Comparative efficacy and safety of atezolizumab and bevacizumab between hepatocellular carcinoma patients with viral and non‐viral infection: A Japanese multicenter observational study

Abstract Aim This study compared the efficacy and safety of atezolizumab and bevacizumab (Atez/Bev) in patients with viral and non‐viral infection in clinical settings. Methods We conducted the retrospective cohort study of 323 BCLC stage B or C hepatocellular carcinoma (HCC) patients with Child‐Pugh class A, and a performance status of 0 or 1 who started Atez/Bev from September 2020 to December 2021 at 22 institutions in Japan. Patients with viral infection was defined as those who were either serum anti‐HCV‐ Ab or HBs‐Ag‐positive, while patients with non‐viral infection was defined as those who were both serum anti‐HCV Ab‐ and HBs‐Ag‐negative. We constructed a propensity‐score‐matched cohort to minimize the risk of observable potential confounders. Results Propensity score matching produced 126 matched pairs for patients with viral versus non‐viral infection. After matching, the significant differences in baseline demographic features did not exist between the two groups. The objective response rate was 20.6% and 24.6% in viral‐ and non‐viral‐related HCC patients, respectively, without a significant difference (p = 0.55). The disease control rate was not also significantly different (68.3% vs 69.0%, p = 1.00). The median progression‐free survival was 7.0 months (95% confidence interval [CI] 6.0–9.6) and 6.2 months (95% CI 5.1–7.8) in patients with viral and non‐viral infection, and the 12‐month survival rates were 65.5% (95% CI 50.8–76.8) and 71.7% (95% CI 57.3–81.9) in those with viral and non‐viral infection, respectively, which were not significantly different (p = 0.33, p = 0.38). No significant difference in treatment‐related adverse events was found between the two groups. Conclusions Our etiology‐based study demonstrated that Atez/Bev showed good efficacy and safety for HCC patient with non‐viral infection as well as those with viral infection.


| INTRODUCTION
(95% confidence interval [CI] 0.52-1.60), 0.51 (95% CI 0.32-0.81), and 0.43 (95% CI 0.22-0.87) for non-viral-, hepatitis B virus (HBV)-, and hepatitis C virus (HCV)-related HCC patients, respectively. 1 Based on the updated report concerning the efficacy and safety of this RCT, the median OS was 17.0 months (95% CI 11.7-22.8) and 18.1 months (95% CI 11.7-26.3) in patients with non-viral infection treated with Atez/Bev and sorafenib, respectively, resulting in an HR of 1.05 (95% CI 0.68-1.63) for death with Atez/Bev. 4 Two meta-analyses of immune checkpoint inhibitor (ICI) trials showed that the pooled HR for the OS in viralrelated HCC patients receiving an ICI was lower than in those receiving standard-care treatment. 5,6 These two studies indicated that patients with viral infection benefitted from ICI treatment, while those with non-viral infection did not. 5,6 Our previous study 7 reported the initial experience of Atez/Bev treatment, showing that the initial therapeutic response (radiologically evaluated at 6 weeks after the starting Atez/Bev) was not significantly differed between the patients with viral and non-viral infection (10.8% vs 10.4%, p = 0.96, n = 171). However, we did not evaluate the PFS, OS and adverse events (AEs) owing to the short observation period. In addition, our previous report 7 did not correct imbalances in baseline characteristics between viral-, and non-viral-related-HCC patients. Accordingly, whether or not underlying liver etiologies can predict the therapeutic response to Atez/Bev remains unclear.
In the current study, we expanded our cohort with long observation period, and compared the therapeutic outcome of Atez/Bev in patients with viral and nonviral infection in clinical settings using propensity score matching.

| Patient population
In the present retrospective cohort study, 399 HCC patients who received Atez/Bev from September 2020 to December 2021 were included. The study enrolled patients at 22 different institutions in Japan. We excluded patients with a poor liver function (that is, Child-Pugh class B or C; n = 23), those with Barcelona Clinic Liver Cancer (BCLC) stage A or D (n = 45), and those with a performance status (PS) ≥2 (n = 8). Accordingly, we included the remaining 323 HCC patients with Child-Pugh class A, BCLC stage B or C, and PS ≤1 in the present study ( Figure 1). HCC was diagnosed based on histological data obtained during the clinical course or the typical findings on radiological imaging including enhanced dynamic computed tomography (CT) and dynamic magnetic resonance imaging (MRI). 8

| Definitions of viral and nonviral infection
When participated patients tested positive for anti-HCV antibody (anti-HCV Ab), cause of HCC was attributed to chronic HCV infection. When patients tested positive for hepatitis B virus surface antigen (HBs-Ag), HCC was attributed to chronic HBV infection. Patients with viral infection were defined as those with either anti-HCV Ab or HBs-Ag seropositivity, while patients with non-viral infection were defined as those with both serum anti-HCV Ab and HBs-Ag seronegativity. If patients with non-viral infection had a presence and/or history of habitual alcohol consumption (≥60 g/day), HCC was attributed to chronic alcoholic liver injury. The diagnosis of non-alcoholic fatty liver disease (NAFLD) was performed based on the pathological examination or clinical findings; that is, findings of fatty liver, medical interview of low habitual alcohol intake (<30 g/day in males and <20 g/day in female), and metabolic comorbidities.

| Treatment procedure
Before Atez/Bev treatment, we evaluated the preserved liver function, tumor extension, and PS to determine the BCLC stage. 2 In addition to the Child-Pugh classification, we also used the albumin-bilirubin (ALBI) score 9 and modified ALBI (mALBI) grade 10 to assess the liver function. Tumor extension was assessed based on findings of radiological imaging including dynamic CT or MRI.
Patients intravenously received 1200 mg of atezolizumab followed by 15 mg/kg of bevacizumab on the same days F I G U R E 1 Flow diagram of the eligible hepatocellular carcinoma patients every 3 weeks. Atez/Bev was continued until disease progression was observed or unacceptable AEs developed. The tumor response after initiation of Atez/Bev was evaluated by Response Evaluation Criteria in Solid Tumors version 1.1 (RECIST v.1.1). We evaluated the degree of AEs according to The Common Terminology Criteria for Adverse Events version 5.0. When treatment-related AEs were observed, we interrupted or discontinued each drug according to the guidelines for Atez/Bev treatment provided by the manufacturer.

| Statistical analyses
Categorial variables were presented as the counts (percentage) and numerical variables as median [interquartile range]. The differences between categorial variables were calculated using the χ 2 test, and Fisher's exact, if appropriate. Numerical variables were compared using Mann-Whitney U-test. Given the differences in the patient characteristics associated with efficacy and safety between the eligible patients with viral and non-viral infection, a propensity-score-matched cohort was identified to minimize the risk of observable potential confounder. The propensity score for each patient was estimated by generating a logistic regression model. The matching variables included the age, gender, BCLC stage, and mALBI grade. We used 1:1 nearest neighbor matching without replacement and adopted caliper width of 0.2 of standard deviations.
The PFS was measured between the starting date of Atez/ Bev and the date of disease progression or death, whichever came first. The OS was calculated from the initiation date of Atez/Bev to death or to the date of last follow-up. The survival outcome was assessed using the Kaplan-Meier method and the log-rank test was used to compare the differences between subgroups. The up-to-seven criteria was defined as the sum of largest tumor size (cm) and number of tumors in BCLC intermediate stage HCC patients. All of the reported p-values were two-sided. If it was less than 0.05, a result was considered significant. All statistical analyses were conducted using EZR V. 1.53 (Saitama Medical Center, Jichi Medical University, Saitama, Japan). 11
The patients with viral infection were younger, had a lower body mass index (BMI), a lower platelet count, higher level of albumin, better ALBI score, and better mALBI grade than those with non-viral infection. Moreover, viral-related HCC patients tended to more often have BCLC intermediate stage than non-viral-related HCC (p = 0.095).
Propensity score matching produced 126 matched pairs for patients with viral versus non-viral infection. After matching, the significant differences in baseline demographic features did not exist between the two groups except for in the BMI and platelet count (Table S1).
In the analysis of the PFS, 128 events (50.8%) were detected during the observation. The median PFS was 7.0 months (95% CI 6.0-9.6) in patients with viral infection whereas it was 6.2 months (95% CI 5.1-7.8) in patients with non-viral infection, respectively, with an HR of 0.96 (95% CI 0.52-1.76). No significant difference between the two groups was not found (p = 0.33; Figure 2). After the end of Atez/Bev treatment, 38 (70.4%) and 56 (78.9%) patients with viral and non-viral infection received any subsequent treatments. The proportion of any subsequent  Figure 3).
Next, we investigate the PFS and OS in the patients treated with Atez/Bev as the first-line treatment and later-line treatment, finding that the significant extents were not differed between viral-, and non-viral-related HCC patients. The Kaplan-Meier curves are shown in Figures S1 and S2a,b. Furthermore, we analyzed the PFS and OS according to each etiology of underlying liver disease, noting no significant differences by etiology (HCV, HBV, alcohol, NAFLD and others). The PFS and OS curves are shown in Figure S2. The results of comparison of NAFLD and other etiologies (HCV, HBV, alcohol, and others) were shown in Figure S4.  Abbreviations: CI, confidence interval; CR, complete response; NE, non-evaluable, PD, progressive disease; SD, stable disease; PR, partial response.

F I G U R E 2
The progression-free survival (PFS) of patients with viral and non-viral infection. The median PFS was 7.0 (95% confidence interval [CI] 6.0-9.6) months and 6.2 (95% CI 5.1-7.8) months in viral and non-viral patients, respectively, with a hazard ratio of 0.96 (95% CI 0.52-1.76). Significant differences did not note between the two groups (p = 0.33) The most commonly experienced AEs in propensityscore-matched cohort were protein urea, secondary to fatigue, appetite loss, hypertension, and liver injury during the observation period. We compared the incidence and severity of each AE in both two groups, resulting in no significant differences (Table 3).

| DISCUSSION
We compared the baseline characteristics between the patients with viral and non-viral infection, revealing that patients with viral infection were significantly younger with a better liver function than those with non-viral infection. In addition, the viral-related-HCC patients tended to have a higher percentage of BCLC advanced stage than those with non-viral infection. We conducted propensity score matching to correct for these imbalances in baseline characteristics, demonstrating that significant differences in pretreatment factors associated with efficacy and safety did not note between the two groups. In addition, we also investigated the PFS and OS in patients receiving Atez/Bev as the first-line treatment and later treatment, indicating that there were not significant differences between viral-, and non-viral HCC patients. We also compared the PFS and OS according to each etiology of liver disease, showing that the significant extent was not found.
The present etiology-based study showed that Atez/ Bev had good efficacy and safety for HCC patient with non-viral infection as well as those with viral infection. This indicates that underlying liver diseases did not affect the clinical outcome of Atez/Bev treatment. A real-world data indicated that underlying liver diseases did not affect the clinical outcome of sorafenib. 12 As for lenvatinib, whether or not liver etiology influenced on the therapeutic outcome remains controversial. Two real-world studies reported that liver etiology did not associate with overall survival 13,14 while another two studies showed the overall survival is better in patients with non-viral infection than in those with viral infection. 15,16 However, no studies had compared Atez/Bev in patients with viral and no-viral infection in real-world settings. As far as we know, this is the first study investigating the association between the clinical outcome of Atez/Bev and the etiology of liver disease based on real-world data.
Pfister et al. 5 carried out a meta-analysis of three RCT trials (CheckMate-459, 17 Imbrave150, 1 and KEYNOTE-240 18 ), indicating that patients with viral infection benefitted from ICI treatment compared to standard-care treatment, while those with non-viral infection did not benefit from this treatment. Haber et al. 6 performed a systematic review of phase-three RCTs (reported from 2002 to 2020) and examined the association between the liver etiology and outcome of ICI treatment and between the liver etiology and outcome of tyrosine kinase T A B L E 3 Incidence and degree of adverse events in viral-, and non-viral-related HCC patients after propensity score matching inhibitor (TKI)/anti-VEGF therapies. The efficacy of ICIs between viral-and non-viral-related HCC significantly differ between viral-and non-viral-related HCC, while the liver etiology was not associated with the outcome of TKI/ anti-VEGF therapies. 6 The present results seem to be in disagreement with these previous studies and the results of the subgroup analysis of Imbrave150. 1,4 While we cannot concretely explain this difference, some possible reasons may be hypothesized. First, the treatment method (ICI monotherapy/ICI plus VEGF inhibitor combination therapy) and treatment settings (first-, and second-line treatment) differed among the three RCT trials. 1,17,18 Second, ICI treatment was compared to sorafenib or best supportive care (BSC) in three RCT trials, 1,17,18 whereas we compared the therapeutic outcome of Atez/Bev in patients with viral and non-viral infection. This means that the control group differed between the three RCT trials (sorafenib or BSC) and the present study (non-viral patients). Third, more than 70% patients with viral infection received antiviral therapy in the current study. The persistent liver inflammation after HCV eradication was mainly associated with hepatic steatosis, 19,20 which were a risk factor of carcinogenesis. 21 Hence, antiviral therapy may affect the present results. Fourth, while the updated subpopulation analysis of Imbrave150 4 showed that the OS in non-viral patients receiving Atez/Bev was comparable to the value in those receiving sorafenib (HR 1.05, 95% CI 0.68-1.63), details concerning baseline characteristics were not clarified. Among non-viral-related HCC patients, the statistical comparison of baseline characteristics between Atez/Bev and sorafenib groups will be required to assess the therapeutic outcome. A further study with a greater focus on the baseline characteristics stratified by the liver disease etiology is therefore suggested.

Variables
Pfister et al 5 reported that NASH-driven HCC was not benefitted from immunotherapy targeted at PD-1 due to impaired immune surveillance, which were presumed to disagree with the present results. This difference is possibly due to the presence of combined anti-VEGF therapy. VEGF inhibited dendric cell maturation, reduced T-cell tumor infiltration, and increase immune-suppressive cells such as myeloid derived suppressor cells and regulatory T cells in the tumor microenvironment, resulting in immune-suppressive effects. 22 In addition, bevacizumab not only enhanced the efficacy of immunotherapies but also showed the antitumor effect for HCC. 23 A further study will be required to investigate whether or not NASHdriven HCC respond to combination of immunotherapy and anti-VEGF therapy.
Recently, the HIMALAYA phase 3 study, which examined the efficacy and safety of tremelimumab and durvalumab in comparison to sorafenib, reported that tremelimumab and durvalumab achieved a good survival in HBV patients (HR 0.64, 95% CI 0.48-0.86), and nonviral patients (HR 0.74, 95% CI 0.57-0.95) but not in HCV patients (HR 1.06, 95% CI 0.76-1.49). 24 According to the PFS and preliminary interim OS results of COSMIC-312 trial 25 evaluating the therapeutic efficacy of cabozantinib and atezolizumab versus sorafenib in first-line treatment, cabozantinib and atezolizumab improved OS in HBV patients (HR 0.53, 95% CI 0.33-0.87) but did not in HCV patients (HR 1.10, 95% CI 0.72-1.68) or non-viral patients (HR 1.18, 95% CI 0.78-1.79). Given these results, further analyses will be needed to investigate whether or not ICI treatment was unlikely to induce a response in non-viralrelated HCC patients.
After matching, the platelet count and BMI were lower in viral patients than in non-viral patients, probably due to disease-specific differences. In general, the platelet count was lower in HBV or HCV patients than that in nonB-, nonC-patients. Indeed, the cut-off value of HCV-related cirrhosis 26 was lower than that of NAFLD-related cirrhosis. 27 Regarding the BMI, this difference was because the carcinogenesis of non-viral-related HCC was strongly associated with metabolic disease, including obesity, hypertension, and diabetes mellitus, resulting in a higher BMI in non-viral HCC patients than in viral HCC patients. Furthermore, while the BCLC stage, 28 mALBI grade, 28 early AFP response, 29,30 neutrophil-to-lymphocyte ratio, 31,32 and early bev interruption 33 were reported to be associated with the outcome of Atez/Bev, whether or not the platelet count and BMI influenced the outcome remains uncertain. Accordingly, we believe that it is not appropriate to correct the imbalance in the platelet count and BMI between the two groups.
In the current study, about 70% patients received any subsequent therapies after the end of Atez/Bev treatment. Given that post-progression survival was strongly correlated with OS, 34 subsequent therapies play an important role in prolongation of survival. While the proportion of receiving the subsequent therapies after sorafenib 35 and lenvatinib 36 are reported to be about 50%, the previous real-world studies 37,38 showed that about 70% patients received it after Atez/Bev treatment, which was consistent to the present study.
While there was no RCT comparing the efficacy and safety of Atez/Bev and lenvatinib in patients with unresectable HCC patients, comparative studies based on realworld data have recently been reported. Kim et al reported that the OS and PFS was not significantly differed between the Atez/Bev and lenvatinib while the objective response was better in lenvatinib-treated patients than Atez/Bevtreated patients with non-viral infection. 39 Two studies reported by Hiraoka et al 40 and Maesaka et al 41 showed that there were no differences in the OS and PFS between patients treated with Atez/Bev and lenvatinib. Because the number of the participated patients was small and the observation period was short, further comparative studies with sufficient cases and follow-up period were warranted.
There are some important limitations in the current study. First, this study was analyzed in a retrospective manner. Second, although we conducted propensity score matching, some unintentional bias may still have been included. Third, there is the relatively small number of objects with a short observation period. For example, although the statistical significance was not observed, the survival curve might be distinct in patients receiving Atez/ Bev as the first-line treatment ( Figure S1). Including more patients with a long-term follow-up might influence the present findings. Fourth, there were about 70% patients who received antiviral therapy in the current study. Because steatohepatitis is associated with the carcinogenesis after the eradication of HCV, as we mention above, antiviral therapy might affect the present results. Another limitation is the high percentage of patients receiving subsequent therapies after the Atez/Bev treatment. In conclusion, the present etiology-based study revealed that Atez/Bev showed good efficacy and safety for HCC patient with non-viral infection as well as those with viral infection.