Short-term and long-term efficacy of 7 targeted therapies for the treatment of advanced hepatocellular carcinoma: a network meta-analysis

Supplemental Digital Content is available in the text


Introduction
Hepatocellular carcinoma (HCC) is one of the most common clinical digestive malignant tumors, [1] whose etiology has not been fully elucidated, yet the hepatitis, cirrhosis, and hepatic carcinoma are considered to be the 3 main reasons for its continued evolution of migration through long-term clinical observation. [2][3][4] Other factors like alcohol and unhealthy living habits may also function as HCC inducement. [5] Currently, multidisciplinary treatments of surgery, molecular targeted therapy, and traditional Chinese medicine are advocated for HCC treatment. [6] Due to the insidious onset, high malignant degree, dissemination, and metastasis of HCC, the diagnosis of pathologically early HCC remains difficult. [7] Therefore, it is common for majority HCC patients to develop advanced hepatocellular carcinoma (AHCC) at initial diagnosis and lose the opportunity of radical surgery and other local treatments. [8,9] Therefore, it is of great importance to explore strategies for AHCC patients in order to further improve the overall efficacy of AHCC treatment.
In recent years, targeting drugs has gradually become a focus of HCC treatment, and there are a variety of targeted drug therapies in clinical trials that have been proven to be effective. [10,11] Wherein, sorafenib, which is based on a category of multitargeted tyrosine kinase inhibitor, has been used in clinical trials for its wide range of antitumor effect. [12] Sorafenib can effectively extend the overall survival time of HCC patients, but its severe side effects may affect the life quality of those patients. [13] In addition to sorafenib, sunitinib, an oral multitargeted tyrosine kinase inhibitor, and brivanib, a selective dual inhibitor of vascular endothelial growth factor (VEGF) and fibroblast growth factor signaling, are the most concerned agents targeted to AHCC management with effective outcomes; [1] whereas, tivantinib, a receptor tyrosine kinase encoded from a proto-oncogene c-Met gene, can cause cell death by acting on the caspase-dependent apoptosis pathway. [14] In addition to tyrosine kinase inhibitors, everolimus, an oral small-molecule serine-threonine kinase inhibitor, demonstrates a good drug resistance of AHCC with fewer adverse reactions though inhibition of certain signaling pathway. [15] Ramucirumab also displays satisfactory clinical results as well for its angiogenesis inhibition of tumor. [16] Currently, clinical assessment shows that sunitinib possess superior effects than sorafenib, which may represent a new generation of targeted regimen. [17] Considering the unsatisfactory results in angiogenesis of HCC, erlotinib has also been reported as a promising target for HCC owing to its ability to inhibit phosphorylation of the intracellular domain of the epidermal growth factor receptor (EGFR). [18] The comparisons on efficacy among different drug treatments cannot be achieved through traditional meta-analysis, but can be accomplished depending on network meta-analysis, which implements a quantified comparison with similar disease interventions for the selection of the optimal treatment strategy. [19] Therefore, this study enrolled 11 randomized controlled trials (RCTs) based upon a network meta-analysis to evaluate the efficacies of 7 targeted drugs, including sorafenib, ramucirumab, everolimus, brivanib, tivantinib, sunitinib, and sorafenib + erlotinib with the expectation to provide supporting evidence for a reasonable choice for AHCC treatment.

Ethics statement
Our study is a network meta-analysis and the ethics statement is not applicable.

Literature search
PubMed, Embase, Cochrane central register of controlled trials, Ovid, EBSCO, and other English databases were searched from the inception of each database to September 2016. The search was conducted using MeSH terms, keywords, and combined words, which include: liver neoplasms, cancer of liver, hepatocellular cancer, hepatic cancer, sorafenib, metuximab, trastuzumab, ramucirumab, cetuximab, matuzumab, panitumumab, sunitinib, everolimus, brivanib, temsirolimus, celecoxib, lapatinib, randomized controlled trial, and so on.

Inclusion and exclusion criteria
The inclusion criteria: study design -RCT; interventionstargeted drug/placebo for AHCC patients; study subjectpatients with AHCC; and end outcomesstable disease (SD), progressive disease (PD), complete response (CR), partial response (PR), disease control rate (DCR), overall response ratio (ORR), and overall survival (OS). The exclusion criteria: studies with insufficient data; non-RCTs; non-AHCC; non-English reference; and duplicated publications by the same author using the same interventions.

Data extraction and quality assessment
Two reviewers extracted data from the enrolled studies using a specifically designed form. Additionally, a 3rd reviewer was consulted if those 2 reviewers failed to reach an agreement. Researchers of 2 or more reviewed the RCTs according to Cochrane risk of bias assessment tools, [20] which include 6 domains, namely, random allocation, allocation concealment, blinding, loss outcome data, selected outcome reports, and other bias. The assessment included a judgment assignment of "yes," "no," or "unclear" for each domain to designate a low, high, or unclear risk of bias. The study was classified as a low risk of bias if one or no domain was deemed "unclear" or "no," a high risk of bias if 4 or more domains are deemed "unclear" or "no" and a moderate risk of bias if 2 or 3 domains were deemed "unclear" or "no." [21] Quality assessment and investigation of publication bias were conducted using Review Manager 5 (RevMan 5.2.3, Cochrane Collaboration, Oxford, UK.

Statistical analysis
Traditional pairwise meta-analyses were performed to directly compare different treatment arms. Then Bayesian network metaanalyses were applied for direct comparisons of different interventions to each other. The results were reported as odds ratios (ORs) or hazard ratio (HR) with 95% CI accounting for study sample sizes. The node-splitting plot statistic was conducted to assess the extent of inconsistency, and ontology consistent model was applied if P > 0.05. Each analysis was grounded on noninformative priors to obtain effect sizes and precision. Convergence and lack of auto correlation were examined and confirmed after 4 chains and a 20,000-simulation burn-in phase. Subsequently, direct probability statements were derived from an additional 50,000-simulation phase. [22] Surface under the cumulative ranking curve (SUCRA) was applied to provide a ranking among the included treatments and indicate which treatment was the optimal one. [23] The network plot of interventions was a representation of the evidence base and conferred a concise description of its characteristics. [24] Cluster analyses were used to group the treatments regarding their similarity on outcomes. [25] All computations were performed using R (V.3.1.

Pairwise meta-analysis
Direct paired comparisons for efficacies of the 8 treatments of AHCC were conducted, suggesting better efficacies of ramucirumab and tivantinib than that of placebo in terms of SD

3.3.
Pooled results of network meta-analysis 3.3.1. Network Plot. A total of 7 kinds of targeted drugs (sorafenib, ramucirumab, everolimus, brivanib, tivantinib, sunitinib, sorafenib + erlotinib) and placebo, were included in this study. Node size in the network plot represented the number of the included subjects, and the width of nodes and lines indicated the accuracy of effect size (the inverse of variance). The network plot of SD and DCR were shown in Fig. 2A, PD in Fig. 2B, CR, PR, and ORR in Fig. 2C, and OS in Fig. 2D.

Inconsistency test.
Six studies among the total 11 constituted a closed loop to implement the inconsistency test (sorafenib vs placebo: 4 studies; brivanib vs placebo: 1 study; and brivanib vs sorafenib: 1 study), which showed no inconsistencies among all the studies in terms of SD, PD, CR, PR, DCR, ORR, and OS (all P > 0.05) ( Fig. 3A-F, Fig. 4A). Therefore, consistency model was applied.   in view of CR, whose SUCRA value was 0.64. Ramucirumab achieved the best efficacy considering PR and ORR, whose SUCRA values were 0.84 and 0.81, respectively. The ranking plot results suggested that tivantinib presented the highest probability to rank 1st in the cumulative probability of SD, PD, and DCR, while ramucirumab exhibited the highest probability to occupy the 1st place in the cumulative probability of CR, PR, and ORR ( Fig. 4B and Fig. 6A-F).

Cluster analysis.
Results of cluster analysis of SUCRA values based on CR versus PR, CR versus DCR, and CR versus ORR indicated that the short-term efficacies of ramucirumab and sorafenib + erlotinib for patients with AHCC were more favorable, and the placebo, everolimus, and tivantinib treatment delivered a poorer short-term efficacy. Results of cluster analysis of SUCRA value based on PR versus DCR, and PR versus ORR revealed that the efficacies of ramucirumab and sorafenib + erlotinib were better, and the placebo, everolimus, and tivantinib presented a relatively dissatisfactory efficacy. Based on DCR versus ORR, it is indicated that the short-term efficacies of ramucirumab and sorafenib + erlotinib were superior, and the placebo, everolimus, and tivantinib treatment were inferior in short-term efficacy. Cluster analysis was not applied for longterm efficacy, because there was only 1 outcome indicator OS for the long-term effects (Fig. 7).

Discussion
Network meta-analysis was applied to evaluate the efficacy of 7 kinds of targeted drugs for AHCC treatment, namely, sorafenib, ramucirumab, everolimus, brivanib, tivantinib, sunitinib, and sorafenib + erlotinib, which have been widely discussed in HCC. [10] Our results indicated that ramucirumab and sorafenib + erlotinib were superior targeted drugs for AHCC management in terms of short-term effects, and sorafenib + erlotinib exhibited better efficacy in the long term. Initially, our analysis revealed that ramucirumab and sorafenib + erlotinib presented more satisfactory results in the short-term efficacy concerning PR and ORR, and brivanib was better in terms of ORR. Ramucirumab has been approved by the Food and Drug Administration to be used for the treatment of metastatic nonsmall cell lung carcinoma. [37] Ramucirumab may be effective in AHCC treatment by contributing to angiogenesis which plays an important role in the invasion and metastasis of solid tumor and in the neovascularization based on VEGF and vascular endothelial growth factor receptor (VEGFR). [16,38,39] As a kind of VEGFR2 antagonists, ramucirumab is capable of specifically binding VEGFR2 extracellular domain, which can prevent the binding of VEGFR2 with its ligand and thereby inhibit the activation of VEGFR2 and the conduction of downstream signaling pathways. [40] Thus, angiogenesis is prevented to block blood supply to tumor cells, which leads to tumor cell apoptosis. [41] Ueda et al confirm the efficacy of ramucirumab in advanced gastric adenocarcinomas achieving best overall responses with SD (n = 5) and PR (n = 1). [42] Quintela-Fandino et al [43] demonstrate in phase I clinical trial of HCC that the combination of sorafenib + erlotinib can inhibit the tumor cell proliferation in a more efficient fashion. The possible explanation for this result might be that as a small molecule tyrosine ammonia kinase inhibitor, erlotinib can prevent the autophosphorylation of EGFR. [18] The inhibition of cell proliferation, angiogenesis, invasion, and metastasis and the inducement of apoptosis eventually can exert antitumor effects. [44] The experiment of Gu et al [45] indicates that the inhibition of tumor growth and metastasis of sorafenib is achieved by blocking mitogen-activated protein kinases/extracellular signal-regulated kinases/signal transducer and activator of transcription 3 and phosphatidylinositol-3-kinase/protein kinase B/signal transducer and activator of transcription 3 signaling pathways, both of which are different in tumor suppression mechanism which plays a synergistic effect in inhibition of AHCC. Another possible explanation revealed that sorafenib + erlotinib can inhibit the tumor growth mainly in the process of synergy in G0/G1 phase, and meanwhile activate caspase3 and regulate the balance of antiapoptotic molecule B-cell leukemia/ lymphoma 2 and pro-apoptotic molecules B-cell leukemia/ lymphoma 2-associated X protein, [46] thus increasing the inhibitory effect of drug combination on tumor treatment. The  www.md-journal.com Table 1 ORs or HR and 95% confidence intervals of 8 treatment modalities under 3 end indicators according to the network meta-analysis.  Table 2 The results of surface under the cumulative ranking curve of 8 treatment modalities.    [47] indicated the efficacy of erlotinib and sorafenib for the treatment of chemotherapy-naive patients with advanced nonsmall cell lung cancer of 24% PR, 50% SD, and ultimately ORR of 28%. Brivanib functions as a dual inhibitor of VEGF receptor tyrosine kinases, [48] which demonstrates potent antiproliferative and antiangiogenic effects on various tumor cells, including hepatoma cells. [49,50] Llovet et al [30] reported an ORR of 10% for brivanib based on modified response evaluation criteria in solid tumors compared to 2% for placebo. Furthermore, it is indicated in our study that sorafenib + erlotinib presented longer OS in the aspect of long-term efficacy. In phase II trial of unselected patients with pretreated advanced nonsmall cell lung cancer, sorafenib as single agent has conferred a promising antitumor effect, with a median OS of  .7 months and a median progression-free survival of 2.7 months with an acceptable toxicity. [51] Sorafenib is the only systemic treatment to indicate a significant but modest OS benefit, resulting in an era of targeted drugs. [10] Gridelli et al [52] demonstrate that combined erlotinib and sorafenib is feasible in elderly patients with advanced nonsmall cell lung cancer and is correlated with a higher 1-year survival rate than the other arm. It is also suggested by Spigel et al [53] that EGFR-negative patients reveal a benefit for the combination of erlotinib and sorafenib compared with erlotinib alone in terms of progression-free survival and OS, [53] which is consistent with our findings. The advantages and disadvantages of different drug efficacy can be achieved through network meta-analysis method, whereby different treatments with similar disease interventions can be implemented with a quantified comparison and thus confers optimal treatment. [19] However, this study has some limitations. First, the included 11 documents are limited in number; 2nd, limited targeted drug treatments may result in selection bias, and insufficient emphasis on the allocation concealment may exaggerate the effects of treatment; ultimately, the included studies fail to indicate the reasons for patients who were lost during follow-up, which may affect the evaluation of the therapeutic effect.
The results of network meta-analysis in this study suggested that the short-term efficacy of ramucirumab and sorafenib + erlotinib for AHCC patients were superior to other drugs, and sorafenib + erlotinib was better in long-term efficacy, which may shed a little light for the clinical treatment of AHCC. However, the limitation of network meta-analysis requires further studies to comprehensively evaluate the efficacy of different targeted drugs taking overall outcomes of patients, survival rate, and costeffectiveness into consideration.