The goals of treatment determine the role of systemic agents. Treatment goals for patients with mCRC can be classified as: (1) curative (sometimes referred to as “resectable” or “operable”); (2) potentially curative; (3) non-curative with active treatment intent (most patients fall into this group); and (4) non-curative with palliative intent (best supportive care)

As previously described, the liver is the most common site of hematogenous metastases for gastrointestinal tumors colorectal liver metastases (CLMs). Other common sites are lungs, peritoneum, lymph nodes, bones, and the central nervous system.

According to the surgical literature, up to 25% of patients diagnosed with CRC are found to have synchronous CLMs[5,6]. Up to 50% of patients without CLM at presentation will develop CLM during their lifetime[7]. In about one third of patients with mCRC, metastatic disease appears confined to the liver; about one third of these patients are deemed resectable by expert liver surgeons. Of those who get surgery, about 25% are cured (live for 10 years and do not have recurrence) while 10% are long-term survivors up to 5 years post-resection with recurrence[8,9].

As a result of the success of modern surgery (staged resections, newer vascular reconstruction techniques, assistance from complex hepatic interventional radiology), the terms resectable and unresectable will probably become outdated as we move towards a clinical decision of who will benefit from surgery (curative intent) and who will not (uncurable). Cure is most commonly defined in the literature as survival of 10 years with no recurrence.

There are three surgical approaches around which systemic treatment will be designed. The classic approach is resection of the colorectal primary, followed by adjuvant chemotherapy, then CLM resection. However, patients can decompensate between surgeries and may miss the opportunity for CLM resection. More commonly used today, the combined approach, includes same-session primary and CLM resection. Finally, the reverse approach places CLM resection before colon tumor resection. Brouquet et al[10] showed no significant difference in survival among these three groups. The combined approach has a higher risk of postoperative complications[11-13].

What happens when neoadjuvant chemotherapy is too successful? The problem of disappearing metastases becomes a big issue in patients who are potentially curable but receiving neoadjuvant therapy. The incidence of disappearing metastases (complete radiological response) can be as high as 38%[14-19]. Unfortunately, the natural history of patients with disappearing metastases is unknown. What is known is that microscopic disease is expected in up to 80% of patients with disappearing metastases[14,16-19] Placement of a fiduciary marker by an interventional oncologist should be considered for patients with metastases at risk of disappearing.

Patients who are resectable will often receive approximately 3 mo of neoadjuvant and 3 mo of adjuvant chemotherapy under the combined surgical approach, for example. Although the number of combinations is vast (Table 1), FOLFOX-4 (folinic acid, fluorouracil, oxaliplatin) is a standard first-line treatment with three recent landmark studies that constitute the frontier for neoadjuvant chemotherapy in patients treated with curative intent.

The most recent landmark study for perioperative chemotherapy was published in 2013. EORTC 40983 (funded in part by Sanofi-Aventis, the makers of oxaliplatin) was a phase II study that asked the question “In patients with resectable CLM, does neoadjuvant plus adjuvant FOLFOX4 improve OS at 8.5 years follow-up” The answer to this was no. However, the FOLFOX4 group did show improved progression-free survival (PFS) in the intergroup trial short-term data from EORTC 40983[20].

Can neoadjuvant chemotherapy facilitate resection? The concept of anti-epidermal growth factor receptor (EGFR) agents combined with chemotherapy was studied in the CELIM phase II study (funded in part by Merck-Serono, the makers of cetuximab) which asked “In patients with resectable CLM, does neoadjuvant cetuximab plus FOLFOX/FOLFIRI increase resectability compared to historic controls?” The answer was yes; resectability rates increased from 32% (22/68 patients) to 60% (41/68) after chemotherapy (P < 0.0001). Cetuximab is a monoclonal antibody that targets EGFR[21].

Moreover, OPUS, a phase II study (funded in part by Merck-Serono, the makers of cetuximab) showed that adding cetuximab to FOLFOX in chemotherapy-naïve patients increases the response rate and time to cancer progression compared to chemotherapy alone. Only patients who were KRAS wild type were found to benefit from cetuximab, but this is expected because KRAS mutant cells send growth signals independent of EGFR activation[22].

Response rates to modern chemotherapy have increased up to 60%-70%, and initially unresectable patients who are closely followed by their specialists during chemotherapy may become surgical candidates.

The concepts of down-sizing and down-staging” follow from the above discussion; if surgery is currently our only modality that is shown to be potentially curative, then we need to work to convert unresectable patients to surgical candidates with neoadjuvant chemotherapy (or interventional oncology methods, to be discussed later).

Importantly, chemotherapy can induce steatosis, fibrosis, and functional liver derangements, which need to be factored in to the regimen planning. For example, even a short 3-mo course of oxaliplatin regimens (e.g., FOLFOX) increase liver morbidity after hepatic resection, as shown by Nordlinger et al[20] in the EORTC Intergroup 4098 trial.

Currently the strongest evidence guiding chemotherapy selection for these patients is randomized phase III data showing that KRAS wild-type subgroups have the highest RECIST (radiological) response rates for EGFR inhibitors (cetuximab). Therefore, triple therapy using cetuximab in addition to irinotecan (CRYSTAL study)[23] is probably the best choice for patients with a goal of downsizing for potential resection. Furthermore, in a randomized controlled trial of 138 unresectable patients, Ye et al[24] found that cetuximab combined with chemotherapy improved resectability (25.7% vs 7.4%, P < 0.01) of liver metastases and improved 3-year OS (41% vs 18%, P = 0.013) compared with chemotherapy alone. Of note, panitumumab is a humanized version of cetuximab that simply has a different dosing schedule and can be used in the case of cetuximab allergy. Panitumumab was established in the PRIME study showing significant PFS increase (9.6 mo vs 8 mo)[25,26].

These patients will most commonly have three or fewer lung metastases or peritoneal implants. This remains an area of controversy. The surgical (or ablative) control of these lesions remains understudied, as does the role of systemic therapy in these patients.

Despite all modern efforts described above, < 5% of all patients with mCRC will be cured; usually due to relapse before reaching 10 years of DFS. Moreover, the 5-year OS for patients with mCRC is 7%[5,26]. The goal of care from these patients moves from cure to control.

The paradigm shift in treating these patients will be demonstrating the impact of liver-directed therapies in the context of extrahepatic disease. It makes sense that the vast synthetic and filtration functions of the liver (a vast understatement of the complex biophysiology of this organ) serve a vital role in patient morbidity and mortality. Stated differently, it makes sense, but has not yet been rigorously proven, that patients with widespread mCRC benefit from a focus on the CLM, whether it be with chemotherapy or locoregional treatments (to be discussed later). Moreover, the liver especially makes sense for mCRC (especially colon cancer metastases) given the predictable metastatic pattern for these tumors.

Globally, contemporary standard of care is the use of doublet or triplet regimens. Doublet regimens involve combination of irinotecan and oxaliplatin (which is not useful alone[27]) with 5-FU/leucovorin (LV)[28-31]. All trials comparing first-line combination superiority between oxaliplatin and irinotecan have shown equivalent survival. Capecitabine[32,33], S1[34], and UFT-tegafur are oral fluoropyrimidines with established non-inferiority to 5-FU, and can be used in doublet/triplet combinations based on toxicity profiles. Specifically, CAPIRI and FOLFIRI had similar efficacy and adverse event profiles in a meta-analysis of the six major trials available[35]. Randomized controlled trials on triplet and quadruplet regimens are forthcoming, but the quadruplet regimens stand the risk of being limited by toxicity.

Unfortunately, de Gramont and the GERCOR group showed that the majority of patients treated with continuous FOLFOX will discontinue for neurotoxicity before progression. OPTIMOX 1 showed that an “oxaliplatin holiday” with just infusional 5-FU could be used to gain maximal time on oxaliplatin and reach progression before abandoning the drug (so called “stop and go” oxaliplatin regimen). OPTIMOX 2 aimed to study whether the infusional 5-FU was even necessary, but remained underpowered with 216 patients once bevacizumab was approved in France, which effectively dismantled OPTIMOX 2. Nonetheless, patients receiving maintenance therapy had improved PFS and OS.

Doublet regimens are often enhanced with a choice of two additional targeted biological agents that benefit from specificity and reduced toxicity profiles (but can be cost-prohibitive). These are an anti-vascular endothelial growth factor (VEGF) monoclonal antibody, bevacizumab, and another monoclonal antibody targeted for the human epidermal growth factor receptor (HER-1 or EGFR).

In the second line, the phase III VELOUR trial[36] showed that FOLFIRI plus aflibercept (a fusion protein of human IgG Fc1 and VEGF receptors 1 and 2 that functions as a VEGF-trap and blocks all human VEGF-A isoforms, VEGF-B, as well as placental growth factor PIGF-2) provided significant survival advantage, which was even seen in patients with resistance at prior bevacizumab (a pure VEGF ligand inhibitor as opposed to a VEGF trap) treatment. The European TML study showed that patients who progressed on first-line chemotherapy plus bevacizumab benefitted from having bevacizumab included in their second-line chemotherapy as well, with significant PFS and OS differences (5.7 mo vs 4.1 mo and 11.2 mo vs 9.8 mo, respectively)[37]. In a related study, the RAISE study (presented at ASCO 2015) showed that second-line FOLFIRI ramucirumab (a human IgG1 monoclonal antibody targeting the extracellular domain of VEGFR-2) also leads to a statistically significant improvement in OS in comparison to FOLFIRI plus bevacizumab (13.3 mo vs 11.7 mo, respectively).

In the salvage setting, the CORRECT trial[38] stands out as a landmark trial, where regorafenib (an oral multikinase inhibitor) showed a significant survival advantage compared to best supportive care. The newest frontier in this setting was announced at 2015 ASCO GI with famitinib, a multi-target receptor tyrosine kinase inhibitor primarily acting against angiogenesis which had favorable phase I data, with further studies forthcoming.

As chemotherapy improves, patients are now living longer. Median survival now exceeds 30 mo, with many patients living up to 4-5 years with advanced disease. In fact, patients are living so long that we must begin to consider the cumulative toxicities from multiple lines of chemotherapy[39].

The contemporary approach for patients who progress on doublet therapy is to “flip the doublet”, that is, irinotecan and oxaliplatin based doublets are interchanged. The E3200 study showed that adding bevacizumab to the second-line in bevacizumab-naïve patients increased progression-free (7.3 mo vs 4.7 mo) and overall (12.9 mo vs 10.8 mo) survival. The TML study has been described above but supports continuation of bevacizumab in the second line. As described above, VELOUR[36] added aflibercept to the list of medications with phase III survival improvement data in the second line. As above, EPIC[40] showed that cetuximab improved PFS (4 mo vs 2.6 mo) in patients with EGFR-expressing tumors.

Data regarding third-line therapies are relatively slim. Regorafenib was studied in the CORRECT trial as detailed above, and shown to improve OS in patients progressing on anti-EGFR therapy (6.4 mo vs 5 mo). This comes at the risk of hand-foot syndrome which can be seen with regorafenib[38].

KRAS and BRAF at a minimum, in the future multi-gene testing systems for targetable mutations.

As CRC approaches designation as a chemosensitive disease, and is increasingly facilitated by palliative endoluminal stenting for concurrent obstruction, surgery will be used less frequently as the first treatment[41,42].

In the correct patient population, there is increasing evidence that, even in systemic disease, removal of the primary tumor improves survival. This is retrospective at this point.

The concept that heavy node-positive disease at primary resection may be treated as if the patient has distant metastases.

The benefits of chemotherapy will become more established for patients with stage II tumors that are high risk (lymphovascular invasion/emergent presentation with obstruction or perforation/poor differentiated/T4 N0). Already there is level I evidence for survival benefit with chemotherapy for stage III tumors.

As the number of agents and patient survival increase (Table 2), choosing an optimal chemotherapy strategy becomes more complex, limited by expense as well as cumulative toxicity profiles.

The test-of-time approach is a crucial concept in the management of mCRC that can spare patients from unnecessary surgery. This concept was popularized by Livraghi et al[44] in 2003, and sought to ask the question: for resectable patients awaiting surgery, what happens if ablation is done first, with the plan of still taking them to surgery? In their study, ablated patients who had complete tumor necrosis with margins [with necrosis defined as non-enhancement at contrast-enhanced computed tomography (CT) on postoperative day 1], 98% (52/58) were spared surgical resection, 23/52 (44%) because they remained disease free, and 29/52 (56%) because they manifested with widespread disease. Had these patients undergone surgical resection, it would have resulted in unnecessary morbidity because they would have developed those new liver metastases anyway. All patients received chemotherapy before radiofrequency ablation (RFA) unless they refused; in total, 70/88 patients (80%) received chemotherapy before RFA.

To summarize, Livraghi et al[44] showed that if RFA with margins can be performed for resectable oligometastatic CLM, 98% of patients are spared unnecessary surgery. The concept that today’s “resectable” will come to mean “ablatable” is a visionary guiding principle for all interventional oncologists seeking to treat patients in a percutaneous minimally invasive fashion.

Ablation is a repeatable and minimally invasive therapy that is directly complementary to surgical resection in principle. Induction chemotherapy does not always result in a complete visual disappearance of all lesions at imaging. The initial “induction” response of first-line chemotherapy can be “consolidated” by percutaneous image-guided ablation in a minimally invasive fashion. Depending on local practice patterns, the eradication of visible tumor may allow for the patient to take a “holiday” from receiving chemotherapy (which can involve frequent visits to the hospital, side effects, and expense, all of which may affect quality of life). As above, this can be used with the test-of-time approach.

In randomized controlled trials, the gold standard endpoint is overall survival. However, this endpoint requires extended follow-up and is often confounded by subsequent lines of treatment. Even though the United States Food and Drug Administration (FDA) will approve treatments based on the endpoint of PFS, it is becoming less controversial to use PFS as a surrogate for OS as a primary endpoint in studies in medical oncology and interventional oncology.

The use of PFS as a surrogate for OS should be cancer-specific. In this regard, for patients with mCRC and CLM, liver progression of disease is what will usually affect survival. In 2002, the FDA approved SIR-spheres for mCRC CLM on the basis of statistically significantly improved PFS. Recently, the FDA published regulation 21CFR813, subpart H allowing the use of PFS in the accelerated approval of new drugs for serious illnesses.

Ablation is defined as the delivery of energy into a tumor to destroy that tumor. Just like a margin of normal tissue must be removed for R0 resection, it makes sense that a margin of normal tissue must be ablated to perform A0 ablation.

The concept of A0 ablation, as well as objective criteria to document the performance of it, is a subject being pioneered at selected centers worldwide, for example, by work including Interventional Oncologists Sofocleous, Erinjeri, and Solomon and colleagues at Memorial Sloan Kettering Cancer Center (MSKCC) in New York City. For example, immediate post-ablation biopsy of the peritumoral margin can be evaluated using YO-PRO-1 as a biomarker of cell death[45].

Strategies to document A0 ablation using immediate procedural imaging are also being studied but will benefit from the aforementioned histological verification prior to rigorous clinical application. Makino et al[46] performed a pilot feasibility study of pre-ablation and post-ablation contrast-enhanced CT [or magnetic resonance imaging (MRI), respectively] images, showing that MR fusion provided technically acceptable image registration for ablation volume comparison in 86/92 (93.5%) while CT did so for 62/92 (68%), noting that shrinkage of the ablation zone during the ≤ 28 d between pre- and post-ablation scans may need attention in future studies. In another feasibility pilot study, Rempp et al[47] showed that ablation performed under MR guidance can be followed with MR thermometry and diffusion-weighted imaging to document tumoricidal temperatures in real time and establish the margins of cellular destruction. Finally, while not FDA-approved for this indication, Mauri et al[48] have performed a substantial pilot study of sulfur hexafluoride microbubble ultrasound intravascular contrast for intraprocedural rapid assessment of ablation volume, showing that contrast-enhanced ultrasound spared retreatment in 29/93 (31%) of patients who had incomplete ablation with an approximately 22% cost reduction for overall interventional treatment. Of note, unfortunately, gas bubbles seen at routine B-mode ultrasound during thermal ablation do not indicate complete ablation of the gas-emitting region[49], and the zone of gas bubbles does not correlate accurately with the zone of necrosis[49].

Guidance options for probe placement include ultrasound, CT, CT fluoroscopy, positron emission tomography (PET)/CT, and MRI (with compatible ablation systems). Augmented reality systems including fusion imaging and volumetric spatial navigation are emerging as adjunct technologies.

Ultrasound is well suited for mCRC in the liver because the lesions are usually conspicuously hypoechoic relative to the surrounding liver parenchyma. Using the sensitivity of diffusion-weighted MRI to detect lesions non-invasively, cognitive fusion with ultrasound provides the benefits of real-time needle/probe tracking with the respiratory cycle. Additionally, ultrasound provides real-time vascular assessment with Doppler ultrasound as well as omniplanar vector planning, which is difficult for CT. Intraoperative ultrasound is especially sensitive for lesion detection. One unsolved dilemma with ultrasound guidance during ablation is that water vapor and nitrogen gas released during tissue boiling (RFA, microwave) or ice formation (cryoablation) cause acoustic scattering, acoustic refraction, and acoustic shadowing of an already-hypovascular lesion. Therefore, given the option to ablate a deep lesion and a superficial lesion in sequence, experienced operators ablate the deeper lesion first to avoid acoustic shadowing of the second lesion.

Body habitus may limit ultrasound capabilities because the subcutaneous adipose layer scatters the otherwise-organized sound waves originating from the cutaneous piezoelectric crystals, distorting image quality. This becomes especially important for deep liver lesions. CT is helpful in this setting. Fluoroscopic CT can be used to safely guide the needle/probe/antenna to the target and confirm the zone of ablation. Iodinated contrast can be given that will provide a time window of 5 min or less for targeting of inconspicuous lesions. PET/CT using fluorodeoxyglucose (FDG) (or any other radiotracer that emits positrons) is a powerful technique to target the metabolically active portions of tumors. Indeed Shyn et al[50] have shown that a reasonable 20-s breath-hold PET acquisition (shorter than that usual 3-min summed breath-hold acquisition) can safely be used for intrahepatic lesion targeting. In practice, it is important to remember that lead aprons do not block positrons, so in-room time should be limited once the FDG dose is given. Finally, not all ablation equipment is MRI compatible but MRI does provide exquisite anatomic detail, with Rempp et al[47] demonstrating technical success of MRI-guided ablation in 210/213 lesions (98.6%) when using wide-bore MR-guided RFA. Especially for patients with large body habitus, the wide-bore magnet is preferable, and in practice the patient can be positioned asymmetrically in the bore to open up space on the right side of the patient.

Electromagnetic tracking-based image fusion is a powerful technique combining the spatial/metabolic/physiological sensitivity and specificity of PET/CT, contrast-enhanced CT or MRI, with the real-time imaging and handheld convenience of ultrasound. Example software provides “plug-and-play” fusion whereby images from any CD-ROM or PACS can be registered to ultrasound by the operator by selecting mutual anatomic landmarks in a series of images. The magnetic field generator tower that tracks ultrasound probe positioning may not be compatible with cardiac pacing devices. Respiratory motion (which can vary later in the case due to sedatives) remains an issue for this fusion - the anatomy is fused at a single point in the respiratory cycle (preferably expiration which lasts longer than inspiration and provides a longer window for needle/probe positioning). Mauri et al[51] have performed one of the largest series regarding fusion guidance during thermal ablation in 295 ablation cases, demonstrating correct tumor targeting and ablation in 266/295 (90.2%) of cases. Of note, this technology is most valuable for lesions that are poorly visualized by ultrasound but require ultrasound for appropriate positioning (for example liver dome lesions).

Navigation is important to differentiate from fusion. Fusion simply refers to overlay of two image sets (usually a real-time ultrasound to a static hybridized PET/CT for example), which are spatially registered to vary in the x, y and z axes of ultrasound scanning. Navigation refers specifically to tracking of the operator’s instrument (needle/probe/antenna) in space toward a target and can be performed even on unfused images. Improved lesion targeting with navigation systems may one day become the standard of care, with emerging studies such as that from Bale et al[52] showing that use of an optical frameless stereotactic navigation system with percutaneous RFA achieves similar OS and DFS rates as surgical resection, directly challenging surgical resection as the first-line treatment of choice for CLM.

After clinical factors have been addressed, technical factors guiding patient selection include the following. (1) the number of tumors is limited, usually ≤ 4 tumors; (2) the size of tumor appropriate for ablation. There is no strict absolute cutoff and tumor histology and ablation technology must be considered. Broadly, 3 cm is not controversial, while ≤ 5 cm is acceptable. A new frontier ripe for study is the combination of embolization and ablation for this patient population, which has already shown promise for HCC based on data from MSKCC; and (3) tumor location near major vessels will designate the ablational at-risk margin regarding incomplete ablation due to heat sink and current sink (detailed below) and should prompt percutaneous temperature probe placement to document tumoricidal temperatures along the at-risk margin.

Contraindications to ablation include: (1) uncorrectable coagulopathy; and (2) no safe window for access vector to the tumor, even after considering strategic patient repositioning and temporary organ displacement maneuvers.

RFA involves delivery of energy with a frequency of less than 900 kHz using needle electrodes of varying geometry. The energy agitates ions resulting in temperature elevation. At 60 °C, coagulation necrosis occurs. At 100 °C, undesirable carbonization occurs, limiting heat distribution throughout the tumor. Tumors near blood vessels pose a challenge for RFA, as the nearby blood flow will remove electrical current and heat from the ablation zone (heat sink and current sink). Nearby biliary structures risk stricture formation as well.

The ideal candidate for RFA is a patient with a solitary CLM < 3 cm in size. Although rigorous evidence is forthcoming, it is believed that A0 ablation requires at least a 5-mm margin of normal tissue (based on imaging follow-up 1-2 mo post ablation).

Although major literature is forthcoming, the available retrospective data suggest that carefully selected RFA patients do just as well as resected patients. For example, Gillams et al[53] showed in their series of 167 patients undergoing percutaneous RFA, for patients with ≤ 5 metastases, maximum diameter ≤ 5 cm and no extrahepatic disease, the 5-year survival from the time of diagnosis was 30%, and the 5-year survival from the time of first RFA was 26%. This compares favorably to the 5-year survival for operable patients of a median of 32%.

Future frontiers with RFA will include pre-medication or pre-embolization (see Bland Embolization section below). For example, Devun et al[54] have shown that systemic pretreatment of mice with the DNA repair inhibitor Dbait improves the efficacy of radiofrequency ablation.

The literature regarding RFA of CLM is highlighted in Table 3.

Cryoablation uses a special probe applying the Joule-Thompson principle to argon gas to cause rapid tissue cooling. Cancer cells contain more water than non-cancer cells. Freezing leads to the formation of intracellular ice crystals. At -40 °C, tissue death occurs. Early generation devices have limited the implementation of this modality for the liver due to adverse event reports of cryoshock, a condition similar to diffuse intravascular coagulation. Despite the availability of effective thermal ablation methods in the liver, cryoablation remains the oldest method of tumor ablation and has been used with success in other organs[55-57]. Furthermore, the ability to visualize a clear ablation margin under CT using cryoablation makes it an attractive option, and for this reason forthcoming literature regarding cryoablation in the liver constitutes a new frontier for mCRC CLM.

The available literature regarding cryoablation of CLM is highlighted in Table 4.

Microwave ablation (MWA) was developed to overcome limitations with RFA. MWA uses high-frequency waves (900 MHz and 2.4 GHz) to oscillate water molecules, creating friction, tissue heating, and tissue destruction by coagulation necrosis. MWA does not rely on electrical current to generate heat, and therefore a current sink is not an issue. Due to the power of the microwave generator, MWA can overcome the heat sink even if it does not fully avoid it. MWA is superior to RFA in treating larger tumors, with lower recurrence rates (as low as 6%)[58].

There is only one randomized trial comparing MWA to resection in mCRC patients with CLM. No statistically significant difference was found as the study was underpowered (40 patients). The mean survival time was greater in the MWA group (27 mo vs 25 mo) while the mean disease-free interval was slightly shorter (11.3 mo vs 13.3 mo)[59]. The available literature regarding MWA of CLM is highlighted in Table 5.

Irreversible electroporation was developed to overcome limitations presented by thermal ablation (RFA and MWA), namely, collateral damage to bile ducts or other important structures that could be damaged due to heat. Irreversible electroporation is a new technology and much of the data is forthcoming. Studies that have included mCRC patients with CLM have shown primary efficacy of up to 100% for tumors adjacent to vascular and biliary structures[60-63]. In this setting, similar to that shown in the microwave setting[59] a tumor size of ≥ 3 cm seems to be an independent risk factor for local recurrence.

The exact timing and modality of follow-up imaging after ablation varies on an institutional basis. There are three main systems used in evaluating treatment response by these patients: the World Health Organization criteria, the Response Evaluation Criteria in Solid Tumors and the Positron Emission Evaluation Response Evaluation Criteria in Solid Tumors.

The treatment team should be aware that there is one major nuance to the application of all of these criteria. With percutaneous ablation, the ablation creates a volume of imaging abnormality larger than the target tumor. Therefore, for proper evaluation of response to treatment and early detection of local tumor progression, a post-ablation scan at 4-8 wk is considered the new baseline for future comparisons. The next scan can be timed for 2-4 mo, and will be used to detect local, near local, and distant disease progression. Ideally the interventional oncologist should see these patients in their clinic and review imaging directly with the patient.

Relying on high-first-pass extraction of chemotherapeutic agent from the bloodstream, hepatic arterial infusion (HAI) of chemotherapy utilizes a subcutaneous injection port with a thin intra-arterial indwelling catheter that has its tip in the proper hepatic artery at the origin of the gastroduodenal artery. The port can be placed surgically or percutaneously. The skeletonization via the percutaneous route is thought to be more complete although no rigorous studies have demonstrated this. Maintenance Tc-99 microalbumin aggregate studies and angiograms are occasionally performed during the dwell time of the device to ensure there is no extrahepatic drug delivery.

HAI improves OS. This finding supports the concept of a liver-directed approach in patients with widespread mCRC, and also supports the concept of organ-directed approaches for metastatic cancers. A meta-analysis of six HAI trials showed improved response rates as well as OS advantage (14.5 mo vs 10.1 mo, P = 0.0009)[8].

It would make sense that treatment of the “field” of radiologically uninvolved parenchyma would optimally be done via the portal vein, where the “normal” parenchyma derives the majority of its blood supply. This procedure is different from portal vein “embolization”, which is done to hypertrophy a contralateral lobe prior to resection. The infusion of portal venous chemotherapy and evaluation of first-pass extraction has been studied. The group SAKK from Switzerland showed that adjuvant portal vein infusion of mitomycin C + 5-FU improved OS but did not reduce the recurrence of liver metastases[68]. A subsequent prospective three-arm randomized multicenter trial of 753 patients with stages I-III colorectal cancer (surgery only vs adjuvant portal vein chemotherapy vs adjuvant peripheral vein chemotherapy) showed that portal vein infusion did not improve DFS and OS. Actually, PVI was shown to have potentially harmful effects with a statistically significant increase in early death in the PVI group[69]. This technique remains investigational at the time of this report.

Bland embolization is the injection of particles with the goal of causing selective tumor ischemia. The practice has evolved from use of PVA toward the use of calibrated microspheres in ascending sequential order of sphere diameter to reach full stasis.

When compared to chemoembolization, there are no level I studies demonstrating superiority of chemoembolization to bland embolization for patients with CLM. Proponents of bland embolization believe that tumor death is caused by anoxia and that chemoembolization achieves its endpoints by this route. Bland embolization is thought to be more repeatable owing to better preservation of the hepatic arterial vasculature. This may be due to reduced caustic effect of chemotherapeutic agents on the intima. Though head-to-head studies are not available, the ability of bland embolization to facilitate extended repeat treatments has been shown[70]. Bland embolization is cheaper than chemoembolization, although this cost may be offset by the brief hospital stay for the management of post-embolization syndrome (categorized by fevers, chills, nausea, and abdominal pain), which is more pronounced for bland embolization than conventional chemoembolization. Conventional chemoembolization still results in a high systemic dose of chemotherapeutic agent, which can result in systemic side effects, although this has changed with the advent of drug-eluting bead technologies. The intra-arterial use of outdated or biologically irrelevant chemotherapeutic regimens (e.g., doxorubicin) intra-arterially has been challenged when these agents are not necessarily used peripherally for the same tumor. However, with the advent of drug-eluting bead irinotecan chemoembolization this has changed specifically to mCRC as described later. Bland embolization can be effectively used to “paint” tumors for post-embolization targeting and ablation under CT guidance, similar to lipiodol in conventional transcatheter arterial chemoembolization (TACE). In the salvage setting, the cumulative toxicities of TACE chemotherapeutic agent with prior lines of chemotherapy, as well as possible resistance of aggressive tumor biology to the chemotherapeutic regimens classically used in TACE, may weigh in favor of the more repeatable bland embolization procedure.

On the new frontiers of bland embolization, in Japan where they reported DEBIRI and radioembolization to be less available, Tanaka et al[71] have recently undertaken a pilot study of questions relevant to the current discussion. First, how does bland embolization, a flow-directed procedure dependent on causing ischemia, fare with tumors that are hypovascular (as colorectal metastases classically are) relative to liver parenchyma? Second, since bland embolization is repeatable and preserves hepatic vasculature, can it be performed less invasively through an implantable port? Tanaka and colleagues have documented the effects of 100-μm microspheres on a patient with metastatic rectal cancer with OS approaching 6 mo, using no other therapies. Future systematic studies might specify outcomes of bland embolization in hypovascular liver metastases.

Another new frontier for bland embolization will be the pre-ablation embolization setting. For example, Tanaka et al[72] have shown in pigs that bland embolization pre-ablation is more effective than bland embolization post-ablation, and that bland embolization with 40-μm microspheres enhances the efficacy of RFA more than bland embolization with 250-μm particles. Future work might include embolization with novel particles.

The literature regarding chemoembolization is the strongest of all transcatheter methods. However, it is important to note that the term chemoembolization is a vast over-simplification given the variety of methods used to perform this procedure [including for example, the choice and dose of chemotherapeutic agents, endpoint of treatment (stasis, near-stasis, or neither), conventional vs drug-eluting bead utilization, and post-treatment embolization with gelfoam or other methods]. This non-standardization has made meaningful meta-analysis of the chemoembolization literature challenging. However, as will be described later, the advent of drug-eluting beads may standardize chemoembolization for patients with mCRC CLM.

The premise of chemoembolization is to combine ischemia and chemotherapeutic penetration for enhanced (ideally, synergistic) tumor destruction. Conventional chemoembolization delivers the agent mixed with lipiodol, which proponents believe penetrates into the deepest vessels of the tumor and allows embolization while slowly leeching chemotherapeutic agent into the tumor. This method also causes failure of the transmembrane pump[73] thought to trap chemotherapeutic agent in cells. This is sometimes followed by a proximal embolization to reduce the pressure head of inflow and prolong the interaction time between the embolic bolus and tumor cells. The chemotherapeutic agent and lipiodol are prepared in the form of an emulsion in a method described by Lo et al[74] in their randomized trial of lipiodol TACE for HCC (not mCRC). However there are other ways to chemoembolize. For example, the original randomized controlled trial providing level 1 evidence for HCC (not mCRC) was performed by Llovet et al[75] and used gelatin sponge with doxorubicin (a technique no longer widely performed). In contradistinction, drug-eluting beads are thought to be a more reproducible mode of chemotherapeutic agent delivery, achieving the endpoint of ischemia as well as a more reliable chemical interaction with the chemotherapeutic agent and reduced systemic leeching of chemotherapy. For example, drug-eluting beads bound to irinotecan were introduced in 2006. DEBIRI has been shown to reduce systemic plasma levels by 75% when compared to intra-arterial irinotecan[76].

Factors that predict adverse events and hospital length of stay after TACE are embolization to complete stasis (P = 0.04), treatment with > 100 mg DEBIRI in a single session (P = 0.03), lack of pre-treatment with hepatic arterial lidocaine (P = 0.005), third or more repeated TACE (P = 0.05), > 50% liver involvement (P = 0.05), and pre-TACE bilirubin of > 2.0 g/dL[77-82].

The only phase 3 randomized controlled trial performed thus far by Fiorentini et al[79] randomized 74 patients with mCRC CLM to DEBIRI vs FOLFIRI. The DEBIRI group had improved survival (22 mo vs 15 mo, P = 0.031) and higher response rate (68.6% vs 20%) and longer life (8 mo vs 3 mo, P < 0.001).

Most recently in 2015, Iezzi et al[83] have shown phase II study results with DEBIRI + capecitabine (PFS 4 mo, OS 7.3 mo) that are comparable to those shown in the CORRECT trial (where regorafenib was compared to best supportive care in 760 patients with PFS 1.9 mo and OS 6.4 mo in the regorafenib group). The available literature regarding TACE of CLM is highlighted in Table 6.

Isolated liver perfusion is a complex open surgical procedure requiring a perfusionist. In contrast, isolated liver infusion, or chemosaturation is a minimally invasive equivalent of isolated liver perfusion that maintains normal hepatic arterial and portal vein inflow. Instead, the hepatic vein drainage is isolated and the chemotherapeutic agent is extracted from the blood (with the newest filtration system extracting approximately 93% of the melphalan) before returning the blood to the systemic circulation. The procedure is performed using triple access (right common femoral and right jugular vein access for blood extraction/filtration and re-entry respectively) and left hepatic artery access (for infusion catheter placement into the hepatic artery).

The proprietary device used for hepatic chemosaturation is a 16 Fr system made by Delcath and has two balloons on either side of a 7-cm fenestrated catheter segment. Delcath performed a Phase II clinical study of melphalan chemosaturation at the National Cancer Institute in the United States enrolling 16 patients with late-stage CLMs. While the safety profile was similar to the Delcath melanoma trial, the efficacy signal from the mCRC study was inconclusive, mainly because melphalan usage in the salvage setting was limited due to cumulative toxicities from prior lines of chemotherapy.

Of note, the isolated hepatic perfusion literature is more robust, using mitomycin C, oxaliplatin, and melphalan with and without tumor necrosis factor-α. For example, Reddy et al[84] studied 120 patients and showed overall response rate of 59% with median time to hepatic progression of 7 mo and median OS of 17.4 mo; patients receiving hepatic artery infusion of floxuridine benefited from a longer time to hepatic progression (13 mo vs 6 mo). Perhaps this and other surgical studies will continue to motivate research into the future outlook of isolated hepatic infusion (chemosaturation).

Kennedy, one of the pioneers of the utilization of radioembolization in CLMs, showed microdosimetry results of intratumoral radiation delivery up to 1000 Gy[84]. This dose is made possible by the limited penetrance of beta radiation. Radioembolization is generally accepted to deliver 100 Gy to intrahepatic tumors; a dose that is impossible using external beam radiation due to radiation-induced liver disease (which occurs at 30 Gy). The curative dose for adenocarcinoma mCRC is 70 Gy. It is important for interventional oncologists to understand that the dosimetry of radioembolization cannot be directly compared to the absorbed dosimetry of external beam radiation due to differences in radiobiology between a continuous low dose-rate beta radiation and intense but brief external beam photon radiation pulsation[85,86].

In 2002, the FDA approved radioembolization with SIR-spheres (20-60-μm spheres, while terminal arterioles are usually 10-40 μm in diameter while capillaries average 8 μm in diameter for humans, and 3 μm in rodents) based on a study by Gray et al[87] that administered resin microspheres through a hepatic arterial infusion pump (whole liver treatment) in the context of floxuridine (FUDR) and compared this to FUDR alone. Tumor volume response (50% vs 24%), carcinoembryonic antigen response (72% vs 47%), median liver time to local progression (16 mo vs 10 mo) and survival (39% vs 29% at 2 years, 17% vs 6.5% at 3 years, and 3.5% vs 0% at 5 years) all reached statistical significance. There were no significant differences in grade 3/4 toxicities or quality of life measures. Unfortunately, the company funding the trial did not extend this trial until reaching the OS endpoint.

The summary of current evidence for yttrium-90 is that there is a strong trend toward prolongation of liver PFS, PFS, and the universal endpoint of OS when yttrium-90 is added to systemic treatments in early and late lines of treatment based on small randomized controlled trials. The safety profile of yttrium 90 treatments, especially regarding gallbladder and biliary complications, has matured over time and this procedure has very low morbidity when done by well-trained operators at experienced centers[88,89].

The only randomized controlled trial comparing SIRT + chemotherapy to systemic chemotherapy alone was done by Van Hazel et al[90] who showed statistically significant doubling in PFS when SIRT was used as a first-line therapy. The forthcoming FOXFIRE and SIRFLOX studies will carry this concept into the modern chemotherapy regimens by using oxaliplatin in combination with SIRT. EPOCH will investigate the use of Theraspheres, glass microspheres that have a higher per-sphere radiation dose and are less embolic than SIR-Spheres resin microspheres. Some hypothesize that Theraspheres may fare better because the treatment is less embolic, which allows consistent delivery of the entire radioactive dose without concern for reflux, and facilitates oxygenation of the irradiated tissue (tumor destruction in radioembolization occurs by oxygen free-radical formation, therefore embolization is actually counterintuitive in this setting). The literature regarding radioembolization for mCRC CLM is highlighted in Table 7.

Herpes simplex virus (HSV)-1 is a virus that has been extensively studied and is known to use host cell machinery to replicate upon gaining entry into cells. Interestingly, the genome of HSV-1 is large but only a few genes are necessary for replication. This leaves space for genetic engineering to create a mutant virus that can selectively infect tumor cells and, if needed, be neutralized by administration of acyclovir (a routinely available antiviral agent).

Specific to CLMs, interventional oncology will continue to play a role in advanced anti-cancer therapy due to the value of the biological payload and the need for organ-directed delivery, especially with the relatively hypovascular CLMs. A phase I open-label dose escalation study at MSKCC demonstrated safety of delivery of neoadjuvant NV1020 via the right hepatic artery in a tumor-specific fashion[91].