Etiology-Based Treatment for Unresectable/Advanced Hepatocellular Carcinoma: Focus on Viral Hepatitis and Metabolic Dysfunction-Associated Steatohepatitis

Abstract

Background

Immune checkpoint inhibitors (ICI) are a keystone in advanced hepatocellular carcinoma (aHCC) therapy. However, preclinical evidence suggests that mice with pure metabolic dysfunction-associated steatohepatitis (MASH)-related HCCs may not benefit from ICIs. This has tremendous implications for both therapy selection and future drug development.

Summary

Viral and MASH-HCCs differ in molecular pathogenesis and immune microenvironment. Preclinical evidence has shown impaired immune surveillance and ICI-induced auto-aggressive T cells in pure MASH-HCC. Phase 3 clinical trials and meta-analyses have conflicting outcomes. For single-ICI regimens with anti-programmed-death 1/L1 (anti-PD1/L1), there was no apparent difference in the overall survival for patients with non-viral HCCs in the CheckMate-459, IMbrave150, COSMIC-312, and KEYNOTE-240 trials for ICI over tyrosine-kinase inhibitors or placebo, while patients with viral hepatitis seemed to have benefited more. Meanwhile, comparable outcomes were seen for patients with viral and non-viral HCCs in the HIMALAYA, RATIONALE-301, LEAP-002, and CARES-310 trials. There was no consistent difference in outcomes between patients with HBV or HCV-HCCs. For dual-ICI regimens with anti-PD1/L1 and anti-cytotoxic T-lymphocyte associated protein-4 (anti-CTLA-4), the HIMALAYA and CheckMate-9DW trials showed similar efficacy for ICI combinations across all etiologies. Important caveats in interpreting current evidence exist. All trial data are from unplanned subgroup analyses only, limiting the level of clinical evidence available. “Non-viral HCCs” are also a heterogenous entity, the composition of which varies from trial to trial. Fundamentally, defining HCC etiologies, especially in mutually exclusive terms, is challenging: Occult hepatitis B infections are inconsistently tested and reported; hence, viral HCCs may be underreported. HCCs are often multifactorial, with steatosis frequently co-existing and interacting with viral hepatitis, creating difficulty in translating findings from pure MASH-HCC mouse models.

Key Messages

Current evidence is insufficient to support individualizing aHCC treatment based on etiology. Preclinical evidence for a lesser benefit with ICIs in MASH-HCCs exists, while data from clinical trials are mixed but limited. Major gaps in knowledge remain, and further studies are required to clarify this important subject.

Introduction

Hepatocellular carcinoma (HCC) is the third leading cause of cancer death worldwide, with over 1 million patients projected to be diagnosed annually by 2025 [1]. There are considerable geographical variations in the cause of HCC, with viral infections by hepatitis B/C predominating in Asia and Africa, and non-viral causes, mainly alcoholic cirrhosis and metabolic dysfunction-associated steatohepatitis (MASH), prevailing in the West (Fig. 1) [2–4]. Metabolic dysfunction-associated steatotic liver disease (MASLD), defined as hepatic steatosis in the setting of at least one cardiometabolic risk factor without other causes of steatosis, can progress to MASH, characterized by histological features of lobular inflammation and hepatocellular ballooning [5]. The global epidemic of metabolic syndrome has contributed significantly to the increasing prevalence of MAFLD [6] which has increased over 50% in the Asia-Pacific region over the past two decades and affects 30–40% of the global population [3, 4, 7]. Corresponding, the burden of MASH-related HCCs has been increasing rapidly [8]. Indeed, although anti-viral therapies have reduced the incidence of viral HCCs, MASH-related HCCs have surged in many parts of the world including the Americas, Eastern Mediterranean, and Southeast Asia [8, 9].

Fig. 1.

Contribution of etiology to incidence of primary liver cancer by region in 2021.

For HCCs diagnosed at an advanced stage (aHCC), systemic therapy using immune checkpoint inhibitors (ICIs) with or without anti-angiogenic agents is widely used. However, recent preclinical models suggest that ICIs are detrimental in MASH and do not control pure MASH-HCCs. This is deeply concerning given the central role ICIs play in aHCC therapy, and that MASH is expected to become the dominant cause of HCCs in many countries. Whether or not the etiology of HCC may impact ICI treatment efficacy has, hence, become a leading issue in the field of HCC research. In this review, which will be narrative in nature, we seek to summarize both available preclinical and clinical evidence and discuss the potential impacts of HCC etiology on the efficacy of ICI, with a particular focus on viral and MASH-related HCCs.

Methods

We conducted searches on electronic databases such as PubMed to identify publications related to preclinical, translational, and clinical data for ICI use in HCCs of different etiologies. Only English language publications were considered. Citations of, and references made by publications yielded from searches were further explored for other literature. The inclusion of individual publications to be referenced was at the discretion of the authors according to relevance and importance to the subjects of discussion.

Molecular and Microenvironment Differences of Viral and MASH-Related HCCs

Distinct natural histories, pathogenesis, pathway perturbations, and immune microenvironments (Fig. 2) support the observation that HCCs of different etiologies are in fact different entities [10]. HBV is capable of hepatocarcinogenesis via insertional mutagenesis within the TERT promoter, resulting in telomerase activation and subsequent hepatocyte transformation [11]. Genomic profiling demonstrated other frequent mutational drivers such as CTNNB1 and TP53, the latter of which is significantly more likely to be mutated in HBV-HCC compared to HBV negative HCCs [12]. Conversely, the HCV RNA virus is incapable of integrating into the host genome but induces chronic hepatic inflammation leading to oxidative stress, fibrogenesis, genetic instability, and DNA damage [13]. HCV-HCCs demonstrate increased TERT promoter mutation and CDKN2A promotor silencing, which leads to downregulation of tumor suppressor p16 [12]. The stepwise progression from chronic liver disease to cirrhosis and subsequent HCC is a hallmark for HCV, while HCCs may occur without cirrhosis in HBV infection [10]. MASH promotes HCC through lipotoxicity, causing oxidative and endoplasmic reticulum stress, which induces inflammation and cell damage [14, 15]. Subsequent reactivation of molecular pathways such as Notch, Hedgehog, and YAP-TAZ in hepatocytes leads to inflammation, cell proliferation, and cancer [16]. In addition to other typical mutations present in viral HCCs such as TERT, CTNNB1, and TP53, MASH-HCC has higher mutation rates in ACVR2A, a potential tumor suppressor gene, compared to HCCs of other etiologies [17].

Fig. 2.

Different pathogenesis and immunopathology of HCC etiologies.

Of greater relevance to ICI therapy (and hence to this review) is the distinct immune microenvironmental profiles in HCCs of different etiologies. Although chronic inflammation is central to the pathogenesis of both viral and non-viral HCCs, the repertoire and functions of tumor infiltrating lymphocytes are dissimilar. When tumor samples from patients with HBV-related HCCs are compared to those with non-viral HCCs, HBV-related HCCs have a more immunosuppressive microenvironment characterized by lower expression of pro-inflammatory markers such as interferon-gamma and tumor necrosis factor alpha. Tumor infiltrating lymphocytes from HBV-related HCCs also express more PD1. Interestingly, both regulatory T cells (Tregs) and resident CD8+ memory T cells (Trm), which have contradicting functions and effects on prognosis, are upregulated in HBV-related HCCs compared to non-viral HCCs, while Tim3+CD8+ T cells and CD244+ natural killer cells are preferentially enriched in the latter [18]. Similarly, HCV-HCC exhibits a significant presence of exhausted CD8+ T cells expressing high levels of PD1 [19]. Anti-PD1 treatment can reactivate PD1+ cytotoxic CD8+ T cells while also attenuating the immunosuppressive functions of PD1+ Treg [18, 20], underscoring the therapeutic relevance of targeting this crucial pathway in viral HCCs. Thus, there is a strong rationale for the efficacy of PD1/PD-L1-targeting ICIs in viral HCCs.

In non-alcoholic steatohepatitis (NASH), immunopathology is a central feature with diverse effects on various T-cell subsets. Importantly, in a series of landmark preclinical studies using mouse models, immune checkpoint blockade seemed to instead exacerbate NASH via CD8+ T cell-mediated mechanisms in livers without HCCs, without corresponding anti-tumor activity in NASH-related HCC [21, 22]. In immunocompetent mice without HCCs, liver resident (CXCR6+), exhausted (PD1 high), effector (granzyme+) CD8+ T cells are accumulated in the livers of mice being fed NASH-inducing diet but not normal diet, likely induced by IL-15 causing FOXO1 inhibition and CXCR6 upregulation [21, 22]. Although this may initially appear to suggest that the PD1/L1 pathway is an attractive target for NASH-related HCC, the opposite was in fact demonstrated: such T cells displayed auto-aggressive behavior against hepatocytes when exposed to fatty acids (specifically acetate), which was abrogated by tissue-necrosis factor (TNF) blockade. Indeed, development of NASH seemed abrogated in mice with TCR deficiency. Accordingly, when mice with NASH but not HCC were given prophylactic anti-PD1, the incidence of HCC increased and NASH activity worsened, and these deleterious effects were abrogated by either CD8 depletion or TNF blockade [22]. Notably, these T cells were also found in liver biopsy specimens of patients with NASH, suggesting potential translational implications for these preclinical findings [21]. Finally, when mice with NASH-HCC were given anti-PD1, no response was obtained despite expanded CD8⁺PD1⁺CXCR6⁺, TOX⁺, and TNF⁺ T cells [22].

Contrary to the persistence and expansion of CD8+ T cells, NASH is also associated with CD4+ T-cell depletion but expansion of Treg in preclinical models [23–25]. CD4+ T cells play the center role in immune surveillance and clearance of premalignant hepatocytes [26]. However, in mice with non-alcoholic fatty liver disease (NAFLD), CD3-high, CD4+ T cells are reduced compared to control mice, with a corresponding increase in hepatocarcinogenesis. Dysregulated lipid metabolism in NAFLD resulted in the accumulation of linoleic acid, a fatty acid which disrupted mitochondrial function, increased reactive-oxidative species (ROS) generation and, hence, oxidative damage. Compared to CD8+ T cells, CD4+ T cells were disproportionately affected due to their higher mitochondrial mass and oxygen consumption rates, the effects of which were abrogated upon blocking ROS. In addition, neutrophil extracellular traps, which are abundant in livers affected by NASH, facilitates mitochondrial oxidative phosphorylation and, thus, differentiation of naïve CD4+ T cells into Tregs, which in turn impairs immunosurveillance of premalignant hepatocytes [24].

Collectively, these preclinical observations suggest that immunosurveillance is impaired in livers affected by MASH, and that anti-PD1 therapy may be deleterious in MASH-HCC due to aberrant T-cell activity, providing the basis for further study of possible impact of HCC etiology on ICI response in clinical settings. However, it must be stressed that these observations are preclinical in nature only, and extrapolations about clinical efficacy in humans should be avoided.

Clinical Evidence: A Critical Appraisal

Summary of Current Evidence

Advanced HCCs

Multiple phase 3 trials have evaluated ICIs in aHCC. These include anti-PD1/L1 alone or in combination with bevacizumab or tyrosine-kinase inhibitors (TKIs), as well as dual-ICI regimens of anti-CTLA-4 combined with anti-PD1/L1 (Fig. 3, 4) in first-line aHCC compared to sorafenib (except for CheckMate-9DW which mostly compared against lenvatinib). Additionally, ICIs have been compared to placebo in second-line aHCC. For the purposes of this review, all subgroups are defined according to the individual trial referenced. Generally, patients with “non-viral HCCs” include those in whom HBV and HCV infections were excluded using trial-specific criteria, and encompasses patients with HCCs of alcoholic, MASH and other heterogenous etiologies.

Fig. 3.

Summary and forest plots of OS with ICI-based regime in phase III trials.

Fig. 4.

Summary and forest plots of progression-free survival with ICI-based regime in phase III trials.

First-Line Trials

Single Agent Anti-PD1/L1. In CheckMate-459, although nivolumab demonstrated numerically higher objective response rates (ORR) compared to sorafenib across all etiology subgroups, the benefits seem more noticeable in patients with viral HCCs. Pre-specified subgroup analyses for overall survival (OS) in patients with HBV or HCV-HCCs showed HRs of 0.77 (95% CI 0.56–1.05) and 0.71 (0.49–1.01), respectively, while that of patients with non-viral HCCs was 0.95 (0.74–1.22) [27]. Meanwhile, the ORRs were 19%, 17%, and 12%, respectively, compared to 7–8% for sorafenib. In contrast, other single agent PD1/L1 trials did not show apparent differences in outcomes for patients with viral and non-viral HCCs. In the single agent durvalumab arm of HIMALAYA, the HR for OS in patients with HBV, HCV, and non-viral HCCs were 0.81 (0.62–1.07), 1.00 (0.73–1.37), and 0.81 (0.64–1.04), respectively [28]. In RATIONALE-301, the HR for patients with HBV, HCV, and non-viral HCCs were 0.91 (0.73–1.14), 0.64 (0.38–1.08), and 0.78 (0.55–1.12) in the tislelizumab arm [29].

Single Agent Anti-PD1/L1 with Bevacizumab. In IMbrave150, atezolizumab-bevacizumab demonstrated consistent anti-tumor activity and survival across etiologies. Post hoc analysis showed no significant difference in OS, PFS, and ORR for patients with NAFLD, HCV, HBV, and alcoholic liver disease in the atezolizumab-bevacizumab arm, and between NAFLD and HBV or HCV patients specifically [30]. When compared to sorafenib arm, the OS benefits for atezolizumab-bevacizumab seemed more pronounced in patients with HBV and HCV-HCC, with HRs of 0.58 (0.4–0.83) and 0.43 (0.25–0.73), while benefit was not apparently seen in non-viral HCC patients with HR 1.05 (0.68–1.63) [31]. Of note, sorafenib performed exceptionally well in patients with non-viral HCCs, with a median OS reaching 18.1 months, considerably longer than that in the SHARP trial (10.7 months) or other sorafenib arms in recent phase three trials (13–15 months), which possibly cause the lack of comparative benefits for atezolizumab-bevacizumab in this subgroup despite preserved absolute benefits [27, 32, 33].

Single Agent Anti-PD1/L1 with Oral TKI. COSMIC-312 is, uniquely, the only ICI trial to further stratify non-viral causes and describe outcomes specifically for patients with MASH-HCC in a pre-specified manner [34, 35]. Although the overall OS outcomes were negative, patients with HBV-HCCs seemed to derive the most benefit, with an HR of 0.63 (0.42–0.95) compared to 1.13 (0.78–1.63), 1.26 (0.71–2.22), and 1.72 (0.83–3.55) for patients with HCV, alcoholic, and NASH/NAFLD-related HCCs, respectively. This pattern was also consistently observed in the PFS outcomes (Fig. 4). Conversely, two other ICI-TKI trials showed no apparent difference in relative efficacy across HCC etiologies. In LEAP-002, outcomes for patients with viral and non-viral HCCs were similar, with HRs for OS being 0.84 (0.67–1.05) and 0.86 (0.66–1.13), respectively [36]. Patients with viral and non-viral HCCs also had similar OS and PFS outcomes in the CARES-310 (Fig. 3, 4) [37].

Dual ICI Regimen. The CTLA-4 pathway affects the priming and proliferation phases of T cells, while the PD1 pathway mainly affects the effector phase [38]. Combined PD1 and CTLA-4 blockade creates distinct biological effects compared to blocking PD1 alone, such as the expansion of memory T cells and differential levels of various plasma chemokines [39, 40]. In fact, we and subsequently other groups have shown that combined blockade has activity in aHCC post-progression on single agent anti-PD1/L1, suggesting a possible distinct clinical effect as well [41–43]. Whether such distinctions translate into more benefits for patients with non-viral HCCs remains to be investigated. However, it is interesting to note that, unlike other drug combinations which had contradicting outcomes across different trials, the two anti-CTLA-4 and anti-PD1/L1 trials published so far both showed consistent benefits for patients with non-viral HCCs. In HIMALAYA, the Single Tremelimumab Regular Interval Durvalumab (STRIDE) regimen demonstrated similar benefits for patients with viral and non-viral HCCs over sorafenib, with HRs of 0.68 (0.51–0.91), 0.94 (0.68–1.29), and 0.75 (0.59–0.96) for patients with HBV, HCV, and non-viral HCCs, respectively [28]. Exploratory analysis using a more strictly defined non-viral etiology (no active viral hepatitis or prior HBV infection) also showed an OS benefit of HR 0.62 (0.47–0.81) [44]. Next, in CheckMate-9DW, benefits were consistent across etiologies for ipilimumab-nivolumab over lenvatinib or sorafenib, with HRs for OS for patients with HBV, HCV, and non-viral HCCs being 0.84, 0.68, and 0.84, respectively, and responses seen regardless of cause of HCC [45].

Second-Line Trials

In KEYNOTE-240, OS benefits seem numerically greater in patients with HBV-HCC (HR 0.57 [0.35–0.94]) treated with pembrolizumab compared to placebo, while the HR of patients with non-viral HCCs was 0.88 (0.64–1.20) [46]. However, the HR for PFS was similar (0.70 [0.44–1.13] and 0.75 [0.56–1.01]) respectively. Only a small number of patients with HCV-HCC were recruited (15.5% of all).

Meta-Analyses of Trials in Advanced HCC

Relevant meta-analyses have largely reflected the conflicting outcomes of individual trials. Haber et al.[47] analyzed three ICI trials (CheckMate-459, KEYNOTE-240, IMbrave150, n = 1,656) and showed that the pooled HR for OS in patients with viral HCCs was significantly better than patients with non-viral HCCs treated with ICIs (0.64 vs. 0.92, p of heterogeneity = 0.0259). Further breakdown of viral etiology into HBV and HCV showed similar benefits, with HRs of 0.64 and 0.68, respectively. This difference was not seen in an analysis of five TKI/anti-VEGF trials (REACH, REACH-2, METIV-HCC, CELESTIAL, JET-HCC, n = 2,083), with similar HRs of 0.74, 0.90, and 0.82 in patients with HBV, HCV, and non-viral HCCs seen. However, in a more recent meta-analysis conducted by Meyer et al. [48] which included other studies mentioned above, a significant OS advantage for both patients with viral and non-viral HCCs was found (HR 0.79, p < 0.001). Patients with HBV-HCCs had the largest predicted benefit (HR 0.70, p < 0.001), followed by patients with HCV-HCCs (HR 0.78, p = 0.04) and patients with non-viral HCCs (HR 0.87, p = 0.02). A meta-analysis of eight studies involving 1,520 patients by Ding et al. [49] also did not demonstrate a significant difference in ORR across patients with viral and non-viral HCCs (OR 1.03, p for heterogeneity 0.152), or between patients with HBV and HCV-HCCs (OR 0.74, p for heterogeneity 0.374).

Intermediate Stage HCC

Single agent anti-PD1/L1 with oral TKIs or bevacizumab and transarterial chemoembolization (TACE) have also been evaluated against TACE alone for patients with intermediate stage HCC. All three trials (LEAP-012 with pembrolizumab-lenvatinib, EMERALD-1 with durvalumab-bevacizumab and TALENTACE with atezolizumab-bevacizumab) have demonstrated PFS benefit without mature OS outcomes [50–52]. Subgroup analysis for etiology has been published for LEAP-012 and EMERALD-1. Interestingly, both trials reported that patients with non-viral HCCs benefitted equally from ICI-containing arms, with HR for PFS in viral and non-viral subgroups in LEAP-012 being 0.68 (0.52–0.90) and 0.52 (0.33–0.83), respectively, and that of HBV, HCV and non-viral subgroups in EMERALDS-1 being 0.82 (0.55–1.23), 0.68 (0.43–1.09), and 0.74 (0.51–1.08) [52].

Implications on Personalization of Therapy

Despite a plethora of trials being conducted in the treatment of aHCC, whether the etiology of HCCs should be considered to individualize treatment decisions is still uncertain. Contrary to preclinical evidence, which suggested that accelerated hepatic injury without corresponding anti-tumor effect may be seen when ICIs were given to MASH-affected mice, anti-tumor activity in terms of ORR is still seen in patients with non-viral HCCs treated with anti-PD1/L1, and no detrimental effects on OS has been observed (despite a lack of data on hepatic toxicities). Therefore, in absolute terms, anti-PD1/L1-based ICIs should still be given to patients currently classified as having “non-viral HCCs”.

A much more difficult question to answer is whether single-ICI regimens with anti-PD1/L1 are unequivocally superior to TKIs in patients with non-viral HCCs in the advanced setting. The evidence is somewhat mixed: ICI-containing regimens given to patients with non-viral HCCs were shown to have no additional OS benefit in CheckMate-459, IMbrave150, and COSMIC-312 over TKIs [27, 31, 35], but similar OS benefits to viral etiologies were seen in other studies. The lack of benefit in some studies is likely at least partially explained by TKIs having more favorable outcomes in patients with non-viral HCCs. Only IMBrave150 and COSMIC-312 reported the median OS in patients with non-viral HCCs [31, 35], but the outcomes for sorafenib were remarkable (18 and 20 months) compared to patients with viral HCCs in the same trials (12–14 months) or to historical data of sorafenib in unselected HCC [32, 33]. Intriguingly, the above differences are not observed in both CheckMate-9DW and HIMALAYA involving dual ICI with anti-CTLA-4, suggesting that dual-ICI regimens consistently convey unequivocal superiority over TKIs across HCCs of different etiologies [28, 45]. Given the available evidence, we suggest that dual-ICIs are probably favored over TKIs in patients with non-viral HCCs, and single-agent TKI remain a viable option if dual-ICIs are less preferred for other reasons. However, the conclusions that can be drawn are fundamentally limited by major issues in both the quality of evidence and in our fundamental framework used to understand the etiology of HCCs, as shall be discussed below. As for the use of ICIs in the intermediate setting, given that none of the trial regimens are globally approved so far, and that important data such as OS are still awaited, more data are needed before a conclusion can be made.

Major Limitations in Available Clinical Evidence

There are major limitations in extrapolating trial results to individualize treatment for patients with HCCs of different etiologies in routine clinical care. First and foremost, most trial-level data are from unplanned subgroup analyses. Such analyses are not powered for statistical comparison, and no formal statistical comparisons between viral and non-viral related subgroups have been made (appropriately so). Available meta-analyses all used trial-level data only as well, which is in turn based on these unplanned subgroup analyses, instead of the more robust individual patient data meta-analyses (or simulated individual patient data). The quality and standardization of data is limited, and between-study heterogeneity may have been considerable. More specifically, the definition of etiologies (such as testing strategies for viral hepatitis) may vary. Additionally, as mentioned above, the size and composition of “non-viral” subgroups may vary from trials to trial depending on where the trial was conducted. All these factors affect the validity of pooled estimates and add difficulties to interpretation.

Next, important accompanying data are not reported, including baseline demographic, tumor characteristics and toxicities. Indeed, major prognostic factors may not have been balanced between subgroups of patients with viral and non-viral related HCCs. Patients with MAFLD/MASH-related HCCs are known to be older and with more cardiovascular comorbidities compared to those with viral hepatitis [53]. Furthermore, as countries where viral hepatitis infections are endemic have instituted large-scale screening programs, increasingly more viral HCCs have been screen-detected [54, 55]. On the contrary, surveillance programs are less prevalent and standardized in MASH-at risk populations [56]. As a result, MAFLD-related HCCs are more likely to be detected outside specific surveillance compared to viral HCCs [57]. HCC screening is known to reduce overall disease-related mortality mainly effected through detecting HCCs at an earlier stage to allow for more curative resections [54, 58]. However, even for patients with advanced HCC, those who were part of screening programs may differ from those who were not in the following ways: more HCCs may have been detected initially in the intermediate stage in the former group, with more TACE being performed. This may result in a different tumor and immune microenvironment as TACE is a known inducer of immunogenic cell death [59]. In addition, because of earlier detection, intrahepatic tumor bulk and extent of vascular or extrahepatic involvement may be smaller in the screened group. In fact, MAFLD-related HCCs are often larger in extent at diagnosis compared to viral HCCs [57]. A larger overall tumor bulk is associated with higher intra-tumoral genomic heterogeneity [60], which is a major driver for therapy resistance due to a larger plethora of potentially resistant clones [61]. Smaller intrahepatic tumor bulk, less vascular involvement, and less advanced cirrhosis may also allow for better liver function preservation and a more favorable TME. Additionally, because they may be younger, have less comorbidity, and more preserved liver function, patients with viral HCCs may possibly have received more post-progression therapies compared to those with MASH-HCCs, resulting in better OS outcomes. These interesting factors could have been further explored but for the unfortunate absence of reporting of ORR, disease control rate (DCR), landmark outcomes, survival curves and post-progression therapy analyses according to tumor etiology. Beyond clinical efficacy, toxicity profiles according to etiology are also of great clinical relevance. As mentioned above, anti-PD1 exacerbates hepatocellular injury via auto-aggressive T cells in mouse models with NASH. Thus, it is of tremendous interest to know if patients with MASH-HCCs indeed experience greater hepatic impairment or progressive fibrosis compared to those with viral HCCs. Unfortunately, none of the above studies reported adverse events stratified according to tumor etiology to allow for exploration of this important question, and further studies are urgently required.

Additionally, the heterogeneity of the trials conducted limits our ability to form a conclusion from the overall body of evidence. The proportion of patients of each etiology varied widely between different trials. As an example, the proportion of patients classified to have non-viral HCC ranged from less than 20% in CARES-310, 30.5% in IMBrave150 to up to 45% in Checkmate-459 [27, 37, 62]. Thus, the subgroup sizes varied widely and limit any cross-trial comparisons, however hypothesis-generating in nature only. Such issues could have been ameliorated by meta-analyses but for the heterogeneity of treatment. As single agent anti-PD1/L1 have not conclusively demonstrated superiority over sorafenib, more contemporary trials have all added either anti-VEGF monoclonal antibodies, TKIs or anti-CTLA-4 antibodies. Thus, it is difficult to evaluate any efficacy or detriment of anti-PD1/L1 in isolation, as well as cross-compare or aggregate the results of different trials with dissimilar second agents.

Furthermore, the definitions of subgroups were imprecise, and the composition varied from trial to trial. In categorizing HCC etiology, the subgroup “non-viral” was usually used, often defined only as having no identifiable HBV/HCV infection as mentioned above. Further stratification of non-viral etiologies was only done in COSMIC-312 and LEAP-002, which reported 22.4% and 31.6% of patients having alcohol-related HCC, respectively – a considerable difference given that patients with non-viral HCCs only constituted 30–40% of the entire trial population; while the incidence of MASH-HCCs was only known in the former [35, 36]. Given that global trials recruit in countries with dissimilar incidences of MASLD/MASH, alcoholism, and aflatoxin exposure (Fig. 1), the “non-viral” subgroups probably consist of different proportions of HCCs due to each etiology across different trials, as the above example has shown [63–65]. This heterogeneity is probably one of the major reasons of the inconsistencies between trial results. It is, thus, not possible to evaluate the efficacy of treatment on HCCs of specific causes, for example, MASH, when such broad categories of variable composition are used without more granular data. Above all, this is also likely the result of perhaps the principal difficulty in research of this topic: the inability to define HCC etiologies in precise and exclusive categories.

Defining HCC Etiologies: A Foundational Issue

Fundamentally, etiology-based personalization of therapy requires precise definition of HCC cause. However, these attempts are significantly complicated by difficulties in defining viral hepatitis infection due to occult hepatitis B infections, defining MASH due to lack of biopsy, the frequent co-existence of steatosis and viral hepatitis, and the multifactorial nature of many HCCs.

Occult Hepatitis B Infections

Accurately determining HBV status can be challenging and inconsistent. Occult hepatitis B infection (OBI) has increasingly been recognized to be hepatocarcinogenic via pathways similar to chronic HBV infection, such as viral integration, production of pro-oncogenic proteins, and persistent necroinflammation [66]. In OBI, replication-competent HBV DNA is present in the liver, while serum hepatitis B surface antigen (HBsAg) is negative, and blood HBV DNA may be absent. In fact, up to 20% of OBIs may be negative for both HBsAg and hepatitis B core antibody (anti-HBc) with existing assays [67, 68]. However, published clinical trials do not uniformly define HBV positivity to allow for detection of OBI. For example, HBV DNA was tested in all patients in HIMALAYA, and HBV infection was defined by positive HBsAg, detectable HBV DNA or anti-HBc [69]. However, patients in IMbrave150 were tested for HBV DNA only if they were HBsAg or anti-HBc positive; while COSMIC-312 defined HBV infection only as HBsAg positive [34, 62]. Therefore, patients labeled as “HBV-negative” or “non-viral” may in fact have OBI in trials, meta-analyses and real life.

Lack of Biopsy for MASH Diagnosis

Steatosis itself is insufficient to cause HCC without inflammation, and only around 20% of patients with MASLD have MASH [70]. MASH is a histological diagnosis consisting of steatosis, hepatocyte ballooning, and lobular inflammation. Imaging alone cannot reliably distinguish MASH from mere steatosis [71]. However, liver biopsy is often not required to diagnose HCC in real life for patients with cirrhosis, creating difficulty in defining MASH-HCC in daily practice [72].

Interaction between Hepatic Steatosis and Viral Hepatitis

Hepatic steatosis is commonly seen in patients with viral hepatitis, and these conditions interact to mediate liver disease, adding difficulty to translating preclinical MASH-HCC evidence to clinical use or dichotomizing HCC etiologies (Fig. 5) [73, 74]. HCV and steatosis synergistically promote hepatic dysfunction and carcinogenesis. Most patients with HCV are overweight, and 40–80% have steatosis, which confers worse outcomes compared to HCV alone [75]. HCV (especially genotype 3) causes steatosis by stimulating de novo lipogenesis for lipoviroparticle assembly and inhibiting mitochondrial fatty acid oxidation [76]. HCV also exacerbates lipotoxicity by increasing liver ROS levels through effects of HCV core proteins and mitochondrial dysfunction [77], as well as causing hepatocytes to become senescent and accumulate fat [78]. Meanwhile, metabolic dysfunction worsens the pathogenic effects of HCV. Hyperinsulinemia exacerbates HCV-initiated fibrosis by stimulating insulin-receptor expressing myofibroblasts to cause fibroblast proliferation and collagen secretion [79]. De novo lipogenesis products also suppress PTEN, an important tumor suppressor of HCC [80]. Lastly, both HCV and MASH promote an immune tolerant environment for hepatocarcinogenesis by causing CD8+ T cells and CD4+ T-cell dysfunction [23, 81, 82], as well as the expansion of Tregs [24, 83].

Fig. 5.

Interaction between viral hepatitis and steatosis.

Unlike HCV, the relationship between HBV and steatosis is more complex. Steatosis is reported in around 30% of patients with HBV, similar to the general population and lower than in patients with HCV [84]. Clinically, steatosis assessed by transient elastography is associated with a 3-fold increase in HBsAg clearance rates, lower incidence of HCC [85–87], but higher risk of fibrosis progression [88]. Immune alterations and metabolic stress related to MASLD, such as toll-like-receptor-4 activation, may enhance anti-viral response and attenuate HBV-related liver disease progression [74, 89]. The same mechanisms may also activate hepatic stellate cells and worsen fibrosis [90]. Comorbid MASH and HBV have also been shown to have significantly reduced interferon gene signature and pathway activity, as well as macrophage gene signatures, compared to HBV or MASH alone, with uncertain effects due to the diverse roles interferons and immune cells play and the complex liver and systemic immune environment [91, 92]. Meanwhile, whether HBV leads to increased steatosis remains controversial: HBV viral load is observed to be inversely correlated to steatosis [93], but HBV has also been implicated in promoting steatosis via induction of IL-13, TGF-B1, myofibroblasts, and JAK/STAT pathways [74] and through inducing mitochondrial dysfunction and peroxisome proliferator-activated receptors by the HBx protein [94]. Although a complex interplay between hepatic steatosis and viral hepatitis exists, the mechanisms and impact of hepatic steatosis on viral-HCC development and therapy response to ICIs are unclear.

Ultimately, many HCCs are in fact multifactorial. Aside from co-existent MAFLD and viral hepatitis, the effects of alcohol on steatosis, accelerated cirrhosis, and hepatocarcinogenesis in patients with viral hepatitis are well established [95, 96]. Aflatoxin also synergistically effects HCC development with viral hepatitis through activation of hepatitis B viral proteins such as HBx and other pathways [97]. In view of the above factors, accurate and clinically relevant definition of HCC etiologies which also accounts for such complex interactions remains elusive.

Implications on Future Trial Design

HCC is a heterogenous disease, and the next generation of trials would be well served to take this into account. The proportion and outcomes of patients with non-viral HCCs may exert a major effect on the overall results and should be a top question when designing a trial. For example, in CheckMate-459, KEYNOTE-240, and COSMIC-312, the apparent reduced benefit from ICIs for patients with non-viral HCCs was likely a major cause for the negative OS outcomes, despite a clear benefit for patients with HBV-HCCs [27, 35, 46]. To provide clinicians with valuable data on treatment selection and ensure the continued relevance of trial outcomes as MASH becomes a leading cause for HCCs as projected [3], specific strategies addressing key knowledge gaps should be considered.

Standardizing Definitions and Reporting of Etiologies

In defining and reporting etiology, the following should be considered: testing strategies for HBV should be unified and include evaluation for OBI as well. Biopsy of tumor, which are often mandatory for trial entry unlike in real life practice, should also evaluate for the presence of steatohepatitis in surrounding liver to allow for accurate definition of MASH-related HCC. Instead of being grouped together as “non-viral HCCs”, the number of patients with MASH and alcoholic-related HCCs should be stated.

Stratifying Outcomes according to Etiologies

In trials involving all-comers, HCC etiology should be considered a key stratification factor in randomization, and enrollment of a pre-specified percentage of patients with MASH-HCC should be done. Reporting of outcomes according to etiology should be done, including clinical responses, survival, and, perhaps most importantly and urgently, toxicities. Ultimately, prospective trials dedicated to patients with HCC of specific etiologies should be considered.

Translational Research and Biomarker Exploration

Finally, in all preplanned translational studies of trials, steatohepatitis should be considered a biomarker to further explore its relationship with ICI efficacy. Acknowledging the considerable heterogeneity of HCCs, future efforts should also be made on identifying novel molecular and immune biomarkers unique to each etiology, which are currently absent. A comprehensive, large-scale comparison of molecular features, tumor microenvironment and immune landscape between viral and MASH-related HCCs would also be illuminating.

Conclusion

Viral and MASH-related HCCs have distinct pathogenesis and immune environments. There is preclinical evidence that ICIs may have deleterious effects in pure MASH-related HCCs. Evidence from clinical trials is limited to unplanned subgroup analyses and paints a mixed picture. Efficacy of ICIs is similar for patients with HBV and HCV-HCCs. For non-viral HCCs, single-ICI trials showed variable results, while dual-ICI trials have shown preserved superiority for ICIs over TKIs. Nevertheless, important issues remain, such as how to define HCC etiology when steatosis frequently co-exists with viral hepatitis, and current evidence is insufficient to recommend an etiology-based treatment approach. Better designed clinical trials are needed to further elucidate the role of HCC etiology on tumor microenvironment, immune landscape and ICI efficacy.

Conflict of Interest Statement

Prof. Assoc. Valerie Chew was a member of the journal’s Editorial Board at the time of submission. Prof. Assoc. Valerie Chew reports funding from National Medical Research Council (NMRC), Singapore (Reference No.: NMRC/OFLCG/003/2018); National Research Foundation, Singapore (Reference No.: NRF-CRP26-2021-0005); and Duke-NUS Khoo Research Fellow Award (Duke-NUS-KPFA/2024/0072).

Funding Sources

This review article did not receive any funding support.

Author Contributions

Prof. Thomas Yau contributed to the conception of the review, drafting and revision of the manuscript, and supervised the process. Frances S.K. Sun and Jeffrey S.L. Wong contributed to the conception, analysis, and drafting and revision of the final manuscript. Frances S.K. Sun also prepared and curated the graphs and figures. L.-Y. Mak, Carmen C.-L. Wong, Valerie Chew, Bryan Li, Roland Leung, and T.T. Cheung provided expert insights and contributed to the writing, analysis, and revision of the manuscript.

Funding Statement

This review article did not receive any funding support.