1 Introduction

Neuroendocrine tumors (NETs) are a heterogeneous group of cancers arising from neuroendocrine cells, most often in the alimentary or respiratory tracts. The incidence of NETs has been constantly increasing, which may be related to improvements in imaging and pathological diagnostic techniques. According to Surveillance, Epidemiology, and End Results Program (SEER) data, the overall incidence of gastro-entero-pancreatic (GEP) NETs increased from 1 case per 100,000 to 3.65 cases per 100,000 from 2003 to 2007 [1]. The prevalence of metastatic disease in NETs is very high (40–50%) and cancer-related death in these patients is generally caused by distant sites of disease [2,3,4]. The liver is the most frequently involved target, representing the sole distant site of the disease (oligo-metastatic presentation) in nearly half of NET patients and in up to 60% of GEP-NETs [5]. According to the classic “anatomical/ mechanical” hypothesis [6], the majority of the alimentary tract – where the majority of primary NETs are located – is drained by the portal venous system, explaining the dominance of liver metastases (MNETs). In a recent publication from the Swedish tumor registry, NETs showed a relatively predominant spread to the liver, irrespective of the primary tumor site [5]. By accurate pathological examination, MNETs are often multifocal, and less than half of the lesions are detectable at radiologic imaging [7]. Tumor grading of the primary tumor is associated with metastatic disease: according to SEER data, 50% of poorly differentiated (G3) or undifferentiated (G4) NETs had synchronous distant metastases, as opposed to 30% of cases in G1-G2 NETs [1]. In metastatic NETs, 5-year overall survival ranges between 19% and 38%, although survival has increased over the last two decades [1]. Given the slow growing nature of these tumors and their long natural history even if untreated, it is difficult to demonstrate the efficacy of a therapy with respect to another in terms of survival benefit. Moreover, especially in functioning tumors, any therapy should aim also at symptom control, thus making both quality of life and survival important primary endpoints for these patients. In this article, the non-pharmacologic options available for treatment of MNETs – liver resection (LR), loco-regional therapies (LRT) and liver transplantation (LT) – will be reviewed, with special consideration for the curative potentials of LT in selected subgroups of MNET patients.

2 Liver-directed therapies: resection and loco-regional treatments

According to the latest ENETS guidelines “it is strongly recommended if not mandatory to discuss NEN patients after accurate imaging and pathology review in a multidisciplinary tumor board for appropriate therapeutic decision making, especially to exploit surgical therapy in potentially resectable NEN patients and explore loco-regional therapies upfront” [8]. These recommendations are based on the results of retrospective series, including comparative studies and case-series (Table 1), rather than on randomized controlled trials, which are very difficult to conduct in this context.

Table 1 Literature review of available treatments for GEP-NET liver metastases (> 20 patients treated)
Full size table

All major guidelines agree on the fact that in patients with resectable disease and no extra-hepatic tumor spread, liver resection should be offered as the first treatment option for MNETs, especially in case of metastases limited to the liver (oligometastatic NETs).

The available evidence on the results of liver resection is mostly based on single-institution small series, resulting in a very heterogeneous population. As a consequence, the reported median overall survival (OS) after hepatic resection largely varies from 52 to 190 months, with a 5-year OS between 41% and 100%. Median disease-free survival (DFS) ranges between 15 and 50 months, and is below 40% at 5 years in most series, reflecting the fact that liver resection offers a chance of definitive cure in a relatively small and selected proportion of patients [9,10,11,12,13,14,15,16]. Cytoreductive debulking surgery can be considered for symptom control in functional tumors as an alternative to or in combination with loco-regional treatments, but there is no evidence that R2 resections improve OS or PFS [13, 15]. In one study addressing the role of cytoreductive surgery [15], 23 patients treated with debulking resections had a mean survival of 32 months compared to 24 months in patients treated with embolization (p < 0.001). In another study [13], 10-year OS after R0 and R2 resections were 90.4% and 51.4%, respectively, as compared to 19.4% in non-resected patients. These results are affected by important selection biases, since non-surgical patients had higher tumor load in the liver with respect to those eligible to loco-regional treatment; therefore, no definite conclusions could be drawn on the relative effectiveness of surgical resection. As a matter of fact, the real entity of liver involvement (tumor burden) is frequently underestimated preoperatively even by modern imaging and patients submitted to potentially curative resection often turn out to be unresectable after surgical exploration and re-staging with intraoperative ultrasound.

In patients with unresectable metastatic disease confined to the liver, or with limited extrahepatic spread, intra-arterial therapies should be offered. The rationale of these treatments stands upon the vascular supply of metastases that is largely arterial, especially in liver metastases from GEP-NETs. The most common intra-arterial techniques are trans-arterial embolization (TAE), trans-arterial chemoembolization (TACE) and Y90-radioembolization (TARE). Since metastatic lesions are often bilobar, two treatment sessions – with one lobe treated at a time – are usually required, in order to reduce toxicity. In the absence of large and well-conducted trials, there are no clear recommendations to guide the choice between available intra-arterial therapies. Several retrospective series of TACE/TAE in NET metastases report a median OS ranging from 23.2 to 55 months, depending on site of the primary tumor, grading and burden of disease, with a median time-to-progression (TTP) ranging from 5 to 24 months [17,18,19,20,21,22,23,24, 41]. The experiences with TARE are more limited with reported OS ranging from 14.6 to 70 months. Possible indications to TARE at present could be the contraindications to an embolic therapy, i.e. portal vein thrombosis and presence of bilio-enteric anastomoses, the latter condition being observed after resection of the primary tumor from the pancreatic head [25,26,27, 42]. Intra-arterial therapies play an important role in symptom control, particularly in case of high tumor load in the liver: in functional tumors these therapies should be used early to prevent carcinoid crisis, as outlined in 2016 ENETS guidelines [8].

Loco-regional therapies are also indicated in case of tumor progression under somatostatin analogs (SSA), which is the recommended fist-line therapy for any MNETs. The commercially available SSA agents, octreotide and lanreotide, are effective for symptom control in functionally active tumors [30,31,32], and demonstrated anti-tumor activity in two randomized controlled trials [29]. For progressive disease despite SSA treatment, the targeted drugs everolimus and sunitinib are available, since three registration trials showed positive results in this setting [34, 36, 43]. For both drugs, median reported TTP ranged between 11 and 12 months, while a median OS of 32 months was observed with everolimus in a field-practice retrospective series [33]. Combination of SSA and targeted drugs is often used in functional NET, associating tumor growth inhibition and symptom control.

Another option for progressive metastatic NET expressing somatostatine receptors (SSTR) is Peptide Receptor Radionuclide Therapy (PRRT) using either Y90 or Lu177-labelled SSA. Retrospective series reported a median OS of 46–94.6 months and a median PFS of 16.3–40 months for patients treated with PRRT [35, 37, 38]. In a recently published registration trial [39], PRRT offered a prolonged PFS with respect to high-dose ocreotide. This has opened the door to a more systematic use of PRRT, especially in case of advanced midgut NETs. The potential limitations of renal and myeloid toxicity have to be considered.

3 Liver transplantation

Although pharmacologic options now available show promising results, surgery – when possible – remains the most efficient and successful approach for the treatment of neuroendocrine liver metastases, offering the longest-lasting benefit in terms of survival and symptom control in comparison with other available treatments. Curative resection of liver metastases (i.e., resection with clear margins and absence of microscopic residual disease, R0) is feasible only in a small proportion of cases, as 80% of MNETs are unresectable due to bilateral liver involvement and multifocal disease at diagnosis [40]. For this reason, total hepatectomy and liver transplantation (LT) is a potential chance of radical cure for patients with MNETs, with the possibility of a complete removal of all the macroscopic and microscopic metastatic tumor load. In most countries, MNETs are an accepted indication for LT, because of the limited biological aggressiveness and slow growth of the disease. However, experience of liver transplantation for MNETs remains anecdotal, representing only 0.3% and 0.2% of liver transplants (European Liver Transplant Registry and United Network for Organ Sharing database, respectively) [44] and is still a debated indication to LT.

3.1 Prognostic factors

The series of LT for NET metastases reporting at least 10 patients are summarized in Table 2. Overall survival at 5 years ranges from 40% to 90% in the most recent series, while 5 years RFS ranges between 20 and 80% [44,44,45,46,47,48,49,50,51,52,53,54,55,57, 59]. This large heterogeneity in terms of long-term outcomes is mostly explained by the large variability of selection criteria applied by different centers. The main recognized prognostic factors are the following.

Table 2 Literature review of liver transplantation for metastases from NETs (> 10 patients)
Full size table

3.1.1 Extrahepatic disease and resection of the primary NET

Extrahepatic tumor locations found at the time of diagnosis is an absolute contraindication to LT, since systemic and peritoneal involvement prevent a complete tumor resection at the time of LT. For this reason, a robust staging using both MRI and gallium68 octreotate PET-CT scan with a specific focus on ruling out bone and lymph node metastases is mandatory prior to enlisting and while on the waitlist [58, 60,60,62]. The resection of the primary tumor should be performed with a R0 curative intent, and should include locoregional and distant lymph node dissection [63]. In a French multicentric study, radicality in resection of the primary tumor was independently associated with post-LT survival, being 5-year OS 9% for patients with a R1-R2 resection vs. 53% for patients who underwent R0 resections [52]. Resection of the primary tumor can be performed at the time of LT or in a staged fashion. A systematic review of the European Liver Transplant Registry including 213 patients showed overall 5-year OS of 52% [48] but a high perioperative mortality of 10% was observed, possibly due to the inclusion of cases in which major procedures were added to LT. In a multicentre retrospective study with data predominantly from the US, Sher et al. recently reported similar survival outcomes (52% 5-year OS) and showed that extent of resections added to LT was a poor prognostic factor [46]. According to these data, simultaneous resection of the primary at the time of LT seems to have a detrimental effect on short and long-term outcomes after LT, and should generally be avoided.

In some patients presenting MNETs, the location of the primary tumor cannot be determined. The French multicentric series reports 14 patients (16%) who underwent LT for MNEts from unknown primary and achieved a 5-year OS of 59% [52]. Based on these findings, it appears that unknown primary location can be discussed and not excluded a-priori for LT consideration.

3.1.2 Site of primary tumor

The site of primary tumor is an independent prognostic factor for post-LT outcomes. In particular, among primary NETs with a portal drainage, significantly better outcomes were reported for GEP-NETs rather than pancreatic NETs (pNETs) [45, 52, 56]. This could be partially explained by the higher postoperative mortality resulting from extended surgical procedures such as pancreato-duodenectomy performed before or in association with LT. Moreover, pNETs usually show higher proliferation indexes with respect to ileo-jejunal NETs and this may account for a higher risk of post-LT recurrence, even though promising results might be achieved with LT if specific selection criteria (i.e., age < 55 years, limited burden of disease in the liver and transplantation performed an a separate operation with respect to LT) are applied to pNET transplant candidates [50].

For what concerns MNETs deriving from primary tumors not drained by the portal system (i.e. bronchial and rectal NETs), it would appear reasonable to exclude these patients from LT candidacy, since the liver might not be the “first level” of hematogenous spread of malignant cells from the primary site [63]. Evidence on this issue is not univocal: in two multicentric series including patients who underwent LT for MNETs from bronchial tumors [48, 52], this location of the primary tumor did not turn to be an independent prognostic factor for OS. Conversely, post-LT recurrence in pNETs is usually detected at lymph-nodal rather than systemic sites, which claims that patient selection on portal drainage may limit hematogenous cancer dissemination while not preventing lymphatic spread. Thus, primary location in the extra-portal sites may be considered a relative rather than absolute contraindication to LT. These patients should undergo a careful preoperative work-up to rule out extrahepatic metastases.

3.1.3 Metastatic tumor burden: Liver involvement

Poor outcomes for patients with extensive metastatic liver involvement are mainly related to early and aggressive post-LT recurrences, even in well-differentiated MNETs [60]. Touzios et al. reported that tumor burden within the liver is a major predictor of 5-year survival, being 8% vs. 67% for patients with liver involvement >50% vs. <50%, respectively [11]. These data have been confirmed by other series [64], and our group has set a cut-off of 50% liver involvement as a selection criteria for LT [63]. Alternatively, other groups evaluated liver involvement by the presence of hepatomegaly based on estimation of liver volume on CT-scans. Le Treut et al. identified hepatomegaly (defined as an explanted liver volume more than 20% greater than standard liver volume) as an independent predictor of poor prognosis [52]; this result was subsequently confirmed in a registry-based European series [48].

3.1.4 Tumor biology and histology

As for other oncologic indications to LT such as hepatocellular carcinoma, tumor biology and histologic grading are closely related to the outcome of LT for MNETs. According to a recent meta-analysis, Ki-67 index (involved in cell proliferation) should be used as a prognostic marker for both survival and tumor recurrence in patient selection for LT [65]. In 2002, a retrospective case-series from the Hannover group reported that patients with tumors in which Ki67 expression was greater than 10% were at higher risk for recurrence and worse survival [57]. Similiarly, Frilling et al. reported a satisfactory disease control rate (5-year RFS of 48%), resulting in an OS of 67% at 5 years by selecting patients with Ki67 < 10% and no evidence of extrahepatic metastatic spread [54]. As for tumor grading, many authors have reported poor outcome after LT for poorly differentiated MNETs [45,45,46,48, 52], confirming its influence on the risk of post-LT recurrence. In a recent series, Grat et al. were able to show an optimal 10-year OS of 78.8% by excluding poorly differentiated disease [47]. Some cautions should be exerted in comparing different series, as the classification system used for tumor grading and morphology in NETs have evolved over time. According to the available evidence, well- to moderately-differentiated MNETs with a Ki67 level below 10% appear the most reasonable selection criteria for LT candidates at good prognosis. Poorly differentiated tumors (NET G3) are unanimously considered as a contraindication for LT.

3.1.5 Pre-transplant observation period

Since 2001 [66] our group suggested a staged approach, with a minimum period of 6 months from the time of primary tumor removal to the time of enlisting: this approach may assure disease stability over an adequate time interval, and may allow for detection of early loco-regional recurrences that would turn to be a contraindication to LT [54]. In a registry-based study on the United Network for Organ Sharing (UNOS) database, wait time to LT turned to be significantly related to post-LT survival. A progressive prognostic advantage for patients with longer waits was found, with 5-year OS of 63% vs. 36% for patients waiting more vs. less than 2 months, respectively. These findings confirm that disease stability for a longer period before LT may be an appropriate selection criteria [44].

3.2 Selection criteria for LT in MNETs

The conventional strategy for the management of patients with metastastic NETs takes into account the LT option as a salvage therapy for patients with very advanced disease. Conversely in recent years, more accurate studies on the prognostic factors summarized above, have led to the reappraisal of liver transplantation as a potentially curative strategy for highly selected cases of metastatic cancer. The results of LT for MNETs may depend on which of these two strategies is followed (Fig. 1).

Fig. 1
figure 1

Natural history of metastases from NETs and timing of LT [67]

Full size image

In a meta-analysis of 103 LT for MNETs [68], Lehnert identified two independent risk factors for poor survival: extended procedures (i.e., LT associated with upper abdominal exenteration) and age over 50 years. The 5-year survival rate was 65% in patients with none of the factors, 32% in patients with one positive factor, and 0% in patients with both factors. The author concluded that LT for MET could be effective in young patients if not associated with extended operations. In 2008, Le Treut et al. reported a multicentric French study in which OS was independently related to pancreatic location and presence of hepatomegaly. Five-year OS were 76%, 66%, and 12% in presence of both, one or none of these factors, respectively. The authors concluded that patients presenting pNETs and hepatomegaly were unlikely to benefit from LT [52]. These data were subsequently confirmed by the multivariate analysis of the data of 213 patients in the European Liver Transplant Registry: the significant risk factors were multivisceral resections or multivisceral grafts, a poorly differentiated tumour and hepatomegaly [48]. If patients with none or one of such risk factors are selected, the rate of 5-year survival can reach 80%.

Based on these and other features our group from Milan proposed the following stringent criteria for potential candidates for LT: a) age less than 60; b) G1/G2 grading according to WHO classification; c) primary tumour drained by the portal system; d) metastatic involvement limited to the liver (oligo-metastatic NETs); e) tumor burden involving no more than 50% of the hepatic parenchyma; f) no tumor progression for 6 months prior to listing [63]. Although these criteria have not been externally validated, their strict application has shown excellent 5-year and 10-year OS of 97.2% and 86.9%, respectively, with an extremely limited recurrence rate (88.8% and 86.9% 5-year and 10-year RFS), as recently reported by our group [45]. Comparable results (90% OS) were obtained by Olausson et al. in unselected patients with older age, higher grading and larger tumour burden [53]. Rosenau et al. reported a 80% 5-year OS in 19 symptomatic patients with higher proliferation index who underwent LT [57]. However, disappointing 5-year DFS rates around 20% were reported in both series, possibly accounting for decreased OS if a longer observation period could be allowed.

Curative potential of LT for MNETs seems to be related to the ability of restrictive criteria to select patients with limited and liver-only metastatic involvement, associated with favourable tumor biology. Restrictive criteria applied to the most radical form of loco-regional treatment of a metastatic disease – as LT can be defined – have the inherent limitation to focus on only one of the possible sites of disease spread (i.e. liver and loco-regional lymph-nodes). Such a strategy may succeed only if extra-hepatic NET deposits are virtually absent. This may result in the prolongation of DFS and, in favourable cases, in cure of disease with annihilation of tumor recurrence.

3.3 Risk of post-LT recurrence

As stated, the risk of recurrence following LT is a relevant issue, possibly exacerbated by immunosuppressive therapy. Due to the pro-carcinogenic effect of immunosuppression, recurrent metastatic NET following LT may display more aggressive biologic features and involve less typical sites such as bone, thyroid, ovary and central nervous system spread. Although results are extremely heterogeneous among different series, the long-term recurrence rates reported in multi-institutional series and LT registries remain high and that questions the curative role of LT in this setting. In the European multicentre study by Le Treut, as low as 30% 5-year RFS rate was reported, in keeping with data from American studies [48]. Similarly, analysis from UNOS by Gedaly showed an unsatisfactory 5-year RFS of 32% with a recurrence rate of 31%, significantly higher than the 10–15% recurrence rate of HCC, the most relevant tumor indication to LT [44]. By following restrictive criteria (the Milan-NET criteria, Table 3), our group reported relatively low rates of recurrence (13.1% and 13.1% at 5 and 10 years, respectively) [45]. Such results are comparable to those achieved for HCC and may justify LT also in the context of MNETs. However, since recurrences may occur very late after LT, a close and long-lasting postoperative follow-up is mandatory for patients undergoing LT for MNETs.

Table 3 Milan selection criteria for liver transplantation in patients with liver metastases from NET
Full size table

3.4 Outcomes of LT in comparison with other therapeutic options

A recent systematic review evaluated the available literature comparing post-LT outcomes with those of non transplant options [67]. There is no prospective randomized trial to show the superiority of LT over resection or non-surgical treatment. Because of the highly selective nature of patients chosen for surgical treatments (resection or transplantation), randomized controlled trials evaluating patient outcomes with non-surgical modalities would be difficult to perform.

In the very few comparative studies [51, 66] including resection or non-surgical treatment as a control arm, the 5-year OS of transplanted patients was about the same as non transplanted patients, but the 5-year DFS was higher in transplanted patients (50% vs. 34%). Considering survival after diagnosis of liver metastases, recent series report 5-year OS ranging between 69 and 97.2% in patients who received LT [45, 52], 34% and 50.9% in patients undergoing non-surgical treatments [4, 45], and 20–30% in untreated patients [1].

In the Milan series [45], after propensity score adjustment, in a group of 42 LT vs. 46 non-LT patients, a significant survival advantage in the transplant group was maintained (HR: 10.67; 95%CI: 3.48–32.72; p < 0.001) with respect to non-transplant options, indicating LT as the best option in selected patients with liver metastases from NETs.

Such good survival results have to be balanced against surgical morbidity of LT, with perioperative mortality as high as 10% in the analysis from the European Liver Transplant Registry by Le Treut [48]. This may have a negative initial survival impact as compared to medical treatment and may have contributed to the poor outcome of some series. However, such morbidity burden has been related to concomitant procedures added to LT (multivisceral transplantation or other resections), as shown also in the US multicenter study by Sher [46]. Furthermore, the long-term survival achieved by LT exposes transplanted patients to the long-term complications of LT and immunosuppression. In the same systematic review from European centers on 213 patients, 17 died because of late morbidity related to LT without any recurrence. De novo neoplasms, accounting for almost 30% of deaths 10 years after LT for other indications, may be a relevant issue in this context of long survival following LT, but no data are available [69]. In any case, such events should be weighed against the side effects of lifelong medical treatments for patients undergoing non-surgical approaches. To date robust evidence from randomized trials comparing different treatment modalities is lacking and current indications are based primarily on retrospective data from case series with no controls.

3.5 Is inclusion of LT in the NET guidelines justified?

European guidelines (ENETS) recognize LT as a treatment option in the management of diffuse unresectable hepatic metastases. However, its role is mainly limited to patients with symptoms due to carcinoid syndrome or other functional NETs, refractory to medical treatment including loco-regional therapies and PRRT [8, 70]. A curative role for LT is stated as exceptional. Guidelines seem to have incorporated recommendations from our and other groups concerning selection criteria. Ki67 of 10% is a suggested cut-off not to be exceeded, also suggesting that only well- and moderately-differentiated tumours (G1/G2) be selected for LT candidacy. Age < 55 is recognized as a favourable prognostic factor as well as primary tumour removal before LT (6 months). An extensive workup including Ga68-PET is recommended to exclude extrahepatic metastatic spread, and a 6-month observation period is suggested to rule out aggressive behaviour. Finally, <50% liver involvement is a widely accepted minimal requirement, although up to 75% involvement can be considered for patients with refractory hormonal symptoms. The American transplant system represented by the UNOS (United Network for Organ Sharing), has included the Milan-NET criteria in the MELD exceptions allowing transplant listing [71], while NANETS guidelines are less precise and consider LT as an option in younger patients without risk factors, with a metastatic carcinoid tumor that is unresectable and limited to the liver [72].

3.6 Is LT for MNETs justified with respect to other transplant indications?

Liver transplantation is a demanding procedure, and its applicability is limited by organ shortage. Because of these limitations, all diseases that include LT as a treatment option are in competition. In this setting, an indication to LT should be justified if two conditions are met:

  • acceptable long term outcomes (utility principle), meaning that the limited resource has been used correctly;

  • proved survival benefit, meaning that, whichever the indication, a positive survival gain offered by LT in comparison with the best alternative therapy should be demonstrated.

Concerning utility of LT for MNTEs, the transplant community generally considers LT as justified if the minimum predicted OS at 5 years is 50% [73]. As shown in Table 2, most of the series of LT for MNETs fulfil this requirement, particularly in the more recent years in which selection criteria for LT have been progressively refined. Moreover, in a registry-based study from the UNOS database, Gedaly et al. compared the results of LT for MNETs metastases with those for HCC during the same period, and found no statistically significant difference between the two indications, supporting LT as an option in selected patients.

As for survival benefit, clear evidence of the role of LT for MNETs as compared to other therapeutic options is lacking due to inconsistency of available data. To date, the only study evaluating the survival benefit of LT with respect to the best available non-surgical therapy in a consecutive series of patients with similar characteristics was performed by the Milan group [45]. We were able to demonstrate that the transplant survival benefit in NET patients significantly increases over time with respect to non-transplant options (p < 0.001), escalating from approximately 7 months after 5 years to about 3.5 years (38.4 months) after a decade of follow-up (Fig. 2). These results now allow enlisting of NET patients for LT as MELD exceptions.

Fig. 2
figure 2

Survival benefit of LT for liver metastases from NETs with respect to other non-transplant available treatments [45]

Full size image

4 Conclusions

In conclusion, in selected patients with non-resectable NET liver metastasis, LT offers the best long-term outcome among any other available treatment for metastatic NET. Considering organ shortage, a careful selection of patients is of utmost importance in order to derive a solid utility and benefit from LT. The use of stringent criteria - such as the Milan-NET criteria - to select patients for LT may guarantee the accomplishment of those aims, even though further studies are needed to evaluate the reproducibility of these results.