This study included a data collection of randomized controlled trials (RCTs), which evaluated adult patients (aged 18 and older) with unresectable HCC to receive combination therapy of intravenous atezolizumab (1200 mg) plus bevacizumab (15 mg/kg) every 3 wk (or periodically). The study dataset was further divided into a control segment consisting of sorafenib monotherapy, atezolizumab monotherapy, or placebo.

The RCTs incorporated primary efficacy outcomes of mortality, measured as a median number of deaths and stratified hazard ratios (HRs). Moreover, the Response Evaluation Criteria in Solid Tumors (RECIST) criteria (Supplementary Table 1) measured secondary outcomes of OS, median OS, and median PFS as tumor response proportions or percentages. An example of the aforementioned was demonstrated in the overall confirmed objective response, confirmed complete and partial responses, stable disease, progressive disease, and disease control rate.

RCTs safety measurements evaluated included patients with adverse events (AEs) from causes that included serious treatment-related AEs and treatment-related mortality events. Additionally, AEs that resulted from drug dosing modifications and/or interruptions were evaluated along with drug withdrawal trials that included participants with Grade 3-5 AEs. Any unfavorable clinical or laboratory result associated with an investigational intervention during the clinical trial was considered an AE; hence, this included any unfavorable and life-threatening medical outcome that required inpatient hospitalization or prolongation of hospitalization.

All RCTs evaluated in the study had documented hepatitis virological status as well as a Child-Pugh classification of A or B (Supplementary Table 2). However, the data excluded trials involving patients who received treatments for medical conditions other than HCC, as well as participants with autoimmune liver disease or any autoimmune conditions and participants in Child-Pugh class C. Non-human studies, non-English and unpublished articles were also excluded.

A comprehensive and extensive literature review of published articles was conducted to identify RCTs that met the inclusion and exclusion criteria using appropriate MeSH terms. This systematic review was performed following the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses. The search was conducted using the Cochrane, Cochrane Central, PubMed Central, Scopus, ScienceDirect, WHO trials, clinicaltrials.gov, Google Scholar, Embase, CINAHL, and MedLine databases. The following terms and Boolean operators were employed in MeSH and free-text searches to identify relevant articles: “Hepatocellular carcinoma,” “liver tumor,” “liver cancer,” “atezolizumab and bevacizumab,” and “sorafenib.”

The data search was conducted until December 27, 2020. The search criteria were broadened by identifying additional studies from the reference lists of selected articles, as well as by the “related articles” function of PubMed. Additionally, the systematic review was registered in PROSPERO, the international prospective registry of systematic reviews of the National Institute for Health Research (CRD42021237736). For transparency, the study was pre-registered on the open science framework (URL: https://osf.io/esvk9), and in PROSPERO.

Selection of studies: All articles used in this document were screened for eligibility via their titles and abstracts. Thereafter, the full-text of all chosen studies was examined in detail. Two independent reviewers were chosen to perform the screening and examination process according to predefined eligibility criteria for the qualitative review.

Two review authors (MKC and MZ) independently extracted the data and summarized the trial characteristics in each table. They were also involved in extracting the baseline characteristics of the participants, study design, geography, settings, methods, types of interventions (dosage, route of administration, regimen protocol), efficacy, and safety outcomes.

This study assessed the risk of bias (RoB) in the included studies by using the revised Cochrane RoB 2 tool for randomized trials. This tool was used to assess the RoB across the following five domains: Bias arising from the randomization process, bias due to deviation from the intended intervention, bias due to missing outcome, bias in the measurement of outcome, and bias in the selection of the reported results. Moreover, the Robvis data software was used to create a RoB traffic light plot (Figure 1) and a RoB summary plot (Figure 2).

Open in New Tab   Full Size Figure   Download Figure

Figure 1 Traffic light plot showing the risk of bias of the two completed studies.

Open in New Tab   Full Size Figure   Download Figure

Figure 2 Summary plot of the risk of bias assessment of the two completed studies.

The studies that measured relevant objective data outcomes (e.g., mean survival rates) according to their study protocols, but included non-retrievable online data, missing content or unclear data, were cleared up by reaching out to the original authors of the published reports. However, if no response was obtained from the original authors, the selected study was excluded from the analyses.

A descriptive analysis of all study results was performed. All continuous variables such as mean and median were analyzed, and all dichotomous outcomes were investigated as proportions and percentages. Furthermore, epidemiological variables (e.g., risk ratio, attributable risk, and numbers needed to treat) to measure certain effects of intervention such as mortality were estimated as deemed necessary.

This review included the two concluded trials in the present analyses (Finn et al[12] and Lee et al[16]) as well as the results of two currently ongoing trials of La Roche[14] and Hack et al[15] that will be used in future updates.

La Roche[14] is a Phase IIIb, single-component, multicenter study of atezolizumab/bevacizumab, which is currently ongoing. Hack et al[15] is also currently ongoing, and is evaluating randomized patients included in an intervention dataset (atezolizumab/bevacizumab) and patients assigned to the control portion of the dataset undergoing active surveillance. Patients included in the control were allowed to crossover to the intervention dataset (atezolizumab/bevacizumab) after confirmed recurrence.

Lee et al[16] is part of an open-label, multicenter, multi-segmental, phase 1b study also known as GO30140 study, which enrolled patients at 26 academic centers and community oncology practices in seven countries worldwide. The study included five cohorts, but only the results of the two HCC cohorts, Groups A (atezolizumab monotherapy) and F, are described within this review article. Finn et al[12] compared the intervention dataset (atezolizumab/bevacizumab) with the control dataset (sorafenib monotherapy) and compared patients from 111 sites in 17 countries. Hence, details of the trials and participants are shown in Table 1.

As La Roche[14] and Hack et al[15] are currently ongoing, information for primary and secondary objectives are incomplete. It should be noted that while incomplete data were not assessed in this document, the authors of this manuscript fully intend to obtain updated data concerning related objectives in the future.

The Finn et al[12] and Lee et al[16] studies encompass a total of 724 patients and have been evaluated as follows. All clinical trials comprised a large sample of patients recruited from more than 310 sites across more than 20 countries. The countries included sites in North America (United States, Canada), Europe (United Kingdom, Germany, France, Italy, Poland, Spain, Russia, Czech Republic), and Asia-Pacific (China mainland, Japan, Republic of Korea, Taiwan, Australia, Hong Kong, Russia, Singapore, and New Zealand). The distribution of sites is shown in Figure 4. The specific patient profiles of La Roche[14] and Hack et al[15] have not yet been published (Table 1).

Open in New Tab   Full Size Figure   Download Figure

Figure 4 Distribution of study sites.

The clinical trials recruited adult volunteers of both genders, with locally advanced metastatic or unresectable HCC (or both). All trials used the American Association for the Study of Liver Disease criteria for histologic, cytologic, and clinical diagnostic confirmation. A documented hepatitis virological status was also required and a history of autoimmune disease was considered an exclusion criterion.

Eligible patients in the trials, who had not previously received systemic therapy for HCC, had an Eastern Cooperative Oncology Group (ECOG) performance score of either 0 or 1 and a Child-Pugh class of A or B (Supplementary Table 2). The studies attributed their exclusion of Child-Pugh class C to the increased risk of patient death due to related underlying cirrhosis. The patients with underlying cirrhosis were excluded from the study to avoid potentially confounding impact on the assessment of treatment-related antitumor efficacy. La Roche[14], Finn et al[12], and Lee et al[16] included patients who had HCC that was measurable as per the RECIST 1.1 (Supplementary Table 1). However, there was no specific mention of RECIST 1.1 in Hack et al[15].

Moreover, Hack et al[15], Finn et al[12], and Lee et al[16] employed BCLC staging (Supplementary Table 3). There was no specific mention of BCLC staging in La Roche[14].

Baseline characteristics of patients across the two completed trials and treatment modalities were adequately balanced (Table 2). Both studies had a median age range of about 63-years-old. Specifically, Finn et al[12] had a median age of 64 (interquartile range, [IQR], 56-71) and 66 (IQR, 59-71) years for its interventional (atezolizumab/bevacizumab) and control (sorafenib monotherapy), respectively. Whereas Lee et al[16] had a median age of 62 (IQR, 23-82) for Group A (atezolizumab/bevacizumab), 60 (IQR, 22-82) for Group F (atezolizumab/bevacizumab) and 63 (23-85) for the control (atezolizumab monotherapy), respectively. Both studies predominantly included the male sex (83%), Asian (62%) and Caucasian (30%) ethnicities, Child-Pugh class A (99%), and advanced BCLC (stage C) (84%) in the treatment and control groups. Both studies reported a higher prevalence of patients with extrahepatic spread, positivity for hepatitis B, and a history of alcohol use. Finn et al[12] included mostly ECOG 0 patients than ECOG 1, while the opposite was observed for Lee et al[16] Only about 35% of patients across the studies had alpha-fetoprotein > 400 ng per milliliter. Regarding PD-L1 status, more patients had tumor cell or immune cell ≥ 1% than any other classification, across treatment and control for both studies. Finn et al[12] showed the number of patients who previously experienced local therapy for HCC, and almost half of the patients had at least one treatment on both the atezolizumab/bevacizumab (48%, 161/336 patients) and sorafenib (52%, 85/165 patients) arms. Whereas Lee et al[16] did not show those patients who had prior local therapy for HCC.

There were no significant differences that were explicitly stated between interventional and control in the published manuscripts of Finn et al[12] and Lee et al[16]. The following baseline characteristics were evaluated for differences: Median age, sex, race, geographic region, Child-Pugh class, ECOG stage, BCLC stage, alpha-fetoprotein levels, macrovascular invasion, extrahepatic spread, hepatitis status, alcohol use, PD-L1 status, and prior local therapy for HCC. Moreover, regarding gastroesophageal varices (current/treated), there were no specific indications of statistically significant differences between the interventional and control groups in Finn et al[12] as well. On the other hand, Lee et al[16] did not report the prevalence of varices but stated that varices were managed when present, according to the standard of care.

Mortality rates: According to Finn et al[12], mortality occurred in 28.6% of patients (96/336) in the atezolizumab/bevacizumab group during a follow-up duration of 8.9 mo at the clinical data cut-off, and was significantly lower than that reported in the sorafenib group (39.4%; 65/165 patients) during a similar surveillance time of 8.1 mo (P < 0.001 by χ²) (Table 3).

The overall mortality reported by Lee et al[16] in the atezolizumab/bevacizumab group was significantly different (higher in Group A, 27% [16/104 patients] and zero in Group F, [0/60 patients]; P < 0.0001 by χ²), and was significantly lower than that in the atezolizumab group (31%; 18/59 patients) at a median follow-up of 12.4 mo (P < 0.0001 by χ²). Lee et al[16] also showed a 7% mortality (7/10 patients) in Group A and did not report deaths related to AEs in Group F.

Using epidemiological analyses, this review estimated the relative risk (RR) of death from the combination therapy vs the monotherapy to be 0.72 (Finn et al[12]) and 0.87 (Lee et al[16]), respectively. The calculated RR reduction was 0.28 and 0.13, respectively, for both studies. The attributable risk was 0.108 (Finn et al[12]) and 0.04 (Lee et al[16]), and the number needed to treat (NNT) 9.2 for atezolizumab/bevacizumab vs sorafenib (Finn et al[12]), whereas the NNT was 25 for atezolizumab/ bevacizumab vs atezolizumab alone (Lee et al[16]).

Hazard ratio for deaths and PFS: According to Finn et al[12], the stratified hazard ratio (HR) for death was 0.58 (95% confidence interval [CI]: 0.42-0.79; P < 0.001), whereas Lee et al[16] did not report on HR for death but rather estimated the HR for PFS stratified HR 0.55 (80%CI: 0.40-0.74; P = 0.011) in Group F (Table 3).

Overall/median survival: According to Finn et al[12], the estimated rates of OS at 6 and 12 mo were 84.8% (95%CI: 80.9-88.7) and 67.2% (95%CI: 61.3-73.1) in the atezolizumab/bevacizumab group, respectively. These results were significantly higher compared to 72.2% (95%CI: 65.1-79.4) and 54.6% (95%CI: 45.2-64.0) in the sorafenib group, respectively (P < 0.001) (Table 3). For Lee et al[16], median OS in Group A atezolizumab/bevacizumab was 17.1 mo (95%CI: 13.8 to not estimable), with only 55% (57 patients) still alive at the end of the surveillance. Median OS was not reached in atezolizumab/bevacizumab Group F. Additionally, neither Group F nor the atezolizumab group had estimable results as follows: (atezolizumab/bevacizumab: 95%CI: 8.3 mo to not estimable; atezolizumab group: 8.2 mo to not estimable).

Median PFS: Both studies reached significantly longer PFS in the atezolizumab/ bevacizumab dataset vs their respective monotherapy data set. In detail, median PFS was 6.8 mo (95%CI: 5.7-8.3) for patients treated with atezolizumab/bevacizumab compared to 4.3 mo (95%CI: 4.0-5.6] for patients treated with sorafenib alone in Finn et al[12] study (P < 0.001). On the other hand, for Lee et al[16], median PFS was 7.3 mo (95%CI: 5.4-9.9) and 5.6 mo (95%CI: 3.6-7.4) in Group A and Group F (atezolizumab/ bevacizumab), respectively, vs 3.4 mo (95%CI: 1.9-5.2) for atezolizumab monotherapy (P < 0.001) (Table 3).

Disease progression or death: At baseline, Finn et al[12] reported higher extrahepatic spread in 212/336 patients (63%) in the atezolizumab/bevacizumab group compared to 93/165 patients (56%) in the sorafenib group (P = 0.0005, by χ²) (Table 2). These percentages were significantly lower than those observed by Lee et al[16], where extrahepatic spread was shown in 91/104 (88%), 47/50 (78%), and 50/69 (85%) patients, for Groups A, F (atezolizumab/bevacizumab), and atezolizumab monotherapy, respectively (P = 0.004 by χ²). Moreover, in the study of Finn et al[12] atezolizumab/bevacizumab and sorafenib groups experienced similar disease progression (58.6% [97/336 patients] vs 66.1% [109/165 patients]; P = 0.10 by χ²) (Table 3). The stratified HR for progression or death was estimated to be 0.58 (95%CI: 0.42-0.79; P < 0.001).

Tumor response rate: A better overall tumor-confirmed objective response with combination therapy compared to the respective monotherapies in the control groups was reported. For Finn et al[12], a significantly higher overall tumor response was observed in 89/336 patients (27.3%; 95%CI: 22.5-32.5) with atezolizumab/ bevacizumab combination therapy than the one obtained with sorafenib in 19/165 patients (11.9%; 95%CI: 7.4-18.0) (P < 0.001 by χ²). Whereas Lee et al[16] detected a better overall tumor response in Group A of the combination therapy compared to Group F and the atezolizumab group, which were similar (36% [95%CI: 26-46] vs 20% [95%CI: 11-32] vs 17% [95%CI: 8-29]; P = 0.011). Further details concerning the other indices of tumor response, including the confirmed partial/complete/ongoing objective responses are shown in Table 3.

Disease control rate: The disease control rate was significantly higher in both trials for atezolizumab/bevacizumab combination therapy than sorafenib or atezolizumab monotherapies (Table 3). The estimates were 73.6% (240/336 patients) vs 55.3% (88/165 patients) for atezolizumab/bevacizumab and sorafenib (P < 0.001 by χ²), and 71% (74/104 patients), 67% (40/60 patients), and 49% (29/58 patients) for Group A, Group F and atezolizumab (P = 0.016 by χ²), respectively.

Overall AEs: Finn et al[12] estimated similar AEs that were contributed from any cause in 98.2% (323/336 patients) and 98.7% (154/165 patients) for the atezolizumab/ bevacizumab vs sorafenib monotherapy groups (P = 0.17 by χ²), respectively. Likewise, according to Lee et al[16], 96% (100/104 patients), 95% (57/60 patients), and 90% (52/58 patients) for Groups A/F (atezolizumab/bevacizumab) and atezolizumab (P = 0.13 by χ²), respectively (Table 3).

Treatment-related serious AEs: Finn et al[12] reported higher treatment-related AEs with atezolizumab/bevacizumab compared to sorafenib monotherapy (38% [125/336 patients] vs 30.8% [48/165 patients]; P < 0.0001 by χ²). Lee et al[16] recorded 24% (25/104 patients), 12% (7/60 patients), and 3% (2/58 patients) for Groups A/F (atezolizumab/bevacizumab) and atezolizumab (P < 0.001 by Fisher’s Exact Test), respectively (Table 3).

Grade 3-5 AEs: Details of the treatment-related deaths and severe AEs as well as other indices of safety are shown in Table 3. Finn et al[12] and Lee et al[16] reported Grade 3-5 AEs in 10% and 15% of participants, respectively. In the Finn et al[12] study, hypertension occurred in 15.2% and 12.2% for the combination therapy and sorafenib monotherapy groups, and 14%, 5%, and 1% in Groups A/F (atezolizumab/ bevacizumab) and atezolizumab, respectively (Lee et al[16]). Furthermore, Proteinuria occurred in 3% and 0.6% in the combination therapy and sorafenib monotherapy groups, respectively, whereas in 7%, 5%, and 0% for Groups A/F (atezolizumab/ bevacizumab) and atezolizumab groups, respectively. Other Grade 3-5 AEs that were reported included abdominal pain, fatigue, rashes, pyrexia, and palmar-plantar erythrodysesthesia syndrome (Table 3). Finn et al[12] registered fewer Grade 5 AEs in the combination therapy compared to sorafenib monotherapy (4.6% vs 5.8%), whereas Lee et al[16] study registered 0% of such events.

The RoB assessment for Finn et al[12] and Lee et al[16] trials resulted in high quality with a low RoB. However, in the two ongoing studies (La Roche[14] and Hack et al[15]), the RoB was not evaluated given that incomplete information existed (Figures 1 and 2).

Both completed trials (Finn et al[12] and Lee et al[16]) showed adequate randomization with a low RoB arising from the randomization process that was performed through an interactive voice-response or Web-response system in permuted blocks. There was also fair allocation concealment. For Lee et al[16], an independent statistician was responsible for generating the randomization sequence, which was subsequently stored within the interactive voice systems.

In both completed studies (Finn et al[12] and Lee et al[16]), open-label trials were implemented, and consequently had neither blinding nor masking of interventions. Lee et al[16] described 26 participants who deviated from the initially assigned atezolizumab/bevacizumab to atezolizumab monotherapy without describing post-crossover efficacy and safety results. Moreover, Finn et al[12] found it less cumbersome to not administer intravenous placebo infusions. Hence, due to the lack of blinding or masking and because of deviations from intended interventions, the two completed studies were estimated to have some RoB concerns.

All RCT results showed a low risk of attrition bias from the missing outcome data. Both studies also had appropriate measurements of survival outcomes (Finn et al[12]). For example, both used PFS and objective tumor response with the RECIST 1.1 (Supplementary Table 1) as well as the HCC-specific mRECIST by investigator assessment and independent faculty review. Thus, they displayed a low risk of measurement bias as well as lower selective reporting bias.

Despite the use of the current standard therapy, the prognosis of unresectable hepatocellular carcinoma (HCC) patients is poor, with median survival times of 40 mo in intermediate HCC (Barcelona Clinic Liver Cancer [BCLC] stage B) and 6–8 mo in advanced HCC (BCLC stage C). Although patients with early-stage HCC are usually suitable for therapies with curative intention, up to 70% of patients may manifest disease relapses at 5-year surveillance. Moreover, no treatment has been demonstrated to be useful in the adjuvant setting.

This systematic review described the evidence for atezolizumab/bevacizumab combination therapy vs sorafenib or atezolizumab monotherapies in improving survival outcomes and reducing disease progression in patients with unresectable HCC.

To evaluate the efficacy and safety of atezolizumab/bevacizumab vs sorafenib or atezolizumab alone, in patients with unresectable HCC with non-decompensated liver disease.

A comprehensive literature review of published articles was conducted to identify studies that met our inclusion criteria using relevant mesh terms. This systematic review was performed according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines and we assessed for risk of bias using the Cochrane ROB tool and Sevis tool to create the traffic light plots and summary plots.

There was an improvement in overall tumor response, reduction of disease progression, and longer progression-free survival in the atezolizumab/bevacizumab group compared to the monotherapy of either sorafenib or atezolizumab.

This study confirms that combination treatment of atezolizumab plus bevacizumab could be a promising alternative to the standard of care sorafenib as a first-line treatment in patients with unresectable HCC and non-decompensated liver disease.

Given the scarcity of randomized controlled trials specifically focusing on this therapeutic strategy, further research is needed to strengthen the current evidence. Two completed clinical trials were analyzed in this research; however, this review will be updated upon the completion of two ongoing trials. Moreover, further evaluation regarding the cost-effectiveness of combination therapy vs monotherapy is still needed valuable information.