In 2008, sorafenib became the first systemic treatment to demonstrate a significant survival benefit in patients with advanced HCC. Sorafenib is a multikinase inhibitor (MKI) that reduces both HCC cell proliferation and angiogenesis by targeting a broad spectrum of protein kinases, including VEGFR, PDGFR, c-KIT and RAF. Two phase 3 trials (SHARP and ASIA-PACIFIC) evaluating sorafenib versus placebo showed a significant increase in median OS in patients with preserved liver function (Child-Pugh A) and advanced HCC (BCLC C or BCLC B with tumor progression after locoregional therapy and naive of systemic therapy) [18, 19]. Diarrhea, hand-foot syndrome, and fatigue were the most frequent adverse events, causing approximately 8% of grade 3–4 events each. Exploratory subgroup analyses of the SHARP study showed that sorafenib increased OS and disease control rate (DCR) relative to placebo regardless of etiology, initial tumor volume, ECOG PS, and previous treatments [23]. The ASIA-PACIFIC study was a mirror clinical trial of the SHARP study in a population of Asian patients [19]. The shorter OS (6.5 versus 4.2 months) observed in the ASIA-PACIFIC study may be explained by the higher frequency of poor prognostic factors in the patients included, with large tumor volumes, high prevalence of HBV infection, and altered ECOG PS [24].Tab1

Summary of positive phase 3 clinical trials of angiogenic inhibitors in patients with advanced hepatocellular carcinoma (HCC)

Following these two pivotal trials, sorafenib obtained worldwide approval and became the standard first-line treatment for advanced HCC. No predictive markers of response had been identified in the translational studies derived from the SHARP study [25]. Since then, several predictive biomarkers have been proposed, including amplifications of fibroblast growth factor 3/4 or VEGF-A, polymorphisms of VEGF-A and VEGF-C, or tissue expression of pERK or VEGFR-2 [17] and imaging criteria [26]. However, none of these biomarkers has been validated for clinical use with antiangiogenics. Combinations of sorafenib with erlotinib [27], doxorubicin [28] or transarterial chemoembolization [29] has been explored in randomized trials, without improvement of OS or progression-free survival (PFS) [27, 28]. The reasons for these failures were limiting toxicities and the absence of patient selection based on molecular markers.

Since the approval of sorafenib, new candidate drugs failed to demonstrate their efficacy as first-line therapies versus sorafenib: they included sunitinib [30], brivanib [31] and linifanib [32]). In 2018, a non-inferiority trial evaluating lenvatinib versus sorafenib was published [20]. Lenvatinib is an angiogenesis inhibitor targeting multiple tyrosine kinase receptors, including VEGF receptors 1 to 3, FGF receptors 1 to 4, PDGF receptor, RET and KIT. This non-inferiority trial in patients with BCLC B or C HCC and Child-Pugh A showed similar efficacy of lenvatinib and sorafenib in terms of median OS (13.6 months versus 12.3 months, respectively), with improved median PFS (7.4 months versus 3.7 months, respectively) and objective response rate (ORR) according to modified RECIST criteria (24% versus 9%, respectively). In addition, the toxicity profile of lenvatinib was more favorable than that of sorafenib (lower incidence of fatigue, diarrhea and hand-foot syndromes). Together, these results led to lenvatinib approval by the Food and Drug Administration.

Several drugs have failed versus placebo in second-line treatment trials after failure of or intolerance to sorafenib, including brivanib [33] or everolimus [34]. In 2016, the RESORCE phase 3 trial showed that regorafenib, a sorafenib derivative whose structure differs by the addition of a fluorine atom, significantly improved median OS by 3 months, as compared to placebo, as second-line treatment after failure of sorafenib to prevent disease progression (hazard ratio (HR) = 0.63; p < 0.0001, 33). The most frequent grade 3–4 adverse events on regorafenib were hypertension (15%), hand-foot syndrome (13%), fatigue (9%) and diarrhea (3%).

The CELESTIAL phase 3 trial subsequently evaluated cabozantinib, an MKI targeting VEGFR 1 to 3, c-MET and AXL, all involved in sorafenib resistance, as second- or third-line therapy in patients previously treated with sorafenib [22]. The study showed a 2-month benefit for median OS in favor of cabozantinib, as compared to placebo (HR = 0.76; p = 0.005). The most common serious adverse events were hand-foot syndrome (17%), hypertension (16%), liver function disorders (12%), fatigue (10%) and diarrhea (10%).

Finally, the REACH-2 phase 3 trial evaluated ramucirumab, an anti-VEGFR-2 monoclonal antibody in patients with advanced HCC pre-treated with sorafenib and with a high alpha-fetoprotein (AFP) level (≥ 400 ng/mL) [35]. This study was designed following the results of the REACH-1 trial; in this phase 2 study, the primary objective was not met in the overall HCC patient population (unselected for AFP), but a benefit was suggested in the subgroup of patients with elevated AFP at the initiation of treatment [36]. REACH-2 showed a modest but significant survival benefit, as compared to placebo: 8.5 months versus 7.3 months, respectively (HR = 0.71; p = 0.019). Thus, ramucirumab is the first biomarker-guided therapy to show efficacy in patients with HCC. Due to the lack of liver metabolism, ramucirumab had a milder toxicity profile, as compared to MKI, inducing mainly hypertension (12% of grade ≥ 3) and hyponatremia (5.6% of grade ≥ 3).

In summary, cabozantinib, regorafenib, and ramucirumab have recently become new therapeutic options in patients with advanced HCC previously treated with sorafenib. Interestingly, the efficacy of these three drugs is within the same range, with a 25% reduction of the risk of death, albeit in non-comparable patient populations. Indeed, REACH-2 included patients with a poorer prognosis than the other trials, due to the selection based on high AFP levels, which may explain the lower survival rate observed in this study. Finally, no biomarker has been identified to guide the choice between these three angiogenesis inhibitors in clinical practice. Furthermore, whether a treatment sequence effect similar to angiogenic blockade beyond progression in colorectal cancer (with angiogenesis being continuously blocked) exists in HCC remains unknown.

Two trials evaluated the safety and efficacy of tremelimumab, a monoclonal antibody inhibiting CTLA-4, alone or in combination with ablation therapy (chemoembolization or radiofrequency), respectively [46, 47]. The first study was a phase 2 trial evaluating tremelimumab in 20 patients with advanced HCC and chronic HCV infection [46]. These patients were heavily pretreated, with non-resectable or metastatic HCC (BCLC C: 57%, portal vein invasion: 28%), high tumor volume, and frequent altered liver function (Child-Pugh B: 43%). Tremelimumab resulted in a partial response rate of 17.6% and a DCR of 76.4%. The second study combined tremelimumab with ablation therapy in patients with advanced HCC in order to induce tumor necrosis, thereby promoting the release of tumor antigens to increase the efficacy of anti-CTLA-4 [47]. Thirty-two patients were included, but only 19 were radiologically evaluable. Patients with radiological benefit (26.3% ORR, 63% DCR) had an increase in intratumoral CD8+ T cells on biopsies performed 6 weeks after the start of treatment. In both studies, antitumoral activity and antiviral activity against HCV (decreased viral load) were observed and tolerance was acceptable without dose-limiting toxicities.Tab3

Summary of clinical trials of immune therapies (single agent or combination with angiogenics inhibitors) in patients with advanced hepatocellular carcinoma (HCC)

The results of these studies were encouraging but should be analyzed with caution, due to the limited sample size with a large proportion of patients not reaching the first radiological evaluation due to early clinical progression. In addition, industrial developments of anti-CTLA-4 now focuses on associations with anti-PD-1/programmed cell death-ligand 1 (PD-L1) antibodies in advanced stage HCCs (NCT03298451, [59]) or in the neo-adjuvant setting (NCT03510871). These combination studies rapidly emerged in the context of data showing the activity of anti-PD-1/PD-L1 monotherapy in advanced HCC, supported by a more favorable toxicity profile than anti-CTLA-4.

While PD-1 receptor is mainly expressed by activated lymphocytes, PD-L1 ligand has been identified at the surface of tumor cells, as well as in the peritumoral stroma of HCCs and its presence is a poor prognostic factor [60]. In a cohort of 217 resected HCCs [61], PD-L1 expression within HCC tumors was found in about 75% of cases, with a wide range of intensity [61]. It has been suggested that PD-L1 expression should be assessed: (i) on tumor cells (threshold of 1%), and (ii) on immune cell clusters infiltrating the tumor [61]. High PD-L1 expression was associated with markers of tumor aggressiveness (high AFP levels, satellite nodules, poor differentiation, macro- and micro-vascular invasion). These observations support the potential therapeutic interest of blocking the PD-1/PD-L1 axis in HCC.

The first results of studies evaluating anti-PD-1/PD-L1 monotherapies as post-sorafenib second-line treatment in advanced HCC appeared promising for durvalumab (anti-PD-L1), pembrolizumab (anti-PD-1), and cemiplimab (anti-PD-1), while nivolumab (anti-PD-1) could be used first- or second-line. The toxicity profiles were similar to those previously described in the literature for anti-PD-1/PD-L1 in other tumor types and did not differ between molecules. No cases of HBV or HCV reactivation have been observed. Efficacy results are summarized in Table 3. However, a recent communication reported that the KEYNOTE-240 phase 3 trial, comparing pembrolizumab to placebo plus best supportive care in 413 patients previously treated with sorafenib, did not meet its co-primary endpoints of OS and PFS [49]. Indeed, despite an improvement in favor of pembrolizumab, these differences did not achieve statistical significance per the prespecified statistical plan. ORR was 16.9% for pembrolizumab vs 2.2% for placebo (P = 0.00001) and responses were durable (median duration of response: 13.8 months). Reasons for this failure could be previous sorafenib treatment, the progressive nature of the tumors in these patients, in whom the immune reserve was depleted and/or subsequent treatments. The KEYNOTE-394 trial is an ongoing mirror clinical trial of KEYNOTE-240 in Asian patients (NCT03062358).

In a multicohort study with durvalumab [52], the clinical benefit was greater in patients with chronic HCV infection, but this observation was limited by the small number of patients.

In a pre-specified exploratory analysis of the phase 2 study with pembrolizumab [48], the authors did not find any criteria predictive of the objective response (age, viral or non-viral etiology, AFP levels, BCLC stage, macrovascular invasion, extrahepatic metastases). Another pre-specified analysis evaluated the association between PD-L1 and the radiological response by proposing an overall expression score combining PD-L1 expression by the immune microenvironment and tumor cells (CPS score). The proposed score was defined by the number of PD-L1+ cells (≥1%) (tumor cells, lymphocytes and macrophages) divided by the total number of tumor cells. PD-L1 positivity in tumor cells did not predict radiological response, although a trend was observed (p = 0.08), or PFS (p = 0.096). In contrast, 42% of patients were positive for PD-L1 according to the CPS score and there was a significant association with ORR (32% versus 20% in CPS-positive versus negative patients, respectively, p = 0.021) and PFS (p = 0.026). Thus, considering PD-L1 expression both in the tumor and immune cell compartments improved the prediction of the response to anti-PD-1 therapy in HCC. In addition, a prospective study in 31 ICI-treated patients showed that WNT/β-catenin pathway alterations are associated with a poorer DCR (0 versus 53%) and OS (9.1 versus 15.2 months) [41]. These results support the establishment of composite scores combining PD-1 expression and molecular alterations in order to properly predict ICI response.

Finally, the CheckMate-040 phase 1/2 trial evaluated nivolumab in 3 cohorts of patients with advanced HCC: HBV-infected, HCV-infected, and non-infected, respectively [50]. This study included both treatment-naive and previously sorafenib-treated patients. No significant differences in the response according to treatment exposure and/or viral infection were observed. In line with the pembrolizumab study, PD-L1+ tumors (≥1% on tumoral cells) showed a non-significant trend for higher ORR as compared to PD-L1- tumors (26 and 19%, respectively). Several studies evaluating nivolumab in patients with advanced HCC are ongoing, including the pivotal CheckMate 459 phase 3 study comparing nivolumab with sorafenib in treatment-naive patients with advanced HCC (NCT02576509). Other studies are currently evaluating nivolumab in combination with chemoembolization (NCT03572582), radioembolization (NCT03033446), TGF-β inhibitors (NCT02423343), indoleamine dioxygenase inhibitors (NCT03695250) in patients with advanced HCC, or in neo-adjuvant and adjuvant settings (NCT03630640, NCT03383458). Other strategies may include the combination of Anti-PD-1 with cytotoxic drugs as recently reported in a trial investigating camrelizumab + FOLFOX4 or GEMOX and showing in 34 treatment naive HCC patients an ORR of 26.5% and a mPFS of 5.5 months [58]. The additive/synergistic effect of immune checkpoints inhibition and cytotoxic agents remains to be established.

First results of immune checkpoint inhibitors combination has been recently communicated. The Checkmate-040 trial included a 3-arm randomized phase II investigating nivolumab + ipilimumab at different dose in sorafenib pretreated and Anti-PD-1 naive patients [53]. One hundred and forty-eighth number patients were randomized and 37% had high grade treatment related adverse event leading in 5% to discontinuation. Overall, the ORR was 31% with 5% (N = 7) of complete response. In the 3 arms the DCR ranged between 43 to 54% with higher benefit in the Nivolumab 1 mg/kg + ipilimumab 3 mg/kg (4 doses) Q3W. This arm displayed an interesting mOS of 23 months compared to the other arms (12 and 13 months). The addition of an Anti-CTLA-4 to the anti-PD-1 seems to improve the response rate but at the price of increased toxicity. Biomarker studies are pending. Current development of such combinations focuses on the neoadjuvant setting (NCT03510871, NCT03222076) or as adjuvant treatment after stereotatic radiotherapy (NCT03203304) or TACE (NCT03638141).