Evaluating the risk-benefit ratio of immunotherapy according to liver-functional reserve in advanced HCC: the dark side of the moon.

Systematic review and meta-analyses aim to accumulate, synthetize, and evaluate evidence across individual studies, with the goal of resolving uncertainties, reducing biases, and informing practice, thus contributing as fundamental tools for evidence-based medicine. However, limitations of meta-analyses, as differences in the design of the included studies, sample size, baseline severity of illness, follow-up (as well as other potentially important confounders) can preclude their placement at the top of the pyramid of evidence.1 In this issue of Hepatology, El Hajra et al.2 aimed to shed light regarding use of immune checkpoint inhibitors (ICIs) in HCC patients with impaired liver function, by conducting an aggregate data meta-analysis of studies including HCC patients with Child-Pugh B/C cirrhosis receiving ICI-based systemic treatments. The authors pooled data from 15 studies (all but 1 observational) including 2311 patients treated with atezolizumab plus bevacizumab, nivolumab plus iplimumab or monotherapy with nivolumab or pembrolizumab. They found a restricted mean survival time of about 8 months for patients with Child-Pugh B/C, with a significantly higher risk of death in these patients compared with those with Child-Pugh A (HR: 1.65, 95% CI, 1.45–1.84). However, all the analyses were affected by a high and statistically significant heterogeneity, and although authors tried to explore it by sensitivity and meta-regression analyses, unfortunately they failed to identify study-level or patient-level covariates able to fully explain the observed heterogeneity. The advent of ICIs has recently revolutionized the landscape of systemic treatments for advanced HCC. Despite antiprogrammed death 1/antiprogrammed death ligand 1 monotherapy showed a promising clinical activity in early phase clinical trials with high objective response rate reaching 20%, it failed to show a significant superiority in overall survival compared with tyrosine-kinase inhibitor sorafenib in phase 3 studies with the exception of durvalumab and tislelizumab showing a noninferior overall survival.3–6 Conversely, combination strategies including antiprogrammed death 1/antiprogrammed death ligand 1 in association with anti-VEGF (atezolizumab plus bevacizumab), with anti-CTLA4 (durvalumab plus tremelimumab or with multikinase inhibitor showed a significant survival benefit in comparison with sorafenib monotherapy in phase 3 randomized controlled trials (RCTs).5,7,8 Although the effectiveness of these combination strategies is clearly higher compared with multikinase inhibitors, it has not yet demonstrated if the liver toxicity profile of ICIs is significantly different from that of multikinase inhibitors; moreover, it is unclear if ICIs are a safe treatment option in patients with impaired liver function, raising doubts regarding the feasibility of these treatments in a real-world setting including patients with clinical characteristics that deviate from stringent inclusion criteria of RCTs. This is particularly relevant for patients with cirrhosis and more severe liver dysfunction (ie, Child-Pugh B or C) usually excluded from trials, as it is difficult to accurately ascertain the cause of death for competing risk related to impairment of liver function. Although it has been suggested that ICI treatment could have a lower negative impact on liver function compared with multikinase inhibitors, up-to-date only a single-arm trial specifically addressed this issue: the study shows a comparable safety profile between patients in Child-Pugh A and those in Child-Pugh B 7–8 without ascites or HE, with these latter having a worse survival.9 Differently from many other neoplasms, in whom prognosis and treatment selection are largely dictated by tumor stage at the time of diagnosis, the clinical scenario of HCC, that is frequently superimposed on a chronic liver disease, results significantly more complex, from both clinical and methodological point of view. It is well known that the functional impairment of underlying liver disease has a significant impact on prognosis, with liver decompensation being the major driver of death in patients with successfully treated early-stage HCC.10 Indeed, in these patients, liver decompensation is associated with a risk of death that is about 3 times higher than that related to HCC recurrence. The impact of liver decompensation on the prognosis of HCC patients is even more relevant in the advanced setting, which includes patients with heavier tumor burden (encompassing macrovascular invasion), with more compromised performance status or with a longer history of chronic liver disease and previous surgical and locoregional treatments for HCC.11,12 It is therefore clear that death is affected by 2 main competing risks related to HCC progression and hepatic decompensation.13 Although death in Child-Pugh A patients with HCC is usually related to cancer progression, it can be difficult to accurately ascertain the cause of death in HCC patients with Child-Pugh B or C. These latter patients with more impaired liver function have been usually excluded from phase 3 trials of systemic treatments, to virtually decrease the “background noise” related to the competing risk of hepatic decompensation. Moreover, RCTs conducted in other oncological settings did not usually include patients with any degree of hepatic impairment or cirrhosis and it is not possible to extrapolate conclusions about the risk of liver function deterioration in patients with cirrhosis from studies conducted in patients with other solid cancers. Therefore, if the inclusion of only well-compensated patients in RCTs avoids the competing risk for death related to hepatic decompensation when estimating the survival of patients with HCC and cirrhosis remains to be established. It should be noted that the risk of liver decompensation has never been specifically and prospectively assessed as a clinical endpoint in RCTs, which usually report liver-related events as adverse events, defined, and classified according to the National Cancer Institute-Common Terminology Criteria for Adverse Events (CTCAE).14 However, CTCAE system is affected by several limitations in the setting of cirrhosis, as it merely describes the increase of transaminases and/or bilirubin during systemic treatment, without considering the severity of baseline liver dysfunction before starting therapy and without considering the occurrence of hepatic decompensation events, such as ascites, esophagogastric varices bleeding, jaundice, or HE. It is relevant as hepatic decompensation often occurs without a significant increase of transaminases and therefore it is not usually captured as adverse event in RCTs. Conversely, an increase of transaminases during ICI treatment, even if classified as severe according to CTCAE, does not translate automatically into a clinically relevant decompensation event. According to these considerations, the evaluation of liver toxicity of systemic treatments for advanced HCC in RCTs may be biased, especially in patients with more advanced liver disease at baseline, and it requires the application of well-known liver function measures (Child-Pugh, ALBI, MELD scores) in new standardized definitions and reporting systems for clinical trials in HCC. In such a complex clinical setting with different competing risks for complications and death related to HCC progression and/or hepatic decompensation, the inaccurate reporting of decompensation events could be an explanation for the heterogeneous and not always satisfactory surrogacy between radiology-based endpoints (ie, progression-free survival) and overall survival.15,16 It has been speculated that progression-free survival evaluation may be biased and overestimated when treatment toxicity leads to the dropout of a significant proportion of patients before that radiological progression could be documented (informative censoring bias).17 This could be the case of HCC setting, although the lack of data from RCTs on the dropout rate because of hepatic decompensation does not allow to reach firm conclusions. Finally, it should be considered that patients from real-world practice are drastically different from the well-selected populations enrolled in RCTs and with the increasing availability of ICI-based combinations for patients with advanced HCC, data on the effectiveness and the safety of these treatments in patients with Child-Pugh B/C are sparse and controversial. In this line, the effort from the authors is noteworthy, as they tried to summarize evidence on a relevant and timely topic, which still represents a burning unmet medical need. Nevertheless, the significantly higher heterogeneity observed in this meta-analysis still hampers an evidence-based recommendation about the use of ICI treatment in patients with Child-Pugh B or C cirrhosis. In this line, we suggest a possible approach to allow a more precise evaluation of the risk-benefit profile of ICI treatment in patients with HCC and cirrhosis in future RCTs (Figure 1) and real-world studies. First, to report hepatic decompensation, as a clinical measure of safety, both as single (time from treatment start to hepatic decompensation) and composite (decompensation-free survival) endpoint, and to estimate survival by competing risks model (evaluating death for HCC progression and/or hepatic decompensation) is urgently needed in both RCTs and prospective real-world studies. In this line, a phase 3b trial evaluating safety as primary outcome in patients treated with atezolizumab plus bevacizumab is ongoing (NCT04487067). Second, an individual patient-data meta-analysis is needed to explore and to explain the heterogeneity observed in the meta-analysis by El Hajra and colleagues, trying to identify a subgroup of patients with liver dysfunction that could benefit from ICI treatment according to a risk profile encompassing etiology of liver disease, previous or current liver decompensation, severity of portal hypertension, performance status and HCC, and liver function evolutionary stages. Finally, we believe in the need of well-conducted real-world studies that, despite the presence of several biases, offer a chance to properly unravel the net-health benefit in terms of effectiveness and toxicity of ICI-based therapies, when administered in real-life patients carrying comorbidities.FIGURE 1: Proposal for a trial design for advanced HCC systemic treatment that includes liver function as stratification factor at baseline and as secondary aim point. Abbreviations: AFP, alpha-fetoprotein; CTCAE, common terminology criteria for adverse events; ECOG-PS, Eastern Cooperative Oncology Group; HCC, hepatocellular carcinoma; MELD, model for end-stage liver disease; ORR, objective response rate; OS, overall survival; PFS, progression-free survival.