3.2.1. Eligibility in the Real-World Clinical Practice
The target population of the concluded ICI studies on HCC includes patients with an advanced-stage (BCLC-C) cancer or an intermediate-stage (BCLC-B) neoplasm not amenable to surgery or locoregional procedures. However, in the real-world clinical practice, not all of these patients are eligible to receive ICIs as monotherapy or combination, as they may have specific contraindications.
Recently, Giannini et al. [
36] explored the Italian Liver Cancer (ITA.LI.CA) database to assess the theoretical applicability of ICIs in field-practice conditions according to the criteria utilised for patient enrolment in clinical trials. The ITA.LI.CA database includes patients with newly diagnosed or recurrent HCC managed in a large number of Italian centres with different levels of specific expertise. This database, due to its heterogeneity in terms of tumour stage, the severity of underlying liver disease and therapeutic approaches, predicts the analysis of the potential utilisation of these drugs. Amongst the 2483 patients (distributed across different BCLC stages), 525 (21.1%) and 268 (10.8%) were theoretically eligible for nivolumab and pembrolizumab, respectively, as frontline therapy [
36]. Considering only the 1514 patients in the advanced-stage or the intermediate-stage, but unresponsive to locoregional procedures, the rate of eligibility raised to 34.7% for nivolumab and 17.7 for pembrolizumab. Child-Pugh class >A (
n = 601), uncontrolled ascites (
n = 380), performance status >1 (
n = 343), active alcohol intake (
n = 323), thrombocytopenia (
n = 299), hepatic encephalopathy (
n = 155), aminotransferase levels >5× (
n = 123) and concurrent autoimmune diseases (
n = 57) were amongst the main limitations to the potential use of ICIs in the front-line setting (34). The eligibility in the second-line setting was even lower, with 5.4% and 8.0% of patients amenable for nivolumab and pembrolizumab, respectively [
36].
We also analysed the same database to verify the applicability of other therapeutic options based on the conventional inclusion/exclusion criteria adopted for the clinical trials of intravenous anti-VEGF agents and TKIs. Overall, 52 additional patients had clinically significant heart disease, ten patients had uncontrolled hypertension, fifteen had chronic non-healing skin ulcerations, and three had non-liver-related coagulative abnormalities increasing the risk of bleeding. Consequently, the rate of patients eligible to the atezolizumab-bevacizumab combination (and, by extension, to a combination of ICIs and TKIs) was 17.9% in the whole HCC population and 29.4% in HCC patients with an advanced HCC or an intermediate tumour not eligible for surgery or locoregional procedures (unpublished data).
Therefore, analyses of a large unselected cohort of HCC patients generated by the real-world clinical practice would indicate that, among potential candidates to immunotherapy, no more than one-third of them are amenable to ICIs as a frontline approach, and this percentage further decreases considering combination therapies with anti-VEGF or TKI agents.
3.2.2. Safety
The inhibition of physiological immune checkpoints may be associated with immune-related AEs (irAEs) targeting the skin, gut, thyroid, adrenal glands, lung and liver [
15]. For monotherapies with PD-1 /PD-L1 inhibitors, the risk of irAEs is dose-independent, with an incidence of 27% for all Grades, and 6% for Grade ≥3 [
37]. Instead, with CTLA-4 inhibitors, the overall incidence of irAEs is dose-dependent and remarkably higher, reaching 72% for all Grades and 24% for Grade ≥3 [
38]. Generally, these events are easily manageable, delaying the subsequent scheduled dose and using corticosteroids in severe or unresponsive cases. A recent meta-analysis reports 42 (0.6%) cases of fatal irAEs among 6528 patients treated with ICIs, with ipilimumab-induced colitis being the leading cause of death [
39]. Furthermore, a minimal number of fatal outcomes due to immune-related pneumonitis [
40] and myocarditis [
41] have been reported.
Despite this acceptable safety profile of ICIs, a justifiable concern on the expected risk/benefit ratio accompanies their use in cirrhotic patients, for different reasons. First, immune-related hepatitis can precipitate an acute-on-chronic liver failure with a high risk of severe liver decompensation and death. Second, the use of corticosteroids to treat severe irAEs is particularly problematic in cirrhosis, especially in terms of increased risk of infections and ascitic decompensation. Third, cirrhosis is known to disrupt the liver’s homeostatic immune function, provoking per se a condition, including both systemic inflammation and immunodeficiency [
42]. Indeed, a study enrolling patients treated with ICIs for different cancers seemed to suggest that hepatic AES were related to a worse prognosis [
43].
Sangro et al. [
14] reported a rate of aminotransferase increase close to 50% in their pivotal trial with tremelimumab. However, these alterations were transient, never associated with liver function impairment, and resolved without corticosteroids [
14]. Fortunately, even the safety reports from subsequent clinical trials testing ICIs in HCC patients were reassuring [
15,
17,
18]. Moreover, the large CheckMate459 and KEYNOTE-240 trials confirmed that the safety profile of ICIs was consistent with that reported in previous studies for melanoma and lung cancer [
11,
20], suggesting that cirrhotic patients have not an increased risk of liver irAEs. The proportion of cases who needed corticosteroid treatment was 6% for durvalumab (18) and 20% for the nivolumab-ipilimumab combination [
30]. It is worth noting than the risk of relevant AEs in HCC studies increased when ICIs were tested in combinations with other agents (
Table 6).
Overall, the comforting data gave support to the use of nivolumab in Child-Pugh B patients. In these particularly frail subjects, treatment-related hepatic AEs were reported in only 4 out of 49 patients, leading to treatment discontinuation in 2 patients [
44].
Nevertheless, both HCC and liver cirrhosis act as confounders for other types of irAEs. Cutaneous toxicities, for instance, are the most common AE reported in clinical trials on ICIs (28). The interpretation of skin toxicities may be difficult when ICIs are prescribed in combination with TKIs as this class of drugs has this potential AE [
19] and the hand-foot skin reaction, typical of TKIs, has not been reported for ICIs. It is also pertinent to note that the TKI-related skin rash usually appears during the first week of treatment and quickly disappears after drug discontinuation due to the short half-life of most TKIs [
13]. In contrast, ICI-related skin toxicities appear later, and in the absence of steroid therapy, requires a long-lasting interruption of treatment to resolve [
39].
Diarrhoea is another common irAE which can difficultly be ascribed to a precise cause. Indeed, cirrhotic patients are often medicated with osmotic laxatives (the dosage of which must be accurately tailored) to prevent hepatic encephalopathy, and in patients treated with both TKIs and ICIs, the same considerations made for dermatological AEs apply to diarrhoea. When diarrhoea is associated with abdominal pain and signs of colonic inflammation, immune-related colitis should be suspected and immediately managed, as it still represents the most frequent cause of death due to ICIs [
39]. Although the diagnosis of immune-related colitis is frequently made based on clinical signs and symptoms, colonoscopy is the diagnostic gold standard and assesses its severity and prognosis [
45].
Immune-related endocrinopathies also pose challenges. Thyroid function is often monitored in patients with advanced HCC as a result of their inclusion in clinical trials or because of a concurrent treatment with TKIs, which can provoke thyroid dysfunctions [
2,
13]. As a consequence, immune-related hypo- and hyperthyroidism are usually detected in a pre-clinical phase. On the contrary, the identification of adrenal failure can be problematic as cirrhotic patients have an intrinsic tendency to hypotension due to the hemodynamic peculiarities of advanced liver disease and slight hyponatremia due to haemodilution and use of diuretics.
Other irAEs are unrelated to the underlying liver disease or concurrent therapies, but require immediate attention as a late diagnosis might cause a dismal prognosis. For instance, the appearance of cough, fever and shortness of breath should prompt immediate investigations to detect an immune-related pneumonitis, which requires an early treatment since acute respiratory failure may rapidly ensue [
40].
3.2.3. Unpredictable Efficacy, the Need for Biomarkers
All trials of anti-PD-1/PD-L1 for HCC consistently identified a subgroup of 15–20% of patients obtaining an objective response (with an increase of this proportion up to 36% using combination regimes) [
27]. These patients also obtained the most important benefit in terms of OS. Therefore, the identification of predictors of response would have a crucial role in optimising the cost-effectiveness of therapy with ICIs. At the same time, predictors of futility might channel patients to other treatments (TKIs, for instance), avoiding the cost and risk of pointless irAES.
Historically, immunostaining of tumour specimen with anti-PD-L1 antibodies was the first approach used to predict the response to ICIs. However, in most studies, PD-L1 expression was not predictive of response. When it was claimed as predictive, different thresholds of PD-L1 were identified in different tumours (from 1% to 50%) [
46]. Moreover, the determination of PD-L1 expression suffers from the intrinsic variability of immunohistochemistry [
47]. Moreover, the biological characteristics of malignancy, including intratumoral heterogeneity and tumour microenvironment, play an essential role in reducing the reliability of this technique [
48]. Notably, Bensch et al. [
49] performed the first-in-human study assessing PD-L1 expression by radionuclide imaging (89 Zr-atezolizumab), finding a good correlation between increased tumour uptake and response to anti-PD1 therapy. However, the reliability of this intriguing non-invasive way to detect PD-L1 expression that avoids sample biases needs further validation.
The role of PD-L1 expression has been evaluated in a patient subgroup of the CheckMate-459 study testing nivolumab vs sorafenib. In the nivolumab arm, the OS did not differ between patients with low and high PD-L1 expression. At the same time, surprisingly, the OS was different in the sorafenib arm, being about 14 months in patients with low PD-L1 expression and eight months in those overexpressing PD-L1 [
11]. Altogether, these results suggest that PD-L1 expression has a negative prognostic effect in HCC patients, and a PD1 blockade (but not sorafenib) can reverse this negative effect on survival. Nevertheless, the benefit of nivolumab cannot be predicted by PD-L1 expression alone, as it did not affect the OS of patients undergoing this treatment.
Meanwhile, other putative biomarkers are under investigation. Several modern pieces of research are exploring the relationship between DNA damage/mutations and tumour immunogenicity. Tumour Mutational Burden (TMB) is a quantitative measure of the total number of nonsynonymous mutations per coding area of the tumour genome and is considered a surrogate marker of tumour immunogenicity reflecting neoantigen load. TMB is usually calculated using next-generation sequencing (NGS) techniques on tumour samples. Moreover, new blood tests (bTMB), exploring a limited number of genes, are under investigation in an attempt to obtain liquid biopsies in patients with tumours inaccessible to biopsies [
50]. TMB determination, however, suffers from the same limitations of PD-L1 staining, namely, the lack of standardised thresholds and variability in quantification methods [
51].
Mutations in the mismatch repair (MMR) system and microsatellite instability (MSI) are other DNA alterations potentially associated with increased tumour immunogenicity. In particular, tumours harbouring an erroneous MMR system will accumulate DNA mutations, which can lead to the presence of high levels of mutation-associated neoantigens [
52,
53], so that anti-PD1 agents are now prescribed to patients with colorectal cancers showing MSI [
54]. The role of these DNA alterations in guiding the HCC treatment remains to be established.
3.2.4. Resistance to Immune Checkpoint Inhibitors
Despite the overall encouraging results of immunotherapy, most HCC patients under anti-PD1/PD-L1 and anti-CTLA4 agents eventually experience a disease progression. Resistance to ICIs can be primary or acquired.
The primary resistance to ICIs is due to a paucity (or even lack) of intratumoral immune infiltrate, which suggest a defective immune cell trafficking. Interestingly, this profile of “immune exclusion” is often associated with an activated Wnt/ß-catenin pathway signalling in HCC [
55], giving support to the role of Wnt/ß-catenin activation as biomarker predictive of resistance to ICIs.
Primary resistance to ICIs may also derive from a more complex alteration of the immune system, with other immune pathways lying outside of the classical PD-1/PD-L1 and CTLA-4 checkpoints. For instance, lymphocyte activation gene-3 (LAG-3) is involved in the inhibition of CD8+ T cell and NK cell functions. Its expression is associated with a poor prognosis in HCC patients [
56]. Moreover, T-cell immunoglobulin and mucin-containing protein-3 (TIM-3) and its ligand galectin-9 can activate a complex cascade that ultimately leads to T-cell exhaustion [
56].
It can be argued that LAG-3, TIM-3 and PD-1 act synergistically, facilitating the HCC immune evasion, and could mediate the resistance to the classical PD-1/PD-L1 blockade [
57,
58]. Some trials are currently investigating the association effects of ICIs combined with TIM-3 (NCT03099109) and LAG-3 (NCT01968109) inhibitors in solid tumours.
Tumour microenvironment (TME) is another possible player in the development of primary resistance to ICIs. TME includes not only immune cells, but also blood vessels, fibroblasts, signalling molecules and the extracellular matrix surrounding the tumour [
59]. Indoleamine-pyrrole 2,3-dioxygenase 1 (IDO-1) is a heme-containing enzyme physiologically expressed in many tissues and cells, which is activated during tumour development, helping malignant cells escape eradication by the immune system (56). Tumours with high IDO1 deplete the essential amino acid tryptophan from TME, resulting in T-cell anergy and immune suppression [
59,
60]. EPACADOSTAT and INCAGN01949 are two drugs targeting IDO and the T-cell costimulatory molecule CD134, which are being tested in combination with ICIs for HCC (NCT02178722, NCT03241173).
The role of TME is strictly related to the phenomenon of the epithelial to mesenchymal transition (EMT), a cellular process that enables epithelial cells to gain mesenchymal features leading to an aggressive and motile phenotype [
61]. Several animal models and in vivo patient studies have shown that the activation of EMT in HCC promotes tumour progression and metastasis [
62]. Moreover, EMT can promote an immunosuppressive TME by recruitment of tumour-associated macrophages, regulation of immune checkpoint molecules and immune resistance to NK cell-mediated lysis [
63,
64]. The association between EMT and immunosuppression has been reported in different cancer types, including HCC [
65]. The role of tumour growth factor beta (TGF beta) is also of particular interest. This multifunctional cytokine plays multiple key activities, because of its role in immune and stem cell regulation and differentiation [
66,
67]. In many cancer cells, the TGF-β signalling is disrupted [
68], and therefore, TGF-β is no longer able to downregulate the cell cycle, causing a simultaneous proliferation of both cancer and surrounding stromal cells in the setting of an immunosuppressive and pro-angiogenic microenvironment [
69]. Additionally, TGF-β can convert effector T-cells into regulatory T-cells [
70] and exerts inhibitory effects on B-cells [
71], turning off the inflammatory reaction and favouring tumour immune escape. Investigation on the combination of nivolumab and the TGF-beta inhibitors galunisertib and ascrinvacumab (NCT02423343, NCT03893695) is currently in progress and will provide valuable information about the ability of these combinations in overcoming resistance to ICIs.
The VEGF signalling pathway can also provoke immune resistance as it induces Fas ligand, leading to cell death in tumour-infiltrating CD8+ T cells [
72]. The results of the ImBrave-150 trial testing atezolizumab-bevacizumab would support the hypothesis of a consistent role of this pathway in HCC progression. However, the design of this trial does not clarify whether the efficacy of atezolizumab-bevacizumab derived from a synergistic or additional effect [
73].
Currently, there are no data suggesting that the cirrhotic microenvironment actually affects the efficacy of ICIs. Indeed, when the first trials of immunotherapy for HCC were designed, this hypothesis (deriving from pre-clinical experiences) [
42] was considered, but subsequent clinical data showed a similar efficacy both in viral and nonviral patients. Even more relevant, no differences between cirrhotic and noncirrhotic patients have been so far demonstrated.
Acquired resistance to ICIs is an even more complex phenomenon and is rapidly becoming a hot topic as its occurrence hampers long-term results in patients responding to immunotherapy. Differently from classical chemotherapies and TKIs, ICIs have not a direct an antitumour effect as they act by enhancing the cytotoxicity of the immune system. The acquired resistance to ICIs probably relies on different events, for which, however, dynamic mutations in tumour cells still play a pivotal role. In particular, mutations in genes codifying for target antigens of the HLA system (resulting in a loss of expression of HLA genes on tumour cells) or in genes involved in the interferon signalling may be involved in this phenomenon [
74]. While some strategies to overcome these events can be hypothesised (i.e., enhancing the natural killer T-cell response in case of HLA loss), mechanistic and clinical studies are needed to highlight these phenomena further. Interestingly, on ongoing trial for the HCC treatment, combining nivolumab and ABX196 relies on the possibility of activating the natural killer T-cells, potentially overcoming the acquired resistance derived from the HLA loss (NCT03419481).
3.2.5. The Radiological Evaluation of Response
Historically, the RECIST 1.1 [
75] have been used as the preferred radiological criteria to assess the response to the systemic drugs for most malignancies, including HCC. However, in the case of HCC, the modified RECIST criteria (mRECIST) assess the response to locoregional treatments and have also been endorsed for the evaluation of systemic therapies [
76]. Whether the information provided by the mRECIST in the systemic setting is superior is still a matter of debate [
77,
78], but the leading regulatory agencies still require a RECIST 1.1-based evaluation.
However, the advent of immunotherapy poses some unique challenges that cannot be addressed by both RECIST1.1 and mRECIST. In early trials of ipilimumab for melanoma, the investigators described an initial disease behaviour meeting the RECIST criteria for progressive disease, followed by marked and durable responses [
79]. This pattern was called “pseudoprogression” and was attributed to a delayed response to ICIs and prompted the RECIST working group to propose new immune-related response criteria (iRECIST) [
80]. According to these criteria, an increase of the tumour burden or even the appearance of new lesions should be classified as unconfirmed progression (iUPD) [
80], and if the patients are clinically stable, ICIs should not be discontinued, and a new imaging assessment should be scheduled in the next 4.-8 weeks. In case of further increase of the tumour burden, radiological progression is confirmed (iCPD), and the treatment should be discontinued. If the tumour remains stable or shrinks, the imaging showing iUPD is regarded as a novel “baseline imaging” for the subsequent evaluations [
80].
The combinations of ICIs with either TKIs or anti-VEGF agents could prevent pseudoprogression, and consequently, the applicability of iRECIST for combination therapies is debatable and should be investigated.