Liver transplantation for hepatocellular carcinoma: from patient selection and downstaging to risk stratification and post-transplant surveillance.

Introduction
According to the 2022 GLOBOCAN report, liver cancer is the sixth most commonly diagnosed cancer and the third leading cause of cancer-related death worldwide, with an estimated 865,000 new cases and 750,000 deaths in 2022. Hepatocellular carcinoma (HCC) accounts for approximately 75%–85% of these cases, highlighting its major global health burden [1]. Among available treatment options, liver transplantation (LT) remains the most effective curative therapy for eligible patients. However, LT is typically reserved for those with early-stage HCC, leaving a substantial proportion of patients ineligible [2–4].
The demand for LT is rising and exceeding the available organ supply [5]. Currently, HCC represents one of the fastest-growing indications for LT, accounting for at least 22% of LT in the U.S. [6]. The underlying etiology of HCC has also shifted in recent years. Whereas hepatitis C virus infection used to be the predominant cause, its incidence has markedly declined with the widespread use of direct-acting antivirals [7]. In contrast, metabolic dysfunction–associated steatotic liver disease (MASLD) has emerged as a major driver of HCC and is expected to become the leading indication for LT in the near future [6, 8, 9]. Importantly, patients with MASLD-related HCC often experience poorer graft and post-LT survival outcomes [10].
This evolving epidemiology has important implications for recipient selection and liver allocation. Patients with MASLD-related HCC frequently have advanced age, obesity, diabetes, and cardiovascular comorbidities, which may influence transplant candidacy, perioperative risk assessment, and post-transplant outcomes. As MASLD-related HCC continues to increase, transplant programs and allocation systems may face growing challenges in balancing oncologic benefit, medical complexity, and equitable organ distribution. This further emphasizes the need for ongoing studies to improve patient selection and management strategies for these cases.
Given these changing trends, a comprehensive understanding of patient selection criteria, downstaging strategies, and post-LT surveillance is necessary. This review aims to summarize key considerations in LT for HCC, from identifying appropriate candidates and applying evidence-based downstaging approaches to optimizing post-LT monitoring for recurrence risk.

Liver transplantation as a curative approach for hepatocellular carcinoma
Patients with HCC are eligible for curative treatments, such as local ablative therapies including radiofrequency ablation (RFA), microwave ablation (MWA), or surgical resection, or LT if presenting with early-stage cancers [2, 3, 11].
For the tumor with BCLC stage 0 or A and well-compensated liver diseases, surgery is considered the treatment of choice [2, 3]. Patients with non-cirrhotic liver diseases or well-compensated cirrhosis and no significant portal hypertension are excellent candidates for curative treatment with surgical resection [12, 13]. The issue is that only about 10% of HCC cases occur without cirrhosis [14]. There is also a high risk of cancer recurrence after resection, with recurrence rates of about 50%–70% within 5 years, most commonly during the first year after surgery [2]. If surgery is contraindicated, early-stage HCC cases can be recommended to receive local ablative treatment. However, there is also a high risk of recurrence after ablative therapy, and the reported recurrence-free survival (RFS) for ablation in tumors ≤3 cm is about 46% [15–17].
For unresectable early-stage HCC, LT is usually recommended, as it not only removes the tumor but also eliminates the cirrhotic liver and the underlying disease that led to HCC [2, 3]. LT is the only treatment that can potentially cure both the tumor and the underlying liver disease. The recurrence rate after LT is much lower than after resection or ablation, at about 10%–15% [18, 19]. Unfortunately, LT can’t be recommended for all HCC cases. On one hand, there is a shortage of donor organs. If tumors are beyond the early stage, not meeting LT criteria, the risk of recurrence is higher, making it a potential waste of a donor liver. Therefore, researchers have tried to determine which patients, based on morphological/biological characteristics, are suitable for LT, aiming to achieve RFS and overall survival (OS) rates.

Selection criteria for liver transplantation
Milan criteria, introduced in 1996, was the first LT selection criteria proposed that patients with a single tumor ≤5 cm or up to three tumors each ≤3 cm, without macrovascular invasion or metastasis, could be considered suitable candidates for LT [20]. These criteria were associated with low recurrence rates and became the standard rule for LT in HCC cases. Only 4% of patients in the original study experienced HCC recurrence, and the reported 4-year survival rate was 75%, similar to that of LT recipients without HCC [20–22].
The main limitation of the Milan criteria was that they were restrictive, excluding many HCC patients from LT eligibility, and only considered morphological characteristics of HCC. To address this, other researchers tested broader criteria. In 2001, the University of California, San Francisco (UCSF) proposed expanded criteria that included one tumor ≤6 cm or up to three tumors each ≤4.5 cm, with a total tumor diameter ≤8 cm [23]. These criteria were associated with 1- and 5-year OS rates of 90% and 75.2%, respectively, and a 2-year recurrence rate of 11.4% in the original study. With the UCSF criteria, more patients could be considered for LT while achieving outcomes similar to those under the Milan criteria. Several criteria have been developed to expand LT eligibility while maintaining acceptable outcomes (Table 1). The ‘Up-to-seven’ criteria, sum of tumor number and diameter ≤7, showed 71.2% OS and 9.1% recurrence [24]. Newer criteria, including Kyoto, French AFP, Hangzhou, Metroticket, and New York/California (NYCA), incorporate biological markers, such as des-gamma-carboxyprothrombin (DCP) and alpha-fetoprotein (AFP) in addition to tumor burden, as detailed in Table 1 and discussed in later sections.
It should be noted that the allocation systems and donor availability shape selection practice and regional and policy differences are important when interpreting and applying transplant criteria. Many Western programs typically emphasize standardized deceased-donor liver transplantation (DDLT) pathways, formal down-staging protocols like Milan criteria, fixed observation intervals, and explicit AFP thresholds for granting transplant priority. By contrast, many centers in East Asia, where living-donor liver transplantation (LDLT) is more common, have adopted expanded, center-based criteria that combine morphologic features with tumor biology [33–35]. Examples include the Kyoto criteria (≤10 nodules, each ≤5 cm, and DCP/protein induced by vitamin K absence/antagonist-II (PIVKA-II) ≤ 400 mAU/mL) [25] and the Japanese ‘5–5-500’ rule (≤5 nodules, largest diameter ≤5 cm, and AFP ≤500 ng/mL) [36, 37]. Both of which have allowed safe extension of LT access in experienced LDLT programs. Direct comparison between Western DDLT-centered policies and Asian LDLT-driven practice requires attention to these structural differences in donor source, selection bias, and allocation systems.

Bridging therapy for hepatocellular carcinoma patients awaiting liver transplantation
In 2001, the Model for End-Stage Liver Disease (MELD) was introduced by Kamath et al. to predict survival in patients with end-stage liver disease [38]. In 2002, the MELD score was adopted in the United States as the primary metric for liver allocation. LT candidates with HCC who meet Milan criteria are eligible for MELD exception points to account for tumor progression risk not reflected by liver function. HCC patients must remain on the waiting list with their laboratory MELD score for at least 6 months before receiving a standardized exception score.
Considering the mandatory 6-month waiting period for LT, different therapies are used to prevent tumor progression and reduce waitlist dropout, a strategy known as bridging therapy. Various locoregional therapies (LRTs) are applied for this purpose, including transarterial chemoembolization (TACE), transarterial radioembolization (TARE), thermal ablation, and external beam radiation therapy (EBRT) [39]. There is currently no firm recommendation favoring one treatment over another. While TACE remains the most common approach, thermal ablation and TARE have become more frequently utilized, especially in patients with larger tumors, higher AFP levels, or better liver function. Importantly, initiating treatment with TARE or ablation was linked to a lower likelihood of dropping off the transplant waiting list compared with starting with TACE, highlighting the potential benefit of these modalities in bridging patients to LT [40].
Tumor progression during bridging treatment while on the waiting list has been associated with poorer post-LT outcomes. Therefore, this mandatory waiting period also provides an opportunity for observation and monitoring to better assess post-LT prognosis and determine the most appropriate LT candidate [41]. Response to LRT based on modified Response Evaluation Criteria in Solid Tumors (mRECIST) criteria, as well as tumor size (>3 vs. ≤3 cm), has been shown to help predict a patient’s risk of recurrence after LT [42]. Additionally, in a large study including 747 patients without LRT and 2,854 patients who received LRT while on the LT waiting list, RFS and post-LT recurrence rates were comparable between the two groups. However, a higher number of LRT sessions and an unfavorable AFP trend during the waiting period were significant predictors of post-LT recurrence, whereas the specific LRT modality did not influence recurrence risk. An important factor associated with improved post-LT outcomes was achieving a complete pathologic response (cPR) [43]. Therefore, identifying pre-LT factors associated with achieving cPR may help predict post-LT outcomes. In an analysis of 501 HCC patients who received LRTs before LT, cPR was associated with lower MELD scores, lower pre-LT AFP levels, single tumor, smaller total tumor diameters, prior ablation, tumors within Milan or UCSF criteria, and a favorable LRT response [44].
A longer interval from LRTs to LT has been associated with achieving cPR [44]. However, waiting time can be a double-edged sword: while adequate observation allows assessment of tumor biology, excessively short or excessively long waits may negatively impact post-LT outcomes. An analysis of 911 HCC patients from three U.S. centers demonstrated that very short (<6 months) or very long (>18 months) wait times from diagnosis to LT were associated with a 1.6-fold higher risk of post-LT HCC recurrence [45]. Therefore, optimizing the timing of LRT and LT for each individual on the waitlist is essential.

Downstaging: expanding the recipient population

Concept and rationale
The concept of downstaging has been introduced to reduce the tumor burden and biological activity through various treatment strategies. If a patient is successfully downstaged to within accepted LT criteria, they may then become eligible for transplantation. High-quality evidence shows that LT after successful downstaging improves both RFS and OS [46]. Therefore, patients should be considered for LT after successful downstaging with complete or partial radiologic response. Even cases with complete radiologic response may harbor microscopic metastases, which can lead to post-transplant recurrence. It is crucial to recognize that the interval between the last LRT and LT should be optimized, as extremes may worsen post-transplant outcomes. Priority should require stability over a predefined observation interval (commonly 3–6 months) with at least two consecutive imaging studies. Concordant biomarker decline, notably AFP falling from >1000 to <500 ng/mL and remaining low for ≥3 months, supports true tumor control, whereas rising or persistently high AFP argues against listing. Transplant decisions should integrate AFP dynamics and durability of imaging response using a ‘wait and see’ policy, absence of radiologic/clinical high-risk features, and treatment modality [46–48]. We will discuss them in detail in the next sections, but two key questions arise in this context: first, which patients should be considered for downstaging, and, if so, what is the limit for downstaging; and second, which treatment modalities are most effective for achieving it?

Patient selection for downstaging
Based on the United Network for Organ Sharing (UNOS) downstaging policy, candidates can be considered for downstaging if their tumor burden meets one of the following criteria: (i) a single tumor larger than 5 cm but ≤8 cm; (ii) two or three tumors, with at least one >3 cm, none >5 cm, and a total tumor diameter ≤8 cm; or (iii) four or five tumors, each <3 cm, with a total diameter ≤8 cm.
After successful downstaging, meeting predefined LT criteria, such as Milan, with no major vascular invasion, no extrahepatic spread, and no lymph node involvement, the patient may be listed for LT. Under current UNOS policy, patients with AFP > 1000 ng/mL must achieve AFP < 500 ng/mL before LT. In addition, patients must remain within Milan criteria for 6 months before receiving MELD exception points, and there should be at least 3 months of radiologic stability between successful downstaging and LT.
Several studies have shown that patients who do not initially meet the Milan criteria but fall within the UNOS downstaging criteria can be successfully downstaged and undergo LT, their outcomes are comparable to those who met the Milan criteria at baseline [49, 50].
In 2015, a study by UCSF group compared 118 patients who met the UNOS downstaging criteria with 488 patients who met the Milan criteria at listing. Among those downstaged to within Milan criteria, 54.2% underwent LT, and 5 patients (7.5%) experienced recurrence. The 5-year OS and RFS rates were 77.8% and 90.8%, respectively, while the 5-year intention-to-treat survival was 56.1%. None of these outcomes differed significantly from those observed in patients within the Milan criteria at baseline. However, dropout rates were significantly higher in the downstaging group at both 1 year (24.1% vs 20.3%) and 2 years (34.2% vs 25.6%) [28].
In 2017, the Organ Procurement and Transplantation Network (OPTN) and UNOS officially established these parameters as the UNOS downstaging protocol. According to this policy, patients who are successfully downstaged to meet the Milan criteria qualify for automatic MELD exception points following a required 6-month observation period [51].
After implementation of these criteria by UNOS, a large multicenter retrospective study in 2020 compared outcomes between patients always within the Milan criteria and those meeting the UNOS downstaging criteria (UNOS-DS). Three-year post-LT survival was 83.2% for Milan and 79.1% for UNOS-DS, with recurrence rates of 6.9% and 12.8%, respectively. Higher AFP levels (≥100 ng/mL) and shorter waiting times were linked to poorer outcomes. These results validated the UNOS-DS criteria, showing comparable outcomes to Milan cases and underscoring the importance of AFP level and waiting time in improving post-LT results [49]. Moreover, a multicenter randomized controlled trial in 2020 from Italy (XXL study) showed that patients who underwent LT after successful downstaging had significantly higher OS rates compared to those who did not receive LT [46].
In 2021, in the first study of Multicenter Evaluation of Reduction in Tumor Size before Liver Transplantation (MERITS-LT), across seven centers and four UNOS regions (2016–2019), 209 HCC patients undergoing downstaging per UNOS-DS criteria were evaluated to determine downstaging success rates, dropout, and post-LT outcomes. Successful downstaging to Milan criteria was achieved in 87.7% of cases, while 37.3% dropped out from the transplant list within 2 years. Pre-treatment lectin-reactive AFP ≥10% was linked to higher dropout risk. Two-year post-LT survival reached 95%, and HCC recurrence occurred in 7.9% of patients [52].
Taken together, these studies support the validity of the UNOS-DS framework as a pragmatic method to expand transplant access without consistently compromising post-LT outcomes. Across cohorts, two recurring determinants of poorer post-transplant results were elevated pre-treatment AFP and shorter waitlist duration, highlighting the prognostic value of biochemical response and observation time. However, most evidence is retrospective and heterogeneous in patient selection, downstaging modalities, and follow-up, which limits generalizability. Therefore, while UNOS-DS provides an evidence-based pathway for downstaging, its application should be accompanied by rigorous patient selection, AFP monitoring, and standardized radiologic response criteria, and interpreted in the context of center-level capabilities and allocation policies. However, the question remains: what happens if a patient initially falls outside UNOS-DS criteria (‘all-comers’ [AC]), is successfully downstaged, and undergoes LT? Several studies have addressed this question from different perspectives.
In a UCSF study in 2019, evaluating downstaging in 74 AC HCC patients, successful conversion to Milan criteria was achieved in 64.8% of cases. Dropout from the transplant list was high, reaching 53.5% at 1 year and 80% at 3 years. Only 13.5% of AC patients ultimately received a LT. Intention-to-treat survival was 77.4% at 1 year and 21.1% at 5 years. Among the 10 AC patients who underwent LT, three experienced HCC recurrence [53].
Similarly, a 2020 retrospective analysis of the UNOS database evaluated 121 AC cases. Vascular invasion was present on the explant in 23.7% of these patients. Three-year post-LT survival was 71.4%, and HCC recurrence occurred in 16.7%, both notably worse outcomes observed in patients initially within Milan or UNOS-DS criteria [49]. More recently, a retrospective study of the UNOS database analyzed 668 AC patients listed for LT between 2010 and 2019, comparing their outcomes with patients always within Milan and those meeting UNOS-DS criteria. The 2-year cumulative dropout rate for AC-DS patients was 30%, and those with AFP >100 ng/mL faced an even higher risk, with a 2-year dropout probability of 52% [54].
The data show that patients with a very high tumor burden (AC group) face a greater risk of waitlist dropout even after downstaging and have a lower likelihood of favorable post-LT outcomes compared with those meeting Milan or UNOS-DS criteria. Previous studies were mostly retrospective, with small sample sizes and limited follow-up. In a more recent 2025 prospective study spanning nine LT centers across five UNOS regions, 326 patients within UNOS-DS criteria and 190 AC patients were analyzed. In the AC group, the success of downstaging declined with increasing tumor burden: 72% for cases with the sum of number plus largest lesion <10, 51% for sum 10–12, and 39% for sum >12. Two years after the initial LRTs, 66% of AC patients were successfully downstaged compared with 82% of UNOS-DS patients. The AC group experienced a higher 5-year post-LT recurrence rate (30% vs 14%), though rates of undergoing LT were similar between groups (40% vs 48%; P = 0.10), as was 5-year post-LT survival (72% vs 74%; P = 0.77) [55]. These results indicate that LT after downstaging in AC patients is achievable, but requires careful, individualized evaluation [2]. Importantly, the likelihood of successful downstaging decreases with greater tumor burden and remains strongly influenced by AFP level. These data suggest that AC patients require individualized risk stratification rather than routine listing after downstaging. Practical selection parameters that emerge from recent multicenter data include lower tumor-burden thresholds (favoring sum <10), durable radiographic response, and sustained AFP decline before granting transplant priority. For AC cases with very high burden or poor biochemical response, conservative management or enrollment in prospective protocols is recommended.
Current evidence supports the use of UNOS-DS criteria with attention to AFP and observation time, while patients outside these criteria may still be considered individually if tumor burden and treatment response are favorable. This aligns with the AASLD guideline [2]. The EASL guideline also recommends that HCC cases beyond LT criteria at each center may be considered for LT only if successfully downstaged to within Milan criteria; otherwise, they are considered poor LT candidates. It does not specify detailed criteria or limits for downstaging [3]. This guideline also notes that AFP > 1000 ng/mL is an absolute contraindication to LT, whereas AFP > 400 ng/mL is a relative contraindication. While both baseline AFP and AFP dynamics are important, EASL emphasizes that the single most prognostic AFP value is the last level measured before LT. It also highlights that successful downstaging with durable radiologic response for ≥3 months after the last treatment, together with sustained AFP improvement, identifies patients eligible for LT with favorable post-transplant outcomes. In response to whether there should be an upper limit for downstaging, the International Liver Cancer Association (ILCA) and the International Liver Transplantation Society (ILTS) note that while carefully selected patients beyond local LT criteria may achieve acceptable post-LT outcomes, a subset is at higher risk of waitlist dropout and poor post-LT survival, particularly those with impaired liver function or AFP > 1000 ng/mL [56]. Therefore, they recommend caution when expanding eligibility to these patients. They also emphasize the importance of an observation period of at least 3 months after successful downstaging to confirm durable biological response before transplantation.

Effective treatment strategies for downstaging

Downstaging using locoregional therapies
AASLD [2], EASL [3], and ILTS-ILCA consensus [56] currently recommend that different forms of LRTs may be used for downstaging/bridging, with no evidence favoring one modality over another. Treatment selection is generally guided by tumor number, size, location, and the center’s experience with each specific LRT. While surgery is not recommended as a downstaging tool, patients who previously underwent surgery, experienced recurrence, and whose tumors later on meet the Milan criteria may be considered eligible for LT [2, 3]. A case report also showed that surgery together with subsequent TARE may also be beneficial in selected cases with macrovascular invasion and potentially increase eligibility for LT. TACE, TARE, RFA, and MWA are currently the most commonly used LRTs for downstaging [57]. Table 2 summarizes the outcomes of downstaging using LRTs for cases with different tumor burdens, including all-comers, UNOS-DS, and Milan criteria, which will be reviewed here in more detail.
A 2015 study of 86 patients compared outcomes after TACE (n = 42) versus Y-90 TARE (n = 44). Despite higher tumor burden in the TARE group, median OS was similar (16.4 months for TARE vs. 18 months for TACE). However, TACE was associated with more adverse events and longer hospitalization, suggesting TARE may be better tolerated with shorter hospital stays [60]. Similarly, a 2016 meta-analysis of five studies including 553 patients with unresectable HCC (284 TACE, 269 TARE) found comparable 4-year survival between the two groups. Response rates, both partial and complete, were also similar. TACE, however, was associated with longer hospital stays and more post-procedure pain, while TARE was mostly outpatient [61]. Comparisons of TARE and TACE in the context of bridging and downstaging have shown similar outcomes. A 2023 meta-analysis included 14 studies (seven downstaging, three bridging, and four mixed). 55 cases received TARE from 204 cases undergoing downstaging and subsequently underwent LT. One of the included studies was a randomized controlled trial (RCT), which reported a higher tumor response with TARE (9/32) compared with TACE (4/34) [62].
Several studies have reported comparable outcomes for TARE and TACE in both UNOS-DS and AC patients, with some even suggesting better response with TARE for downstaging or bridging purposes [63, 64]. One study focusing on UNOS-DS patients found that the choice between TACE and TARE as the initial downstaging approach did not significantly impact treatment response, time or probability of achieving successful downstaging, likelihood of waitlist dropout, or LT [52]. But even a better trend in 3-year survival was observed in the TARE group (92.9% vs. 75.7%; P = 0.052). Microvascular invasion was also less frequent with TARE (3.6% vs. 27%), and patients in the TARE group required fewer treatment sessions compared to TACE [64]. TARE may also be a more suitable option for patients with portal vein tumor thrombosis [65, 66] and even in AC cases demonstrated higher downstaging success rates [55].
Taken together, available evidence indicates that both TACE and TARE are effective for downstaging HCC, with largely comparable rates of tumor response, successful downstaging, and post-LT outcomes across UNOS-DS and AC patients. TARE may offer practical advantages, including fewer treatment sessions, shorter hospital stays, lower peri-procedural morbidity, and potentially improved response in patients with portal vein involvement or higher tumor burden. Nonetheless, most studies are retrospective, with heterogeneous patient populations and varying LRT protocols, limiting the ability to definitively recommend one modality over another. Treatment selection should therefore be individualized based on tumor characteristics, vascular involvement, center expertise, and patient tolerance, while awaiting prospective trials to establish optimal strategies.

Downstaging using systemic therapies
Systemic therapies, including anti-angiogenic agents (such as multikinase inhibitors and monoclonal antiangiogenic antibodies) and immune checkpoint inhibitors (ICIs, including programmed cell death protein 1 [PD-1] inhibitors like nivolumab and pembrolizumab, programmed death-ligand 1 [PD-L1] inhibitors, such as atezolizumab and durvalumab, and cytotoxic T-lymphocyte-associated protein 4 [CTLA-4] inhibitors, like ipilimumab and tremelimumab), have been explored for potential use in downstaging or bridging to LT. However, current evidence is limited, and these approaches are still considered investigational. Existing guidelines do not recommend systemic therapies for downstaging or bridging outside of clinical trials, but they also do not advise excluding patients who have received such treatments [2, 3].
While data on using tyrosine kinase inhibitors (TKIs) in pre-LT settings are limited, several studies have reported their use, either alone or in combination with LRTs, for downstaging or bridging purposes. Most of these reports are small case series or case reports, primarily describing experiences with sorafenib [67–73] and lenvatinib [74–76]. Additionally, a phase III, randomized, double-blind trial tested whether adding sorafenib to TACE improves tumor control in HCC patients within Milan criteria on the transplant waiting list. Fifty patients were randomized to sorafenib (n = 24) or placebo (n = 26). Tumor progression occurred in 14 patients (7 per group), and response rates, progression-free survival, time to transplant, and adverse events were similar between groups [77]. Another trial assessed neoadjuvant sorafenib in 12 HCC patients within UCSF criteria before LT [78]. Most patients needed dose adjustments and treatment breaks, indicating limited tolerability of sorafenib in this context. In a retrospective 2022 study of 327 patients with HCC, 62 (19%) received sorafenib therapy [79]. Half of these cases involved tumor progression following LRT, while the remaining half could not undergo further LRT. Overall, 58% were removed from the transplant list due to disease progression, whereas 42% ultimately underwent LT. Those who received sorafenib because further LRT was not feasible showed reasonable 5-year overall and post-LT survival, but patients treated after LRT failure due to tumor progression had higher dropout rates and poorer overall outcomes. Finally, in a cohort of 128 HCC patients who underwent LT, two groups were compared: those treated with TACE alone and those who received TACE plus sorafenib. The combination group included a higher proportion of patients beyond Milan criteria. Disease-free survival was 67.2% in the TACE-only group and 100% in the combination group (borderline significance, P = 0.07). Five-year OS was 61.5% vs. 77.8% (P = 0.51). Among patients within the Milan criteria, the mean tumor necrosis percentage was significantly higher in the TACE plus sorafenib group compared with the TACE alone group (69.6% vs. 43.8%, P = 0.03) [80].
In summary, data on TKIs for downstaging or bridging are very limited, with no large controlled prospective studies, and most reports describing small case series where dose reductions or treatment interruptions were frequently required. Response to therapy remains unpredictable, and concerns exist that TKIs could impair tissue regeneration, which may negatively impact post-LT outcomes. Overall, the role of TKIs in pre-LT downstaging remains uncertain and investigational, with unclear optimal dosing or timing.
Another class of systemic agents that may contribute to downstaging or bridging is ICIs. Although data on their use in this setting remain limited, current evidence appears more promising compared to other systemic therapies. Definitive recommendations for their use before LT cannot yet be made. Currently, ICIs as neoadjuvant or bridging therapy remain investigational, with no standardized protocol, and further studies are needed to establish their safety and efficacy.
In a small pilot randomized trial including 22 patients, 11 assigned to the pembrolizumab plus lenvatinib (PLENTY) group and 11 to the control group, tumor-specific RFS at 30 months was 37.5% in the PLENTY group versus 12.5% in the control group. Moreover, the 12-month post-LT tumor-specific RFS reached 85.7% compared to 37.5%, respectively. Notably, no cases of graft rejection occurred in either group [81]. While ICIs can enhance antitumor immunity, their use in the pre-LT setting raises concern, as post-LT management requires immunosuppression to prevent graft rejection. This creates a potential conflict between immune activation before LT and immune suppression afterward. Several case reports, case series, and small cohort studies have been published to assess the risk of rejection and post-LT outcomes in this context.
An individual patient data meta-analysis published in 2025, including 91 cases who received ICIs before LT, reported during a median follow-up of 690 days that 26.4% experienced rejection, ∼10% died, and ∼10% had tumor recurrence. The rejection rate was similar to that observed in non-LT settings. Importantly, the interval between the last ICI dose and LT was found to be associated with rejection risk, with a median cutoff of 94 days corresponding to a rejection probability of ≤20%, suggesting a safe interval of about 3 months between the final ICI dose and LT. The study also found that patients who developed HCC recurrence had received fewer ICI cycles before LT and were less likely to have tumors within Milan criteria [82]. Later, in 2025, a prospective U.S. study evaluated 117 patients with HCC listed for LT, all of whom received ICIs. Among them, 65 cases were successfully downstaged to meet Milan criteria, and 43 eventually underwent LT (18 always within Milan and 23 after successful downstaging). The 3-year cumulative risk of waitlist dropout was around 28% for those within Milan criteria and 48% for those initially beyond. The 3-year intention-to-treat survival rates were approximately 74% and 70% for within and beyond Milan groups, respectively, while the 3-year post-LT survival reached 85%. Seven rejections occurred, six within three months of the last ICI dose, causing one graft loss [83]. A more recent international retrospective study also evaluated 119 HCC patients, all of whom received ICIs before LT. The rejection rate was 20.2%, and the study results were consistent with previous findings regarding the impact of the washout period. Specifically, washout intervals shorter than 30 days and those between 30 and 50 days were associated with an increased risk of rejection, while a washout period longer than 50 days appeared to be safe in terms of rejection risk [84].
Based on emerging and limited clinical data, post-LT management of patients previously exposed to ICIs should involve heightened vigilance for early rejection. As mentioned, current evidence suggests that younger age, shorter ICIs washout periods, and immune-related adverse events are predictors of allograft rejection and these factors may help identify patients at particularly high allograft rejection risk after LT [85]. Such patients may require more intensive immunosuppressive strategies, including higher dose and number of agents, prolonged induction therapy and a longer transition from induction to maintenance immunosuppression [86]. Maintenance immunosuppression typically includes early initiation of calcineurin inhibitors (CNIs), often combined with mycophenolate, while careful steroid tapering is recommended. Emerging data suggest that the use of triple immunosuppressive regimens, particularly those incorporating CNIs-based maintenance therapy, may reduce the risk of allograft rejection in this high-risk population [82, 86, 87]. Data on anti-T cell induction are limited, but their use during LT may help eliminate ICI-bound T cells and reduce rejection risk [83]. Given the competing risks of rejection, infection, and HCC recurrence, immunosuppressive strategies should be individualized and guided by close clinical and biochemical monitoring.
ICIs may also provide benefits in patients with TACE-unsuitable intermediate-stage HCC who are beyond the up-to-seven criteria [88]. IKF-035/ABC-HCC study showed that compared to TACE, ICIs may yield better outcomes in intermediate-stage HCC cases [89]. Using systemic treatment like ICIs may help target micrometastases that cannot be treated with LRTs alone [90]. Another area where ICIs may help is in cases with macrovascular invasion, which traditionally exclude patients from LT. Macrovascular invasion is associated with more than a twofold lower 5-year OS and 2-year RFS [91]. The anatomical extent of portal vein tumor thrombus (PVTT) is important: lobar or main-trunk PVTT carries very high recurrence and poor survival, whereas segmental or branch PVTT may achieve acceptable post-LT outcomes [92]. After successful downstaging, outcomes are no longer significantly different between patients with and without macrovascular invasion [91]. LT, particularly with careful patient selection, remains the preferred approach for downstaged macrovascular invasion. Currently, LRTs, particularly TARE, are considered for treating such cases [93, 94] and as downstaging tools for them [65, 66], and emerging evidence suggests that ICIs may also be effective in treating patients with macrovascular invasion [95, 96], potentially improving downstaging success and reducing recurrence risk. Therefore, combining ICIs with LRTs may theoretically enhance downstaging and decrease post-LT recurrence. Several studies (EMERALD-1, LEAP-012, TALENTACE) have already shown promising results for combining these approaches in intermediate-stage HCC, both with and without macrovascular invasion, in the non-LT setting [97–99]. Prospective studies with ICIs are needed to identify which patients with limited macrovascular invasion can safely undergo LT.
We still lack definitive studies defining how ICI-treated patients should be selected for LT. However, based on current evidence from downstaging strategies using LRTs, together with emerging data on ICI-based downstaging and bridging, candidate selection should be structured and multidisciplinary. In addition to tumor burden like Milan status at LT, selection should incorporate the durability of radiologic response and concordant biomarker trends, particularly sustained AFP decline. Several ICI-specific factors should also be considered, including drug class, number of treatment cycles, and the washout interval before LT. As mentioned earlier, the latter has been consistently associated with rejection risk, immune-related adverse events may also signal both enhanced ICI response [100] and higher rejection risk [85], and greater cumulative ICI exposure may be associated with lower recurrence risk [82]. Together, these tumor- and treatment-related factors may ultimately support development of more formalized selection criteria for ICI-treated LT candidates, pending validation in prospective studies. Several clinical trials are underway [101], and many key questions remain unanswered. So far, most studies have focused primarily on rejection, while our understanding of how ICIs affect HCC recurrence, predictors of recurrence, and OS after LT is still limited. Factors, such as ICI characteristics, along with patient and tumor-related variables, may influence both pre- and post-LT response. Identifying predictors of better ICI response before LT could help refine patient selection and improve post-LT outcomes, highlighting the need for further studies.

Risk stratification and prediction of post-liver transplantation hepatocellular carcinoma recurrence

Pre-liver transplantation biomarkers: AFP, dynamic AFP changes, AFP-L3, and DCP
Even among patients transplanted within the Milan criteria, HCC recurs in about 10%–15%. Because post-LT recurrence has a poor prognosis (median survival ≈ 1 year) and is the leading cause of death after HCC transplant, efforts have been made to identify additional potential predictors of post-LT recurrence beyond tumor size and number [102–105].
Pre-LT tumor markers help assess recurrence risk before transplant and AFP is most widely used. High AFP at transplant predicts recurrence, while rising AFP on the waitlist adds the risk further. High pre-LT AFP levels (>1000 ng/mL) are associated with poorer survival and higher recurrence in HCC patients [106]. Reducing AFP below 500 ng/mL before LT significantly improves outcomes, with AFP ≤100 ng/mL linked to the most favorable results, highlighting the importance of both AFP level and its dynamic response in pre-LT management. In one study, an AFP level > 25.5 ng/mL at LT was associated with a roughly 3-fold increase in recurrence (standardized hazard ratio 3.3; P = 0.001). Dynamic AFP changes (the AFP slope) are also important: this study found that not just absolute AFP but its rise on the waitlist was prognostic and a > 20.8% rise in AFP during waiting also significantly signaled recurrence [107].
Other markers, including DCP and AFP bound to Lens culinaris agglutinin (AFP-L3), have been investigated. In a prospective cohort study of patients undergoing LT, pre-LT AFP, AFP-L3, and DCP were assessed for their ability to predict RFS. Overall, 6.3% of cases (18 patients) experienced recurrence. AFP-L3 and DCP showed better predictive performance (C-statistics 0.81 and 0.86, respectively) than AFP alone (0.74). Moreover, a combination of DCP ≥7.5 and AFP-L3 ≥ 15% predicted 61.1% of recurrences, whereas only 2.6% of recurrence cases occurred in patients who did not meet these thresholds. Specifically, patients meeting those thresholds had a 3-year RFS of only ∼44%, versus ∼97% for others [108]. Interestingly, the combination of DCP ≥ 7.5 (3 points) and AFP-L3 ≥ 15% (2 points) has been incorporated into a modified version of the Risk Estimation of Tumor Recurrence After Transplant (RETREAT) score (mRETREAT) and help enhanced the original RETREAT score performance, which is further discussed in the next section. In an evaluation of 284 LT patients, RFS for patients with mRETREAT scores ≥4 was 73.2%, compared to 97.8% for those with scores <4. By comparison, the original RETREAT score showed RFS of 80.0% for scores ≥2 and 98.0% for scores <2. The area under the curve (AUC) for mRETREAT (0.86) was also superior to the original RETREAT score (0.82), indicating improved predictive performance [109]. Collectively, these studies support using both static and dynamic measurements of AFP, DCP, or AFP-L at pre-LT to better stratify HCC patients.

Integrating biological biomarkers and morphology for liver transplantation selection and outcomes
Considering that morphological characteristics (tumor size and number) and pre-LT biological markers, such as AFP, DCP, and AFP-L3, have been shown to be helpful in predicting post-LT outcomes, several criteria have been developed combining morphological and biological factors to better predict post-LT outcomes and identify more eligible candidates for LT, some of which are summarized in Table 1.
One of the first such criteria is the Kyoto Criteria [25, 110]. Cases meeting these criteria were associated with a 5-year OS rate of 82% and a recurrence rate of 4.4%. The French AFP model (or AFP score) refines transplant selection by integrating AFP level with tumor size and number [26]. It assigns points based on the largest tumor diameter, number of nodules, and AFP concentration. Patients with a total score ≤ 2 show very low recurrence rates. Even for patients beyond Milan criteria, an AFP score ≤ 2 identified those with AFP <100 ng/mL and low 5-year recurrence (14% vs. 47%). Conversely, within the Milan criteria, an AFP score >2 marked patients with AFP >1000 ng/mL who had a higher 5-year recurrence risk (37% vs. 13%). In a validation cohort of 574 non-French recipients, 5-year recurrence was about 13% for AFP ≤2 versus roughly 50% for AFP >2. Even among those within Milan criteria, higher AFP scores identified a subgroup with poorer outcomes (∼32% vs. 13% recurrence for AFP score >2 vs. ≤2) [111]. This model allows selection of patients beyond Milan who still have favorable tumor biology, effectively identifying low-risk candidates.
The Metroticket 2.0 model introduced by Mazzaferro et al. is a quantitative prognostic tool derived in Italy from a large international transplant cohort [30]. It predicts HCC-specific survival (and by extension RFS) by summing the number of tumors and their largest diameters (both treated as continuous rather than dichotomous variables) along with log-transformed AFP. Metroticket 2.0 showed that patients with AFP < 200 ng/mL should have (size + number) ≤7 to achieve about 70% 5-year HCC survival. By contrast, higher tumor burden or AFP strongly lowers predicted survival. In validation study, Metroticket 2.0 outperformed traditional binary criteria. Mazzaferro et al. reported that its concordance was significantly higher than Milan or UCSF criteria, meaning it more accurately predicted which patients would recur or die of HCC. Because it uses continuous tumor metrics, it avoids arbitrary cutoffs and can finely stratify risk across a spectrum. Several groups in Europe and North America have since demonstrated that Metroticket 2.0 (or its continuous equivalent) better discriminates recurrence risk than Milan alone.
The HALT-HCC (Hazard Associated with Liver Transplantation for Hepatocellular Carcinoma) model, published in 2017 in the U.S., was developed to assess a continuous risk score predicting OS after LT [112]. This retrospective cohort analysis included 420 patients and incorporated input variables, such as MELD-Na, tumor burden score (TBS), AFP, LT year, underlying cause of cirrhosis, neutrophil-to-lymphocyte ratio (NLR), history of LRTs, and Milan criteria status. The risk equation was defined as (1.27 × TBS) + (1.85 × lnAFP) + (0.26 × MELD-Na), generating HALT-HCC scores ranging from 2.40 to 46.42 in the. Validation using nationwide data from the Scientific Registry of Transplant Recipients (SRTR; n = 13,717) confirmed its predictive utility: higher HALT-HCC scores were associated with progressively worse OS, with 5-year survival of 78.7%, 74.5%, 71.8%, and 61.5% across increasing quartiles. The score correlated with OS (HR 1.06 per point, 95% CI 1.050–1.07) and achieved a C-index of 0.613 (95% CI 0.602–0.623). Notably, among 12,754 patients within Milan criteria, 2,714 were identified as having poor prognosis using a HALT-HCC cutoff of 17, while among 963 patients beyond Milan, 287 also exhibited unfavorable outcomes, highlighting the model’s ability to refine risk stratification beyond conventional criteria.
Finally, the NYCA score, which incorporates a dynamic AFP response, has been validated as a reliable predictor of recurrence and survival across an international cohort, outperforming traditional selection tools, such as the Milan Criteria, French AFP model, and Metroticket 2.0 [31, 32]. Incorporating a dynamic AFP response into transplant eligibility criteria allows safe expansion of candidate selection and improves access to curative LT for patients who might otherwise be excluded.

Outcome prediction using pre-liver transplantation markers and explant pathology findings
In addition to biological markers like AFP and tumor burden or morphological characteristics, explant pathology findings, including viable tumor count and size, microvascular invasion, and tumor differentiation, are strong predictors of post-LT recurrence. Differentiation is graded by the Edmondson–Steiner system (I–IV), and microvascular invasion; is defined by tumor cells in small vessels, classified as minor (<5 vessels and local) or major [113]. These features complement pre-LT markers and enhance post-LT risk prediction. Several multivariable scores were developed to improve risk prediction beyond tumor size and number. Well-known examples include the RETREAT score, the MORAL score (pre- and post-LT versions), and a UCLA clinicopathologic nomogram; these widely cited models combine pre-LT markers with explant pathology (Table 3). Below, we review commonly used risk scores, outlining their variables and prognostic performance. These tools identify small LT subgroups at very low or very high recurrence risk and help guide post-LT surveillance and potential adjuvant therapy.
The RETREAT score, developed from a multicenter U.S. cohort, incorporates three explant-based variables: microvascular invasion, AFP level at LT, and the sum of the largest viable tumor diameter plus the number of viable tumors [105]. Each factor contributes points (e.g. higher AFP, greater tumor burden, or invasion), producing a total score ranging from 0 to ≥5. RETREAT score greatly stratifies recurrence risk: in the original study, 5-year HCC recurrence was <3% for patients with RETREAT = 0 (no viable tumor or invasion and AFP < 20 ng/mL) compared with ∼75% for those with RETREAT ≥ 5. In validation study, the RETREAT score demonstrated strong predictive accuracy (C-index ∼0.77–0.86) and outperformed the explant Milan criteria regarding recurrence risk [105]. The score can be calculated soon after transplant once explant pathology is available, allowing clinicians to individualize surveillance: patients with low scores may need less frequent imaging, while those with high RETREAT require closer monitoring and may be candidates for adjuvant therapy trials. Additionally, as discussed in the earlier section, a modified RETREAT score incorporating DCP and AFP-L3 has been developed, which outperformed this original RETREAT score in patient stratification accuracy. However, the RETREAT score was derived from a U.S. multicenter cohort and depends only on explant pathology, so it cannot be used pre-transplant. Its performance may also differ across regions with varying donor types, wait-list practices, LRT use, and HCC etiologies, highlighting the need for local validation or recalibration.
The MORAL score (Model of Recurrence After Liver Transplantation) integrates both clinical and tumor-related factors to predict post-LT recurrence [102]. Developed by Halazun et al., it includes two complementary models: a pre-LT and a post-LT version. The pre-MORAL assigns points for NLR ratio ≥ 5, AFP > 200 ng/mL, and tumor >3 cm, while the post-MORAL adds explant variables, such as poor differentiation (grade IV), vascular invasion, tumor size >3 cm, and more than three tumors. Each variable is weighted by its hazard ratio to generate a composite score. In the original study, patients with high pre- and post-MORAL scores had markedly worse RFS. Both models demonstrated superior discrimination compared with the Milan criteria (C-index ≈ 0.82–0.87 vs. 0.63). The combo-Moral model for recurrence prediction also showed a C-statistic of 0.91. Clinically, MORAL enables risk assessment using routinely collected data. A Korean version, the MoRAL score, focuses on AFP and PIVKA-II, showing similar predictive performance [115]. Across studies, higher MORAL scores, whether in the Columbia or Korean versions, consistently correlate with greater recurrence risk, identifying both very high-risk cases (up to 80%–90% recurrence) and a broader low-risk group than Milan. Overall, the approach of integrating systemic inflammation, tumor markers, and pathology into a unified prognostic model is now well recognized in LT oncology. However, the results are based on retrospective cohorts, and the combined pre- and post-transplant model again includes explant findings, limiting its role to post-LT prognostication rather than pre-transplant selection. Components, such as systemic inflammation markers (like NLR), can also be influenced by intercurrent illness or treatments like infection or steroids, which may reduce score specificity in some settings.
The UCLA nomogram, developed by Agopian et al. from a large single-center cohort of 865 HCC transplant recipients, integrates multiple clinicopathologic factors into a graphical tool to predict recurrence [103]. Key variables include poor differentiation (grade IV), macro- and microvascular invasion, tumor status beyond Milan on imaging, pre-LT NLR ratio, peak AFP, and even serum cholesterol. Each parameter contributes to a cumulative score reflecting recurrence risk. In the original UCLA study, this model showed excellent accuracy (C-index ≈ 0.85) and surpassed both Milan and UCSF criteria. In clinical practice, the nomogram provides individualized post-LT risk estimates, classifying patients into low-, intermediate-, or high-risk categories and higher scores correlate with greater recurrence probability and poorer survival. However, it was developed in a single-center cohort and relies on explant pathology, which limits pre-transplant use and generalizability across different programs. Independent multicenter validation is recommended.
A novel, high-tech approach is the TRIUMPH model (Toronto Recurrence Inference Using Machine learning for Post-transplant HCC), which applies machine learning to predict HCC recurrence. In a 2025 multicenter study (n = 2,844), a regularized Cox model analyzed a broad set of pre-LT clinical, lab, and imaging variables [116]. TRIUMPH achieved the highest predictive accuracy of current scores, with a C-index of ∼0.71, better than MORAL or AFP-only models (∼0.61) and comparable, but not significantly superior to HALT-HCC (0.67). Decision-curve analysis also showed greater net benefit across clinically relevant thresholds, indicating improved distinction between patients who will recur versus those who will not, potentially informing organ allocation and post-LT management. TRIUMPH illustrates how complex predictor interactions can be leveraged by modern methods. Key inputs included AFP, tumor number/size, inflammation markers, and regional donor factors. Developed and validated across diverse populations (deceased and living donors, Asia/Europe/North America), it demonstrates broad applicability. While still emerging, machine learning–based models like TRIUMPH complement established scores, such as RETREAT and MORAL, representing the forefront of HCC transplant prognostication. It should be noted that machine-learning models require large, well-curated datasets, transparent reporting, and accessible calculators for clinical implementation. They may be susceptible to overfitting, less interpretable than traditional scores, and their performance can drop if predictor definitions or data availability differ between development and target centers.
In summary, transplant centers now have multiple validated tools to estimate HCC recurrence risk after LT. Each incorporates tumor biology in addition to size/number: for example, AFP levels, NLR, and explant pathology are common elements. No single score is universally ‘best’; rather, they are used to stratify patients into risk tiers. Low-risk patients like RETREAT = 0 or low AFP score may need only routine surveillance, while high-risk patients may be monitored intensively or considered for adjuvant therapy trials. It should be noted that most prognostic scores were derived retrospectively from cohorts with heterogeneous geographic and temporal characteristics and differ in whether they are intended for pre- or post-transplant use. Future refinements, incorporating imaging biomarkers or genomic data, may further improve prediction, but current scores already provide far more nuance than tumor diameter alone.

Post-liver transplantation hepatocellular carcinoma surveillance
After considering LT criteria and incorporating risk scores to predict HCC recurrence, it’s important to carefully monitor patients, especially those at higher recurrence risk based on prognostic scoring models. Most HCC recurrences develop extrahepatically (50%–60%), commonly in the lungs and bones, while other recurrence patterns may involve both intrahepatic and extrahepatic sites, or occur within the liver alone. Since more than 75% of recurrences appear within the first two years after LT, regular imaging surveillance during this period is necessary [117, 118]. This should include imaging modalities that evaluate both the liver and lungs.
Multiphasic computed tomography (CT), contrast-enhanced magnetic resonance imaging (MRI), and positron emission tomography–computed tomography (PET-CT) are imaging modalities that can be used for this purpose [118]. Cross-sectional imaging is generally preferred over ultrasound because of its higher sensitivity. Typically, abdominal CT or MRI scan with concurrent imaging of the lungs is performed [119]. Hepatobiliary MRI like using gadoxetic acid is particularly valuable for detecting small intrahepatic lesions and is more sensitive than CT in this regard [120]. In cases with equivocal CT/MRI findings or elevated tumor markers (such as AFP), dual-tracer PET-CT using fluorodeoxyglucose and acetate tracers can help identify occult lesions and guide treatment decisions [121]. Abdominal ultrasound alone is insufficient for post-LT HCC recurrence surveillance due to limited sensitivity [118, 119].
Three key factors in imaging surveillance for HCC recurrence after LT are the type of imaging modality, frequency of examinations, and timing of surveillance. In a survey of 48 U.S. transplant centers, most centers reported performing cross-sectional imaging of both the abdomen and chest [122], consistent with current guideline recommendations [2]. Some centers limited imaging to the abdomen, while others also included bone scans in addition to chest and abdominal imaging [122]. More than two-thirds of centers have incorporated AFP monitoring into their post-LT surveillance programs [122].
Regarding timing, there might be no significant difference between imaging every 3 months versus every 6 months [123, 124]. Most centers adopt a schedule of every 3 to 4 months during the first year, every 6 months during the second year, and every 6 to 12 months for the following 3 years [122]. However, based on recurrence risk predicted by models, such as RETREAT, shorter intervals may be considered for high-risk patients.
A 2020 study introduced the concept of cumulative exposure to surveillance (CETS), showing that achieving sufficient imaging coverage, approximately 252 days (equivalent to three surveillance scans) during the first 24 months, was associated with a higher likelihood of receiving curative treatments of surgery or ablation after recurrence and improved post-recurrence survival [123]. In other words, increasing the number of surveillance scans during the first two years after LT can contribute to better overall outcomes.
Despite these data, the optimal frequency and duration of imaging surveillance remain uncertain. Centers may individualize intervals based on patient-specific factors, such as recurrence risk scores, immunosuppression status, and clinical course. As discussed in the section on ICI use as a downstaging strategy, immunosuppression regimens influence rejection, infection, and recurrence risk; therefore, balancing immunosuppression intensity is an integral component of surveillance planning. Modulation of immunosuppression should be incorporated into risk-adapted post-LT management because higher immunosuppressive exposure, particularly high CNIs levels, has been associated with increased recurrence risk. Early CNI minimization like through antibody induction to permit lower tacrolimus exposure, is therefore commonly used in higher-risk patients [125, 126]. mTOR inhibitor-based regimens (sirolimus/everolimus) have also shown signals of reduced recurrence and improved RFS in meta-analyses and observational studies, and may be considered in selected high-risk cases [127–129]. Steroid-sparing strategies or early steroid withdrawal appear generally safe and may further mitigate oncologic risk when the likelihood of rejection is low [130]. Both the AASLD guideline and the ILTS–ILCA consensus acknowledge potential oncologic benefits of immunosuppression optimization, including reduced CNIs exposure and use of mTOR inhibitors, for lowering post-transplant recurrence risk (although not specifically for induction therapy). However, the phase 3 SiLVER trial did not demonstrate a sustained improvement in RFS beyond 5 years with sirolimus [131]. Notably, sirolimus was associated with improved OS and RFS during the 3 to 5 years after LT [131], and subgroup analyses suggested greater benefit among patients with AFP ≥10 ng/mL and those treated with sirolimus for more than 3 months [132]. Ultimately, the choice and intensity of immunosuppression should be individualized using validated recurrence-risk tools like RETREAT and balanced against rejection risk and comorbidities, recognizing that most evidence remains observational [2].
The AASLD guideline recommend using multiphasic contrast-enhanced abdominal CT or MRI and chest CT scan and using prognostic tools like the RETREAT score to estimate 5-year recurrence risk and determine optimal imaging intervals [2]. The ILTS-ILCA consensus also recommends using validated HCC recurrence risk stratification models and performing cross-sectional contrast-enhanced imaging of the chest, abdomen, and pelvis every 6 months in individuals at intermediate to high risk [56]. They also advise periodic AFP monitoring in high-risk patients with elevated pre-transplant AFP levels, although no specific interval is specified. Overall, integrating clinical parameters like AFP with high-resolution imaging modalities can facilitate early detection of recurrence and improve treatment outcomes. Ultimately, imaging surveillance should be tailored and performed alongside comprehensive clinical assessment and close monitoring of immunosuppression. In cases of elevated AFP or other tumor markers without clear radiologic findings, alternative modalities, such as PET-CT may be considered for further evaluation.

Novel biomarkers, artificial intelligence, and future direction
Several molecular and genetic biomarkers are being evaluated to improve patient stratification and, more importantly, surveillance accuracy. It has been shown that circulating tumor DNA (ctDNA) can serve as a useful biomarker for cancer recurrence in several malignancies, including colorectal and lung cancers. Several studies have also evaluated its potential role as a biomarker for HCC, suggesting that ctDNA may help detect minimal residual disease and predict post-LT recurrence. In one study of resected HCC, detectable postoperative ctDNA tumor mutational burden predicted recurrence far better than AFP. While ctDNA positivity alone was not associated with recurrence, high tumor mutational burden within ctDNA strongly correlated with shorter RFS [133]. In another study of 38 LT recipients with HCC or cholangiocarcinoma, personalized tumor-informed ctDNA assays demonstrated high feasibility and excellent specificity (100%) with 75% sensitivity for recurrence detection, performing compared to AFP and carbohydrate antigen 19–9. ctDNA may identify recurrence earlier than imaging, showing strong potential as a noninvasive adjunct for post-LT surveillance [134]. In a large real-world study of 125 HCC patients (including post-resection and post-LT cases), personalized tumor-informed ctDNA testing accurately predicted recurrence and treatment response. ctDNA positivity within 2–12 weeks after surgery strongly correlated with relapse (hazard ratio up to 18), and serial testing detected recurrence much earlier than AFP (median lead time 7.9 vs 2.2 months), highlighting its value for early surveillance and monitoring [135]. Together, these results suggest serial ctDNA testing could complement standard biomarkers, potentially flagging recurrence earlier than imaging alone.
Other liquid biopsy markers are also under investigation. For example, one study identified a drop in a serum exosomal microRNA (miR-718) as predictive of post-LT recurrence [136]. Likewise, the presence of circulating tumor cells (CTCs) before LT was strongly associated with recurrence and CTC-positive patients had significantly worse post-LT disease‐free survival [137]. These findings highlight that blood‐based tumor markers, like ctDNA and CTCs, may noninvasively signal occult recurrence.
As post-LT recurrence remains a major challenge, patient selection for LT, post-LT surveillance, and treatment guidance need to be refined. Artificial intelligence (AI), together with currently available morphological, biological, and genomic biomarkers, may help improve these processes. AI models have been used to predict both recurrence and survival among HCC patients. A recent study developed an AI-based predictive model for HCC recurrence after surgical resection using data from 958 patients across multiple centers. The model, built with a multilayer perceptron, identified key pre-surgical risk factors, including liver function, disease etiology, ethnicity, and modifiable metabolic risks, and achieved high accuracy in cross-validation and external testing. An online tool was created for real-time, individualized risk prediction [138]. Another retrospective study of 393 HCC patients evaluated machine learning methods to predict survival across all disease stages. Using demographic, clinical, pathological, and laboratory data, key mortality predictors were identified with feature selection techniques. Machine learning models achieved up to 91% recall for 6-month survival in early-stage patients and up to 92% accuracy for 3-year survival in advanced-stage patients [139]. These findings highlight the potential of AI-driven models and machine learning for personalized risk stratification and targeted interventions to reduce HCC recurrence and increase survival.
AI-driven imaging analyses are being developed for use specifically in the LT setting. CT-based radiomics signatures extracted from pre-LT scans using machine learning have shown high accuracy in predicting recurrence. In one study of HCC patients after LT, CT-based radiomics features, particularly from the arterial phase, were used to construct a predictive signature for recurrence. A combined model integrating the radiomics signature with clinical risk factors demonstrated good accuracy for RFS (C-index ∼0.79), supporting radiomics as a noninvasive biomarker for identifying aggressive tumors and predicting post-LT recurrence [140]. In a more recent 2025 study of 245 HCC patients undergoing LT, a CT-based deep learning radiomics nomogram (DLRN) was developed using tumor and peritumor features combined with clinical variables. The DLRN accurately predicted early recurrence with AUCs of 0.884 (training) and 0.829 (validation), and high scores conferred a 16-fold increased recurrence risk. Combining DLRN with Metroticket 2.0 criteria further improved prediction (AUC up to 0.936), demonstrating its value as a noninvasive tool for risk stratification and post-LT surveillance [141].
Future research should evaluate these AI models for patient risk stratification and tailoring imaging surveillance. Emerging tools, such as ctDNA assays, advanced liquid biomarkers, and AI-enhanced imaging, hold promise for early detection of HCC recurrence after LT.

Conclusion
LT is the most effective curative option for selected HCC patients, with lower recurrence rates than resection or ablation. LT selection criteria have evolved from strict morphologic cutoffs like Milan or UCSF criteria, to integrated models combining tumor morphology, biomarkers, and treatment response. Expanded frameworks (Metroticket 2.0, French AFP, HALT-HCC, and NYCA) allow more accurate prognostication and broader LT eligibility in patients with favorable tumor biology.
Downstaging with LRTs effectively converts many patients initially beyond criteria to LT eligibility, yielding post-LT survival comparable to those within Milan, though dropout rates remain higher for larger or AC tumors. TACE and TARE show similar downstaging efficacy, with TARE potentially requiring fewer sessions and causing less morbidity, but study heterogeneity limits firm conclusions. Systemic therapies, particularly ICIs, show promise for downstaging and may enhance outcomes when combined with LRTs. However, they carry notable post-LT rejection risks linked to shorter washout intervals. A 3-month washout appears safer, though evidence remains limited and prospective studies are needed to define optimal protocols.
Consensus is needed on standardized post-LT surveillance timing and modalities, typically combining AFP and imaging every 3 to 6 months in the first 2 years. With expanding LT criteria, individualized risk stratification based on explant pathology, validated scores like RETREAT or MORAL score and emerging tools, such as ctDNA and AI/radiomics, is essential to guide surveillance intensity and identify candidates for adjuvant trials.
Best practices for LT in HCC should include: (i) integrated selection combining morphology, biomarkers, and treatment response; (ii) individualized downstaging with LRTs based on tumor features and center expertise; (iii) cautious use of ICI therapies with standardized washout protocols before LT; and (iv) risk-adaptive, multimodal post-LT surveillance using validated scores, imaging, and emerging liquid biomarkers. Prospective multicenter studies are needed to standardize downstaging algorithms, establish safe pre-LT ICI protocols, and validate combined biomarker–imaging surveillance strategies to minimize recurrence and improve long-term graft and patient outcomes.

Authors’ contributions
Concept and Design: J.D.Y., M.S.R-Z. Data Acquisition and/or Interpretation: M.S.R-Z, J.D.Y. Drafting the Manuscript: M.S.R-Z. Critical Revision: J.D.Y. Final Approval and Agreement: J.D.Y, M.S.R-Z.