Long-Term Results

Since the studies of Wood and Wagner and colleagues, numerous large, single institutional and multicenter reports from around the world have demonstrated that liver resection is safe and results in long-term survival (Table 81A.2). In recent series, 5-year survival rates approach 40%, and median survival exceeds 40 months (Fong et al, 1999; Choti et al, 2002); this represents a dramatic improvement from the seminal study by Foster (1978) on liver resection, when the reported 5-year survival was 20%.

Based on the favorable data reported by these studies, liver resection has now become the standard treatment for metastatic lesions from colorectal primaries. Even in the premodern chemotherapy era—in studies from the 1990s, before FOLFOX and FOLFIRI were given routinely—long-term survival in patients undergoing resection for metastatic disease far exceeded the survival of unresected patients. This improvement in survival was striking, even in patients with unresectable, low-volume metastatic disease: 20% to 40% 5-year survival with metastasectomy versus less than 10% for patients with unresected solitary metastases.

With newer multimodal treatments and careful patient selection, it is anticipated that 5-year survival approaching 70% can be achieved after resection (Nikfarjam et al, 2009), and comparable outcomes are likely to be reported in larger studies in the near future. Many patients with CRC at stage IV are cured by surgery. A recent review of actual 10-year survivors indicated that one third of actual 5-year survivors will eventually succumb to a cancer-related death; however, 10-year survival is tantamount to a cure in all but the rarest cases (<1%) (Tomlinson et al, 2007).

No prospective, randomized trials have been done that directly compare surgery with nonoperative therapy for metastatic CRC, and such a trial is unlikely to ever occur. The strikingly favorable outcome for surgical patients in retrospective studies compared with patients managed nonoperatively renders such a trial unethical.

Outcome and Surgical Volume

A number of studies have correlated perioperative outcome to hospital volume for hepatectomy. Choti and colleagues (1998) and Glasgow and colleagues (1999) used the state registries in Maryland and California, respectively, and found that hepatectomy performed at a high-volume center resulted in improved outcomes. Surgery at high-volume centers was also associated with reductions in perioperative mortality, length of hospital stay, and cost. A recent study by Fong and colleagues (2005) based on a national Medicare database provides further support for superior outcomes at high-volume centers. The authors demonstrated that perioperative outcome correlated with the expertise of the institution, and the survival advantage persisted after the perioperative period. These data lend support to the concept of regionalization of hepatectomy to high-volume centers.

Characteristics of the Patient and Primary Tumor

Age has not been found to be a significant prognostic variable for long-term survival, although analyses examining the role of age are confounded by the fact that older patients are carefully selected for surgery (Cady et al, 1992; Ballantyne & Quin, 1993). The location of primary cancer has not been shown to affect outcome, as metastatic rectal and colon cancer have similar prognoses (Doci et al, 1991). On the other hand, the stage of the primary tumor is quite useful in staging metastatic disease. Patients who had positive lymph nodes associated with their primary CRC do worse, as do patients who present with synchronous liver disease. Histologic grade of the primary cancer is not a significant predictor of long-term survival in patients with metastatic disease (Doci et al, 1991; Ballantyne & Quin, 1993).

Clinical Characteristics of Liver Metastases

The most important clinical predictor of outcome related to the liver metastases themselves is the disease-free interval: the shorter the interval, the worse the prognosis (Rosen et al, 1992; Scheele et al, 1995). Other negative factors include multiple liver metastases—four lesions is a commonly used cutoff—bilateral disease, tumor size greater than 5 cm, and a markedly elevated preoperative carcinoembryonic antigen (CEA >200 ng/mL; Nordlinger et al, 1996).

The prognostic value of response to neoadjuvant chemotherapy is controversial, and opinions vary in the literature. In a study by Adam and colleagues (2004), the 5-year disease-free survival was so poor following progression during neoadjuvant chemotherapy—3% versus 20% in patients who had a response or stable disease—the authors proposed that such a finding should preclude liver resection. This phenomenon was confirmed in our own experience (Allen et al, 2003). Although other groups have not found this to be true, such studies were based on patients who received older chemotherapy regimens. Roughly half of the patients only received 5-FU, which might account for the high rate of progression (Gallagher et al, 2009; Neumann et al, 2009). With modern chemotherapy, the expected rates of progression would be less than 20% compared with 60% in the studies mentioned above that used 5-FU (Nordlinger et al, 2008).

Operative and Pathologic Characteristics of Liver Metastases

Higher operative blood loss and transfusion requirements have been found to be associated with an increased number of perioperative complications and mortality but not with a worse long-term survival (Table 81A.3; Kooby et al, 2003). Anatomic resections are believed to indirectly improve long-term outcomes, because this approach is associated with a significantly lower rate of positive margins. In a retrospective analysis from our institution, patients who had an anatomic segmental resection had a better outcome than patients subjected to wedge resections (DeMatteo et al, 2000). In an anatomic resection, planes are developed along major hepatic veins, and inadvertent disruption of the tumor is avoided. Wedge resections are usually guided by palpation and can be complicated by inadequate exposure or retraction. The attendant problems of poor visibility and impaired tactile sense at the depth of the resection will often lead to tearing of the specimen along the hard tumor–soft liver interface, thus resulting in a positive margin.

Perhaps the most informative surgical prognostic factor is the finding of extrahepatic disease (Rosen et al, 1992; Fong et al, 1999; Abdalla et al, 2006). The existence of nonpulmonary extrahepatic disease has long been considered by many surgeons to be a contraindication to hepatectomy (Fong et al, 1999; Abdalla et al, 2006); however, the approach to extrahepatic disease has been reevaluated in recent years, as some studies have demonstrated acceptable survival in selected patients with limited extrahepatic disease managed with resection, particularly in the setting of low-volume pulmonary metastases. The two largest studies on this question evaluated combination resection of liver and lung metastases and noted 5-year survival rates of 30% (Headrick et al, 2001; Miller et al, 2007).

The presence of metastatic disease to the portal lymph nodes remains a particularly poor prognostic factor. Portal metastases are found in 3% to 30% of patients with metastatic CRC to the liver (Abdalla et al, 2006), and disease recurs in virtually all patients with resected portal lymph nodes (Carpizo et al, 2009). A consensus paper on the management of extrahepatic disease concluded that no convincing data suggest that therapeutic portal lymphadenectomy for metastatic CRC improves survival (Abdalla et al, 2006). In contrast to portal lymph node metastases, there appears to be a role for hepatic resection in conjunction with metastases to other sites. The question has been examined in the form of large retrospective studies that included resections of the peritoneum or mesentery, retroperitoneal lymph nodes, and a variety of solid organs (Minagawa et al, 2000; Elias et al, 2003; Carpizo et al, 2009). The 5-year survival rates reported in these series range from 20% to 28%. Based on these studies, hepatectomy combined with surgical resection of extrahepatic disease is believed to be appropriate in selected patients.

A macroscopic positive margin, or R2 resection, is widely accepted as a poor prognostic factor. The presence of microscopic disease at the resection margin (R1 resection) or the precise thickness of the negative margin (R0 resection) is perhaps more controversial. For instance, the preponderance of data on the impact of an R1 resection suggests that a microscopically positive margin is associated with worse survival. The 5-year survivals are consistently above 30% for negative margins but are 20% or less with microscopic disease at the margin (Fong et al, 1999; Pawlik et al, 2005; Nuzzo et al, 2008); however, multivariate analyses that include R1 resection status often do not find it to be a significant predictor of outcome after adjusting for other competing risk factors, which suggests that a positive microscopic margin is merely a marker of poor biology (Pawlik et al, 2005; Nuzzo et al, 2008). As adjuvant therapy improves, patients with R1 resections will likely have outcomes that more closely approach patients with negative resection margins. Indeed, a recent report suggests that a microscopic positive resection margin obtained out of necessity does not negatively impact the 5-year survival, although the recurrence rate is increased (de Haas et al, 2008).

With regard to the extent of negative margin, a 1-cm rim of normal liver parenchyma surrounding a resected tumor was traditionally espoused in the older literature (Shirabe et al, 1997); however, recent studies suggest that subcentimeter margins are acceptable (Are et al, 2007) and are perhaps even associated with equivalent outcomes to R0 resections (Scheele et al, 1995; Pawlik et al, 2005; Hamady et al, 2006; Figueras et al, 2007). Low rates of micrometastastic disease, or satellitosis, were found in resection specimens in two Japanese studies, further supporting this idea (Yamamoto et al, 1995; Kokudo et al, 2002). Anatomic segmental resection, as opposed to wedge resections (DeMatteo et al, 2000), and neoadjuvant chemotherapy are two proposed strategies to improve R0 resection rates in patients with colorectal liver metastases (Parikh et al, 2003).

In summary, the prospect of a positive microscopic or close microscopic margin is not a contraindication to liver metastasectomy in appropriately selected patients with modern chemotherapy. In addition to a macroscopically positive resection margin, the findings of hepatic satellite lesions and intrabiliary invasion have also been shown to predict tumor recurrence (Gayowski et al, 1994; Scheele et al, 1995; Okano et al, 1999; Povoski et al, 2000; Kubo et al, 2002).

Predictive Models and Clinical Risk Scores

Four large studies have enabled robust multivariate analyses of prognostic factors with metastatic CRC and design of useful predictive models for favorable survival after metastasectomy (Nordlinger et al, 1996; Fong et al, 1999; Kattan et al, 2008; Rees et al, 2008). Nordlinger and colleagues (1996) reported on a multicenter series of more than 1500 patients. Fong and colleagues (1999) reported on a single institutional series of 1001 patients and later on a cohort of 1477 patients (Kattan et al, 2008), and Rees and colleagues (2008) evaluated long-term survival in 929 patients from a tertiary referral center in the United Kingdom.

In the series by Fong and colleagues (1999), seven parameters were found to be independent predictors of an unfavorable prognosis: 1) the presence of extrahepatic disease, 2) a positive resection margin, 3) nodal metastases associated with the primary CRC, 4) a short disease-free interval, 5) largest liver metastasis greater than 5 cm, 6) more than one liver metastasis, and 7) a serum CEA greater than 200 ng/mL. The presence of extrahepatic disease and a postive resection margin are generally determined intraoperatively. When presumed preoperatively, these two variables are often considered relative contraindications to metastasectomy. A preoperative clinical risk score (CRS) system was therefore created using the last five factors (Box 81A.1), with each positive criterion counting as one point. This CRS is a simple and facile staging system used to classify patients with liver-exclusive metastatic CRC (Fig. 81A.1).

Box 81A.1 Prognostic Scoring System for Hepatic Colorectal Metastases

Node-Positive Primary Tumor

Disease free interval <12 months between colon resection and appearance of metastases

Size of largest lesion >5 cm

>1 Tumor

Carcinoembryonic antigen >200 ng/dL

Sum of points, with a point assigned for each positive criterion

FIGURE 81A.1 Survival as related to clinical risk score after hepatic resection (P < .00001). Open box: score = 0 (n = 52); solid triangle: score = 1 (n = 262); open circle: score = 2 (n = 350); solid circle: score = 3 (n = 243); solid box: score = 4 (n = 80); open triangle: score = 5 (n = 14).

Copyright 1999 Lippincott Williams & Wilkins. All rights reserved. Reprinted with permission. (Fong Y, et al, 1999: Clinical score for predicting recurrence after hepatic resection for metastatic colorectal cancer: analysis of 1001 consecutive cases. Ann Surg 230[3]:314.)

The presence of any one of these characteristics was still associated with a 5-year survival of 24% to 41%; therefore no single criteria can be considered a contraindication to resection; rather, the total score out of 5 is highly predictive of outcome, with a score of 2 or less suggestive of a particularly favorable prognosis—the optimal candidate for liver metastasectomy. Patients with a CRS of 3 or 4 have less favorable outcomes and may be appropriate for clinical trials involving adjuvant chemotherapy. Long-term disease-free survival is rare in patients with a CRS of 5, and these patients are also typically best served in clinical adjuvant chemotherapy trials. Importantly, actual 10-year survival has been observed in patients with even the highest CRS, as patients with a CRS of 3 to 5 have a 10% disease-specific survival (DSS). This is roughly equal to the cure rate of treatment in these patients; therefore a high CRS should not be considered a contraindication to hepatic resection (Tomlinson et al, 2007).

Preoperative Investigations

The selection of patients for hepatic metastasectomy depends upon patient operability and tumor resectability. The criteria for patient operability are similar to those considered for any major laparotomy. A significant history of cardiac and pulmonary disease must be investigated, as these patients are at significant risk for perioperative complications. Pulmonary compromise as a result of a long upper abdominal incision and the development of a right sympathetic pleural effusion may predispose to pneumonia. In addition, dramatic intravascular fluid shifts and electrolyte imbalances may occur perioperatively that can precipitate cardiac ischemia and arrhythmia. Finally, any previous liver disease that might have impaired hepatic function should be evaluated, as this will determine the volume of liver than can be safely resected.

Preoperative investigations prior to resection of metastatic colon cancer are directed at 1) establishing the diagnosis, 2) anatomic definition of the liver lesion for surgical planning, and 3) staging to rule out extrahepatic disease. A confirmatory biopsy of hepatic lesions is only indicated to confirm the diagnosis when the clinical picture is unclear. The differentiation between metastatic tumors and benign hepatic lesions can usually be done with imaging modalities, including ultrasound (US), magnetic resonance imaging (MRI), and PET scan as discussed below (see Chapters 13, 15, and 17). The risk of tract seeding from percutaneous fine needle aspiration appears to be quite small, with only a few case reports in the literature. The workup to determine the extent of disease includes imaging, as outlined below, and recent colonoscopy (within 6 months). Although a detailed discussion of each imaging modality is beyond the scope of this chapter, the following sections will focus on the practical aspects of imaging for preoperative workup of patients with hepatic colorectal metastases.

Computed Tomographic Scans

Computed tomographic (CT) scans have become indispensable in the staging of patients with metastatic CRC (see Chapter 16). A contrast-enhanced CT scan of the chest, abdomen, and pelvis are routinely obtained, although the yield from scanning the chest with regard to detecting metastatic disease is limited to just a few percent (Kronawitter et al, 1999). The most important images in the evaluation of the liver in the setting of metastatic CRC are obtained in the portal venous phase, because the lesions are not typically well vascularized. Arterial phase images are principally used to distinguish metastatic disease from benign vascular lesions, such as hemangiomas, or to better define the arterial anatomy of the liver. Such a step is necessary when planning for the placement of hepatic arterial infusion pumps. Standard CT scanning involves a triple-phase protocol through the liver that includes arterial, portal venous, and delayed phases. Slices are usually 5 to 7 mm thick through the liver, but these are reduced to 1.5-mm thick overlapping slices when high-resolution hepatic angiography is desired. Coronal and sagittal reconstructions are performed to further delineate the anatomy.

Metastatic lesions from colorectal primaries tend to push structures away, rather than invade directly, respecting the liver capsule and intersegmental planes (Baer et al, 1989; Scheele, 1989). Even when large lesions appear to involve the inferior vena cava on preoperative imaging, they often do not on surgical exploration; therefore such a finding on CT should not preclude an operation. Diaphragmatic invasion is a rare event as well (Weinbren & Blumgart, 1986).

Magnetic Resonance Imaging

MRI is most useful to evaluate indeterminate hepatic lesions and to better define the relationship of tumors to the hepatic vasculature and biliary tree using magnetic resonance cholangiopancreatography (MRCP; see Chapter 17). In particular, MRI is excellent at distinguishing between benign lesions such as hemangiomas, adenomas, and fibronodular hyperplasia. Moreover, MRI is better than CT for investigating pathology in the background of a fatty liver, such as in patients who have a history of obesity, diabetes, alcoholism, or previous chemotherapy treatment (Bipat et al, 2005; Fernandez et al, 2005). Because MRI is much more expensive than CT, this modality is not used routinely; however, it is warranted to better characterize indeterminate lesions or for the patient with an elevated CEA and a fatty liver with no evidence of disease on CT.

Ultrasonography

Ultrasonography is a relatively inexpensive test that can provide detailed information about the number, extent, and anatomic relationships of liver metastases (see Chapter 13). Duplex US can define the proximity of the tumor to the portal vessels, hepatic veins, and inferior vena cava, and US is particularly useful for characterizing cystic lesions within the liver. In expert hands, it can be very sensitive at detecting small metastatic lesions, although the test is highly operator dependent.

Hepatic Angiography and Computed Tomographic Portography

Direct hepatic angiography has largely been replaced by CT and MR angiography, which are less invasive alternatives (see Chapter 19). The main role for direct angiography in the modern era of high-resolution cross-sectional imaging is CT portography. In this technique, contrast is administered directly into the splanchnic bed by injection into the superior mesenteric artery, and a CT portal venogram of the liver is acquired after a suitable delay. This technique is highly sensitive at detecting small metastatic tumors (Sica et al, 2000); however, because the test requires arterial catheterization and is associated with a relatively high false-positive rate because of perfusion abnormalities, its primary role is in the setting of a planned portal vein embolization (PVE; see Chapter 93A, Chapter 93B ).

Positron Emission Tomography

The role of whole-body positron emission tomography as an adjunct to CT in the evaluation of metastatic CRC has been growing in recent years (see Chapter 15). In the United States, the use of FDG PET for staging and diagnosis of CRC was approved by Medicare in 2001 (Kelloff et al, 2005). The imaging test utilizes an intravenously administered radioactive tracer, which in most cases is ¹⁸F-FDG. The radioactive glucose analogue cannot proceed down the glycolytic pathway and therefore accumulates within glucose-avid cancer cells. PET capitalizes on this phenomenon, as ¹⁸F-FDG reveals the distribution of metabolically active metastases overexpressing cell surface GLUT1 transporters (Fig. 81A.2). The limitations of PET include poor sensitivity for lesions less than 1 cm, that some larger lesions are not FDG avid, and that the test lacks a high degree of anatomic detail. Moreover, false positives occur in the setting of inflammation, which may exist early after surgery and radiotherapy, or in the context of an infectious process. Integration of FDG PET imaging and CT addresses some of these limitations, particularly with regard to anatomic details.

FIGURE 81A.2 A patient with two potentially resectable hepatic metastases in the right lobe of the liver. An FDG PET scan identified abnormal uptake in the left ligula of the lung, which was biopsied and found to represent an extrahepatic colorectal cancer metastasis.

PET has emerged as a relevant test in several areas of metastatic CRC management. Reported applications include patient selection for metastasectomy, evaluation of a patient with a suspected recurrence, radiotherapy planning, response assessment, and the discovery of incidental colorectal lesions (Herbertson et al, 2007). With regard to the initial staging of patients with metastatic CRC, FDG PET imaging leads to a change in management in 2% to 36% of patients. For instance, Strasberg and colleagues (2001) examined the utility of PET in evaluating patients with hepatic colorectal metastases for possible hepatic resection. Out of 43 patients, the authors identified 6 with unresectable disease detected by PET imaging, which amounted to a 14% rate of occult disease detection by PET. Futile surgery was avoided in these patients. This improved detection rate is comparable to the experience described above by Schussler-Fiorenza and colleagues (2004) in the section on uses of the CRS. Interestingly, Strasberg and colleagues (2001) report quite favorable outcomes after metastasectomy, which they attribute to improved patient selection from the use of preoperative PET imaging: the resectability rate, 3-year disease-free survival, and 3-year overall survival were 95%, 40%, and 77% respectively. A meta-analysis on the subject found that the sensitivity and specificity of FDG PET were 91.5% and 95.4% for extrahepatic disease compared with 60.9% and 91.1% for CT (Wiering et al, 2005). The advantage of FDG PET over CT alone in the evaluation of a patient with a suspected recurrence, in the setting of a rising serum CEA, is similar to the experience described above for the patient undergoing a complete staging workup for stage IV disease (Huebner et al, 2000). FDG PET may also be particularly useful at identifying omental or peritoneal disease. Finally, the prospect of using functional imaging to assess treatment response is a promising application of PET technology. For instance, multiple studies have demonstrated that early metabolic response to chemotherapy correlates with eventual RECIST response, based on CT imaging, for both CRC and other cancer types (Findlay et al, 1996; Weber et al, 2003; Juweid & Cheson, 2006). One can imagine treatment adjustments based on FDG PET scans after just a few cycles of chemotherapy.

The National Comprehensive Cancer Network (NCCN) guidelines do recommend that an FDG PET scan be considered in the follow-up of a patient with CRC in the setting of CEA elevation and suspected recurrence and also in the setting of a resectable synchronous or metachronous liver metastasis. The utility of preoperative staging with FDG PET in metastatic CRC is the focus of a currently accruing randomized Phase III trial (PET START). Enrollment for this 400-patient study is nearing completion, with results expected in the near future.

Intraoperative Staging and the Role of Intraoperative Ultrasound (See Chapters 21 and 95)

As with any laparotomy for cancer, the abdomen should be carefully explored for evidence of extrahepatic metastases. In particular the celiac axis and portocaval and hilar lymph nodes must be palpated, and any suspicious nodes should be removed and examined by frozen section. Most surgeons routinely use intraoperative US after mobilization of the liver. With experience this practice may detect 5% to 10% of lesions missed on noninvasive imaging with CT, MRI, and transabdominal US (Boldrini et al, 1987; Machi et al, 1987, 1991; Olsen, 1990; Stone et al, 1994). In addition, intraoperative US is useful to delineate the interior anatomy of the liver, such as the intrahepatic vessels; therefore hepatic resection can be performed more safely with an anatomic approach. Intraoperative US is also useful throughout parenchymal transection, so that the resection line can be viewed in relation to the lesion and blood vessels. Prior to the development of PET imaging, several reports demonstrated that surgical management of hepatic tumors was influenced by intraoperative US in 30% to 50% of operations (Castaing et al, 1986; Rifkin et al, 1987).

Recently, interest is mounting with regard to contrast-enhanced intraoperative liver US with contrast agents that consist of microbubbles of air or gas that create greater contrast between tumors and normal liver parenchyma. Typically, the contrast is delivered in the form of a 4-mL intravenous bolus of SonoVue, sulfur hexafluoride gas stabilized by a phospholipid shell (Bracco SpA, Milan, Italy), followed by a flush of normal saline. In a series of 60 patients from the United Kingdom with metastatic CRC, the acccuracy of preoperative MRI/CT, standard intraoperative US, and contrast-enhanced US were 74%, 79%, and 96% respectively. The median size of lesions identified on contrast-enhanced US, but not on preoperative imaging, was 0.8 cm. In 17 patients (29%), findings on intraoperative contrast-enhanced US changed the surgical plan (Leen et al, 2006). In a second study from the United Kingdom, contrast-enhanced US altered the management plan in 4 (19%) of 21 patients compared with standard intraoperative US (Shah et al, 2010).

Immediate Postoperative Care (See Chapter 22)

Metabolic derangements are common in the postoperative period following liver resection and are due to both hepatic insufficiency and hepatic regeneration. Hypoproteinemia and hypoglycemia may indicate early hepatic insufficiency, but they are usually transient. Hypokalemia and hypophosphatemia occur on the second postoperative day, as the liver begins to consume these electrolytes for regeneration. Such metabolic abnormalities should be treated supportively. We prefer maintenance intravenous fluid with 5% dextrose and half-normal saline with the addition of 15 to 30 mEq/L of potassium phosphate. Alkaline phosphatase and transaminase elevations are common, and hyperbilirubinemia may occur. Abnormalities in coagulation or platelets may also occur and should be treated when the international normalized ration (INR) exceeds 1.5, or the platelets are less than 50,000/µL. Hepatic regeneration occurs rapidly after surgery. DNA synthesis is initiated in hepatocytes within 10 to 12 hours after surgery, and most of the hepatic mass returns within 1 week. Subsequent architectural rearrangements continue over the next few weeks, and in the majority of cases, this process is complete in 6 weeks.

Perioperative Mortality and Morbidity

For a surgical approach to be widely accepted, it must be efficacious, safe, and feasible. Without a doubt, major liver resection can be performed with acceptable mortality and morbidity. Using proper patient selection, preoperative staging, and surgical expertise, metastatic hepatectomies can be performed safely and result in prolonged disease-free survival and cure.

Follow-up After Resection

No consensus exists regarding the extent and frequency of follow-up after surgery for primary CRC, as no specific strategy has been shown to conclusively improve survival (Makela et al, 1995; Ohlsson et al, 1995; Kjeldsen et al, 1997; Schoemaker et al, 1998). Similarly, no conclusive data demonstrate improved survival with close follow-up after hepatic resection for CRC metastases (Metcalfe et al, 2004). Nevertheless, patients who have undergone hepatic resection are usually monitored closely with the goal to identify early and potentially curable resectable recurrences. Our paradigm is as follows: physical examination, serum CEA level, and CT of the abdomen and pelvis every 3 to 4 months for the first 2 years and then every 6 months thereafter, and patients undergo chest imaging annually. We continue surveillance for more than 5 years, because roughly a third of 5-year survivors will still recur (Tomlinson et al, 2007).

Patterns of Recurrence

Although the cure rate for liver metastasectomy is now considered to be roughly 20% (Tomlinson et al, 2007), the majority of patients with metastatic CRC to the liver die from recurrent disease. It stands to reason that these patients harbored occult disease at surgery that ultimately progressed and led to their demise. The most common sites for failure include the liver and lung (Table 81A.5). Indeed, the recurrences occur within the liver in 60% of patients (de Jong et al, 2009) and exclusively in the liver in up to 40% (Nordlinger et al, 1987).

Adjuvant Systemic Chemotherapy

A number of nonrandomized studies have evaluated the utility of perioperative systemic chemotherapy in the setting of hepatic resection for CRC metastases with equivocal results. Most of these studies compared patients receiving 5-FU–based chemotherapy with or without irinotecan or oxaliplatin. A large retrospective cohort study from European and American centers compared 792 patients who underwent hepatic resection alone to hepatic resection with perioperative 5-FU chemotherapy. The basis for receiving chemotherapy was largely a function of where the patients were treated, thereby removing the bias toward chemotherapy in higher risk patients, which confounds other studies (Hewes et al, 2007). In this study, chemotherapy was associated with an improved 5-year overall survival (37% vs. 31%; Parks et al, 2007). Postoperative systemic 5-FU chemotherapy was evaluated in two prospective, randomized, controlled trials, the Fédération Francophone Cancerologie (FFCD) trial 9002 and the ENG trial, but both studies were closed early as a result of poor accrual. The pooled results included 278 patients randomized to surgery alone or surgery with 6 months of 5-FU. The authors demonstrated a marginal improvement in PFS (18.8 vs. 27.9 months; P = .058) without an improvement in overall survival at 3 or 5 years (Mitry et al, 2008).

More recently, Nordlinger and colleagues (2008) reported the results of the European Organization for the Research and Treatment of Cancer (EORTC) intergroup trial 4098. This prospective and randomized trial included 364 patients with resectable hepatic colorectal metastases who were treated by surgery alone or surgery with perioperative FOLFOX chemotherapy, six preoperative cycles and six postoperative cycles (Nordlinger et al, 2008). Based on an intention-to-treat analysis, an insignificant improvement was seen in PFS at 3 years of 7.3% (28.1% vs. 35.4%; P = .058); however, when only eligible patients were analyzed, the absolute difference in 3-year PFS was 8.1% and reached statistical significance (P = .041). The difference was 9.2% when just patients who underwent surgical resection were included (P = .025). This was technically a negative study, because the groups were statistically equivalent in the intent-to-treat analysis, but this is frequently quoted as a study demonstrating that perioperative chemotherapy improves PFS in patients undergoing hepatic resection for CRC metastases. Furthermore, one cannot comment on the role of neoadjuvant therapy from this study, because patients received chemotherapy before and after surgery in the adjuvant arm. No prospective and randomized study to date has examined the role of newer, biologic agents in the adjuvant setting in patients with resectable metastatic CRC.

Adjuvant Hepatic Arterial Infusion Chemotherapy (See Chapter 86)

Although regional liver-directed chemotherapy is not universally used, some liver surgeons and gastrointestinal oncologists support a role for this strategy in the setting of metastatic CRC to the liver. The rationale for this approach is that the liver is the most common site for tumor recurrence after liver resection, and it is often the only site of recurrence. Moreover, most metastatic liver tumors preferentially derive their blood supply from the hepatic artery, as opposed to normal hepatic tissue, which relies on the portal venous blood supply. In addition, the hepatic parenchyma can extract and metabolize chemotherapy drugs to less toxic byproducts, which allows the administration of very high doses to tumor cells with minimal systemic toxicity. For a more detailed discussion of the technical aspects of regional chemotherapy, refer to Chapter 86.

The most extensively studied agent used in hepatic artery infusional therapy is 5-FU-2-deoxyuridine (FUDR). This agent is a 5-FU analogue, which is concentrated 100 to 400 times in the liver as a result of its 95% hepatic extraction ratio (Ensminger et al, 1978). Most of the data on regional hepatic chemotherapy focuses on unresectable metastatic CRC. Randomized trials have clearly demonstrated that the technique is safe and is associated with a higher response rate than systemic chemotherapy alone; responses are 48% to 62% versus 0% to 21%, respectively (Chang et al, 1987; Kemeny et al, 1987; Hohn et al, 1989; Martin et al, 1990; Rougier et al, 1992); however, improved treatment response has translated into an improved survival in just a single study (Chang et al, 1987). Four relatively large randomized trials of adjuvant regional chemotherapy have been completed, and three were positive trials in favor of regional chemotherapy (Kemeny et al, 1999; Lygidakis et al, 2001; Kemeny et al, 2002). The fourth trial by Lorenz and colleagues (1998) found no difference in outcome for patients treated with resection alone versus resection plus regional 5-FU chemotherapy in an intention-to-treat analysis; however, the patients randomized to regional therapy had a very high operative mortality (8%), and 5-FU was used instead of FUDR, which has a much higher hepatic extraction rate and therefore less systemic toxicity. In addition, implanted ports were used as opposed to pump devices. Technical complications prevented chemotherapy administration in many of the patients randomized to regional therapy, so that just 74% and 30% of patients in the experimental arm initiated and completed treatment, respectively. These findings highlight the importance of technical expertise with regional liver chemotherapy (Lorenz et al, 1998).

Two of the positive adjuvant trials were large American studies. The Eastern Oncology Cooperative Group/Southwest Oncology Group (ECOG/SWOG) study consisted of 109 patients randomized to surgery alone or surgery plus hepatic arterial infusion (HAI) with FUDR. The 4-year hepatic disease-free survival was significantly better in the HAI group (67% vs. 43%; P = .03) (Kemeny et al, 2002). No difference in overall survival was found, although the study was insufficiently powered to test this end point. In a larger single-institution study from Memorial Sloan-Kettering Cancer Center (MSKCC), 156 patients were randomized to resection plus systemic 5-FU or resection plus combined systemic 5-FU and HAI with FUDR. The patients who were treated with regional therapy had a significantly higher 2-year survival (86% vs. 72%; P = .03) and markedly improved liver disease control, and 10-year follow-up demonstrated significantly improved PFS associated with regional therapy (31.3 vs. 17.2 months; Fig. 81A.3). Survival rates at 10 years were 41.1% and 27.2% in the two groups, although the difference was not statistically significant (P = .1) (Kemeny & Gonen, 2005).

FIGURE 81A.3 Overall survival among patients with metastatic colorectal cancer treated with hepatic arterial infusion plus systemic chemotherapy (combined therapy) or systemic chemotherapy alone (monotherapy).

(From Kemeny N, Gonen M, 2005: Hepatic arterial infusion after liver resection N Engl J Med 352[7]:734.

Future prospective trials should explore regimens combining HAI chemotherapy with FUDR and modern systemic chemotherapy with agents such as irinotecan (CPT-11), oxaliplatin, bevacizumab, and cetuximab. Phase I and II studies combining systemic CPT-11 and HAI-FUDR or systemic oxaliplatin with HAI-FUDR are presently ongoing. As second-line therapy in the palliative setting for unresectable hepatic colorectal metastases, systemic CPT-11 and/or oxaliplatin combined with HAI-FUDR was associated with a response rate of over 70% (Kemeny et al, 2001; Kemeny & Gonen, 2005). This compares quite favorably to second-line systemic chemotherapy, which has a response rate of only 20%.

Recent Phase I and II studies of HAI with floxuridine and dexamethasone combined with systemic irinotecan (Kemeny et al, 2003) or oxaliplatin (Kemeny et al, 2009) demonstrated safety and feasibility in the adjuvant setting. Active investigation is now directed at determining the efficacy of this strategy. While awaiting the results of larger scale trials, regional FUDR should be considered an option in the adjuvant setting in patients at high risk for liver recurrence, particularly after failure of first-line systemic therapy.

Neoadjuvant Chemotherapy for Downstaging of Unresectable Disease

Several practical and conceptual advantages attend neoadjuvant chemotherapy for metastatic CRC involving the liver. Improved efficacy of modern chemotherapy has the potential to convert unresectable metastatic CRC into resectable disease, or it may reduce the extent of resection required to remove the hepatic disease (see Chapters 65 and 87). In addition, neoadjuvant chemotherapy can treat micrometastatic disease early as opposed to addressing it in a delayed fashion. Finally, this strategy may function as an in vivo test of chemoresponsiveness to a particular agent. Bismuth and colleagues (1996) were the first to publish this observation in a study of 330 consecutive patients with unresectable disease. They found that a combination of 5-FU/LV and oxaliplatin converted 53 patients to resectability (16%). Furthermore, the long-term outcome of these individuals after resection was similar to patients who were initially resectable. Since then, the potential to downstage unresectable into resectable disease with neoadjuvant chemotherapy has been validated in a number of studies and ranges between 3% and 41% of patients (Giacchetti et al, 1999; Adam et al, 2001; Wein et al, 2001; Clavien et al, 2002; Adam et al, 2004; Pozzo et al, 2004; Vauthey et al, 2006; Adam et al, 2009). Patients who undergo complete resection have a median survival between 30 and 60 months. Of course, patient selection plays a vital role when estimating the potential benefit of neoadjuvant treatment. In the largest prospective study to date, 138 of 1104 unresectable patients (12%) became resectable after a course of neoadjuvant chemotherapy using either FOLFIRI or FOLFOX. Factors associated with a poor long-term outcome in these patients included large tumors, an increased number of tumors (≥3), increased preoperative CEA, and periportal or celiac lymph node metastases (Adam et al, 2004).

Regional chemotherapy alone or in conjunction with systemic chemotherapy may also convert unresectable disease into resectable disease. For example, 6 (26%) of 23 patients became resectable after HAI-FUDR treatment (Clavien et al, 2002). The 3-year actual survival rate was 84% for the patients who responded to HAI treatment compared with 40% for nonresponders. As part of a Phase I trial conducted at MSKCC, 21 patients with unresectable metastastic CRC received systemic oxaliplatin and CPT-11 in conjunction with HAI-FUDR. Seven of these patients (33%) became resectable after treatment (Kemeny et al, 2005).

Chemotherapy-Induced Liver Damage (See Chapters 65 and 87)

Neoadjuvant chemotherapy has been shown to cause liver injury in some patients. Specifically, oxaliplatin is associated with hepatic sinusoidal obstruction (Rubbia-Brandt et al, 2004), and CPT-11 is associated with hepatic steatosis (Kooby et al, 2003; Parikh et al, 2003); these agents can even cause severe steatohepatitis (Fernandez et al, 2005). Obese patients are at increased risk, and a preoperative liver biopsy should be considered to evaluate the liver parenchyma in such patients. When a major hepatic resection is planned in a patient suspected of having impaired liver function after exposure to systemic chemotherapy, preoperative portal vein embolization may be warranted to induce hypertrophy in the future liver remnant. In some instances, concern about the function of the future liver remnant should push the liver surgeon to proceed straight to surgery, and plan for chemotherapy postoperatively, after the liver has regenerated. In addition to steatohepatitis, the treated liver may become fibrotic and may develop perivascular adhesions, which can complicate a planned resection.

Despite well-documented histologic changes in the liver that can occur with chemotherapy, the actual clinical impact is not entirely clear. Indeed some studies have demonstrated a negative impact of oxaliplatin or irinotecan chemotherapy on postoperative morbidity (Karoui et al, 2006; Vauthey et al, 2006; Nakano et al, 2008) and mortality (Vauthey et al, 2006), whereas other studies have not confirmed these findings (Hewes et al, 2007; Mehta et al, 2008).

Disappearing Hepatic Metastases

A radiologic complete response, defined as the disappearance of a lesion on cross-sectional imaging, occurs in approximately 6% to 9% of lesions treated with neoadjuvant chemotherapy (Benoist et al, 2006; Elias et al, 2007; Auer et al, 2010). A lesion that disappears during chemotherapy may represent a pathologic complete response or a reduction in the sensitivity of cross-sectional imaging secondary to chemotherapy-induced steatosis. When a hepatic metastasis is present in a segment of the liver that cannot be removed, disappearance of this lesion during chemotherapy may preclude wedge resection or ablation and thereby jeopardize a complete resection. Moreover, in some patients, resection is only a consideration because previously apparent liver disease had disappeared during chemotherapy. When approaching such a patient, surgeons must understand that removal of visible disease is likely to be ineffective if occult and viable tumor remains in the foci that correspond to the “vanished” lesions.

Three institutions have published retrospective studies on the significance of liver metastases that disappear with chemotherapy (Benoist et al, 2006; Elias et al, 2007; Auer et al, 2010), and although all three studies agree that a complete radiologic response does not often equate to a true clinicopathologic response, the strength of this association varied across the studies. Benoist and colleagues (2006) found the lowest rate of true clinicopathologic response. The authors identified 38 patients with at least one or more disappearing liver metastases, a total of 66 complete radiologic responses, by CT scan (Benoist et al, 2006). All of these patients eventually underwent surgical exploration after a lesion disappeared. In 20 of 66 instances, macroscopic disease was appreciated at the site of disappearance. In 15 cases, the area of disappearance was resected, even though no residual disease was observed at surgery, and viable tumor cells were seen on histologic examination in 12 of these specimens. The remaining 31 sites were left in situ, and 23 recurred within 1 year of surgery. In summary, a true and durable response was only appreciated in 17% of the disappearing lesions.

Elias and colleagues reviewed data from 16 patients who had complete disappearance for at least 3 months during HAI or systemic chemotherapy in at least one liver metastasis in the future liver remnant on cross-sectional imaging. The disappearing lesion reappeared in 6 patients (38%) but represented a complete response in 10 patients (62%) based on a minimum of 2 years of follow-up. Interestingly, none of the lesions that disappeared with HAI therapy reappeared in follow-up. In the most recent study by Auer and colleagues, 38 patients had 118 disappearing liver metastases; a total of 68 disappearing lesions were resected, and 50 were followed clinically (Auer et al, 2010). In the resected liver that harbored the 68 vanished lesions, a complete pathologic response was observed in 44 patients (65%). Similarly, there were 31 recurrences (62%) in the 50 disappearing lesions that were followed clinically for a median of 41 months. Factors associated with a true complete response included HAI chemotherapy, inability to visualize the disappearing lesion on MRI, and normalization of serum CEA with treatment.