Real‐World Outcomes of Atezolizumab–Bevacizumab in Hepatocellular Carcinoma: The Prospective French CHIEF Cohort

ABSTRACT

Background and Aims

Atezolizumab–Bevacizumab (AtezoBev) was the first immunotherapy approved for hepatocellular carcinoma (HCC) in France, with initial trials primarily involving patients with viral‐related liver disease. This prospective study aimed to evaluate the efficacy of AtezoBev in a French HCC population predominantly affected by non‐viral liver disease.

Methods

Data from 545 HCC patients treated with AtezoBev as first‐line systemic therapy were collected from 32 French centres in the CHIEF cohort between July 2020 and January 2023. Kaplan–Meier analysis evaluated overall survival (OS) and progression‐free survival (PFS), while log‐rank tests assessed the impact of baseline characteristics.

Results

Median age was 69, with 81% Child‐Pugh A and 19% Child‐Pugh B. Liver disease was primarily alcohol‐related (30%) or viral (16%), with mixed aetiology with at least alcohol consumption in 58%. At AtezoBev initiation, 72% of cases were treatment‐naive, 31% were BCLC‐B and 64% were BCLC‐C. Median OS was 23.1 months, with a 12‐month survival rate of 66%. OS was higher in BCLC‐B patients (27.8 months) compared to BCLC‐C (17.2 months, p = 0.0043) and in Child‐Pugh A (26.4 months) compared to Child‐Pugh B (10.6 months, p < 0.001). Median PFS was 5.2 months, with BCLC‐B patients showing significantly longer PFS (6.7 months vs. 3.7 months for BCLC‐C, p = 0.05).

Conclusion

Real‐world data from the CHIEF cohort demonstrate AtezoBev's effectiveness in a large French HCC population, showing survival and response rates comparable to the IMbrave150 study. These findings validate AtezoBev as effective in routine practice across diverse clinical profiles.

Summary.

• This prospective multicentre study in the CHIEF cohort evaluated the real‐world efficacy of Atezolizumab‐Bevacizumab (AtezoBev) for hepatocellular carcinoma (HCC) in a French population with predominantly non‐viral liver disease.

• Among 545 patients, with 72% treatment‐naive, 31% in the intermediate stage (BCLC‐B), and 64% in the advanced stage (BCLC‐C), the median overall survival (OS) was 23.1 months, with a 12‐month OS rate of 66%.

• Outcomes were better for patients in the BCLC‐B stage and those with preserved liver function (Child‐Pugh A). Progression‐free survival (PFS) was also significantly longer in BCLC‐B patients.

• These findings align with the pivotal IMbrave 150 trial, affirming AtezoBev's effectiveness in routine clinical settings, even with diverse HCC etiologies and advanced disease features.

Abbreviations

95% CI

95% confidence interval

AFP

alpha‐fetoprotein

ALBI grade

albumin‐bilirubin grade

ALD

alcohol‐related liver disease

AtezoBev

atezolizumab bevacizumab

BCLC

barcelona clinic liver cancer

CT

computed tomography

EV

oesophageal varices

HCC

hepatocellular carcinoma

HR

hazard ratio

ICI

immune checkpoint inhibitors

INR

international normalised ratio

MASLD

metabolic dysfunction associated liver disease

MELD

model for end‐stage liver disease

MR

magnetic resonance

OS

overall survival

PFS

progression‐free survival

PT

prothrombin time

PVTT

portal vein tumour thrombosis

SD

standard deviation

1. Introduction

While the landscape of hepatocellular carcinoma (HCC) management has shown promising advancements with the introduction of Atezolizumab‐Bevacizumab (AtezoBev) combination therapy as first‐line in patients with advanced HCC [1, 2, 3], there remains a need for real‐life data to complement the existing evidence derived from clinical trials. Notably, the initial results from the IMbrave150 study have demonstrated encouraging outcomes, prompting a paradigm shift in the approach to advanced HCC treatment. However, the study predominantly enrolled patients with viral aetiology‐associated HCC [1, 2, 3], whereas the prevalence of metabolic hepatopathies, including alcohol‐related and nonalcoholic steatohepatitis (MASLD) in the French population, raises pertinent questions regarding the generalisability of these findings in western countries.

Real‐world data are crucial to appreciate the broader efficacy and safety profiles of AtezoBev in diverse patient populations, considering the varying underlying liver pathologies and the potential impact of these differences on treatment response. Furthermore, the translation of trial outcomes into routine clinical practice necessitates an understanding of the treatment's impact on patient quality of life, as well as its feasibility and tolerability in a less controlled, real‐world setting. Thus, a comprehensive assessment of the response, tolerability and quality of life outcomes within a real‐world context assumes paramount significance to inform clinical decision‐making and optimise patient care strategies.

In this prospective study, our objectives were: (i) to evaluate OS and PFS in a large prospective multicentric cohort of patients representing the French territory, the CHIEF cohort and (ii) to identify prognostic factors for the response to AtezoBev treatment.

2. Patients and Methods

2.1. Presentation of the CHIEF Cohort

A prospective cohort of patients with HCC in France (the ‘CHIEF’ cohort) was launched on September 1, 2019. The study sponsor was Amiens‐Picardie University Hospital. Thirty‐two centres are participating in this cohort study: 27 university tertiary centres (the hepato‐gastroenterology departments of CHU Amiens‐Picardie, CHU Bordeaux, CHU Bobigny Avicennes, CHU Clichy Beaujon, CHU Créteil Henri Mondor, CHU Grenoble, CHU Lille, CHU Lyon, CHU Montpellier, CHU Nancy, CHU Nice, CHU Rennes, CHU Rouen, CHU Toulouse, CHU Tours, CHU Paul Brousse Paris, CHU Poitiers and, since May, 2021 CHU Angers, CHU Brest, CHU Caen, CHU Dijon, CHU Guadeloupe Antilles Française, CHU Limoges, CHU Pitié‐Salpêtrière Paris, CHU Saint Antoine Paris, CHU Reims and CHU Besançon), four general hospitals (CHG Corbeille‐Essonnes, CHG Creil, CHG intercommunal de Créteil and CH Auxerre), and one private hospital (Hôpital Saint‐Joseph, Marseille).

The eligibility criteria for the CHIEF cohort encompassed patients with either a recent diagnosis of HCC or a previously diagnosed case of HCC who had initiated locoregional or systemic treatment within the past 6 weeks. Treatments included chemoembolization (TACE), selective internal radiation therapy (SIRT), sorafenib, regorafenib, cabozantinib, lenvatinib, ramucirumab and immunotherapy (atezolizumab, pembrolizumab, nivolumab and durvalumab), irrespective of the treatment line. Every patient was required to provide informed consent and be covered by social security. Patients concurrently enrolled in another clinical trial were also considered. Exclusion criteria comprised the presence of another life‐threatening condition, a different tumour type (such as cholangiocarcinoma or hepatocholangiocarcinoma), pregnancy, patients under 18 years of age, absence of consent for participation or the inability to obtain such consent.

The CHIEF cohort study, registered under ClinicalTrial.gov identifier # NCT04348838, falls under category 2, signifying minimal risks and constraints as per the Jardé law in France. Prior approval from the regional ethics committee (CPP Sud‐Est IV) was obtained, and the study adheres to the guidelines outlined by the French ‘Informatique et Liberté’ law (CNIL) with a classification of MR001 for data protection. Conforming to the principles of the Declaration of Helsinki 2000, the study strictly adheres to good clinical research practices. All participating patients were comprehensively informed and provided explicit consent for the confidential utilisation of their clinical and biological data.

2.2. Patients Selection Among CHIEF Cohort

Eligible patients for this study fulfilled the following criteria: (1) HCC diagnosed by histology or non‐invasive criteria using contrast‐enhanced computed tomography (CT) scan or magnetic resonance (MR) imaging according to EASL criteria (3); (2) patients not eligible for, or presenting progression after locoregional treatment; (3) proposal of a treatment with AtezoBev as first‐line therapy by the local tumour multidisciplinary meeting (1).

Exclusion criteria were: (1) a combined locoregional and systemic treatment; (2) a previous systemic treatment; (3) Child‐Pugh C score; (4) BLCL‐D stage; and (5) a previous liver transplantation.

2.3. Data Collection

Demographic data, medical history, usual treatment, cause of liver disease, history of liver decompensation, liver function assessment (Child‐Pugh score, MELD [model for end‐stage liver disease] score and ALBI [Albumin‐Bilirubin] Grade), biology and HCC features (number and size of lesions, tumour vascular invasion and extrahepatic spread at imaging, BCLC [Barcelona clinic liver cancer] grade, AFP) were collected at the initiation of AtezoBev treatment. MASLD was diagnosed following expert panel criteria, which required current or past evidence of hepatic steatosis on biopsy or imaging alongside at least one cardiometabolic risk factor—including BMI > 30 kg/m², type 2 diabetes, hypertension, or dyslipidemia and the absence of another cause of chronic liver disease. Alcohol‐related liver disease (ALD) was defined by the presence of alcohol intake higher than 50 g daily in patients.

Patients underwent a prospective follow‐up with clinical, biological and triple‐phase contrast material–enhanced CT or MRI, assessed every 3 months after the beginning of treatment.

2.4. Follow‐Up and Endpoints

Disease progression using RECIST1.1 was assessed on the follow‐up CT and MR imaging by a radiologist specialised in abdominal imaging every 3 months, and then validated during the local tumour multidisciplinary meeting.

The primary endpoint of the study was the overall survival (OS = time from initiation of AtezoBev treatment to death of any cause) and its prognostic risk factors.

The secondary endpoints were (1) the progression‐free survival (PFS = time from initiation of AtezoBev treatment to disease progression or death according to RECIST1.1), and its predictive risk factors; and (2) the rate of objective response at the first evaluation to AtezoBev (the percentage of patients with a confirmed complete or partial response).

2.5. Statistical Analysis

Patient characteristics were expressed as medians [interquartile range] for continuous variables and as numbers (percentages) for categorical data. Statistical comparisons for continuous and categorical variables were conducted using the nonparametric Mann–Whitney U test and Fischer's exact test, respectively.

Time‐to‐event analyses were initiated from the day of AtezoBev treatment, and event incidences were estimated using the Kaplan–Meier method. The log‐rank test was employed to evaluate differences between various groups. To investigate associations between variables and events, univariate Cox proportional hazards regression models were utilised. All variables with a p‐value < 0.05 in the univariate analysis were further considered in a multivariate Cox regression model, using a backward stepwise elimination approach to compute hazard ratios (HR) along with their corresponding 95% confidence intervals (CI). Statistical analyses were performed using R (version 4.4.0) for survival analysis and GraphPad Prism. A two‐sided p‐value < 0.05 was considered statistically significant.

3. Results

3.1. Baseline Characteristics of the Patients

Between July 2020 and September 2022, a total of 3292 patients were included from 32 French centres in the CHIEF cohort, at continuous and stable rates. Of these, 2747 were secondarily excluded from the analysis after the revision of individual data, owing either to another diagnosis (n = 71), or they had another treatment (n = 2269), or they did not receive AtezoBev in first line as systemic treatment (n = 400) or due to Child‐Pugh C cirrhosis (n = 6) and BCLC‐D classification (n = 1) (Figure S1). Analyses were thus performed in 545 patients with a median follow‐up of 18.91 months (IQR: 13.42–23.50). The patients were mainly male 86%, with a median age of 69 years. Liver disease was due to MASLD in 14%, ALD in 30%, and viral infection in 16% of patients. Mixed aetiology was observed in 21% of the patients. The baseline characteristics of the population are extensively described in Table 1. Most of the patients presented with compensated cirrhosis (81% Child‐Pugh A, median MELD score of 8 [IQR: 7–11]). A multinodular tumour, vascular invasion and extrahepatic spread of the tumour were present in 37%, 29% and 20% respectively. Median AFP was 55 ng/mL and 64% of patients were BCLC‐C.

TABLE 1.

Characteristics of the population.

Baseline characteristics		Atezolizumab‐Bevacizumab
		Available data	Total N = 545
Gender (male) a		545	471 (86)
Age (years) b		545	69 [62–75]
Age > 70 years a		545	267 (49)
Cirrhosis a		541	390 (72)
BMI > 30 kg/m2 a		393	63 (16)
ECOG‐PS	ECOG‐PS 0 a	496	292 (59)
	ECOG‐PS 1 a		167 (34)
	ECOG‐PS 2 a		37 (7)
Etiologies of liver disease	ALD only a	545	161 (30)
	MASLD only a	545	76 (14)
	Viral infection only a	545	88 (16)
	Mixed etiologies a	545	116 (21)
	Mixed etiologies with at least alcohol consumption a	545	315 (58)
	Mixed etiologies with at least alcohol viral infection a	545	50 (9)
Liver function	Child‐Pugh A a	510	413 (81)
	Child‐Pugh B a		97 (19)
	MELD b	466	8 [7–11]
	Albi Grade 1 a	490	166 (34)
	Albi Grade 2 a		295 (60)
	Albi Grade 3 a		29 (6)
Portal hypertension features	Previous history of AVB a	224	13 (6)
	Previous history of liver decompensation a	250	50 (20)
	No EV a	241	111 (46)
	Small size EV (Grade 1) a		78 (32)
	High‐risk EV (Grade 2–3) a		52 (22)
	Hepatic encephalopathy at baseline a	539	5 (1)
	Ascites at baseline a	539	24 (4)
	Platelet count < 150.000/mm3 a	509	198 (39)
Biology at baseline	Total bilirubin (μmol/L) b	512	14 [9–21]
	Albumin (g/L) b	491	37 [32–40]
	Creatinine (μM) b	511	74 [62–88]
	INR a	439	1.1 [1.0–1.2]
	Platelets count (×103/mm3) a	509	176 [122–246]
	CRP b	309	10 [5–26]
	NLR b	503	3.1 [2.1–4.8]
	NLR > 2.77 a	503	295 (59)
HCC features	BCLC‐A a	507	24 (5)
	BCLC‐B a		156 (31)
	BCLC‐C a		327 (64)
	Multinodular (≥ 3 nodules) a	509	187 (37)
	Tumour size (largest nodule) (mm) b	153	20 [13–35]
	Vascular Invasion a	523	152 (29)
	Extrahepatic spread a	490	96 (20)
	Serum AFP (ng/mL) b	475	55 [7–721]
	Serum AFP ≥ 400 ng/mL a	475	151 (32)
	Serum AFP ≥ 200 ng/mL a	475	186 (39)

Abbreviations: AFP, alpha‐fetoprotein; ALBI grade, albumin‐bilirubin grade; ALD, alcoholic liver disease; AVB, acute variceal bleeding; BCLC, Barcelona Clinic Liver Cancer classification; ECOG‐PS, Eastern Cooperative Oncology Group – performance status; EV, oesophageal varices; HCC, hepatocellular carcinoma; INR, international normalised ratio; MASLD, metabolic‐associated liver disease; MELD score, model for end‐stage liver disease; NLR, neutrophiles lymphocyte ratio.

^a Number (percentage).

^b Median [interquartile₁ – interquartile₃].

3.2. Overall Survival

The median OS was 23.1 months in the cohort with a 12‐month OS of 66% (Figure 1A). Causes of death were deterioration in liver function (63%), HCC progression (26%), sepsis (4%) and other causes (6%). In Cox multivariate analysis, only Child‐Pugh B score (HR = 1.93 [95% CI: 1.38–2.70], p < 0.001) was associated with a higher risk of mortality (Table 2). Mortality was also significantly higher in Child‐Pugh B compared to Child‐Pugh A patients (median OS of 10.6 months vs. 26.4 months, p < 0.001) in Log‐Rank analysis (Figure 1B). Compared to Child‐Pugh A patients, Child‐Pugh B patients presented more frequently with mixed etiologies with at least alcohol consumption (68% vs. 56%, p = 0.023), more history of liver decompensation (40% vs. 13%, p < 0.001), a higher rate of ALBI Grade 3 (25% vs. 1%, p < 0.001), a higher MELD score (12 vs. 8, p < 0.001) and more ascites at baseline (25% vs. 0%, p < 0.001). HCC features were similar between Child‐Pugh A and B patients except for a higher rate of extrahepatic spread in Child‐Pugh A patients (p = 0.023) (Table S1). ALBI grade or the presence of ascites in Child‐Pugh B patients were not associated with the worst outcome (Figure S2).

FIGURE 1.

Overall survival (OS) in patients treated by Atezolizumab‐Bevacizumab. (A) OS in the global cohort. (B) OS according to the Child‐Pugh score. (C) OS according to the ALBI grade. (D) OS according to the BCLC classification. (E) OS according to the ECOG‐PS. (F) OS according to the presence of MASLD in patients treated by AtezoBev, compared to patients with virus infection only and ALD only. ALD, alcoholic liver disease; MASLD, metabolic dysfunction associated liver disease; OS, overall survival. Results represented using the Kaplan–Meier Method with the log‐rank test. The numbers of patients at risk are figured under the x‐axis.

TABLE 2.

Univariate and multivariate analysis of factors associated with mortality in patients treated with Atezolizumab‐Bevacizumab.

Variable	Atezolizumab‐Bevacizumab (n = 545)
	Alive (n = 306)	Death (n = 239)	Univariate analysis		Multivariate analysis
			HR (95% CI)	p a	HR (95% CI)	p a
Age b	69 [62–75]	69 [60–76]	1.00 (0.99–1.01)	0.99
Age > 70 years c	152 (50)	115 (48)	1.02 (0.79–1.32)	0.85
Gender (female) c	42 (14)	32 (14)	0.99 (0.69–1.46)	0.98
Viral infection only c	52 (17)	36 (15)	0.78 (0.54–1.11)	0.17
MASLD c	40 (13)	36 (15)	1.07 (0.75–1.53)	0.71
ALD only c	86 (28)	75 (31)	1.19 (0.91–1.58)	0.20
Diabetes c	124 (41)	93 (39)	0.93 (0.72–1.21)	0.59
BMI ≥ 30 kg/m2 c	43 (19)	20 (12)	0.66 (0.41–1.05)	0.08
Presence of EV c	63 (21)	67 (28)	1.36 (1.03–1.81)	0.032	1.25 (0.90–1.73)	0.19
Presence of high‐risk EV c	27 (8)	25 (10)	1.25 (0.82–1.89)	0.29
Presence of ascites c	10 (3)	14 (6)	1.58 (0.92–2.71)	0.10
History of liver decompensation c	29 (20)	21 (19)	1.09 (0.67–1.75)	0.72
Child‐Pugh score B c	39 (13)	58 (26)	1.90 (1.41–2.56)	< 0.001	1.93 (1.38–2.70)	< 0.001
Meld a , d	8 [7–10]	9 [8–11]	1.03 (1.01–1.06)	0.002
Albi Grade 1 c , d Albi Grade 2 c , d ALBI Grade 3 c , d	115 (41) 157 (56) 9 (3)	51 (24) 138 (66) 20 (10)	— 1.81 (1.31–2.50) 3.64 (2.17–6.12)	— < 0.001 < 0.001
ECOG‐PS 0 c	169 (60)	123 (57)	—	—	—	—
ECOG‐PS 1 c	99 (35)	68 (32)	1.02 (0.76–1.37)	0.89	1.00 (0.73–1.38)	0.99
ECOG‐PS 2 c	14 (5)	23 (11)	1.89 (1.21–2.95)	0.005	1.48 (0.92–2.39)	0.10
NLR > 2.77 c	145 (51)	150 (68)	1.82 (1.37–2.43)	< 0.001	—	—
Platelets b	179 [128–248]	172 [120–243]	1.00 (1.00–1.00)	0.21
Platelets < 150 000/mm3	111 (38)	87 (39)	1.02 (0.78–1.34)	0.89
Total bilirubin a , d	13 [9–20]	16 [11–24]	1.01 (1.00–1.01)	< 0.001
Albumin a , d	38 [33–41]	36 [32–39]	0.95 (0.92–0.97)	< 0.001
Creatinine b	73 [62–87]	76 [63–91]	1.00 (1.00–1.003)	0.85
AFP b	29 [6–557]	92 [10–1249]	1.00 (1.00–1.00)	0.006
Serum AFP ≥ 200 ng/mL c	96 (35)	90 (45)	1.35 (102–1.78)	0.0357	1.25 (0.93–1.68)	0.13
Serum AFP ≥ 400 ng/mL c	79 (29)	72 (36)	1.3 (0.98–1.73)	0.077
BCLC‐A c	17 (6)	7 (3)	—	—	—	—
BCLC‐B c	102 (35)	54 (25)	0.66 (0.49–0.90)	0.009	1.03 (0.44–2.4)	0.94
BCLC‐C c	170 (59)	157 (72)	1.60 (1.2–2.16)	0.002	1.69 (0.74–3.86)	0.21
Nodular HCC (=1 nodules) c	89 (36)	56 (32)	—	—
Multinodular HCC (=2 nodules) c	44 (18)	45 (25)	1.41 (0.95–2.09)	0.09
Multinodular HCC (≥ 3 nodules) c	111 (45)	76 (43)	0.89 (0.66–1.20)	0.46
Vascular Invasion c	77 (26)	75 (33)	1.3 (0.97–1.71)	0.07
Extrahepatic metastasis c	56 (20)	40 (19)	0.97 (0.69–1.37)	0.86

Note: Bold values correspond to significant results in the univariate analysis.

Abbreviations: AFP, alpha‐fetoprotein; ALBI grade, albumin‐bilirubin grade; ALD, alcoholic liver disease; BCLC, Barcelona Clinic Liver Cancer classification; CI, confidence interval; ECOG‐PS, Eastern Cooperative Oncology Group – performance status; EV, oesophageal varices; HCC, hepatocellular carcinoma; HR, hazard ratio; INR, international normalised ratio; MASLD, metabolic‐associated liver disease; MELD score, model for end‐stage liver disease; NLR, neutrophiles lymphocyte ratio.

^a Likelihood ratio test, Score (logrank) test.

^b Median [interquartile₁–interquartile₃].

^c Number (percentage).

^d MELD score, ALBI, albumin and total bilirubin were not entered in the multivariate analysis in order to avoid collinearity with Child‐Pugh score.

Considering ALBI score, ALBI Grade 2/3 patients had a lower OS rate compared to ALBI Grade 1 patients in Log‐Rank analysis (median OS of 17.2 months vs. not reached, p < 0.001) (Figure 1C). ALBI Grade 2/3 patients presented more mixed etiologies with at least alcohol consumption (61% vs. 51%, p = 0.03), more high‐risk EV (25% vs. 16%, p = 0.04), more ascites at baseline (7% vs. 1%, p = 0.003) but less extrahepatic spread (15% vs. 28%, p = 0.002) (Table S2).

Mortality was also higher in BCLC‐C patients compared to BCLC‐B patients (median OS of 17,2 months vs. 27.8, p = 0.0043) in Log‐rank analysis (Figure 1D). Compared to BCLC‐B patients, BCLC‐C patients were younger (68 vs. 71 years old, p = 0.001), with more MASLD (25% vs. 16%) and viral infection etiologies (30% vs. 21%, p = 0.019). More multinodular HCC (56% vs. 28%, p < 0.001) were observed in BCLC‐B patients and more aggressive HCC features in BCLC‐C patients such as vascular invasion (45% vs. 0%, p < 0.001), extrahepatic spread (30% vs. 0%, p < 0.001) and higher levels of AFP (125 vs. 21 ng/mL, p < 0.001) (Table S3).

Mortality was significantly higher in ECOG‐PS 2 patients compared to ECOG‐PS 0 and 1 (p = 0.016) (Figure 1E). No impact of the aetiology on mortality was noticed in univariate analysis (Table 2) as well as in log‐rank analysis (Figure 1F).

3.3. Progression‐Free Survival

The median PFS was 5.2 months in the global cohort AtezoBev (Figure 2A). In Cox multivariate analysis, Albi Grade 3 (HR = 2.1985 [95% CI: 1.24–3.87 9–3.1], p = 0.007), AFP ≥ 200 ng/mL (HR = 1.35 [95% CI: 1.04–1.76], p = 0.0236) and Platelets (HR = 1.002 [95% CI: 1.00–1.003], p = 0.024) were associated with progression (Table 3). PFS was significantly higher in BCLC‐B patients compared to BCLC‐C patients (p = 0.05) (Figure 2B), but was similar according to the Child‐Pugh score (Figure 2C), the ALBI grade (Figure 2D), the ECOG‐PS status (Figure 2E) and the aetiology of the underlying liver disease (Figure 2F). Median treatment duration was 3.7 months (IQR: 1.6–7.7). The main reasons for treatment discontinuation were HCC progression (55%), deterioration in liver function (11%) and in general condition (16%). To note, 39 patients continued AtezoBev treatment despite radiological progression, due to perceived clinical benefit.

FIGURE 2.

Progression‐free survival (PFS) in patients treated by Atezolizumab‐Bevacizumab. (A) PFS in the global cohort. (B) PFS according to BCLC classification. (C) PFS according to Child‐Pugh score. (D) PFS according to ALBI grade. (E) PFS according to ECOG‐PS. (F) PFS according to the presence of MASLD in patients treated by AtezoBev, compared to patients with virus infection only and ALD only. ALD, alcoholic liver disease; MASLD, metabolic dysfunction associated liver disease; PFS, progression‐free survival. Results were represented using the Kaplan–Meier Method with the log‐rank test. The numbers of patients at risk are figured under the x‐axis.

TABLE 3.

Univariate and multivariate analysis of factors associated with progression in patients treated with Atezolizumab‐Bevacizumab.

Variable	Atezolizumab‐Bevacizumab (n = 545)
	No progression (n = 429)	Progression (n = 116)	Univariate analysis		Multivariate analysis
			HR (95% CI)	p a	HR (95% CI)	p a
Age b	70 [63–76]	66 [59–73]	0.99 (0.98–1.01)	0.46
Age > 70 years c	222 (52)	45 (39)	0.88 (0.69–1.12)	0.30
Gender (female) c	56 (13)	18 (15)	1.28 (0.91–1.78)	0.15
Viral infection only c	58 (14)	30 (26)	1.10 (0.80–1.52)	0.55
MASLD c	65 (15)	11 (09)	1.12 (0.80–1.57)	0.49
ALD only c	128 (30)	33 (28)	1.14 (0.88–1.48)	0.30
Diabetes c	173 (40)	44 (38)	0.96 (0.75–1.22)	0.72
BMI ≥ 30 kg/m2c	51 (16)	12 (16)	1.30 (0.85–2.00)	0.23
Presence of EV c	106 (25)	24 (21)	1.08 (0.82–1.43)	0.57
Presence of high‐risk EV c	44 (10)	8 (7)	1.19 (0.78–1.81)	0.42
Presence of ascites c	22 (5)	2 (2)	0.89 (0.51–1.56)	0.68
History of liver decompensation c	39 (20)	11 (21)	1.20 (0.75–1.90)	0.45
Child‐pugh score B c	85 (21)	12 (11)	1.22 (0.90–1.65)	0.19
MELD score b	9 [7–11]	8 [7–10]	1.04 (1.01–1.07)	0.01
ALBI Grade 1 c	127 (333)	39 (38)	—	—	—	—
ALBI Grade 2 c	234 (60)	61 (60)	1.24 (0.93–1.64)	0.14	1.32 (0.99–1.78)	0.06
ALBI Grade 3 c	27 (7)	2 (2)	2.20 (1.29–3.75)	0.004	2.19 (1.24–3.87)	0.007
ECOG‐PS 0 c	22 (57)	70 (66)	—	—
ECOG‐PS 1 c	137 (35)	30 (28)	1.07 (0.82–1.40)	0.62
ECOG‐PS 2 c	31 (8)	6 (6)	1.16 (0.75–1.81)	0.50
NLR > 2.77 c	233 (58)	65 (62)	1.56 (1.21–2.02)	0.0007	—	—
Platelets b	175 [120–247]	179 [128–243]	1.00 (0.99–1.00)	0.016	1.002 (1.00–1.003)	0.024
Platelets < 150 000/mm2	162 (40)	36 (34)	0.90 (0.70–1.15)	0.41
Total bilirubin b	14 [9.7–22]	14 [09–20]	1.01 (1.00–1.02)	0.002
Albumin b	37 [32–41]	38 [34–41]	0.98 (0.96–1.01)	0.10
Creatinine b	74 [62–90]	74 [64–83]	0.99 (0.99–1.003)	0.83
AFP b	51 [7–676]	87 [9.5–1362]	1.00 (1.00–1.00)	0.004
Serum AFP ≥ 200 ng/mL c	147 (39)	39 (41)	1.36 (1.05–1.76)	0.018	1.35 (1.04–1.76)	0.023
Serum AFP ≥ 400 ng/mL c	120 (31)	31 (33)	1.40 (1.07–1.84)	0.014
BCLC‐A c	19 (5)	5 (5)	—	—
BCLC‐C c	259 (64)	68 (67)	0.66 (0.34–1.29)	0.23
BCLC‐B c	127 (31)	29 (28)	0.88 (0.46–1.66)	0.69
Nodular HCC (=1 nodules) c	111 (33)	34 (38)	—	—
Multinodular HCC (=2 nodules) c	70 (21)	19 (21)	1.10 (0.76–1.59)	0.61
Multinodular HCC (≥ 3 nodules) c	151 (45)	36 (40)	0.74 (0.54–1.01)	0.06
Vascular invasion c	122 (30)	30 (27)	1.16 (0.89–1.51)	0.26
Extrahepatic metastasis c	76 (20)	20 (19)	1.1 (0.80–1.50)	0.57

Note: Bold values correspond to significant results in the univariate analysis.

Abbreviations: AFP, alpha‐fetoprotein; ALBI grade, albumin‐bilirubin grade; ALD, alcoholic liver disease; BCLC, Barcelona Clinic Liver Cancer classification; CI, confidence interval; ECOG‐PS, eastern cooperative oncology group – performance status; EV, oesophageal varices; HCC, hepatocellular carcinoma; HR, hazard ratio; INR, international normalised ratio; MASLD, metabolic‐associated liver disease; MELD score, model for end‐stage liver disease; NLR, neutrophiles lymphocyte ratio.

^a Likelihood ratio test, Score (logrank) test.

^b Median [interquartile₁ – interquartile₃].

^c Number (percentage).

^d MELD score and total bilirubin were not entered in the multivariate analysis in order to avoid collinearity with ALBI grade.

3.4. Predictive Factors of Best Response to Treatment

The objective response rates were 34% at the best treatment response according to RECIST 1.1, and 6% reached complete response (Table 4). The progression rate was significantly higher in BCLC‐C patients compared to BCLC‐B patients (25% vs. 11%, p < 0.001), but no difference was observed according to AFP level, Child‐Pugh score, ALBI grade and aetiology of the liver disease (Table 4). OS and PFS were significantly higher in patients with objective response rates compared to patients with stable disease or progression (Figure 3A,B). Compared to patients with objective response rates and stable disease, patients with progressive disease were significantly younger (p = 0.024), with a lower rate of mixed etiologies with at least alcohol consumption (p = 0.046), higher CRP (p = 0.022) and NLR (p = 0.024) at baseline, but also more BCLC‐C HCC (p = 0.027) and higher AFP levels (p = 0.019) (Table S4).

TABLE 4.

Best treatment response.

	Objective response rate	Stable disease	Progressive disease	p
Global cohort (n = 331) a	112 (34)	152 (46)	67 (20)	< 0.001
According to BCLC classification
BCLC‐B (n = 112) a	41 (37)	59 (52)	12 (11)	0.007
BCLC‐C (n = 196) a	62 (32)	84 (43)	50 (25)
According to AFP > 200 ng/mL
AFP < 200 ng/mL (n = 191) a	69 (36)	93 (49)	29 (15)	0.18
AFP ≥ 200 ng/mL (n = 121) a	43 (35)	50 (41)	28 (23)
According to Child‐Pugh score
Child‐Pugh A (n = 278) a	91 (33)	129 (46)	58 (21)	0.48
Child‐Pugh B (n = 52) a	21 (40)	23 (44)	8 (15)
According to Albi score
Albi Grade 1 (n = 115) a	45 (39)	48 (42)	22 (19)	0.74
Albi Grade 2 (n = 195) a	62 (32)	94 (48)	39 (20)
Albi Grade 3 (n = 14) a	5 (36)	7 (50)	2 (14)
According to the aetiology
MASLD only (=44) a	16 (36)	19 (43)	9 (20)	0.12
ALD only (n = 91) a	37 (41)	36 (40)	18 (20)
Viral only (n = 52) a	12 (23)	21 (40)	19 (37)
According to the presence of EV
No EV (n = 78) a	21 (27)	48 (62)	9 (11)	0.13
Presence of EV (n = 78) a	25 (32)	39 (50)	14 (18)

Note: Pearson's Chi‐squared test. Bold values correspond to significant results in the univariate analysis.

Abbreviations: AFP, alpha‐fetoprotein; ALBI grade, albumin‐bilirubin grade; ALD, alcoholic liver disease; BCLC, barcelona clinic liver cancer classification; EV, oesophageal varices; MASLD, metabolic‐associated liver disease.

^a Number (percentage).

FIGURE 3.

Progression‐free survival and overall survival in patients treated by Atezolizumab‐Bevacizumab according to best response status during follow‐up. (A) OS according to response status in patients treated by AtezoBev. (B) PFS according to response status in patients treated by AtezoBev. OS, overall survival; PFS, progression‐free survival. Results were represented using the Kaplan–Meier Method with the log‐rank test. The numbers of patients at risk are figured under the x‐axis. The objective response to AtezoBev was defined by the patients with a confirmed complete or partial response according to RECIST 1.1 criteria.

4. Discussion

In this prospective multicenter real‐world CHIEF cohort analysis, we demonstrate that the effectiveness of AtezoBev treatment mirrors the outcomes observed in the Phase III IMbrave 150 study clinical trial. Notably, despite a distinct patient profile characterised by a predominance of non‐viral liver disease typical in France, AtezoBev maintains comparable efficacy. Additionally, we observed notable enhancements in OS, PFS and treatment response among BCLC‐B patients, indicating the potential suitability of AtezoBev for this subgroup of HCC patients.

In our cohort, the median OS was 23.1 months, likely influenced by the higher proportion of BCLC‐B patients (31%) in the CHIEF cohort. Notably, within our cohort, BCLC‐B patients exhibited a significantly higher median OS (27.8 months) compared to BCLC‐C patients (17.2 months, p = 0.0043). The latest update of the IMbrave 150 study reported a median OS of 25.8 months in BCLC‐B compared to 15.5 months for BCLC‐C patients. These competitive findings in the BCLC‐B population prompt further investigation into immunotherapy trials for patients with intermediate‐stage HCC. This avenue of research could hold particular significance for patients at risk of lower response rates to locoregional therapy or heightened susceptibility to liver decompensation following such therapy, potentially preventing them from accessing subsequent HCC treatments. Indeed, locoregional therapy, such as TACE, might not be effective for some patients and can lead to liver function deterioration [4, 5]. This decline is significant since liver function is a critical prognostic factor for systemic therapy, and deterioration due to repeated TACE might reduce the access to further HCC treatments. In addition, bilobar tumours with over 50% liver involvement, infiltrative or poorly defined nodular tumours and large vessel vascular invasion have been associated with less effectiveness using TACE treatment [4]. Thus, systemic therapy such as AtezoBev could be a good alternative for selected BCLC‐B patients.

As already described [6], median OS was lower in patients with Child‐Pugh B score (10.6 months) compared to Child‐Pugh A patients (26.4 months, p < 0.001) and Child‐Pugh B score was independently associated with mortality in Cox analysis. However, PFS and response rate were similar between the two groups of patients, as already showed in real life data survey [6, 7, 8]. Even if the use of ICIs in Child‐Pugh B patients is debated, some series demonstrated better outcome in Child‐Pugh B7/B8 patients treated by ICIs compared to best supportive care [9]. Phase 2 trials focusing on patients with altered liver function are currently ongoing to better describe the outcome of ICIs in this specific population of patients, but AtezoBev could be considered case by case in Child‐Pugh B patients [10]. More data are currently available about the use of immunotherapy as a downstaging strategy before liver transplantation, and systemic therapies could be discussed in patients with liver dysfunction who could benefit from liver transplantation.

Objective response rate occurred in 34% of the patients and was close to the one observed in Mbrave 150 trial. Interestingly, we also observed similar increased OS and PFS in patients with objective response rate and stable disease compared to patients with progressive disease. Responses were previously rarely seen before the era of immunotherapy. Such responses might also lead to new strategies, including offering loco‐regional treatment after responses. With further follow‐up, we will describe in the CHIEF cohort the sequences of treatment after AtezoBev and other immunotherapy regimens.

Concerns have been raised regarding the efficacy of ICIs in patients with MASLD. Early reports suggested that non‐viral HCC, particularly MASLD‐HCC, might demonstrate reduced responsiveness to immunotherapy [11, 12]. However, larger‐scale meta‐analyses and real‐life retrospective cohorts have not confirmed these findings [13, 14]. In addition, a recent post hoc analysis of IMbrave150 found no differences in PFS and OS based on aetiology [15]. In our CHIEF cohort, 14% presented with MASLD only, and 58% exhibited mixed aetiologies involving alcohol consumption, reflective of the French population. Conversely, approximately 60% of the IMbrave150 population had viral infections. Notably, we observed no differences in OS, PFS, and response at 3 months between patients with MAFLD, ALD and viral aetiologies. These findings challenge prevailing concerns regarding the diminished efficacy of immunotherapy in HCC patients with MASLD [11, 12]. While the intricate interplay between metabolic factors and the immune response in HCC requires further investigation, treatment decisions for systemic therapies in clinical practice should not be solely based on aetiology.

Our study has notable strengths, including a substantial patient cohort from 32 different centers prospectively recorded and rigorous data analysis. However, the median follow‐up time remains short and long‐term studies are needed, and some data are missing which prevent us from proposing accurate data regarding tolerance and side effects of the treatment. In addition, due to the observational design of the study, some assessments may have been missed, or adherence to the evaluation schedule may have varied. A detailed analysis focusing on safety outcomes will be conducted in a separate dedicated study with a longer follow‐up period.

In conclusion, the efficacy of AtezoBev in the French cohort CHIEF shows similar outcomes compared to the IMbrave150 trial despite different patient profiles, with better outcomes in BCLC‐B patients, suggesting that they might benefit from such treatment.

Author Contributions

Drs. Allaire, Costentin, Ducournau, Nguyen‐Khac and Thiam had full access to all data in the study and take responsibility for the integrity of data and the accuracy of data analysis. Study concept and design: Drs. Allaire, Costentin, Ducournau, Nguyen‐Khac and Thiam. Acquisition of data: Drs Allaire, Amaddeo, Anty, Baron, Bourliere, Blanc, Bronovicki, Brusset, Cattan, Costentin, Decaens, Ducournau, Edeline, Ganne, Ganry, Heurgue, Loustaud‐Ratti, Lequoy, Manfredi, Merle, Nguyen‐Khac, Nousbaum, Oberti, Olliver‐Hourmand, Pascale, Peron, Pageaux, Riachi, Silvain, Thiam, Uguen, Villing and Ziol. Analysis and interpretation of data: Drs. Allaire, Amaddeo, Bouattour, Brusset, Costentin, Edeline, Thiam and Ziol. Drafting of the manuscript: Drs. Allaire, Costentin, Ducournau, Nguyen‐Khac and Thiam. Critical revision of the manuscript for important intellectual content: Drs Allaire, Amaddeo, Anty, Baron, Bourliere, Blanc, Bronowicki, Cattan, Costentin, Decaens, Ducournau, Edeline, Ganne, Ganry Brusset, Heurgue, Loustaud‐Ratti, Lequoy, Manfredi, Merle, Nguyen‐Khac, Nousbaum, Oberti, Ollivier‐Hourmand, Pascale, Peron, Pageaux, Riachi, Silvain, Thiam, Uguen, Villing and Ziol. Statistical analysis: Dr. Thiam. Study supervision: Drs. Allaire, Costentin, Ducournau and Nguyen‐Khac.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Flow chart of the study.

Figure S2: Overall survival in Child‐Pugh B patients treated by Atezolizumab‐Bevacizumab according ALBI grade and the presence of ascite.

Table S1: Characteristics of the population according to Child‐Pugh score.

Table S2: Characteristics of the population according to ALBI grades.

Table S3: Characteristics of the population according to BCLC classification.

Table S4: Characteristics of the population according to best response.

Allaire M., Thiam E. M., Amaddeo G., et al., “Real‐World Outcomes of Atezolizumab–Bevacizumab in Hepatocellular Carcinoma: The Prospective French CHIEF Cohort,” Liver International 45, no. 10 (2025): e70337, 10.1111/liv.70337.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.