Hepatotoxicity of systemic therapies for unresectable hepatocellular carcinoma

The number of effective systemic therapies for the treatment of unresectable hepatocellular carcinoma (uHCC) is rapidly increasing and the advent of immunotherapy changed the treatment paradigm for these patients, leading to a significant improvement in survival outcomes. While sorafenib, a tyrosine‐kinase inhibitor monotherapy, remained the only effective treatment for almost a decade, the combination of atezolizumab, an immune checkpoint inhibitor (ICI) targeting programmed death‐ligand 1, plus bevacizumab, an antiangiogenic agent targeting vascular endothelial growth factor, now represents the new standard of care for patients with uHCC. Moreover, several further clinical trials are ongoing to evaluate novel combinations between ICIs with other drugs, belonging to the same class or to other classes. As HCC occurs in most cases in the setting of cirrhosis, the evaluation of the risk/benefit ratio of systemic treatments represents a critical point. The underlying liver disease significantly influences the safety and the effectiveness of current and future systemic treatments for uHCC. For this reason, the hepatotoxicity profile and impact on liver function of these molecules should be carefully assessed in both clinical trials and in the real‐world setting. Here, we review hepatotoxicity data on systemic treatments for uHCC and offer suggestions on monitoring and managing liver‐related adverse events occurring during the treatment.

features that could explain the wide prognosis spectrum observed in clinical trials and in the real-world practice. 3 Owing to its dismal prognosis, HCC represents the leading cause of death even in patients with compensated cirrhosis. A quite large amount of curative and palliative therapies exists, being differently applied according to the stage of the tumour and the underlying liver disease. Treatment options include curative options (liver transplant, surgical resection and thermo-ablation), treatments for intermediate stages, such as transarterial chemoembolization, and for advanced stages, such as systemic therapy. It should be highlighted that, unless liver cirrhosis is compensated, there are no available options to actively treat HCC.
Moreover, the occurrence of liver decompensation represents the main driver of death in patients with early HCC and it has also a significant impact on the survival of patients with more advanced HCC stages. 4,5 HCC is notoriously resistant to chemotherapy and other systemic treatment modalities; in fact, the drug-metabolizing properties of the liver, in addition to elevated levels of multidrug resistance proteins expressed by HCC cells, likely contributes to the limited efficacy of chemotherapeutics in the treatment of HCC. 6 Over the years, a huge amount of drugs have been tested, and the oral tyrosine kinase inhibitor (TKI) sorafenib has represented the first approved drug in 2008, according to the results of the SHARP trial. 7  showed to be more effective than sorafenib in phase III IMBrave 150 trial, 9 changing the paradigm of the treatment of unresectable (uHCC) and establishing an infusional ICI-based combination therapy as the new standard of care. Similarly, also the number of available drugs for second-line treatment has recently increased. Among TKIs, Regorafenib 10 showed efficacy in patients who tolerated prior therapy with sorafenib, becoming the first TKI approved after sorafenib for second-line therapy for uHCC. Thereafter, cabozantinib 11 was approved in both progressor and intolerant to sorafenib, having both been compared with placebo. Ramucirumab, 12,13 an anti-VEGF agent, significantly improved survival compared to placebo in patients with alpha-fetoprotein (AFP) higher than 400 ng/Ml. Among ICIs, nivolumab 14 and pembrolizumab 15 (both agents targeting PD1) have been tested as monotherapy in the second-line setting, reaching an accelerated approval by Food and Drug Administration (FDA) in the USA according to preliminary results of phase I and II trials as well as the combination of nivolumab plus ipilimumab 16 (an agent targeting Cytotoxic T-Lymphocyte Antigen-4 ).
In this evolutionary scenario, the quantitative assessment of the safety of these novel agents represents a crucial point to optimize the outcomes of patients in terms of net health benefit. 17 As HCC mainly occurs in the setting of cirrhosis, the potential hepatotoxicity of these drugs should be carefully evaluated, as well as their potential impact on liver decompensation, which represents an event that makes the patient no longer treatable. This issue is increasingly relevant as more sequential treatment lines are now available. Moreover, the clinical interpretation of the occurrence of liver injury in the setting of uHCC could be challenging, as it could depend on different factors including tumour progression, the decompensation of the underlying liver disease, drug-induced hepatotoxicity (including autoimmune mechanisms for ICIs) and viral (HBV or HCV) reactivation. Here, we reviewed the available data on the hepatotoxicity of systemic treatments for uHCC and we provided recommendations for monitoring and managing liver-related adverse events (AEs).

| First line
Drug-induced liver injury (DILI) represents one of the most common AEs in HCC patients undergoing systemic therapies. Even though TKI are generally more tolerated than other chemotherapies, they have a broad spectrum of AEs, including among others liver, skin, cardiovascular system and thyroids disorders. 18 In the SHARP trial, 7 602 HCC patients with preserved liver function were included and were randomized to receive sorafenib or placebo. Most of the included patients (95%) had Child-Pugh class A (284 in the sorafenib group vs 297 in the placebo group). The primary outcome of the trial was overall survival (OS) and the trial was prematurely interrupted for efficacy: median OS was 10.7 months, as compared with 7.9 months in the placebo group (HR 0.69; 95% CI 0.55-0.87). The incidence of AEs related to treatment was 80% in the sorafenib group and 20% in the placebo group. A similar discontinuation rate as a result of treatment was found in the two groups (38% vs 37%). Liver dysfunction was one of the most frequent treatment-related AEs (trAEs) leading to discontinuation in patients treated with sorafenib (5%). Among the most common serious AEs of any cause, liver dysfunction (7% vs 5%) and ascites (5% vs 4%) were described. A similar rate of hepatobiliary events was found in the two groups (11% and 9%).
Subsequently, a trial with a similar design was conducted in the Asia-Pacific region, and similar results of efficacy and consistent safety profile were observed. 19 Overall, 226 patients with HCC, of whom 97% were Child-Pugh A, were assigned to receive sorafenib or placebo. Median OS was 6.5 months in the sorafenib group, compared with 4.2 months in the placebo group, with an HR similar to that observed in the SHARP trial (HR 0.68 95% CI 0.50-0.93). Anyway, the two cohorts of patients enrolled showed significant differences in treatment duration (5 months for SHARP trial vs 3 months for Asia-Pacific), likely reflecting nuances in the severity of the background liver disease associated with liver cancer. 20 It should be underlined that the underlying chronic liver disease has a significant impact on the feasibility, the efficacy and the safety of treatments for HCC, but the need for strict selection criteria in terms of liver function in registrative trials prevents a true evaluation of survival benefit in the whole cirrhotic population. 21 After sorafenib approval, a number of real-life experiences described the safety of sorafenib and validated its effectiveness in different settings. 20,22,23 Randomized clinical trials (RCTs) obviously are the best level of evidence because of the precise design of this study and the accurate evaluation of drug efficacy and safety profile, 24 but they could fail to detect adverse events that could be severe because of the short-term follow-up and the inclusion of selected patients. 25 This is particularly true in the setting of cirrhosis, in which patients treated in real-life clinical practice could be different from those included in clinical trials. Indeed, more comprehensive and complementary findings often derive from properly designed observational studies giving a better assessment and a more widespread understanding of both efficacy and safety.  proved albumin, bilirubin, age and ECOG to be significant baseline predictive factors of early discontinuation of sorafenib because of intolerance. Although this field-practice study showed sorafenib to be a safe and effective treatment for uHCC, a higher rate of drug discontinuation compared to the SHARP trial caused by liver impairment was defined, probably caused by the higher presence of Child-Pugh B patients. A subsequent real-life study from the same group showed that discontinuation of sorafenib because of worsening liver function and cirrhosis decompensation represented the worst-case scenario with a median post-sorafenib survival (mPSS) of 1.8 months compared to patients who discontinued sorafenib for progressive disease or for AEs, respectively, with an mPSS of 4.6 months and 7.3 months. 20 GIDEON study 23 was a prospective, observational registry study evaluating the safety of sorafenib and treatment practices in a large cohort of 3275 patients. Enrolled patients were in Child-Pugh class A in 61%, Child-Pugh B in 21% and Child-Pugh C in 2% of cases. Although sorafenib dosing was similar irrespective of Child-Pugh score and the initial sorafenib dose and the proportion of patients with a dose reduction or increase were also comparable, the median duration of treatment was longer in Child-Pugh A. In line with previous studies, median overall survival (OS) was longer in Child-Pugh A patients (13.6 months) than in Child-Pugh B patients (5.2 months) and Child-Pugh C patients (2.6 months), but the rate of the most common adverse events (AEs) was also broadly comparable between Child-Pugh groups. Anyway, even if the global incidence of AEs was similar between different Child-Pugh classes, higher rates of serious AEs were observed in the Child-Pugh B population. The incidence of serious AEs increases with the deterioration of liver function. Moreover, a higher rate of AEs leading to definitive drug discontinuation was shown (40%) in Child-Pugh B patients, compared to Child-Pugh A (29%). Liver dysfunction secondary to sorafenib treatment was comparable among the different subgroups of Child-Pugh classes. Grades 3 and 4 liver dysfunction have been reported only in 1% of patients taking sorafenib [10]. It is worthy to note that the higher rates of AEs in Child-Pugh B and C patients might be due not only to intrinsic drug toxicity but also to the tumour progression and on the basis of poor liver function when the TKI was started.
It should be underlined that the outcomes of sorafenib improved over time and that sorafenib arms representing the control group in most recent clinical trials 8,9,28 displayed a higher OS compared to previous trials. The reasons for these observations could be related to better management of sorafenib dosing, better management of AEs and better accuracy of prediction of liver decompensation. Regarding the sorafenib dose, an Italian field-practice study showed that the adjustment of sorafenib dose to 400 mg/day had higher cost-effectiveness compared to the standard dose of 800 mg/day. 22 These data were subsequently confirmed by a larger study on 4903 patients from Veterans Health Administration demonstrating that the dose adjustment was associated with a lower rate of discontinuation because of AEs and with a similar OS compared to standard dose. 29 These findings are not surprising, considering that the standard dose was borrowed from evidence coming from patients with renal cancer and that the underlying liver dysfunction could modify sorafenib pharmacokinetics. Reduced liver function results in impaired drug metabolism and reduced albumin production that can increase the plasma concentration of TKIs, reducing their tolerability. 26,[30][31][32] Congruently with the above data, it is important to precisely assess hepatic function in patients with HCC before and during treatment with TKIs. The albumin-bilirubin (ALBI) grade has been evaluated in the setting of HCC. 33 Compared to Child-Pugh class, it could more accurately predict mortality without the need for subjective determinants of liver failure such as ascites and encephalopathy. 34 ALBI grade resulted in good objective hepatic reserve estimation across each BCLC stage of HCC. Also in patients treated with sorafenib, ALBI grade was shown to be a better tool than Child-Pugh. 35 Finally, in 2018, the non-inferiority of lenvatinib, compared to sorafenib was demonstrated in terms of OS (13.6 months vs A worsening of the ALBI grade was also observed during lenvatinib. 37 Similarly, a study conducted on 156 cirrhotic patients further showed that ALBI grade was better than Child-Pugh score in predicting TTP, and only hepatic reserve function at the time of lenvatinib introduction was a significant risk factor for a decline to worse Child-Pugh class. 38

| Second-line
Several drugs failed to show a significant OS benefit compared to placebo after the first-line progression. According to the RESORCE trial, 10 Regorafenib was the first approved second-line treatment in patients who progressed to sorafenib. Included patients must have tolerated sorafenib (at least 400 mg daily for at least 20 of the 28 days before discontinuation) and received their last sorafenib dose within 10 weeks of randomization. A well-tolerated safety profile was described and the most common Grade 3 or 4 AEs included hypertension, hand-foot skin reaction, fatigue and diarrhoea.
All the patients in the regorafenib group experienced at least one treatment-emerging AE. AST elevation was the most common AE leading to drug discontinuation, with no differences between the two groups. New ascites development rate of any grade was comparable in both arms; conversely, hepatobiliary disorders were more common in the placebo group (18% [34/193] one case of fatal liver failure has been described. 43 Data on hepatotoxicity associated with TKI treatments for uHCC reported in clinical trials are shown in Table 1. Ramucirumab is an anti-VEGF and a VEGF-receptor-2 monoclonal antibody and it was studied in two phase III RCTs. 12,13 In the REACH trial, ramucirumab was tested as second-line therapy in patients previously treated with sorafenib, either progressed or intolerant. In this study, 565 patients were randomly assigned to receive ramucirumab or placebo. Treatment with ramucirumab did not show a significant improvement in OS. Median OS for ramucirumab was 9.2 months vs 7.6 months in the placebo group (HR 0.87 [95% CI 0.72-1.05]; P = .14]. Child-Pugh B patients were initially enrolled in this study, although an increased rate of liver adverse events was noted in this subgroup, therefore protocol was amended not to include patients beyond Child-Pugh A and those patients already enrolled were not included in the intention-to-treat analysis. Ascites We recommend assessing liver function before and during TKI treatment with physical examination (to detect clinical signs of cirrhosis decompensation), liver function tests (to early capture changes in ALBI grade and Child-Pugh score), abdominal ultrasound (to detect macrovascular invasion or mild/moderate ascites) and upper endoscopy (to assess portal hypertension) (see Table 2).
When possible, etiological treatments for underlying liver disease should be considered, such as nucleos(t)ide analogues (NUCs) for HBV, direct-acting antiviral agents for HCV and alcohol withdrawl.
When starting sorafenib in patients with more advanced liver disease (ie Child-Pugh class B without ascites), the dose adjustment to 400 mg/d may be considered, together with a toxicity-guided approach. Transient worsening of transaminases or bilirubin may be managed with dose adjustment during TKI treatment, while significant changes in ALBI grade and Child-Pugh score or the occurrence of decompensating events should lead to treatment interruption and to the identification and treatment of any precipitant factors. The mechanism of action of these molecules is to restore T-cell activity against tumour. 44 They had an important role in maintaining immune homeostasis and the alteration of this stability can lead to a range of immune-related AEs (irAEs). The most commonly involved organs are the skin, liver, gastrointestinal tract and endocrine glands.

| HEPATOTOXI CIT Y OF I CI -BA S ED THER APIE S IN UNRE S EC TAB LE H CC
Adverse events against the liver are defined as hepatic irAEs or immune-mediated hepatitis. [45][46][47] While irAEs have been extensively studied in solid neoplasms other than HCC treated with ICIs, data on hepatotoxicity of these drugs in patients with HCC and cirrhosis are less conspicuous. As previously described, patients with cirrhosis and HCC treated with systemic therapies, including ICIs, can have multiple causes of liver injury and the differential diagnosis is often challenging, leading to different therapeutic approaches. 48 ICI trials showed that the incidence of liver toxicity is mainly linked to the type of drug, the posology and the combination of them. 47 The incidence of liver toxicity is higher in patients who received combination therapy than in those under monotherapy, but it remains lower compared to other organ toxicities. 48
a higher incidence of elevations in transaminases and bilirubin levels. show significant superiority of sorafenib in terms of OS. 28 In the real-world setting, a multicentre observational study confirmed the safety profile of nivolumab observed in clinical trials 50 showing an overall trAE rate of 26.6% and grade > trAE rate of 6.4%, without significant differences in trAE rates between patients in Child-

| ICI-based combination therapies
The combination of an anti-PDL-1 (atezolizumab) and an anti-VEGF (bevacizumab) was assessed in phase III Imbrave 150 trial. 9   Data on hepatotoxicity associated with ICI-based treatments for uHCC reported in clinical trials are shown in Table 3.

| Management of hepatotoxicity associated with ICIs
Immune-mediated hepatitis associated with ICIs is a diagnosis of exclusion. The causality between the drug exposure and the onset of liver injury can be tested through the Roussel Uclaf causality assessment method that has been validated in both retrospective and prospective studies. 57   We recommend that viral hepatitis serology, transaminase levels (ALT and AST), bilirubin, albumin and INR should be tested prior to starting therapy with checkpoint inhibitors ( could be useful to improve the collaborative decision-making process, involving physicians and patients, in the choice of the optimal tailored treatment for each patient. 67 In conclusion, it is expected that the number of effective ICI-based combination therapies will increase in the next future, potentially leading to their use also in neoadjuvant or adjuvant settings. Nevertheless, awareness of their hepatotoxicity profile and the management of liver-related AEs are increasingly relevant. A more accurate assessment of the safety of novel systemic treatments for HCC and an improved diagnostic and therapeutic management of AEs are urgently needed to better inform clinical decision making and to optimize the outcomes of patients with HCC.

ACK N OWLED G EM ENTS
All the authors take full responsibility for the preparation of the manuscript and approved the final draft manuscript.