone, the combination therapy had significantly better 0.5-, 1.5-, 2-, and 2.5-year survival, but similar 1and 3-year survival [65]. Additionally, the combination therapy was superior to the mono-therapy in terms of total effective rate, complete necrosis rate, recurrence, AFP reduction, and CD4 improvement.

TACE + radiotherapy versus TACE alone
Two meta-analyses compared the outcomes of TACE in combination with radiotherapy versus TACE alone [68,80]. Both of them demonstrated that the combination therapy had significantly better 1-, 2-, 3-, and 5-year survival than TACE alone [68,80]. Additionally, one of them showed that the combination therapy significantly increased the tumor response, but did not influence the development of adverse events, such as nausea/vomit, leukocyte count declined, alanine aminotransferase level increased, and total bilirubin level increased [80].
RCT studies were included in the meta-analyses by Meng (n = 5) and Liao (n = 3).
The meta-analysis by Meng had a larger number of included studies than that by Liao (17 versus 7) (Supplementary Table S33). All studies which were included in the meta-analysis by Liao were also included in the meta-analysis by Meng. The results regarding the OS were completely consistent among them. In details, TACE plus radiotherapy should be favored in term of OS.

TACE + 3D-CRT versus TACE alone
Only one meta-analysis compared the outcomes of TACE in combination with 3D-CRT versus TACE alone [68]. Regardless of study design, the combination therapy was superior to TACE alone in terms of 1-and 3-year survival [68].
Only one RCT study was included in the metaanalysis by Liao.

TACE + TCMs versus TACE alone
Six meta-analyses compared the outcomes of TACE in combination with TCMs versus TACE alone [18-19, 79, 81, 108, 122]. Three of them favored the combination therapy in term of OS [18,79,108]; one favored the combination therapy in terms of 1-, 2-, and 3-year survival, but not 6-month survival [19]; one favored the combination therapy in term of 2-year survival, but not 1-year survival [122]; one did not report the survival data [81].
Five of them favored the combination therapy in term of tumor response [18-19, 79, 81, 122]. Another one did not report the relevant data [108].
Four of them favored the combination therapy in term of quality of life [18-19, 79, 108]. Another two did not report the relevant data [81,122].
RCT studies were included in the meta-analyses by Cheung (n = 67), Cho (n = 30), and Meng (Explore (NY), 2011) (n = 11), but not in the meta-analyses by Sun and Wu. The information regarding the inclusion of RCTs was not reported in the meta-analysis by Meng (J Altern Complement Med, 2008).
Given the superiority in the number of RCTs, the meta-analysis by Cheung might be more reliable. In details, TACE plus TCMs should be favored in terms of OS, tumor response, and quality of life.

TACE + CIK cell therapy versus TACE alone
Two meta-analyses compared the outcomes of TACE in combination with CIK cell therapy versus TACE alone [14,63]. The combination therapy was beneficial in terms of OS, RFS, TTP, quality of life, and liver and immune function [14,63]. Additionally, one of them evaluated whether or not adjunctive CIK cell therapy could improve the outcomes of TACE in combination with RFA [63]. Adjunctive CIK cell therapy was beneficial in terms of OS and RFS [63].
RCT studies were included in the meta-analyses by Chen (n = 9) and Li (n = 6).
The meta-analysis by Li had a larger number of included studies than that by Chen (11 versus 9) (Supplementary Table S35). However, not all included studies were completely overlapped between them.
The results regarding the OS were completely consistent among them. In details, TACE in combination with CIK cell therapy should be favored.

Sorafenib
Seven meta-analyses compared the outcomes of sorafenib versus placebo (Supplementary Table S1) [22,28,101,118,141,143,160]. The use of sorafenib was beneficial in terms of OS, TTP, and disease control rate [22,101,118,141,143,160]. However, the time to symptomatic progression was statistically similar between the two groups [22,118,160]. The incidence of adverse events was significantly increased by the use of sorafenib [28,101,118,141,143,160].
RCT studies were included in the meta-analyses by Shen Table S36). All studies which were included in the meta-analysis by Wang, Duffy, Zhang T (Anticancer Drugs, 2010), Zhang X (Hepatobiliary Pancreat Dis Int, 2012), and Zou were also included by Shen. In the meta-analysis by Cinco, the included studies were not reported.
The results were completely consistent among them. In details, sorafenib should be favored.
Five meta-analyses evaluated the DFS/RFS [49,56,103,112,154]. Four of them favored the use of antiviral therapy in term of DFS/RFS [56,103,112,154]. Another one meta-analysis was performed according to the study design and type of viral hepatitis. In the subgroup analysis of RCTs, the DFS/RFS was statistically similar between the two groups regardless of HCV or HBV [49]. In the subgroup analysis of non-RCTs, antiviral therapy improved the DFS/RFS by in HCV patients, but not HBV patients [49].
Given the superiority in the number of RCTs, the meta-analysis by Zhuang (PLoS One, 2013) and Huang might be more reliable. In details, interferon therapy after curative treatment should be favored.

Vitamin
Five meta-analyses compared the outcomes of vitamin versus placebo (Supplementary Table S3) [21,82,99,112,150]. Two of them favored the use of vitamin in term of OS [112,150]; two favored the use of vitamin in term of 2-year survival, but not 3-year survival [21,82]; one showed that the 1-and 2-year survival were statistically similar between the two groups [99].
One meta-analysis favored the use of vitamin in term of RFS [112].
Two meta-analyses favored the use of vitamin in term of 1-year recurrence, but another two did not [21,82]. Four meta-analyses favored the use of vitamin in terms of 2-and 3-year recurrence [21,82].
The meta-analysis by Zhong had a larger number of included studies than those by Chu, Wang, Riaz, and Meng (7 versus 6, 6, 5, and 4) (Supplementary Table  S38). In the meta-analysis by Meng, the included studies were not reported. However, not all included studies were completely overlapped between them.
Given the superiority in the number of RCTs, the meta-analyses by Wang and Zhong should be more reliable. In details, the use of vitamin should be favored in term of OS. However, its benefit was weak.

Octreotide
Three meta-analyses compared the outcomes of octreotide versus placebo (Supplementary Table S4) [29,43,52]. As for the 6-and 12-month survival, one of them favored the use of octreotide [52], but another two did not show any significant difference between the two groups [29,43]. As for the 24-month survival, two of them showed that the survival was statistically similar between the two groups [29,43], but another one did not report the relevant data [52].
RCT studies were included in the meta-analyses by Ji (n = 9) and Guo (n = 6).
The meta-analysis by Ji had a larger number of included studies than those by Guo and Estanislao (11 versus 6 and 3) (Supplementary Table S39). In the meta-analysis by Estanislao, the included studies were not reported. All studies which were included in the metaanalysis by Guo were also included by Ji. Given the superiority in the number of RCTs, the meta-analyses by Ji and Guo might be more reliable. In details, the benefit of octreotide remains uncertain.

TCM
One meta-analysis compared the outcomes of kanglaite injection plus hepatic arterial intervention versus hepatic arterial intervention alone (Supplementary Table S5) [34]. The combination therapy was beneficial in terms of tumor response, Karnofsky score improvement, and pain relief [34]. But neither OS nor DFS/RFS was evaluated [34].
One meta-analysis compared the outcomes of Chinese herbal medicine plus chemotherapy versus chemotherapy alone [104]. The combination therapy was beneficial in terms of OS and tumor response [104].
One meta-analysis compared the outcomes of TCM versus other treatment [121]. TCM was superior to other treatments in terms of OS and tumor response [121].

CIK cell therapy
One meta-analysis compared the outcomes of CIK cell therapy versus other treatment (Supplementary Table  S6) [76]. CIK cell therapy was superior to other treatments in terms of OS, PFS, disease control rate, tumor response, and quality of life [76].

Tamoxifen
Two meta-analyses compared the outcomes of tamoxifen versus placebo or no treatment (Supplementary  Table S7) [73,90]. Both of them demonstrated that the OS was statistically similar between the two groups [73,90].
Although the meta-analysis by Nowak had a larger number of included studies than that by Llovet (10 versus 7) (Supplementary Table S40), the included studies were not similar between them.
The results were completely consistent among them. In details, tamoxifen should not be favored.

Antibiotics
One meta-analysis compared the outcomes of antibiotics versus no antibiotics after hepatic transarterial therapy [113]. The incidence of fever, bacteremia, septicema, and sepsis were not significantly improved by antibiotics [113].

DISCUSSION
AASLD and EASL guidelines recommend BCLC staging algorithm for the management of HCC. Only 5 treatment modalities have been considered in the current guidelines. In details, the therapeutic modalities of HCC include the LT, surgical resection, and RFA for HCC in the stage 0 and A, TACE for HCC in the stage B, sorafenib for HCC in the stage C, and supportive treatment for HCC in the stage D. However, the BCLC staging algorithm is not flawless and needs to be persistently updated. Nowadays, more and more novel treatment modalities have been widely produced and adopted. Their efficacy and safety have been gradually established. In this circumstance, our study was worthwhile, because it attempted to collect the relevant evidence as many as possible and to provide an overview of outcomes of novel and well-established treatment modalities for HCC based on the results of metaanalyses. More notably, we found that lots of combination therapy might be more effective and safe. For example, the meta-analyses of RCTs demonstrated that RFA plus TACE was superior to mono-therapy, and that surgical resection plus post-operative TACE was superior to surgical resection alone. Given the quality of such metaanalyses, the guidelines should be updated regarding the use of combination therapy.

Limitations
This was a time-consuming work, because a large number of relevant meta-analyses were included. Several limitations should be acknowledged. First, we must clarify that only the results of meta-analyses, but not the accuracy of meta-analyses, were systematically reviewed. Because we cannot repeat every meta-analysis, we cannot guarantee that their findings were accurate. Second, we did not consider the heterogeneity among included studies in every meta-analysis. A significant heterogeneity could affect the stability of a meta-analysis. Third, we arbitrarily evaluated the reliability of meta-analyses according to the number of RCTs and non-RCTs.

Recommendations LT
1. LDLT has lower DFS than DDLT (grade of recommendation: low). 2. Short-and long-term outcomes may be comparable between primary and salvage LT (grade of recommendation: low). 3. Sirolimus-based immunosuppression should be recommended after LT (grade of recommendation: low).

Surgical resection
1. Surgical resection margin aiming at 2 cm may be superior to 1 cm for the improvement of long-term outcomes (grade of recommendation: moderate).

Survival benefit may be comparable between
laparoscopic and open resection. Additionally, laparoscopic resection had less blood loss, blood