Immune checkpoints are receptors or ligands expressed on T cells, tumor cells, and antigen-presenting cells that maintain immune tolerance by inhibiting T-cell activity [23] ; however, their overexpression on tumor cells often leads to immune evasion [24] . ICB therapy employs immune checkpoint inhibitors (ICIs) to disrupt these receptor-ligand interactions and restore antitumor immunity. The programmed death-1/programmed death-ligand 1 (PD-1/PD-L1) axis is the most frequently targeted checkpoint, and monoclonal antibodies against PD-1 or PD-L1—such as nivolumab, pembrolizumab, and atezolizumab—have been approved for treatment of HCC. Nonetheless, objective response rates to PD-1/PD-L1 monotherapy remain modest [25] , and combination regimens have been explored to enhance efficacy.

Currently, the combination of atezolizumab and bevacizumab (a VEGF inhibitor) serves as the first-line systemic therapy for advanced HCC, as demonstrated in the IMbrave150 trial [26] . IMbrave050 was the first positive Phase III trial for postoperative adjuvant therapy in HCC: 668 high-risk patients post-ablation (RFA or MWA) or resection were randomized (334 per arm) to receive adjuvant atezolizumab plus bevacizumab or active surveillance. At 12 months, recurrence-free survival (RFS) was 78% in the combination arm versus 65% in the surveillance arm, corresponding to a 28% reduction in the risk of recurrence or death (hazard ratio (HR) 0.72, adjusted 95% confidence interval (CI) 0.53 - 0.98; P = 0.012). This study first confirmed the regimen’s efficacy as postoperative adjuvant therapy in HCC, advancing immunotherapy into earlier settings. However, treatment-related adverse events occurred in 98% of combination-treated patients (vs. 62% in controls), including immune-related toxicities in 63% (vs. 18%), indicating that optimizing administration strategies may help balance efficacy and safety [27] . Wen et al. evaluated 48 patients with recurrent HCC and found that RFA plus toripalimab yielded a lower tumor progression rate than RFA alone (45% vs. 80%) and higher 18-month RFS (48.7% vs. 18.8%) [28] . Wang et al.’s retrospective analysis of 127 recurrent HCC patients demonstrated that RFA combined with PD-1 inhibitors (camrelizumab or sintilimab) significantly prolonged 1-year RFS (32.5% vs. 10.0%), median RFS (38.6 vs. 16.7 weeks, P = 0.001) and median overall survival (OS) (50.9 vs. 47.3 weeks, P = 0.016) [29] . Zhou et al. conducted a multicenter, randomized Phase I/II trial. In the first stage, 48 patients with previously treated, unresectable HCC were randomized to RFA or MWA plus toripalimab versus toripalimab alone. The combination arm achieved a higher objective response rate (ORR) than monotherapy, and initiating toripalimab on day 3 post-ablation (Schedule D3) yielded better responses than starting on day 14 (Schedule D14). In the second stage, 98 eligible patients were randomized to Schedule D3 versus toripalimab alone. Schedule D3 significantly improved median progression-free survival (PFS) (7.1 vs. 3.8 months; HR = 0.57; P < 0.001) and median OS (18.4 vs. 13.2 months; HR = 0.58; P = 0.005) compared with monotherapy [30] . Lyu et al. reported that combining anti-PD-1 therapy (nivolumab or pembrolizumab) with RFA or MWA markedly increased ORR and extended both PFS and OS in advanced HCC, with manageable toxicity [31] . These studies indicate that combining ICB therapy with RFA exerts a more favorable antitumor effect in HCC than monotherapy. Moreover, several large Phase III trials are underway to validate adjuvant PD-1/PD-L1 inhibition post-RFA in HCC, including CheckMate 9DX (NCT03383458; nivolumab) [32] , KEYNOTE-937 (NCT03867084; pembrolizumab) [33] , and EMERALD-2 (NCT03847428; durvalumab ± bevacizumab) [34] .

Another key target is cytotoxic T-lymphocyte-associated protein 4 (CTLA-4), which inhibits T-cell activation by competitively binding CD80/CD86 [35] . Its inhibitors, such as ipilimumab and tremelimumab, are clinically approved. Sangro et al. were the first to conduct relevant research. They treated 21 HCC patients with chronic HCV using tremelimumab, successfully inducing antiviral immune responses and reducing viral loads [36] . Duffy et al. further combined tremeli-mumab with RFA in 32 BCLC B/C HCC patients; on day 36 ± 96 hours, they performed subtotal RFA or cryoablation, with transarterial chemoembolization (TACE) for BCLC B. They observed partial responses in 5 patients, significant HCV viral load reductions in 12, and increased intratumoral CD8⁺ TILs in responders [37] . Agdashian et al. used a similar design in 39 refractory advanced HCC patients, administering tremelimumab followed by subtotal RFA/ cryoablation/TACE on day 35; median OS was 10.9 month, and peripheral activated T-cell subsets rose markedly (CD4⁺HLA-DR⁺ + 65%, CD8⁺HLA-DR⁺ + 106%, CD4⁺ICOS⁺ + 444%, CD8⁺ICOS⁺ + 528%; all P < 0.0001), while CD3⁺ tumor-infiltrating lymphocytes increased from 0.78% to 2.45% (P = 0.012) [38] . These data highlight CTLA-4 blockade’s immunomodulatory potential, but its later development and higher systemic toxicity have limited its clinical use compared with PD-1/PD-L1 inhibitors [39] , and large Phase III results are still pending.

Cell therapy involves ex vivo activation and expansion of immune cells, which are then reinfused to precisely target tumor cells. While cell therapy has achieved remarkable success in hematologic malignancies, its application in solid tumors, including HCC, has expanded in recent years, and its combination with RFA has drawn considerable interest.

DC-based therapy is a major focus in this context. Kitahara et al. enrolled 30 HCV-related HCC patients who had undergone RFA. 14 received infusions of DCs activated with the immunostimulant OK-432 (OK432-DC), while 16 received unmodified “baseline” DCs as controls. The OK432-DC group exhibited significantly longer RFS (24.8 vs. 13.0 months; P = 0.003), although OS did not differ between groups (67.8 vs. 73.4 months; P = 0.780). Among 17 HLA-A24⁺ patients, 35.3% mounted an IFN-γ response to at least one TAA peptide, indicating that DC infusion aids in the generation of tumor-associated antigen-specific cytotoxic T lymphocytes [40] . Building on ex vivo DC activation, DC vaccines can be further loaded with TAAs to generate specificity. Peng et al. performed whole-exome sequencing on tumor specimens from 10 HCC patients treated by resection or RFA, synthesized personalized neoantigen peptides, and manufactured a personalized DC vaccine plus neoantigen-activated T cells. Patients received 18 cycles of DC vaccine + adoptive T-cell therapy, yielding encouraging results: 70% developed novel neoantigen-specific T-cell responses, and 71.4% of responders remained recurrence-free at 2 years [41] . Conversely, Lee et al. reported that DC vaccines loaded with TAAs reduced recurrence in non-RFA patients but paradoxically increased recurrence in RFA-treated patients—possibly due to more rapid DC exhaustion in the RFA setting. The underlying mechanisms remain to be elucidated [42] .

Another notable cell therapy is cytokine-induced killer (CIK) cell therapy, in which peripheral blood mononuclear cells are stimulated ex vivo with cytokines to generate highly cytotoxic effectors bearing both T-cell and NK-cell characteristics, then reinfused to target tumor cells. Lee et al. conducted a Phase III multicenter randomized trial in 230 postoperative HCC patients (after RFA, resection, or percutaneous ethanol injection), randomizing 115 to CIK and 115 to control. The CIK arm achieved significantly longer median RFS (44.0 vs. 30.0 months; HR 0.63; 95% CI 0.43 - 0.94; P = 0.010) and prolonged OS (HR 0.21; 95% CI 0.06 - 0.75; P = 0.008). Treatment-emergent adverse events were predominantly grade 1 - 2, and no treatment-related deaths occurred in the CIK group [43] . Yoon et al. compared 59 Phase I/II HCC patients (post-resection or RFA), finding that median RFS was not reached in the CIK arm versus 29.8 months in controls (HR 0.42; P = 0.006). Safety was favorable, with no ≥ grade 3 AEs reported [44] . Ji et al. evaluated CIK combined with RFA + TACE in HCC patients. The RFA + TACE + CIK group achieved a mean OS of 42.1 ± 5.6 months versus 37.8 ± 4.8 months in the RFA + TACE only arm; 5-year OS rates were 29.3% versus 13.8% (log-rank P = 0.045) [45] .

Emerging cell-based therapies such as T cell receptor-engineered T cells and chimeric antigen receptor T cells have shown preliminary efficacy in HCC immunotherapy, but none have been investigated in combination with RFA [46] - [48] .

The core mechanism of tumor vaccines is to deliver TAAs to elicit tumor-specific T-cell immunity. Vaccine platforms include protein/peptide, nucleic acid, cellular, viral-vector vaccines, etc. In HCC, the glypican-3 (GPC3) peptide vaccine is among the most well-studied: as a TAA highly expressed in HCC, GPC3 peptides can induce HCC-specific immune responses [49] . Taniguchi et al. treated 35 postoperative HCC patients with the GPC3 peptide vaccine and observed that the 1-year recurrence rate was 25.7% versus 42.4% in controls; 5-year OS was 70.6% versus 57.6%; and 8-year OS was 67.1% versus 38.9%. Among GPC3 IHC-positive patients, 60.0% mounted a specific CTL response versus 16.7% of IHC-negative patients [50] . In a Phase I/II trial, Sawada et al. showed that adjuvant GPC3 vaccination after resection or RFA reduced the 1-year recurrence rate in GPC3-positive patients from 48% in controls to 24% in the vaccine arm (P = 0.047), with robust induction of tumor-specific immunity [51] [52] . Other vaccine approaches are also in development. Löffler et al. conducted a multicenter Phase I/II study in 22 HLA-matched, early- to intermediate-stage HCC patients (including six post-RFA). After low-dose cyclophosphamide preconditioning, patients received subcutaneous IMA970A, a 17-peptide HCC TAA vaccine, combined with the CV8102 RNA adjuvant. Among 19 evaluable patients, this regimen induced HLA-I peptide responses in 37%, HLA-II responses in 53%, and vaccine-specific T-cell responses in 68.4% [53] . DC-based vaccines also constitute a form of tumor vaccination; their progress has been detailed above under cell-based therapies.

The greatest limitation of RFA is incomplete RFA (iRFA), a major driver of increased local recurrence. This stems from RFA’s reliance on tissue conductivity: as surrounding tissue heats and chars, its conductivity drops sharply, constraining the ablation zone, a phenomenon known as the roll-off effect [54] . When tumors abut large intrahepatic vessels, convective heat loss further restricts the ablation margin, a phenomenon known as the heat-sink effect [55] . Together, these effects limit the efficacy of RFA for larger lesions (typically >3 - 5 cm). iRFA not only fails to eradicate the tumor but may also enhance the aggressiveness of residual disease, leading to tumor progression [56] - [58] . Even more concerning, iRFA induces persistent local inflammation and accumulation of immunosuppressive cells, amplifying post-RFA immunosuppressive factors, creating a suppressive immune microenvironment and hindering subsequent immunotherapy. For example, Zhang et al. showed that although iRFA transiently activated antitumor immunity, CD4⁺ and CD8⁺ T-cell and DC infiltrates precipitously declined thereafter, while immunosuppressive tumor-associated macrophages rose to 1.92  ±  0.50-fold and 4.12  ±  0.64-fold of baseline at days 3 and 8 post-iRFA, respectively [59] . Similarly, Shi et al. found that Tregs, MDSCs, and M2 macrophages continued to accumulate after iRFA, suppressing T-cell function and abrogating PD-1 blockade—thereby establishing a vicious cycle in which incomplete ablation leads to immunosuppression, which in turn drives accelerated tumor progression [19] .

To enlarge the effective ablation zone and achieve complete necrosis, strategies beyond conventional saline infusion (which temporarily enhances conductivity) have been explored. Mocan et al. developed PEG-coated gold nanoparticles that are injected directly into the ablation zone via the electrode; their superior conductivity and thermal diffusivity substantially mitigate the roll-off effect compared with saline infusion [60] . Combining TACE with RFA has also been investigated to counteract heat-sink effects near large vessels. Zhang et al. compared TACE + RFA versus surgical resection for small recurrent HCC and found that, although OS was similar, the TACE + RFA arm had a significantly lower complication rate [61] . A meta-analysis further demonstrated that TACE + RFA yields superior recurrence-free and OS compared with RFA alone [62] . With the advent of various new technologies, the roll-off and heat-sink effects of RFA are expected to be mitigated.

Another major challenge is treatment-related toxicity, especially immune-related adverse events (irAEs) triggered by ICIs. Common irAEs include hepatotoxicity, rash, pneumonitis, colitis, etc. In the IMbrave050 study of high-risk HCC patients after curative resection or ablation, any-grade adverse events occurred in 98% of patients treated with atezolizumab plus bevacizumab versus 62% under active surveillance; grade 3 - 4 events occurred in 41% versus 13% and grade 5 events in 2% versus < 1%. 88% of events were deemed treatment-related, including proteinuria in 46%, hypertension in 38%, thrombocytopenia in 20%, elevated AST in 16%, ALT in 14% and hypothyroidism in 14%—all substantially higher than in the control arm [26] . Because many HCC patients have underlying liver disease such as cirrhosis, their hepatic reserve is limited and hepatotoxic irAEs (for example, immune-mediated hepatitis) are particularly dangerous. A meta-analysis of thirty real-world studies encompassing 3867 HCC patients treated with atezolizumab plus bevacizumab found AST and ALT elevations in 35 percent and 20 percent of patients respectively—rates exceeding those observed in IMbrave050 [63] . This heightened liver toxicity may reflect checkpoint blockade disrupting intrahepatic immune-tolerance pathways, leading to excessive T-cell activation and acute inflammatory damage [64] .

The ASCO guideline classifies irAEs into four grades by severity. Grade 1 (mild) generally does not require treatment interruption. Grade 2 (moderate) warrants holding immunotherapy and initiating low-dose corticosteroids. Grade 3 (severe) should prompt permanent discontinuation of the ICI, administration of systemic high-dose steroids and a slow taper once symptoms improve. Grade 4 (life-threatening) necessitates permanent ICI cessation and consideration of additional immunosuppressive agents alongside steroids [65] .

Aside from RFA, other ablation methods, such as MWA, cryoablation, HIFU and IRE, also elicit antitumor immune responses.

MWA uses electromagnetic energy to heat and destroy tumor cells; like RFA, it relies on percutaneous electrodes and is subject to both roll-off and heat-sink effects. Clinical trials have shown no significant differences between MWA and RFA in local efficacy, complication rates, or long-term survival [66] . By thermally destroying tumor cells and releasing TAAs and DAMPs, MWA can activating antitumor immunity [67] . MWA combined with immunotherapy in HCC has also emerged as a major focus of current clinical research [68] [69] .

Cryoablation employs cycles of rapid freezing and thawing to form intracellular ice crystals, disrupting cell membranes and inducing apoptosis or necrosis. Its low-temperature mechanism better preserves intact antigen structures and can more effectively provoke antitumor immunity compared with RFA or MWA [70] [71] . Despite this promise, the complexity and cost of cryoablation equipment have limited its adoption in HCC.

HIFU is a noninvasive technique that focuses ultrasonic energy to generate localized thermal and mechanical damage, releasing TAAs and DAMPs to trigger immune activation [72] . However, obstruction of ultrasound transmission to the liver by the rib cage prevents HIFU’s effective use in HCC [73] ; it is thus not included in HCC treatment guidelines, and its main current applications are in uterine fibroids and adenomyosis.

IRE is a newer, nonthermal ablative method that delivers brief high-voltage pulses to create permanent nanopores in cell membranes, inducing programmed cell death without heat [74] . IRE avoids heat-sink effects and can effectively ablate tumors adjacent to major vessels. Preclinical studies demonstrate that IRE elicits even stronger immune activation than RFA, and may surpass cryoablation in this regard [71] [75] . Nonetheless, IRE’s application in HCC remains exploratory, and its technical complexity and cost have so far limited widespread clinical adoption.

Future research should further explore the clinical application of cryoablation, IRE, and HIFU in HCC, elucidating their immunologic mechanisms and working to reduce cost and procedural complexity. It should also compare RFA with these alternative ablation methods in combination with immunotherapy, assessing both immunogenicity and clinical efficacy. Patient-specific factors, such as tumor size, location, and liver function, must guide the choice of the optimal ablation-immunotherapy regimen. Such comparative, personalized studies will be essential to tailor multimodal strategies for HCC.

The spatiotemporal dynamics of RFA-induced immune responses remain poorly defined, and the optimal sequencing of immunotherapy with RFA requires further investigation. To date, most studies have evaluated immunotherapy as an adjuvant post-RFA [27] - [31] [40] - [45] [50] - [53] , although some have explored its use in a neoadjuvant setting, administering RFA during the course of immunotherapy [37] [38] . In clinical practice, immunotherapy for HCC must be tailored and precise, based on disease stage. For intermediate- or advanced-stage patients who are not candidates for curative resection, neoadjuvant immunotherapy can be used to downsize tumors before RFA, thereby reducing overall tumor burden. Conversely, in early-stage patients with a solitary lesion, upfront RFA followed by adjuvant immunotherapy may eradicate microscopic residual disease and lower the risk of recurrence or metastasis.

Moreover, to date no single biomarker has been endorsed by major guidelines for accurately selecting HCC patients most likely to benefit from immunotherapy. Traditional tissue biomarkers, such as PD-L1 expression, tumor mutational burden (TMB), and microsatellite instability-high (MSI-H), have limited use in HCC due to inconsistent predictive value or low prevalence [76] - [78] . In contrast, liquid biopsy approaches exemplified by circulating tumor DNA (ctDNA) are garnering attention for their noninvasive nature and ability to provide dynamic monitoring [79] , while routine serum markers like alpha-fetoprotein (AFP) and the albumin-bilirubin (ALBI) score offer important prognostic information despite limited specificity [80] . Concurrently, studies of the gut microbiome and resistance-related pathways such as Wnt/β-catenin are yielding mechanistic insights into differential treatment responses [81] [82] . It is foreseeable that integrated, multidimensional biomarker models will enable a robust predictive framework for immunotherapy response in HCC and thus facilitate personalized treatment optimization. Progress toward this goal is underway, but further foundational research and clinical validation remain essential.

Emerging therapies, such as tumor vaccines and cell therapy, combined with RFA have shown early promise in reducing recurrence rates. However, these trials have generally been small and nonrandomized, limiting their ability to inform clinical practice. Moreover, the high cost and technical complexity of these advanced immunotherapies restrict their application. For example, Peng et al.’s personalized DC vaccine approach involves whole-exome sequencing and custom synthesis of neoantigen peptides, which under current technology can only be evaluated in small cohorts [40] . Future Large-scale, randomized controlled trials are needed to define the ideal combination strategies, identify the patient subgroups most likely to benefit, and establish long-term efficacy and safety.