Selection of first-line systemic therapies for advanced hepatocellular carcinoma: A network meta-analysis of randomized controlled trials.

BACKGROUND: The majority of clinical trials of first-line systemic treatments for hepatocellular carcinoma (HCC) used placebo or sorafenib as comparators, and there are limited data providing a cross comparison of treatments in this setting, especially for newly-approved immune checkpoint inhibitor and vascular endothelial growth factor inhibitor combination treatments. AIM: To systematically review and compare response rates, survival outcomes, and safety of first-line systemic therapies for advanced hepatocellular carcinoma.
METHODS: We searched PubMed, Science Direct, the Cochrane Database, Excerpta Medica Database, and abstracts from the American Society of Clinical Oncology 2020 annual congress. Eligible studies were randomized controlled trials of systemic therapy enrolling adults with advanced/unresectable HCC. Risk of bias was assessed with the Cochrane risk of bias tool for randomized controlled trials. A network meta-analysis was used to synthesize data and perform direct and indirect comparisons between treatments. P value, a frequentist analog to the surface under the cumulative ranking curve, was used to rank treatments.
RESULTS: In total, 1398 articles were screened and 27 included. Treatments compared were atezolizumab plus bevacizumab, brivanib, donafenib, dovitinib, FOLFOX4, lenvatinib, linifanib, nintedanib, nivolumab, sorafenib, sunitinib, vandetanib, 11 sorafenib combination therapies, and three other combination therapies. For overall response rate, lenvatinib ranked 1/19, followed by atezolizumab plus bevacizumab and nivolumab. For progression-free survival (PFS), atezolizumab + bevacizumab was ranked 1/15, followed by lenvatinib. With the exception of atezolizumab + bevacizumab [hazard ratios (HR)PFS = 0.90; 95% confidence interval (CI): 0.64-1.25], the estimated HRs for PFS for all included treatments vs lenvatinib were > 1; however, the associated 95%CI passed through unity for bevacizumab plus erlotinib, linifanib, and FOLFOX4. For overall survival, atezolizumab plus bevacizumab was ranked 1/25, followed by vandetanib 100 mg/d and donafinib, with lenvatinib ranked 6/25. Atezolizumab + bevacizumab was associated with a lower risk of death vs lenvatinib (HRos = 0.63; 95%CI: 0.44-0.89), while the HR for overall survival for most other treatments vs lenvatinib had associated 95%CIs that passed through unity. Vandetanib 300 mg/d and 100 mg/d were ranked 1/13 and 2/13, respectively, for the lowest incidence of treatment terminations due to adverse events, followed by sorafenib (5/13), lenvatinib (10/13), and atezolizumab + bevacizumab (13/13).
CONCLUSION: There is not one single first-line treatment for advanced HCC associated with superior outcomes across all outcome measurements. Therefore, first-line systemic treatment should be selected based on individualized treatment goals. ©The Author(s) 2021. Published by Baishideng Publishing Group Inc. All rights reserved.

Core Tip: The present network meta-analysis is the first to compare data from randomized trials of all first-line systemic therapies for hepatocellular carcinoma including chemotherapy, targeted drugs, immunotherapy, and combination therapies. Furthermore, the analysis represents a comprehensive cross comparison of outcomes, including tumor response rates, survival, and safety and included a sub-analysis in patients with hepatitis B virus infection. Our results showed that atezolizumab plus bevacizumab was ranked first for progression-free survival and overall survival but also had the highest rate of discontinuations due to adverse events. Lenvatinib ranked first for overall response rate and second for progression-free survival.

INTRODUCTION

Liver cancer is the sixth most common cancer globally, accounting for 4.7% of all new cancer cases in 2018, and represents the third most common cause of cancer-related death worldwide behind lung and colorectal cancer[1]. Of the primary liver cancers, hepatocellular carcinoma (HCC) is the most prevalent histological subtype and accounts for 80%-85% of cases[2]. Surgical resection and liver transplant are associated with the best survival outcomes for patients with HCC, and are potentially curative treatments[3]. Locoregional therapies including arterially directed therapies, ablation, and radiotherapy are also associated with good survival outcomes in patients with unresectable disease confined to the liver[4]. However, over 50% of patients with HCC are diagnosed at an advanced stage or with other characteristics that preclude surgical or locoregional treatment[5]. For these patients, systemic therapy is usually the recommended treatment option[4,6].

Over the past 3 years, the number of approved first-line systemic therapies for patients with HCC has expanded greatly, and numerous drugs and drug combinations have been evaluated in this setting[7]. Between 2007 and 2018, sorafenib was the only approved systemic treatment for HCC based on the results of the Phase III SHARP trial, which showed a survival benefit for sorafenib vs placebo[8]. In the decade following the approval of sorafenib, numerous unsuccessful trials of systemic therapies in advanced HCC were conducted until the approval of lenvatinib in 2018[9]. Lenvatinib was approved for first-line use in advanced HCC following the successful outcome of the Phase III REFLECT trial. In this trial, lenvatinib showed a non-inferior overall survival (OS) vs sorafenib for the treatment of advanced HCC[10]. Since the approval of lenvatinib, the immunotherapy drugs nivolumab and pembrolizumab, as well as other tyrosine kinase inhibitors (TKI), have been approved for the second-line treatment of HCC. Most recently, combination therapy with the anti-PD-L1 agent atezolizumab plus bevacizumab demonstrated better OS and progression-free survival (PFS) than sorafenib in the Phase III IMbrave 150 trial[11].

The expansion of first-line treatment options for advanced HCC represents a significant advance in the treatment of this disease. However, further data would be useful to inform treatment selection. Most clinical trials of first-line therapies for HCC used placebo or sorafenib as comparators and there are limited data providing a cross comparison of the efficacy and safety of drugs in this setting. Furthermore, although lenvatinib is widely seen as a standard of care in real clinical practice and is a recommended first-line therapy in most international treatment guidelines[4,12,13], there are limited head-to-head data comparing lenvatinib with other systemic therapies. Finally, although historically systemic treatments for HCC were associated with low tumor response rates, recently approved therapies have been associated with response rates > 30%[14]. This has led to renewed interest in tumor response rates in HCC, and investigation of downstaging and conversion therapy strategies. A comparison of response rates for all currently available therapies would therefore be of clinical value.

This network meta-analysis was conducted to systematically review and compare the response rates, survival outcomes, and safety reported by randomized trials of first-line systemic therapies in patients with advanced unresectable HCC, and to provide a comparison between lenvatinib and other systemic therapies in this setting. Two recent meta-analyses have investigated a similar topic to the present study; however, one did not include data on atezolizumab plus bevacizumab and excluded non-targeted therapies[15], and a more recent analysis focused on treatment sequencing by investigating survival outcomes only[16]. Therefore, although there is some overlap with the present analysis, these studies are complementary to each other. In particular, the present analysis is the first to include data on donafenib, a Chinese drug that has shown a superior OS to sorafenib in a Phase III trial[17]. Furthermore, our analysis includes data on survival, response rate, and safety, which in combination are important for treatment decision-making, particularly for patients who may be candidates for downstaging. Finally, the present meta-analysis included a sub-group analysis of patients with HBV infection, which is an important population in the Asia-Pacific region and has not been covered by other current meta-analyses.

MATERIALS AND METHODS

The analysis methods and inclusion criteria for this study were specified in advance and the protocol was prospectively submitted for registration in the PROSPERO database on May 26, 2020. This report has been written in line with the PRISMA guidelines for network meta-analyses.

Eligibility criteria

This analysis included randomized controlled trials conducted in adult patients (age ≥ 18 years) with advanced or unresectable HCC not eligible for, or with disease progression after, surgical or locoregional therapies. Eligible studies included patients with Child-Pugh Class A or B liver function, ≥ 1 measurable lesion, and no evidence of untreated brain or meningeal metastases. Eligible studies were also required to report at least an assessment of tumor response, survival [OS, PFS, or time to progression (TTP)], and safety. The analysis excluded studies including patients with Child-Pugh Class C liver function, patients receiving anticoagulation therapy or antiretroviral therapy for HIV, and patients who had received previous systemic treatment. These broad eligibility criteria covered a number of trials reporting negative results vs sorafenib. Although the analysis therefore includes multiple therapies that failed clinical trials in HCC, this allowed the collection of data for sorafenib from studies conducted over a wide time range, which improved the precision of the analysis.

Information sources, search strategy, and study selection

Studies were identified by searching the following electronic databases: PubMed, Science Direct, and the Cochrane Database, and Excerpta Medica Database Abstracts from the American Society of Clinical Oncology (ASCO) 2020 annual congress were also searched. The search was completed on May 21, 2020 using the search terms shown in Figure 1.

Figure 1

Flow chart of study selection and search terms. Different therapeutics were searched for using individual searches to allow easier processing of the results; NOT was used for databases allowing use of this operator. ASCO: American Society of Clinical Oncology.

The Cochrane risk of bias tool for randomized controlled trials was used to assess the quality and risk of bias of studies included in the analysis[18].

Data extraction

Data were independently extracted by two evaluators (Luo JF and Huang XQ) and cross-checked. In the case of disagreement, the original documents were checked and the correct data confirmed. General information extracted included journal name, document title, publication time, author, country, region where the lead author was located, and the country and region where the research was conducted. Demographics and baseline characteristics extracted were patient age, gender, Barcelona clinic liver cancer classification, Eastern Co-operative Oncology Group performance status, prevalence of HBV infection, and presence of extrahepatic vascular infiltration and extrahepatic metastasis. Details of interventions extracted included dosage and dose schedule. Efficacy and safety endpoints extracted (where available) were overall response rate [ORR; assessed by response evaluation criteria in solid tumours (RECIST) v1/1.1 for all included studies], OS, PFS, TTP, incidence of Grade ≥ 3 adverse events (AE), incidence of treatment interruption due to adverse events (AEs), and incidence of dose reductions due to AEs.

Statistical analysis

For OS, PFS, TTP, and other survival endpoints, hazard ratios (HR) were estimated to compare treatments. For discrete variables such as ORR, and incidence of AEs, estimated risk ratios were calculated to compare treatments. Selection of a fixed effect or random effect model was based on the level of heterogeneity in the data, assessed using the Higgins I statistic and defined as I ≤ 50% and P > 0.1. If no obvious data heterogeneity was found, a fixed effect model was adopted, otherwise a random effect model was utilized. For endpoints reported in a relatively small number of studies (< 6), a fixed effects model was adopted.

A network meta-analysis was used to synthesize information from the included studies, and perform direct and indirect comparisons using a method based on the frequency school of Rücker et al[19,20]. The Q statistic was used to assess the consistency of direct and indirect evidence in the treatment network(s) studied. If no obvious inconsistency (P > 0.1) was found, a fixed effect model was adopted, otherwise a random effect model was utilized. P value, a frequentist analog to the surface under the cumulative ranking curve, was used to rank treatments[21]. A funnel chart was used to evaluate publication bias; a symmetrical graph indicates a low influence of publication bias and an asymmetric graph indicates possible publication bias. A post-hoc analysis of all studies reporting data from patients with HBV-related HCC was also included to assess OS, PFS, and safety in these patients.

All statistical analyses were performed using Rv3.6. The Robias toolkit was used for evaluation of literature quality and Netmeta was used for the network meta-analysis.

RESULTS

Studies included in the analysis

In total, 1398 articles were screened: PubMed/MEDLINE, n = 114; Science Direct, n = 312; Cochrane Database, n = 355; Excerpta Medica Database, n = 561; and the ASCO 2020 abstract book, n = 12 (Figure 1). After removing duplicates and top-line screening of abstracts for suitability, a total of 86 articles were reviewed in detail, of which 27 met the full inclusion criteria (Table 1). These 27 articles corresponded to 27 different studies (Supplementary Figure 1).

Table 1

Details of included studies

Ref.	Year	Experimental arm(s)	Comparator arm	Primary endpoint	Analysis timing	Survival outcomes, mo
Yen et al[47]	2018	Nintedanib	Sorafenib	TTP	PFS: Final; OS: Final	PFS: 2.7 vs 3.7; OS: 10.2 vs 1.1
Ciuleanu et al[48]	2016	Mapatumumab + sorafenib	Placebo + sorafenib	TTP	PFS: Final; OS: Final	PFS: 3.2 vs 4.3; OS: 10.0 vs 10.1
Finn et al[11]	2020	Atezolizumab + bevacizumab	Sorafenib	OS and PFS	PFS: Final; OS: Final	PFS: 6.8 vs 4.3; OS: NE vs 13.2
Abou-Alfa et al[49]	2010	Doxorubicin + sorafenib	Doxorubicin + placebo	TTP	PFS: Final; OS: Final	PFS: 6.0 vs 2.7; OS: 13.7 vs 6.5
Cheng et al[50]	2016	Dovitinib	Sorafenib	OS and TTP	TTP: Final; OS: Final	TTP: 4.1 vs 4.1; OS: 8.0 vs 8.4
Cheng et al[28]	2015	Tigatuzumab (6 + 2) + sorafenib; Tigatuzumab (6 + 6) + sorafenib	Sorafenib	TTP	TTP: Final; OS: Final	TTP: 3.0 vs 3.9 vs 2.8; OS: 8.2 vs 12.2 vs 8.2
Hsu et al[51]	2012	Vandetanib 300 mg/d; Vandetanib 100 mg/d	Placebo	Tumor stabilization rate	PFS: Final; OS: Final	PFS: 1.1 vs 0.7 vs 1.0; OS: 6.0 vs 5.8 vs 4.3
Johnson et al[22]	2013	Sorafenib	Brivanib	OS	PFS: No; OS: Final	PFS: 4.1 vs 4.2; OS: 9.9 vs 9.5
Cainap et al[24]	2015	Linifanib	Sorafenib	OS	PFS: Final; OS: Final	PFS: 4.2 vs 2.9; OS: 9.1 vs 9.8
Kudo et al[10]	2018	Lenvatinib	Sorafenib	OS	PFS: No; OS: Final	PFS: 7.4 vs 3.7; OS: 13.6 vs 12.3
Yau et al[23]	2019	Nivolumab	Sorafenib	OS	PFS: Final; OS: Final	PFS: 3.7 vs 3.8; OS: 16.4 vs 14.7
Cheng et al[29]	2013	Sunitinib	Sorafenib	OS	PFS: Final;OS: Final	PFS: 3.6 vs 3.0; OS: 7.9 vs 10.2
Zhu et al[26]	2015	Sorafenib + erlotinib	Sorafenib + placebo	OS	TTP: Final; OS: Final	TTP: 3.2 vs 4.0; OS: 9.5 vs 8.5
Llovet et al[52]	2008	Sorafenib	Placebo	OS and TTP	TTP: Final; OS: Final	TSP: 5.5 vs 2.8; OS: 10.7 vs 7.9
Cheng et al[25]	2009	Sorafenib	Placebo	-	TTP: Final; OS: Final	TTP: 2.8 vs 1.4; OS: 6.5 vs 4.2
Palmer et al[53]	2018	Nintedanib	Sorafenib	TTP	PFS: Final; OS: Final	PFS: 5.3 vs 3.9; OS: 11.9 vs 11.4
Thomas et al[54]	2018	Bevacizumab + erlotinib	Sorafenib	OS	PFS: No; OS: Final	PFS: 4.4 vs 2.8; OS: 8.6 vs 8.6
Abou-Alfa et al[55]	2019	Sorafenib + doxorubicin	Sorafenib	OS	PFS: Final; OS: Final	PFS: 4.0 vs 3.7; OS: 9.3 vs. 9.4
Tak et al[27]	2018	Sorafenib	Sorafenib + resminostat	TTP	TTP: Final; OS: Final	TTP: 2.8 vs 2.8; OS: 14.1 vs 11.8
Jouve et al[56]	2019	Sorafenib + pravastatin	Sorafenib	OS	PFS: Final; OS: Final	PFS: 5.0 vs 5.4; OS: 10.7 vs 10.5
Lee et al[57]	2016	AEG35156 + sorafenib	Sorafenib	PFS	PFS: Final; OS: Final	PFS: 4.0 vs 2.6; OS: 6.5 vs 5.4
Assenat et al[58]	2019	Sorafenib + GEMOX	Sorafenib	PFS	PFS: Final; OS: Final	PFS: 6.2 vs 4.6; OS:13.5 vs 14.8
Azim et al[59]	2018	Sorafenib + tegafur–uracil	Sorafenib	TTP	PFS: Final; OS: Final	PFS: 6.0 vs 6.0; OS: 8.2 vs 10.5
Koeberle et al[60]	2016	Sorafenib	Sorafenib + everolimus	PFS	PFS: Final; OS: Final	PFS: 6.6 vs 5.7; OS: 10.0 vs 12
Bi et al[17]	2020	Donafinib	Sorafenib	OS	PFS: Final; OS: Final	PFS: 3.7 vs 3.6; OS: 21.1 vs 10.3
Qin et al[61]	2013	FOLFOX4	Doxorubicin	OS	PFS: Final; OS: Final	PFS: 2.9 vs 1.8; OS: 6.4 vs 5.0
Yeo et al[62]	2005	Doxorubicin	PIAF	OS	PFS: No; OS: Final	OS: 6.8 vs 8.7

FOLFOX4: Oxaliplatin/folinic acid/5-fluorouracil; GEMOX: Gemcitabine and oxaliplatin; NE, Not reported; OS, Overall survival; PFS, Progression-free survival; PIAF, Cisplatin/interferon α-2b/doxorubicin/5-fluorouracil; TTP, Time to progression.

Study characteristics

Of the 27 studies included, 25 investigated targeted treatment regimens (nintedanib, mapatumumab + sorafenib, atezolizumab + bevacizumab, doxorubicin + sorafenib, dovitinib, tigatuzumab + sorafenib, vandetanib, brivanib, linifanib, lenvatinib, nivolumab, sunitinib, sorafenib + erlotinib, sorafenib (two studies), nintedanib, bevacizumab + erlotinib, sorafenib + doxorubicin, sorafenib + resminostat, sorafenib + pravastatin, AEG35156 (a second-generation synthetic antisense oligonucleotide inhibitor of cellular expression of the X-linked inhibitor of apoptosis protein) + sorafenib, sorafenib + gemcitabine and cisplatin, sorafenib + tegafur–uracil, sorafenib + everolimus, and donafinib) and two investigated combination chemotherapy regimens [oxaliplatin/folinic acid/5-fluorouracil (FOLFOX4) and cisplatin/interferon α-2b/doxorubicin/5-fluorouracil] (Table 1). Twenty-one of the included studies used sorafenib as the comparator treatment, three used doxorubicin, and three studies were placebo controlled (including the two Phase III studies of sorafenib). The majority of the studies had OS (n = 12) or PFS/TTP (n = 10) as the primary endpoint, and almost all had reported final data for these endpoints.

Quality assessment

Study design characteristics are summarized in Supplementary Table 1. In brief, all 27 studies selected for inclusion were randomized controlled studies (20 provided details of the randomization scheme used and seven articles did not specify), seven of the studies used double blinding and 20 were open label, and 24 included a data flow chart. Overall, the quality of the included studies was considered relatively high (Figure 1).

Patient description

All of the studies included patients with advanced HCC who had not received previous treatment. Overall, the total of 10256 patients included in the analysis were predominantly male and had median ages ranging from 49 to 68 years, and most of the studies included > 50% of patients with extrahepatic metastasis (Table 2).

Table 2

Patient characteristics in the included studies

Ref.	Year	Treatments	n	Age, median	Males, %	ECOG 0/1–2, %	Extrahepatic disease, %
Yen et al[47]	2018	Nintedanib	63	58	91	55.6/44.5	68.3
		Sorafenib	32	62	81	56.3/43.8	68.3
Ciuleanu et al[48]	2016	Mapatumumab + sorafenib	50	60	52	36.0/64.0	66.0
		Placebo + sorafenib	51	61	77	33.3/66.6	49.0
Finn et al[11]	2020	Atezolizumab + bevacizumab	336	64	82	62.0/38.0	63.0
		Sorafenib	165	66	83	62.0/38.0	56.0
Abou-Alfa et al[49]	2010	Doxorubicin + sorafenib	47	66	66	-	51.1
		Doxorubicin + placebo	49	65	86	-	79.6
Cheng et al[50]	2016	Dovitinib	82	56	89	63.0/37.0	-
		Sorafenib	83	56	81	64.0/35.0	-
Cheng et al[28]	2015	Tigatuzumab (6 + 2) + sorafenib	53	63	85	60.4/39.6	-
		Tigatuzumab (6 + 6) + sorafenib	54	63	83	57.4/42.6	-
		Sorafenib	55	66	80	54.5/45.5	-
Hsu et al[51]	2012	Vandetanib 300 mg/d	19	55	95	-	-
		Vandetanib 100 mg/d	25	61	68	-	-
		Placebo	23	56	87	-	-
Johnson et al[22]	2013	Sorafenib	578	60	84	61.0/39.0	62.0
		Brivanib	577	61	84	64.0/36.0	63.0
Cainap et al[24]	2015	Linifanib	514	59	86	62.8/37.2	59.7
		Sorafenib	521	60	84	66.2/33.8	56.8
Kudo et al[10]	2018	Lenvatinib	478	63	85	-	-
		Sorafenib	476	62	84	-	-
Yau et al[23]	2019	Nivolumab	371	65	85	-	-
		Sorafenib	372	65	85	-	-
Cheng et al[29]	2013	Sunitinib	530	59	82	52.5/46.8	78.9
		Sorafenib	544	59	84	52.9/46.7	76.3
Zhu et al[26]	2015	Sorafenib + erlotinib	362	60	82	61.3/38.7	56.6
		Sorafenib + placebo	358	61	80	60.3/39.7	61.2
Llovet et al[52]	2008	Sorafenib	299	65	87	54.0/46.0	53.0
		Placebo	303	66	87	54.0/46.0	50.0
Cheng et al[25]	2009	Sorafenib	150	51	85	25.3/74.6	68.7
		Placebo	76	52	87	27.6/72.4	68.4
Palmer et al[53]	2018	Nintedanib	62	66	77	51.6/48.4	64.5
		Sorafenib	31	64	84	58.1/33.0	67.7
Thomas et al[54]	2018	Bevacizumab + erlotinib	47	61	NR	32.0/68.0	40.0
		Sorafenib	43	61	NR	40.0/60.0	25.0
Abou-Alfa et al[55]	2019	Sorafenib + doxorubicin	180	62	85	36.1/63.9	-
		Sorafenib	176	62	87	39.8/60.2	-
Tak et al[27]	2018	Sorafenib	84	62	87	-	56.0
		Sorafenib + resminostat	86	65	80	-	51.8
Jouve et al[56]	2019	Sorafenib + pravastatin	162	68	96	-	29.0
		Sorafenib	161	68	88	-	30.4
Lee et al[57]	2016	AEG35156 + sorafenib	31	61	87	3.2/96.8	-
		Sorafenib	17	54	88	11.8/88.3	-
Assenat et al[58]	2019	Sorafenib + GEMOX	39	62	86	-	77.0
		Sorafenib	44	65	92	-	61.0
Azim et al[59]	2018	Sorafenib + tegafur–uracil	36	59	86	69.4/30.6	52.8
		Sorafenib	38	59	90	65.8/34.2	47.4
Koeberle et al[60]	2016	Sorafenib	46	65	87	72.0/28.0	57.0
		Sorafenib + everolimus	59	66	81	59.0/41.0	54.0
Bi et al[17]	2020	Donafinib	328	53	86	61.3/38.7	-
		Sorafenib	331	53	88	66.8/33.2	-
Qin et al[61]	2013	FOLFOX4	184	50	90	-	-
		Doxorubicin	187	49	87	-	-
Yeo et al[62]	2005	Doxorubicin	94	54	90	87.2/12.8	-
		PIAF	94	49	93	92.6/7.4	-

ECOG: Eastern Co-operative Oncology Group; FOLFOX4: Oxaliplatin/folinic acid/5-fluorouracil; GEMOX: Gemcitabine and oxaliplatin; NR: Not reported; PIAF: Cisplatin/interferon α-2b/doxorubicin/5-fluorouracil.

Evaluation of efficacy

Overall response rate: A total of 18 studies reported ORR, including 19 interventions and allowing 20 comparisons (Figure 2A). No significant heterogeneity was detected between the studies (tau-squared = 0; I = 0%; P = 0.9502) and a fixed effect model was selected. P value for ORR showed that lenvatinib was associated with the best ORR among all treatments included in the analysis (P = 0.9042) (Figure 2B). Atezolizumab + bevacizumab ranked second (P = 0.8045) and nivolumab ranked third (P = 0.7834). Using lenvatinib as the comparator, all treatments included in the analysis had an estimated risk ratio for ORR (RRORR) of < 1, except for AEG35156 + sorafenib, which had an estimated RRORR of 1.3451 [95% confidence interval (CI): 0.07-25.21] (Figure 2B).

Figure 2

Response rates of first-line systemic therapy in patients with advanced hepatocellular carcinoma. A: Network diagram; B: Interventions ranked by P value with risk ratios and 95% confidence interval for overall response rate for each treatment vs lenvatinib. CI: Confidence interval; GEMOX: Gemcitabine and oxaliplatin; ORR: Overall response rate; RR: Risk ratio.

Progression-free survival: A total of 15 studies reported PFS, including 15 interventions and allowing 15 comparisons (Figure 3A). No significant heterogeneity was detected (tau-squared = 0; I = 0%; P = 0.7361) and a fixed effect model was selected. Atezolizumab + bevacizumab was ranked first for PFS (P = 0.9501), followed by lenvatinib (P = 0.9041). Nivolumab ranked sixth (P = 0.558) (Figure 3B). With the exception of atezolizumab + bevacizumab (HRPFS = 0.90; 95%CI: 0.64-1.25), the estimated HRs for PFS for all included treatments vs lenvatinib were > 1; however, the associated 95%CI passed through unity for bevacizumab plus erlotinib, linifanib, and FOLFOX4.

Figure 3

Survival outcomes in patients with advanced hepatocellular carcinoma following first-line systemic therapy. A, C, and E: Network diagrams; B: Interventions ranked by P value with hazard ratios for progression-free survival, D: Time to progression and F: overall survival for each treatment vs lenvatinib. CI: Confidence interval; FOLFOX4: Oxaliplatin/folinic acid/5-fluorouracil; HCC: Hepatocellular carcinoma; HR: Hazard ratio; OS: Overall survival; PIAF: Cisplatin/interferon α-2b/doxorubicin/5-fluorouracil; TTP: Time to progression.

Time to progression: A total of 17 studies reported TTP, including 17 interventions and allowing 19 comparisons (Figure 3C). No significant heterogeneity was detected between studies (tau-squared = 0; I = 0%; P = 0.9028) and a fixed effect model was selected. Lenvatinib was ranked first for TTP (P = 0.9888) followed by linifanib (P = 0.9067) and sorafenib + doxorubicin (P = 0.7344) (Figure 3D). When compared with lenvatinib, all other treatments in the analysis had an estimated HRTTP > 1, although the associated 95%CI passed through unity for linifanib and sorafenib plus tegafur–uracil.

Overall survival: A total of 24 studies reported OS, including 25 interventions and allowing 28 comparisons (Figure 3E). No significant heterogeneity was detected between studies (tau-squared = 0; I = 0%; P = 0.9802) and a fixed effect model was selected. Atezolizumab + bevacizumab was ranked highest for OS (P = 0.9651) followed by vandetanib 100 mg/d (P = 0.8653), donafinib (P = 0.7958), and nivolumab (P = 0.7701) (Figure 3F). Lenvatinib ranked sixth (P = 0.6675). Atezolizumab + bevacizumab was associated with a lower risk of death vs lenvatinib (HRos = 0.63; 95%CI: 0.44-0.89), and the HRos for most other treatments vs lenvatinib had associated 95%CIs that passed through unity.

Outcomes in patients with HBV infection: Ten studies included sub-analyses of patients with HBV infection, including data on the following treatments: atezolizumab + bevacizumab[11], brivanib[22], nivolumab[23], lenvatinib[10], linifanib[24], sorafe-nib[25], sorafenib + erlotinib[26], sorafenib + resminostat[27], tigatuzumab + sorafenib[28], and sunitinib[29]. A total of three studies reported PFS in patients with HBV infection, including four interventions and three comparisons (Supplementary Figure 2A). A fixed effect model was selected for the analysis. Lenvatinib ranked first for PFS (P = 0.8786) followed by atezolizumab + bevacizumab (P = 0.7746) and donafinib (P = 0.2972) (Figure 2B). A comparison of HRs for PFS vs lenvatinib is shown in Supplementary Figure 2B. A total of nine studies reported OS, including ten interventions and allowing nine comparisons (Supplementary Figure 2C). A random effect model was selected for the analysis. Atezolizumab + bevacizumab ranked first (P = 0.9751), followed by lenvatinib (P = 0.8308) and nivolumab (P = 0.7732) (Supplementary Figure 2D). Comparison with lenvatinib revealed that atezolizumab + bevacizumab had an estimated HROS < 1 and all other interventions had an HROS > 1 (Supplementary Figure 2D).

Safety

Grade ≥ 3 adverse events: In total, 17 studies reported data on the incidence of Grade ≥ 3 AEs, including 19 interventions and allowing 21 comparisons (Supplementary Figure 3A). No significant heterogeneity was detected between studies (tau-squared = 0; I = 0%; P = 0.4493) and a fixed effect model was selected. Nivolumab ranked 2/19 (P = 0.9351), sorafenib ranked 8/19 (P = 0.5040), atezolizumab + bevacizumab ranked 11/19 (P = 0.4167), and lenvatinib ranked 16/19 (P = 0.2468) for incidence of Grade ≥ 3 AEs (higher ranking indicated a lower incidence of AEs) (Supplementary Figure 3B).

Treatment termination due to adverse events: A total of 13 studies reported the incidence of treatment termination due to AEs, including 13 interventions and allowing 15 comparisons (Supplementary Figure 3C). A degree of heterogeneity was detected between studies (tau-squared = 0.1536; I = 65%; P = 0.0573) and a random effect model was selected. After ranking all interventions from the lowest to highest incidence of terminations due to AEs, vandetanib 300 mg/d and 100 mg/d were ranked first and second (P = 0.8036 and 0.7252, respectively), sorafenib ranked 5/13 (P = 0.5372), nintedanib ranked 8/13 (P = 0.4251), lenvatinib ranked 10/13 (P = 0.3907), and atezolizumab + bevacizumab ranked 13/13 (P = 0.2584) (Supplementary Figure 3D).

DISCUSSION

Following an expansion of first-line systemic treatment options for HCC over the past decade, international treatment guidelines now recommend sorafenib, lenvatinib, and atezolizumab plus bevacizumab in this setting, as well as nivolumab and FOLFOX (off-label use in many countries, but approved by the China National Medical Products Administration) for selected patients[4,12,13]. Numerous other therapies and combinations of therapies have also been unsuccessfully investigated in first-line advanced HCC management. However, most trials of systemic therapy for HCC used sorafenib as the comparator, as it was the only approved systemic therapy available at the time, and this limits the clinicians’ ability to compare currently available treatment options. The present study represents one of the most comprehensive systematic reviews and meta-analyses of first-line systemic treatments for advanced unresectable HCC conducted to date, and compares the treatment outcomes and safety of lenvatinib with multiple other systemic therapies, including immunotherapy (nivolumab) and combined therapy with immunotherapy and a TKI (atezolizumab + bevacizumab).

Our results show that atezolizumab plus bevacizumab is associated with the best OS outcomes of all therapies included in the analysis. This result is supported by findings from a recent meta-analysis that investigated optimal treatment sequencing for HCC and also reported that atezolizumab plus bevacizumab had a higher OS benefit vs lenvatinib (HROS = 0.63; 95%CI: 0.44-0.89), nivolumab (HROS = 0.68; 95%CI: 0.48-0.98), and sorafenib (HROS = 0.58; 95%CI: 0.42-0.80)[16]. Atezolizumab plus bevacizumab is the first combined immunotherapy and vascular-targeted regimen to be recommended as a first-line treatment option in the National Comprehensive Cancer Network HCC guidelines[4]. The long OS associated with atezolizumab plus bevacizumab may be related to the ‘long tail’ effect characteristic of immune checkpoint inhibitors, which was also observed in the Phase III Checkmate 459 study of nivolumab. A number of studies have identified several mechanisms by which angiogenesis-related processes can enhance immune checkpoint inhibitors therapy, including vascular normalization, reduction of hypoxia, and increasing tumor infiltrating lymphocytes[30]. Although bevacizumab monotherapy failed Phase II trials in unresectable HCC, in combination with atezolizumab it led to superior efficacy compared with bevacizumb monotherapy[31]. However, consideration of treatment safety and tolerability is also an important factor in clinical decision-making. Our analysis revealed that atezolizumab plus bevacizumab was associated with the highest incidence of discontinuation due to AEs. This may be associated with the relatively long time to progression and duration of treatment reported for atezolizumab plus bevacizumab, but as treatment discontinuations due to AEs usually involve uncontrolled Grade ≥ 3 AEs, this would likely be a weak association. In addition, the prescribing information for bevacizumab highlights a possible risk of bleeding, and requires termination of bevacizumab at least 4 wk before surgery[32]. Therefore, in patients with high risk of gastric esophageal varices and patients with the potential to undergo any surgical procedures, atezolizumab plus bevacizumab should be used carefully, to manage the risk of bleeding events. Atezolizumab plus bevacizumab may be more suitable for patients who are unsuitable for surgery but with good liver function and limited cirrhosis, who have the potential to achieve a long-term survival benefit with systemic therapy.

The results of this meta-analysis show that there is currently not one single systemic treatment for advanced HCC associated with superior outcomes across all outcome measurements (ORR, OS, PFS, and safety). This highlights the importance of individualized treatment selection based on specific treatment goals. For example, a number of studies have shown that lenvatinib or lenvatinib combination therapy[33] can allow patients to achieve downstaging and become eligible for surgery[34-36]. For patients with HCC ineligible for surgical intervention at diagnosis, we are of the opinion that treatment selection should be objective based. In patients without serious underlying liver disease and for whom surgery may be possible, systemic treatments with the highest ORR are the optimal choice. Conversely, for patients with poor liver function, underlying liver disease, or local advanced HCC, selection of therapies based on longer OS may provide the most benefit.

In our analysis, lenvatinib had superior short-term efficacy compared with all other systemic therapies investigated. Lenvatinib ranked first for ORR and TTP, and second for PFS after atezolizumab plus bevacizumab. This finding is supported by the results of another recent network meta-analysis presented at the ASCO Gastrointestinal Symposium 2021 that also ranked atezolizumab plus bevacizumab first for OS but lenvatinib first for ORR[37]. In addition, although direct comparison of the ORRs (RECIST v1.1) reported for atezolizumab plus bevacizumab and lenvatinib in the IMbrave 100 and REFLECT studies appears to show a moderately higher ORR for atezolizumab plus bevacizumab (27% vs 18%)[10,11], our network analysis provides a more robust comparison of the two therapies by comparing both to sorafenib. There are several possible mechanistic explanations for this finding. First, preclinical studies show that lenvatinib has multiple targets including VEGFR1-3, FGFR1-4, PDGFRα, RET, and KIT, and this broad spectrum of activity may be one factor explaining the high response rates associated with this therapy[38]. Furthermore, lenvatinib is a type V TKI with fast binding and relatively slow dissociation compared with other TKIs[39]. In addition to anti-vascular effects, lenvatinib also has a regulatory effect on the immune microenvironment of liver cancer[40]. Preclinical research has shown that, compared with sorafenib, lenvatinib has a significant anti-tumor effect in immunodeficient mice, suggesting that lenvatinib may activate immune function by decreasing the number of tumor-associated macrophages, increasing the proportion of activated CD8+ cells[40], and increasing activation and infiltration of natural killer cells[41].

HBV-related liver cancer is particularly prevalent in Asian populations, especially in China where 69%-80% of liver cancers have an HBV etiology[42,43]. Our meta-analysis of data from patients with HCC and HBV infection suggested that, in terms of OS, atezolizumab plus bevacizumab, lenvatinib, and nivolumab are the three most effective treatments in this patient population. For PFS, lenvatinib ranked first over atezolizumab plus bevacizumab. This finding supports previous meta-analyses that have shown lenvatinib to have a particularly strong anti-tumor effect vs sorafenib in patients with HBV-related HCC[44,45]. It is unclear why lenvatinib may have a particularly good anti-tumor effect in HBV-related HCC, but it may be due to the impact of lenvatinib on the immune microenvironment, as described above. In addition, the China National Health and Health Commission guidelines for diagnosis and treatment of primary liver cancer in China (2019 edition) recommend lenvatinib as a systemic therapy with good efficacy in patients with HBV-related liver cancer[46].

This network meta-analysis had several possible limitations. First, the quality of studies included in the analysis had some heterogeneity; for example, the analysis included both large Phase III clinical trials, such as REFLECT and Checkmate 459, and smaller Phase II clinical studies. Second, there was also heterogeneity in the patient populations included in the analysis, including patients from different geographic regions, of different races, and different proportions of patients with HBV infection. Additionally, it should be noted that second-line therapeutic options for HCC have greatly improved over the past decade. As a result, estimates of first-line OS from older studies are generally shorter than those from more recent studies. However, among the therapies included in this analysis that are currently approved for first-line HCC, only sorafenib has OS data old enough to potentially be biased by this phenomenon. Fortunately, our analysis was based on pooled data from the pivotal study of sorafenib in 2008 and comparator arms of trials conducted between 2008 and 2020, which limits the potential effect of bias from improvements in second-line therapies[47-62]. Finally, because multiple interventions were included in the analysis, several had data from only one study and therefore a relatively small sample size, which may have led to bias.

CONCLUSION

This network meta-analysis of first-line systemic therapies for advanced HCC revealed that atezolizumab plus bevacizumab is associated with the best OS and PFS, but also with a high incidence of discontinuation due to AEs. The results also showed that lenvatinib is associated with the best ORR of all systemic therapies included in the analysis, as well as a relatively high PFS, particularly in patients with HBV-related liver cancer in whom lenvatinib ranked first for PFS, over atezolizumab plus bevacizumab. Therefore, in patients with unresectable advanced HCC, systemic treatment should be selected based on the individualized treatment goals of each patient.

ARTICLE HIGHLIGHTS

Research background

The recent expansion of first-line systemic therapy options for patients with advanced hepatocellular carcinoma represents a significant advance in the treatment of this disease. However, the majority of clinical trials in first-line hepatocellular carcinoma management used placebo or sorafenib as comparators, and there are limited data providing a cross comparison of the efficacy and safety of treatments in this setting, especially for newly-approved immune checkpoint inhibitor and vascular endothelial growth factor inhibitor combination treatments.

Research motivation

Clinical trials of recently-approved therapies for hepatocellular carcinoma have revealed differing profiles of efficacy and safety, and comparative data to inform selection of first-line treatments for individual patients are limited. Furthermore, although lenvatinib is widely seen as a standard of care in real clinical practice, and is a recommended first-line therapy in most international treatment guidelines, there are limited head-to-head data comparing lenvatinib with other systemic therapies.

Research objectives

The objectives of this network meta-analysis were to systematically review and compare the response rates, survival outcomes, and safety of first-line systemic therapies for advanced hepatocellular carcinoma, and to provide a comparison between lenvatinib and other systemic therapies in this setting. The study also included a sub-group analysis of patients with hepatitis B virus infection, which is an important population in the Asia-Pacific region and has not been covered by other current meta-analyses.

Research methods

We searched PubMed, Science Direct, the Cochrane Database, Excerpta Medica Database, and abstracts from the American Society of Clinical Oncology 2020 annual congress. Eligible studies were randomized controlled trials of systemic therapy enrolling adults with advanced/unresectable hepatocellular carcinoma. A network meta-analysis was used to synthesize data and perform direct and indirect comparisons between treatments for endpoints including (where available) overall response rate, overall survival, progression-free survival, time-to-progression, incidence of Grade ≥ 3 adverse events, incidence of treatment interruptions due to adverse events, and incidence of dose reductions due to adverse events. P value, a frequentist analog to the surface under the cumulative ranking curve, was used to rank treatments.

Research results

Treatments included in the analysis were atezolizumab plus bevacizumab, brivanib, donafenib, dovitinib, FOLFOX4, lenvatinib, linifanib, nintedanib, nivolumab, sorafenib, sunitinib, vandetanib, 11 sorafenib combination therapies, and three other combination therapies. Atezolizumab plus bevacizumab was ranked first for progression-free survival and overall survival but also had the highest rate of discontinuations due to adverse events. Lenvatinib ranked first for overall response rate and second for progression-free survival. Our findings show that first-line systemic treatment should be selected based on individualized treatment goals and provide valuable comparative data that can help to inform treatment decisions.

Research conclusions

Our findings suggest that there is no one single first-line treatment for advanced hepatocellular carcinoma associated with superior outcomes across all outcome measurements. Therefore, first-line systemic treatment should be selected based on individualized treatment goals.

Research perspectives

Future research should continue to evaluate new therapeutic strategies for hepatocellular carcinoma in the context of existing treatments, and provide further information to support treatment selection for individual patients.