Multimodal management of neuroendocrine liver metastases

Abstract

Background:

The incidence of neuroendocrine tumours (NET) has increased over the past three decades. Hepatic metastases which occur in up to 75% of NET patients significantly worsen their prognosis. New imaging techniques with increasing sensitivity enabling tumour detection at an early stage have been developed. The treatment encompasses a panel of surgical and non-surgical modalities.

Methods:

This article reviews the published literature related to management of hepatic neuroendocrine metastases.

Results:

Abdominal computer tomography, magnetic resonance tomography and somatostatin receptor scintigraphy are widely accepted imaging modalities. Hepatic resection is the only potentially curative treatment. Liver transplantation is justified in highly selected patients. Liver-directed interventional techniques and locally ablative measures offer effective palliation. Promising novel therapeutic options offering targeted approaches are under evaluation.

Conclusions:

The treatment of neuroendocrine liver metastases still needs to be standardized. Management in centres of expertise should be strongly encouraged in order to enable a multidisciplinary approach and personalized treatment. Development of molecular prognostic factors to select treatment according to patient risk should be attempted.

Introduction

Neuroendocrine tumours (NET) represent a heterogeneous group of neoplasms with distinct morphological and biological manifestations.^1–5 They are defined as either non-functioning tumours with symptoms related to mass effects and malignant tumour disease or functioning tumours with specific hormones/neuropeptides autonomously secreted in sufficient amounts to induce specific clinical syndromes. A further hallmark of the tumour cells is the ability to take up amine precursors and/or express somatostatin receptors (SSTRs). Most common anatomical sites of their origin are the gastrointestinal tract and the bronchopulmonary system.^2–11 Neuroendocrine tumours can develop either in a sporadic form or as a component of inherited endocrine tumour susceptibility syndromes such as multiple endocrine neoplasia, von Hippel-Lindau (VHL) syndrome and neurofibromatosis-type 1.^12,13

Historically, NET have been considered as rare tumours comprising approximately 0.5% of all malignant conditions.^14,15 More recent series encompassing large prospective tumour registries, e.g. Surveillance, Epidemiology and End Results (SEER), report on the linearly increasing overall incidence and changing spectrum of NET manifestation over the past four decades.^5,16–23 With a global incidence of approximately 5–7 cases per 100 000, gastroenteropancreatic NET have the second highest prevalence of all gastrointestinal cancers. However, it remains a matter of debate if an increased awareness of the disease, widespread use of advanced diagnostic techniques or true changes in incidence are responsible for these observations. Within the intestinal group, rates for appendiceal NET have decreased whereas those for NET localized within the stomach, small intestine and rectum have increased.^5,16–18 Women with both diabetes and a positive family history for cancer have been shown to harbour an increased risk for developing gastric NET.²⁴ Conversely, NET patients have an estimated risk of 22.4% to have a second malignant non-neuroendocrine tumour, most often adenocarcinoma of the small bowel.⁵

Survival rates differ between the anatomic tumour locations being the worst for pancreatic tumours and more favourable for tumours arising in the respiratory tract or in the appendix and for localized rectal NET.^5,16,25–29 Beside the site and size of the primary tumour,^30,31 proliferative activity and the mitotic index,³² presence of distant metastases,^5,19,33 age,^16,30 and several other factors, such as depth of tumour invasion,³⁰ vascular and lymphatic invasion,³³ plasma chromogranin A concentration,^34,35 urinary 5-hydroxyindoleacetic (5-HIAA) levels²⁸ and cellular atypia, all influence tumour prognosis³²

In contrast to the traditional opinion that NET represent an indolent disease, two large population-based studies from the United States⁵ and the United Kingdom²⁰ encompassing 13 715 and 4104 patients, respectively, disclosed a rather poor overall 5-year survival of 67.2% regardless of tumour location and 5.2–57% when considering different histological subsets of digestive NET. Of note, no survival improvement over the last three decades could be registered in either survey. In 12.9% of the patients, metastases are already detectable at the time of initial tumour diagnosis⁵ and a substantial number of patients (5–10%) presents with metastases and primary of unknown origin. Beside regional lymph nodes, the liver is the predominant site of NET metastases. Up to 75% of patients with small bowel NET and 30–85% of those with tumours localized within the pancreas present with liver metastases (LM) either at initial evaluation or during the course of their disease (Fig. 1a–b).^5,36,37 An additional 5–10% of NET patients present with LM with unknown primary tumour site. Based on the NET cases included in the SEER 9 registry data of 1977–2004, Yao and colleagues observed a significantly better survival rate in patients treated since 1988.²¹ The authors hypothesize that implementation of octreotide in the management of NET in 1987 might have contributed to this development.

Figure 1.

Intraoperative finding in a patient with a 1 cm in size ileal neuroendocrine tumor, necrotic mesenterial lymph node metastases (a), and bilobar liver metastases (b)

This review was performed to elucidate the prognostic impact of LM in patients with neuroendocrine tumours and to address the role of the large variety of treatment options now available.

Impact of liver metastases on prognosis of neuroendocrine tumours

Hepatic metastases are the most powerful prognosticator of survival in patients with NET regardless of the primary site.³⁸ A 13–54% 5-year survival in historical cohorts of patients with untreated neuroendocrine LM compared with 75–99% in those free of hepatic deposits underlines the unique molecular genetics of malignant NET and, at the same time, clearly delineates them from their non-endocrine counterparts.^{13,31,39–41} For patients with gastrinoma without LM a 95% survival at 20 years has been reported in contrast to 15% 10-year survival in the presence of diffuse metastatic bilobar hepatic involvement.⁴² A 5-year survival in small bowel NET decreases from 60% to 70% in the absence of LM to 50–60% if LM occur.^5,31,43 Colorectal NET rarely present with LM; however, once metastasized, 5-year survival is less than 30% compared with 75–88% in localized tumours.^5,34,44 Progression of LM associated with manifestation of bone metastases, as frequently seen in poorly differentiated tumours, accounts as an indicator of a very aggressive disease. Not only the presence of LM, but also the pattern of their distribution, is a crucial determinant of prognosis.^34,45–47 It is not unexpected that limited hepatic tumour load, in general, amenable for treatment, is associated with a favourable survival. Aside from the impact of treatment on the prognosis of patients with neuroendocrine LM, it was postulated that the pattern of hepatic metastatic spread corresponds to the diverse tumour biology.^46,48,49

Beside rarity and heterogeneous biological features of NET, lack of a uniform and generally adopted classification system burdens a reliable prognostic estimation of the metastasized condition. Supplementing the World Health Organization (WHO) classification of NET⁵⁰ and its further refinement,⁵¹ efforts were made to introduce new staging and grading systems that would serve as a clinical tool for risk and prognosis assessment of an individual patient. The European Neuroendocrine Tumor Society (ENETS) adopted the current tumour-node-metastasis (TNM) WHO classification and proposed a new grading system which considered the mitotic count and Ki-67 index^52–54 (Table 1). A United States group introduced a new TNM staging system by incorporating tumour size and depth of invasion to the T category.^55–58

Table 1.

The ENETS criteria for assessing the prognosis of non-functioning (neuro)endocrine pancreatic tumours^{52–54,239,240}

Biological behaviour	WHO classification	Metastases	Invasion	Histological differentiation	Tumour size (cm)	Angioinvasion	Ki-67 index (%)	Mitotic count (10 HPF)
Benign (low risk)	Group 1	−	−	Well differentiated	≤2	−	≤2	<2
Benign or low-grade malignant (intermediate risk)	Group 1	−	−	Well differentiated	>2	±	≤2	<2
Low-grade malignant	Group 2	+	+	Well differentiated	usually >3	+	3–20	2–20
High-grade malignant	Group 3	+	+	Poorly differentiated	any	+	>20	>20

10 HPF: high power field = 2 mm², at least 40 fields (at 40× magnification) evaluated in areas of highest mitotic density.

Ki-67 index determined with MIB1 antibody; % of 2000 tumour cells in areas of highest nuclear labelling.

Clinical manifestations

Depending upon the degree of endocrine activity and the local effects of hepatic tumour load, patients with metastasized NET may stay asymptomatic for a longer time or develop incapacitating hormone-excess symptoms in terms of carcinoid syndrome and/or local discomfort. Vague clinical symptoms often delay diagnosis of LM particularly in those originating from non-functioning tumours. Even in patients with extensive metastatic hepatomegaly, hepatic malfunction and hepatic failure are seldom seen.

With regard to morphologic manifestation of neuroendocrine LM, in general, three growth types can be discriminated: single metastasis of any size (type I); isolated metastatic bulk accompanied by smaller deposits, with both liver lobes always involved (type II); and disseminated metastatic spread, with both liver lobes always involved, a single lesion of varying size and virtually no normal liver parenchyma (type III).^46,59 The three growth types reflect a different grade of aggressiveness of the disease and overlap with the staging system for gastrointestinal NET encompassing the tumour grading categories G1–G3. While G1 and G2 tumours are considered to correspond to well-differentiated NET of different level of aggressiveness, G3 tumours correspond to a poorly differentiated neuroendocrine carcinoma that follows a highly aggressive course.^52,53 In accordance with this hypothesis, Sutliff et al. determined three different growth rates of untreated hepatic metastatic gastrinoma: 26% of their patients demonstrated no growth over a 29-month mean observation period, 32% showed slow growth and 42% had a rapid growth (>50% volume increase per month) over an 11-month follow-up.⁴⁹

Diagnosis

Biochemical diagnosis

Liver metastases may be the first manifestation of NET. To classify the tumour tissue as neuroendocrine, both biochemistry and specific imaging procedures are helpful.

The biochemical diagnosis of NET is based on the evaluation of specific tumour markers. Plasma chromogranin A (CgA), one of the acidic proteins stored within secretory granules of neuroendocrine cells and secreted by a large variety of NET, is a widely accepted tumour marker with respect to diagnosis, prognosis and monitoring of the treatment.^60–63 Its sensitivity depends upon the NET type and tumour burden. Patients with LM tend to have significantly higher CgA concentrations than those without metastases.⁶⁴ False-positive results can be obtained with renal insufficiency,⁶⁵ hypergastrinemia per se or as a result of proton pump inhibitor therapy, hepatic failure and inflammatory bowel disease. The main urinary serotonin metabolite, 5-HIAA serves as an additional sensitive tumour marker for diagnosis and follow-up. It is increased in patients with LM of NET from the small intestine or the pancreas. For detection of tumours with low serotonin production or those lacking DOPA decarboxylase, measurement of platelet and urine serotonin, respectively, has been recommended, but it is not widely applied in clinical practice.^28,66,67 Of note, while a serotonin-rich diet may lead to a false-positive urinary 5-HIAA concentration, the determination of platelet serotonin will not.⁶⁸ Other circulating non-specific peptide markers for NET are synaptophysin, neuron-specific enolase, pancreatic polypeptide, human chorionic gonadotropin, parathyroid hormone-related protein or calcitonin. However, all of them are of lower clinical relevance.^69,70 In functioning tumours, those hormones related to the specific syndrome, e.g. insulin, gastrin, glucagons or vasoactive intestinal polypeptide should be determined.

Morphologic imaging

After its introduction in the 1990s, somatostatin receptor scintigraphy (SRS) has rapidly evolved as the gold-standard imaging procedure for NET expressing somatostatin receptor subtype 2. Beside the advantage of total-body imaging with the potential of simultaneous visualization of the primary tumour and metastatic deposits (Fig. 2), SRS offers the possibility to identify those patients who might be candidates for somatostatin receptor-based radiotherapy.^71–73 The initially used somatostatin analogue [indium -111 (111 In)-diethylene-triamine-penta-acetic acid (DTPA)-D-Phe1]-octreotide (111 In-DTPAOC) (Octreoscan®)^74–77 has been replaced in several units for the benefit of more sensitive and highly specific radiopharmaceuticals such as 1,4,7,10-tetra-azacyclododedcan-4,7,10-tricarboxy-methyl-1-yl-acetyl-D-Phe1(DOTA) Try3 -octreotide (DOTATOC) and DOTA-1-Nal3-octreotide (DOTANOC) with high affinity to SSTR-2 and SSTR-2, -3, and -5, respectively.⁷⁸ Moreover, the labelling of DOTATOC or DOTANOC with the positron emitter gallium-68(68 Ga) (68 Ga-DOTATOC or 68 Ga-DOTANOC) provides the option of positron emission tomography (PET) imaging with a diagnostic sensitivity up to 30% higher than achieved by standard scanning.^79–82 Precise fusion of functional 68 Ga-DOTATOC PET images with a morphologic image modality, such as CT (68 Ga-DOTATOC PET/CT), facilitates additional information.^83–87 For patients with NET showing negative 68 Ga-DOTATOC or DOTANOC PET/CT results or those with elevated plasma serotonin levels, serotonin and catecholamine precursors 18 F-dihydroxy-phenyl-alanine and 11 C-hydroxy-tryptophan, respectively, as tracers are recommended.^88–90 Moreover, it has been shown that SRS frequently revised the primary tumour staging, detected tumour recurrence after resection of LM at an earlier stage than standard imaging in one-third of cases,⁹¹ and changed the clinical management in 33–77% of NET patients.^46,72 For the assessment of the additional benefit of more sensitive 68 Ga-DOTATOC PET/CT more clinical data are still needed.

Figure 2.

⁶⁸Ga-DOTATOC PET/CT revealing bilobar liver metastases and a neuroendocrine primary localized within the small bowel (arrow). The spleen shows physiological radionuclide accumulation

Contrast-enhanced abdominal ultrasound and multiphasic helical computed tomography (CT) serve as the standard initial imaging procedures. Utilizing advanced CT protocols and fusioning of CT with single positron emission tomography, a sensitivity of 94–100% has been noted.^74,92,93 As an alternative to CT, T2-weighted magnetic resonance imaging (MRI) using gadolinium, or in some institutions, mangafodipir trisodium as liver-specific contrast agents is recommended for routine staging, assessment of disease progression and monitoring of the therapy. MRI allows reducing excessive radiation burden.^94–96

Histopathology

The diagnostic workup of LM should encompass tissue acquisition for histopathological examination, especially so, as the staging of NET depends on markers of proliferation (Ki-67 and/or the mitotic index) and histopathological evaluation of vascular or neural invasiveness. Tumour staging predicts the prognosis and tailors the therapeutic strategy^59,97 particularly in patients who are not candidates for complete resection. As the cytological diagnosis of neuroendocrine carcinoma remains unreliable, a core needle biopsy or a biopsy via a laparoscopic approach with the additional option to assess perihepatic lymph nodes may be more useful.⁹⁸ Recently, the group from Yale University introduced polymerase chain reaction (PCR)-based measurement of CgA gene expression as a promising new tool for detection of neuroendocrine tumour cells in apparently normal lymph nodes or equivocal liver lesions.⁹⁹

In a substantial number of patients with neuroendocrine LM, palliative therapy is the only possible approach. Thus, quality-of-life assessment at the initial diagnosis and throughout treatment is recommended in order to balance the benefits and potential side-effects of the therapy applied.^100–103Figure 3 illustrates diagnostic evaluation and treatment decisions for patients with well-differentiated neuroendocrine LM.

Figure 3.

Evaluation and treatment decision for patients with well-differentiated neuroendocrine liver metastases⁴⁷. LM, liver metastases; CRR, cytoreductive resection; RFA, radiofrequency ablation; LT, liver transplantation; TACE, transcatheter arterial chemoembolization; SIRT, selective internal radiotherapy; PRRT, peptide receptor radionuclide therapy; EHD, extrahepatic disease; REHD, resection of extrahepatic disease; US, ultrasound; FNB, fine needle biopsy

Treatment

The best strategy for treatment of neuroendocrine LM is still poorly defined despite the existence of various guidelines^{59,100,104–107} and comprehensive reviews addressing this topic.^97,108–114 This is in part because of the lack of prospective studies comparing different treatment modalities in homogenous cohorts of patients. Two recent Cochrane reviews on the optimal management of LM originating from gastrointestinal NET revealed that no evidence for optimal therapeutic strategies exists.^115,116

Selection of patients for surgery with curative intent

Meticulous patient selection is the key issue in hepatic surgery with the potential for cure. When selecting candidates for the resectional approach, the following prerequisites should be considered^59,117: (i) well-differentiated NET (G1 or G2 status, Table 1),¹¹⁸ (ii) reasonable performance status and degree of comorbidities, (iii) exclusion of non-resectable extrahepatic disease, (iv) possibility of R0 resection with a remnant functional liver parenchyma of ≥30%, (v) absence of advanced stages of carcinoid heart disease as increased central venous pressure can trigger severe haemorrhage during liver surgery,^109,119,120 (vi) if indicated, cardiac valve surgery prior to hepatic resection in patients with advanced stages of carcinoid heart disease,¹²⁰ and (vii) acceptable morbidity (<15%) and mortality (<1–1.5%) of the anticipated procedure. Ideally, the primary tumour should already be eliminated prior to liver surgery or considered as resectable synchronously with LM in selected patients suitable for aggressive surgery.^121–123 In cases of unknown primary at the time of liver surgery, close post-operative follow-up with the aim of identifying the primary location or in rare cases, ultimately classifying the resected hepatic disease as a primary of hepatic origin, must be assured.

In centres dealing with neuroendocrine LM, not only advanced skills in hepatic surgery but also special techniques to achieve resectability in patients initially considered as technically unresectable, should be available. Three-dimensional (3D) imaging software and 3D CT angiography have proved to be valuable for visualization of biliary and vascular structures. In addition, these techniques allow for the calculation of post-resectional liver remnant during pre-operative planning. In those cases where the projected liver remnant is below 30%, portal vein embolization (PVE) or right portal vein ligation can be performed to induce hypertrophy of the future hemiliver.^{117,121,124,125} In patients with bilobar metastases requiring extended surgery, radiofrequency ablation (RFA) or cryotherapy can be applied as an adjunct to liver resection, most commonly contralaterally to the site of hepatectomy.^117,118 If complete tumour elimination is not feasible, two-stage hepatectomy, either as a sole procedure or in combination with PVE, may be an option in selected patients.^126,127 After a two-step surgical strategy in 19 patients with synchronous bilobar LM of digestive NET, Belghiti and his group achieved a 5-year overall and disease-free survival of 94% and 50%, respectively.¹²⁸ Their strategy encompassed the primary tumour resection and limited non-anatomical resection of the left liver lobe including ligation of the right portal vein. After hypertrophy of the tumor-free left liver lobe, a right or extended right hepatectomy was carried out.

Surgery

Management of the primary tumour in patients with LM

Ideally, the primary tumour, including regional and distal lymph nodes, should be radically resected prior to treatment of LM. This condition enables adequate immunohistological tumour classification and grading, reliable assessment of resectional margins and exploration of the abdominal cavity for additional extrahepatic tumour manifestation.

Assuming the primary negatively influences survival in patients with LM, surgery of the primary tumour has been recommended even in the presence of unresectable LM.^129–131 Apart from the resection of the primary tumour, Ki-67 and age at diagnosis have been identified as additional independent prognostic factors in the presence of LM originating from mid-gut NET.¹³² Among 84 patients reviewed by Givi et al. the median survival time in the group with resected primary was 159 months, in contrast to only 47 months in the unresected group.¹³⁰ In patients with symptomatic non-functioning pancreatic NET and unresectable LM, Hung et al. demonstrated a marked reduction of symptoms and a lower rate of progression in those with resection of the primary compared with the unresected cohort.¹³¹ In addition to the prognostic and symptomatic benefit, primary tumour resection including enlarged perihilar and mesenteric lymph nodes reduces the risk of intestinal obstruction, portal vein obstruction and ischaemia, as a result of encasement and compression of the mesenteric root.^109,133,134 For patients with unresectable LM who might be candidates for somatostatin analogue treatment or hepatic artery embolization, preventive cholecystectomy was suggested to limit the risk of future cholecystolithiasis and ischaemic cholecystitis.¹⁰⁹

Liver resection

The evaluation of the effectiveness of the surgical approach is hindered by the limited number and heterogeneity of patients reviewed, as well as by the lack of uniform terminology for the completeness of hepatic surgery. The definition of complete and potentially curative resection ranges from: R0 resection,^{46,124,135–137} complete resection without information on R status,^138–141 removal of all visible gross disease,¹⁴² resections performed in combination with RFA^{121,143–145} and/or transarterial chemoembolization¹⁴³ or cryoablation¹⁴⁴ or results after RFA treatment.¹⁴⁴ For palliative surgical procedures determinants such as R1 or R2 resections, incomplete resections, resections with gross residual disease, cytoreductive surgery, partial cytoreductive surgery, palliative cytoreduction, debulking and debulking of at least 90% of the hepatic tumour mass are applied.

There is common consent that radical surgery including resection of the primary tumour and liver metastases may prolong survival and offer symptom control in patients with resectable neuroendocrine LM. Hepatic resection should not be attempted when only incomplete elimination of the metastatic disease is to be expected.^111,117 The only exception is in patients with incapacitating symptoms as a result of functioning LM. In these patients, debulking may improve symptoms and prolong survival.

Overall survival after hepatic resection has been reported in 46–86% at 5 years and 35–79% at 10 years (Table 2). These wide ranges likely reflect diverse patient selection and the heterogeneity of the underlying tumour biology. Median overall survival in patients in whom hepatic resection was feasible ranged from a median survival time that was not reached during a follow-up of 27 months¹³⁹ up to 78 months¹⁴⁵ compared with 27 months¹³⁹ and 17 months¹⁴⁵ in those with unresectable tumour burden. The decision for liver surgery is hampered by a reported complete resection (R0) rate of only 20–57%^{46,124,139,140,144,146} and a 5-year local recurrence rate of up to 97% even when complete resection has been achieved.^{91,124,137,138}

Table 2.

Results of liver resection for neuroendocrine liver metastases: a reviewof the recent series

Author	Year	Patients (n)		Extent of resection (n)		Overall survival after resection			Survival after curative resection
		Total	Resected	Curative (R0)	Palliative	3 years	5 years	10 years	5 years
Dousset141	1996	36	17	12	5	87% (2-y)	46%		NS
Chamberlain47	2000	85	34	15	19	83%	76%		85% OS
Grazi146	2000	28	19	16	3	NS	93% (4-y)	79%	NS
Pascher135	2000	41	26	13	13	85% (2-y)	76%		NS
Nave136	2001	31	31	10	21	NS	47%		86% OS
Jaeck121	2001	13	13	NS	NS	91%	68% (6-y)		NS
Yao140	2001	36	16	16	0	NS	70%		70% OS
Elias124	2003	112	47	25a,b	17		71%	35%	74% OS 66% DFS
Sarmiento138	2003	170	170	75	95	NS	61%	35%	NS of OS 24% DFS
Norton122	2003	16	16	16a	0	NS	82%		82% OS
Touzios143	2005	60	30	18a	12	NS	72%		NS
House145	2006	31	26	26a	0	84%	65%		65% OS
Musunuru144	2006	48	13	9a	4	83%	60%		NS
Mazzaferro169	2007	36	36	36	0	NS	85%	59%	85% OS 34% DFS
Gomez137	2007	18	18	15	3a	94% (2-y)	86%		90% DFS
Landry150	2008	54	23	7	16a	100%	75%		NS
Kianmanesh128	2008	23c	19	19	0	94% (2-y)	94%	79% (8-y)	94% 50% DFS
Frilling46	2009	119	27	23	4a	94%	94%	85%	100% OS 96%DFS

^aSeries in which resection has been performed in combination with additional treatment modalities such as RFA, TACE, cryotherapy, or yttrium-90 microspheres.

^bIncluded are also patients with complete resection of peritoneal carcinomatosis who underwent surgery followed by intraperitoneal chemotherapy.

^cOnly patients considered for 2-step surgery.

DFS, disease-free survival; NS, not stated; OS, overall survival.

The largest experience with hepatic surgery for neuroendocrine LM has been documented in the three series from the Mayo Clinic.^138,147,148 Of the 37 patients reported on in the initial series in 1990, 17 (46%) underwent curative resection, whereas in 20 patients gross residual tumour remained. In the curative group, 75% of the patients were disease-free at a median follow-up of 19 months. In 16 patients undergoing palliative resection of hormonally active tumours, complete relief of symptoms was obtained in eight patients with a median duration of response of 6 months. One patient died post-operatively. Six patients had incomplete relief of symptoms and in one patient the symptoms remained unchanged. The authors suggested that curative resection should be the mainstay of treatment for patients with tumours amenable to complete elimination. They stressed that refinement of radiological imaging is necessary in order to further reduce the number of patients undergoing surgery with pre-operatively understaged disease. Despite their inferior results with palliative resections, they recommended debulking for symptom relief, however, only if at least 90% of the metastatic mass can be eliminated safely.¹⁴⁷ In the subsequent report on the cohort of 74 resected patients, some of them overlapping with those considered in the initial series, curative resection was obtained in 28 (37.8 %). The post-operative complication rate was 24% and two deaths occurred (2.7%). Overall survival was 75% at 4 years, with no significant difference between the completely and incompletely resected patients. The median time to recurrence was 19.3 months and symptomatic response was achieved in 90% of patients.¹⁴⁸ In the last series encompassing 170 surgically treated patients, the group attempted to define the role of liver resection as the benchmark against increasingly applied non-surgical treatment options.¹³⁸ The 5-year overall survival was 61% with a median survival of 81 months in the entire study population. Morbidity was documented in 14% and two patients died post-operatively. In total 84% and 94% of the patients developed tumour recurrence at 5 years and 10 years, respectively, with a median time to recurrence of 21 months. In the 44% of patients with complete tumour resection, the 5-year recurrence was 76% with a median time to recurrence of 30 months, compared with 91% and 16 months in the incompletely resected group. Owing to the fact that virtually all patients developed disease recurrence, the authors considered hepatic surgery for neuroendocrine LM as unlikely to be curative, however, they consider it of benefit with regard to survival and symptom relief.^109,138 These discouraging results question the effectiveness of surgery and call for consideration of adjunct therapies. The role of adjuvant treatment after R0 resection is poorly assessed. Adjuvant therapy with streptozotocin and 5-fluorouracil failed to improve disease-free survival in patients with digestive NET.¹⁴⁹

In the report of Elias et al., 36 out of 47 (77%) patients required extrahepatic tumour resection at the time of hepatectomy.¹²⁴ The respective morbidity and mortality rates of 45% and 5% in this series most likely reflected the extent of the procedures. Twelve patients with hepatectomy and complete resection of peritoneal carcinomatosis were offered immediate additional intra-peritoneal chemotherapy. While at the time of the report the authors considered this as a possible curative approach, they revised their opinion in a later report, as only 10% of these patients remained disease-free at 5 years after surgery.¹²⁴

Pascher et al. reviewed 41 patients with LM (40.2%) from a cohort of 102 NET patients seen during a 10-year period.¹³⁵ In 90% of cases, LM were multifocal, and in 75%, additional extrahepatic disease was evident. Hepatic surgery marginally prolonged median survival to 60 months in the resected patients compared with 46.5 months in the non-resected group. The prognostic relevance of R0 resection has been pointed out by Gomez et al.¹³⁷ In their group of 18 resected patients, the overall 5-year recurrence rate was 34%. Five-year recurrence was only 10% in patients with tumour-free resection margins, in contrast to 75% when resection margins were involved.

Effectiveness of surgical resection in comparison to hepatic artery embolization and medical therapy was first studied by Chamberlain et al. in a group of 85 patients with additional extrahepatic disease in 38 of these patients.⁴⁷ Three- and five-year survival rates for the entire cohort were 61% and 53%, respectively. Analysis of the three treatment groups yielded 3- and 5-year survivals for patients with surgery, hepatic artery embolization or medical therapy of 83%, 83% and 39% and 76%, 50% and not available, respectively. Curative resection was associated with prolonged survival; however, surgery was beneficial only in patients with less than 75% tumorous liver involvement. In the series of Yao et al. focusing on hepatic resection vs. chemoembolization, the 5-year survival for the resected patients was 70% whereas it dropped to 40% when chemoembolization was performed.¹⁴⁰

Touzios et al. compared survival in 60 patients with either invasive [resection/ablation or transarterial chemoembolization (TACE) ± resection/ablation] or non-invasive (chemotherapy, radiation, octreotide) treatment strategies. They found a significantly better median and 5-year survival with invasive treatment (>96 months and 72% for resection/ablation, 50 months and 50% for TACE ± resection/ablation) then for non-invasive therapy (20 months and 25%, respectively).¹⁴³ However, invasive treatment was beneficial only for those patients with less than 50% liver involvement.

The preponderance of cytoreductive surgery over embolization for symptomatic neuroendocrine LM has been documented by Osborne et al.¹⁴² Of a total of 120 patients, those 61 patients who underwent surgery had a mean duration of symptom relief of 32 months and mean survival of 43 months, as compared with 22 and 24 months, respectively, in the group with embolization therapy. Based on their experience, these authors suggested hepatic cytoreduction rather than embolization, even if complete disease resection may not be feasible.

Other groups demonstrated benefits of surgical therapy, either complete or cytoreductive, in comparison with (i) hepatic artery embolization and medical treatment,¹⁴⁴ (ii) medical therapy,¹³³ and (iii) various liver-directed non-resectional interventions¹⁵⁰ or, in case of cytoreductive resection, as a component of a multimodality approach including hepatic artery embolization, RFA, chemotherapy and peptide receptor radionuclide therapy (PRRT).^150–153 Reduction of a CgA level of ≥80% was shown to reliably indicate post-operative symptom relief and stabilization of disease after cytoreductive surgery.¹⁵⁴ For prolongation of symptom-free survival, adjuvant treatment with octreotide long-acting release was recommended.¹⁵⁵

Liver transplantation

As a result of their less aggressive biological behaviour when compared with other secondaries, non-resectable neuroendocrine LM have been accepted as an indication for orthotopic liver transplantation (OLT). Since the first reports in 1988 by O'Grady et al.¹⁵⁶ and Iwatsuki et al.¹⁵⁷ on two and five transplanted patients, data on approximately 170 transplanted patients have been collected worldwide. Lack of stringent generally adopted pre-transplant selection criteria and of uniform follow-up protocols hamper critical assessment of the published results (Table 3).

Table 3.

Summary of results of liver transplantation for hepatic metastases of neuroendocrine tumours

Author	Year	Patients(n)	Overall survival			Disease free survival
			1-year	3-years	5-years	3-years	5-years
Makowka242	1989	5	60%	NS	NS	NS	NS
Curtiss247	1995	3	100%	NS	NS	NS	NS
Anthuber245	1996	4	25%	0	0	0	0
Pascher135	2000	4	100%	NS	50%	NS	NS
Ringe244	2001	5	74%	56%	NS	31%	NS
Rosenau166	2002	19	89%	89%	80%	35%	8%
Amarapurkar243	2003	14	64%	50%	14%	25%	12%
Fernandez246	2003	5	80%	80% (2-y)	NS	40% (2-y)	NS
Florman170	2004	11	73%	NS	36%	NS	NS
van Vilsteren165	2006	19	88%	NS	NS	80% (1-y)	NS
Frilling167	2006	15	78%	NS	67%	48% (2-y)	48%
Mazzaferro169	2007	24	NS	NS	90%	NS	77%
Marin241	2007	10	86%	57%	NS	38%	NS
Olausson168	2007	15a	NS	NS	90%	70% (1-y)	20%
Le Treut164	2008	85b	72%	59%	47%	37%	20%

In order to avoid data overlap, only the most recent report of an institution is considered.

^aIncludes five multivisceral transplantations.

^bMulticentric study.

NS, not stated.

In the initial series encompassing rather unselected patients, overall and recurrence-free 5-year survival of 36–57% and 24–45%, respectively, has been reached.^{141,158–162} Based on their experience with nine transplanted patients, Dousset et al. demonstrated for the first time that better patient selection, e.g. removal of the primary tumour prior to OLT, with the option of concomitant exclusion of extrahepatic tumour spread and carcinoid rather than non-carcinoid primaries, may contribute to better treatment results.¹⁴¹ Bearing in mind the immanent scarcity of donor organs and the potential risk of immunosuppression after transplantation for a oncological condition, the same group further recommended considering patients with rapid tumour growth unresponsive to alternative treatment options rather than those with quiescent disease as transplant candidates.¹⁶³ The first multicentric study from France encompassing 31 patients confirmed significantly better survival for carcinoid as opposed to non-carcinoid tumours (69% at 5 years vs. 8% at 4 years).¹⁶¹ In a more recent multicentric report from the same group analysing 85 patients, duodenopancreatic primary localization and extensive hepatic tumour load have been suggested as factors indicating poor prognosis.¹⁶⁴ While the 5-year survival rate was 68% in patients with limited hepatic disease and non-duodenopancreatic tumours, it dropped to 12% in the case of hepatomegaly and primary tumour localized within the duodenum or pancreas. This has been confirmed by other groups as well.¹⁶⁵

The prognostic relevance of Ki-67 and E-cadherin, a marker for cell proliferation and metastatic potential of tumours, respectively, has been demonstrated by Rosenau and co-workers.¹⁶⁶ In transplanted patients with Ki-67 ≥5% and/or aberrant E-cadherin, expression survival was significantly decreased compared with those with lowed expression of Ki-67 and regular E-cadherin staining results (7-year survival 0% vs. 100%, respectively). The importance of Ki-67 as a selection criterion, in addition to the diagnosis of extrahepatic tumour spread prior to transplantation by octreotide scintigraphy, has been pointed out by several investigators.^165,167,168 On the other hand, Olausson et al. achieved an excellent 5-year overall survival and 1-year recurrence-free survival of 70% despite expanding their criteria for transplantation towards an older age (64 years) and higher Ki-67 (10–15% in 6 of the 15 patients).¹⁶⁸ In cases of NET localized within the pancreatic head, the group followed a very aggressive approach and removed the primary tumour, lymph nodes and liver followed by multivisceral transplantation (liver, stomach, pancreas, duodenum and upper third of the small bowel or entire small bowel). The close post-transplant follow-up and the aggressive treatment approach in case of tumour recurrence (median time to recurrence of 1.9 years, recurrence-free survival of 20% at 5-years) resulted in a 5-year overall survival of 90%, despite the inclusion of some high-risk patients.

In our own experience in 17 patients transplanted for NET LM, Ki-67, SRS-based exclusion of non-resectable extrahepatic tumour manifestation and carcinoid heart disease proved to be a particularly powerful selection criteria.^46,167 The following criteria for the indication for transplantation were identified by Mazzaferro et al.: functioning or non-functioning low-grade NET, a primary tumour drained by the portal system and removed with a curative resection preceding transplantation, ≤50% metastatic involvement of the liver, good response or stable disease for a minimum of 6 months prior to transplantation and age ≤50 years.¹⁶⁹ After this policy they achieved a 90% overall survival and a 77% recurrence-free survival at 5 years. In general the indication for liver transplantation for metastatic NET has been considered either as a curative treatment option or as palliation for patients with intractable symptoms.^{59,165,167,168,170} The group from Milan, in contrast, demanded a 5-year recurrence-free survival of at least 50% as a benchmark to justify the intervention. Further developments are the recommendation of living donor liver transplantation for neuroendocrine tumour patients to spare the deceased donor pool for patients with benign indications.^{164,165,167,170,171}

A critical review of the data on surgical treatment of neuroendocrine LM is burdened by the small number of patients eligible, the different prognoses of pancreatic and mid-gut tumours and the lack of data from prospective studies.

Locally Ablative Techniques

Radiofrequency ablation

Radiofrequency ablation can be used as an image-guided percutaneous procedure, laparoscopically or through an open surgical approach, either as a single procedure or in combination with other techniques.^172,173 The method is amenable to repeated applications. Evidence exists that laparoscopic and open techniques provide superior local tumour control compared with percutaneous approaches, regardless of the size of the lesion.^174,175 In the largest series on RFA in neuroendocrine LM, Mazzaglia et al. treated 452 liver lesions laparoscopically in 63 patients.¹⁷⁶ The mean number of lesions treated at the first RFA session was six (range, 1–16) and mean tumour size was 2.3 cm (range, 0.5–10 cm). The procedure-associated morbidity was 5% and there was no 30-day mortality. Median survival was 3.9 years calculated from the first RFA session. On multivariate analysis, gender and the size of the dominant LM were predictive of survival. Median survival was most favourable in patients with a diameter of the dominant liver lesions of <3 cm. The lowest local recurrence among different primary and secondary liver tumours was achieved in neuroendocrine LM with <3 cm and at least a 1-cm circumferential post-ablation margin.¹⁷⁷ In a separate study, the Cleveland group showed that RFA is associated with an increased risk of liver abscess formation in patients with a previous Whipple procedure (40% vs. 0.4%) as a result of the colonization of the bile duct system. This finding might be important for surgical candidates with pancreatic NET.¹⁷⁸

Hellman et al. performed RFA percutaneously or intra-operatively in 21 patients with endocrine tumours and 43 LM.¹⁵³ At a mean follow-up of 2.1 years (range, 3 months–4 years), complete tumour ablation was documented in 96% of the tumours. Post-interventional bile leakage and pleural effusion accompanied by fever were seen in one patient each. In a group of 25 patients with 189 neuroendocrine LM, Gillams et al. performed 30 and 36 percutaneous treatment sessions with a solid-state laser and RFA, respectively.¹⁷⁹ At a median follow-up of 21 months (range, 4–75 months) local tumour control and symptom relief were evident in 75% and 69% of patients, respectively. Wessels and Schell successfully applied RFA as salvage treatment for carcinoid hepatic metastases refractory to hepatic artery embolization.¹⁸⁰

Other locally ablative measures, such as laser therapy,¹⁸¹ cryotherapy^182–184 and ethanol injection,^173,185 successfully achieved symptom control and positive tumour response in smaller case studies. However, they were gradually replaced by RFA in most centres because of better treatment results. Using cryotherapy in 13 patients with 52 neuroendocrine LM, Seifert et al. documented a treatment response in all seven symptomatic patients.¹⁸⁶ Of note, however, was a morbidity rate of 31% encompassing bleeding caused by coagulopathy, acute renal failure and pulmonary embolism. The well-recognized problem of an increased risk of haemorrhage related to large-diameter probes might be resolved by the newly developed small-caliber probes.¹¹³ Ethanol injection under ultrasound guidance has been shown to be useful in selected patients with smaller lesions adjacent to biliary or vascular structures vulnerable to the heat-sink effect.^173,187 Recently introduced microwave coagulation therapy utilizing electromagnetic microwaves, which agitate water molecules in the surrounding tumour tissue causing cellular death through coagulation necrosis, may be superior to RFA in cases when ablation of larger and/or multifocal tumours is attempted.¹⁸⁸

Percutaneous liver-directed techniques

Hepatic transcatheter arterial embolization and transcatheter arterial chemoembolization

Induction of a selective ischaemic necrosis after surgical ligation or transient occlusion of the hepatic artery accounts for the initial vascular occlusive therapy for palliation of symptomatic patients with metastatic NET.¹⁸⁹ Utilizing this concept hepatic transcatheter arterial embolization (TAE)^190,191 and transcatheter arterial chemoembolization (TACE)¹⁹² have been consecutively introduced. Although both techniques have been widely adopted, it remains debatable if the addition of cytotoxic drugs to embolization material increases the effectiveness of bland embolization alone, particularly when performed selectively.^{113,172,193–197} In a multicentric retrospective review on 100 patients treated with either TAE or TACE, Pitt et al. revealed no difference between the two techniques with regard to median overall survival, symptom improvement, morbidity and mortality. Controversies exist regarding the extent of liver volume that should be embolized during a single treatment session, the effectiveness of different chemotherapeutic regimens, the timing of the procedure during the course of the disease (‘early’ vs. ‘late’ embolization) and the timing of subsequent embolization procedures. A post-embolization syndrome, i.e. transient fever, abdominal pain and elevation of liver enzymes, is frequently seen. Portal vein thrombosis and hepatic insufficiency are considered exclusion criteria for both TAE and TACE.^198,199

In one of the largest single-centre studies including 69 patients with carcinoid tumours and 54 patients with pancreatic neuroendocrine carcinoma, Gupta et al. identified prognostic factors for progression-free survival after TAE or TACE.²⁰⁰ Comparing the results of patients with carcinoid tumours vs. those with pancreatic NET, patients with carcinoid tumours had a longer overall cumulative survival at 2 and 5 years (68.6% and 28.6% vs. 48.7% and 13.7%, respectively), better radiological response rate (66.7% vs. 35.2%) and longer median progression-free survival (22.7 months vs. 16.1 months). Male gender was the only independent risk factor for unfavourable survival in carcinoid patients. A persistent primary tumour, ≥75% liver involvement and bone metastases were predictive for poor survival in patients with pancreatic NET. While in the carcinoid group, TAE was associated with a significantly higher response rate than TACE, no such difference was obvious for pancreatic NET. Post-embolization syndrome occurred in nearly all cases and major complications were documented in 19 patients, of them most with large-volume liver involvement. Interestingly, the difference between the patients with carcinoids and those with islet cell tumours after treatment with TAE or TACE was not confirmed by Ho et al.²⁰¹ Among the 122 patients with metastasized carcinoids treated with TACE by Bloomston et al. peri-interventional morbidity and mortality were seen in 23% and 5%, respectively.²⁰² A radiographic tumour response was documented in 82%, accompanied with disease stabilization in 12%. Symptom palliation with a median duration of 13 months was achieved in 92% of the patients. In a most recent review on interventional treatment of neuroendocrine LM, Vogl et al. recommended TAE or TACE for patients with multifocal disease and higher tumour burden, whereas surgical resection or local ablative are preferred techniques for those with single metastatic deposits.²⁰³

Radioembolization

Selective internal radiation therapy (SIRT), combining the effect of interstitial high-dose radiation and embolization of malignant microvasculature with micro spheres, has evolved as a promising treatment option for hepatic malignancies, among them also neuroendocrine metastases.^204–206 In a retrospective multicentric study on 148 patients with LM of various NET treated with yttrium-90 (90Y) labelled resin microspheres, a complete or partial response on imaging studies was seen in 2.7% and 60.5% of patients, respectively. Stable or progressive disease was documented in 22.7% and 4.9%, respectively. The median survival time was 70 months. The authors concluded that SIRT has the potential to deliver high doses of radiation to neuroendocrine LM, yielding a treatment effect and safety profile comparable to other local techniques. King et al. combined SIRT with a 7-day systemic infusion of 5-fluorouracil (5-FU) in 34 patients with unresectable neuroendocrine LM, of them 59% with additional extrahepatic disease. They achieved a radiological response in 50% of cases, with 18% and 32% showing a complete or partial response, respectively.²⁰⁷ A symptomatic response accompanied by a decrease of the CgA serum concentration was seen in 55% and 50% of the patients at 3 and 6 months, respectively. The mean survival was 27.6 ± 2.3 months. The number of treatment-related side-effects was low. As an alternative to SIRT, selective hepatic arterial administration of radionuclide-labelled somatostatin analogues were shown to provide radiological tumour response and symptomatic relief in single patients with large-volume somatostatin receptor positive LM.^208,209

The time line for palliative non-surgical strategies, i.e. before or after systemic therapy remains unclear. Treatment strategies have to be individualized according to experience, patient preference and quality of life considerations.

Systemic treatment options

Peptide receptor radionuclide therapy

The peptide receptor radionuclide therapy (PRRT), a somatostatin-analogue-based targeted therapy, is an emerging new field in the palliative treatment of NET with very encouraging data. The treatment modality utilizes radioactive substances that are conjugated with various somatostatin analogues such as octreotide or lanreotide. Intravenously administrated, these drugs bind to the somatostatin receptor localized on the tumour cells. The ligand–receptor complex is internalized and the attached radiation has the potential to destroy the tumour cells.^107,210 The side-effects are few and mostly mild, in particularly when treatment protocols consider renal protective agents. Initially used 111In-DTPA-[D-Phe1]-octreotide achieved partial remission rates up to 8%, whereas stable disease was seen in 42% to 81% of patients.^{101,211–213} Despite these positive results, it has been recognized early that because of the limited tissue penetration range of 111In, 111In-coupled peptides are rather ineffective for PRRT and their utilization should be restricted for therapeutic purposes.

More suitable somatostatin analogues such as DOTATOC and [DOTA,Tyr3]octreotate (DOTATATE)] labelled with 90Y or lutetium-177 (177Lu) achieved better results in terms of response and imaging-assessable tumour mass reduction. Another approach, using the more lipophilic somatostatin analogue lanreotide (90Y-DOTA-lanreotide, DOTALAN) induced disease regression in 14% and stable tumour disease in 41% of patients.²¹⁴ In various studies with 90Y-DOTATOC in metastasized intestinal NET using different protocols, the overall response rate varied between 24% and 33%.^215–218 In our own experience with 90Y- and 177Lu-DOTATOC treatment in 20 patients with metastasized NET, all with LM, partial remission or stable disease were achieved in 5 and 11 patients, respectively, while tumour progression was documented in four patients.¹⁵² In a recent study from Rotterdam on 177Lu-DOTATATE in 310 patients with intestinal NET, tumour response of various degrees was found in 46%, stable disease was seen in 35% and progressive disease in 20%.²¹⁹

Chemotherapy

The efficacy of systemic cytotoxic therapy, traditionally based on streptozotocin in combination with 5-FU or doxorubicin, in well-differentiated mid-gut NET is low, yielding response rates of not more than 10–15%.^122–124 In contrast, pancreatic NET trials of streptozotocin plus 5-FU/doxorubicin or dacarbazine as a mono-agent demonstrated objective tumour response rates of 39% and 33%, respectively, and an improved overall survival.^220,221 Using temozolomide and thalidomide Kulke et al. achieved a response rate of 45% in pancreatic NET, but only 7% in mid-gut NET.²²² Patients with poorly differentiated NET, regardless of the primary tumour localization, are candidates for cisplatin and etoposide treatment. Response rates of 55–80% are reported, with a median duration of response of 8–11 months.²²³

Biotherapy and antiproloferative therapy

Biotherapy with somatostatin analogues or alpha interferon and, more recently, antiproliferative therapy with tyrosine kinase inhibitors, anti-angiogenic substances and mammalian target of rapamycin (mTOR) inhibitors, have been implemented as systemic treatment in NET.^224–226 Somatostatin analogues, either as single agents or rarely in combination with interferon, effectively relieved symptoms of functional tumour disease.^227–231 Recently, the PROMID study comparing somatostatin analogue octreotide long-acting repeatable (LAR) vs. placebo in patients with well-differentiated mid-gut NET reported a significantly increased median time to progression (14.4 vs. 6 months, respectively).²³² The ability of NET to express vascular endothelial growth factor (VEGF) and its receptor (VEGFR) make them suitable candidates for novel therapeutic strategies based on anti-angiogenic agents.²³³ A recent trial of sunitinib, a tyrosine kinase inhibitor with activity against various receptors, among them VEGFRs, reported a 16.7% overall response rate and 68% stable disease in pancreatic NET patients compared with 2.4% and 83% for non-pancreatic NET.²³⁴ In a phase II study, in which 44 patients on stable doses of octreotide were treated either with VEGF antibody bevacizumab or pegylated interferon alpha-2b, bevacizumab treatment was associated with an objective tumour response, reduction of tumour blood flow and longer progression-free survival when compared with the alternative treatment arm.²³⁵ mTOR is a serine/threonine kinase involved in cell cycle, cell proliferation and angiogenesis. Its inhibitor, RAD001 (everolimus) alone or in combination with octreotide LAR has generated a partial response in 22% of patients with low-to intermediate-grade NET with stabilization of the disease in 70% of patients.²³⁶

Follow-up

Imaging studies with ultrasound, CT, MRI, SRS, in addition to repeated determination of the tumour marker CgA, are used for the follow-up of patients with neuroendocrine LM. In functioning tumours' cases, tumour-specific hormones are used. For poorly differentiated NET, neurospecific enolase serves as an additional useful marker. The time frame for follow-up is individualized, depending on the different tumour subtypes, the grade of differentiation and the treatment performed. Patients with G1 and G2 tumours should be followed every 6–12 months and 6 months, respectively, whereas for G3 tumours 3-months intervals are recommended.²³⁷

Conclusions

The number of neuroendocrine tumours seen in a clinical practice is continuously increasing. An as yet unanswered question is whether this represents a true increase of biological incidence or is an increase because of our growing percipience of the disease. Hepatic metastases, which frequently occur in NET patients, significantly worsen their prognosis. While little progress in the treatment was achieved in the past, new specific imaging techniques with increasing sensitivity enabling tumour detection at an early stage and promising novel therapeutic options offering targeted approaches are under evaluation. Management in centres of expertise should be strongly encouraged in order to appropriately assess and classify the tumour burden and provide personalized multimodal treatment.²³⁸

Conflict of interest

None declared.