Non-colorectal hepatic metastases

Pathophysiology and molecular basis of liver metastases

Achieving cancer cure requires the complete eradication of all tumour cells. Thus, for most solid tumours, complete surgical excision is the cornerstone of treatment, often with adjuvant systemic treatment to treat microscopic disease. In the presence of metastases, there is an apparent contradiction in using a local therapy, i.e. surgery, to treat what is considered disseminated disease.

The rationale behind a surgical approach to metastatic disease is based on the concept of site-specific metastases. First proposed by Paget in 1889, this ‘seed and soil’ hypothesis argues that solid tumours have a distinct pattern of distant organ involvement created by the target organ microenvironment. Ewing proposed a ‘mechanical’ theory in which the metastatic pattern is determined by the venous drainage of the primary tumour. Neither theory takes into account the complexity of the metastatic process, which requires that a cancer cell gains specific invasion and metastatic potential before it can disseminate. The clonal selection model of the metastatic process suggests that heterogeneity develops within a population of cancer cells through mutational events, allowing a subpopulation to randomly acquire the necessary traits to disseminate successfully. Alternatively, it has been argued that within cancers of the same pathological type, i.e. breast cancer, some tumours are a priori more likely to develop metastases than others. This is supported by gene expression data where specific molecular signatures have been found to accurately predict prognosis in breast cancer, ovarian cancer and melanoma. Similarly, in CRC the genotype of microsatellite instability correlates with a decreased likelihood of metastatic spread.

A recent refinement to Paget’s hypothesis, based on molecular genetic research, suggests that the primary tumour is itself capable of preparing the ‘soil’ by creating a ‘premetastatic niche’. Every cancer has a type-specific pattern of cytokine expression that appears to direct both malignant and non-malignant cells to specific distant organs. The influx and clustering of bone marrow-derived haematopoietic cells is one of the earliest events in the development of a metastatic deposit. This is closely followed by local inflammation and the release of matrix metalloproteinases. These local events appear to mediate remodelling of the extracellular matrix, creating a more permissive microenvironment for the eventual deposition and growth of malignant cells. Thus, the primary tumour both chooses and alters the sites to which it metastasises. For reasons not yet understood, many solid tumours preferentially metastasise to the liver.

If the site-specific hypothesis of metastatic spread is correct, complete surgical excision of liver metastases can remove the only site of disease and offer a chance for cure. Nonetheless, residual micrometastatic disease may exist within the liver, and hepatic recurrences are a common cause of treatment failure following hepatectomy. Even in the presence of micrometastases, the removal of all macroscopic disease may have immunological benefits. The immune-suppressing effects of cancers are well accepted: malignant cells can induce both adaptive and innate immune suppression, facilitating tumour growth. The degree of immune suppression correlates with the tumour burden and if all gross metastatic disease can be removed, host defences may attack micrometastatic deposits more effectively. The use of neoadjuvant or adjuvant chemotherapy may improve cure rates by controlling micrometastases.

The advent of next-generation sequencing technologies and high-density oligonucleotide arrays has further deepened our understanding of the metastatic process. Whereas the ability of a cancerous cell to metastasise was once believed to occur following the accumulation of multiple somatic mutations in many cancer-causing genes, new findings, specifically in pancreatic cancer, have challenged this belief. Studies by Yachida et al. and Campbell et al. describe the existence of multiple subclones within a primary pancreatic cancer, each containing a unique genetic signature corresponding to an eventual site of metastatic spread. These subclones are present many years before an eventual metastasis is clinically detected, when disease is at an early stage. Furthermore, metastases seen in different organs share many common genetic mutations as well as site-specific changes that confer a selective growth advantage in the respective tissue. Alternatively, Notta et al. in 2016 have challenged the timing of tumour evolution of pancreatic cancer whereby a ‘single cataclysmic’ event, such as chromothripsis, spawns a highly metastatic clone capable of seeding multiple organs very rapidly.

Studies investigating single-nucleotide polymorphisms in genes associated with tumour dormancy and immune response checkpoints have associated certain mutations with survival outcomes in patients undergoing resection of CRLM. ^, These findings could assist oncology teams in better risk-stratifying patients for surgical resection and theoretically similar studies could be performed on patients with non-CRLM, though at present there are few validated data available. Future studies on the biology of metastases are likely to improve our understanding of this complex process, translating into more effective therapy.

Clinical approach to non-colorectal liver metastases

Routine clinical, radiological and serological assessments for liver metastases should be guided by the propensity for liver metastases of each specific tumour type and the ability of potential treatments to alter the outcome of the metastatic disease. In imaging the liver, the choice of transabdominal ultrasound (US), contrast-enhanced ultrasound (CEUS), contrast-enhanced triphasic computed tomography (CT), magnetic resonance imaging (MRI) and positron emission tomography (PET) will be dictated by tumour type as well as local availability and expertise.

Functional imaging uses radiolabelled somatostatin analogues, such as indium pentetreotide scintigraphy (Octreoscan), and can detect NETs expressing somatostatin receptors with 80–90% sensitivity. Additionally, whole-body PET using a somatostatin analogue, Gallium (DOTA-TOC, DOTATATE or DOTANOC), has been found to be accurate for the detection of new metastases in NETs. Occasionally, the initial presentation of a NET will be a liver metastasis from an unidentified primary. Biopsy will demonstrate NET and the investigative focus will be aimed at localisation of the primary tumour. When the primary site remains unknown despite imaging and endoscopy, extended immunohistochemistry staining for neuroendocrine markers or gene expression classifiers can distinguish small bowel from pancreatic NETs. Positive CDX2 staining suggests a midgut primary, whereas positive PAX6 or ISL1 staining suggests a pancreatic primaryA. Some patients can be assessed for recurrence of non-CRLM using more targeted techniques and biochemical markers (e.g. CA-125 for epithelial ovarian cancer, chromogranin A, pancreastatin, neurokinin A, polypeptide, substance P, and neuron-specific enolase for NETs).

Certain tumours, such as gastric, breast and ovarian cancer, have a predilection for intraperitoneal spread. Although CT is the preferred modality for diagnosing peritoneal carcinomatosis, its accuracy is still limited. For many of these equivocal cases, diagnostic laparoscopy has been recommended and found to result in a change in management in 20% of cases and may be used selectively in preoperative staging.

While the discussion that follows will review management and outcomes of liver resection for hepatic metastases based on primary tumour type, general considerations as proposed by Adam et al. in 2006 remain relevant. In their review of 1452 patients from 41 centres undergoing liver resection for non-colorectal and non-neuroendocrine primary tumours, age > 60 years, presence of extrahepatic disease, R2 resection and major hepatectomy were associated with decreased OS. These factors should be considered when contemplating liver resection for these patients.

Treatment strategies

Several treatment modalities exist for metastatic disease, and the therapeutic approach must be tailored to the tumour type, the performance status of the patient and the extent of disease. Treatment decisions in this context can only be determined properly by a multidisciplinary oncology team. Non-surgical ablative strategies and systemic or locally delivered chemotherapy can be used as adjuncts to resection.

RFA has been reported to be safe and successful at achieving local control in patients with liver metastases from breast cancer, ovarian cancer and NETs. The major limitation of RFA is the difficulty in achieving complete necrosis for tumours > 3 cm, as well as limited utility when tumours are close to major vascular or biliary structures. Novel techniques such as microwave ablation (MWA), stereotactic radiotherapy and irreversible electroporation (IRE) need to be assessed in prospective randomised trials as they may supplant surgical resection in some cases.

Transarterial embolisation (TAE) takes advantage of the differential blood supply of liver metastases, which depends mainly on the hepatic arteries, and the normal parenchyma, which relies more heavily on the portal vein. Transarterial chemoembolisation (TACE) involves the local delivery of a drug prior to occluding the artery and allows prolonged exposure of the tumour to the agent without increasing systemic toxicity. Both TAE and TACE have been well described for the treatment of unresectable hepatocellular carcinoma and the symptomatic relief of NETs.

Finally, it behooves the surgical oncologist to keep abreast of the explosion of recent advances in systemic therapy, such as immunotherapy, to ensure proper selection of patients who will benefit from resection.

Neuroendocrine tumours

NETs represent a diverse group of tumours originating throughout the GI tract from cells of the neuroendocrine system. They are classified by site of origin and include GI and pancreatic histological subtypes. NETs arise most commonly in the midgut and produce neurosecretory granules, express characteristic neuroendocrine differentiation markers, and may secrete vasoactive substances responsible for carcinoid syndrome.

About 30% of the patients with NETs present with metastatic disease, preferentially to the liver, and in many patients the liver remains the only site of metastatic disease for a prolonged period of time. ^, The majority of patients have multifocal, bilobar disease, of which less than 20% are candidates for surgery. Additionally, these patients have worse prognosis than patients with isolated locoregional disease. ^{, ,} ( Fig. 8.1 ) Depending on the primary site, extent of the disease, and tumour characteristics, different treatment modalities such as liver resection, non-surgical liver-directed therapies, and systemic therapies are available. Liver resection may be performed with curative intent, symptom control or prolongation of survival in the palliative setting.

The choice of treatment for NET hepatic metastases is largely dependent on underlying tumour biology and pattern of metastatic spread. In general, resection or cytoreduction is associated with improved survival, offers relief from hormonal symptoms and prevents sequelae of carcinoid syndrome. Resection has a clear advantage over medical treatment in regards to symptom relief and OS in patients that can be optimally cytoreduced. In 2009, Frilling et al. classified the metastatic pattern of spread in the liver for NETs into three morphological subtypes : (I) ‘restricted metastases’ involving one lobe or two adjacent segments; (II) ‘dominant lesion with bilobar metastases’ whereby a single major focus is accompanied by multiple contralateral satellite lesions; (III) diffuse, multifocal liver metastases affecting multiple segments within and between lobes. These three groups differ significantly in terms of treatment strategies and clinical outcomes. Hence, patients with type I or II disease (25% and 15% of cases, respectively), in the absence of extrahepatic metastases, can be considered for curative surgical resection. The aim of liver resection with curative intent in NETs is to leave no residual disease (R0 resection) in both primary and secondary sites, and this may be associated with 5-year survival rates of up to 85%.

Hepatic resection for metastatic NETs results in improved OS compared with those receiving supportive care. There is still a debate regarding the resection margin. Some series show that there is improved progression-free survival (PFS) in patients who underwent complete resection. Sarmiento et al. showed that R0 resection was associated with longer PFS than R1/R2 resections (median PFS 30 months vs 16 months, P < 0.001). However, few other studies have shown that margin status is not associated with risk of progression. ^{, ,} Furthermore, R1 and R2 resections result in 5-year survival rates of 70% and 60%, respectively, challenging the dogma that surgery should be reserved only for patients most likely to have an R0 resection. ^, The recommendation currently is to pursue surgical cytoreduction for both functional and non-functional tumours even when complete cytoreduction cannot be obtained.

Historically, cytoreduction aimed to reduce tumour volume by at least 90% in situations where R0 resection was not feasible. Although there are no data from randomised trials, large series using historical controls or contemporary cases matched for stage have demonstrated that liver resection with optimal cytoreduction results in improved survival. Recent data suggest that cytoreduction of 70% or greater of tumour burden effectively palliated symptoms and was associated with improved PFS. ^{, ,} Therefore, resection should be considered when a 70% debulking threshold is possible including in patients with multiple NET liver metastases, Ki-67 less than 20% and even in patients with extrahepatic disease as long as this threshold is achieved. ^, Importantly, patients with extensive liver replacement with tumour burden > 50–70% or high-grade NET liver metastases may not benefit from surgery. Additionally, patients with poor performance status and significant comorbidities, severe hepatic insufficiency or carcinoid heart disease may also not benefit from surgical intervention.

Cytoreduction can also offer effective and durable palliation from symptoms for patients with functional tumour syndromes. As a result, surgical debulking has been advocated for both functional and non-functional tumours. An aggressive approach, sometimes combining liver resection with other ablative strategies, is warranted ( Fig. 8.2 ). Liver transplantation is an emerging option for NET liver metastases that are otherwise unresectable but is controversial.

In regard to the question of whether or not to resect the primary tumour in patients with unresectable metastatic disease, guidelines recommend evaluating the patients’ symptoms, medical condition, grade and location of the tumour, and potential to improve response to peptide receptor radionuclide therapy (PRRT).

Despite resection, hepatic recurrence occurs in up to 84% of patients at 5 years post-surgery. ^, Sarmiento et al. showed a 5-year recurrence rate of 76% even in patients who had R0 resection. This is partially due to underestimation of the tumour burden on preoperative imaging. Recurrence is suspected by the elevation of tumour markers such as 5-hydroxyindoleacetic acid (5-HIAA) and chromogranin A. Chromogranin A is more sensitive than 5-HIAA in identifying disease progression, and high levels have been shown to predict poorer outcomes. A reduction in chromogranin A levels of > 80% predicts a good outcome following cytoreductive hepatectomy, even when complete resection has not been achieved.

Non-surgical treatment modalities used for NET metastases include RFA, TAE and TACE. Ablation can be performed as an adjunct to resection and can minimize the loss of normal liver tissue and allows improved cytoreduction. RFA in isolation can achieve symptomatic relief and local control of variable duration in up to 80% of patients with NET hepatic metastases. Although studies comparing RFA with other modalities are limited, RFA has been advocated in patients with bilobar disease with up to 14 hepatic lesions of < 7 cm in diameter, involving up to 20% of liver volume. ^, Ablation seems to be effective for palliation of symptoms and may provide survival benefit with relatively low rate of complications. TAE and TACE are more useful options for improving hormonal symptoms and slowing tumour growth in patients with NET liver metastases. Both appear to deliver comparable results and thus one modality is not favoured over the other. Embolisation is usually indicated for more extensive hepatic disease or for tumours in close proximity to biliary structures precluding RFA. ^, Duration of response is typically short as the tumour rapidly develops collaterals and thus repeat treatments are often required. Embolisation is contraindicated in patients with 50–75% liver involvement due to the risk of precipitating acute hepatic failure. In general, aggressive multimodal therapy with embolic, ablative and systemic strategies is recommended to debulk or downstage metastatic NETs. Symptom control with non-surgical approaches, recognising that repeat treatments may be required, can limit the need for cytoreductive surgery in many patients.

Liver transplantation has been advocated for patients with extensive, unresectable liver metastases with no extrahepatic disease. The two largest reports from US and European registries reported only 150 cases in 20 years and 213 cases in 27 years, respectively. ^, The 5-year survival rates were relatively poor at 52% with a 5-year disease-free survival (DFS) of approximately 30% in both studies. Additionally, a systematic review found that recurrence after liver transplantation for NET liver metastases ranged from 31% to 57%. In 2016, Mazzaferro et al. conducted a retrospective, non-randomized study of patients with NET considered for liver transplantation from 1995 to 2010. Transplant eligibility included a primary tumour with curative resection of all extrahepatic disease along with criteria for the best transplant outcomes: low-grade (G1-2) histology, <50% tumour involvement of the liver, stable disease for more than 6 months and age up to 60 years. They reported significantly improved long-term outcomes in the transplantation group compared with controls, with 5-year survival rates of 97% versus 88% and 10-year survival rates of 51% versus 22%, respectively. Concerns remain regarding tumour recurrence in the context of immunosuppression, so while liver transplantation does appear to confer long-term survival in carefully selected patients, optimal patient selection criteria remain in evolution. ^,

NETs are graded based on mitotic rate and Ki-67 proliferative index on a scale from 1 to 3. Grade 1 and 2 tumours are considered to have indolent growth patterns. Despite this benign description, 46–93% of patients with NETs will have liver involvement at the time of diagnosis, with a 5-year untreated survival of 0–20%. Systemic chemotherapy with platinum-based regimens has shown a response rate of up to 67% in grade 3 NETs and is generally indicated for patients with high-grade NETs. However, overall the survival benefit of chemotherapy is limited and associated with significant toxicity and response rates of approximately 30%. ^,

Somatostatin analogues such as octreotide and lanreotide can achieve symptomatic relief in 70–80% of patients with functional tumours. Both have also been shown to confer improvement in PFS compared with placebo as well as stabilized tumour size. There is emerging evidence that pasireotide, a somatostatin analogue that binds more somatostatin receptors than octreotide or lanreotide, may have an even greater antiproliferative effect. Furthermore, molecularly targeted therapies, such as the vascular endothelial growth factor inhibitor sunitinib and the mammalian target of rapamycin (mTOR) inhibitor everolimus, have shown promise in patients with metastatic NETs. Furthermore, PRRT is a treatment option for patients who progress on somatostatin analogues.

Gastrointestinal stromal tumours

Liver metastases may be observed in approximately 16% of patients with retroperitoneal sarcomas and about 60% of patients with visceral sarcomas. Gastrointestinal stromal tumours (GISTs) are the most common GI mesenchymal malignancies and the most common source of hepatic metastases. GISTs originate from the interstitial cells of Cajal and approximately 70–80% of GISTs harbour a mutated c-Kit proto-oncogene, which results in the constitutive activation of the receptor tyrosine kinase leading to unregulated cell growth. Two-thirds of c-Kit mutations are located on exon 11. c-Kit exon 9 and platelet-derived growth factor receptor α (PDGFRA) mutations, encompassing a wild-type kinase domain that modulates receptor inhibitor sensitivity, account for another 5–10% of GISTs.

Primary GISTs represent 1% of all GI malignancies, and arise in the stomach (55–60%), small intestine (30–35%), colon/rectum (5–10%) and oesophagus (5%). ^, The primary tumour can be classified into four prognostic categories ranging from very low risk to high risk, according to site of the primary lesion, size of the primary lesion, and the number of mitotic figures identified on histology. Resection remains the standard treatment of primary GIST. For GIST metastatic to the liver, the therapeutic modalities include systemic therapies such as chemotherapy and targeted therapy, including tumour ablation, TAE and resection.

GIST metastatic to the liver is usually unresponsive to cytotoxic agents with response rates as low as 5%. However, targeted therapy directed against the oncogenic KIT and PDGFR tyrosine kinases has been shown to be very effective. Imatinib mesylate is a selective tyrosine kinase inhibitor (TKI) that has revolutionised the treatment of GIST. Despite complete surgical resection with microscopic negative margins, recurrence (local or distant) occurs in 50% of patients. The use of imatinib in the adjuvant setting was investigated in the phase III ACOSOG placebo-controlled trial (Z9001) for patients with resected GIST > 3 cm in size. A statistically significant 1-year recurrence-free survival (RFS) of 98% in the treatment group versus 83% in the placebo group was observed, prompting the inclusion of imatinib as an adjuvant treatment modality in patients with moderate- to high-risk primary tumours. Response to imatinib is greatest in tumours that harbour the c-Kit exon 11 mutation, with resistance rates higher in patients harbouring exon 9 or PDGFRA mutations. Demetri et al. showed that the median OS was 57 months with imatinib in a study population of which 95% had liver metastasis. Importantly, approximately 18% had primary resistance to imatinib as well as secondary resistance that developed at a median of 2 years due to development of secondary mutations. ^, Second- (e.g. sunitinib) and third-line agents (e.g. nilotinib and masitinib) have shown promise in patients resistant to imatinib.

Percutaneous ablation and intra-arterial therapies are used for palliative purposes or combined with resection. There are very scarce data regarding these modalities for the treatment of liver metastasis from GIST.

In the era of imatinib as an effective systemic therapy for GIST, the primary goal is to remove macroscopic disease. The efficacy and low side-effect profile of imatinib prompted initial enthusiasm for the combined use of surgery and imatinib in the management of metastatic GIST. Additionally, resection is sometimes performed in order to remove disease that is progressing while leaving macroscopic disease that is responding to therapy with imatinib. Overall, surgery is recommended after 6–9 months from the initiation of imatinib for maximal treatment response and is usually appropriate for patients with response to preoperative imatinib treatment.

Although evidence guiding surgical management in metastatic GIST is limited, a single-institution randomized controlled trial combining neoadjuvant imatinib with surgery and adjuvant imatinib in patients with previous R0 resection of the primary tumour has shown favourable 3-year OS (90% vs 60%). A recent Dutch study evaluated 48 patients who underwent liver resection for GIST metastases, 36 of whom received TKI therapy either pre- or postoperatively. Median survival was 7.5 years and 5-year OS was 76%. Multivariate analysis demonstrated that R0 resection of the liver metastasis was the only significant predictor of survival. A multicentre European study retrospectively evaluated 239 patients undergoing hepatic metastasectomy, all of whom received adjuvant imatinib. R0/R1 resection was found to be a significant predictor of survival and median survival in this group was 8.7 years. Other significant predictors of survival were female sex and metastases confined to the liver (compared to patients with liver and peritoneal disease). Metastatic disease to the liver, peritoneum and/or other sites had a median time to progression of 3 months following surgery.

Disease progression is managed by imatinib dose escalation followed by second- and third-line agents. In the event of tumour rupture or haemorrhage, surgery or hepatic artery embolisation may be performed in an emergency setting. Imatinib therapy for 6–12 months is recommended for patients with unresectable hepatic metastases and if the tumour responds, resection can be considered if an R0 resection can be anticipated.