Treatment Decision-Making in Unresectable Hepatocellular Carcinoma: Importance of Understanding the Different Response Patterns between IO plus Anti-VEGF and IO plus IO Regimens

Two immuno-oncology (IO)-based combination regimens for hepatocellular carcinoma (HCC) treatment exist: IO plus anti-vascular endothelial growth factor (VEGF) and IO plus IO. The IO plus anti-VEGF regimen involves atezolizumab plus bevacizumab (Atezo+Bev) [1, 2], which is currently approved worldwide. Sintilimab plus bevacizumab biosimilar (Sin+Bev) [3] and camrelizumab plus rivoceranib (Cam+Ribo) [4] are also approved in China. The IO plus IO regimen, on the other hand, involves durvalumab plus tremelimumab (Dur+Tre) [5], which is currently approved globally, and nivolumab plus ipilimumab (Nivo+Ipi), which is expected to be approved soon based on results from the CheckMate 9DW trial [6]. This editorial discusses the differences in overall survival (OS) Kaplan-Meier (KM) curves and response patterns between these two regimens.

In brief, the IO plus anti-VEGF regimen exhibits a “sustained anti-tumor effect with continued OS benefit from early treatment initiation.” Conversely, the IO plus IO regimen is characterized by an “all-or-nothing response pattern, where responders demonstrate a quick, deep, and durable response, resulting in a long-tail OS benefit.”

Figure 1 illustrates OS KM simulation curves for lenvatinib (LEN) monotherapy from the LEAP-002 trial [7]and LEN (85%) or sorafenib (SOR, 15%) monotherapy from CheckMate9DW [6], as well as LEN plus pembrolizumab (LEN+PEM) from the LEAP-002 trial [7], Cam+Rivo [4], Sin+Bev [3], Atezo+Bev [2], Nivo monotherapy from the CheckMate 459 trial [8], Dur monotherapy [5], Nivo+Ipi [6], and Dur+Tre [5] combination immunotherapy [9]. The IO plus Anti-VEGF regimens and tyrosine kinase inhibitor (TKI) monotherapies demonstrated fewer early deaths within 6–12 months compared to the IO plus IO regimens and IO monotherapies (Fig. 1).

Atezolizumab plus bevacizumab provides a clear OS benefit compared to Nivo or Dur monotherapies, as well as Dur+Tre or Nivo+Ipi, particularly during the first 12 months of treatment (early OS benefit) [9] (Fig. 1). This favorable early OS KM curve phenomenon is also observed with LEN monotherapy and LEN (85%)/SOR (15%) monotherapy (Fig. 1). One plausible explanation for this early OS benefit is the pronounced anti-tumor effect of molecular targeted agents containing anti-VEGF activity such as LEN, Rivo, and Bev [10‒14], evident from the initiation of treatment. These agents’ direct anti-tumor effects across a broad patient population contribute to the early OS benefits seen with LEN monotherapy and LEN (85%)/SOR (15%) monotherapy. Furthermore, in combination with IO, anti-VEGF/TKIs enhance the immune microenvironment, yielding a favorable long-tail OS effect (Fig. 1a).

VEGF promotes immunosuppressive cells, such as regulatory T cells (Tregs), myeloid-derived suppressor cells, and tumor-associated macrophages, while inhibiting dendritic cell maturation and CD8⁺ T-cell infiltration into tumors. Anti-VEGF/TKI activity improves the immune microenvironment by counteracting these mechanisms. Consequently, the IO plus Anti-VEGF/TKI regimen is more effective in inflamed tumors, even within an immunosuppressive microenvironment. By contrast, LEN monotherapy or LEN (85%)/SOR (15%) monotherapy yields early OS benefits but lacks the tail plateau effect observed in IO plus anti-VEGF/TKI regimens (Fig. 1).

In contrast, the IO plus IO regimen, which lacks anti-VEGF activity, demonstrates limited anti-tumor effects in some patients, potentially increasing early mortality within 6–12 months (Fig. 1b). Among IO plus IO regimens, Nivo+Ipi shows superior OS beyond 1 year compared to IO plus Anti-VEGF/TKI regimens, likely due to a higher overall response rate (ORR 36%) [6], contributing to the long-tail plateau effect in OS KM curves [6].

The IO plus Anti-VEGF/TKI regimen exerts a direct anti-tumor effect early, enhancing the cancer-immunity cycle by maturing dendritic cells, infiltrating activated CD8⁺ T cells into the tumor, and inhibiting immunosuppressive cells. This mechanism is effective across a broader patient population, including active and exhausted immune subclasses.

Conversely, the Dur+Tre Phase I/II study (Study 22) [15] suggests that IO plus IO’s mechanism of action involves a single peak of activated CD4⁺ and CD8⁺ T cells in the blood on day 15 after anti-CTLA-4 antibody administration, leading to a quick and deep response in certain patients. The HIMALAYA study [16] also noted a single peak of immune-mediated adverse events (imAE) following a peak of activated CD8⁺ T cells [17]. This suggests the early appearance of imAEs from the widespread distribution of activated CD8⁺ T cells to all the organs in addition to the tumor tissue, rather than sustained memory T-cell effects.

The response pattern of IO plus IO regimens features quick, deep, and durable responses. This drastic response is determined by the following three factors: (1) the amount of CD8⁺ T cells released into the systemic circulation, (2) the immune subclass being a “hot or cold tumor subtype,” and (3) Treg depletion via antibody-dependent cellular cytotoxicity activity (ADCC) by macrophages or natural killer (NK) cells [18‒20] (Fig. 2). However, it is worth noting that the progressive disease (PD) rate is high (40% in Dur+Tre), making IO plus IO an “all-or-nothing” regimen. Despite this, anti-CTLA-4 antibodies in responders provide a “quick, deep, and long-lasting” response, providing a long-term survival shared by Dur+Tre and Nivo+Ipi regimens.

Exploratory analysis of the IMbrave150 trial [1] demonstrates a clear relationship between tumor depth of response (−60% to −100% reduction, −30% to −60% reduction, 0% to −30% reduction, 0% to +20% increase, and >20% increase) and OS, with well-stratified OS KM curves [21]. Similarly, Five-year OS by tumor response data from the HIMALAYA trial [16] showed a clear relationship between depth of response and OS, with a deep response (−50% to −100% tumor reduction) showing a 5-year survival rate of extremely high (62.4%), while stable disease (SD) patients with tumor size increases (0% to +20%) had OS KM curves similar to those with PD. This highlights a significant OS KM curve difference between IO plus anti-VEGF and IO plus IO regimens.

The IO plus IO regimen’s efficacy is associated with the tumor immune microenvironments, such as high CD8⁺ T cell/Treg ratios in tumor tissues and the amount of activated CD8⁺ T cells released into the circulation [15]. In HCC, anti-CTLA-4 antibodies may deplete Tregs via Fcγ receptor-dependent ADCC activity [18‒20] (Fig. 2). While the augmentation mechanism of binding affinity between the Fcγ region of Anti-CTLA-4 and Fcγ receptor of macrophage/NK cells remains unclear, providing this could enable predictive biomarker development for IO plus IO response. For now, however, early identification of responders to the IO plus IO regimen is essential to guide clinical treatment decision-making: continuation of the IO plus IO regimen (biomarker responders) or switching to the other regimen since image-based response is much more delayed than biomarker response (non-biomarker responders).

Real-world timing for switching to second-line treatment after IO plus IO regimen should differ from the phase 3 clinical trial protocols. With IO plus IO regimens, delayed response assessments may risk early mortality, as shown in Figure 1. Biomarkers such as AFP, PIVKA-II, and AFP-L3 fractions, effective for systemic therapy response monitoring, provide earlier clinical decision-making than imaging. For Dur+Tre, responders and non-responders can be distinguished within 2–3 weeks post-treatment initiation, and AFP/PIVKA-II responses precede imaging-based evaluations by 6–12 weeks [22‒24]. Timely switching to anti-VEGF-based regimens can mitigate early mortality.

When selecting first-line and second-line regimens for unresectable HCC, particularly for BCLC-C and BCLC-B patients with high tumor burden, it is crucial to first evaluate whether or not combination immunotherapy is suitable. Subsequently, assessing the patient's ability to tolerate grade 3 or 4 imAEs associated with anti-CTLA-4 antibody is essential. Patients tolerable for grade 3 or 4 imAEs typically exhibit good performance status (PS), good liver function, relatively younger age, absence of severe or active comorbidities such as cardiovascular disease, and lack of Vp3 or Vp4 vascular invasion. If these criteria are not met, the recommended first-line treatment is an IO plus anti-VEGF regimen (Fig. 3).

On the other hand, if the patient is speculated to be tolerable for grade 3/4 imAE and has a relatively very high tumor burden or extensive extrahepatic spread affecting multiple organs, an IO plus IO regimen is recommended as the first-line regimen (Fig. 3). In cases of locally advanced HCC (with or without vascular invasion) with no or limitted extrahepatic spread, an IO plus anti-VEGF regimen is preferred due to its broad benefits and lower toxicity [25]. Furthermore, IO plus anti-VEGF therapy has the potential to achieve deep responses leading to curative conversion after tumor shrinkage, making it an optimal choice for locally advanced HCC [26].

According to previous reports, using Durva plus Treme as a second-line treatment following Atezo plus Bev results in inferior response rates, progression-free survival, and OS compared to its use as a first-line treatment [27]. In that sense, even for locally advanced HCC, selecting an IO plus IO regimen as the first-line regimen can be viable, aiming for CR through systemic therapy alone. On the other hand, IO plus IO may still be effective after Atezo plus Bev as Treg depletion cases can achieve sufficient response even in the post-Atezo plus Bev. Therefore, preserving IO plus IO as a treatment option after Atezo plus Bev is also a reasonable treatment approach.

Sequential therapy involving IO plus IO followed by IO plus anti-VEGF, or vice versa, is feasible since their mechanisms of action differ significantly. The only difference is that Atezo plus Bev has late responders, necessitating continuation of treatment until imaging PD before transitioning to the next line of therapy (Fig. 3). In contrast, IO plus IO requires early switching to anti-VEGF containig regimens for non-biomarker responders (Fig. 3).

Additionally, according to ASCO, ESMO, EASL, and NCCN guidelines [28‒31], tyrosine kinase inhibitors (TKIs), especially lenvatinib, are recommended as second-line treatments after combination immunotherapy. Lenvatinib’s strong anti-VEGF and anti-FGF effects make it a viable second-line option. Considering patient status (such as age, PS, comorbidities, liver function and tumor burden) and each regimen’s different modes of action, it is theoretically possible to select “the most appropriate regimen at the most appropriate timing for the most suitable patient.”

The IO plus Anti-VEGF regimen provides “continuous tumor response with early OS benefit,” characterized by sustained responses during the entire treatment course. In contrast, the IO plus IO regimen features an “all-or-nothing response pattern with drastic responses in responders.” IO plus IO regimen use should capitalize on its quick, deep, and durable response (drastic response) and long-tail effect advantages in responders. Using AFP, PIVKA-II, or AFP-L3, a biomarker-based responder selection [9] is crucial for early treatment decision-making to rescue early mortality with IO plus IO regimen. Switching to Atezo+Bev or LEN within 3–4 weeks after initiation of the IO plus IO regimen can prevent early mortality among non-responders, emphasizing the importance of timely therapeutic clinical decision-making.

No statement is needed because this study was based exclusively on published data.

Lectures: Chugai, Eisai, Eli Lilly, Takeda, AstraZeneca. Grants: Otsuka, Taiho, Chugai, GE Healthcare, Eisai, AbbVie. Advisory Consulting: Chugai, Chugai Roche, Eisai, AstraZeneca. Masatoshi Kudo is the Editor-in-Chief of Liver Cancer.

There was no funding for this editorial.

Masatoshi Kudo conceived, wrote, and approved the final manuscript.

Data availability is not applicable because this is not a research article.