INTRODUCTION — Approximately 146,970 Americans are diagnosed with colon or rectal cancer annually, and 49,920 will die from this disease, most with metastatic tumors [1]. Hepatic metastatic disease from colorectal cancer (CRC) is a significant clinical problem. The liver is the dominant metastatic site for patients with CRC, and although two-thirds of affected patients have extrahepatic spread, some have disease that is isolated to the liver. For patients with isolated liver metastases, regional treatment approaches may be considered as an alternative to systemic chemotherapy.

The available regional treatments for hepatic metastases from CRC include surgical resection, local tumor ablation (ie, instillation of alcohol or acetic acid directly into the metastatic lesions, cryotherapy, radiofrequency ablation [RFA]), regional hepatic intraarterial chemotherapy or chemoembolization, and radiation therapy (RT). Among these treatments, only surgery is associated with a survival plateau.

Although hepatic resection used to be reserved for patients with a maximum of three lesions in the same lobe if it was possible to achieve 1 cm margins and those without portal lymph node metastases, all of these "rules" have been challenged in the modern era, particularly with advancements in both surgical technique and systemic therapy [2]. Profound improvements in the outcomes of patients with metastatic CRC over the past 15 years have been attributed to increased use of hepatic resection in selected patients and more effective chemotherapy [3]. (See "Systemic chemotherapy for metastatic colorectal cancer: Completed clinical trials".)

As a result, the criteria for defining which patients are suited for surgical therapy have evolved, and most clinicians take an aggressive stance in the management of hepatic metastases.

This topic review will focus on resection and the role of systemic chemotherapy. Methods and results for local ablation, regional chemotherapy into the hepatic artery, radiation therapy, and management of the primary tumor in patients who present with stage IV disease are addressed elsewhere. (See "Nonsurgical treatment strategies for colorectal cancer liver metastases" and "Surgical management and palliation in patients who present with stage IV colorectal cancer".)

BIOPSY CONFIRMATION — A biopsy may be indicated to confirm the diagnosis, depending upon the clinical picture. The risk of tract seeding from percutaneous fine needle aspiration (FNA) biopsy appears small (only a handful of case reports are described in journals listed in Medline [4-10]), although this complication was described in five of 51 cases of biopsy-proven hepatic colorectal metastases in one series [11]. Needle tract seeding more often complicates FNA biopsy of primary liver tumors. (See "Clinical features and diagnosis of primary hepatocellular carcinoma", section on 'Histopathology'.)

SURGICAL RESECTION — Resection offers the greatest likelihood of cure for patients with liver-isolated CRC. In surgical case series, five-year survival rates after resection range from 24 to 58 percent, averaging 40 percent (table 1), and surgical mortality rates are generally <5 percent [4-18]. Subgroups with advanced age, comorbid disease, and synchronous hepatic and colon resection may have higher procedure-related mortality and worse long-term outcomes. As an example, five-year survival rate was only 25 percent in a population-based retrospective report of 3957 US Medicare enrollees undergoing hepatic resection for colorectal cancer liver metastases [19]. Even so, five-year survival rates with the most active systemic chemotherapy regimen (FOLFOX4) are only 10 percent. (See "Systemic chemotherapy for metastatic colorectal cancer: Completed clinical trials".)

Approximately one-third of five-year survivors suffer a cancer-related death, while those who survive 10 years appear to be cured [20]. In an analysis of 612 consecutive patients who underwent resection of CRC liver metastases and were followed for at least 10 years, there were 102 actual 10-year survivors (17 percent), and only one patient experienced a disease-specific death after 10 years of survival.

Because of its clear survival impact, surgical resection is the treatment of choice when feasible. Unfortunately, no more than 20 percent of patients with isolated hepatic metastases are amenable to potentially curative resection. Most are not surgical candidates because of tumor size, location, multifocality, or inadequate hepatic reserve.

However, some patients with initially unresectable disease may become resectable after induction chemotherapy. Updated guidelines from the National Comprehensive Cancer Network (NCCN) suggest that patients who appear to have initially unresectable metastatic colorectal cancer be categorized as potentially convertible or unconvertible [21]. Patients whose disease is felt to be unconvertible should be referred for alternative treatment (most often palliative chemotherapy), while induction chemotherapy is appropriate for those whose disease is potentially convertible. (See 'Neoadjuvant chemotherapy' below.)

Patient selection — The optimal selection of patients for hepatic resection is evolving, and the criteria for resectability differ among individual liver surgeons regarding borderline cases, from center to center and from country to country. One consensus statement defined absolute unresectability as nontreatable extrahepatic disease, unfitness for surgery, or involvement of more than 70 percent of the liver or six segments (figure 1) [22]. (See 'The OncoSurge decision model for selecting patients for neoadjuvant therapy' below.)

Risk scoring systems (such as the clinical risk score [12]) to predict which patients with metastatic CRC are most likely to benefit from resection are of limited clinical utility [23,24].

Modern multidisciplinary consensus defines resectable CRC liver metastases simply as tumors that can be resected completely, leaving an adequate liver remnant [25]. For resection to be considered, most surgeons would require that there be no radiographic evidence of involvement of the hepatic artery, major bile ducts, main portal vein, or celiac/paraaortic lymph nodes [26], and adequate predicted functional hepatic reserve postresection. Preoperative liver MRI and intraoperative ultrasound offer the optimal assessment of the number, size, and proximity of tumors to key vascular and biliary structures. Complete resection must be feasible. There should be no unresectable extrahepatic sites of disease, and the primary tumor should have been resected for cure [21].

Concurrent resection of hepatic and extraahepatic disease in well-selected patients is associated with a possibility of long-term survival, particularly if the extrahepatic disease is surgically resectable lung or ovarian metastases [27]. (See "Surgical resection of pulmonary metastases: Outcomes by histology" and "Surgical management and palliation in patients who present with stage IV colorectal cancer".)

The modern approach to resection of CRC liver metastases is outlined in the table (table 2) [2].

Selecting patients for diagnostic laparoscopy — Although diagnostic laparoscopy can identify occult intraperitoneal metastases, thereby preventing unnecessary laparotomy, it is not necessary in all patients. Laparoscopy is usually reserved for those thought to be at the highest risk for occult metastatic disease.

Some data suggest that these patients can be identified using a clinical risk score, derived from five preoperative criteria (nodal positivity, relapse free interval <12 months, more than one hepatic tumor, CEA level >200 ng/mL within one month of surgery, and size of largest hepatic tumor >5 cm) which were significant predictors of outcome from hepatic resection in multivariate analysis [12,28].

However, the clinical risk score is not used by the vast majority of hepatobiliary surgeons, mainly because it was derived from a database of patients treated in the 5-FU/LV era, before chemotherapy was routinely administered prior to liver resection, and before resection of numerous liver lesions became commonplace. In modern treatment paradigms, laparoscopy is infrequently performed, particularly since many patients have undergone surgical exploration of the peritoneum at the time of resection of a synchronous primary tumor.

We use diagnostic laparoscopy in patients with a suggestion of small volume carcinomatosis on radiographic imaging studies (ie, CT, MRI, or PET), and for selected other cases at high risk (eg, a metachronous presentation with several liver metastases that do not respond to chemotherapy).

PET scans — Whole body PET scans can identify radiographically occult extrahepatic disease and optimize the selection of appropriate candidates for hepatic resection, mainly by reducing nontherapeutic laparotomy rates [4,29-31]. The benefits of PET can be illustrated by the following studies:

• The superiority of PET over CT for detection of extrahepatic disease was suggested in a systematic overview of retrospective data which utilized a scoring system to weigh the individual studies according to the quality of the data and the clinical impact of the radiographic findings [29]. For the six articles judged to be of the highest quality [32-37], the pooled sensitivity and specificity for PET were 80 and 92 percent, respectively, for hepatic disease, and 91 and 98 percent, respectively, for extrahepatic disease [29]. The corresponding values for CT were 83 and 84 percent, respectively, for hepatic metastases, and 61 and 91 percent, respectively, for extrahepatic metastases. The percent change in clinical management from the performance of PET ranged from 20 to 32 percent (average 25 percent).
• The benefit of adding PET to the staging strategy was subsequently shown in a randomized trial in which 150 patients with CRC liver metastases selected for hepatic resection by CT were randomly assigned to triple-phase contrast-enhanced CT imaging only or CT plus a separate PET scan [30]. The primary outcome measure was the number of futile laparotomies (any laparotomy that did not result in complete tumor treatment or that did not result in a disease-free survival period of at least six months). In a preliminary report, the addition of PET significantly reduced the number of futile surgeries (28 versus 45 percent) and prevented unnecessary surgery in one of every six patients.

Two caveats must be considered when interpreting these results. First, this study used separate contrast-enhanced CT and PET images and not the increasingly popular integrated PET/CT imaging, in which both PET and CT are performed sequentially during a single visit on a hybrid PET/CT scanner. The CT component of integrated PET/CT imaging is performed in most institutions without the use of intravenous contrast material, which compromises the detection of small metastases both within and outside of the liver. At some institutions, PET/CT is carried out with intravenous contrast, but this practice is not widespread.

Secondly, the results of restaging PET scans (particularly if negative) must be interpreted in the context of recent therapy. Chemotherapy may alter the sensitivity of PET for the detection of metastases, thought due to decreased cellular metabolic activity [38]. In a series of 42 patients undergoing surgery for hepatic CRC metastases, 15 of 41 liver lesions (37 percent) were undetected by PET among the 13 who had received preoperative chemotherapy. Three of the 13 patients had lesions confirmed pathologically (one single, two multiple liver lesions) that were all undetectable by PET. In contrast, of the 29 patients who did not receive preoperative chemotherapy, 16 of 69 lesions (23 percent) were undetected by PET.

Guidelines from the NCCN recommend a staging PET scan for patients with potentially surgically curable metastatic colorectal cancer [21]. In the US, CRC restaging is a Medicare-reimbursable diagnosis for PET.

Number and location of metastases — The optimal selection of patients who are candidates for resection is evolving. In the past, contraindications to liver resection were defined by retrospective series and recapitulated in a large multi-institutional review of 856 patients [10]. Based upon these analyses, surgeons have been reluctant to offer resection to patients with more than three lesions, those with bilobar distribution (ie, tumor involving any segments of the left and right hemi-liver), those in whom it was not possible to achieve 1 cm margins, patients with portal lymph node metastases, and those with liver metastases from cancers other than colorectal tumors [39].

However, these relative contraindications have been challenged [2]. As examples:

• With safer surgery and better chemotherapy, hepatic resection is playing an increasing role in patients with multiple, even bilobar tumors [40-43]. In a report of 484 patients with multiple metastases (136 with ≥4, 36 with ≥8) five-and ten-year survival was 47 and 29 percent [42]. Even patients with ≥8 metastases had a five-year survival rate of 24 percent.

Others emphasize the importance of modern systemic chemotherapy to achieve favorable long-term outcomes from hepatic resection in patients who have multiple metastases because of the high risk of recurrence after surgery (over 50 percent) [44-46]. (See 'Neoadjuvant chemotherapy' below and 'Therapy after resection of liver metastases' below.)

• In a multicenter report of 557 patients who underwent resection for CRC liver metastases, there was no difference in five-year overall survival rate or in the rate of tumor recurrence in the liver whether the tumor-free margin was 1 to 4, 5 to 9, or >10 mm [47]. Only patients with a positive margin had worse survival and a higher intrahepatic recurrence rate. Similar results have been noted by others [46,48].
• One of the classic contraindications to hepatectomy is the presence of portal lymph node metastases, thought to indicate systemic disease that could not be successfully treated surgically. This view has been challenged [26,49-51]. Jaeck et al reported outcomes of complete hepatic pedicle lymphadenectomy in patients with CRC liver metastases [49]. Nineteen percent of patients survived three years despite portal node metastases. Outcome was more favorable if nodal involvement was limited to the porta as compared to along the common hepatic artery (three-year survival 38 versus 0 percent, respectively).

An important caveat is that this was a retrospective review of highly selected patients who did not have enlarged portal nodes detected preoperatively. Whether this principle can be extrapolated to patients with grossly enlarged nodes is unclear.

Largely based upon these data, modern criteria for resectability have been expanded to include any patient in whom all disease can be resected with negative margins and who has adequate hepatic volume/reserve [51].

NEOADJUVANT CHEMOTHERAPY — The availability of increasingly effective systemic chemotherapy has prompted interest in preoperative or neoadjuvant systemic chemotherapy as a means of "downstaging" liver metastases prior to resection. Preoperative hepatic intraarterial chemotherapy is discussed below. (See 'Preoperative hepatic intraarterial chemotherapy' below.)

Conversion therapy for initially unresectable liver metastases — The term "conversion therapy" has been proposed to designate the use of induction chemotherapy in patients with isolated but initially unresectable CRC liver metastases [52]. Between 12 and 33 percent of such patients have a sufficient downstaging response to permit a subsequent complete (R0) resection (table 3) [53-69]. Five-year survival rates average 30 to 35 percent, results that are substantially better than expected using chemotherapy alone (4 to 9 percent five-year survival, even with the most active regimens). (See "Systemic chemotherapy for metastatic colorectal cancer: Completed clinical trials".)

In the largest prospective study, 184 patients with initially unresectable liver metastases underwent resection after downstaging chemotherapy with 5-FU/leucovorin alone (18 percent), or with oxaliplatin (62 percent), irinotecan (6 percent), or both (9 percent) [54]. Overall survival at 5 and 10 years was 33 and 27 percent, respectively, while the corresponding disease-free survival rates were 19 and 15 percent, respectively.

Among the 148 patients who survived five years or longer, 24 (16 percent) were considered "cured", six after a repeat resection of recurrent disease. Among the independent predictors of cure were maximal size of metastases 30 mm at diagnosis, three or fewer metastases at hepatectomy, and a complete pathologic response to downstaging chemotherapy.

The addition of cetuximab or bevacizumab to a chemotherapy backbone containing oxaliplatin or irinotecan may increase the number of patients potentially eligible for resection and improve outcomes [70-73]. Although there is a correlation between objective response to neoadjuvant chemotherapy and prognosis [74], the number of chemotherapy cycles should be limited because of the potential for liver toxicity (see below).

Even with the most effective regimens, the complete pathologic response rate after neoadjuvant chemotherapy is only 4 to 9 percent [54,74,75]. The majority of radiographic completely responding lesions (83 percent in one series [76]) contain viable tumor. Thus, even in the setting of a complete clinical response, resection is still needed. Issues surrounding the accuracy of PET scans in patients who undergo chemotherapy are discussed above. (See 'PET scans' above.)

Choice of regimen — The optimal regimen to be used in the neoadjuvant setting is not established. We consider that the following regimens have been associated with high objective response rates, and are reasonable choices: FOLFOX with or without bevacizumab, FOLFOXIRI, or FOLFIRI or FOLFOX plus cetuximab (for patients whose tumors lack K-ras mutations) (table 4). (See "Systemic chemotherapy for metastatic colorectal cancer: Completed clinical trials", section on 'Molecular determinants of response'.)

On the other hand, guidelines from the NCCN suggest any of the following regimens [21]:

• FOLFOX or CAPOX or FOLFIRI with or without bevacizumab or
• FOLFOX or CAPOX or FOLFIRI plus cetuximab (wild-type K-ras only)
• FOLFOXIRI

Guidelines from the NCCN further recommend that patients be reevaluated for conversion to resectable disease every two months [21].

Liver toxicity from neoadjuvant chemotherapy — There are an increasing number of reports of steatosis (nonalcoholic steatohepatitis, NASH), vascular injury, and nodular regenerative hyperplasia in the livers of patients treated with preoperative irinotecan or oxaliplatin-containing chemotherapy regimens [54,77-87]. The following represents the range of findings:

• European studies (and at least two American reports [83,87]) have described hepatic sinusoidal abnormalities similar to those that characterize veno-occlusive disease, predominantly in patients receiving preoperative oxaliplatin [80,84,86,88,89]. (See "Pathogenesis and clinical features of hepatic sinusoidal obstruction syndrome (veno-occlusive disease) following hematopoietic cell transplantation".)

In one study, such changes were present in 44 (51 percent) of 87 hepatic specimens from patients who had received preoperative chemotherapy, 77 percent oxaliplatin-containing [80]. None of these changes were seen in the 66 patients undergoing liver resection who had not received preoperative chemotherapy.

• In addition to the vascular changes, nodular regenerative hyperplasia (precluding subsequent resection) has been reported in three patients treated with neoadjuvant oxaliplatin plus 5-FU for liver metastases [85]. (See "Nodular regenerative hyperplasia of the liver".)

• In contrast, irinotecan-containing regimens have more often been associated with steatosis and steatohepatitis [79,83,87]. An early report compared outcomes from hepatic metastasectomy in 325 patients with steatosis and 160 patients without steatosis [79]. Of the patients with markedly steatotic livers, 66 percent had received preoperative chemotherapy with a fluoropyrimidine, leucovorin, and/or irinotecan. Although mortality was not significantly worse in patients with steatotic as compared to nonsteatotic livers, marked steatosis was an independent predictor of perioperative infectious complications.

A second series of evaluated histologic liver sections from patients undergoing hepatic resection for CRC liver metastases who had received no chemotherapy (n = 158), 5-FU only (n = 63), 5-FU/irinotecan (n = 94) or 5-FU/oxaliplatin (n = 79) [83]. The median duration of preoperative chemotherapy was 16 weeks. Compared to the group not receiving chemotherapy, patients receiving oxaliplatin had significantly higher rates of sinusoidal dilatation (19 versus 2 percent) but not steatohepatitis (6.3 versus 4 percent), and those receiving irinotecan had significantly higher rates of steatohepatitis (20 versus 4 percent) but not sinusoidal dilatation. Patients with steatohepatitis had a significantly higher 90-day mortality rate as compared to those without steatohepatitis (14.7 versus 1.6 percent, respectively).

Steatohepatitis has been observed more frequently after chemotherapy in patients with a higher body mass index, which may explain why this complication is reported more frequently in US studies [81], while vascular lesions are seen more commonly in patients treated in Europe [77,79,80,84,86,88,89].

There are conflicting data as to whether postchemotherapy vascular changes such as sinusoidal dilatation or steatohepatitis increase the risk of perioperative morbidity/mortality [83,90,91]. Others suggest that these risks are related to the duration and/or timing of preoperative therapy [84,92,93]. In one report, administration of more than 12 weeks of chemotherapy or an interval of four or fewer weeks between stopping chemotherapy and resection predisposed patients to more postsurgical complications, higher rates of reoperation, and longer hospitalization [92].

Emerging data suggests utility for superparamagnetic iron oxide-enhanced MRI for the preoperative detection of sinusoidal obstruction syndrome in patients who have received neoadjuvant chemotherapy. In one report, 24 of 60 patients treated with neoadjuvant chemotherapy (predominantly 12 or more weeks of an oxaliplatin-containing regimen) were suspected of having moderate to severe sinusoidal obstruction on MRI, and 23 cases were confirmed at surgery [86]. The sensitivity, specificity, positive predictive value, and negative predictive value were 87, 89, 83, and 92 percent, respectively. However, whether patients who are identified noninvasively as having this pattern of liver damage after neoadjuvant chemotherapy can be more safely resected after a delay and how much of a delay is needed to allow healing of the liver is uncertain.

This MRI technique is not widely available.

Bevacizumab-related toxicity — Biologic agents such as bevacizumab have specific toxicities that could interfere with metastasectomy such as bleeding, impaired wound healing, and possibly impaired hepatic regeneration, although there is little evidence to support this concern [94-96]. The feasibility and safety of metastasectomy in this setting has been addressed in at least five retrospective series [95,97-100], none of which suggest an excess of problems with bleeding or wound healing in patients who received bevacizumab prior to resection. Whether bevacizumab impairs hepatic regeneration after portal vein embolization (which may be needed to increase the volume of the future hepatic remnant) is unclear; the available data are conflicting [96,101].

Because of the long half-life of bevacizumab (20 hours), it is commonly recommended that six to eight weeks should elapse between the administration of bevacizumab and elective hepatic resection [98,99,102,103]. As an example, in a preliminary report from the community-based BRiTE observational cohort of 622 patients who had surgery after bevacizumab, the incidence of serious wound complications in patients who had their last dose 0 to 13, 14 to 27, 28 to 59, or ≥60 days before surgery was 6.5, 3.4, 5.4, and 1.8 percent, respectively [103].

However, the available data correlating the incidence of postoperative complications and the time between last bevacizumab dose and surgery are conflicting:

• At least some data suggest that the interval between last bevacizumab dose and surgery can be shortened to five weeks without an increase in perioperative complications [95].
• Others have failed to find an association between postoperative complications and days from last bevacizumab dose to surgery (≤60 versus >60 days) [100].

The addition of bevacizumab to a neoadjuvant oxaliplatin-based regimen may reduce the incidence and severity of oxaliplatin-related liver injury [104,105]. This finding needs to be prospectively and independently validated.

Preoperative hepatic intraarterial chemotherapy — The rationale for administering chemotherapy into the hepatic artery (regional therapy) for hepatic metastases is discussed in detail below. (See 'Regional therapy' below.)

The benefit of regional rather than systemic therapy to downstage hepatic metastases is unclear. The following data are available:

• In a series of 22 patients with hepatic metastases from CRC who received HIA chemotherapy followed by exploration, 10 of 18 patients who received surgical treatment could be completely resected [106]. However, 15 of the 18 had developed recurrent disease at a median follow-up of 17 months and the remaining three died of other causes (one of postoperative complications) within seven months of surgery.
• A combined approach of hepatic intraarterial floxuridine plus systemic chemotherapy with oxaliplatin plus irinotecan was explored in a single institution study of 49 patients with initially unresectable CRC liver metastases [107]. Overall, 92 percent had either a complete or partial response rate to chemotherapy, and 23 of the 49 patients (43 percent) were able to undergo a later resection, 19 with negative margins. The median overall survival from pump placement for the entire cohort was 40 months.

While these results seem promising, this approach is not used by many clinicians outside of New York City. The only way to assess the contribution of intrahepatic arterial chemotherapy to neoadjuvant systemic chemotherapy in this setting is in a randomized controlled trial. Until such trials are completed, when neoadjuvant therapy is required, systemic rather than regional therapy is indicated.

Initially resectable liver metastases — For patients with initially resectable liver metastases, the question of whether perioperative chemotherapy improves survival was addressed in an EORTC trial in which 364 patients with up to four metastases without prior exposure to oxaliplatin were randomly assigned to liver resection with or without FOLFOX4 chemotherapy (oxaliplatin plus short-term infusional leucovorin-modulated 5-FU [the de Gramont regimen]) [68]. Six cycles (12 weeks) of chemotherapy were administered prior to surgery and six cycles were administered postoperatively. The key findings were as follows:

• Sixty-seven of the 182 patients assigned to chemotherapy had an objective response (four complete), while 11 progressed, eight of whom were no longer considered resectable. Overall, 83 percent of patients were successfully resected, similar to the number who were successfully resected in the surgery alone group (84 percent). Thus, the fear that initial chemotherapy would cause resectable liver metastases to become unresectable was not realized in this study. Furthermore, initial chemotherapy improved patient selection for hepatic resection. Among those who went directly to surgery, 18 of 170 (11 percent) had a nontherapeutic laparotomy, compared with only 8 of 159 (5 percent) in the initial chemotherapy group.
• The postoperative complication rate was significantly higher in the chemotherapy group (25 versus 16 percent). Patients receiving perioperative chemotherapy had higher rates of hepatic failure (7 versus 5 percent), biliary fistulas (8 versus 4 percent), and intraabdominal infection (7 versus 2 percent). However, the postoperative mortality rate was not higher than with surgery alone (1 versus 2 deaths).
• At a median follow-up of 3.9 years, there was a nonstatistically significant trend in three-year progression-free survival favoring the chemotherapy group (35 versus 28 percent, hazard ratio [HR] 0.79, p = 0.058). When ineligible patients were excluded from the analysis, the difference was statistically significant. Overall survival was not reported.

Thus, whether there is a net benefit for upfront chemotherapy as compared to immediate resection followed by adjuvant chemotherapy for patients with potentially resectable colorectal cancer liver metastases remains uncertain. Although we generally prefer immediate resection followed by chemotherapy for patients with potentially resectable disease, neoadjuvant chemotherapy is an acceptable approach, especially for borderline resectable disease [21].

The OncoSurge decision model for selecting patients for neoadjuvant therapy — As is evident from the above discussion, there are no widely accepted guidelines for determining which patients with colorectal cancer (CRC) liver metastases should undergo immediate surgery and when more complex care, such as neoadjuvant chemotherapy, is indicated. The increasing reports of liver injury following neoadjuvant chemotherapy as well as the data from the EORTC Intergroup trial 40983 have brought this issue to the forefront. (See 'Initially resectable liver metastases' above.)

Some, including the authors, recommend initial surgery rather than neoadjuvant chemotherapy for low-risk (medically fit, four or fewer lesions), potentially resectable patients, followed by a course of postoperative chemotherapy [81]. We consider neoadjuvant chemotherapy to be reasonable for those who have higher risk, borderline resectable or unresectable liver metastases.

A computer program has been developed to aid in selecting the best treatment option for individual patients with CRC liver metastases (the OncoSurge model) [22]. The model uses the RAND/University of California at Los Angeles (UCLA) appropriateness method (RAM), which has been used to evaluate other medical procedures and therapies. The RAM integrates a comprehensive literature review of clinical evidence with a subsequent consensus-based categorization of therapeutic options for specified case scenarios by a panel of experts [108].

A multidisciplinary panel of 16 experts in oncology, radiology, and liver surgery developed hypothetical patient profiles, and then rated the options of resection, chemotherapy, or local ablation using a health benefit-to-negative-consequence ratio on a scale of 1 to 9 (algorithm 1). Each panelist independently rated specific interventions as appropriate (expected benefit outweighs possible consequences), inappropriate (no health benefits), or uncertain (benefits and risks nearly equal, or panel members disagreed), and the group rating was taken as the median value of all participants.

The following represent the general recommendations from this panel of experts that were then incorporated into a decision matrix:

• Resection was always preferred, if possible, over local ablation strategies (cryosurgery, radiofrequency ablation [RFA], laser techniques).
• Resection was absolutely contraindicated in the presence of unresectable extrahepatic disease, extensive liver involvement (>70 percent, more than six segments, or involvement of all three hepatic veins), major liver insufficiency, or other conditions causing the patient to be unfit for surgery.
• Immediate resection was appropriate if adequate margins could be radiographically defined, there was no portal lymph node involvement, and four or fewer lesions. Resection could be considered for more than four lesions if they were localized to a single lobe.
• For patients with more than four metastases or bilobar involvement, resection was considered appropriate only after tumor shrinkage using neoadjuvant chemotherapy. The panelists considered 5-FU/leucovorin to be only rarely appropriate whereas 5-FU in combination with either irinotecan or oxaliplatin was generally appropriate. The panel did not address the utility of adding biologic agents (bevacizumab, a monoclonal antibody targeting vascular endothelial growth factor, or cetuximab, an anti epidermal growth factor receptor monoclonal antibody which is appropriate only for patients with Kras-wild-type tumors) to the upfront chemotherapy regimen. (See "Systemic chemotherapy for metastatic colorectal cancer: Completed clinical trials", section on 'Bevacizumab' and "Systemic chemotherapy for metastatic colorectal cancer: Completed clinical trials", section on 'Cetuximab'.)

Summary — Neoadjuvant chemotherapy has the potential to convert some patients with initially unresectable large or critically located liver metastases to resectable disease. The optimal selection criteria, specific regimen and duration of neoadjuvant chemotherapy, and the best way in which chemotherapy should be interdigitated with surgery in patients who present with synchronous metastatic disease have not been defined.

The growing number of reports describing liver toxicity and higher rates of perioperative morbidity in patients undergoing resection after receiving oxaliplatin or irinotecan-based neoadjuvant chemotherapy has somewhat tempered enthusiasm for this approach, particularly in patients with a small number of initially resectable liver metastases.

 We recommend the following approach:

• For low-risk (medically fit, four or fewer lesions), potentially resectable patients, initial surgery rather than neoadjuvant chemotherapy should be chosen, followed by postoperative chemotherapy.
• For patients who have higher risk, borderline resectable or unresectable disease, neoadjuvant chemotherapy is the preferred approach.

The optimal regimen has not been established. We consider the following combinations to have the highest objective response rates, and all are reasonable choices: FOLFOX with or without bevacizumab, FOLFOXIRI, or FOLFIRI plus cetuximab (for patients whose tumors lack K-ras mutations) (table 4). Guidelines from the National Comprehensive Cancer Network (NCCN) suggest any of the following regimens [109]:

• FOLFOX or CAPOX or FOLFIRI with or without bevacizumab or
• FOLFOX or CAPOX or FOLFIRI plus cetuximab (wild-type K-ras only)

Regardless of the specific regimen chosen, the duration of neoadjuvant chemotherapy should be limited, radiographic response assessment performed at approximately six-week intervals, and surgery undertaken as soon as the metastases become clearly resectable. Liver resection should be delayed at least four weeks after completion of chemotherapy, six to eight weeks if bevacizumab was a component of therapy. (See 'Liver toxicity from neoadjuvant chemotherapy' above.)

LOCAL OPTIONS FOR INCOMPLETELY RESECTED METASTATIC DISEASE — Radiofrequency ablation (RFA) or cryosurgery is sometimes applied following macroscopically incomplete resection of CRC liver metastases or if there are incidentally found small lesions that are surgically inaccessible. This subject is discussed in detail elsewhere. (See "Nonsurgical treatment strategies for colorectal cancer liver metastases", section on 'Tumor ablation'.)

For patients with a macroscopically complete resection in whom a tumor-free resection margin cannot be obtained because of vascular proximity or multinodularity, further locoregional therapy may not be necessary, particularly if an ultrasonic dissector is used for the hepatic resection [47]. This was shown in a collective series of 436 patients undergoing macroscopically complete, potentially curative resection for CRC liver metastases at three institutions; 234 were microscopically complete (R0), while 202 had microscopically positive margins (R1 resection) [110]. Further locoregional therapy was not given to those with an R1 resection (with the exception of argon beam or bipolar coagulation of the cut resection margin), but the majority (88 percent) had postoperative chemotherapy (as did 78 percent of the patients who underwent R0 resection).

At a median follow-up of 40 months, there were more intrahepatic recurrences among the patients who had an R1 resection, but the rate of recurrence at the surgical margin was not higher in this group. Furthermore, five- and ten-year overall survival rates were not dissimilar between the two groups (61 and 43 percent in the R0 group, versus 57 and 37 percent in the R1 group). In multivariate analysis, preoperative CEA level and major hepatectomy (≥3 segments) but not R1 resection were independent predictors of poor outcome.

These data support the view that the inability to obtain clear resection margins is not necessarily a strict contraindication to resection of liver metastases (see 'Patient selection' above. However, a complete (R0) resection must always be the goal of any resection. Current data do not support planning for surgical resection with foreknowledge that residual disease will be left behind.

TIMING OF HEPATECTOMY IN PATIENTS PRESENTING WITH METASTASES — A controversial issue is the timing of hepatic resection in patients who have liver metastases at initial presentation. Some reports indicate a poor prognosis in such patients, at least in part attributable to the failure to resect clinically occult micrometastatic liver disease [5,111].

In theory, delaying hepatic surgery by three to six months would permit the biological behavior of the metastatic disease to become evident, thus improving the selection of patients for whom hepatic metastasectomy might be curative. Delayed resection does not seem to increase the risk of patients becoming unresectable due to growth of the initial metastases [112], although if patients are untreated during this interval, it can increase the volume of resected liver, a significant predictor of postoperative complications.

One potential solution is to perform percutaneous RFA during the interval between diagnosis and hepatic metastasectomy [113]. Another is to administer chemotherapy during this period, an approach that might also allow some patients with initially unresectable or borderline resectable liver metastases to undergo successful later hepatic resection. (See 'Neoadjuvant chemotherapy' above and "Nonsurgical treatment strategies for colorectal cancer liver metastases".)

Simultaneous resection of the primary and metastatic disease is clearly preferable from the patient's perspective, and several surgical case series have failed to confirm inferior survival or greater morbidity for patients who undergo a one-stage procedure as compared to delayed (staged) hepatic resection, unless major hepatic resection (three or more segments) is needed [112,114-120]. One-stage surgery is probably a reasonable option for patients who present with low-volume (four or fewer, or all in the same lobe) resectable hepatic metastases. Because of the incidence of synchronous second primary colorectal cancers (approximately 3 to 5 percent), complete colonoscopy prior to surgical resection should be undertaken, if feasible. (See "Surveillance after colorectal cancer resection", section on 'Diagnosing second cancers and polyps'.)

On the other hand, if there are five or more simultaneous potentially resectable hepatic metastases (unless all are located in one lobe), bilobar involvement, or if disease is borderline resectable due to location, initial chemotherapy followed by reassessment and delayed resection is probably a better strategy. (See 'Neoadjuvant chemotherapy' above.)

THERAPY AFTER RESECTION OF LIVER METASTASES — As noted previously, there is a clear survival benefit from resection in patients with limited hepatic metastases from colorectal cancer (CRC). The role of systemic or regional therapy following metastasectomy is far less certain.

Systemic chemotherapy — A limited number of studies have explored the benefit of systemic 5-FU-based chemotherapy following resection of hepatic metastases; clear evidence of a survival benefit compared to observation alone has not yet emerged.

Two randomized trials of similar design were initiated in the early 1990s (the French FFCD 9002 and EORTC/NCIC trials), but both were closed prematurely because of slow accrual [121,122]. The FFCD trial randomly assigned 173 of a planned 200 patients to six months of postoperative systemic 5-FU and leucovorin (with both drugs administered for five consecutive days once per month for six months) versus observation alone following hepatic resection [122]. At five years, patients receiving chemotherapy had a significantly better disease-free survival, which was the primary endpoint (34 versus 27 percent at five years), but only a trend toward better survival (51 versus 41 percent).

The EORTC trial used the same chemotherapy regimen, but the results have not been published separately. In a combined analysis of both trials (totaling 278 patients), the differences in median progression-free survival (28 versus 19 months, p = 0.058) and overall survival (62 versus 47 months), while potentially clinical meaningful, were not statistically significant [121].

These data provide a proof of principle of the benefit of postresection adjuvant chemotherapy in this population. However, the chemotherapy used in these trials is considered inferior by modern standards. For patients with unresectable metastatic CRC, the introduction of newer drugs such as oxaliplatin, irinotecan, bevacizumab, and cetuximab has been accompanied by a marked improvement in median survival from six to seven months to 20 to 24 months. (See "Systemic chemotherapy for metastatic colorectal cancer: Completed clinical trials".)

There are limited data available from randomized trials on the benefits of these modern chemotherapy regimens following resection of colorectal cancer liver metastases [68,123]:

• A randomized trial from the EORTC evaluating perioperative FOLFOX chemotherapy (six cycles preoperatively, six postoperatively) versus observation alone in patients with initially resectable liver metastases showed that chemotherapy was associated with a trend toward improved three-year progression-free survival relative to surgical resection alone [68]. The difference was statistically significant when the analysis was restricted to eligible patients [68]. Survival was not reported. (See 'Initially resectable liver metastases' above.)

• The benefit of adding irinotecan to 5-FU and leucovorin was tested in a multicenter trial in which 321 patients undergoing complete surgical resection for liver-isolated metastatic disease were randomly assigned to short-term infusional 5-FU plus leucovorin every other week for 24 weeks without or with irinotecan (180 mg/m2 every other week) [123]. At a median follow-up of 42 months, there was no significant disease-free survival advantage for adding irinotecan (median 25 versus 22 months).

Of note, at least two trials of adjuvant irinotecan plus short-term infusional 5-FU have failed to demonstrate a progression-free survival benefit over 5-FU plus leucovorin alone in patients with resected stage II or III disease. Thus, administration of postoperative irinotecan-containing chemotherapy should not be considered a standard approach following resection of liver-isolated metastatic disease. (See "Adjuvant therapy for resected colon cancer", section on 'Irinotecan'.)

NCCN recommendations — Despite the paucity of data, updated guidelines from the NCCN recommend six months of an active systemic chemotherapy regimen for patients who have undergone resection of hepatic metastases from colorectal cancer [109]. Acceptable options include:

• FOLFOX or CAPOX or FOLFIRI with or without bevacizumab or
• FOLFOX or CAPOX or FOLFIRI plus cetuximab (wild-type K-ras only)

They also state that observation alone or a shortened course of chemotherapy is reasonable if the patient received neoadjuvant therapy.

In view of the newly available data regarding irinotecan plus 5-FU presented at the 2008 ASCO meeting [123], we would consider all of these regimens to represent acceptable options except FOLFIRI. However, based on indirect evidence, others consider FOLFOX to be the reference regimen for this group of patients [124].

Regional therapy — The liver is the dominant site of recurrence in over one-half of patients undergoing potentially curative resection. This observation, coupled with the proven efficacy of adjuvant systemic 5-FU-based chemotherapy in patients with node-positive colon cancer, led to the study of regional treatment.

Hepatic intraarterial chemotherapy alone — The fact that liver metastases derive their blood supply predominantly from the hepatic artery provides the rationale to apply regional intrahepatic arterial (HIA) chemotherapy following metastasectomy.

Despite initially encouraging data from small randomized studies conducted in fewer than 40 patients [125,126], larger randomized controlled trials were disappointing. In one American Cooperative Group study that randomized patients to resection only versus resection plus HIA FUDR for a solitary or multiple resectable liver metastases, patient accrual was inadequate to address the utility of adjuvant therapy [127].

A later German trial was closed prematurely when interim analysis suggested a worse outcome from HIA [128]. This study randomly assigned 226 patients to receive HIA 5-FU plus leucovorin or no treatment following hepatic resection of colorectal metastases, and treated patients had a worse median survival (35 versus 41 months), and similar median time to progression (14.2 versus 13.7 months) compared to those undergoing surgery alone.

The observation that many of these patients were failing outside of the liver led to efforts combining regional and systemic 5-FU-based chemotherapy.

HIA plus systemic therapy — The benefit of combined systemic and intrahepatic arterial chemotherapy was evaluated in an Intergroup study that randomly assigned completely resected patients to observation versus a combination of HIA FUDR and infusional 5-FU following resection. Unfortunately, only 109 patients were accrued during nine years [129]. In the final analysis, which was limited to 75 eligible patients (45 controls and 30 chemotherapy-treated patients), combined systemic and regional therapy was associated with significantly longer time to recurrence, a better four-year hepatic recurrence-free survival (67 versus 43 percent), and overall recurrence-free survival (46 versus 25 percent). However, there was no benefit in terms of median overall survival.

A slightly different approach was employed in a study of 156 patients with metastatic CRC who were randomly assigned to receive either six months of systemic therapy with leucovorin-modulated 5-FU or six cycles of combination HIA with FUDR plus systemic chemotherapy with 5-FU and leucovorin following surgical resection of liver metastases [130]. Combination therapy was associated with a significantly better two-year survival rate (86 versus 72 percent) and two-year freedom from tumor recurrence in the liver (90 versus 60 percent).

A later report of this series with over 10 years of follow-up suggested that benefit might be limited to the subgroup of patients with the highest risk for recurrence following resection alone [131]. Using the clinical risk score defined by Fong et al (assigning one point for each of the following: node-positive primary, disease-free interval from primary to metastases <12 months, more than one hepatic metastasis, largest hepatic tumor >5 cm, and CEA level >200 ng/ml [12]), outcomes were not significantly different with combined therapy versus surgery alone for those with a clinical risk score of 0 to 2 (median survival 83 months in both groups). However, outcomes were significantly better with combined therapy in those with a clinical risk score of 3 to 5 (median survival 60 versus 38 months, 10-year survival 39 versus 16 percent). This series demonstrates how well some patients can do with resection that is done in a center with substantial experience, and that regional therapy can improve the rate of hepatic tumor control.

More recent studies are exploring regimens that combine HIA chemotherapy and intravenous irinotecan and oxaliplatin [132-135]. Although treatment is tolerable and early results are promising, the ultimate proof of benefit will require a randomized controlled trial. Such a trial, NSABP C-09, is underway, comparing systemic capecitabine plus oxaliplatin alone or with alternating HAI FUDR after resection of colorectal cancer liver metastases.

Routine use of HIA chemotherapy after liver resection has not gained widespread acceptance. Placement of the HIA pump increases the complexity of the operation, and in one series, only 19 percent of the patients randomized to receive regional plus systemic chemotherapy completed the prescribed course of treatment [130]. Concerns about hepatobiliary toxicity of FUDR delivered by HIA combined with high response rates with modern systemic chemotherapy regimens will limit the use of HIA chemotherapy after hepatic resection. Nevertheless, updated guidelines from the NCCN indicate that hepatic arterial infusion with or without systemic 5-FU and leucovorin is a reasonable approach after liver resection at institutions with experience in both the surgical and medical aspects of this therapy [109].

HIA plus other local approaches — Other investigators are exploring the role of HIA FUDR-based chemotherapy in conjunction with partial debulking of liver metastases, either via surgical resection or with cryosurgery [136]. Until the results are known from such pilot studies, the use of HIA chemotherapy in that setting should be considered investigational.

Portal vein infusion — Because HIA FUDR carries a risk for biliary sclerosis, administration into the portal vein has been explored as an alternative. (See "Nonsurgical treatment strategies for colorectal cancer liver metastases", section on 'Optimizing chemotherapy'.) The rationale is based upon the observation that in contrast to clinically detectable metastases, which derive 90 percent of their blood supply from the hepatic artery, hepatic micrometastases (as well as the biliary tree) are primarily dependent on the portal vein for their blood supply. Like HIA infusion, portal vein infusion (PVI) carries with it a significant regional exposure advantage.

The potential benefit of adjuvant PVI with FUDR after resection or ablation of isolated hepatic metastases was evaluated in two trials conducted at the City of Hope Medical Center [137]. Systemic administration of 5-FU and leucovorin was given in conjunction with PVI FUDR, which was administered for 14 days on and 14 days off at a dose approximately twofold higher than that used with HIA FUDR. Although there was a low incidence of hepatic drug-induced toxicity, overall and disease-free survival (at three years 42 and 19 percent, respectively) were somewhat lower than has been reported with HIA FUDR and systemic 5-FU plus leucovorin following resection of hepatic metastases [129,130]. Thus, the role for this approach appears to be limited.

Systemic radioimmunotherapy — A somewhat different approach was exploited by investigators examining the utility of a radiolabeled humanized monoclonal antibody against the tumor marker carcinoembryonic antigen. This approach relies upon the anti-CEA antibody to target the radioimmunoconjugate to areas of tumor involvement, and "crossfire" radiation from the CEA-positive cancer cells targeted by the radiolabeled antibody to deliver tumoricidal doses to surrounding intrahepatic tumor cells.

An early report included 23 patients who received a single 30-minute infusion of 131-I-labetuzumab (two patients were retreated at the time of disease recurrence) after undergoing resection for colorectal cancer liver metastases [138]. With a median follow-up of 64 months, the median survival from the first liver resection was 68 months, and the five-year survival rate 51 percent. Treatment was well tolerated with the exception of transient myelosuppression.

Although these results seem promising compared to expected outcomes from partial hepatectomy alone, definitive proof of benefit will require a randomized trial.

The use of external beam radiotherapy and internal application of radiation therapy through the use of yttrium-labeled microspheres is discussed in detail elsewhere. (See "Nonsurgical treatment strategies for colorectal cancer liver metastases".)

REPEAT RESECTION FOR RECURRENT METASTASES — For patients who undergo hepatic resection for apparently isolated hepatic metastases, the liver is the most common site of recurrence; it is the only site of recurrence in approximately 40 percent [10,139]. Although randomized trials have not been conducted to prove benefit, repeat hepatic resection may be considered in selected patients who have no evidence of extrahepatic disease, and a good performance status. In several reported series, perioperative mortality rates were less than 5 percent, and relapse-free survival rates ranged from 20 to 43 percent at two to five years (table 5) [16,50,140-152].

Patients with a relapse-free interval of longer than one year appear to have a more favorable outcome from reresection [144]. Other factors associated with a poor outcome include synchronous resection for the first liver metastases, and the presence of multiple lesions at second hepatectomy [145,146,150].

SUMMARY AND RECOMMENDATIONS — The only potentially curative option for patients with liver-isolated metastatic colorectal cancer is surgical resection. For appropriately selected patients with four or fewer metastases, five-year relapse-free survival rates average 30 percent; in at least four contemporary series, five-year overall survival rates are approximately 58 percent [4,14,47,153].

The optimal selection of patients for resection is evolving. However, in general, surgical exploration should only be ruled out in the following situations (see 'Patient selection' above:

• Extensive unresectable extrahepatic disease as detected by CT and/or PET scans
• Radiographic evidence of involvement of the hepatic artery, major bile ducts, or main portal vein
• Extensive liver involvement (>70 percent, more than six segments (figure 1), or involvement of all three hepatic veins)
• Inadequate postresection functional hepatic reserve

We suggest not using a clinical risk score to select patients for diagnostic laparoscopy (Grade 2C). We perform an initial diagnostic laparoscopy only in patients with a suspicion of low-volume carcinomatosis based on preoperative radiographic imaging and for selected other cases at high risk for intraperitoneal metastatic disease (eg, a patient with a metachronous presentation of the primary and metastases who has several liver metastases that are not responding to chemotherapy). (See 'Selecting patients for diagnostic laparoscopy' above.)

We suggest immediate surgical resection for medically fit patients with four or fewer isolated hepatic metastases (Grade 2B). For patients with a good performance status who have more than four metastases (unless all are localized to a single lobe), radiographic suspicion for portal node involvement, or bilobar disease (ie, tumor involving any segments of the left and right hemi-liver), we suggest initial systemic chemotherapy followed by surgical reevaluation (Grade 2C). We prefer a bevacizumab plus oxaliplatin or irinotecan-based combination regimen.

Others disagree, instead recommending two to three courses of preoperative chemotherapy in patients with potentially resectable liver metastases, in part to select those patients who are most likely to benefit from resection [154]. However, if preoperative chemotherapy is selected, the number of courses should be minimized. Radiographic response assessment should be performed at six week intervals, and surgery undertaken as soon as the metastases are clearly resectable.

The optimal regimen in the neoadjuvant setting is not established. The following regimens have been associated with high objective response rates, and all are reasonable choices: FOLFOX with or without bevacizumab, FOLFOXIRI, or FOLFIRI plus cetuximab (for patients whose tumors lack K-ras mutations) (table 4).

Guidelines from the NCCN suggest two to three months of therapy with any of the following regimens [109]:

• FOLFOX or CAPOX or FOLFIRI with or without bevacizumab or
• FOLFOX or CAPOX or FOLFIRI plus cetuximab (wild-type K-ras only)

(See 'Choice of regimen' above.)

For patients who have hepatic metastases at initial presentation, the optimal timing of liver resection is uncertain. We suggest one-stage surgery, if feasible (Grade 2C). If not feasible, resection of liver metastases can follow six to eight weeks after resection of the primary tumor. (See 'Timing of hepatectomy in patients presenting with metastases' above.)

Following complete resection of liver metastases, the best postoperative strategy is uncertain. Although a modest survival benefit of combined hepatic intraarterial (HIA) and systemic chemotherapy compared to systemic leucovorin-modulated 5-FU alone was shown in a United States Cooperative Group trial, the control arm did not include either irinotecan or oxaliplatin, two agents that clearly add benefit to 5-FU and leucovorin in the setting of advanced disease. Thus, the relative benefit and indications for HIA chemotherapy in patients with hepatic metastases from CRC remains uncertain. Although this approach is followed in some institutions, we suggest that it not be used (Grade 2C) (see 'HIA plus systemic therapy' above.

In the absence of published randomized trials to guide clinical practice following metastasectomy, we suggest a six month course of systemic chemotherapy containing oxaliplatin (Grade 2C). We recommend against the use of an irinotecan-based regimen in this setting (Grade 1B). (See 'Therapy after resection of liver metastases' above.)

Guidelines from the NCCN suggest a shorter course of postoperative therapy or observation alone for patients who have received neoadjuvant chemotherapy, although this recommendation is based more on expert opinion than prospective data [109].