Alpha-Fetoprotein and Des-Gamma-Carboxy Prothrombin-Based Tumor Marker Score for First-Line Immunotherapy Selection in Hepatocellular Carcinoma.

Introduction
Systemic pharmacotherapy for unresectable advanced hepatocellular carcinoma (HCC) has made remarkable progress since the introduction of sorafenib in 2007. Among these advancements, the atezolizumab plus bevacizumab (Atez/Bev) regimen, introduced in 2020, and the durvalumab plus tremelimumab (Dur/Tre) regimen, introduced in 2022, have had a significant impact on HCC treatment. These regimens have resulted in cases exhibiting the characteristic “tail plateau” associated with immune checkpoint inhibitors (ICIs), with increasing reports of long-term survival.
The IMbrave150 trial [1] and the HIMALAYA trial [2] have provided evidence demonstrating the superiority of Atez/Bev and Dur/Tre over sorafenib, respectively, leading to their approval as first-line treatments for HCC. The survival data from these two trials suggest comparable overall survival (OS) between Atez/Bev and Dur/Tre. However, no direct comparative studies have been conducted. Moreover, the criteria for selecting between these two treatment options remain unclear.
Previously, we reported that a tumor marker score composed of alpha-fetoprotein (AFP), AFP-L3, and des-gamma-carboxy prothrombin (DCP) could stratify OS and progression-free survival (PFS) in patients treated with Atez/Bev [3]. However, no such reports exist for Dur/Tre, and its usage remains unknown. This study was designed to evaluate whether a modified tumor marker (mTM) score can serve as a useful criterion for selecting between Atez/Bev and Dur/Tre and in stratifying treatment outcomes of unresectable HCC.

Methods

Patients
Patients diagnosed with unresectable advanced HCC between 2020 and 2024 and treated with either Atez/Bev or Dur/Tre were included in this study. Eligible patients had no prior exposure to ICIs, were classified as Barcelona Clinic Liver Cancer (BCLC) stage B or C, and had baseline levels of AFP and DCP assessed before treatment initiation.
Clinical data were collected from the participating institutions, including sex, age, body mass index, Eastern Cooperative Oncology Group Performance Status (ECOG-PS), underlying liver disease, presence of extrahepatic metastases, vascular invasion, BCLC stage, and laboratory parameters such as white blood cell count, neutrophil-to-lymphocyte ratio, platelet count, albumin, total bilirubin, aspartate aminotransferase, alanine aminotransferase, lactate dehydrogenase, creatinine, C-reactive protein, AFP, and DCP. The number of intrahepatic tumors was categorized as follows: 0 (no intrahepatic tumor), 1 (solitary), 2–3 (oligonodular), >4 (multinodular), diffuse (uncountable lesions), and unknown. The size of intrahepatic tumors was categorized as follows: none, ≤2 cm, 2.1–5 cm, >5 cm, unmeasurable, and unknown. Patients categorized as having “no intrahepatic tumor” were those with extrahepatic metastases only without any detectable liver lesions. “Unmeasurable” was defined as cases in which the intrahepatic lesion(s) could not be clearly delineated or measured because of diffuse infiltration or technical limitations.

Assessment Tool for Hepatic Reserve Function
In addition to the conventional Child-Pugh classification [4], hepatic reserve function was assessed in greater detail using the modified albumin-bilirubin (mALBI) grading system [5]. This system refines the ALBI grade [6] by subdividing grade 2 at an ALBI score of −2.27 into two subgrades, allowing for a more precise evaluation of liver function.

Therapeutic Strategy for HCC
The treatment strategy for HCC was determined in accordance with the Japan Society of Hepatology guidelines [7] and was finalized by the attending physician in shared decision-making with the patient. The decision to initiate Atez/Bev or Dur/Tre was made by the attending physicians at each participating institution.
Atez/Bev was administered as an intravenous injection of 1,200 mg of atezolizumab and 15 mg/kg of bevacizumab every 3 weeks [8]. Dur/Tre was administered as a single dose of 300 mg of tremelimumab and 1,500 mg of durvalumab on day 1, followed by 1,500 mg of durvalumab every 4 weeks [2]. Treatment was discontinued in cases of unacceptable severe adverse events or clinically confirmed tumor progression.

Evaluation Methods of Therapeutic Effects
Atez/Bev and Dur/Tre were continued until progressive disease (PD) was determined or treatment could no longer be sustained owing to adverse events (AEs). Antitumor efficacy was assessed at each institution using the Response Evaluation Criteria in Solid Tumors version 1.1 [9], classifying responses as complete response (CR), partial response (PR), stable disease (SD), or PD.
The objective response rate (ORR) was defined as the sum of CR and PR, whereas the disease control rate (DCR) was defined as the sum of CR, PR, and SD. Unless otherwise specified, The ORR and DCR were evaluated based on the best overall response. PFS was defined as the time from the initiation of Atez/Bev or Dur/Tre to the occurrence of PD. OS was defined as the time from the initiation of Atez/Bev or Dur/Tre to the patient’s death. AEs were evaluated using the Common Terminology Criteria for Adverse Events version 5.0 (https://ctep.cancer.gov/protocoldevelopment/electronic_applications/ctc.htm).

Development of mTM Score
The mTM score was calculated based on AFP and DCP, both measured before the initiation of Atez/Bev or Dur/Tre therapy. According to previous reports [3, 10], elevations in AFP (≥100 ng/mL) and DCP (≥100 mAU/mL) levels were each assigned 1 point. The total prognostic score was determined by summing the number of elevated markers, with all patients categorized into three groups (score 0, 1, or 2) based on their TM score.

Statistical Analysis
Categorical variables were presented as numbers (percentages) and analyzed using the chi-squared test or Fisher’s exact test. Continuous variables were expressed as medians (interquartile range) and analyzed using the Mann-Whitney U test.
Survival curves were estimated using the Kaplan-Meier method and compared with the log-rank test. Multivariable analysis was performed using Cox regression analysis.
To assess the relationship between AFP, DCP, and PFS for each treatment, a restricted cubic spline regression was applied. Inverse probability of treatment weighting (IPTW) was employed to minimize potential confounding factors in overall comparisons by adjusting for sex, ECOG-PS, mALBI grade, BCLC stage, treatment line, and observation period.
All reported p values were two-sided, and statistical significance was defined as p < 0.05. Statistical analyses were conducted using EZR version 1.68 (Jichi Medical University Saitama Medical Center, Saitama, Japan) [11].

Results

Patient Characteristics
A total of 1,313 patients were included (Atez/Bev: 1,157; Dur/Tre: 156). Baseline characteristics are summarized in Table 1. The median age was 74 years in both groups, and the majority of patients were male. Compared with the Atez/Bev group, the Dur/Tre group had a higher proportion of patients with poorer liver function (Child-Pugh score ≥7, mALBI grade 2b or higher), and a slightly higher frequency of extrahepatic metastases and macrovascular invasion. Non-viral etiology was more common in the Dur/Tre group. Atez/Bev was used more frequently as first-line therapy, whereas TKI was used more frequently in the Atez/Bev group. The median observation period was significantly shorter in the Dur/Tre group. The AFP and DCP levels were slightly but not significantly higher in the Dur/Tre group.

Efficacy of Atez/Bev and Dur/Tre
The ORR and DCR were 29.8% (326 of 1,157 patients) and 77.4% (847 of 1,157 patients), respectively, in the Atez/Bev group, and 27.1% (38 of 156 patients) and 67.1% (94 of 156 patients), respectively, in the Dur/Tre group. Although no significant difference was observed in ORR (p = 0.556), DCR was significantly higher in the Atez/Bev group than in the Dur/Tre group (p = 0.011) (Table 2).
The PFS was 7.5 months (95% CI, 6.8 to 8.1) in the Atez/Bev group and 6.0 months (95% CI, 4.4 to 7.2) in the Dur/Tre group, with no significant difference (p = 0.086). This result remained unchanged after IPTW analysis (p = 0.074) (Fig. 1a). At the time of analysis, disease progression was observed in 836 patients (72.3%) in the Atez/Bev group and in 91 patients (58.3%) in the Dur/Tre group.
The OS was 21.3 months (95% CI, 19.2 to 24.7) in the Atez/Bev group and 17.9 months (95% CI, 14.1 to not available) in the Dur/Tre group, with no significant difference (p = 0.414). This finding remained consistent after IPTW analysis (p = 0.436) (Fig. 1b). At the time of analysis, death had occurred in 613 patients (53.0%) in the Atez/Bev group and in 45 patients (28.8%) in the Dur/Tre group.
In the first-line treatment, the median PFS was 7.9 months (95% CI, 7.1 to 9.0) with Atez/Bev and 6.1 months (95% CI, 5.4 to 7.8) with Dur/Tre, with a significantly longer PFS in the Atez/Bev group (p = 0.028). However, there was no significant difference in OS between the two groups (p = 0.439) (online suppl. Fig. 1a, b; for all online suppl. material, see https://doi.org/10.1159/000547519). Conversely, in patients with a history of prior TKI treatment, PFS and OS did not differ significantly between the two treatment groups (p = 0.616 and p = 0.605, respectively) (online suppl. Fig. 1c, d).

mTM Score
In the Atez/Bev group, 465 patients (40.2%) had AFP levels of 100 ng/mL or higher, and 728 patients (62.9%) had DCP levels of 100 mAU/mL or higher. In the Dur/Tre group, 68 patients (43.6%) had AFP levels of 100 ng/mL or higher, and 111 patients (71.2%) had DCP levels of 100 mAU/mL or higher.
The distribution of the mTM scores, classified as 0, 1, or 2, was as follows. In the Atez/Bev group, 324 patients (28.0%) had a score of 0, 473 patients (40.9%) had a score of 1, and 360 patients (31.1%) had a score of 2. In the Dur/Tre group, 35 patients (22.4%) had a score of 0, 63 patients (40.4%) had a score of 1, and 58 patients (37.2%) had a score of 2. There was no significant difference in the score distribution between the groups (p = 0.209).
For patients with AFP levels below 100 ng/mL, PFS was significantly longer in the Atez/Bev group than in the Dur/Tre group. The median PFS was 9.1 months (95% CI, 8.0 to 10.5) in the Atez/Bev group and 5.6 months (95% CI, 4.1 to 7.3) in the Dur/Tre group (p = 0.004).
Similarly, OS was significantly longer in the Atez/Bev group. The median OS was 26.1 months (95% CI, 24.7 to 29.8) in the Atez/Bev group and 17.9 months (95% CI, 12.5 to not available) in the Dur/Tre group (p = 0.033).
For patients with DCP levels below 100 mAU/mL, PFS was significantly longer in the Atez/Bev group than in the Dur/Tre group. The median PFS was 9.6 months (95% CI, 8.0 to 11.5) in the Atez/Bev group and 5.0 months (95% CI, 2.4 to 5.9) in the Dur/Tre group (p < 0.001). Similarly, OS tended to be longer in the Atez/Bev group, with a median OS of 26.0 months (95% CI, 24.7 to 30.6) in the Atez/Bev group and 17.0 months (95% CI, 11.4 to not available) in the Dur/Tre group (p = 0.060).
In contrast, for patients with AFP levels of 100 ng/mL or higher, there were no significant differences between the two groups in terms of PFS or OS. The median PFS was 5.3 months (95% CI, 4.5 to 6.3) in the Atez/Bev group and 6.8 months (95% CI, 3.1 to 10.4) in the Dur/Tre group (p = 0.399).
The median OS was 16.7 months (95% CI, 14.3 to 19.4) in the Atez/Bev group, whereas the median OS in the Dur/Tre group was not available (95% CI, 13.6 to not available) (p = 0.308). Similarly, for patients with DCP levels of 100 mAU/mL or higher, there was no significant difference in the PFS or OS between the two groups. The median PFS was 6.3 months (95% CI, 5.6 to 7.5) in the Atez/Bev group and 7.0 months (95% CI, 4.4 to 10.4) in the Dur/Tre group (p = 0.531).
The median OS was 18.1 months (95% CI, 16.3 to 20.3) in the Atez/Bev group and 18.2 months (95% CI, 13.6 to not available) in the Dur/Tre group (p = 0.711). Since the survival curves for both AFP levels below 100 ng/mL and DCP levels below 100 mAU/mL effectively stratified prognoses between Atez/Bev and Dur/Tre, patients who met both criteria (TM score of 0) were classified as having a low tumor marker score (mTM low). Patients with TM scores of 1 or 2 were classified as having a high tumor marker score (mTM high).

Prognostic Stratification Ability of the mTM Score in Atez/Bev Cases
In patients treated with Atez/Bev, radiological response showed no significant difference in the ORR between the mTM low and mTM high groups. However, the DCR was significantly higher in the mTM low group at 84.3% than in the mTM high group at 74.8% (p = 0.001). At the time of the initial response evaluation, the ORR and PD rates were 24.3% (73/301) and 15.3% (46/301), respectively, in the mTM low group, and 18.5% (144/778) and 24.6% (191/778), respectively, in the mTM high group. A greater proportion of patients in the mTM low group achieved the ORR at the initial assessment, while the mTM high group had a higher proportion of patients with PD (p = 0.004).
The duration of Atez/Bev administration was significantly longer in the mTM low group (median of 7.7 months [95% CI, 3.2 to 16.6]) than in the mTM high group (5.8 months (95% CI, 2.4 to 12.7) (p < 0.001). PFS was significantly longer in the mTM low group, with a median PFS of 10.6 months (95% CI, 9.0 to 13.1), than in the mTM high group with a median of 6.6 months (95% CI, 5.8 to 7.5) (p < 0.001) (Fig. 2a). Similarly, OS was significantly longer in the mTM low group, with a median OS of 29.7 months (95% CI, 25.0 to 32.7), than in the mTM high group, with a median OS of 18.5 months (95% CI, 16.8 to 20.8) (p < 0.001) (Fig. 2b).

Prognostic Stratification Ability of the mTM Score in Dur/Tre Cases
In patients treated with Dur/Tre, the ORR was significantly higher in the mTM high group at 31.5% than in the mTM low group at 12.5% (p = 0.041). However, there was no significant difference in the DCR between the two groups (p = 0.833).
At the time of the initial response evaluation, the ORR and PD rates were 6.2% (2/32) and 37.5% (12/32), respectively, in the mTM low group, and 27.8% (30/108) and 32.4% (35/108), respectively, in the mTM high group. There was no significant difference in the proportion of patients with PD at the time of the initial response assessment; however, the proportion of patients who achieved an objective response was significantly higher in the mTM high group (p = 0.038).
The duration of Dur/Tre administration was 2.6 months (95% CI, 1.1 to 6.6) in the mTM low group and 2.4 months (95% CI, 1.0 to 6.5) in the mTM high group, with no significant difference between the two groups (p = 0.899). PFS was 4.5 months (95% CI, 2.4 to 6.2) in the mTM low group and 6.5 months (95% CI, 4.4 to 9.7) in the mTM high group, indicating better outcomes in the mTM high group (p = 0.072) (Fig. 2c). This finding contrasts with the trend observed in the Atez/Bev group. OS was 17.0 months (95% CI, 11.4 to not available) in the mTM low group and 18.2 months (95% CI, 14.2 to not available) in the mTM high group, with no significant difference between the groups (p = 0.703) (Fig. 2d).

The mTM Score Can Stratify PFS and OS in Patients with HCC Undergoing Atez/Bev and Dur/Tre
The patient background characteristics for the mTM low and mTM high groups across treatment regimens are summarized in online supplementary Table 1. In the mTM low group, the ORR was significantly higher in the Atez/Bev group than in the Dur/Tre group. The ORRs were 33.8% in the Atez/Bev group and 12.5% in the Dur/Tre group (p = 0.015). Similarly, the DCR was significantly higher in the Atez/Bev group, with rates of 84.3% and 65.6%, respectively (p = 0.014).
In contrast, among the patients in the mTM high group, there was no significant difference in radiological response between the two treatment regimens (Table 3). Regarding PFS, in the mTM low group, the median PFS was 10.6 months (95% CI, 9.0 to 13.1) in the Atez/Bev group, which was significantly longer than 4.5 months (95% CI, 2.4 to 6.2) in the Dur/Tre group (p < 0.001). This trend remained consistent even after IPTW adjustment, with a p value of <0.001 (Fig. 3a).
In the mTM high group, the PFS was similar between the Atez/Bev and Dur/Tre groups. The median PFS was 6.6 months (95% CI, 5.8 to 7.5) in the Atez/Bev group and 6.5 months (95% CI, 4.4 to 9.7) in the Dur/Tre group, with a p value of 0.780. The results remained unchanged after IPTW adjustment, with a p value of 0.446 (Fig. 3b).
Regarding OS, in the mTM low group, the median OS was 29.7 months (95% CI, 25.0 to 32.7) in the Atez/Bev group, which was significantly longer than 17.0 months (95% CI, 11.4 to not available) in the Dur/Tre group (p = 0.023). This trend remained consistent after IPTW adjustment, with a p value of 0.029 (Fig. 3c).
In the mTM high group, OS was comparable between the Atez/Bev and Dur/Tre groups. The median OS was 18.5 months (95% CI, 16.8 to 20.8) in the Atez/Bev group and 18.2 months (95% CI, 14.2 to not available) in the Dur/Tre group, with a p value of 0.793. This result remained unchanged after IPTW adjustment, with a p value of 0.680 (Fig. 3d).

Factors Associated with PFS and OS
A Cox proportional hazards regression analysis was performed to identify factors influencing PFS. In the mTM low group, Atez/Bev was identified as an independent factor associated with improved PFS, along with mALBI grade. In contrast, in the mTM high group, only mALBI grade and macrovascular invasion were significantly associated with PFS, whereas the treatment regimen was not a significant factor (Table 4).
For OS, analysis showed that in the mTM low group, Atez/Bev was an independent factor for improved OS, along with mALBI grade 1 and 2a and the absence of macrovascular invasion. In contrast, in the mTM high group, only mALBI grade and macrovascular invasion were independent prognostic factors, and the treatment regimen was not associated with OS (Table 5).

Relationship between AFP, DCP, and PFS in Each Treatment Group
The mTM high group was further subdivided into score 1 and score 2 for analysis. In the score 1 subgroup, the PFS was 7.9 months (95% CI, 7.0 to 9.0) in the Atez/Bev group and 6.0 months (95% CI, 4.1 to 9.7) in the Dur/Tre group, with no significant difference (p = 0.346) (online suppl. Fig. 2a).
In the score 2 subgroup, the median PFS was 4.5 months (95% CI, 3.8 to 5.6) in the Atez/Bev group and 7.0 months (95% CI, 2.9 to 10.8) in the Dur/Tre group. Although PFS tended to be longer in the Dur/Tre group, the difference was not statistically significant (p = 0.154) (online suppl. Fig. 2b). As the score increased, PFS decreased in the Atez/Bev group and it tended to increase in the Dur/Tre group. Given this trend, a restricted cubic spline regression analysis was performed to examine the relationship between AFP, DCP, and PFS in each treatment regimen.
In the Atez/Bev group, an increase in AFP was significantly associated with a shorter PFS (p < 0.001). The AFP level at which the log relative hazard reached zero was 89 ng/mL. Similarly, an increase in DCP was also associated with a shorter PFS (p < 0.001), with the DCP level at which the log relative hazard reached zero being 402 mAU/mL (Fig. 4a, b).
In contrast, in the Dur/Tre group, an increase in DCP was significantly associated with a longer PFS (p = 0.037). The beneficial effect of increased DCP on PFS continued to rise, with the zero-log relative hazard point at 400 mAU/mL.
For AFP, the log relative hazard fell below zero in the range of 12 to 14,600 ng/mL. However, nonlinear analysis did not demonstrate a significant association between AFP levels and PFS (p = 0.356) (Fig. 4c, d).

Survival Outcomes Stratified Radiological Response
To further evaluate the association between radiological response and prognosis, Kaplan-Meier curves for OS were plotted according to the radiological response within each treatment group and mTM score subgroup. In the Atez/Bev cohort, among patients with low mTM expression, OS was not reached (NA) in those with CR (95% CI, 32.0 to NA), 32.7 months in those with PR (95% CI, 25.9 to NA), 26.0 months in those with SD (95% CI, 23.9 to 32.2), and 14.8 months in those with PD (95% CI, 10.8 to 25.0), showing a significant difference based on the radiological response (p < 0.001) (online suppl. Fig. 3a). Similarly, in the high mTM expression group, OS was NA in the CR group (95% CI, NA to NA), 33.6 months in the PR group (95% CI, 27.6 to 39.3), 19.2 months in the SD group (95% CI, 16.9 to 22.3), and 9.5 months in the PD group (95% CI, 7.5 to 10.4), similarly demonstrating significant differences (p < 0.001) (online suppl. Fig. 3b). In contrast, in the Dur/Tre cohort, among patients with low mTM expression, OS was NA in the CR group (95% CI, NA to NA), 11.4 months in the PR group (95% CI, 11.4 to NA), 17.0 months in the SD group (95% CI, 17.0 to NA), and 10.4 months in the PD group (95% CI, 5.1 to NA), with a significant difference observed (p = 0.011) (online suppl. Fig. 3c). Among patients with high mTM expression, OS was NA in the CR and PR groups (95% CI, NA to NA for both), 19.2 months in the SD group (95% CI, 13.6 to NA), and 10.5 months in the PD group (95% CI, 5.1 to NA), with only the PD group demonstrating poor outcomes (p < 0.001) (online suppl. Fig. 3d).

Reasons for Treatment Discontinuation and Subsequent Therapy
Treatment discontinuation occurred in 994 patients (85.9%) in the Atez/Bev group and 119 patients (76.3%) in the Dur/Tre group. Discontinuation due to adverse events tended to be more frequent in the Dur/Tre group. In the Atez/Bev group, approximately 10% of the patients discontinued treatment because of a decline in performance status or worsening liver function. The rate of transition to subsequent therapy was higher in the Dur/Tre group (77.0%) than in the Atez/Bev group (65.4%). Among subsequent therapies, TKIs were most commonly used after Atez/Bev, whereas alternative ICI regimens were more frequently administered following Dur/Tre (online suppl. Table 2).

Safety Profile
Among the 1,157 patients in the Atez/Bev group, 729 patients (63.0%) experienced treatment-related AEs of any grade. Similarly, among the 156 patients in the Dur/Tre group, 98 patients (62.8%) experienced treatment-related AEs. There was no significant difference between the two groups (p = 1.000).
Immune-related AEs (irAEs) of any grade occurred in 409 patients (35.4%) in the Atez/Bev group and in 88 patients (56.4%) in the Dur/Tre group, with a significantly higher incidence in the Dur/Tre group (p < 0.001). Similarly, Grade 3 or higher irAEs were more frequent in the Dur/Tre group, occurring in 30 patients (19.4%) compared with 85 patients (7.3%) in the Atez/Bev group.
In the Atez/Bev group, proteinuria was observed in 353 patients (30.5%) and hypertension in 187 patients (16.2%), both occurring more frequently than in the Dur/Tre group.
In contrast, in the Dur/Tre group, rash was reported in 42 patients (26.9%), colitis/diarrhea in 27 patients (17.3%), and pulmonary toxicity in 9 patients (5.8%), all occurring more frequently than in the Atez/Bev group (online suppl. Table 3).
There was no significant difference in these trends between the mTM low and mTM high groups. A total of seven Grade 5 adverse events were reported, all of which occurred in the Atez/Bev group (p = 1.000). These included four cases of drug-induced pneumonitis, 2 cases of liver dysfunction, and 1 case of colitis.

Discussion
This study revealed three novel findings. First, in patients with unresectable advanced HCC, there was no significant difference in the PFS or OS between Atez/Bev and Dur/Tre. When the analysis was restricted to first-line treatment, Atez/Bev was associated with a longer PFS than Dur/Tre, although the OS remained comparable between the two regimens. Second, among the patients in the mTM low group, Atez/Bev demonstrated superior efficacy relative to Dur/Tre. Third, the therapeutic efficacy of Atez/Bev decreased with increasing AFP and DCP levels, whereas the therapeutic efficacy of Dur/Tre appeared to be less influenced by AFP, and an elevated DCP level was associated with prolonged PFS.
Since its approval in 2020, Atez/Bev has demonstrated superior PFS and OS compared with conventional TKIs. The IMbrave150 trial reported an ORR of 29.8% for Atez/Bev, substantiating its clinical benefit in advanced HCC [1]. This high response rate has provided significant therapeutic advantages for patients. Hatanaka et al. [12] reported that patients who underwent conversion therapy following Atez/Bev treatment achieved survival outcomes comparable to those who attained CR with systemic therapy. The introduction of Dur/Tre in 2022 expanded the therapeutic landscape of HCC by providing an alternative ICI-based regimen. To the best of our knowledge, no prior studies have directly compared these two regimens. This study represents the first large-scale investigation to evaluate the comparative efficacy of Atez/Bev and Dur/Tre. Although Atez/Bev was associated with a marginally longer PFS in treatment-naïve patients, OS was not significantly different between the two regimens, suggesting that both may provide comparable clinical benefits as first-line ICI-based therapies.
To date, limited evidence has been available regarding treatment selection based on predictive biomarkers for therapeutic efficacy, and treatment decisions have largely relied on physician discretion. In clinical practice, patients with contraindications to vascular endothelial growth factor (VEGF) inhibitors, such as severe proteinuria or high-risk esophageal varices, are generally recommended to receive Dur/Tre instead of Atez/Bev. However, in patients eligible for VEGF inhibitors, no validated biomarkers have been established to guide therapeutic decision-making. Hiraoka et al. [13] reported unfavorable prognoses in patients who received Dur/Tre following Atez/Bev failure, yet no prior studies have identified predictive markers for selecting the optimal first-line treatment.
Our previously reported tumor marker score, which included AFP, AFP-L3, and DCP, showed no significant correlation between AFP-L3 and PFS [3]. Therefore, in the present study, AFP-L3 was excluded, and the tumor marker score was redefined as an mTM score, based only on AFP and DCP. The mTM score provided insights into the distinct antitumor effects of Atez/Bev and Dur/Tre. The findings for Atez/Bev were consistent with our previous reports, where higher tumor marker levels were associated with poorer treatment outcomes. Additionally, both AFP and DCP levels independently and linearly correlated with shorter PFS in patients receiving Atez/Bev.
In contrast, AFP levels were not associated with PFS in patients treated with Dur/Tre, and elevated DCP was correlated with improved treatment outcomes. A higher tumor burden has generally been associated with diminished ICI efficacy. Previous studies in lung cancer, malignant melanoma, and head and neck squamous cell carcinoma have reported inferior ICI efficacy in patients with greater tumor volume [14–16]. Given that AFP and DCP are widely recognized surrogate markers for tumor burden [17, 18], patients with high mTM scores may be presumed to have a greater tumor burden than patients with low mTM scores. The observed decrease in Atez/Bev efficacy in mTM high patients is consistent with findings in other malignancies and may be attributed to deterioration of the immune microenvironment as tumor burden increases [19], as well as direct effects of elevated AFP and DCP levels.
Several mechanisms may underlie the association between elevated AFP and reduced efficacy of Atez/Bev. First, in HCC, high AFP levels correlate with poor tumor differentiation, which has been linked to diminished ICI efficacy [20]. Second, HCC with high AFP expression exhibits upregulation of VEGF, particularly VEGF-B [21]. VEGF-B binds to VEGFR1, suppressing the NF-κB pathway and inhibiting DC maturation [22]. Additionally, increased VEGF-B may reduce VEGF-A binding to VEGFR1, thereby promoting VEGFR2 activation [21]. Moreover, VEGF upregulates PD-1 and myeloid-derived suppressor cells while polarizing tumor-associated macrophages toward an immunosuppressive M2 phenotype [23, 24]. Third, Takahashi et al. [25] reported that in AFP-producing gastric cancer, high AFP levels were associated with increased vascular density and VEGF expression. Furthermore, administration of anti-AFP antibodies suppressed tumor angiogenesis, suggesting that AFP itself may contribute to tumor neovascularization.
DCP has been implicated in tumor angiogenesis in HCC and may attenuate the efficacy of VEGF inhibitors [26]. However, in the present study, Dur/Tre demonstrated improved PFS in patients with high mTM scores, particularly in patients with elevated DCP levels. Anti-CTLA-4 antibodies such as tremelimumab not only enhance T-cell priming in the cancer immunity cycle but also exert a regulatory Treg-depleting effect [27]. Additionally, CTLA-4 is expressed on DCs, where it inhibits DC maturation and antigen-presenting capacity, thereby suppressing T-cell priming [28]. The observed association between elevated DCP and improved response to Dur/Tre suggests that DCP may serve as a biomarker indicative of a tumor microenvironment responsive to anti-PD-L1 plus anti-CTLA-4 therapy, warranting further investigation.
The present findings suggest that the mTM score may serve as a potential biomarker for treatment selection in first-line immunotherapy for advanced HCC. Among patients with low mTM scores, Atez/Bev demonstrated superior ORR and DCR compared with Dur/Tre (ORR: 33.8% vs. 12.5%, p = 0.015; DCR: 84.3% vs. 65.6%, p = 0.014). PFS (11.5 months vs. 4.4 months) and OS (30.6 months vs. not reached) were also significantly longer with Atez/Bev. These findings support the use of Atez/Bev as the preferred first-line therapy for patients with low mTM scores. Conversely, in patients with high mTM scores, the efficacy of Atez/Bev decreased with increasing AFP and DCP levels, with a log relative hazard ratio for PFS exceeding 0 at AFP ≥100 ng/mL and DCP ≥400 mAU/mL. In contrast, for Dur/Tre, the log relative hazard ratio for PFS was below 0 when DCP >400 mAU/mL, suggesting that Dur/Tre may be the preferred first-line option in such patients (Fig. 4).
However, the optimal treatment strategy for patients with AFP ≥100 ng/mL and DCP <400 mAU/mL remains unclear. Combination therapy with Atez/Bev and transarterial chemoembolization may represent a promising approach. Saeki et al. [29] proposed that early changes in AFP and DCP levels following Dur/Tre treatment may predict therapeutic response, underscoring the importance of early assessment strategies and timely treatment modification in non-responders. Kudo et al. [30] stated that Dur/Tre has an all-or-nothing response pattern, and that early switching (within 3–4 weeks) based on the lack of an early biomarker response using tumor markers such as AFP and DCP may be key to improving prognosis. Sequential therapy from Dur/Tre to Atez/Bev or vice versa may improve the prognosis of patients in this group.
Actually, in the Dur/Tre cohort, 43.7% of patients in the mTM low group and 60.2% in the mTM high group exhibited an all-or-nothing response pattern. Notably, in the mTM low group, the majority of cases exhibiting an all-or-nothing pattern showed PD, whereas in the mTM high group, the cases were more evenly split between objective response (ORR) and PD.
Although AFP-L3 was not included in the present analysis, Kudo et al. have emphasized its importance in assessing treatment response to Atez/Bev [31]. Future studies should also evaluate AFP-L3, given its potential clinical significance.
The incidence of irAEs was higher with Dur/Tre, and irAEs were the most common reason for treatment discontinuation in the Dur/Tre group. However, the transition rate to subsequent treatments was favorable in the Dur/Tre group. Proper management of irAEs may help salvage the prognosis of HCC patients who begin treatment with Dur/Tre. Additionally, TKIs were frequently used as subsequent therapy after Atez/Bev, whereas Atez/Bev was commonly used after Dur/Tre. This difference is likely attributable to the historical context in which Atez/Bev was the only available ICI treatment between 2020 and 2022.
This study has several limitations. First, although this was a large-scale multicenter study, the Dur/Tre cohort was relatively small. Despite efforts to minimize bias using IPTW analysis, residual bias cannot be entirely excluded. Furthermore, the number of patients in the Dur/Tre group with mTM low status was relatively limited (n = 35), and a substantial proportion of these patients had non-viral etiologies, which differed from the Atez/Bev group. Although IPTW adjustment was applied, the small sample size and etiologic imbalance may have affected the robustness of the subgroup analyses.
Second, the follow-up period may have been insufficient to draw definitive conclusions. Prospective studies with longer observation periods are needed to confirm these findings. Third, the cutoff values for AFP and DCP identified through restricted cubic spline regression have not yet undergone external validation, raising concerns of overfitting. Large-scale validation studies are necessary to establish optimal cutoff values for distinguishing between Atez/Bev and Dur/Tre responders. Additionally, we adopted 100 ng/mL for AFP and 100 mAU/mL for DCP as the cutoff values to define the mTM score. These thresholds were chosen based on their frequent use in prior studies and their clinical interpretability. Notably, our restricted cubic spline analysis revealed similar inflection points (AFP: 89 ng/mL, DCP: 402 mAU/mL), supporting the appropriateness of these cutoffs. However, given the exploratory nature of this analysis, further prospective studies are warranted to validate the optimal thresholds for treatment stratification.
Fourth, treatment availability changed during the study period. Atez/Bev was approved in 2020, whereas Dur/Tre became available only after 2022. Therefore, patients enrolled earlier in the study period were more likely to receive Atez/Bev, which may have introduced a temporal bias in treatment selection.

Conclusion
The mTM score, composed of AFP and DCP, is a simple and intuitive tool that effectively identifies patients who are more likely to benefit from Atez/Bev. In cases with high DCP levels, initiating treatment with Dur/Tre may lead to improved prognosis in patients with HCC.

Acknowledgment
We thank Dr. Edward Barroga (https://orcid.org/0000-0002-8920-2607), Medical Editor, Professor at Showa Medical University, Japan, and former Professor at St. Luke’s International University, Japan, for editing the manuscript.

Statement of Ethics
The entire study protocol was approved by the Institutional Ethics Committee of NHO Takasaki General Medical Center (IRB No. TGMC2024-03). After receiving official approval, this study was conducted as a retrospective analysis of database records based on the Guidelines for Clinical Research issued by the Ministry of Health and Welfare of Japan. All procedures were performed in accordance with the Declaration of Helsinki. The data were made anonymous before analysis to protect patient privacy. Written informed consent was obtained from all patients before treatment. Informed consent for the analysis was obtained from each patient prior to Clinical Research Committee approval with an opt-out option, while written informed consent was obtained from each patient after approval.

Conflict of Interest Statement
Atsushi Hiraoka, MD, PhD, received lecture fees from Eli Lilly, AstraZeneca, and Chugai. Toshifumi Tada, MD, PhD, received support from AbbVie, Eisai, AstraZeneca, and Chugai. Satoru Kakizaki, MD, PhD, received lecture fees from AbbVie. Hidenori Toyoda, MD, PhD, received lecture fees from AbbVie, Gilead Sciences, Takeda, Eisai, Kowa, Terumo, Fujifilm WAKO, Chugai, AstraZeneca, and Bayer. Kazuhito Kawata, MD, PhD, received support from AbbVie. Hidekatsu Kuroda, MD, PhD, received lecture fees from Eisai, Chugai, and AstraZeneca. Kazuhiro Nouso, MD, PhD, received lecture fees from AbbVie, Aska Pharmaceutical, AstraZeneca, Bayer, Century Medical, Chugai, Covidien, Eisai, Gilead, Kowa, Lilly, and Otsuka and research funding from CureApp, Denka, Fujifilm, and Medtronic. Masatoshi Kudo, MD, PhD, received honoraria from Bayer, Chugai, Eisai, Eli Lilly, MSD, and Takeda and research funding from AbbVie, EA Pharma, Eisai, GE Healthcare, Gilead Sciences, Otsuka, Sumitomo Dainippon Pharma, Taiho, and Takeda. Kazuhiro Nouso and Masatoshi Kudo were members of the journal’s Editorial Board at the time of submission.

Funding Sources
The authors conducted this investigation in an independent manner. No funding specifically related to this study was received.

Author Contributions
Kazunari Tanaka, Kunihiko Tsuji, and Takashi Kumada conceived the study, and participated in its design and coordination. Kazunari Tanaka, Kunihiko Tsuji, Atsushi Hiraoka, Toshifumi Tada, Masashi Hirooka, Kazuya Kariyama, Joji Tani, Masanori Atsukawa, Koichi Takaguchi, Ei Itobayashi, Shinya Fukunishi, Toru Ishikawa, Kazuto Tajiri, Hironori Ochi, Hidenori Toyoda, Yuichi Koshiyama, Chikara Ogawa, Hiroki Nishikawa, Takashi Nishimura, Takeshi Hatanaka, Satoru Kakizaki, Hidenao Noritake, Kazuhito Kawata, Atsushi Naganuma, Hisashi Kosaka, Kosuke Matsui, Tomomitsu Matono, Hidekatsu Kuroda, Yutaka Yata, Hironori Tanaka, Tomoko Aoki, Hideyuki Tamai, Fujimasa Tada, Hideko Ohama, Yuki Kanayama, Kazuhiro Nouso, Asahiro Morishita, Akemi Tsutsui, Takuya Nagano, Norio Itokawa, Tomomi Okubo, Taeang Arai, Osamu Yoshida, Michitaka Imai, Shinichiro Nakamura, Hirayuki Enomoto, Masaki Kaibori, Masatoshi Kudo, Yoichi Hiasa, and Takashi Kumada performed data curation. Kazunari Tanaka and Kunihiko Tsuji performed statistical analyses and interpretation and drafted the text. All authors have read and approved the final version of the manuscript.