Hepatocellular carcinoma with macrovascular invasion: Review and survival meta-analysis of initial local therapy using minimal prognostic criteria

INTRODUCTION

Macrovascular invasion (MaVI) is a key prognostic driver defining advanced hepatocellular carcinoma (HCC) within the Barcelona Clinic Liver Cancer (BCLC) staging system.^1,2 Since the first BCLC version, the BCLC-C stage included patients with preserved liver function, mild tumor-related symptoms (ECOG-PS 0 to 2), and extrahepatic spread, defined by the presence of MaVI and/or metastases to other organs.

Over the past 2 decades, the introduction of systemic therapies for advanced HCC—such as sorafenib³ and lenvatinib⁴—and, more recently, immunotherapy-based regimens^5–8 has improved overall survival (OS) in this population and has intensified the discussion about outcomes across different levels of MaVI involvement.⁹ Indeed, the REFLECT⁴ and HIMALAYA⁶ trials excluded patients with adverse prognostic features—for example, REFLECT⁴ excluded those with HCC occupying ≥50% of liver volume, clear bile duct invasion, or main portal vein invasion (Vp4), while HIMALAYA⁶ excluded patients with main portal vein thrombosis. Other common exclusion criteria across pivotal studies included impaired liver function (Child–Pugh class B or C) and poor performance status (ECOG ≥2). While pivotal trials and real-world data have validated the survival benefits of immunotherapy, evidence supporting other treatment options—including resection, transplantation, locoregional therapies, and radiotherapy—remains highly heterogeneous. This heterogeneity underlies the discrepancies observed across guidelines from scientific associations.

Indeed, whereas western clinical practice guidelines (CPGs) for HCC recommend systemic therapy for patients with advanced disease—or for those who are not candidates for other therapies but retain preserved liver function and performance status^10–13—guidelines from China, Japan, Korea, and, in highly selected cases, Italy, support the use of surgery or locoregional therapies as first-line options in this population.^14–17

These discrepancies highlight the need for a deeper analysis of the reported survival data in this population, with particular attention to the limitations and potential insights of the published evidence. Recognizing this gap, we designed a systematic review and meta-analysis to assess survival at 1, 2, 3, and 5 years in HCC patients with MaVI who received at least one locoregional therapy or surgery as initial treatment and no systemic therapy.

METHODS

RESULTS

DISCUSSION

This systematic review and meta-analysis show that evidence on OS with non-systemic treatments for HCC with MaVI is insufficient to deliver firm recommendations to provide clinical benefit as compared with systemic therapy. Across modalities, the only endpoint with consistently complete and comparable data was 1-year OS; longer-term estimates were undermined by missing data. Most included studies were from Asia—often surgical series—and even there, only 1-year outcomes were free of missing data. Given that untreated MaVI carries a median OS of ~10–12 months,^19–22 1-year survival remains the most informative benchmark.

Within this context, TARE was the only modality with reproducible results, showing low heterogeneity (I ²=0%; Q=3.19; p=0.527; τ ²=0.000) and a pooled 1-year OS of 34% (95% CI, 20–48%). All other approaches displayed substantial heterogeneity (I ²>80%), consistent with treatment-selection effects rather than true differences in efficacy.

Two randomized trials have not demonstrated TARE superiority over sorafenib.^23,24 With immunotherapy now standard, head-to-head trials of TARE versus immunotherapy—stratified by MaVI extent and baseline prognostic factors—are urgently needed, furthermore considering the technical and dosimetry advances in recent years as was reported by Garin et al.²⁵ At present, evidence specific to MaVI under immunotherapy derives primarily from subgroup or exploratory analyses of pivotal trials (eg, HIMALAYA, IMbrave150)^26–28 and from observational cohorts.

Among recent observational data, 2 Japanese cohorts of patients treated with atezolizumab–bevacizumab are informative.^29,30 In a JLCA/JSHBPS-based analysis, Matono et al.²⁹ reported OS of 86.2%, 65.8%, 44.1%, and 35.0% at 6, 12, 18, and 24 months in patients with MaVI defined as Vp2–3, Vv2, or B2–3; rates were 73.3%, 43.6%, 18.7%, and 18.7% in those with major vascular invasion (Vp4, Vv3, B4). Komatsu et al.³⁰ found 1- and 2-year OS of 62.7% and 49.3%, with strong modulation by liver function (2-year OS 78.6% for mALBI 1/2a vs. 20.0% for mALBI 2b/3; p=0.0041). These data underscore liver function as a key determinant of outcome and help explain the rising rates of missing data for long-term OS in the present meta-analysis (0% at 1 y; 17.3% at 2 y; 32.7% at 3 y; 70.2% at 5 y).

Although the 12-month TARE estimate in the present meta-analysis appears lower than the OS rates reported by Matono et al.²⁹ and Komatsu et al.,³⁰ differences in cohort composition, treatment modality, and outcome definitions—combined with wide confidence intervals—preclude valid indirect comparisons. At most, our findings suggest a lower TARE point estimate for 1 year that should be interpreted with caution.

Selection bias also complicates comparisons between surgery and systemic therapy. Two large observational series illustrate this problem.^30,31 Famularo et al.³¹ reported marked baseline imbalances—portal vein thrombosis grade (Vp1–Vp3), cirrhosis prevalence, and ECOG performance status—between patients treated with resection versus sorafenib; inverse-probability-of-treatment weighting (IPTW) was required to improve balance and approximate comparable cohorts. Likewise, Komatsu et al.³⁰ analyzed patients after excluding extrahepatic metastases and restricting the cohort to MaVI with multinodular disease. Despite a more favorable baseline profile in the resection group, median OS was 14.8 months after hepatectomy versus 31.9 months with first-line systemic therapy (lenvatinib; atezolizumab–bevacizumab; durvalumab–tremelimumab). Taken together, these data suggest that apparent survival advantages with surgery largely reflect treatment selection and tend to dissipate when disease extent and liver function are adequately accounted for, even after propensity-score methods (PSM/IPTW).

In the absence of randomized comparative evidence, uncertainty should be explicitly acknowledged, and adjusted estimates should not be overinterpreted. Retrospective analyses remain intrinsically vulnerable to residual and unmeasured confounding—even after propensity-score matching or IPTW—because treatment allocation follows center-specific pathways: resection is prioritized for operable candidates, whereas systemic therapy is selected when tumor location, vascular-invasion pattern, liver function, comorbidities, or patient preference preclude surgery. These nonrandom processes violate exchangeability and can persist despite sophisticated adjustment, limiting the causal interpretability of between-modality comparisons and justifying the design of an individual patient data meta-analysis. This is precisely the type of scenario for which the 2025 BCLC update³² recommends applying the CUSE framework (Complexity, Uncertainty, Subjectivity, Emotion) to make uncertainty explicit and structure shared, evidence-anchored decisions.

Methodologically, we excluded 425 studies (85%) at the identification stage for lacking minimal baseline data on ECOG, performance status, and liver function, yet residual confounding remains likely. Some cohorts may not have been fully treatment-naïve, others do not report subsequent therapy, and missing OS at later time points may reflect unregistered deaths or incomplete reporting. Both scenarios highlight that current evidence for non-systemic strategies is constrained by data completeness and selection. These limitations are explicitly recognized in the BCLC 2025, which included the CUSE (Complexity, Uncertainty, Subjective and Emotions) framework.^32,33

The CUSE framework links evidence to clinical reality. It prioritizes treatments aligned with patient goals and context, accounts for patient factors and availability, and acknowledges emotions without letting them drive decisions. Consequently, until reliable biomarkers and randomized, stratified head-to-head trials are available, treatment selection should be guided by objective liver function and disease anatomy within well-defined MaVI strata. Non-systemic approaches should be considered cautiously, given the risk of not matching the survival achievable with immunotherapy.

Our study has some limitations. As previously discussed, the results of this aggregate-data meta-analysis do not allow for definitive conclusions regarding the efficacy of different locoregional or surgical treatments in patients with MaVI. The substantial heterogeneity observed across most treatment approaches, except for TARE, likely reflects differences in treatment selection based on patient-specific factors.

CONCLUSIONS

Among patients with advanced HCC and MaVI, 1-year OS remains the most reliable endpoint for cross-modality comparisons. Transarterial radioembolization shows consistent 1-year results but no proven benefit over sorafenib in randomized trials, and superiority or non-inferiority to contemporary immunotherapy-based regimens remains unproven. Pending stratified head-to-head comparisons versus immunotherapy-based regimens and validated predictive markers, use of TARE should be individualized to objective liver function and vascular anatomy within predefined MaVI strata, ideally within clinical trials or multidisciplinary care pathways, and after considering systemic therapy as the standard of care.

Supplementary Material