Early reports described dismal outcomes, with 20%–38% 5-year OS and 53%–84% recurrence rates, despite anecdotal cases of long-term survival in early-stage node-negative patients, and a controversial role of primary sclerosing cholangitis (PSC) [33–36].

In 1993, the Mayo Clinic [37] described a novel protocol proposing LT for pCCA after thorough selection and aggressive neoadjuvant chemo-radiotherapy (Figure 1; Table 1). The first large case series [38] demonstrated excellent outcomes. Of 184 enrolled patients, 172 completed chemoradiation and underwent staging surgery, and 126 ultimately received LT. The 5-year intention-to-treat (ITT) survival was 54%, reaching 61% in patients with underlying PSC and 42% in those with de novo pCCA. Recurrence occurred in 21 patients (18%) after a mean time of 25 months.

These findings were confirmed by a multicenter study [41] involving 287 patients from 12 high-volume transplant centers across the USA. In this cohort, 71 patients dropped out before undergoing LT, 5-year ITT survival was 53%, and the recurrence-free survival (RFS) was 65%.

Following these encouraging results several groups in Europe and the US started following the Mayo protocol or Mayo-like protocols with similar inclusion criteria and slight modifications in the neoadjuvant treatment. However, rather small case series were reported.

A long-term analysis from the Mayo Clinic [42] reported 349 patients (1993–2018), of whom only 60% ultimately underwent LT. OS at 5- and 10-year was 69% and 62% in the per-protocol analysis and 51%, and 46% in the ITT analysis. Interestingly, a significant difference in survival was observed between patients with PSC-associated and those with de novo pCCA (5-year OS 76% vs. 58%).

A recent meta-analysis [43] of 20 studies comprising 428 patients reported a 31.6% pooled 5-year OS following LT without neoadjuvant therapy, compared to 65.1% in patients who completed neoadjuvant chemoradiation. Furthermore, 3-year recurrence was significantly lower among those who received neoadjuvant therapy (24% vs. 51.7%). However, despite the Mayo Clinic cohort being the largest (152 patients), marked heterogeneity was observed in the application of the protocol across studies. While most studies selected patients with unresectable tumors smaller than 3 cm and excluded those with prior resection or biopsy, only a subset included patients with PSC, elevated CA19-9 levels, or malignant stricture in the absence of positive cytology. Furthermore, some groups (including the Mayo Clinic during its early experience) excluded patients with tumor extension beyond the origin of the cystic duct, to avoid pancreaticoduodenectomy. Although preoperative staging was universally performed, the extent of lymph node sampling varied considerably. Only three studies strictly followed the original Mayo chemoradiation regimen, while others introduced modifications such as substituting 5-fluorouracil with capecitabine or gemcitabine/cisplatin, or omitting chemotherapy or brachytherapy altogether.

In line with these principles, the recent Milan consensus [6] recommended LT only after neoadjuvant treatment with Mayo chemo-radiation regimen, and in presence of an unresectable pCCA <3.0 cm, with no evidence of nodal or distant metastases, no previous surgical manipulation nor transperitoneal biopsy. Interestingly, the jury supports LT also in case of borderline or dubious preoperative resectability.

The strength of the Mayo Protocol lies in rigorous criteria and locally-aggressive neoadjuvant treatment. However, the relative contribution of these two components to the overall success remains uncertain. An ELITA-ELTR study showed that a subgroup of patients who met Mayo criteria but did not receive neoadjuvant treatment before LT had similar excellent long-term oncological results (5-year OS 59%) and fared significantly better than patient outside criteria (5-year OS 21%) [44].

Although some centers [45], question the utility of neoadjuvant treatment, available data suggest that this approach increases the risk of positive margins and disease recurrence, meanwhile preventing a test of time on biological aggressiveness.

An Italian survey [39] showed that several patients underwent LT for pCCA without receiving neoadjuvant therapy, due to concerns regarding the use of radiotherapy and its short- and long-term complications, along with the requirement to deviate from current SOC chemotherapy.

These concerns are shared by the Mayo Clinic group [46], who, despite reporting excellent outcomes in the per-protocol cohort, also observed a worrisomely high 31% dropout rate (41% for de novo and 15% for PSC-associated pCCA). They further highlight the high toxicity of chemoradiation: nearly all patients develop recurrent cholangitis, while vascular friability often results in ischemic cholangiopathy and strictures, frequently progressing to liver failure in the absence of transplantation [38, 42, 47].

Even assuming a therapeutic effect of the neoadjuvant regimen, the considerable dropout rate implies that a several patients who ultimately did not undergo transplantation received suboptimal chemotherapy while being exposed to treatment-related complications, without any survival benefit [45]. It has been argued that radiotherapy (and consequently radiosensitizing fluorouracil) could be avoided unless their role in improving post-transplant outcomes is definitively established, as they are not currently included among standard treatments for advanced pCCA [4, 5].

Primary sclerosing cholangitis (PSC) is a major predisposing condition for pCCA. Management is particularly challenging due to diffuse biliary involvement, impaired liver function, and a pro-oncogenic field that favors multifocal and synchronous neoplastic transformation [48]. Surgical resection is technically demanding and often associated with high morbidity and incomplete oncologic clearance [49]. When performed according to the Mayo Clinic protocol, LT yields superior outcomes in PSC-associated pCCA compared to de novo cases, reflecting earlier diagnosis, less aggressive tumor biology, and the concurrent treatment of both the malignancy and the underlying cholangiopathy [42, 43]. Reported results show 5-year overall survival of 65%–70% and recurrence rates of 20%–24%, making neoadjuvant therapy and LT the treatment of choice in candidable patients with PSC-associated pCCA [48].

Hepatopancreatoduodenectomy (HPD) is an extremely complex and technically demanding procedure associated with high morbidity. The technique has been mainly developed and reported by Japanese groups [50–54], who also provided most of the available outcome data [52, 55–59]. A recent meta-analysis [60] reports a 90-day mortality of 10% and morbidity of 64%, although mortality can approach zero in highly experienced centers [61]. Outcomes show marked geographic variability, with 90-day mortality of 26% in North America [62], 13%–17% in Europe [63], and <5% in Japan [58, 61].

The combination of total hepatectomy, pancreaticoduodenectomy (PD) and LT has been poorly explored. PD may be performed simultaneously with transplantation or delayed by weeks to months, but evidence is limited to small series and case reports [35, 42, 64–68], with long-term survival mainly driven by CCA recurrence [64]. The addition of pancreaticoduodenectomy increases morbidity, particularly due to technical complexity and to the impact of immunosuppression on pancreatic healing. Pancreatic fistula, reported in up to 24% of cases [64], is especially critical in the transplant setting because of the risk of vascular anastomotic injury or compression; total pancreatectomy or a two-stage approach may be considered in case of complications.

Although potentially advantageous for optimizing transplant timing in the neoadjuvant setting, LDLT was limited by concerns regarding the risk of arterial thrombosis related to perihilar irradiation. Although jump-grafts to the aorta can be used, and the middle-colic or right gastroepiploic artery were employed to avoid performing anastomoses in the irradiated field, these strategies remain technically demanding [69, 70]. Recent neoadjuvant protocols that omit pre-transplant radiotherapy [39, 71, 72] have renewed interest in the use of LDLT for pCCA.

A retrospective analysis [73] by the Mayo Clinic compared 73 cases of LDLT performed for pCCA (66% PSC-associated) with 173 LDLTs for other indications. The pCCA group showed higher requirement for arterial or portal vein reconstruction and Roux-en-Y choledochojejunostomy. Rates of early hepatic artery thrombosis were similar between the two groups (5.4% vs. 7.6%), whereas late arterial (18.9% vs. 4.1%) and portal (37.8% vs. 8.7%) complications were more frequent in the pCCA group, although these did not affect long-term survival. 5-year OS was significantly lower in the overall pCCA cohort (66.5% vs. 87%), and differed between de novo (47.5%) and PSC-associated (75.9%) cases. The Mayo Clinic tried to address the issue of operating in an irradiated field by introducing technical modifications, particularly within LDLT protocols [73]. These include nonstandard arterial reconstruction (avoiding irradiated hepatic artery, use alternative inflow sources with interposition grafts, anastomosis to the infrarenal or supraceliac), portal reconstruction (using jump grafts or anastomosis to the superior mesenteric vein or splenic vein confluence below the irradiated field) and systematic biliary reconstruction with Roux-en-Y choledochojejunostomy [42, 47].

The University of Kyoto [70] drafted a modified protocol for LDLT in pCCA, consisting of GCS chemotherapy administered for more than 2 months, followed by external-beam radiotherapy only in case of disease stability. In their initial report on 10 patients, only five proceeded to LT, achieving 100% 1-year survival rate, with one recurrence after 10 months. Hepatic artery thrombosis and delayed gastric emptying occurred in two and three patients, respectively.

The excellent long-term outcomes after LT, contrasting with persistently poor results after liver resection, raised the issue whether LT should be extended beyond unresectable disease to include borderline-resectable or even resectable cases. Only few studies addressed this issue, and case series are small and heterogeneous. A 2019 systematic review and metanalysis [74] of studies comparing LT and LR suggested a trend towards longer OS after LT, although not statistically significant. Their analysis, however, showed comparable mortality rates, but shorter hospital stay and higher rates of R0 margins after LT. In contrast, the most recent report from the Mayo Clinic [46] focusing on de novo pCCA, demonstrated superior results of LT compared with resection (with or without vascular resection) in terms of OS (78 vs. 25.8 vs. 58.2 months) and perioperative mortality (4% vs. 8% vs. 7%) in the per-protocol analysis. However, the high dropout rate (31% in the LT group, 28% in the surgical group) had a substantial impact on the ITT analysis, which failed to demonstrate a significant survival advantage of LT over resection.

Dropout rate is a crucial and underestimated factor. The randomized TRANSPHIL trial (NCT02232932), comparing neoadjuvant chemoradiation and LT with liver resection for resectable pCCA, reported poor long-term survival and a dropout rate exceeding 50%, ultimately leading to early termination for ethical reasons. Exposing resectable patients to both the toxicity of chemoradiation and the high likelihood of dropout and futility may ultimately condemn them to poor outcomes associated with chemotherapy alone, rather than the still unfavorable but comparatively better results of resection. These findings warrant caution and at the same time support efforts to improve consistency in preoperative management, favoring SOC chemotherapy over chemoradiation [39].

A benchmark study [75] involving 134 patients from 17 high-volume centers provided several important insights. Ideal cases were defined as treated at high-volume centers (≥50 LT/year), who underwent neoadjuvant chemoradiotherapy, had tumors <3 cm, negative lymph nodes, and no significant comorbidities. Benchmark thresholds included 90-day mortality rate ≤5.2%, 1-year Comprehensive Complication Index (CCI) ≤33.7, ≤66.7% grade ≥3 complications, and R0 resection margins rate ≥80.0%. For long-term outcomes, the benchmarks for 5-year disease-free survival (DFS) and OS were ≥43.8% and ≥60.0%, respectively. Authors advocate for recognizing unresectable pCCA treated with neoadjuvant chemoradiotherapy as a formal indication for LT, and propose extending its use to resectable cases based on the observation that benchmark outcomes of LT for pCCA not only exceed those of LT for other indications [76] but also those of surgical resection [21]. In this study, benchmark LT cases were also directly compared with a matched cohort of curatively resected, node-negative Bismuth IV patients, demonstrating significantly superior 5-year DFS (50.2% vs. 17.4%) and OS (56.3% vs. 39.9%) in the LT group, with no significant difference in major complications (72.7% vs. 74.6%) and a higher 3-month mortality rate in the resection group.

To address the limitations of the Mayo protocol, several groups shifted to a neoadjuvant treatment based on SOC chemotherapy, with various adoption of radiotherapy or transarterial radioembolization (Figure 2).

An undergoing Italian trial (LITALHICA, NCT06125769) maintains Mayo selection criteria but replaces neoadjuvant radiochemotherapy with SOC chemotherapy to avoid altering the patient’s therapeutic pathway solely due to trial inclusion [77].

Ongoing trials are summarized in Table 2, the main clinical studies and their key outcomes are reported in Table 3, and the main statements with supporting studies, corresponding levels of evidence, and relevant guideline recommendations are provided in Supplementary Table 1.

In cirrhotic patients with severe portal hypertension and small unresectable iCCA, LT may simultaneously treat the tumor and the chronic liver disease.

A 2014 Spanish retrospective study [83] reported 62% 5-year OS after transplantation among cirrhotic patients with small incidental iCCA. A subsequent international study [84] showed that similarly defined “very early” iCCA (<2 cm) had better outcomes compared to “advanced (multiple or >2 cm) tumors (5-year OS 65% vs. 45%). Risk of recurrence at 5 years was also lower in the very early group (18% vs. 65%), although tumor size was not a predictor of tumor recurrence at multivariate analysis.

A metanalysis [85] of 18 studies including 355 cases, showed that cirrhosis, was positively associated with RFS, and at subgroup analysis patients with very early iCCA had superior pooled 5-year RFS compared to advanced iCCA (67% vs. 34%). To be noticed, incidental diagnosis was not associated with either prolonged OS or RFS.

However, real life applicability of the 2 cm cutoff may be difficult, as pre-transplant confirmation such small unresectable iCCA is quite uncommon. Both HCC and CCA can develop on cirrhosis, as long as mixed HCC-CCA forms, and preoperative differential diagnosis can be challenging [86–88]. Indeed, a prospective trial on LT for early iCCA (NCT02878473) by the Toronto group was terminated because of low accrual.

Attaining R0 resection for locally advanced iCCA can be challenging even in the non-cirrhotic [20]. To this respect, total hepatectomy followed by LT represents a resection with the highest potential for radicality, provided that there are no lymphnode involvement and extrahepatic spread. Evidence suggest, however, that patients should be selected based on surrogates of favorable tumor biology, namely response to neoadjuvant chemotherapy and test of time of disease control.

The group from UCLA [89, 90] in a 24-year single center experience on 35 cases, highlighted how patients receiving LT had significantly better outcomes than those receiving resection (5-year RFS 33% vs. 0%). Moreover, in the LT group, patients receiving neoadjuvant and adjuvant chemotherapy had better survival compared to those receiving no therapy or adjuvant therapy alone (5-year RFS 47% vs. 20% vs. 33%). On multivariate analysis, recurrence was not associated with tumor size, but rather with factors biology-related factors like multifocality, infiltrative pattern, pernieural and lymphovascular invasion, history of PSC, neoadjuvant and adjuvant therapy. In 2022, the same group reported their 30-year experience [91] (19 pCCA and 30 iCCA), confirming excellent oncological results for LT even for large size CCAs compared to patients not receiving preoperative treatment, particularly when adopting a multimodal chemotherapy and loco-regional neoadjuvant approach (5-year OS 100% vs. 41%).

The Houston Methodist-MD Anderson group developed a protocol offering LT to patients with unresectable iCCA, without evidence of macrovascular or lymph node involvement, who had sustained tumor stability with gemcitabine-based neoadjuvant therapy for more than 6 months [92]. Their latest series [93] (18 patients) showed post-LT OS of 71%, and 57% at 3 and 5 years respectively. Tumor recurred in 39% of patients after a median time of 11 months after LT, being treated with further systemic therapy and surgery. Interestingly, transplanted patients had a median number of 2 iCCA tumors and a median cumulative tumor diameter of 10.4 cm, confirming that acceptable OS can be achieved independently from size in presence of good response to therapy and disease stability. Next-generation sequencing was performed in most cases, using liquid biopsy, percutaneous biopsy, or explant tumor tissue. Known genetic alterations were identified, including FGFR (27%), CDKN2A (7%), IDH1 (35%), BRAF (19%), and TP53 (19%), but univariate analysis showed no association with outcomes. In selected patients, the presence of targetable alterations enabled the use of targeted therapies, including the FGFR inhibitor pemigatinib (1 case), the IDH1 inhibitor ivosidenib (2 cases), and the PARP inhibitor olaparib (4 cases).

Patients with iCCA were traditionally excluded from LDLT because of insufficient expected OS and RFS to justify the donor’s risk. However, the evolving diffusion of the concept of transplant benefit as gain in life-years quality-adjusted over alternative available therapies is now changing such a perception, provided the achievement of a minimal 5-year survival to avoid futility.

Literature remains limited [94, 95]. A multicenter study from Japan [96] on 19 LDLT recipients incidentally diagnosed with iCCA showed 46% 5-year OS. A similar study from Pakistan [97] including patients with incidental pCCA and iCCA, collected 16 patients, reporting 47% 3-year RFS (64% for well differentiated tumors). The largest series [98], focusing on LDLT outcomes for primary sclerosing cholangitis, included 55 out of 805 cases with iCCA, with OS reaching 81.9%.

The 2024 ILTS–ILCA Consensus [16] recommends considering LDLT for iCCA within institutional study protocols, particularly for patients with early-stage disease. At the same time, the Toronto group is currently leading a multicentric trial (NCT04195503) to validate LDLT’s efficacy for advanced iCCA.

Early reports showed markedly worse survival after LT compared with resection [99, 100], although patient matching was frequently biased [94].

A propensity score–matched (PSM) analysis by Jung et al., (16 LT vs. 100 resections), showed comparable OS and recurrence rates [101]. Several groups subsequently analyzed data from the US-NCDB [102–105], consistently reporting similar OS between LT and resection. However, in a multivariate analysis [104], LT was associated with a significantly reduced risk of death compared with matched resection cases.

Huang [106] recently analyzed the US-SEER database, including 2538 patients with iCCA treated with curative surgery (2425 resections, 113 LT) and 5048 LT for HCC. PSM between resected and transplanted iCCA groups corrected the baseline imbalance, since patients with early stage, smaller tumors, well-differentiated histology, and cirrhosis were more likely to be selected for LT. LT patients had significantly longer survival than those who underwent LR in the matched cohorts (median OS: 23 vs. 18 months; 5-year OS 52.8% vs. 29.9%). Interestingly, a subgroup analysis showed that patients who met recommended selection criteria (i.e. very early iCCA on cirrhosis or locally advanced iCCA after chemotherapy) had a 5-year OS of 43.8% and 61.7% respectively.

Preliminary data [107] from the TESLA trial report 5 patients showing excellent perioperative course after LT after neoadjuvant treatment, although two experienced disease recurrence within 12 months (Figure 3; Table 4).

In Italy, the LIRICA trial, is enrolling patients with unresectable iCCA after 6 months of SOC chemotherapy. As for LITALHICA trial, unresectability is assessed by a dedicated multidisciplinary tumor board and patients are listed for LT only after 6 months of documented disease stability.

The Milan-INT group is investigating the neoadjuvant combination of chemotherapy and transarterial radioembolization (Y90-TARE). Preliminary data from 13 patients revealed a 69% dropout rate due to disease progression or inadequate response. Only four patients proceeded to transplantation, showing favorable early outcomes [108].

Ongoing trials are summarized in Table 4, the main clinical studies and their key outcomes are reported in Figure 3 and Table 5, and the main statements with supporting studies, corresponding levels of evidence, and relevant guideline recommendations are provided in Supplementary Table 2.

Tumor burden cutoffs and patterns are already included in most LT protocols for patient selection, both in pCCA [6, 38] and iCCA [84].

Even though not yet integrated in patient selection processes, molecular biology might play a crucial role. In particular, iCCA exhibits marked heterogeneity, with significant diagnostic, prognostic, and therapeutic implications. Small-duct and large-duct subtypes, differ in morphology, cellular origin, clinical behavior, and molecular characteristics. Small-duct subtype generally confers better prognosis and is often associated with targetable genetic alterations (IDH1/IDH2, FGFR2). In contrast, large-duct iCCA more commonly harbors classical adenocarcinoma genetic alterations, such as in KRAS and TP53 [11]. Next-generation sequencing enables detection of actionable mutations, and is gradually transforming oncologic care [111]. A large meta-analysis of 1,481 resected iCCA cases demonstrated that patients with tumors harboring KRAS, TP53, and/or SMAD4 mutations had significantly worse OS and RFS compared to those with FGFR2 fusions, IDH mutations, BAP1 mutations, or no major genetic abnormalities [112]. Additional mutations in RB1, ERBB2, and BAP1 are also frequently observed in iCCA, and up to 70% of patients harbor potentially targetable genetic alterations [113]. Pemigatinib and infigratinib, showed progression-free survival (PFS) of 6.9 and 5.8 months, respectively, in patients with FGFR2 fusions [114–116]. Ivosidenib (IDH1 inhibitor) improved both PFS and OS in the ClarIDHy trial and was approved as palliative treatment [117, 118].

Ca19-9 is elevated in 60%–80% [19, 119] of cases, and its diagnostic accuracy is limited by false positives in biliary obstruction and cholangitis. High preoperative and postoperative values of CA19-9, particularly when associated with increased carcinoembryonic antigen (CEA), are linked with advanced disease and worse OS and RFS [120–122]. On the contrary, a >50% reduction in CA19-9 after systemic therapy is strongly associated with radiologic response [120] and improved survival, and could serve as therapeutic objective.

Liquid biopsy is a non-invasive technique [2, 123] enabling detection of circulating tumor-derived material, including circulating tumor cells, ctDNA, ctRNA, microRNAs, and extracellular vesicles. ctDNA has emerged as a key biomarker for genomic profiling and tumor burden assessment [123–125], showing high concordance with tissue mutation profiles [126–128], with variant allele frequencies correlating with tumor load and supporting its role as a dynamic indicator of disease status [124, 129, 130]. ctDNA is detectable across all disease stages and carries prognostic value, with ctDNA-positive patients showing poorer progression-free survival both pre- and postoperatively. During surveillance, ctDNA detection is associated with significantly worse relapse-free survival and identifies recurrence in 93.8% of cases, with a mean lead time of 3.7 months over imaging [131].

Bile-derived cfDNA appears particularly promising due to direct tumor contact, detecting driver mutations in 54% of cases compared with 17% in plasma [132, 133].

However, limitations include lower sensitivity for gene fusions compared with tissue RNA-based assays, variability in ctDNA shedding depending on tumor burden and site, lack of standardization in extraction methods, platforms and timing, need for prospective interventional validation, and cost and reimbursement issues [123–125]. ctDNA remains complementary rather than a replacement for tissue testing, increasing actionable variant detection by 14.3% when used concurrently, and is particularly valuable when tissue is insufficient, unavailable, or when rapid or serial assessment is required [126, 128, 134].

Radiomics-based machine learning shows excellent diagnostic accuracy [135, 136] in CCA, particularly when integrated into clinical–radiomic models [137], often achieving performance comparable to postoperative pathology. Its main strength lies in improving diagnosis and preoperative prediction of microvascular invasion [138–140], gene mutations, perineural invasion [141], and lymph node metastasis [142, 143]. This stratification may guide surgical decision-making and enable prediction of early recurrence [144, 145] and survival [146]. Emerging deep learning models enable multimodal integration of radiology, pathology, and molecular data, but remain limited by data heterogeneity, poor interpretability, lack of standardization, and the need for prospective multicenter validation. Despite expert-level performance, clinical translation is hindered by regulatory constraints, cost sustainability, algorithmic bias, and insufficient validation [137, 147, 148].

Functional and metabolic assessment could provide useful insights as shown in CRLM transplantation setting [149]. PET is not recommended [48] for tumor diagnosis due to limited accuracy, but shows good performance in detecting lymph node and distant metastases [150–154]. SUVmax is an independent prognostic factor for disease-free and overall survival [155–159]. While 18F-FDG PET/CT is established for detecting metastatic disease and recurrence with high specificity, PET/MRI provides superior staging accuracy, particularly for T and N staging [160–162]. A new tracer 68Ga-FAPI PET/CT, targeting cancer-associated fibroblasts in the tumor microenvironment, shows high positive predictive value, high detection rates and better outcomes compared to 18F-FDG in terms of detection of primary tumors, lymph nodes, and distant metastases [163–168].

Integrating the patient’s clinical trajectory into the selection algorithm provides a longitudinal perspective on disease behavior [169]. The “test of time” itself reflects an indolent tumor biology, characterized by disease confinement within the liver and the absence of systemic or circulating tumor spread. Similarly, response to therapy serves as a surrogate marker of favorable tumor biology and informs postoperative management, as patients who respond to treatment before transplantation are more likely to maintain therapeutic sensitivity thereafter. For these reasons, as in the setting of CRLM [170], these principles have been incorporated into most modern neoadjuvant protocols, emphasizing refined patient selection over procedural acceleration [39, 93].