Efficacy and Safety Associated With Immune Checkpoint Inhibitors in Unresectable Hepatocellular Carcinoma

Key Points Question What are the efficacy and safety associated with immune checkpoint inhibitors (ICIs) vs standard therapies in unresectable hepatocellular carcinoma (HCC)? Findings In a meta-analysis of 3 randomized clinical trials totaling 1657 patients, ICIs were associated with significantly improved overall survival, progression-free survival, and overall response rate compared with standard therapies. In addition, the rate of grade 3 or 4 treatment-related adverse events was lower with ICIs than with sorafenib. Meaning These findings suggest that ICIs should be the new standard of care in systemic therapy of unresectable HCC.


Introduction
Hepatocellular carcinoma (HCC) is one of the most lethal malignant neoplasms, ranking as the fourth most common cause of cancer-related death in the world. 1 Approximately 20% of patients have advanced disease at presentation, which portends a poor prognosis, with an estimated 5-year overall survival (OS) rate of 2%. 2 The cornerstone of the treatment of unresectable or metastatic HCC is systemic therapy. Since 2008, sorafenib has been the standard of care, based on the pivotal phase 3 SHARP trial, which demonstrated an improvement in the median OS vs placebo (10.7 months vs 7.9 months; hazard ratio [HR], 0.69, 95% CI, 0.55 to 0.87; P < .001). 3 In the past few years, other antiangiogenic agents, either tyrosine-kinase inhibitors or monoclonal antibodies, have been found to be effective for both patients who are treatment-naive (lenvatinib, 4 donafenib 5 ) and those who are resistant to sorafenibbased treatment (cabozantinib, 6 regorafenib, 7 apatinib, 8 and ramucirumab 9 ). Nevertheless, none has shown superiority over sorafenib.
Immune checkpoint inhibitors (ICIs) have ushered in a new era in cancer therapy, but their efficacy in HCC is uncertain. Single-arm phase 2 studies with patients who are resistant to sorafenib suggested clinical activity of nivolumab, 10 pembrolizumab, 11 and the combination of ipilimumab plus nivolumab, 12 all of which have been approved by the US Food and Drug Administration. When evaluated as monotherapy in randomized clinical trials (RCTs), anti-programmed cell death protein 1 (PD-1) and anti-programmed cell death 1 ligand 1 (PD-L1) inhibitors did not demonstrate superiority compared with standard care. Nivolumab and sorafenib yielded similar OS rates in previously untreated patients (16.4 months vs 14.7 months; HR, 0.85; 95% CI, 0.72 to 1.02; P = .08), 13 and pembrolizumab did not meet the co-primary end point of OS in a placebo-controlled trial (13.9 months vs 10.6 months for placebo; HR, 0.78; 95% CI, 0.61 to 0.99; P = .02). 14 The combination of atezolizumab, an anti-PD-L1 inhibitor, with bevacizumab was compared with sorafenib as first-line treatment of unresectable HCC in the phase 3 IMbrave150 trial. 15 The combination was superior to sorafenib in OS and progression-free survival (PFS), the co-primary end points (12- Owing to the conflicting results of the studies with immunotherapy in HCC, we performed, to our knowledge, the first meta-analysis of RCTs addressing the overall outcomes associated with ICIs compared with the standard of care in unresectable HCC.

Methods
This meta-analysis did nor require institutional review board approval because the sources of the analyzed data are public and the analysis will not make the data individually identifiable. This review was conducted according to the Preferred Reporting Items for Systematic Reviews and Metaanalyses (PRISMA) reporting guideline and is registered in PROSPERO (CRD42020162599). ORs, 95% CIs, and P values). The HRs of time-to-event variables (OS and PFS) were directly extracted from the original studies or were estimated indirectly using the reported number of events and the corresponding P value for the log-rank statistics. Data were extracted using an assessment form that was designed specifically for the topic of this study.

Statistical Analysis
Outcomes of interest included OS, PFS, ORR, and grade 3-4 TRAEs. A summary HR of OS and PFS was calculated using a 95% CI by a random-effects model. The likelihood of ICIs being associated with ORR and TRAEs was expressed by OR and 95% CI using a random-effects model and presented in forest plots. The treatment outcome for each study was expressed as a ratio of the ICIs group vs the standard care group. The quality of the trials was evaluated by the Cochrane risk-of-bias tool for randomized trials. 17 Statistical heterogeneity in the results of the trials was assessed by the χ 2 test and was expressed by the I 2 index, as described by Higgins and colleagues. 18 Publication bias was evaluated with Egger test. 19 Statistical analysis of summary data was performed with RevMan software version 5.3 (Cochrane). Hypotheses tests were 2-sided and P < .05 was deemed as statistically significant. Crossover was also allowed.   (Figure 3). Significant heterogeneity was found among the studies in the analysis of PFS (I 2 = 82%).     the placebo group and 18% in the pembrolizumab arm group (P < .001). 14 Six patients (2.2%) who received the ICI reached complete response.

JAMA Network Open | Oncology
Likewise, in CheckMate-459, the ORR was higher in the nivolumab group (15%) than in the sorafenib group (7%) (HR, 2.41; 95% CI, 1.48-3.92). 13 Patients with PD-L1 expression at 1% or greater had a higher ORR than their patients without PD-L1 expression: 28% among the nivolumab group vs 9% among the sorafenib group of patients with PD-L1 expression, compared with 12% among the nivolumab group vs 7% the sorafenib group among patients without PD-L1 expression.
The ORR was also higher in patients treated with atezolizumab plus bevacizumab than in those treated with sorafenib: 27% vs 12% (P < .001). 15, 16 The difference in ORR between groups was higher when response was evaluated per HCC-specific modified RECIST: 33% vs 13% (P < .001). Based on RECIST, 5.5% of the patients who received combination therapy reached complete response, compared to 10.2% based on HCC-specific modified RECIST.
No significant heterogeneity was found among the studies (I 2 = 0%).

TRAEs
The rate of grade 3 or 4 TRAEs was lower in patients who received ICIs than in those who received sorafenib. In CheckMate-459, the rate of grade 3 or 4 TRAEs was 22% in patients who received nivolumab monotherapy and 49% in those treated with sorafenib. 13  In KEYNOTE-240, the rate of grade 3 or 4 TRAEs was 8% in the placebo group and 19% in the pembrolizumab group. 14 Aspartate aminotransferase increase(13% of patients) and hyperbilirubinemia (8% of patients) were the most frequent grade 3 or 4 AEs in the pembrolizumab The diamond indicates best estimate of the true (pooled) outcome (with width indicating 95% CI); ICI, immune checkpoint inhibitor; OR, odds ratio. Risks of bias are present (+) or absent (-), with A indicating random sequence generation (selection bias); B, allocation concealment (selection bias); C, blinding of participants and personnel (performance bias); D, blinding of outcome assessment (selection bias); E, incomplete outcome data (attrition bias); F, Selective reporting (reporting bias); and G, other bias. group. KEYNOTE-240 data were not included in the overall analysis of TRAEs owing to the comparison with placebo.

Discussion
In this meta-analysis of 3 RCTs evaluating 1657 patients with unresectable HCC, in both the first-line and second-line setting, ICIs were shown to be associated with significantly better ORR, PFS, and OS, which had an overall reduction of 25% in the relative risk of death, compared with standard care (sorafenib or placebo). In addition, ICIs had a safer toxicity profile, with lower rates of grade 3 or 4 TRAEs.
The main driver of the overall benefit associated with ICIs found in this meta-analysis was the data obtained from the IMbrave150 study, which evaluated the combination of anti-PD-L1 plus anti-

Limitations
Our meta-analysis has some limitations. The strict inclusion criteria of our systematic review allowed for the selection of only 3 studies. Despite being an RCT addressing the role of ICIs, the GO30140 trial 21 did not meet the criteria for being included in our meta-analysis because the comparator group (atezolizumab as monotherapy in the first-line setting) is not standard care. There is no minimum number of studies to be analyzed in a meta-analysis. However, meta-analyses with few studies or with small sample sizes compromise the precision of the tests that measure heterogeneity, such as τ 2 , χ 2 , and I 2 . 25 On the other hand, since ours is the first meta-analysis, to our knowledge, to analyze the efficacy and safety of ICIs compared with standard care in unresectable HCC, which is an innovative field, it is expected to find a low number of RCTs.
The diversity of the RCTs analyzed in our meta-analysis prompted us to choose the randomeffects model, since we do not expect a fixed effect in clinical trials with different control (sorafenib or placebo) and experimental (ICIs alone or ICI plus bevacizumab) groups. The inclusion of the IMbrave150 was the main source of between-study heterogeneity, as can be seen through the analysis of the funnel plot (eFigure 1 in the Supplement), which helps us understand the superior efficacy associated with the combination therapy compared with ICIs alone.

Conclusions
In this meta-analysis, dual therapy was evaluated in patients who were treatment naive with liver function scale Child-Pugh A, ECOG performance status 0 to 1, and treated gastroesophageal varices.
The efficacy of dual therapy in patients previously treated with VEGF inhibitors, like sorafenib or lenvatinib, is not known. Therefore, atezolizumab plus bevacizumab should be preferentially used in the first-line setting in eligible patients. The best choice after unsuccessful dual therapy is not clear. In