Changing Treatment Landscape of Hepatocellular Carcinoma: A Real-World Retrospective Analysis of Treatment Algorithms and Outcomes of Patients over 10 Years

Abstract

Background

Hepatocellular carcinoma is one of the most common cancers worldwide and the third leading cause of cancer-related mortality. The therapeutic landscape of hepatocellular carcinoma, especially systemic therapies, is constantly changing and characterized by a multimodal approach.

Objective

The aim of this study was to investigate whether and how treatment sequencing according to tumor stage and liver function has changed over time and if the introduction of new therapeutic options has led to improved survival rates.

Methods

Therapy concepts were retrospectively recorded and evaluated in a cohort of 1288 patients with hepatocellular carcinoma with a first diagnosis between January 2013 and December 2022 who were divided into two cohorts with a cut-off for dividing in 2020 owing to the approval of atezolizumab and bevacizumab for the treatment of hepatocellular carcinoma by the European Medicines Agency (Cohort A initial diagnosis 2013–19, Cohort B initial diagnosis 2020–22). All patients were treated at Hannover Medical School, Germany.

Results

Baseline characteristics differed significantly between cohorts, with Cohort B presenting at earlier tumor stages and with better preserved liver function. Across all treatment modalities, liver function emerged as the primary determinant of prognosis, independent of tumor stage and performance status. Over time, a treatment shift from transarterial chemoembolization towards systemic therapies was observed. Median overall survival from initial diagnosis did not differ between cohorts. In unadjusted analyses, patients treated with immune checkpoint inhibitor-based regimens demonstrated longer median overall survival compared with those receiving tyrosine kinase inhibitors. However, after multivariable adjustment for baseline imbalances, no independent survival difference between treatment groups remained.

Conclusions

This study provides a detailed longitudinal assessment of evolving treatment patterns and outcomes, highlighting a clear shift in therapeutic strategies over time with the increasing adoption of systemic therapies. Despite evolving treatment strategies and increased use of systemic therapies, liver function remained the dominant prognostic factor across treatment settings. Observed survival differences between treatment modalities were strongly influenced by baseline patient characteristics and treatment selection.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s11523-026-01239-8.

Key Points

The therapy algorithm for hepatocellular carcinoma in a real-world setting has substantially changed over 10 years especially reflected by a shift from intra-arterial approaches towards systemic therapies.

Across all therapies, a preserved liver function remains the key prognostic factor in hepatocellular carcinoma and should therefore be the main focus in treatment selection.

Introduction

Hepatocellular carcinoma (HCC) is the sixth most common malignancy worldwide and the third leading cause of cancer-related mortality [1]. A significant proportion of patients present with advanced-stage disease, often accompanied by impaired liver function. These factors not only restrict the feasibility of curative treatment approaches but also contribute to the high mortality associated with HCC [2]. Therapeutic strategies for HCC are closely linked to both tumor stage and underlying liver function, underscoring the importance of accurate staging to guide optimal treatment selection [3]. The Barcelona Clinic Liver Cancer (BCLC) algorithm is widely accepted as the standard framework for treatment allocation in HCC [4]. However, in light of the recent advances in effective systemic therapies, the application of the BCLC algorithm should not be viewed as overly rigid. In selected cases, these therapies may enable downstaging, allowing patients to subsequently receive curative-intent treatments.

For many years, transarterial chemoembolization (TACE) was regarded as the standard-of-care treatment for patients with intermediate-stage HCC [5, 6]. However, in recent years, systemic therapy options have advanced considerably, particularly with the advent of immune checkpoint inhibitor (ICI)-based treatment regimens, which are increasingly relevant for patients with intermediate-stage disease as well. Based on phase III trial data, atezolizumab in combination with bevacizumab (Atezo/Bev) was approved in 2020, followed by tremelimumab with durvalumab in 2022, and most recently, ipilimumab with nivolumab [7–9]. These ICI-based combinations have achieved overall response rates of up to 36% and median overall survival (mOS) approaching 2 years, representing a significant improvement over tyrosine kinase inhibitor (TKI)-based therapies previously used in advanced HCC [9].

Looking ahead, the therapeutic landscape of HCC is expected to continue evolving, driven not only by the development of novel systemic therapies but also by the integration of systemic treatments with interventional approaches [10]. In this context, the EMERALD-1 and LEAP-012 trials represent important milestones as the first positive phase III studies evaluating the combination of TACE with ICI-based therapies and either TKIs or vascular endothelial growth factor inhibitors [11, 12]. Both trials demonstrated significantly improved progression-free survival and the objective response rate in the combination treatment arms compared with TACE alone. However, these improvements in progression-free survival did not translate into prolonged mOS for patients (LEAP-12, press release), a critical point for the future integration of these combination strategies into standard HCC management.

Although the treatment landscape for intermediate and advanced HCC is rapidly evolving, with the benefits of new and effective systemic therapies clearly demonstrated in clinical trials, it is important to recognize that real-world patient populations often differ substantially from those enrolled in clinical studies. In routine clinical practice, factors such as comorbidities, impaired liver function, and advanced tumor burden frequently impact treatment decisions and influence patient outcomes. Consequently, real-world data are essential for understanding how the integration of evolving therapies translates into patient benefit within the context of daily clinical care.

In this study, we evaluated a large cohort of 1288 patients with HCC treated at Hannover Medical School over a 10-year period. The primary objective was to assess whether and how treatment sequencing according to tumor stage and liver function has changed over time in real-life clinical practice. Additionally, we aimed to explore whether evolving treatment strategies were associated with changes in survival outcomes in a real-world setting.

Materials and Methods

Population and Data Collection

Based on the International Statistical Classification of Diseases and Related Health Problems, 10th revision, German Modification code C22.0 (liver cell carcinoma, HCC), 1288 patients with histological or image-based diagnosis of HCC between January 2013 and December 2022 treated at Hannover Medical School were identified. The study protocol followed the ethical guidelines of the 1975 Declaration of Helsinki (ethical approval ID 940-2011).

Baseline characteristics included, sex, etiology of the underlying liver disease, liver cirrhosis, tumor size at initial diagnosis, and tumor-specific characteristics such as the tumor stage at diagnosis according to the BCLC. In addition, liver function was determined using the albumin-bilirubin (ALBI) score. Based on the approval of Atezo/Bev for the treatment of HCC in 2020, the study population was split into two cohorts for a subsequent comparative analysis: Cohort A (initial diagnosis 01.01.2013–31.12.2019) and Cohort B (initial diagnosis 01.01.2020–31.12.2022). Median overall survival was defined as the time from either the initial diagnosis of HCC or the initiation of the respective treatment line to death or last follow-up, whichever occurred first. The data cut-off for follow-up was 1 December, 2025. Two patients were lost to follow-up. Information on mortality was obtained from residents’ registration offices.

Data Analysis

Statistical analyses were performed using IBM SPSS Statistics Version 30.0 (SPSS Inc., Chicago, IL, USA). Baseline characteristics were expressed as numbers and percentages. Sankey diagrams were generated using SankeyMATIC. Continuous variables were analyzed using the Student’s t-test if sample means were normally distributed. If population variances were unequal, the Welch t-test was used. Categorial variables were analyzed using the Chi-square test or Fisher exact test, as appropriate. The Kaplan–Meier method was used to estimate mOS and 95% confidence intervals (95% CIs), which were compared using the log-rank test. A probability (p) value less than 0.05 was considered significant. Hazard ratios (HRs) for clinical variables were analyzed using the Cox proportional hazards model and reported with a 95% CI.

Multivariable Cox proportional hazards regression models were constructed to identify independent prognostic factors for OS in the overall cohort, in patients receiving systemic therapies as first-line treatment and in patients receiving TACE. Covariates included age, sex, presence of cirrhosis, liver function (ALBI grade), tumor stage (BCLC stage), and etiology.

Multiple sensitivity analyses were performed to assess the robustness of the findings, including: models excluding ALBI, tumor-based models including macrovascular invasion and extra-hepatic spread, restriction of follow-up duration (e.g., 48 months), complete-case analyses, and analyses limited to patients with known etiology. In addition, alternative model specifications incorporating ALBI as a categorical variable including missing values were evaluated. Treatment modality was not included in the multivariable model evaluating overall survival (OS) from diagnosis because treatment allocation is strongly influenced by baseline prognosis and treatments are initiated at different timepoints during the disease course, making them susceptible to indication bias and time-dependent confounding.

To account for potential baseline imbalances between treatment groups, a propensity score-based analysis using inverse probability of treatment weighting was performed as a sensitivity analysis. The propensity score was estimated using baseline covariates including age, sex, cirrhosis, ALBI grade, BCLC stage, etiology, and Eastern Cooperative Oncology Group (ECOG) performance status. Weighted Cox regression models were then applied to assess the association between treatment type and OS.

Because of missing data, a complete case analysis was applied for multivariable models.

Prespecified subgroup analyses were conducted to explore potential differences in treatment effects according to liver function (ALBI grade) and tumor stage (BCLC stage B and C). Separate multivariable Cox models were fitted within each subgroup. To formally assess effect modification by tumor stage, an interaction term between treatment type (ICI vs TKI) and BCLC stage (B vs C) was included in the multivariable Cox regression model restricted to patients with BCLC stage B or C disease.

Results

Population Characteristics

Between January 2013 and December 2022, a total of 1288 patients were diagnosed with HCC at Hannover Medical School. For our analysis, patients were divided into two cohorts based on the availability of ICI-based systemic therapies: cohort A consisted of 986 patients diagnosed between 2013 and 2019, while cohort B included 302 patients diagnosed between 2020 and 2022.

As expected, HCC was more frequently diagnosed in male patients in both cohorts, with 70.4% in cohort A and 78.1% in cohort B, reflecting the known sex distribution of HCC. Chronic viral hepatitis remained the most common underlying liver disease across both groups, although its prevalence significantly decreased over time, from 35.6% in Cohort A to 27.1% in Cohort B (p = 0.0007). Among patients with viral hepatitis, hepatitis C was the most common etiology in both groups (Cohort A: 23.1%; Cohort B: 17.2%). There were no significant differences between the cohorts regarding the prevalence of alcoholic liver disease (cohort A 23.3% vs 19.9% in cohort B; p = 0.238) or metabolic dysfunction-associated steatotic liver disease (cohort A 11.4% vs 10.9% in cohort B; p = 0.917). In cohort B, more patients were included with unknown liver disease compared with cohort A (42.1% vs 29.7, p = 0.056). Distinguishing non-viral etiologies in real-world studies is challenging, particularly in the absence of prospective and standardized data collection and is further complicated by the multifactorial nature of HCC. As a result, metabolic-related and alcohol-related liver diseases may be incompletely captured in this retrospective dataset, contributing to the substantial proportion of cases classified as “unknown.”

In both cohorts, most HCC were diagnosed in patients with clinically documented liver cirrhosis, but significantly more patients in cohort A had documented liver cirrhosis compared with cohort B (cohort A 93.7%, cohort B 79.5%; p < 0.001). Liver function at the time of diagnosis according to ALBI ranged widely: most patients in both cohorts were classified as ALBI grade 2, indicating moderately impaired liver function (cohort A 47.7%; cohort B 46.4%). Patients with ALBI grade 3 were equally split between the two cohorts. The proportion of patients with ALBI grade 1, reflecting well-preserved liver function, was higher in cohort B compared with cohort A (35.1% vs 16.4%, respectively; p < 0.001). It is important to note that for 22.3% of patients in cohort A, ALBI grade data were not available at the time of diagnosis.

At the time of initial diagnosis, the predominant proportion of patients were classified as BCLC-A (cohort A 39.9%, cohort B 40.1%) and BCLC-B (cohort A 29.3%; cohort B, 36.8%). There were 27.7% of the patients from cohort A and 20.9% of cohort B diagnosed with BCLC-C, while only a few patients were considered BCLC-0 (cohort A 0.6%; cohort B 1.7%) or BCLC-D (cohort A: 0.4%; cohort B 0.3%).

In both cohorts, approximately half of all patients presented with a maximum tumor diameter of ≤5 cm at the time of diagnosis (p = 0.739). Regarding lesion number, unifocal tumors were more prevalent in cohort B, with 65.9% of patients presenting with a single lesion compared with 49.3% in cohort A (p < 0.001). Baseline characteristics are summarized in Table 1.

Table 1.

Patient characteristics at initial diagnosis

Variable	2013–19 (N = 986)	2020–22 (N = 302)	p-value
Median age, years	66	66	0.913
Male sex	694 (70.4)	236 (78.1)	0.010
Dead	768 (77.9)	170 (56.3)	< 0.001
Etiology
ALD	230 (23.3)	60 (19.9)	0.238
Viral	351 (35.6)	82 (27.1)	0.0007
HCV	228 (23.1)	52 (17.2)
HBV	98 (9.9)a	26 (8.6)
HBV+HCV	15 (1.5)	1 (0.3)
MASLD	112 (11.4)	33 (10.9)	0.917
Other/unkown	293 (29.7)	127 (42.1)	0.056
Cirrhosis	924 (93.7)	240 (79.5)	< 0.001
ALBI grade
1	162 (16.4)	106 (35.1)	< 0.001
2	470 (47.7)	140 (46.4)	0.739
3	134 (13.6)	42 (13.9)	0.964
n.a.	220 (22.3)	14 (4.6)
BCLC stage
0	6 (0.6)	5 (1,7)	0.143
A	393 (39.9)	121 (40.1)	0.99
B	289 (29.3)	111 (36.8)	0.018
C	273 (27.7)	63 (20.9)	0.022
D	4 (0.4)	1 (0.3)	0.99
n.a.	21 (2.1)	1 (0.3)
Number of lesions
Unifocal	486 (49.3)	199 (65.9)	< 0.001
Multifocal	460 (46.7)	100 (33.1)	< 0.001
n.a.	40 (4.1)	3 (1.0)
Tumor size
≤ 5 cm	529 (53.7)	158 (52.3)	0.739
> 5 cm	346 (35.1)	118 (39.1)	0.233
n.a.	111 (11.3)	26 (8.6)

ALBI albumin-bilirubin, ALD alcoholic liver disease, BCLC Barcelona Clinic Liver Cancer, HBV hepatitis B virus, HCV hepatitis C virus, HDV hepatitis D virus, MASLD metabolic dysfunction-associated steatotic liver disease, n.a. not available

^aIncluding one patient with HDV coinfection; statistically significant p-values are in bold

First-Line Therapies

Eight hundred and fifty-nine patients (87.1%) in cohort A and 265 patients (87.7%) in cohort B received at least one tumor‑directed therapy. In cohort A, TACE was the most common first-line treatment modality (279 patients; 28.3%), whereas in cohort B, TACE ranked only third (13.6%, n = 41) [p < 0.001]. Surgical resection as front-line treatment was performed at similar rates in cohort A and B (25.9% vs 26.8%; p = 0.797), making surgery the second most common and the most common treatment modality, respectively (Fig. 1A). Use of systemic therapy as first-line treatment notably increased over time from 9.6% (n = 95) in cohort A to 21.2% (n = 64) in cohort B (p < 0.001). There was no significant difference in the use of radiofrequency or microwave ablation (p = 0.378). Similarly, the proportion of patients treated with transarterial radioembolization remained constant, (5.5% vs 3.6%, p = 0.261). Baseline treatment patterns across both cohorts are summarized in Fig. 1A and Table 2.

Fig. 1.

Treatment sequencing for first-line and second-line therapy (A) and for first-line therapy according to Barcelona Clinic Liver Cancer (BCLC) stage (B). BSC best supportive care, LTX liver transplantation, MWA microwave ablation, RFA radiofrequency ablation, TACE transarterial chemoembolization, TARE transarterial radioembolization

Table 2.

First-line treatment according to year of diagnosis

First-line treatment	2013–19 (n = 986)	2020–22 (n = 302)	p-value
Liver transplantation	12 (1.2)	5 (1.7)	0.567
Resection	255 (25.9)	81 (26.8)	0.797
RFA/MWA	101 (10.2)	37 (12.3)	0.378
TACE	279 (28.3)	41 (13.6)	< 0.001
TARE	54 (5.5)	11 (3.6)	0.261
Systemic therapy	95 (9.6)	64 (21.2)	< 0.001
Others/n.a.	122 (12.4)	36 (11.9)	0.913
BSC	68 (6.9)	27 (8.9)	0.288

BSC best supportive care, MWA microwave ablation, RFA radiofrequency ablation, TACE transarterial chemoembolization, TARE transarterial radioembolization, n.a. not available, statistically significant p-values are in bold

The proportion of patients receiving best supportive care did not differ between the two time periods (p = 0.288). Treatment allocation to first-line therapies according to BCLC stage is presented in Fig. 1B. A notable shift in treatment patterns was observed in patients with BCLC stage B, with the use of TACE decreasing from 52.6% in cohort A to 13.5% in cohort B (p < 0.001), while the use of systemic therapies simultaneously increased from 9.0% to 23.4% (p < 0.001). In BCLC stage C, systemic therapy was the predominant first-line treatment in cohort B, administered to 49.2% of patients, and thus contrasting the 19.7% of patients treated with front-line systemic therapy in cohort A (p < 0.001), while best supportive care was provided to 22.2% of patients in this group. In contrast, patients in cohort A with BCLC stage C were more frequently allocated to other treatment modalities, including resection (15.8%), TACE (9.9%), and transarterial radioembolization (11.7%).

Second-Line Treatment

A total of 492 patients received second-line treatment during the study period, including 379 patients in cohort A (representing 43.9% of those who received first-line therapy) and 113 patients in cohort B (42.6%) [p = 0.73]. Details of second-line treatment patterns are illustrated in Fig. 1A.

In cohort A, a subsequent TACE procedure was the most commonly administered second-line therapy in 27.9% of patients. In contrast, thermal ablation (radiofrequency ablation or MWA) and systemic therapies were the two equally utilized second-line options in cohort B, overall accounting for 35.4% of patients in second line. Thermal ablation in cohort B was primarily employed following initial TACE treatment, reflecting its use as part of a combined curative-intent approach. In this context, TACE in cohort B was frequently integrated into combination strategies, with 51.2% of TACE procedures performed in this cohort used within a curative-intent treatment framework.

Survival Analysis

As anticipated, the longest mOS was observed in patients who underwent liver transplantation (mOS not reached), followed by those treated with surgical resection (mOS 59 months; 95% CI 45.9–72.2) and radiofrequency ablation/MWA (mOS 50.9 months; 95% CI 37.7–64.1) [see Fig. I of the Electronic Supplementary Material (ESM)]. Survival analysis according to underlying liver disease revealed the longest survival for patients with MASLD, with a mOS of 30.2 months (95% CI 21.8–38.6), followed by patients with chronic viral hepatitis (mOS of 23.8 months; 95% CI 18.6–28.5). Patients with alcoholic liver disease had a shorter mOS of 15 months (95% CI 11.6–18.4). This group included a higher proportion of patients with impaired liver function, which may partially account for the reduced survival observed in this subgroup (see Fig. II of the ESM, Table 1). These unadjusted analyses are presented for descriptive purposes only to illustrate real-world outcomes associated with different treatment strategies in an unselected HCC population, which differs substantially from the highly selected populations enrolled in prospective clinical trials. They should not be interpreted as estimates of treatment efficacy or causal treatment effects and are therefore regarded as descriptive and hypothesis generating only.

The mOS from the time of initial diagnosis was similar between the two cohorts (HR 0.97; 95% CI 0.83–1.13; p = 0.721) [see Fig. III of the ESM]. No significant differences in mOS were observed between the cohorts when stratified by BCLC stage or by liver function categories (ALBI grade 1 and ALBI grade ≥2) [see Figs. IV–VIII of the ESM). As patients in cohort B were diagnosed significantly more often at earlier tumor stages and had better liver function, a multivariable analysis was performed to adjust for these potential confounders. After adjustment for baseline characteristics, cohort B remained associated with shorter OS (HR 1.27, 95% CI 1.07–1.51; p = 0.007). However, this finding should be interpreted cautiously given potential residual confounding and differences in treatment allocation over time (Fig. 2). Tumor stage and liver function emerged as independent prognostic factors. Etiology and cirrhosis were also associated with survival, with alcohol-related liver disease showing worse outcomes compared with viral etiology (HR 1.34, p = 0.002).

Fig. 2.

Forest plot of adjusted hazard ratios (HRs) since first diagnosis. ALBI the albumin-bilirubin, ALD alcoholic liver disease, BCLC Barcelona Clinic Liver Cancer, MASLD metabolic dysfunction-associated steatotic liver disease

Sensitivity analyses supported the robustness of these findings across multiple model specifications, including complete-case analysis, tumor-based models, and restricted follow-up periods (see Table S1 of the ESM). Given the higher proportion of missing ALBI values in cohort A, an additional sensitivity analysis was conducted treating missing ALBI as a separate category. This approach allowed inclusion of the full cohort of 1288 patients and yielded consistent results, with cohort B remaining associated with worse OS compared with cohort A (HR 1.28, 95% CI 1.08–1.51, p = 0.004). These findings indicate that the observed cohort effect was not solely driven by exclusion of patients with missing ALBI data.

Role of Systemic Therapy

Considering the evolving role of systemic therapies in the treatment landscape of HCC, we performed a detailed analysis focusing on patients who received systemic treatment. To address a potential confounding by indication and immortal time bias, we only included patients in this analysis, who received a systemic treatment in the first-line setting (either as first-line therapy at all or after tumor recurrence following a curative treatment). As we were particularly interested in the effect of ICI-based therapy in our cohort, we divided the patients into two groups: those who received ICI-based therapy (n = 77) and those who received TKI-based therapy (n = 95). In the ICI-based therapy, almost all patients received the combination of Atezo/Bev (n = 73) and only three patients received tremelimumab and durvalumab (n = 3). In the TKI group, more patients received sorafenib (n = 78) than lenvatinib (n = 16). The baseline characteristics of those patients are summarized in Table 3. Patients treated with ICI-based therapy had a lower prevalence of cirrhosis (ICI 79.2% vs TKI 90.5%; p = 0.039) and were consequently more likely to present with preserved liver function (ALBI 1 26.0% vs 15.8%; p = 0.048). A significantly higher proportion of patients receiving ICI-based therapy were treated at earlier tumor stages (BCLC B: 40.3% vs 27.4% for TKI; p = 0.031) and had a better performance status (ECOG 0: 51.9% vs 30.5%; p = 0.008). To account for these imbalances, a multivariable Cox regression analysis and a propensity score-based sensitivity analysis using inverse probability of treatment weighting (IPTW) were performed. After adjustment for age, ECOG performance status, liver function, tumor stage, cirrhosis, etiology, and prior therapies, no significant difference in OS was observed between TKI-based and ICI-based treatment (HR 1.18, 95% CI 0.76–1.82; p = 0.452) [Fig. 3B]. Similarly, IPTW-based analyses yielded consistent results. These findings support the consistency of the adjusted analyses across different statistical approaches and suggest that observed survival differences in unadjusted analyses were strongly influenced by baseline patient characteristics and treatment allocation.

Table 3.

Baseline characteristics for patients receiving systemic treatment as a first-line treatment

Variable	TKI (N = 95)	ICI (N = 77)	p-value
Median age, years	66	67
Male sex	81 (85.3)	58 (75.3)	0.118
Dead	94 (98.9)	64 (83.1)	< 0.001
Etiology
ALD	30 (31.6)	14 (18.2)	0.052
Viral	28 (29.5)	21 (27.3)	0.78
MASLD	6 (6.2)	8 (10.4)	0.31
Other/unkown	31 (32.6)	34 (44.2)	0.11
Cirrhosis	86 (90.5)	61 (79.2)	0.039
ALBI grade
1	15 (15.8)	20 (26.0)	0.048
2 (≥ grade 2)	56 (58.9)	37 (48.1)	0.18
n.a.	24 (25.3)	20 (26.0)	0.91
BCLC stage
A	10 (10.5)	10 (13.0)	0.63
B	26 (27.4)	31 (40.3)	0.031
C	56 (58.9)	36 (46.8)	0.11
D	3 (3.2)	0 (0.0)	0.25
Macrovascular infiltration
Yes	43 (45.3)	25 (32.5)	0.089
No	52 (54.7)	52 (67.5)	0.089
Extra-hepatic spread
Yes	27 (28.4)	12 (15.6)	0.061
No	67 (70.5)	65 (84.4)	0.061
ECOG
0	29 (30.5)	40 (51.9)	0.008
1	46 (48.4)	30 (39.0)	0.19
≥ 2	16 (16.8)	5 (6.5)	0.041
n.a.	4 (4.2)	2 (2.6)	0.65

ALBI albumin-bilirubin, ALD alcoholic liver disease, BCLC Barcelona Clinic Liver Cancer, ECOG Eastern Cooperative Oncology Group, ICI immune checkpoint inhibitor, MASLD metabolic dysfunction-associated steatotic liver disease, n.a. not available, TKI tyrosine kinase inhibitor; statistically significant (p)-values are in bold

Fig. 3.

A Unadjusted Kaplan–Meier analysis of overall survival from start of first-line systemic treatment according to used substance and B forest plot of adjusted hazard ratio (HR) since the start of systemic treatment in the first-line setting. CI confidence interval, ICI immune checkpoint inhibitor, mOS median overall survival, OS overall survival, TKI tyrosine kinase inhibitor

Liver function (ALBI ≥2 vs 1: HR 2.59, p = 0.001) and impaired performance status (ECOG ≥2 v. 0: HR 2.44, p = 0.014) emerged as independent prognostic factors for survival under systemic treatment regardless of treatment type (Fig. 3B). In unadjusted Kaplan–Meier analyses, patients receiving ICI-based therapy demonstrated longer median OS compared with patients treated with TKI-based therapy (14.4 vs 5.9 months; p = 0.002) [Fig. 3A]. Survival outcomes were comparable across the different TKIs used. A total of 37 patients (21.5%) received more than one line of systemic therapy, and a trend toward improved survival was observed in this group, irrespective of the type of first-line treatment (p = 0.054). It should be noted that the majority of these patients (n = 23, 61.2%) received ICI-based therapy as first-line treatment.

Thirty-eight patients experienced a worsening of the ALBI grade (e.g., from ALBI 1 to ALBI 2) during systemic treatment with either ICI or TKI; however, mOS did not differ significantly between these patients and those patients with a stable ALBI grade during therapy (13.1 months vs 12.4 months; p = 0.446). As more patients in the ICI-based group were diagnosed in an earlier tumor stage, we additionally assessed outcomes within BCLC stage B and C. In both subgroups, unadjusted Kaplan–Meier analyses demonstrated significantly prolonged OS for patients receiving ICI-based therapy. The unadjusted Kaplan–Meier analyses are in Figs. IX–XIII of the ESM. However, these analyses should be interpreted cautiously owing to the substantial differences in baseline characteristics between treatment groups.

Subgroup Analyses

Exploratory subgroup analyses were performed to further explore potential differences between TKI and ICI treatment (Table 4). In patients with preserved liver function (ALBI grade 1), TKI treatment was associated with inferior survival compared with ICIs (HR 1.44, 95% CI 1.00–2.07, p = 0.049). In contrast, no significant difference was observed among patients with impaired liver function (ALBI ≥2; HR 1.25, 95% CI 0.76–2.05, p = 0.375).

Table 4.

Subgroup analysis of TKIs versus ICIs

Subgroup	n (TKI/ICI)	HR (TKI vs ICI)	95% CI	p-value
ALBI 1	15 / 20	1.44	1.00–2.07	0.049
ALBI ≥2	56 / 37	1.25	0.76–2.05	0.375
BCLC B	26 / 31	1.31	0.69–2.47	0.408
BCLC C	56 / 36	2.25	0.83–6.07	0.109

ALBI albumin-bilirubin, BCLC Barcelona Clinic Liver Cancer, CI confidence interval, ICI immune checkpoint inhibitor, TKI tyrosine kinase inhibitor; statistically significant p-values are in bold

When stratified by tumor stage, no statistically significant differences were observed between treatment groups. However, in BCLC stage C, a numerical trend toward worse outcomes with TKI treatment was observed (HR 2.25, 95% CI 0.83–6.07), although this did not reach statistical significance (p = 0.109) because of a wide variation of survival in this group. These subgroup analyses should be interpreted carefully, as, because of the small subgroups, the risk of overemphasizing potentially unstable findings is relatively high.

Sensitivity Analyses

Sensitivity analyses confirmed the robustness of the primary findings (Table S2 of the ESM). Restricting follow-up to 24 months did not materially change the results (HR 1.14, p = 0.584), nor did adjustment with a propensity score (HR 0.89, p = 0.562). However, adjustment for calendar time revealed a significant association favoring ICI-based therapy (HR 0.38, 95% CI 0.16–0.89, p = 0.026), suggesting a potential impact of temporal confounding.

Role of Transarterial Chemoembolization (TACE)

Transarterial chemoembolization has long been the standard treatment for patients with BCLC stage B HCC, with evolving practice patterns towards more selective approaches to preserve non-tumorous liver tissue and maintain liver function. In addition, there is an increasing trend towards using systemic therapies also in intermediate stages, reflected by decreased use of TACE as first-line treatment in cohort B (Fig. 1A). We were therefore interested whether these changes in clinical practice were associated with a better outcome in the more recent patients. In the complete case (n = 303) multivariable Cox regression analysis, age, liver function and tumor stage were identified as independent prognostic factors (Table 5). Patients with impaired liver function (ALBI grade ≥2) had significantly worse survival compared with those with ALBI grade 1 (grade 2: HR 1.51; grade 3: HR 1.75). No significant association with survival was observed for sex, cirrhosis, or etiology.

Table 5.

Complete case multivariable Cox regression analysis for overall survival from first TACE (N = 303)

Variable	Category	N		HR	95% CI	p-value
TACE period	2020–22 vs 2013–19	52/251	0.84	0.57–1.23	0.366
Age	Per year	303	1.03	1.01–1.04	< 0.001
Sex	Male vs female	222/81	1.09	0.82–1.45	0.550
ALBI grade	Overall	303	–	–	0.022
	Grade 2 vs 1	201/73	1.51	1.10–2.07	0.010
	Grade 3 vs1	29/73	1.75	1.05–2.93	0.033
BCLC stage	Overall	303	–	–	0.002
	B vs A	159/121	–	–	–
	C vs A	21 121	–	–	–
Cirrhosis	Yes vs no	289/14	1.18	0.59–2.37	0.638
Etiology	Overall	303	–	–	0.540
	ALD vs viral	87/94	1.00	0.72–1.38	0.993
	MASLD vs viral	38/94	0.74	0.48–1.14	0.175
	Other vs viral	84/94	0.94	0.66–1.34	0.724

ALBI albumin-bilirubin, ALD alcoholic liver disease, BCLC Barcelona Clinic Liver Cancer, CI confidence interval, HR hazard ratio, MASLD metabolic dysfunction-associated steatotic liver disease, TACE transarterial chemoembolization; statistically significant (p)-values are in bold

Notably, however, the mOS from the time of first TACE was comparable between the two cohorts (A mOS 22.5 months vs B mOS 20.3 months; HR 0.98; 95% CI 0.67–1.43; p = 0.80) (see Fig. XIV of the ESM). There was also no significant difference in mOS between the cohorts for patients according to ALBI Grade (see Fig. XV/XVI of the ESM).

Discussion and Conclusions

This large real-world study provides comprehensive insights into evolving treatment patterns and outcomes in patients diagnosed with HCC over a decade, with a particular focus on the impact of novel systemic therapies. The approval of Atezo/Bev in 2020 marked a pivotal shift in the treatment landscape for intermediate and advanced HCC, followed by the integration of additional ICI-based combinations such as tremelimumab/durvalumab and, most recently, nivolumab/ipilimumab [9, 13–15]. These therapies have demonstrated mOS rates approaching 2 years in clinical trials; however, trial populations often include a higher proportion of patients with preserved liver function than typically seen in real-world settings [16, 17].

In our study, we observed a substantial shift in treatment allocation over time, characterized by a decreased use of TACE and a significant increase in use of systemic therapies. This reflects the rapid integration of ICI-based regimens into clinical practice following their regulatory approval. Despite the introduction of ICIs in 2020, no significant improvement in OS from the time of diagnosis was observed across the entire cohort in the post-2020 period. This may be attributable to the relatively low proportion of patients receiving ICI-based therapies during the study period, with only 8.4% of patients treated with ICIs, as well as the limited follow-up, as almost 30% of patients in cohort B were still alive at the time of analysis.

In unadjusted analyses, patients receiving ICI-based therapy demonstrated longer mOS from 14.4 months versus 5.9 months for patients receiving ICI-based therapy compared with TKI-based therapy. However, patients treated with ICI had more favorable baseline characteristics, including better liver function, lower tumor stage, and better performance status. After multivariable adjustment and IPTW-based sensitivity analyses, no independent survival difference between treatment groups remained. In the ICI group, 20.8% of the patients showed a long-term survival over 36 months. As anticipated in a real-world setting, the observed mOS in our study remained lower than the mOS observed in studies such as IMbrave150, HIMALAYA, CARES-310, and CheckMate 9DW [15]. It is noteworthy that in our real-world cohort, most patients receiving systemic therapy were treated with Atezo/Bev and were classified as ALBI grade 2, which may contribute to the shorter-than-expected survival outcomes compared with those reported in clinical trials. In line with recent studies, we saw a trend to a prolonged mOS for patients who received more than one therapy line, defining a need for further investigation of optimal sequence therapy [26].

Notably, we did not see any difference in mOS for patients with worsening of ALBI score undergoing systemic therapy in our cohort. In other studies, early hepatic decompensation has been identified as a key negative prognostic factor in patients receiving Atezo/Bev [20]. Unfortunately, hepatic decompensation in our retrospective cohort is not well reported and results may be biased by under-reporting.

Despite these findings, no significant difference in mOS was observed between ICI-based and TKI-based treatments after multivariable adjustment, a result confirmed by propensity score-based IPTW analyses, suggesting that baseline imbalances are unlikely to fully explain this observation. Subgroup analyses indicated a differential effect by liver function, but not by tumor stage; no significant interaction between treatment type and BCLC stage (B vs C) was observed. This suggests that liver function and patient-related factors, rather than tumor stage alone, are key determinants of outcomes under systemic therapies. Given the limited sample sizes, these subgroup findings should be interpreted cautiously and considered exploratory only. Further studies are required to determine whether specific patient subgroups derive differential benefit from ICI-based therapies.

In patients with preserved liver function (ALBI grade 1), ICI-based therapy was associated with improved survival compared with TKIs, whereas no benefit was observed in patients with impaired liver function. The importance of liver function in determining treatment efficacy is further supported by a post-hoc analysis of the IMbrave150 study, where a clear survival benefit for Atezo/Bev was observed in patients with ALBI grade 1, while no significant benefit was seen in those with ALBI grade 2 compared to sorafenib. In contrast, trials such as HIMALAYA, CARES-310, and CheckMate 9DW reported consistent survival benefits for ICI-based therapies across both ALBI grade 1 and 2 subgroups [15, 18, 19]. These observations suggest that preserved liver function may influence outcomes under systemic therapy and remains an important consideration in treatment allocation and patient selection [21–23].

Regarding the etiology of the underlying liver disease, in our study, we did not observe an increase in the number of patients with MASLD, although this would generally be expected based on other studies [27]. However, in approximately 40% of patients in cohort B, the underlying liver disease etiology could not be clearly distinguished, and a single predominant cause could not be reliably assigned. Given the retrospective nature of this analysis, documentation of MASLD may be less consistent than that of other etiologies in routine clinical practice, potentially leading to an underestimation of its prevalence in this cohort. Nonetheless, we observed a significant decline in viral hepatitis over time, accompanied by improved liver function at the time of initial HCC diagnosis. This trend may, at least in part, reflect the increasing availability and effectiveness of direct-acting antiviral therapies for hepatitis C [28]. When analyzing survival outcomes by liver disease etiology, we found that patients with MASLD had the longest mOS of 30.2 months, followed by those with viral hepatitis at 23.8 months. These findings align with recent reports indicating that patients with HCC related to MASLD generally have survival outcomes comparable to those with viral etiologies with reported mOS around 29 months and 5-year survival rates of 18–20% [24]. In contrast, alcoholic liver disease-related HCC is associated with poor outcomes, with 5-year survival rates around 14% and shorter mOS, reflecting the higher prevalence of advanced liver dysfunction and ongoing alcohol use at the time of diagnosis [25]. Together, our observations underscore the impact of underlying liver disease etiology on survival outcomes in patients with HCC.

Regarding interventional treatment, we observed a significant decrease in the use of TACE as first-line treatment following evolving recommendations for a more cautious use of intra-arterial therapies in patients with impaired liver function [28]. However, no significant difference in adjusted OS between the cohorts was observed in patients undergoing TACE. These findings should be interpreted cautiously given potential treatment-selection bias and differences in patient characteristics between time periods. In contrast, front-line use of systemic therapy increased over time, becoming the second most common first-line treatment in cohort B after surgery. Ongoing prospective trials such as HCC-ABC are expected to provide valuable evidence on the comparative efficacy of systemic therapy versus TACE, while also aiding in refining treatment allocation strategies between these modalities for patients with intermediate-stage HCC. In our cohort, liver function remained the most significant prognostic factor for patients treated with TACE. The mOS for ALBI grade 1 patients of 30.8 months underscores the continued relevance of TACE as a treatment option for patients with BCLC stage B HCC with preserved liver function.

Results from trials such as LEAP-012, EMERALD-1, and TALENTACE may further support an evolving role of TACE in combination with systemic therapies [10–12, 29]. While these trials have successfully met their primary endpoints, the integration of these combination strategies into routine clinical practice has been limited to date. Further data on long-term OS and quality-of-life outcomes will be essential to guide the adoption of these treatment combinations in the management of intermediate-stage HCC.

Several limitations should be acknowledged. The retrospective single-center design introduces the potential for selection bias and unmeasured confounding. Although multiple sensitivity analyses were performed, residual confounding cannot be excluded. In particular, treatment allocation was strongly influenced by tumor stage, liver function, performance status, treatment eligibility, and therapeutic era, introducing the potential for indication bias and treatment-selection bias. Moreover, because systemic therapies were initiated at different timepoints during the disease course, immortal time bias and time-dependent confounding cannot be fully excluded despite multivariable adjustment and sensitivity analyses. Therefore, comparisons between treatment groups should be interpreted cautiously and regarded as exploratory and hypothesis generating rather than causal. Additional limitations include the lack of detailed adverse event and quality-of-life data, incomplete data on hepatic decompensation during systemic treatment, and potential selection bias in treatment allocation. Moreover, the limited proportion of patients receiving ICI-based therapies may affect the interpretation of survival outcomes.

Nevertheless, our data demonstrate a clear shift in treatment strategies over time, with increasing use of systemic therapies in patients with intermediate-stage HCC. However, this evolution has not yet translated into improved OS at the population level. Across all analyses, liver function consistently emerged as the most important determinant of outcome. Although unadjusted analyses suggested longer survival with ICI-based therapy, these differences were not confirmed after adjustment for baseline imbalances. Across all analyses, liver function remained the dominant prognostic factor in patients with HCC.

Supplementary Information

Below is the link to the electronic supplementary material.

Funding

Open Access funding enabled and organized by Projekt DEAL. No external funding was used in the preparation of this article.

Declarations

Conflicts of Interest

Melanie Bathon: honoraria and/or travel support from AstraZeneca, Falk, Galapagos, Jazz Pharma, MSD, and Pfizer. Nele Binder, Sabrina Welland, and Christoph Gerdes have no conflicts of interest that are directly relevant to the content of this article. Thomas Wirth: consultancies: Ipsen, Roche, Novartis, BMS, MSD, Daiichi Sankyo, Servier, Merck, AstraZeneca, Pierre Fabre, Viatris, Celltrion, Dr. Falk Pharma, Delcath, Immunocore, Astellas, Jazz Pharmaceuticals, and Novocure; research funding: Hookipa Biotech and BMS. Kristina I. Ringe: advisory role and honoraria: AstraZeneca, Bayer Healthcare, Boston Scientific Siemens Healthineers, and Ipsen. Markus Quante has no conflicts of interest that are directly relevant to the content of this article. Arndt Vogel: consultancy and advisory role: Roche, AstraZenca, Böhringer-Ingelheim, Ipsen, Incyte, Cogent, EISAI, Zymeworks, Biologix, BMS, Terumo, Elevar, Servier,MSD Tahio, Jazzpharma, Medivir, AbbVie, and Tyra. Anna Saborowski: honoraria and/or travel support from BMS, Roche, Servier, Ipsen, Lilly, AstraZeneca, MSD, Eisai, Amgen, Jazz Pharma, Taiho, and Pierre-Fabre.

Ethics Approval

The study protocol followed the ethical guidelines of the 1975 Declaration of Helsinki (ethical approval ID 940-2011).

Consent to Participate

Because of the retrospective design of the study, informed consent was waived according to institutional regulations.

Consent for Publication

Not applicable.

Availability of Data and Material

The datasets generated and/or analyzed during the current study are not publicly available because of institutional data protection regulations and patient confidentiality requirements but are available from the corresponding author on reasonable request. Access to anonymized data may require approval by the local ethics committee in accordance with institutional policies.

Code Availability

Not applicable.

Authors’ Contributions

MB: data curation, formal analysis, writing (original draft). NB: data curation, formal analysis, writing (original draft). SW: data curation. CG: data curation. TW: conceptualization, supervision. KIR: methodology, radiological assessment. MQ: methodology, surgical expertise, data interpretation. AV: conceptualization, supervision, writing (review and editing). AS: conceptualization, supervision, writing (review and editing).