Assessment of Macrovascular Invasion in Advanced Hepatocellular Carcinoma: Clinical Implications and Treatment Outcomes with Systemic Therapy Open Access

Hepatocellular carcinoma (HCC) is the most common form of primary liver cancer, with an increasing incidence worldwide [1, 2]. Despite guidelines recommending surveillance for those at high risk, such as individuals with chronic hepatitis or cirrhosis [3‒7], many are still diagnosed at an advanced stage, with macrovascular invasion (MVI) or extrahepatic metastasis. Recent advancements in viral hepatitis treatments have shifted the primary causes of HCC from hepatitis B and C to lifestyle-induced fatty liver disease [8, 9]. This shift in etiology has resulted in an increasing number of surveillance targets owing to the vast number of patients affected by fatty liver, which makes early detection more challenging [10, 11]. Consequently, HCC continues to exhibit a high mortality rate as many patients are diagnosed at advanced stages, in which curative treatments are rarely available, and systemic therapy is the preferred treatment option.

HCC originates in the liver, a vital organ for human survival, and is associated with inflammation and/or fibrosis [1, 2, 12]. As HCC progresses, the cancer grows within the liver and develops intrahepatic and extrahepatic metastasis. A unique feature of HCC resides in its propensity to invade the three primary vascular systems of the liver: portal vein, hepatic vein, and bile duct [13, 14]. This invasion can severely impair liver function, resulting in a poor prognosis for patients, making vascular invasion a significant negative predictor of HCC prognosis [15].

Recent advances in systemic therapy for HCC have been remarkable, with many international phase III trials successfully integrating various therapeutic agents into clinical practice [16‒23]. The Response Evaluation Criteria in Solid Tumors (RECIST) serves as a standard for evaluating the efficacy of systemic therapy. It considers the tumor diameter in target lesions and the presence of new lesions through radiological assessments [24]. For HCC, these criteria are used to assess the efficacy of systemic therapy in the majority of clinical trials, including the aforementioned phase III trials, as well as real-world clinical cohort studies. In certain scenarios, a modified RECIST (mRECIST) is specifically employed for HCC for the exclusive assessment of tumor portions that exhibit arterial blood flow using dynamic computed tomography/magnetic resonance imaging, as changes in arterial blood flow within the liver are closely associated with therapeutic response [25]. However, the existing criteria primarily focus on tumor size when evaluating the effect of systemic therapy, making it challenging to assess MVI, which extends into the vasculature. To our knowledge, no existing guidelines explicitly evaluate MVI progression in HCC, and there is a paucity of studies evaluating the radiological impact of MVI on systemic therapy for HCC. Therefore, we conducted a multicenter study using retrospective data to establish clear criteria for determining the progression of MVI and how the progression of MVI diagnosed using these criteria affects the prognosis of patients with advanced HCC receiving systemic treatment.

We conducted a retrospective analysis of patients with advanced HCC with MVI treated with sorafenib, lenvatinib, or atezolizumab plus bevacizumab as first-line systemic therapy between June 2009 and December 2021. Clinical data for each of these three regimens were obtained from a database at Chiba University Hospital. Data were also collected from patients treated with atezolizumab plus bevacizumab for advanced HCC with MVI at Asahi General Hospital, Nippon Medical School Chiba Hokusoh Hospital, and Kimitsu Chuo Hospital during the same period. Details of the treatments, radiological assessment during treatments, description of the data collected in this study, and statistical analysis methods are described in Supplementary Methods. These data were integrated to create a dataset for analysis. This study was approved by the Research Ethics Committee of the Graduate School of Medicine, Chiba University (No. 3091) and received subsequent approvals from three other institutions. Given the nature of this study and in accordance with Japan’s Ethical Guidelines for Medical and Biological Research Involving Human Subjects, formal consent via written signature was not required. Instead, participating sites provided detailed study information to the subjects, giving them the option to opt out if they did not wish to participate (Statement of Ethics is also described below).

Sorafenib was administered orally at a dose of 400 mg (two 200-mg tablets) twice daily as per the established protocol. Lenvatinib was administered orally once daily, with a dose of 12 mg for patients weighing 60 kg or more at baseline and 8 mg for those weighing less than 60 kg. Although standard doses were generally used, the physician in charge had the discretion to initiate treatment with reduced doses in certain patients as necessary. During the treatment period, doses were adjusted based on patient condition, including adverse events. Atezolizumab plus bevacizumab was administered intravenously at a standard dose of 1,200 mg of atezolizumab plus 15 mg per kilogram of body weight of bevacizumab every 3 weeks. In the case of adverse events or other factors that complicated administration, treatment initiation was postponed. If an adverse event clearly related to bevacizumab occurred, atezolizumab was administered alone. Any of three treatments were continued until one of the following conditions was met: the physician deemed the drug clinically ineffective, an adverse event occurred that prevented further treatment, the patient’s general condition declined to a degree that made further treatment impossible, or the patient requested to discontinue treatment.

We ensured that all patients underwent a dynamic computed tomography or dynamic magnetic resonance imaging scan immediately before the start of treatment. The maximum radiological imaging interval allowed prior to treatment initiation was 4 weeks, and patients who did not meet this requirement were excluded. Radiological assessments to evaluate the effectiveness of drug treatment during the treatment period were conducted in accordance with the protocols of each center. The imaging intervals for the general population ranged from 4 to 8 weeks.

We retrospectively collected clinical data, including baseline demographics, such as age at the start of treatment, sex, etiology, Child-Pugh class, albumin-bilirubin (ALBI) grade, site and classification of vascular invasion, the presence of extrahepatic metastases, alpha-fetoprotein, and treatment regimen. Data on the date of death or last follow-up were collected to estimate cumulative survival. We separately recorded the date of determination of disease progression using Response Evaluation Criteria in Solid Tumors (RECIST v1.1) to assess tumor enlargement other than MVI and appearance of new lesions, as well as MVI progressive disease (MVI-PD) by radiological imaging evaluation.

The localization of MVI in patients with advanced HCC was determined according to the classification system proposed by the Liver Cancer Study Group of Japan [26]. This system categorizes portal vein tumor thrombus based on its location in the portal vein: distal to the secondary portal branches (Vp1), in the secondary portal branches (Vp2), in the primary portal branches (Vp3), or in the main portal trunk or bilateral portal branches (Vp4). Hepatic vein tumor thrombus may be classified into three categories based on the location of the tumor thrombus: in the peripheral hepatic veins (Vv1), in the major hepatic veins (Vv2), or in the inferior vena cava (Vv3). Our study focused on the assessment of portal and hepatic vein invasion, which is classified as MVI under the BCLC staging system [27]. However, the assessment of bile duct invasion was excluded from this study due to the inherent difficulties in imaging evaluation.

Radiological images captured during the treatment period for eligible cases were evaluated by two HCC treatment experts (M.I. and K.K.). Initially, they identified the target and non-target lesions based on the RECIST v1.1. They then evaluated the localization of MVI using the aforementioned classification. This study’s MVI evaluation protocol included specific radiologic imaging phases designed to track the progression of portal and hepatic vein tumor thrombosis. For portal vein tumor thrombosis, assessments were primarily performed during the portal phase, which provides the best contrast for portal blood flow. In contrast, hepatic vein tumor thrombosis was assessed in the delayed phase to improve contrast visualization of the hepatic vein and inferior vena cava. If vessel contrasts timing between cases, alternative phases were used to ensure an accurate assessment. Furthermore, to distinguish between viable MVI and thrombosis, arterial phase enhancement characteristics were used to identify MVI, while the portal venous phase or delayed phase was used to detect washout. PD on radiological images was determined by MVI progression, in addition to increased target lesions, exacerbation of non-target lesions, and appearance of new lesions. The evaluation of PD based on tumor enlargement apart from MVI and the appearance of new lesions was referred to as RECIST v1.1, whereas PD of MVI based on our criteria was designated as MVI-PD, and each was assessed separately. MVI-PD was defined as tumor progression into one or more additional branches and/or more than 10 mm in the longitudinal direction of the vessel compared to the baseline radiological image (Fig. 1).

Overall survival (OS) was defined as the date of death from the start of first-line treatment, and the censor date was the date of last observation. Progression-free survival (PFS) was defined as the time from the start date of first-line treatment to the date when progression was observed by radiological imaging or the date of death, with the censor date being the date of the last radiological imaging. Post-progression survival (PPS) was defined as the time from the date when progression was observed by radiological imaging to the date of death, with the censor date being the date of the last observation. The Kaplan-Meier method was used to estimate survival functions for OS and PFS, and the log-rank test was used for intergroup comparisons. Landmark analysis was used to analyze OS based on the presence or absence of MVI-PD and RECIST v1.1-PD up to 3 months after treatment initiation. OS from the start of treatment was estimated in patients alive for more than 3 months. The Cox proportional hazards model was used to calculate hazard ratios for risk factors associated with OS. Kendall’s τ was used to assess the association between OS and PFS/PPS, using values in the interval [−1, 1]. A value of 0 indicated that OS and other outcomes were independent, whereas a value of 1.00 indicated a perfect correlation. The Frank copula model or inverse probability of censoring weighting method, which accounts for censoring, was used to estimate τ. A p value <0.05 was considered statistically significant. All statistical analyses were performed using SPSS version 25 (IBM, Chicago, IL, USA) and R software (version 4.0.3; R Foundation for Statistical Computing, Vienna, Austria).

Using a database from four institutions, we identified 207 patients with advanced HCC and MVI who had received systemic therapy. Eighteen patients were excluded from the imaging study due to the timing of baseline imaging or a lack of follow-up imaging. As a result, 189 patients were included in the study and evaluated using the MVI-PD and RECIST v1.1-PD criteria. In addition, a landmark analysis was used to compare survival times between groups with different imaging assessments at the 3-month mark. Of the 189 patients with sufficient imaging data, 163 patients who were still alive at 3 months were included in the landmark analysis.

The demographic and baseline characteristics of the patients are summarized in Table 1. First-line treatment consisted of sorafenib, lenvatinib, or atezolizumab plus bevacizumab in 118 (62.4%), 43 (22.8%), and 28 (14.8%) patients, respectively. The proportions of portal and hepatic vein invasion based on the baseline radiological evaluation were 86.2% (163 patients) and 23.3% (44 patients), respectively. A detailed breakdown of the demographic and baseline characteristics specific to each first-line treatment regimen is shown in online supplementary Table S1 (for all online suppl. material, see https://doi.org/10.1159/000539380). Data lock was conducted at the end of November 2022, with a median observation period of 6.0 months (range: 0.4–109.1). The follow-up period for each treatment regimen was 5.7 months (range: 0.6–65.6) for sorafenib, 11.2 months (range: 2.0–54.1) for lenvatinib, and 12.2 months (range: 2.0–24.7) for atezolizumab plus bevacizumab. The median OS and median treatment duration of first-line systemic therapy for the study population were 7.5 months (95% confidence interval [CI], 6.0–12.3) and 2.6 months (95% CI, 2.3–3.0), respectively (Fig. 2a). When stratified by first-line treatment regimen, the median OS outcomes were 5.7 months (95% CI, 4.6–6.3) for sorafenib, 13.3 months (95% CI, 5.7–15.2) for lenvatinib, and 22.1 months (95% CI, 9.1 – not applicable) for atezolizumab combined with bevacizumab, with a statistically significant difference (p = 0.002) (Fig. 2b).

As shown in Figure 1, we evaluated MVI-PD and RECIST v1.1-PD during the entire duration of first-line systemic therapy by reviewing all radiological images. Of the 189 patients for whom imaging was available, 95 (50.3%) were diagnosed with MVI-PD during the entire first-line treatment period and 85 were confirmed to have MVI-PD within 3 months of starting treatment initiation. Similarly, of the 115 patients (60.8%) diagnosed with RECIST v1.1-PD during the same period, 85 were confirmed to have RECIST v1.1-PD within 3 months of beginning treatment. Median time to MVI-PD and RECIST v1.1-PD was 2.7 months (95% CI, 2.0–6.3) and 2.8 months (95% CI, 2.5–4.0), respectively (online suppl. Fig. S1). Simultaneous occurrence of MVI-PD and RECIST v1.1-PD was observed in 40 patients (21.2%) of the entire study population. During the treatment period, MVI-PD occurred prior in 51 patients (27.0%), whereas RECIST v1.1-PD occurred prior in 61 patients (32.3%). The incidence of MVI-PD during the treatment regimen was observed in 69 patients (58.5%) treated with sorafenib, 16 patients (37.2%) treated with lenvatinib, and 10 patients (35.7%) treated with the atezolizumab plus bevacizumab regimen, with a statistically significant difference (p = 0.014).

We also compared the post-treatment outcomes of patients with and without MVI-PD. Among those without MVI-PD, 30 patients (31.9%) received best supportive care (BSC), 26 received systemic therapy (27.7%), 5 received HAIC (5.3%), and 15 received TACE (16.0%). Conversely, in the cohort with MVI-PD, BSC was administered to 54 patients (56.8%), systemic therapy to 15 patients (15.8%), HAIC to 9 patients (9.5%), and TACE to 11 patients (11.6%). Patients with MVI-PD were more likely to undergo BSC following treatment, with a significantly smaller number transitioning to systemic therapy than their counterparts without MVI-PD.

The landmark time was established at 3 months from the initiation of first-line systemic therapy to determine the relationship between the incidence of MVI-PD and/or RECIST v1.1-PD within this period and OS. The analysis included 163 patients with a confirmed survival of at least 3 months. Figure 3 shows the Kaplan-Meier estimates for OS stratified by the presence or absence of MVI-PD and RECIST v1.1-PD within 3 months. Of the patients who did not experience both MVI-PD and RECIST v1.1-PD within 3 months, the longest OS of 19.7 months (95% CI, 15.0–25.7) was observed. Conversely, patients who experienced both MVI-PD and RECIST v1.1-PD within 3 months showed the poorest OS of 5.4 months (95% CI, 4.3–6.0). For patients with only RECIST v1.1-PD or MVI-PD or RECIST v1.1-PD occurring within 3 months, OS was 7.2 months (95% CI, 5.0–12.3) and 7.4 months (95% CI, 4.7–13.3), respectively (p < 0.001).

We conducted a multivariable analysis using a Cox proportional hazards model to determine the impact of MVI-PD and RECIST v1.1-PD within 3 months of initiating first-line systemic therapy on OS. Two distinct models were established, one considering the Child-Pugh class as a liver function factor and the other incorporating the ALBI grade (online suppl. Tables S2, S3). In both models, the occurrence of MVI-PD within 3 months (p < 0.001) and RECIST v1.1-PD within 3 months (p < 0.001) emerged as independent unfavorable prognostic factors.

The median PFS, which was calculated based on the occurrence of MVI-PD or death during first-line systemic therapy, was 2.7 months (95% CI, 2.1–5.4). The median PFS, considering RECIST v1.1-PD or death during first-line systemic therapy, was 2.8 months (95% CI, 2.5–3.7). When MVI-PD and RECIST v1.1-PD were combined as a composite endpoint, PFS was only 1.8 months (95% CI, 1.4–2.4). PPS was calculated from the date of MVI-PD, RECIST v1.1-PD, or combined PD (online suppl. Table S4). The correlations between the calculated PFS, PPS, and OS using the three different methods were analyzed (Fig. 4). With respect to PFS and OS, the correlation between PFS and OS based on MVI-PD (Kendall’s τ = 0.515; 95% CI, 0.434–0.595) and between PFS and OS using RECIST v1.1-PD (Kendall’s τ = 0.498; 95% CI, 0.417–0.579) was similar. Interestingly, when MVI-PD and RECIST v1.1-PD were combined as a composite endpoint, the correlation with OS did not show a significant improvement compared to PFS with MVI-PD or RECIST v1.1-PD alone (composite of MVI-PD and RECIST v1.1-PD: Kendall’s τ = 0.403; 95% CI, 0.315–0.490). Although the correlation between PPS and OS constantly demonstrated a stronger association than that between PFS and OS, the correlation between PPS and OS was based on MVI-PD (Kendall’s τ = 0.864; 95% CI, 0.825–0.903), RECIST v1.1-PD (Kendall’s τ = 0.821; 95% CI, 0.777–0.865), and composite PD (Kendall’s τ = 0.860; 95% CI, 0.829–0.891) were similar.

In this study, we developed a definition of MVI-PD to evaluate the progression of MVI during systemic therapy for advanced HCC with the aim of enhancing our understanding of MVI progression. MVI represents a unique form of HCC progression and is associated with the worst prognosis; however, RECIST v1.1 criteria, commonly used to evaluate the efficacy of systemic therapy in various cancers, lack specific guidelines for assessing MVI, which is a distinct feature of HCC progression [24]. Historically, surgery, such as hepatic resection and direct transarterial approaches, such as TACE and HAIC, were considered effective treatments for removing or controlling MVI, even before the advent of systemic therapy [28‒33]. Despite the establishment of systemic therapy for HCC, there are no reports demonstrating the efficacy of systemic therapy specifically targeting MVI in HCC patients.

In the present study, we used two approaches to assess the effect of MVI-PD on OS. First, we established a 3-month landmark time following treatment initiation and evaluated the effects of MVI-PD and RECIST v1.1-PD on OS within this timeframe. We demonstrated that MVI-PD within 3 months was a prognostic factor along with RECIST v1.1-PD within the same period. Landmark analysis is a survival analysis method used to address immortal time bias by considering the treatment response at a specific time [34, 35]. It can exclude subjects who were lost to follow-up or died before the landmark time; however, careful selection of this time is important to avoid bias. Setting a short time may misclassify responders, whereas a long time may limit the sample size and generalizability of the results. In our cohort, the majority had at least one radiological assessment within 3 months, and approximately 90% of MVI-PD occurrences throughout the entire treatment period were observed within this timeframe. Therefore, setting the landmark time at 3 months was considered reasonable.

Next, we performed PFS calculations using two distinct methods for detecting PD specifically defined in this study: RECIST v1.1-PD, which is disease progression that does not consider the presence or absence of MVI-PD, and MVI-PD, which does not consider the presence or absence of RECIST v1.1-PD. Analysis of PFS and OS using these two approaches revealed that the correlation between PFS calculated using RECIST v1.1-PD and OS was almost identical to that between PFS calculated using MVI-PD and OS. Furthermore, there was a comparable correlation between PPS and OS when calculated using both methods. Previous studies have established that PFS for systemic therapy in advanced HCC demonstrates a moderate correlation with OS, whereas PPS exhibits a stronger correlation with OS [36‒38], consistent with the correlation coefficients obtained in our study for PFS, PPS, and OS, which were comparable to those reported in previous studies. The finding that PFS, when considering only MVI progression for systemic therapy of advanced HCC, had a similar impact to that of PFS evaluated using the RECIST v1.1 criteria suggests that controlling MVI is an important factor in the treatment of advanced HCC.

We defined MVI-PD as the progression of the tumor plug into one or more branches on the central side of the vasculature and/or an increased tumor plug >10 mm in the longitudinal direction of the vasculature compared with the baseline radiological image. The definition used in our study was originally developed because, to our knowledge, there have been no reports showing the progression of MVI itself during systemic therapy for advanced HCC. We postulate that MVI-PD, as defined here, can be adequately evaluated using a 5-mm-thick dynamic CT, which is commonly imaged at general institutions worldwide. The RECIST v1.1 criteria assess the effectiveness of systemic therapy based on changes in tumor diameter and the emergence of new lesions. However, a significant challenge has been the inability to evaluate tumors that have invaded the portal and hepatic veins, which are common in HCC. A previous study by Reig et al. [36] focused on progression patterns and outcomes after progression of systemic therapy for advanced HCC, emphasizing the poor prognosis for patients with new extrahepatic lesions, including new-onset MVI. However, they did not investigate the progression of preexisting MVI at the time of systemic therapy initiation, which differs from our methodological approach. Based on our findings, establishing criteria for evaluating MVI as a non-target lesion is necessary to predict an MVI-specific treatment response in patients with advanced HCC with MVI receiving systemic therapy for “overall response” in the RECIST v1.1 criteria.

This study has several limitations. First, its retrospective study and small sample size limit the generalizability of the proposed MVI-PD determination method, emphasizing the need for validation in larger, multicenter, prospective cohorts. Second, this study did not address MVI regression, a rare but clinically significant event during drug therapy, emphasizing the difficulty of developing objective criteria for its evaluation. Regardless of the difficulty in assessing tumor thrombosis reduction, recognizing any non-worsening condition has significant clinical value, especially in the management of liver function deterioration caused by tumor progression. Third, comparing the efficacy of various treatment regimens for MVI is difficult due to differences in sample sizes, observation periods, and patient demographics between groups. For example, differences in OS and MVI control rates observed between sorafenib, lenvatinib, and atezolizumab plus bevacizumab may reflect biases of the retrospective cohort, such as patient background and subsequent treatments. Therefore, caution should be exercised when interpreting these findings, given the limitations of the study in adjusting for these biases.

In conclusion, MVI progression during systemic therapy for advanced HCC with MVI appears to be strongly associated with poor OS. However, objective evaluation of MVI progression, which is a distinct progression pattern in HCC, using RECIST v1.1 criteria is challenging. Therefore, it is necessary to establish a method for assessing the efficacy of MVI to accurately evaluate the effectiveness of systemic therapies for HCC with MVI.

The authors are grateful to Satomi Nakamura, Yuka Iwase, and Ryoko Arai for their contributions to data management.

The study protocol conformed to the ethical guidelines of the 2013 Declaration of Helsinki, the 2018 edition of the Declaration of Istanbul, and was approved by the Research Ethics Committee of the Graduate School of Medicine, Chiba University (No. 3091), and by three other institutions (Asahi General Hospital, Nippon Medical School Chibahokusoh Hospital, and Kimitsu Chuo Hospital). Formal consent by written signature was not required for this type of study based on the Ethical Guidelines for Medical and Biological Research Involving Human Subjects in Japan.

Sadahisa Ogasawara received honoraria from Bayer, Leverkusen, Germany; Eisai, Tokyo, Japan; Eli Lilly, Indianapolis, IN, USA; Chugai Pharma, Tokyo, Japan; AstraZeneca, Cambridge, UK; Merck & Co., Inc., Kenilworth, NJ, USA; consulting or advisory fees from Bayer, Eisai, Merck & Co., Inc., Chugai Pharma, Eli Lilly, and AstraZeneca; and research grants from Bayer, AstraZeneca, and Eisai. Kengo Nagashima received consulting and advisory fees from SENJU Pharmaceutical Co., Ltd., Toray Industries, Inc., and Kowa Company, Ltd. Masanori Atsukawa has received a research grant from Eisai. Naoya Kato received honoraria from Bayer, Eisai, Sumitomo Dainippon Pharma, Tokyo, Japan, Merck & Co., Inc.; consulting or advisory fees from Bayer and Eisai; and research grants from Bayer and Eisai. The other authors have no conflicts of interest to declare.

This study was not supported by any sponsor or funder.

Masanori Inoue, Kazufumi Kobayashi, and Sadahisa Ogasawara contributed substantially to the study conceptualization. Masanori Inoue, Sadahisa Ogasawara, Tomomi Okubo, Norio Itokawa, Masamichi Obu, Kentaro Fujimoto, Hidemi Unozawa, Sae Yumita, Kisako Fujiwara, Miyuki Nakagawa, Hiroaki Kanzaki, Keisuke Koroki, Kazufumi Kobayashi, Soichiro Kiyono, Masato Nakamura, Naoya Kanogawa, Takayuki Kondo, Shingo Nakamoto, Kengo Nagashima, Ei Itobayashi, Masanori Atsukawa, Yoshihiro Koma, and Ryosaku Azemoto contributed substantially to data acquisition, analysis, and interpretation. Masanori Inoue and Sadahisa Ogasawara contributed to manuscript drafting. Naoya Kato supervised this study. All the authors critically reviewed and revised the draft manuscript and approved the final version for submission. Sadahisa Ogasawara and Kazufumi Kobayashi shared co-first authorship.

All data generated or analyzed during this study are included in this article and its online supplementary material files. Further enquiries can be directed to the corresponding author.