First-line targ veted therapies of advanced hepatocellular carcinoma: A Bayesian network analysis of randomized controlled trials.

PURPOSE: A variety of targeted drug were developed and proved effective and safe in clinical trials. Our study aims to compare the efficacies and safety of different targeted drugs in advanced hepatocellular carcinoma (HCC) for first-line treatment using a Bayesian network meta-analysis approach.
METHODS: PubMed, Embase, and Cochrane library were searched for randomized controlled trials (RCTs) of advanced HCC patients that treated with different targeted drugs. Time to progress (TTP), overall survival (OS) and progress-free survival (PFS) were calculated as hazard ratios (HRs). Objective response rate (ORR) and the proportion of Grade 3-5 adverse events (G3-5AE) were expressed as odds ratios (ORs). We pooled study-specific HRs and ORs using Bayesian network meta-analyses, and ranked first-line drugs by the surface under the cumulative ranking curve (SUCRA).
RESULTS: A total of 22 RCTs with 9288 patients were enrolled. Brivanib, linifanib, lenvatinib and sorafenib showed a significant improvement on TTP compared to placebo (HR range, 0.45-0.72). Sunitinib (HR = 1.99) and nintedanib (HR = 2.17) showed a significant decline on TTP compared to lenvatinib. Vandetanib (HR = 0.44) and sorafenib (HR = 0.73) showed a significant improvement on OS compared to placebo. There was no significant difference in PFS, ORR and G3-5AE across different drugs. According to cluster rank analysis, vandetanib was the drug with both more effective (OS) and more secure (G3-5AE) compared to Sor followed by nintedanib.
CONCLUSIONS: This network meta-analysis shows that vandetanib, linifanib, lenvatinib and nintedanib potentially may be the best substitution of sorafenib against advanced HCC as first-line targeted drugs. Vandetanib seems to be the best choise with low quality of evidence. For better survival, novel targeted treatment options for HCC are sorely needed.

Introduction

An estimated 42,220 new cases and 30,200 new deaths of hepatocellular and intrahepatic bile duct cancers occurred in the U.S. in 2018 [1]. The majority of these deaths are due to hepatocellular carcinoma (HCC), the most common primary hepatic cancer [2]. Globally liver cancer is the fourth causes of cancer death for mortality [3]. HCC is most commonly associated with chronic hepatitis B virus or hepatitis C virus infections, especially with cirrhosis, which limits the feasibility of surgical resection [4]. Liver transplantation and surgical resection still remain the most effective treatment for early stage HCC in good surgical candidates. Unfortunately, the vast majority of patients are in advanced stages with unresectable tumors when they were diagnosed as HCC. In the past, the prognosis of advanced HCC was poor and its treatment was limited to transarterial chemoembolization, radiofrequency ablation, radiotherapy, and systemic pharmacotherapy [5].

In the European SHARP Trial, the multi-targeted small molecule tyrosine kinase inhibitor (TKI) sorafenib was demonstrated to improve median survival over placebo for unresectable HCC patients for the first time [6]. Subsequently, more targeted drugs were developed and proved effective and safe in their phase II or III clinical trials [7]. Although the effectiveness and safety of these drugs have been compared to sorafenib or placebo, they have not been compared to each other head-to-head [8]. In order to further assess the evidence on the efficacy and safety of targeted drugs for the treatment of HCC patients, we performed this Bayesian network meta-analysis (NMA) to compare the survivals, objective response rates (ORRs) and adverse events (AEs) among different targeted drugs for HCC.

Materials and methods

This review was performed following the preferred reporting items for the systematic reviews incorporating network meta-analyses [9] (S1 File). This network meta-analysis has been registered in the PROSPERO public database (CRD42019145188; http://www.crd.york.ac.uk/PROSPERO).

Eligibility criteria

We included randomized controlled trials (RCTs) of adult patients with advanced or unresected hepatocellular carcinoma. To avoid the influence of other treatments, the key inclusion criteria for included study populations were as follows: First, it should last more than 4 weeks since most recent local therapy or no local therapy. Second, the patients did not receive prior systemic therapy. The interventions of interest were the targeted drugs for HCC: Bevacizumab plus erlotinib (Bev + Erl), brivanib (Bri), cabozantinib (Cab), codrituzumab (Cod), dovitinib (Dov), erlotinib plus sorafenib (Erl + Sor), everolimus plus sorafenib (Eve + Sor), lenvatinib (Len), linifanib (Lin), nintedanib (Nin), orantinib (Ora), regorafenib (Reg), sorafenib (Sor), sunitinib (Sun), tigatuzumab (Tig), vandetanib (Van). The efficacy and safety outcomes assessed were time to progress (TTP), overall survival (OS), progress-free survival (PFS), objective response rate (ORR), and the proportion of Grade 3–5 adverse events (G3-5AE).

Search strategy and study selection

Two researchers (W.D. & Y.T.) systematically searched Pubmed, Embase and the Cochrane Library using a well-developed search strategy without language restriction from inception to Jun 30th, 2019 (S2 Table). Additionally, relevant references were also searched. Unpublished literatures and conference abstracts were not included.

Two reviewers (W.X. & Y.W.) independently screened out the candidate articles via scanning all titles, abstracts and full-texts. A third reviewer (W.D.) made the final decision of the disagreements on candidate articles through consensus.

Data extraction

Two reviewers (W.D. & Y.T.) extracted relevant data including study author, post time, region, sample size, patient characteristics (age, gender, Eastern Cooperative Oncology Group [ECOG] score, Barcelona Clinic Liver Cancer [BCLC] stage), mode, dose and duration of treatments, and outcomes of interest, independently. A third reviewer (X.X.) made the final decision of the disagreements were via discussion.

Quality assessment

The quality and the risk of bias of RCTs in this study was assessed using the quality criteria of the Cochrane Collaboration’s tool (S1 Table) [10]. The Grading of Recommendations Assessment, Development and Evaluation (GRADE) Working Group approach was used to assess the quality of evidence (QoE) in each of the direct, indirect, and NMA estimates [11, 12]. For direct comparison, we graded evidence from the five aspects; risk of bias, inconsistency, indirectness, imprecision and publication bias, using the standard GRADE approach. For indirect comparison, we rated evidence according to the lower grades of direct comparisons and intransitivity. For NMA estimates, we rated evidence according to the higher grades of the direct and indirect comparisons and incoherence.

Data synthesis and analysis

Results regarding the OS, PFS and TTP are expressed as hazard ratios (HRs) with 95% confidence intervals (CIs). Results regarding ORR and G3-5AE are expressed as odds ratios (ORs) with 95% CIs. If HRs could not be acquired directly, they were extracted from Kaplan-Meier curves using the method described by Parmar et al [13]. If there were different HRs or ORs based on different evaluation criteria in the same article, we selected the result according to the latest criteria. We did direct pairwise meta-analyses of head-to-head comparisons with RevMan version 5.3.0 (Cochrane Collaboration). The evaluation of heterogeneity among studies was performed by Cochran's Q test and Higgin's I2 statistics. The heterogeneity among all included studies was suggested significant when I2>50% and/or P<0.05, then a random-effect model was used (DerSimoniane-Laird method); otherwise, a fixed-effect model (Mantel-Haenszel method) was used.

We did Bayesian network meta-analysis with the package ‘rjags’ version 4–9 and the package ‘GeMTC’ version 0.8–2 in R version 3.6.1 (https://www.r-project.org). The merged HRs and/or ORs of relative treatment effects are reported as the median and accompanying 95% credibility intervals (95% CrI) of the posterior distribution. We drew network diagrams with Stata/MP version 14.0 (4905 Lakeway Drive, College Station, TX77845, USA). Hierarchical Bayesian modeling of the present network meta-analysis conformed to the National Institute for Health and Excellence Decision Support Units (NICEDSU) guidelines [14]. To confirm the transitivity and the loop-specific consistency assumption, pairwise direct and indirect effect estimates of closed loops of evidence were inspected for any disagreement [15]. The transitivity was assessed by examining the patient baseline characteristics across studies (age, gender, performance status and tumor stage), treatment stage and treatment protocol [16]. The global test for inconsistency assumption was conducted with the consistency and inconsistency (unrelated mean effects) models. The consistency between direct and indirect comparison was assessed via using a node-splitting test within each network with a loop [17]. The heterogeneity of network meta-analysis was evaluated with the posterior median of the between-trials standard deviation (σ) [14], while comparison-adjusted funnel plot was used to detect the presence of small-study effects or publication bias.

We undertook Markov Chain Monte Carlo (MCMC) simulation as Bayesian inference to calculate the posterior distributions of the interrogated nodes within the framework of the chosen models and likelihood function on the basis of prior assumptions. We used four different sets of initial values to fit the model, yielding 400,000 iterations (100,000 per chain) to obtain the posterior distributions of model parameters then used 50,000 burn-ins and a thinning interval of 10 for each chain. Autocorrelation function was used to assess the convergence of iterations. Global model fit and parsimony was compared between different fitted models to decide on the most accurate model. The posterior mean of the total residual deviance and deviance information criterion (DIC) was used to choose a more appropriate model [18, 19]. The model with a lower DIC was considered as a more appropriate model. The threshold for the statistical significance was chosen as a two-tailed alpha = 0.05.

In order to determinate intervention rankings for outcomes, rank probabilities were extracted from the network meta-analysis. By merging the rank probabilities of different drugs, we generated the surface under the cumulative ranking curve (SUCRA) to simplify the ranking information as a few numbers [20]. It ranks from 0 to 1. It would be 1 when a treatment is certain to be the best and 0 when a treatment is certain to be the worst. To simultaneously compare the efficacy and safety of each drugs, we jointly presented the SUCRA value of OS and G3-5AE on the clustered ranking plot.

Results

Of 2,808 articles were collected from the databases mentioned above. After removing all duplicate articles and checking all titles and abstracts, 26 studies remained. After further full-texts screening, four researches were excluded (one study [21] was lack of control group, three studies [22-24] were the Sub-studies for previous trials). Finally, a total of 22 RCTs including 9288 patients from all over the world were included in this network meta-analysis (Fig 1) [6, 25–45].

Fig 1

Flow diagram of study selection.

Study characteristics

The main characteristics of the included studies were summarized in Table 1. The median age in the 22 RCTs ranged from 51 to 70 years with a majority of male participants. The sample size ranged from 67 to 1155 patients. The majority of ECOG scores were 0–1. The majority of BCLC stages were B-C. The included RCTs compared thirteen different drugs (bevacizumab, erlotinib, brivanib, dovitinib, erlotinib, everolimus, lenvatinib, linifanib, nintedanib, orantinib, sorafenib, sunitinib, tigatuzumab, vandetanib), which were only compared to sorafenib or placebo. The targeted drug treatment programs and their abbreviations are shown in S4 file. The main characteristics of the included studies are shown in Table 1. As shown in S1 Table, only twelve studies [25–29, 32, 34, 35, 37, 38, 41, 42] were considered with high risk of bias at blinding of participants and personnel due to their open-label design. There was no evidence of substantial imbalance in the distribution of the effect modifiers across trials in the network. A connected network diagram formed by all evidences is provided in Fig 2. The dosage regimen modes of the same drugs across studies were consistent. By examining the patient baseline characteristics, treatment stage and protocol, there was no evidence that the transitivity assumption was violated in any of the networks.

Table 1

Main characteristics of included studies.

Study	Year	Intervene	Samples	Age	Gender (M/F)	ECOG (0/1/2)	BCLC stage (A/B/C/D)	HBV infection	White	Black	Asian
Yen 2018 [25]	2018	Nin	63	58	57/6	35/27/1	1/9/53/0	40	0	0	63
		Sor	32	62	26/6	18/14/0	1/1/30/0	20	0	0	32
Xu 2018 [26]	2018	Sun	51	60	42/9	29/22/0	6/11/34/0	41	0	0	51
		Sor	53	62	41/12	25/28/0	5/16/32/0	44	0	0	53
Thomas 2018 [27]	2018	Bev + Erl	47	61	NR	15/32/0	1/14/32/0	NR	28	NR	NR
		Sor	43	61	NR	17/25/1	4/11/28/0	NR	31	NR	NR
Palmer 2018 [28]	2018	Nin	62	66	48/14	32/28/2	1/15/45/1	4	57	0	0
		Sor	31	64	24/7	18/10/3	1/7/23/0	7	24	1	1
Kudo Finn 2018 [29]	2018	Len	478	63	405/73	304/174/0	0/104/374/0	NR	135	NR	334
		Sor	476	62	401/75	301/175/0	0/92/384/0	NR	141	NR	326
Kudo Cheng 2018 [30]	2018	Ora	444	67	363/81	401/43/0	158/209/74/0*	170	0	0	444
		Pla	444	66	364/80	406/38/0	135/229/72/0*	202	0	0	444
Meyer 2017 [31]	2017	Sor	157	65	139/18	98/58/0*	NR	15	157	0	0
		Pla	156	68	138/18	97/58/0*	NR	14	156	0	0
Lee 2017 [32]	2017	Sor	36	60	30/6	NR	9/27/0/0	NR	0	0	36
		Pla	36	62	32	NR	15/21/0/0	NR	0	0	36
Lencioni 2016 [33]	2016	Sor	154	64.5	135/19	154/0/0	0/154/0/0	60	78	NR	59
		Pla	153	63	126.27	153/0/0	0/153/0/0	51	79	NR	57
Koeberle 2016 [34]	2016	Eve + Sor	59	66	48/18	35/24/0	0/15/44/0	10	59	0	0
		Sor	46	65	40/15	33/13/0	0/14/32/0	8	46	0	0
Cheng 2016 [35]	2016	Dov	82	56	73/9	52/30/0	0/2/80/0	NR	0	0	82
		Sor	83	56	67/16	53/29/0*	0/2/81/0	NR	0	0	83
Zhu 2015 [36]	2015	Erl + Sor	362	60.5	295/67	222/140/0	0/60/302/0	122	186	NR	88
		Sor	358	60	286/72	216/142/0	0/48/310/0	133	183	NR	90
Cheng 2015 [37]	2015	Tig 6mg + Sor	54	62.5	45/9	31/23/0	NR	45	NR	NR	53
		Tig 2mg + Sor	53	63	45/8	32/21/0	NR	33	NR	NR	52
		Sor	55	66	44/11	30/25/0*	NR	25	NR	NR	54
Cainap 2015 [38]	2015	Lin	514	59	444/70	323/191/0	0/81/433/0	251	NR	NR	339
		Sor	521	60	436/85	344/176/0	0/102/418/0	257	NR	NR	350
Kudo 2014 [39]	2014	Bri	249	57	206/43	201/48/0	65/129/54/1	158	22	NR	216
		Pla	253	59	216/37	203/50/0	57/150/44/2	168	23	NR	218
Johnson 2013 [40]	2013	Bri	577	61	483/94	361/216	37/95/444/0	254	134	NR	346
		Sor	578	60	484/94	352/226	30/97/449/0	258	135	NR	372
Inaba 2013 [41]	2013	Ora	50	NR	39/11	45/5/0	21/24/5/0	2	NR	NR	50
		Pla	51	NR	43/8	49/2/0	22/27/2/0	4	NR	NR	51
Cheng 2013 [42]	2013	Sun	530	59	436/94	278/248/0*	0/67/462/0	290	111	6	411
		Sor	544	59	459/85	288/254/0*	0/89/454/0	288	112	10	418
Hsu 2012 [43]	2012	Van 300mg	19	54	18/1	NR	0/4/15/0	14	NR	NR	19
		Van 100mg	25	61	17/8	NR	0/4/21/0	16	NR	NR	25
		Pla	23	56	20/3	NR	0/5/18/0	17	NR	NR	23
Kudo 2011 [44]	2011	Sor	229	69	174/55	201/28/0	NR	47	NR	NR	229
		Pla	229	70	168/61	202/27/0	NR	52	NR	NR	229
Chen 2009 [45]	2009	Sor	150	51	127/23	38/104/8	0/0/143/0*	106	NR	NR	150
		Pla	76	52	66/10	21/51/4	0/0/73/0*	59	NR	NR	76
Llovet 2008 [6]	2008	Sor	299	NR	260/39	161/114/24	0/54/244/0	56*	NR	NR	NR
		Pla	303	NR	264/39	164/117/22	0/51/252/0	55*	NR	NR	NR

* Data were not available for all patients; NR: Not report.

Fig 2

Network diagram of all studies.

Time to progress

Seventeen RCTs [6, 25, 26, 28–31, 33, 34, 36–40, 42, 44, 45] reporting information on TTP were included for meta-analysis. Direct meta-analyses (S1 Fig) confirmed a significant improvement on TTP compared to sorafenib (HR: 0.73; 95%CI: 0.61–0.89) and brivanib (HR: 0.61; 95%CI: 0.48–0.78) over placebo. A connected network diagram formed by TTP is provided in S2 Fig. According to the node-splitting analysis, there was not any significant inconsistency between the direct and indirect comparisons (Pla vs. Bri, P = 0.54; Sor vs. Bri, P = 0.54; Sor vs. Pla, P = 0.54), as shown in S3 Fig. The NMA heterogeneity was low (σ = 0.17; 95%CrI: 0.03–0.43), as shown in S2 Table. The NMA synthesis showed that four drugs (brivanib, lenvatinib, linifanib and sorafenib) achieved a significant benefit on TTP over placebo (HR range, 0.45–0.72). According to SUCRA, three highest ranking drugs were lenvatinib (0.94), linifanib (0.84) and brivanib (0.67), which were in red in Table 2.

Table 2

Network meta-analyses for TTP (Findings are expressed as HR (95% CrI), use of random-effect model).

SUCRA	Drugs	Bri	Erl+Sor	Eve+Sor	Len	Lin	Nin	Ora	Pla	Sor	Sun	Tig 2mg + Sor	Tig 6mg + Sor
0.67	Bri	Bri	1.21 (0.67, 2.21)	1.05 (0.54, 2.08)	0.67 (0.37, 1.21)	0.80 (0.45, 1.47)	1.45 (0.80, 2.66)	1.27 (0.71, 2.33)	1.48 (1.04, 2.14)	1.06 (0.75, 1.53)	1.33 (0.84, 2.33)	1.19 (0.62, 2.28)	1.22 (0.65, 2.32)
0.43	Erl+Sor	0.83 (0.45, 1.50)	Erl+Sor	0.87 (0.41, 1.82)	0.55 (0.28, 1.07)	0.66 (0.34, 1.30)	1.20 (0.61, 2.38)	1.05 (0.52, 2.11)	1.22 (0.73, 2.06)	0.88 (0.54, 1.41)	1.10 (0.63, 2.12)	0.99 (0.47, 2.02)	1.01 (0.49, 2.06)
0.58	Eve+Sor	0.95 (0.48, 1.84)	1.15 (0.55, 2.43)	Eve+Sor	0.64 (0.30, 1.32)	0.77 (0.37, 1.60)	1.38 (0.65, 2.93)	1.21 (0.56, 2.59)	1.41 (0.76, 2.59)	1.01 (0.57, 1.78)	1.27 (0.66, 2.56)	1.13 (0.51, 2.52)	1.16 (0.53, 2.55)
0.94	Len	1.50 (0.83, 2.67)	1.82 (0.93, 3.53)	1.57 (0.76, 3.29)	Len	1.21 (0.62, 2.38)	2.17 (1.12, 4.27)	1.91 (0.95, 3.81)	2.22 (1.33, 3.71)	1.59 (1.00, 2.53)	1.99 (1.15, 3.78)	1.79 (0.86, 3.64)	1.84 (0.90, 3.73)
0.84	Lin	1.24 (0.68, 2.22)	1.50 (0.77, 2.95)	1.30 (0.63, 2.74)	0.83 (0.42, 1.61)	Lin	1.80 (0.92, 3.53)	1.58 (0.78, 3.18)	1.84 (1.09, 3.11)	1.32 (0.82, 2.12)	1.65 (0.95, 3.15)	1.48 (0.71, 3.04)	1.52 (0.74, 3.14)
0.23	Nin	0.69 (0.38, 1.24)	0.83 (0.42, 1.63)	0.72 (0.34, 1.53)	0.46 (0.23, 0.89)	0.56 (0.28, 1.09)	Nin	0.88 (0.43, 1.77)	1.02 (0.60, 1.73)	0.73 (0.45, 1.19)	0.92 (0.52, 1.74)	0.82 (0.39, 1.71)	0.84 (0.41, 1.74)
0.37	Ora	0.79 (0.43, 1.40)	0.95 (0.48, 1.93)	0.82 (0.39, 1.78)	0.52 (0.26, 1.05)	0.63 (0.31, 1.27)	1.14 (0.57, 2.32)	Ora	1.16 (0.73, 1.84)	0.83 (0.50, 1.39)	1.05 (0.58, 2.05)	0.94 (0.44, 1.97)	0.96 (0.46, 2.01)
0.16	Pla	0.68 (0.47, 0.96)	0.82 (0.48, 1.37)	0.71 (0.39, 1.32)	0.45 (0.27, 0.75)	0.54 (0.32, 0.92)	0.98 (0.58, 1.66)	0.86 (0.54, 1.37)	Pla	0.72 (0.58, 0.89)	0.90 (0.61, 1.43)	0.80 (0.44, 1.45)	0.83 (0.47, 1.47)
0.61	Sor	0.94 (0.66, 1.33)	1.14 (0.71, 1.84)	0.99 (0.56, 1.76)	0.63 (0.39, 1.00)	0.76 (0.47, 1.22)	1.37 (0.84, 2.23)	1.20 (0.72, 2.00)	1.40 (1.13, 1.74)	Sor	1.26 (0.91, 1.90)	1.12 (0.64, 1.95)	1.15 (0.68, 1.98)
0.29	Sun	0.75 (0.43, 1.19)	0.91 (0.47, 1.59)	0.79 (0.39, 1.51)	0.50 (0.26, 0.87)	0.61 (0.32, 1.05)	1.09 (0.58, 1.93)	0.96 (0.49, 1.71)	1.11 (0.70, 1.63)	0.80 (0.53, 1.10)	Sun	0.89 (0.44, 1.66)	0.92 (0.47, 1.69)
0.45	Tig 2mg + Sor	0.84 (0.44, 1.62)	1.02 (0.50, 2.13)	0.88 (0.40, 1.98)	0.56 (0.28, 1.16)	0.68 (0.33, 1.40)	1.22 (0.59, 2.57)	1.07 (0.51, 2.29)	1.24 (0.69, 2.27)	0.89 (0.51, 1.55)	1.12 (0.60, 2.26)	Tig 2mg + Sor	1.03 (0.66, 1.61)
0.41	Tig 6mg + Sor	0.82 (0.43, 1.54)	0.99 (0.49, 2.03)	0.86 (0.39, 1.87)	0.54 (0.27, 1.11)	0.66 (0.32, 1.34)	1.18 (0.58, 2.46)	1.04 (0.50, 2.18)	1.21 (0.68, 2.15)	0.87 (0.51, 1.47)	1.09 (0.59, 2.15)	0.97 (0.62, 1.52)	Tig 6mg + Sor

The values in red shading were the highest three SUCRAs. The values in green shading were statistically significant. The texts in yellow shading were targeted drugs.

Progression-free survival

Eight RCTs [25, 26, 28, 29, 31, 38, 41, 43] reporting information on PFS were included for meta-analysis. Direct meta-analyses (S4 Fig) confirmed a significant improvement on PFS compared to Lenvatinib (HR: 0.66; 95%CI: 0.56–0.77) and Linifanib (HR: 0.81; 95%CI: 0.69–0.95) over sorafenib. A star-shaped network diagram formed by PFS is provided in S5 Fig. For no closed loop, node-splitting test of studies for PFS was not applicable. The NMA heterogeneity was low (σ = 0.18; 95%CrI: 0.01–0.43), as shown in S2 Table. The NMA synthesis showed that there was no significant difference on PFS among drugs. According to SUCRA, three highest ranking drugs were lenvatinib (0.77), vandetanib (0.77) and orantinib (0.68) which were in red in Table 3.

Table 3

Network meta-analyses for PFS (Findings are expressed as HR (95% CrI), use of random-effect model).

SUCRA	Drugs	Len	Lin	Nin	Ora	Pla	Sor	Sun	Van 100mg	Van 300mg
0.77	Len	Len	1.23 (0.58, 2.59)	1.70 (0.84, 3.44)	1.07 (0.39, 2.94)	1.53 (0.70, 3.26)	1.51 (0.90, 2.55)	2.16 (0.95, 4.89)	0.97 (0.35, 2.66)	1.08 (0.40, 2.91)
0.58	Lin	0.81 (0.39, 1.72)	Lin	1.38 (0.67, 2.83)	0.87 (0.32, 2.40)	1.24 (0.58, 2.66)	1.23 (0.72, 2.09)	1.76 (0.77, 3.97)	0.79 (0.29, 2.17)	0.88 (0.32, 2.37)
0.26	Nin	0.59 (0.29, 1.20)	0.72 (0.35, 1.49)	Nin	0.63 (0.24, 1.68)	0.90 (0.43, 1.88)	0.89 (0.55, 1.46)	1.27 (0.57, 2.84)	0.57 (0.21, 1.56)	0.64 (0.24, 1.68)
0.68	Ora	0.93 (0.34, 2.53)	1.15 (0.42, 3.11)	1.58 (0.60, 4.24)	Ora	1.43 (0.75, 2.74)	1.41 (0.60, 3.33)	2.01 (0.70, 5.79)	0.91 (0.36, 2.33)	1.01 (0.41, 2.53)
0.32	Pla	0.65 (0.31, 1.42)	0.81 (0.38, 1.73)	1.11 (0.53, 2.35)	0.70 (0.37, 1.34)	Pla	0.99 (0.57, 1.75)	1.42 (0.61, 3.28)	0.64 (0.32, 1.26)	0.71 (0.37, 1.35)
0.35	Sor	0.66 (0.39, 1.11)	0.81 (0.48, 1.38)	1.12 (0.68, 1.83)	0.71 (0.30, 1.67)	1.01 (0.57, 1.76)	Sor	1.43 (0.75, 2.69)	0.64 (0.27, 1.53)	0.72 (0.30, 1.66)
0.11	Sun	0.46 (0.20, 1.06)	0.57 (0.25, 1.30)	0.79 (0.35, 1.74)	0.50 (0.17, 1.43)	0.71 (0.30, 1.65)	0.70 (0.37, 1.33)	Sun	0.45 (0.15, 1.33)	0.50 (0.17, 1.44)
0.77	Van 100mg	1.03 (0.38, 2.83)	1.26 (0.46, 3.46)	1.75 (0.64, 4.73)	1.10 (0.43, 2.81)	1.57 (0.80, 3.08)	1.56 (0.65, 3.72)	2.23 (0.75, 6.54)	Van 100mg	1.11 (0.65, 1.89)
0.66	Van 300mg	0.93 (0.34, 2.50)	1.14 (0.42, 3.09)	1.57 (0.59, 4.17)	0.99 (0.40, 2.47)	1.41 (0.74, 2.69)	1.40 (0.60, 3.29)	2.00 (0.69, 5.79)	0.90 (0.53, 1.54)	Van 300mg

The values in red shading were the highest three SUCRAs. The values in green shading were statistically significant. The texts in yellow shading were targeted drugs.

Overall survival

Twenty RCTs [6, 25, 27–32, 34–45] reporting information on OS were included for meta-analysis. Direct meta-analyses (S6 Fig) confirmed a significant improvement on OS compared to sorafenib (HR: 0.72; 95%CI: 0.54–0.94) and Vandetanib 100 mg (HR: 0.44; 95%CI: 0.22–0.87) over placebo. A connected network diagram formed by OS is provided in S7 Fig. According to the node-splitting analysis, there was not any significant inconsistency between the direct and indirect comparisons (Pla vs. Bri, P = 0.62; Sor vs. Bri, P = 0.61; Sor vs. Pla, P = 0.62), as shown in S8 Fig. The NMA heterogeneity was low (σ = 0.15; 95%CrI: 0.01–0.49), as shown in S2 Table. The NMA synthesis showed that two treatments (Vandetanib 100 mg and sorafenib) achieved a significant benefit on OS over placebo (HR range, 0.44–0.73). According to SUCRA, three highest ranking interventions were tigatuzumab 6mg (0.73), vandetanib 100mg (0.92) and vandetanib 300mg (0.70), which were in red in Table 4.

Table 4

Network meta-analyses for OS (Findings are expressed as HR (95% CrI), use of random-effect model).

SUCRA	Drugs	Bev+Erl	Bri	Dov	Erl+Sor	Eve+Sor	Len	Lin	Nin	Ora	Pla	Sor	Sun	Tig 2mg + Sor	Tig 6mg + Sor	Van 100mg	Van 300mg
0.62	Bev+Erl	Bev+Erl	1.21 (0.60, 2.53)	1.38 (0.59, 3.15)	1.01 (0.47, 2.18)	1.20 (0.51, 2.81)	1.00 (0.46, 2.17)	1.14 (0.52, 2.48)	0.99 (0.45, 2.14)	1.59 (0.75, 3.40)	1.48 (0.76, 2.92)	1.08 (0.57, 2.03)	1.41 (0.64, 3.01)	1.35 (0.58, 3.06)	0.91 (0.39, 2.09)	0.65 (0.26, 1.68)	0.89 (0.35, 2.31)
0.41	Bri	0.82 (0.39, 1.67)	Bri	1.14 (0.58, 2.12)	0.83 (0.45, 1.48)	0.99 (0.49, 1.90)	0.82 (0.45, 1.46)	0.94 (0.51, 1.66)	0.81 (0.45, 1.43)	1.32 (0.78, 2.16)	1.23 (0.83, 1.75)	0.90 (0.61, 1.25)	1.16 (0.62, 2.03)	1.11 (0.56, 2.10)	0.75 (0.38, 1.42)	0.54 (0.25, 1.12)	0.73 (0.34, 1.54)
0.29	Dov	0.73 (0.32, 1.68)	0.88 (0.47, 1.73)	Dov	0.73 (0.36, 1.52)	0.87 (0.39, 1.92)	0.72 (0.35, 1.48)	0.83 (0.40, 1.70)	0.72 (0.35, 1.47)	1.16 (0.58, 2.33)	1.08 (0.60, 1.98)	0.79 (0.46, 1.37)	1.02 (0.49, 2.09)	0.98 (0.45, 2.11)	0.66 (0.31, 1.44)	0.47 (0.20, 1.16)	0.65 (0.27, 1.58)
0.65	Erl+Sor	0.99 (0.46, 2.15)	1.20 (0.68, 2.23)	1.37 (0.66, 2.81)	Erl+Sor	1.19 (0.57, 2.46)	0.99 (0.51, 1.92)	1.13 (0.58, 2.20)	0.98 (0.51, 1.90)	1.59 (0.85, 2.99)	1.47 (0.87, 2.52)	1.08 (0.67, 1.73)	1.40 (0.71, 2.69)	1.33 (0.64, 2.72)	0.91 (0.44, 1.85)	0.65 (0.28, 1.50)	0.88 (0.38, 2.05)
0.44	Eve+Sor	0.84 (0.36, 1.95)	1.01 (0.53, 2.02)	1.15 (0.52, 2.53)	0.84 (0.41, 1.75)	Eve+Sor	0.83 (0.40, 1.74)	0.96 (0.46, 1.98)	0.82 (0.40, 1.71)	1.33 (0.65, 2.74)	1.24 (0.67, 2.32)	0.91 (0.51, 1.61)	1.17 (0.57, 2.44)	1.12 (0.51, 2.47)	0.76 (0.34, 1.69)	0.54 (0.22, 1.36)	0.74 (0.31, 1.84)
0.66	Len	1.00 (0.46, 2.18)	1.21 (0.69, 2.22)	1.38 (0.68, 2.85)	1.01 (0.52, 1.97)	1.20 (0.58, 2.52)	Len	1.14 (0.60, 2.21)	0.99 (0.52, 1.88)	1.60 (0.86, 3.02)	1.49 (0.89, 2.53)	1.09 (0.68, 1.72)	1.41 (0.73, 2.70)	1.35 (0.66, 2.75)	0.91 (0.45, 1.89)	0.65 (0.29, 1.50)	0.89 (0.39, 2.05)
0.49	Lin	0.88 (0.40, 1.91)	1.06 (0.60, 1.97)	1.21 (0.59, 2.47)	0.88 (0.45, 1.71)	1.05 (0.51, 2.19)	0.87 (0.45, 1.68)	Lin	0.87 (0.45, 1.64)	1.40 (0.76, 2.65)	1.30 (0.78, 2.22)	0.95 (0.60, 1.52)	1.23 (0.63, 2.36)	1.18 (0.57, 2.39)	0.80 (0.39, 1.61)	0.57 (0.25, 1.32)	0.78 (0.34, 1.80)
0.65	Nin	1.01 (0.47, 2.21)	1.23 (0.70, 2.24)	1.39 (0.68, 2.84)	1.02 (0.53, 1.96)	1.21 (0.58, 2.52)	1.01 (0.53, 1.93)	1.15 (0.61, 2.23)	Nin	1.61 (0.88, 3.03)	1.50 (0.90, 2.52)	1.10 (0.69, 1.74)	1.42 (0.75, 2.71)	1.36 (0.67, 2.75)	0.92 (0.45, 1.89)	0.66 (0.29, 1.52)	0.90 (0.40, 2.09)
0.13	Ora	0.63 (0.29, 1.32)	0.76 (0.46, 1.28)	0.86 (0.43, 1.72)	0.63 (0.33, 1.18)	0.75 (0.36, 1.53)	0.63 (0.33, 1.16)	0.72 (0.38, 1.32)	0.62 (0.33, 1.14)	Ora	0.93 (0.65, 1.32)	0.68 (0.44, 1.03)	0.88 (0.46, 1.63)	0.84 (0.42, 1.65)	0.57 (0.29, 1.13)	0.41 (0.20, 0.85)	0.56 (0.27, 1.16)
0.18	Pla	0.67 (0.34, 1.31)	0.82 (0.57, 1.21)	0.93 (0.51, 1.67)	0.68 (0.40, 1.14)	0.81 (0.43, 1.48)	0.67 (0.40, 1.12)	0.77 (0.45, 1.28)	0.67 (0.40, 1.11)	1.07 (0.76, 1.53)	Pla	0.73 (0.57, 0.92)	0.95 (0.56, 1.57)	0.91 (0.50, 1.62)	0.61 (0.34, 1.10)	0.44 (0.23, 0.84)	0.60 (0.31, 1.16)
0.57	Sor	0.92 (0.49, 1.74)	1.11 (0.80, 1.64)	1.27 (0.73, 2.20)	0.93 (0.58, 1.49)	1.10 (0.62, 1.95)	0.92 (0.58, 1.47)	1.05 (0.66, 1.68)	0.91 (0.58, 1.44)	1.47 (0.97, 2.26)	1.37 (1.09, 1.75)	Sor	1.30 (0.81, 2.05)	1.24 (0.72, 2.14)	0.84 (0.49, 1.45)	0.60 (0.30, 1.20)	0.82 (0.41, 1.67)
0.25	Sun	0.71 (0.33, 1.55)	0.86 (0.49, 1.61)	0.98 (0.48, 2.02)	0.72 (0.37, 1.40)	0.85 (0.41, 1.76)	0.71 (0.37, 1.38)	0.81 (0.42, 1.58)	0.70 (0.37, 1.33)	1.13 (0.61, 2.16)	1.05 (0.64, 1.79)	0.77 (0.49, 1.23)	Sun	0.95 (0.47, 1.95)	0.65 (0.32, 1.33)	0.46 (0.21, 1.07)	0.63 (0.28, 1.47)
0.31	Tig 2mg + Sor	0.74 (0.33, 1.73)	0.90 (0.48, 1.77)	1.02 (0.47, 2.20)	0.75 (0.37, 1.56)	0.89 (0.40, 1.97)	0.74 (0.36, 1.52)	0.85 (0.42, 1.74)	0.74 (0.36, 1.50)	1.19 (0.60, 2.39)	1.10 (0.62, 2.01)	0.81 (0.47, 1.39)	1.05 (0.51, 2.13)	Tig 2mg + Sor	0.68 (0.43, 1.07)	0.48 (0.20, 1.17)	0.66 (0.28, 1.63)
0.73	Tig 6mg + Sor	1.10 (0.48, 2.55)	1.33 (0.70, 2.62)	1.51 (0.70, 3.27)	1.10 (0.54, 2.27)	1.31 (0.59, 2.92)	1.09 (0.53, 2.24)	1.25 (0.62, 2.54)	1.08 (0.53, 2.21)	1.75 (0.89, 3.49)	1.63 (0.91, 2.96)	1.19 (0.69, 2.05)	1.54 (0.75, 3.11)	1.47 (0.93, 2.32)	Tig 6mg + Sor	0.72 (0.30, 1.72)	0.97 (0.41, 2.38)
0.92	Van 100mg	1.53 (0.59, 3.90)	1.86 (0.89, 3.92)	2.11 (0.86, 5.04)	1.54 (0.67, 3.51)	1.84 (0.74, 4.48)	1.53 (0.67, 3.46)	1.75 (0.76, 3.94)	1.51 (0.66, 3.43)	2.45 (1.18, 5.05)	2.27 (1.19, 4.33)	1.66 (0.83, 3.29)	2.15 (0.93, 4.88)	2.06 (0.86, 4.91)	1.39 (0.58, 3.33)	Van 100mg	1.37 (0.85, 2.20)
0.70	Van 300mg	1.12 (0.43, 2.89)	1.36 (0.65, 2.91)	1.55 (0.63, 3.71)	1.13 (0.49, 2.60)	1.35 (0.54, 3.27)	1.12 (0.49, 2.54)	1.29 (0.55, 2.91)	1.11 (0.48, 2.52)	1.79 (0.86, 3.74)	1.67 (0.86, 3.20)	1.22 (0.60, 2.43)	1.58 (0.68, 3.59)	1.51 (0.61, 3.60)	1.03 (0.42, 2.44)	0.73 (0.46, 1.17)	Van 300mg

The values in red shading were the highest three SUCRAs. The values in green shading were statistically significant. The texts in yellow shading were targeted drugs.

Objective response rates

Thirteen RCTs [6, 26, 28, 29, 31–33, 35, 36, 38–40, 45] reporting information on ORR were included for meta-analysis. Direct meta-analyses (S9 Fig) confirmed that ORR was better in case of lenvatinib (HR: 3.11; 95%CI: 2.14–4.52) or linifanib (HR: 1.72; 95%CI: 1.09–2.72) than sorafenib, and ORR was bad in case of brivanib (HR: 0.21; 95%CI: 0.14–0.31) or sunitinib (HR: 0.42; 95%CI: 0.19–0.93) than sorafenib. A connected network diagram formed by ORR is provided in S10 Fig. According to the node-splitting analysis, there was not any significant inconsistency between the direct and indirect comparisons (Pla vs. Bri, P = 0.13; Sor vs. Bri, P = 0.13; Sor vs. Pla, P = 0.13), as shown in S11 Fig. The NMA heterogeneity was low (σ = 0.72; 95%CrI: 0.31–1.45), as shown in S2 Table. The NMA synthesis showed that there was no significant difference on ORR among drugs. According to SUCRA, three highest ranking interventions were lenvatinib (0.88), erlotinib plus sorafenib (0.73) and linifanib (0.73) which were in red in Table 5.

Table 5

Network meta-analyses for ORR (Findings are expressed as OR (95% CrI), use of random-effect model).

SUCRA	Drugs	Bri	Dov	Erl+Sor	Len	Lin	Pla	Sor	Sun
0.19	Bri	Bri	1.41 (0.13, 16.54)	4.83 (0.53, 48.67)	8.54 (0.97, 76.55)	4.73 (0.54, 43.77)	1.35 (0.36, 4.97)	2.72 (0.76, 10.54)	1.14 (0.12, 11.73)
0.34	Dov	0.71 (0.06, 7.83)	Dov	3.39 (0.22, 54.05)	6.01 (0.41, 92.30)	3.35 (0.22, 50.15)	0.95 (0.10, 8.80)	1.93 (0.25, 15.29)	0.80 (0.05, 13.44)
0.73	Erl+Sor	0.21 (0.02, 1.88)	0.29 (0.02, 4.54)	Erl+Sor	1.76 (0.14, 22.92)	0.98 (0.08, 12.74)	0.28 (0.03, 2.02)	0.56 (0.09, 3.53)	0.24 (0.02, 3.17)
0.88	Len	0.12 (0.01, 1.03)	0.17 (0.01, 2.46)	0.57 (0.04, 7.15)	Len	0.55 (0.05, 6.77)	0.16 (0.02, 1.09)	0.32 (0.06, 1.90)	0.13 (0.01, 1.75)
0.73	Lin	0.21 (0.02, 1.86)	0.30 (0.02, 4.61)	1.02 (0.08, 12.87)	1.81 (0.15, 21.15)	Lin	0.29 (0.04, 1.95)	0.58 (0.10, 3.38)	0.24 (0.02, 3.16)
0.29	Pla	0.74 (0.20, 2.78)	1.05 (0.11, 9.76)	3.57 (0.50, 29.14)	6.32 (0.92, 47.13)	3.49 (0.51, 25.69)	Pla	2.02 (0.88, 5.08)	0.84 (0.11, 7.22)
0.58	Sor	0.37 (0.09, 1.32)	0.52 (0.07, 3.97)	1.77 (0.28, 11.07)	3.14 (0.53, 17.66)	1.73 (0.30, 9.84)	0.50 (0.20, 1.14)	Sor	0.42 (0.06, 2.76)
0.26	Sun	0.88 (0.09, 8.42)	1.25 (0.07, 19.77)	4.24 (0.32, 60.22)	7.52 (0.57, 95.11)	4.16 (0.32, 54.27)	1.19 (0.14, 8.86)	2.41 (0.36, 15.64)	Sun

The values in red shading were the highest three SUCRAs. The texts in yellow shading were targeted drugs.

The proportion of Grade 3–5 adverse events

Eleven RCTs [6, 25, 28, 34–36, 38–40, 43, 45] reporting information on G3-5AE were included for meta-analysis. Direct meta-analyses (S12 Fig) confirmed that brivanib (HR: 0.14; 95%CI: 0.10–0.21) and nintedanib (HR: 0.23; 95%CI: 0.10–0.52) than sorafenib. A connected network diagram formed by G3-5AE was provided in S13 Fig. According to the node-splitting analysis, there was not any significant inconsistency between the direct and indirect comparisons (Pla vs. Bri, P = 0.25; Sor vs. Bri, P = 0.25; Sor vs. Pla, P = 0.25), as shown in S14 Fig. The NMA heterogeneity was low (σ = 0.99; 95%CrI: 0.42–1.92), as shown in S2 Table. The NMA synthesis showed that there was no significant difference on G3-5AE among drugs. According to SUCRA, three highest ranking interventions were vandetanib (vandetanib 100 mg twice daily [0.89]; vandetanib 300 mg twice daily [0.82]) and nintedanib (0.67), which were in red in Table 6.

Table 6

Network meta-analyses for G3-5AE (Findings are expressed as OR (95% CrI), use of random-effect model).

SUCRA	Drugs	Bri	Dov	Erl+Sor	Eve+Sor	Lin	Nin	Pla	Sor	Van 100mg	Van 300mg
0.62	Bri	Bri	5.72 (0.28, 123.97)	5.35 (0.25, 115.35)	5.37 (0.26, 111.72)	7.44 (0.37, 154.93)	0.83 (0.06, 11.06)	0.60 (0.09, 3.66)	3.98 (0.62, 25.71)	0.19 (0.01, 4.27)	0.29 (0.01, 6.58)
0.25	Dov	0.17 (0.01, 3.57)	Dov	0.94 (0.03, 27.07)	0.93 (0.03, 26.50)	1.30 (0.04, 36.79)	0.14 (0.01, 2.98)	0.10 (0.01, 1.75)	0.69 (0.06, 7.67)	0.03 (0.00, 1.55)	0.05 (0.00, 2.29)
0.26	Erl+Sor	0.19 (0.01, 3.97)	1.07 (0.04, 32.27)	Erl+Sor	1.00 (0.03, 28.79)	1.39 (0.05, 38.78)	0.15 (0.01, 3.09)	0.11 (0.01, 1.91)	0.74 (0.07, 8.14)	0.04 (0.00, 1.63)	0.05 (0.00, 2.52)
0.26	Eve+Sor	0.19 (0.01, 3.77)	1.08 (0.04, 33.43)	1.00 (0.03, 29.28)	Eve+Sor	1.38 (0.05, 40.13)	0.15 (0.01, 3.10)	0.11 (0.01, 1.87)	0.74 (0.07, 7.98)	0.04 (0.00, 1.63)	0.05 (0.00, 2.48)
0.19	Lin	0.13 (0.01, 2.73)	0.77 (0.03, 23.24)	0.72 (0.03, 21.74)	0.73 (0.02, 20.36)	Lin	0.11 (0.01, 2.25)	0.08 (0.00, 1.31)	0.53 (0.05, 5.87)	0.03 (0.00, 1.13)	0.04 (0.00, 1.70)
0.67	Nin	1.21 (0.09, 16.40)	6.95 (0.34, 155.71)	6.50 (0.32, 131.89)	6.51 (0.32, 129.54)	8.94 (0.44, 183.46)	Nin	0.72 (0.06, 8.01)	4.82 (0.77, 31.28)	0.24 (0.01, 7.85)	0.35 (0.01, 12.15)
0.74	Pla	1.67 (0.27, 11.07)	9.62 (0.57, 190.57)	8.99 (0.52, 159.81)	8.98 (0.53, 157.59)	12.41 (0.76, 220.52)	1.39 (0.12, 15.75)	Pla	6.63 (1.45, 33.65)	0.32 (0.02, 4.32)	0.49 (0.03, 6.67)
0.31	Sor	0.25 (0.04, 1.61)	1.44 (0.13, 16.96)	1.35 (0.12, 14.62)	1.36 (0.13, 13.90)	1.87 (0.17, 20.16)	0.21 (0.03, 1.29)	0.15 (0.03, 0.69)	Sor	0.05 (0.00, 0.95)	0.07 (0.00, 1.44)
0.89	Van 100mg	5.16 (0.23, 122.61)	29.84 (0.64, 1511.71)	28.01 (0.61, 1342.11)	27.66 (0.61, 1326.10)	38.28 (0.89, 1848.26)	4.23 (0.13, 156.02)	3.08 (0.23, 40.53)	20.56 (1.05, 438.34)	Van 100mg	1.48 (0.10, 21.26)
0.82	Van 300mg	3.48 (0.15, 88.32)	19.97 (0.44, 1062.10)	18.75 (0.40, 923.34)	18.75 (0.40, 949.56)	25.76 (0.59, 1255.14)	2.88 (0.08, 107.45)	2.06 (0.15, 29.58)	13.79 (0.70, 308.28)	0.68 (0.05, 9.82)	Van 300mg

The values in red shading were three highest SUCRA. The texts in yellow shading were targeted drugs.

Cluster rank analysis

According to the meta-analysis performed above, ten interventions (Bri, Dov, Erl + Sor, Eve + Sor, Lin, Nin, Pla, Sor, Van 100mg and Van 300mg) compared to each other head-to-head on both OS and G3-5AE. According to cluster rank analysis, Van 100mg was the drug with both more effective (OS) and more secure (G3-5AE) compared to Sor followed by Nin (Fig 3).

Fig 3

Clustered ranking plot on OS and G3-5AE both expressed as SUCRAs.

The plot guides readers with respect to the trade-off between safety (G3-5AE) and effectiveness (OS) across the interventions. Interventions in the right upper corner tend to be more secure (higher SUCRA for G3-5AE) and more effective (higher SUCRA for OS) than those in the left lower corner of the plot.

Consistency, heterogeneity and quality of evidence

The detection of inconsistency in evidence networks was conducted by evaluating the agreement between the consistency and inconsistency (unrelated mean effects) models (S3 Table). The results of comparisons in both consistency and inconsistency models were roughly consistent. The result showed a robust and homogeneous network of evidence. Additionally, the node-splitting approach also showed a good consistency between the direct and indirect comparisons (S3, S8, S11 and S14 Figs). Though application of a fixed-effect model would provide similar numerical results with shorter credible intervals, random-effect model was more appropriate according to the residual deviance and DIC criteria (S2 Table). There was no obvious asymmetry at visual inspection of funnel plots to suggest publication bias as shown in S16 Fig. According to GRADE approach, the direct, indirect, and NMA Estimates for OS and G3-5AE were shown in S4 and S5 Tables. The quality of the most evidence was low.

Discussion

The SHARP trial was the first study to demonstrate efficacy (HR = 0.69; 95% CI 0.55–0.87, for sorafenib vs placebo, on OS) of targeted therapy for patients with unresectable HCC [6]. Subsequently, an Asia-Pacific study also confirmed the same conclusion (HR = 0.68, 95% CI 0.50–0.93) [45]. Based on the results of the two trials, sorafenib, a multi-targeted TKI, became the standard systemic treatment, approved by the regulatory authorities around the world, for patients with advanced unresectable HCC [46]. However, the advantages of survival and the improvements of symptom or living quality in these two trials were modest. In order to find more effective targeted drugs, several clinical trials ensued. Disappointingly, most of the results were negative.

Several targeted drugs were compared with sorafenib directly in this review [25–29, 34–38, 40, 42]. For TTP, only Len (HR = 0.63, 95% CI 0.54–0.74) and Lin (HR = 0.76, 95% CrI 0.64–0.91) performed better than sorafenib while others comparisons showed no statistical difference. For PFS, also Len (HR = 0.66, 95% CrI 0.56–0.77) and Lin (HR = 0.81, 95% CrI 0.69–0.95) performed better than sorafenib while others comparisons showed no statistical difference. For OS, no targeted drugs were superior to sorafenib while Sun performed worse than sorafenib with statistical difference. These direct comparison results are disappointing. Gratifyingly, a RCT verified that Van 100mg was superior in improving OS compared to placebo, although it didn’t indicated that Van 100mg was better than sorafenib.

To see the results of different targeted drugs comparing to each other, we performed this Bayesian network analysis. In this meta-analysis, brivanib, lenvatinib and linifanib were superior in improving TTP compared to placebo. However, they showed non-superiority in terms of both PFS and OS compared with placebo. Sorafenib was superior in improving both TTP and OS, while Van 100mg was also superior in improving OS. Although Tig 6mg + Sor, Van 300mg and Van 100mg were the three highest ranking interventions, they showed non-superiority in terms of OS compared with sorafenib. For ORR and G3-5AE, there was no significant difference across all targeted drugs. In general, sorafenib appeared to remain superior in the present analysis.

There are some potential reasons for failure to meet the primary endpoints of prolonging OS in HCC trials. First, the inclusion criteria of clinical trials are mainly based on Child-Pugh scores and BCLC stages. However, this screening method couldn’t eliminate the histologic heterogeneity in HCC. Therefore, several biomarkers (e.g., c-MET, RAS and FGF19) were recently used as bases for screening [47, 48]. Lack of predictive biomarkers was also one of the reasons for the failure. Second, by analyzing the target of included drugs, most of the drugs were anti-angiogenic multikinase inhibitors sharing some common pathways [49]. For these trials, there must be only marginal differences relative to sorafenib. To avoid similar targets, several trails tested a new drug in combination with sorafenib vs sorafenib alone, for instance, erlotinib targeting epidermal growth factor receptor, and everolimus targeting mammalian target of rapamycin. However, none of these combinations were superior in improving OS compared to sorafenib. Therefore, there still must be some other reasons for failure in HCC trials. Third, the end point OS is affected by advanced cirrhosis since advanced HCC is often accompanied by severe cirrhosis. The differences in curative effects among targeted drugs may not enough to cause major improvements in survival. To some extent, TTP may more suitable as an endpoint in advanced hepatocellular carcinoma treated with molecular targeted therapy [50]. Fourth, liver cirrhosis is frequently associated with hypohepatia. Due to the insufficiency of liver’s synthesis and metabolism function, the expected drug effect may not be exerted. Meanwhile, the side effects of drugs often lead to treatment interruption.

According to the cluster rank analysis, Van 100mg, Van 300mg and Nintedanib were more effective and more secure compared to Sorafenib, although the advantages were not statistically significant. Although vandetanib has limited clinical activity and was not warranted to be further developed as first-line therapy for advanced HCC [43], the correlational research of vandetanib in HCC had not stopped. Vandetanib-eluting radiopaque beads for locoregional treatment of HCC were under development [51]. Recent studies showed that nintedanib might have similar efficacy comparing to sorafenib in patients with advanced HCC, but with a manageable safety profile [25].

As we know, this is the first network meta-analysis of all RCTs to evaluate the efficacy and safety of targeted drugs for the treatment of HCC patients. Several limitations should be taken into consideration. First, the distributions of BCLC stages in different studies were not in full accord. Patients with B or C stage often had worse prognosis than those with A stage. The BCLC criteria for the patients could have an impact on the overall survival. Fortunately, the vast majority of patients include in this analysis were in stage B or C. Second, cirrhosis is also an important correlation factor in survival. Third, some HRs [26] were obtained by calculating the data extracted from the survival curves when they could not be acquired from the original article directly. Forth, both Response Evaluation Criteria in Solid Tumors (RECIST) v1.0, RECIST v1.1 and Modified RECIST (mRECIST) were used in the included studies. Both National Cancer Institute Common Terminology Criteria for Adverse Events, Version 3.0 and Version 4.0 were used in the included studies.

Our study also has several superiorities. First, we performed a comprehensive literature search to provide a summary of targeted therapies on HCC as detailed as possible. Second, in contrast to previous meta-analyses, the included studies were all RCTs that ensured the reliability of evidences. Third, we performed the cluster rank analysis considering both efficiency and safety in order to support clinical decision.

Conclusion

Taken together, our network meta-analysis suggests that vandetanib, linifanib, lenvatinib and nintedanib potentially may be the best substitution of sorafenib against advanced HCC. For OS, Van (100 and 300mg), seem to be the best options with low and moderate quality of evidence, respectively. For G3-5AE, Van (100 and 300mg), seem to be the best interventions, with low and very low quality of evidence all of them. Further studies are necessary to explore the curative effect of certain subgroup in HCC patients, especially the subgroup classified as BCLC stage, Child-Pugh score and Hepatitis B infection status. For better survival, novel targeted treatment options for HCC are sorely needed.

PRISMA 2009 flow diagram.

(DOCX)

Click here for additional data file.

Detailed search strategy.

(DOCX)

Click here for additional data file.

Targeted drug treatment programs.

(DOCX)

Click here for additional data file.

Risk of bias of included studies.

(DOCX)

Click here for additional data file.

Heterogeneity and model fit.

(DOCX)

Click here for additional data file.

Inconsistency analysis of treatment effects (random effects models—95% CrI).

(DOCX)

Click here for additional data file.

Direct, indirect, and NMA estimates for OS with the GRADE assessment.

(DOCX)

Click here for additional data file.

Direct, indirect, and NMA estimates for G3-5AE with the GRADE assessment.

(DOCX)

Click here for additional data file.

Forest plot (random effects) of direct meta-analyses for TTP.

(TIF)

Click here for additional data file.

Network diagram of studies for TTP.

(TIF)

Click here for additional data file.

Node-splitting test of studies for TTP.

(TIF)

Click here for additional data file.

Forest plot (random effects) of direct meta-analyses for PFS.

(TIF)

Click here for additional data file.

Network diagram of studies for PFS.

(TIF)

Click here for additional data file.

Forest plot (random effects) of direct meta-analyses for OS.

(TIF)

Click here for additional data file.

Network diagram of studies for OS.

(TIF)

Click here for additional data file.

Node-splitting test of studies for OS.

(TIF)

Click here for additional data file.

Forest plot (random effects) of direct meta-analyses for ORR.

(TIF)

Click here for additional data file.

Network diagram of studies for ORR.

(TIF)

Click here for additional data file.

Node-splitting test of studies for ORR.

(TIF)

Click here for additional data file.

Forest plot (random effects) of direct meta-analyses for G3-5AE.

(TIF)

Click here for additional data file.

Network diagram of studies for G3-5AE.

(TIF)

Click here for additional data file.

Node-splitting test of studies for G3-5AE.

(TIF)

Click here for additional data file.

(TIF)

Click here for additional data file.

Comparison-adjusted funnel plots for all comparisons.

(TIF)

Click here for additional data file.

26 Sep 2019

PONE-D-19-22935

First-line targeted therapies of advanced hepatocellular carcinoma: A Bayesian network analysis of randomized controlled trials

PLOS ONE

Dear Mr Xu,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

We would appreciate receiving your revised manuscript by Nov 10 2019 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Ivan D. Florez

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

http://www.journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and http://www.journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

1. Thank you for including your funding statement; "The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript."

Please provide an amended Funding Statement that declares *all* the funding or sources of support received during this specific study (whether external or internal to your organization) as detailed online in our guide for authors at http://journals.plos.org/plosone/s/submit-now.

Please state what role the funders took in the study.  If any authors received a salary from any of your funders, please state which authors and which funder. If the funders had no role, please state: "The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript."

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

Additional Editor Comments (if provided):

Dear Authors,

Your manuscript has been evaluated by two reviewers who have raised a number of points that need to be addressed before considering it for publication. Please provide a response to each one fo the points described below. In particular, make emphasis on the following points which may be the most problematic, and are esssential for any NMA:

- Protocol: Reviewer 1 has raised a crucial point regarding the protocol for this nma. One criterion for any systematic review to be of high quality is the presence of a protocol before starting the review. You haven't provided evidence that that protocol existed. At least a register in a database for Systematic reviews should have existed (e.eg., PROSPERO database). Please, either: provide evidence that the protocol/registration existed and was available before stariting the project or explicitly state in the mauscript that the review was not registered in PROSPERO database and did not have an online protocll available, and provide a brief explanation.

- Inconsistency. Reviewer #1 has raised the point on the inconsistency tests. You should explicitly provide the description of the global and the local tests performed to define the inconsistency (per loop), and provide the results of all the tests: the global and each one of the local tests perfomed for each outcome, and for each loop (fo the local). Depending on the tests used: P values, or ROR, should be shown, in the manuscript or in appendices

- Intransitivity: Reviewer #1 has pointed out that you made conclusions on transitivity, that are not supported in the manuscript, To conclude that you need to be clear about how this assumption was evaluated, considering the variables you describe that were used.

- Certainity of the evidence:Reviewer #1 has recommended to apply GRADE to determine what would be the certainity of the evidence (Also called quality of the evidence) with GRADE. We support that recommendation.

- Heterogeneity: Reviewer #1 pointed out as well that you didn't mention anything on the heterogeneity assessment. Heterogeneity should be evaluated and stataisitically assessed: Both: for the whole netwrok, and for pairwise comparisons. Also, in cases of high heterogeneity, special analyses such as Meta-regression and subgroup analyses shpuld have been performed.

- Reviewer #2 has described how your conclusions do not 100% reflect the results.

We expect you could provide appropriate answers to the concerns I have highlighted above along with the reviewers comments which are key to consider your manuscript.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Review for First-line targeted therapies of advanced hepatocellular carcinoma: A Bayesian

network analysis of randomized controlled trials

The authors evaluated first-line targeted therapies of advanced hepatocellular carcinoma via a network meta-analysis. This is the first NMA adding important information in the field of advanced hepatocellular carcinoma. However, I have concerns due to the lack of NMA methodological aspects and I think that authors should conduct a lot of things aiming to provide a comprehensive NMA analysis.

My comments focus as follows:

Abstract

I suggest to provide more information in methods section.

Please revise this sentence. “Direct and indirect evidence were 29 combined to time to progress (TTP), overall survival (OS), progress-free 30 survival (PFS), objective response rate (ORR), and the proportion of 31 Grade 3-5 adverse events (G3-5AE) and surface under the cumulative 32 ranking curve (SUCRA) of patients with advanced HCC.”

Provide the measure of outcomes and SUCRA values in a more informative and clear way in the sentence.

Please provide also that you implement the network meta-analysis model and the effect size used (OR and HR).

Materials and methods

Please refer about your NMA protocol. Have you a published protocol of the analysis?

Data synthesis and analysis

I advise authors to start this section reporting that they performed the NMA model and the measures used for treatment effects estimates.

Authors need to provide a global test for the assessment of inconsistency, such as the random-effects design-by-treatment interaction model and the local test (node-splitting). Then, authors could report in results section that it not applicable as they have star NMAs.

Missing details for Bayesian NMA. Authors report about posterior distributions but no further details for assumptions, for example prior distribution for model parameters. What priors for model parameters distributions? What prior distribution for heterogeneity? What assumption about heterogeneity, did they used a common between-study standard deviation across all treatment comparisons in each network? What about the assessment of transitivity? Authors need to compute measures for heterogeneity, for example I squared. Sensitivity analyses need to be conducted when it is needed. Rating the quality of the evidence in the estimates using GRADE criteria can also be conducted. Pairwise meta-analyses can also be conducted and I also suggest to conduct the comparison adjusted funnel plot.

Study characteristics

“There was no evidence that the transitivity assumption was violated in 177 any of the networks.” How authors conclude to this? Authors should refer about transitivity in later section of the analysis.

Results

I recommend to provide the results section taking into account all the methodological suggestions are given in Methods.

Reviewer #2: Authors perfored a network meta-analysis of the available RCTs comparing the efficacy of the various targeted therapes used in the management of HCC. This appeared to be a well conducted study with appropriately evaluted outcomes. I have a few comments:

1) Does the BCLC criteria of the included patients inflence the outcomes especially the overall survival? Is there a scope to perform sub-group analysis matching the patient and tumour charateristics in the current or future studies?

2) Are the conclusions reflecting the results appropriately? Soraferenib appears to remain superior from your analysis. THis should be one of the conclusions.

3)in the eligibility criteria: 'The key inclusion

85 criteria for study populations: more than 4 weeks since most recent local

therapy or no local therapy; no prior systemic therapy.' what does this mean?

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

18 Nov 2019

Replies to Reviewers

First of all, we thank both reviewers and editors for your positive and constructive comments and suggestions.

Replies to Reviewer 1:

1. QUESTION: “Abstract

I suggest to provide more information in methods section.

Please revise this sentence. “Direct and indirect evidence were combined to time to progress (TTP), overall survival (OS), progress-free survival (PFS), objective response rate (ORR), and the proportion of Grade 3-5 adverse events (G3-5AE) and surface under the cumulative ranking curve (SUCRA) of patients with advanced HCC.”

Provide the measure of outcomes and SUCRA values in a more informative and clear way in the sentence.

Please provide also that you implement the network meta-analysis model and the effect size used (OR and HR).”

Answer: We have revised the sentence. It has been replaced with “Time to progress (TTP), overall survival (OS), progress-free survival (PFS), were calculated as hazard ratios (HRs) and 95% credible intervals (CIs). Objective response rate (ORR) and the proportion of Grade 3-5 adverse events (G3-5AE) were expressed as odds ratios (ORs) and 95% CIs. We pooled study-specific HRs and ORs using fixed-effects network meta-analyses, and ranked first-line drugs by the surface under the cumulative ranking curve (SUCRA).”

2. QUESTION: “Materials and methods

Please refer about your NMA protocol. Have you a published protocol of the analysis?”

Answer: We were registered our NMA protocol on July 29, 2019. But the records are being assessed. So, we didn’t have a published protocol.

3. QUESTION: “Data synthesis and analysis

I advise authors to start this section reporting that they performed the NMA model and the measures used for treatment effects estimates.

Authors need to provide a global test for the assessment of inconsistency, such as the random-effects design-by-treatment interaction model and the local test (node-splitting). Then, authors could report in results section that it not applicable as they have star NMAs.

Missing details for Bayesian NMA. Authors report about posterior distributions but no further details for assumptions, for example prior distribution for model parameters. What priors for model parameters distributions? What prior distribution for heterogeneity? What assumption about heterogeneity, did they used a common between-study standard deviation across all treatment comparisons in each network? What about the assessment of transitivity? Authors need to compute measures for heterogeneity, for example I squared. Sensitivity analyses need to be conducted when it is needed. Rating the quality of the evidence in the estimates using GRADE criteria can also be conducted. Pairwise meta-analyses can also be conducted and I also suggest to conduct the comparison adjusted funnel plot.”

Answer: We rewrote this paragraph. We reused R language to perform statistical calculations. We had reported the NMA model and the measures used for treatment effects estimates. We had added the detailed description of Bayesian NMA. We dad provided consistency checks for the assessment of inconsistency. We had added heterogeneity test for both pairing comparison and NMA. The transitivity was assessed by examining the patient baseline characteristics across studies (age, gender, performance status and tumor stage), treatment stage and treatment protocol. The GRADE evaluation system and funnel plots were also added into this NMA.

4. QUESTION: “Study characteristics

‘There was no evidence that the transitivity assumption was violated in any of the networks.’ How authors conclude to this? Authors should refer about transitivity in later section of the analysis.”

Answer: We had added the description of transitivity in later section of the analysis. “To confirm the transitivity and the loop-specific consistency assumption, pairwise direct and indirect effect estimates of closed loops of evidence were inspected for any disagreement. The transitivity was assessed by examining the patient baseline characteristics across studies (age, gender, performance status and tumor stage), treatment stage and treatment protocol.”

5. QUESTION: “Results

I recommend to provide the results section taking into account all the methodological suggestions are given in Methods.”

Answer: New results were provided according to the new calculation.

Replies to Reviewer 2:

1. QUESTION: “Does the BCLC criteria of the included patients inflence the outcomes especially the overall survival? Is there a scope to perform sub-group analysis matching the patient and tumour charateristics in the current or future studies?”

Answer: The BCLC criteria of the included patients dose influence the outcomes especially the overall survival. However, the included studies didn’t group by the BCLC criteria. So, we could not perform sub-group analysis. Fortunately, the tumour charateristics of included patients were relatively consistent.

2. QUESTION: “Are the conclusions reflecting the results appropriately? Soraferenib appears to remain superior from your analysis. THis should be one of the conclusions.”

Answer: The conclusion had been supplemented by the affirmation of sorafenib. “For the moment, sorafenib was still as a first-line drug of first choice.”

3. QUESTION: “in the eligibility criteria: 'The key inclusion criteria for study populations: more than 4 weeks since most recent local therapy or no local therapy; no prior systemic therapy.' what does this mean?”

Answer: It means that it should last more than 4 weeks for the included study populations since most recent local therapy or no local therapy. It was to avoid the influence of other treatments.

We appreciate for editors/reviewers’ warm work earnestly, and hope that the correction will meet with approval.

Thank you and best regards.

Yours sincerely,

Xuezhong Xu

E-mail: xxzdoctor@163.com.

Submitted filename: R2R.docx

Click here for additional data file.

17 Dec 2019

PONE-D-19-22935R1

First-line targeted therapies of advanced hepatocellular carcinoma: A Bayesian network analysis of randomized controlled trials

PLOS ONE

Dear Mr Xu,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

We would appreciate receiving your revised manuscript by Jan 31 2020 11:59PM. When you are ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter.

To enhance the reproducibility of your results, we recommend that if applicable you deposit your laboratory protocols in protocols.io, where a protocol can be assigned its own identifier (DOI) such that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). This letter should be uploaded as separate file and labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. This file should be uploaded as separate file and labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. This file should be uploaded as separate file and labeled 'Manuscript'.

Please note while forming your response, if your article is accepted, you may have the opportunity to make the peer review history publicly available. The record will include editor decision letters (with reviews) and your responses to reviewer comments. If eligible, we will contact you to opt in or out.

We look forward to receiving your revised manuscript.

Kind regards,

Ivan D. Florez

Academic Editor

PLOS ONE

Additional Editor Comments (if provided):

Thanks for submitting a revised version. Your revised manuscript has incorporated many of reviewers comments. However, in addition to reviewer #1 comments (which are below) there are a couple of points that hasn’t been completely addressed from reviewer #2, and additional issues identified by myself as Academic Editor:

Reviewer #2 points to be addressed:

1. Reviewer #2 pointed out how the BCLC criteria for the patients could have had an impact on the overall survival. You have pointed out that a subgroup analyses is not possible since the authors didn’t group patients by BCLC criteria. However, there are alternatives to this, considering that you agree that this stage could have an impact on the outcome, and therefore it can be a potential effect modifier.

- If authors provided a proportion of patients with a specific BCLC, you could group studies according to specific proportions

- If a subgroup analyses is not feasible at this point you should address this as a limitation in a paragraph at the end of the discussion section of your manuscript, and discuss there how were the tumor characteristics of the patients and how this may impact on the results.

2. Regarding reviewer#2 comment about the conclusions, you provide the next sentence: “For the moment, sorafenib was still as a first-line drug of first choice.” We think that according to the results, Srafenib should not be highlighted as the best approach. I think In conclusions you should emphasize in those interventions that were superior, BUT adding the quality of evidence for those. See the following sentence as an example:

For OS, Van (100 and 300mg), seem to be the best options with Low and moderate quality of evidence, respectively. For G3 5AE, Van 100 and 300mg), and play seem to be the best interventions, with low quality of evidence all of them (Just as an example, you need to apply GRADE methodology as I suggest below, and find the final quality assessment).

3. Also, regarding this comment from reviewer #2:

In the eligibility criteria: 'The key inclusion criteria for study populations: more than 4 weeks since most recent local therapy or no local therapy; no prior systemic therapy.' what does this mean?”
You have provided a response, However, it is not reflected in the paper. This part should be clear enough for reader so they could have the same question as Reviewer #2. Thus, please detail in this section, what does that mean and the reasons for that decision.

Major and minor Comments from Editor:

MAJOR:

You have explained your GRADE approach. However, major issues in your GRADE approach have been identified:

- You state that you used the methods recommended by: Puhan et al (2014). However, using only this method is not appropriate as it was the first approach and it has had at least 2 updates: Brignardello-Petersen et al. J Clin Epidemiol 2018, Brignardello-Petersen et al. J Clin Epidemiol 2019, Consider applying the full current approach to assess the quality of evidence.

- According to GRADE approach (Puhan’s and Brignardello-Petersen’s articles) you need to assess the quality of the direct, the indirect and the network evidence for each comparison and each outcome. Although you may present only your Network estimates and your network quality assessments in your main manuscript. However, you need to provide evidence that you conducted all the process appropriately and for that you need to provide all the direct, indirect and NMA estimates + quality of evidence. We cannot identify in your appendices these estimates (Except by some of them that are provided in the Forest plots, and without each GRADE assessment) for readers that would be interested in them.

- Also in page 35 you state:

“Therefore, in the present NMA, the GRADE score was downgraded universally due to the diversity of patient characteristics and the disunity of evaluation standard”. This sentence is very problematic, because this is not what GRADE guidance recommends. GRADE working group never recommends to downgrade comparisons “globally”. This shows a misunderstanding of the approach. If you apply GRADE, you should follow all the recommended steps by Puhan’s and Brignardello-Petersen’s articles which are the official GRADE working group Guidance.

- in the same page you state:

“Second, it was further downgraded in certain comparisons for no direct evidence”. That may be true. But again, the only way to identify what comparison were downgraded and for what reasons (GRADE Criteria that were use to downgrade in each direct and indirect and NMA estimates)), is to provide in appendices a full list of all the available estimates (D, I, and NMA) for all the outcomes, along with the corresponding GRADE assessment, following the GRADE guidance, and indicating the reasons for downloading each one of them, according to the 7 criteria recommended by the GRADE approach.

- Also, in the same page, you state: “In addition, it was further downgraded in some relevantcomparisons because of their IF below 50%”. Could you explain how this criterion match with the GRADE approach recommended by Puhan’s and Brignardello-Petersen 2018 and 2019, One more time, there should be evidence of what comparisons were downgraded by any particular criterion

- Finally, as you state that you calculated I2. There is no information in the manuscript nor the appendices that shows the i2 for all the direct comparisons. You should provide this data as you described you calculated them to assess the heterogeneity of direct comparisons

MINOR:

Additional minor changes to consider are:

- You excluded one study and the reason provided is that it was data deficient. please clearly explain what does that mean.

- Page 14, line 80, Please write PRISMA in Capital letters.

- Page 14, line 80. There is a specific PRISMA Guidance for NMA, it seems you followed the PRISMA for regular Sr and MA. You need to follow Guidance that is specific for NMA (PRISMA NMA). You also need to provide the appropriate reference for the PRISMA-NMA Guidance.

- Page 14, line 82. Since you cannot provide a PROSPERO registration number at this time, lease indicate the date you submitted the register to the PROSPERO database. Also indicate what was the state of this PROSPERO submission by the time you submitted this manuscript to PlosOne

- FIGURE 4 is of very low quality, you need to provide a better image

- Page 35: They “IF” acronym has not ben explained before and has not been detailed in the methods as a key approach

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: I am happy that the authors provided a registered protocol in PROSPERO database. I recommend authors to revise the sentence “It did not have an online protocol…..under review” with “The protocol has been registered in PROSPERO database and it is now under review”

The authors addressed most of the concerns and suggestions about the NMA methodology and the manuscript has been improved. Although, I have some comments and concerns for inconsistency checking.

Based on review comments, authors should have provided a global test for inconsistency assumption, eg. Design-by-treatment interaction model (DBT) and if inconsistency was detected they should have provided a node-splitting approach.

I don’t agree with the interpretation provided for the results of the Node-splitting method. Node-splitting approach checks if direct and indirect is in agreement (consistency). Authors concluded to “The result showed a robust and homogeneous network of evidence”. This should be revised. The results of all tests (P-values) for each outcome could be reported in the manuscript with S7, S13, S16 and S19 figures.

All the forest plots provided for NMA should be renamed in figures and tables with Comparison-adjusted funnel plot. Moreover, dashed lines in plots are missing and should be provided (Figure S20).

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files to be viewed.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email us at figures@plos.org. Please note that Supporting Information files do not need this step.

14 Jan 2020

Dear Editors and Reviewers:

Thank you very much for your comments and suggestions.

We have revised the manuscript, according to the comments and suggestions of reviewers and editor, and responded, point by point to, the comments as listed below.

I would like to re-submit this revised manuscript to PLOS ONE, and hope it is acceptable for publication in the journal.

Looking forward to hearing from you soon.

With kindest regards,

Yours Sincerely

Xuezhong Xu.

Replies to Editors

1. QUESTION: Reviewer #2 pointed out how the BCLC criteria for the patients could have had an impact on the overall survival. You have pointed out that a subgroup analyses is not possible since the authors didn’t group patients by BCLC criteria. However, there are alternatives to this, considering that you agree that this stage could have an impact on the outcome, and therefore it can be a potential effect modifier.

- If authors provided a proportion of patients with a specific BCLC, you could group studies according to specific proportions.

- If a subgroup analyses is not feasible at this point you should address this as a limitation in a paragraph at the end of the discussion section of your manuscript, and discuss there how were the tumor characteristics of the patients and how this may impact on the results.

ANSWER: We thank very much for the editor's advice. However, the proportions of patients with different BCLC stages were not uniform. So, the subgroup analysis is not feasible. So, we added some discussion at the end of the discussion section. “The distributions of BCLC stages in different studies were not in full accord. Patients with B or C stage often had worse prognosis than those with A stage. The BCLC criteria for the patients could have an impact on the overall survival. Fortunately, the vast majority of patients include in this analysis were in stage B or C.”

2. QUESTION: Regarding reviewer#2 comment about the conclusions, you provide the next sentence: “For the moment, sorafenib was still as a first-line drug of first choice.” We think that according to the results, Srafenib should not be highlighted as the best approach. I think In conclusions you should emphasize in those interventions that were superior, BUT adding the quality of evidence for those. See the following sentence as an example: For OS, Van (100 and 300mg), seem to be the best options with Low and moderate quality of evidence, respectively. For G3 5AE, Van 100 and 300mg), and play seem to be the best interventions, with low quality of evidence all of them (Just as an example, you need to apply GRADE methodology as I suggest below, and find the final quality assessment).

ANSWER: We fully endorse the editor's suggestion. We had revised the sentence with “For OS, Van (100 and 300mg), seem to be the best options with low and moderate quality of evidence, respectively. For G3-5AE, Van (100 and 300mg), seem to be the best interventions, with low and very low quality of evidence all of them.”

3. QUESTION: Also, regarding this comment from reviewer #2: In the eligibility criteria: 'The key inclusion criteria for study populations: more than 4 weeks since most recent local therapy or no local therapy; no prior systemic therapy.' what does this mean?” You have provided a response, However, it is not reflected in the paper. This part should be clear enough for reader so they could have the same question as Reviewer #2. Thus, please detail in this section, what does that mean and the reasons for that decision.

ANSWER: We had revised the sentence with “To avoid the influence of other treatments, the key inclusion criteria for included study populations were as follows: First, it should last more than 4 weeks since most recent local therapy or no local therapy. Second, the patients did not receive prior systemic therapy.”

4. QUESTION: - You state that you used the methods recommended by: Puhan et al (2014). However, using only this method is not appropriate as it was the first approach and it has had at least 2 updates: Brignardello-Petersen et al. J Clin Epidemiol 2018, Brignardello-Petersen et al. J Clin Epidemiol 2019, Consider applying the full current approach to assess the quality of evidence.

- According to GRADE approach (Puhan’s and Brignardello-Petersen’s articles) you need to assess the quality of the direct, the indirect and the network evidence for each comparison and each outcome. Although you may present only your Network estimates and your network quality assessments in your main manuscript. However, you need to provide evidence that you conducted all the process appropriately and for that you need to provide all the direct, indirect and NMA estimates + quality of evidence. We cannot identify in your appendices these estimates (Except by some of them that are provided in the Forest plots, and without each GRADE assessment) for readers that would be interested in them.

- Also in page 35 you state:“Therefore, in the present NMA, the GRADE score was downgraded universally due to the diversity of patient characteristics and the disunity of evaluation standard”. This sentence is very problematic, because this is not what GRADE guidance recommends. GRADE working group never recommends to downgrade comparisons “globally”. This shows a misunderstanding of the approach. If you apply GRADE, you should follow all the recommended steps by Puhan’s and Brignardello-Petersen’s articles which are the official GRADE working group Guidance.

- in the same page you state:“Second, it was further downgraded in certain comparisons for no direct evidence”. That may be true. But again, the only way to identify what comparison were downgraded and for what reasons (GRADE Criteria that were use to downgrade in each direct and indirect and NMA estimates)), is to provide in appendices a full list of all the available estimates (D, I, and NMA) for all the outcomes, along with the corresponding GRADE assessment, following the GRADE guidance, and indicating the reasons for downloading each one of them, according to the 7 criteria recommended by the GRADE approach.

- Also, in the same page, you state: “In addition, it was further downgraded in some relevant comparisons because of their IF below 50%”. Could you explain how this criterion match with the GRADE approach recommended by Puhan’s and Brignardello-Petersen 2018 and 2019, One more time, there should be evidence of what comparisons were downgraded by any particular criterion.

- Finally, as you state that you calculated I2. There is no information in the manuscript nor the appendices that shows the i2 for all the direct comparisons. You should provide this data as you described you calculated them to assess the heterogeneity of direct comparisons

ANSWER: Thanks very much to the editors for pointing out my shortcomings. We looked through the literature “Brignardello-Petersen R, Bonner A, Alexander PE, Siemieniuk RA, Furukawa TA, Rochwerg B, et al. Advances in the GRADE approach to rate the certainty in estimates from a network meta-analysis. J Clin Epidemiol 2018; 93:36–44. https://doi.org/10.1016/j.jclinepi.2017.10.005 PMID: 29051107”. We have relearned the GRADE approach and rewritten this paragraph. The method was described as follows “For direct comparison, we graded evidence from the five aspects; risk of bias, inconsistency, indirectness, imprecision and publication bias, using the standard GRADE approach. For indirect comparison, we rated evidence according to the lower grades of direct comparisons and intransitivity. For NMA estimates, we rated evidence according to the higher grades of the direct and indirect comparisons and incoherence.” The direct, indirect, and NMA Estimates for OS with the GRADE Assessment were shown in S22-23 Tables. The I2 for all the direct comparisons were shown in S5 Figure, S9 Figure, S11 Figure, S14 Figure and S17 Figure, and they were also shown in S22-23 Tables. We have deleted the “IF” criterion.

5. QUESTION: - You excluded one study and the reason provided is that it was data deficient. please clearly explain what does that mean.

- Page 14, line 80, Please write PRISMA in Capital letters.

- Page 14, line 80. There is a specific PRISMA Guidance for NMA, it seems you followed the PRISMA for regular Sr and MA. You need to follow Guidance that is specific for NMA (PRISMA NMA). You also need to provide the appropriate reference for the PRISMA-NMA Guidance.

- Page 14, line 82. Since you cannot provide a PROSPERO registration number at this time, lease indicate the date you submitted the register to the PROSPERO database. Also indicate what was the state of this PROSPERO submission by the time you submitted this manuscript to PlosOne

- FIGURE 4 is of very low quality, you need to provide a better image

- Page 35: They “IF” acronym has not been explained before and has not been detailed in the methods as a key approach

ANSWER: - The excluded study was lack of control group.

- Page 14, line 80, we have revised PRISMA in Capital letters. We have rewritten the list following the guidance of PRISMA-NMA checklist.

- Page 14, line 82. This network meta-analysis has been registered in the PROSPERO public database (CRD42019145188)

- FIGURE 4 had been deleted and been replaced by S22-23 Tables.

- Page 35: The “IF” acronym means “information fraction”. It was used to assess statistical power and strength of evidence for each treatment comparison. However, we decide to remove this assessment because the GRADE approach was enough.

Replies to Reviewers

First of all, we thank both reviewers and editors for your positive and constructive comments and suggestions.

Replies to Reviewer 1:

1. QUESTION: Based on review comments, authors should have provided a global test for inconsistency assumption, eg. Design-by-treatment interaction model (DBT) and if inconsistency was detected they should have provided a node-splitting approach.

Answer: Thanks very much for the advice. We did lack a check for global consistency. However, design-by-treatment interaction model (DBT) was a frequentist NMA model that considers both heterogeneity between studies and inconsistency between study designs according to White IR (White IR. Network meta-analysis. Stata J 2015; 15: 951-985.). Since what I used was bayesian model, design-by-treatment interaction model (DBT) was not suitable for me. In Konstantinos’s study (Konstantinos K, Panagiotis K, Stavros S, et al. Comparative effectiveness of different transarterial embolization therapies alone or in combination with local ablative or adjuvant systemic treatments for unresectable hepatocellular carcinoma: A network meta-analysis of randomized controlled trials[J]. PLOS ONE, 2017, 12(9): e0184597), the author used unrelated mean effects model to evaluate the inconsistency. So, we also used this method to estimate the inconsistency. And the modification has been added in the revised manuscript. The results of comparisons in both consistency and inconsistency models were roughly consistent. The results were shown in S20 Table.

2. QUESTION: I don’t agree with the interpretation provided for the results of the Node-splitting method. Node-splitting approach checks if direct and indirect is in agreement (consistency). Authors concluded to “The result showed a robust and homogeneous network of evidence”. This should be revised. The results of all tests (P-values) for each outcome could be reported in the manuscript with S7, S13, S16 and S19 figures.

Answer: The sentence has been revised with “The node-splitting approach also showed a good consistency between the direct and indirect comparisons”. The results of all tests (P-values) for each outcome has been added in the manuscript.

3. QUESTION: All the forest plots provided for NMA should be renamed in figures and tables with Comparison-adjusted funnel plot. Moreover, dashed lines in plots are missing and should be provided (Figure S20).

Answer: In S21 Figure, the titles have been revised as Comparison-adjusted funnel plot. And we have added the dashed lines.

We appreciate for editors/reviewers’ warm work earnestly, and hope that the correction will meet with approval.

Thank you and best regards.

Yours sincerely,

Xuezhong Xu

E-mail: xxzdoctor@163.com.

Submitted filename: Response to Reviewers.docx

Click here for additional data file.

10 Feb 2020

First-line targeted therapies of advanced hepatocellular carcinoma: A Bayesian network analysis of randomized controlled trials

PONE-D-19-22935R2

Dear Dr. Xu,

We are pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it complies with all outstanding technical requirements.

Within one week, you will receive an e-mail containing information on the amendments required prior to publication. When all required modifications have been addressed, you will receive a formal acceptance letter and your manuscript will proceed to our production department and be scheduled for publication.

Shortly after the formal acceptance letter is sent, an invoice for payment will follow. To ensure an efficient production and billing process, please log into Editorial Manager at https://www.editorialmanager.com/pone/, click the "Update My Information" link at the top of the page, and update your user information. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, you must inform our press team as soon as possible and no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

With kind regards,

Peter Starkel, M.D., Ph.D.

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

All comments have been addressed. No further remarks.

Reviewers' comments:

12 Feb 2020

PONE-D-19-22935R2

First-line targ veted therapies of advanced hepatocellular carcinoma: A Bayesian network analysis of randomized controlled trials

Dear Dr. Xu:

I am pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please notify them about your upcoming paper at this point, to enable them to help maximize its impact. If they will be preparing press materials for this manuscript, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

For any other questions or concerns, please email plosone@plos.org.

Thank you for submitting your work to PLOS ONE.

With kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr Peter Starkel

Academic Editor

PLOS ONE