Chapter 86 Regional chemotherapy for liver tumors
Systemic Chemotherapy
Responses to systemic chemotherapy (SYS) vary with the type of tumor; for example, patients with liver metastases from a breast primary may have high response rates with SYS (Harris et al, 1997), in contrast to metastases from a gastric or pancreatic primary tumor. Despite 20% response rates and median survivals less than 1 year, fluoropyrimidines represented the standard of care for metastatic CRC for decades (Thirion et al, 2004). More recently, patients with metastatic colon carcinoma have had response rates greater than 30%. With the development of newer drugs, such as irinotecan and oxaliplatin, response rates have increased to 35% to 40% with median survivals of 15 to 19 months (Goldberg et al, 2002; Saltz et al, 2000). The use of targeted agents, such as bevacizumab and cetuximab (Cunningham et al, 2004), has further improved outcomes, and bevacizumab demonstrated overall survival of 20 months (Hurwitz et al, 2004). Many chemotherapy trials do not differentiate patients with liver-only metastases to show how these patients respond to chemotherapy. In tumors with high response rates, such as breast carcinoma, a reasonable response to SYS is found even in patients with liver metastases, although the response is lower with liver metastases than with soft-tissue metastases (George & Hoogstraten, 1978). In patients with CRC, the liver is the most common site of dissemination, and 60% of patients (Daly & Kemeny, 1986) develop liver metastases during the course of their disease.
Rationale for Hepatic Arterial Chemotherapy
1 Liver metastases are perfused almost exclusively by the hepatic artery, whereas normal hepatocytes derive their blood supply from the portal vein and the hepatic artery (Breedis & Young, 1954). After injection of floxuridine (FUDR) into either the hepatic artery or the portal vein, mean liver concentrations of the drug do not differ based on the route of injection; however, mean tumor FUDR levels are significantly increased (15-fold) when the drug is injected via the hepatic artery (Sigurdson et al, 1987).
2 The use of drugs that are largely extracted by the liver during first-pass metabolism results in high local concentrations of drug with minimal systemic toxicity. Ensminger and colleagues (1978) showed that 94% to 99% of FUDR is extracted by the liver during the first pass compared with 19% to 55% of 5-fluorouracil (5-FU). FUDR is therefore an ideal drug for hepatic arterial infusion (HAI), with a 400-fold increase in hepatic exposure with FUDR. The pharmacologic advantage of various chemotherapeutic agents for HAI is summarized in Table 86.1 (Ensminger & Gyves, 1983).
3 Drugs with a steep dose-response curve are more useful when given by the intrahepatic route, because small increases in the concentration of administered drug result in a large improvement in response. FUDR follows linear kinetics and is not saturated at high dose rates.
4 Drugs with a high total body clearance are more useful for hepatic infusion. The area under the concentration-versus-time curve is a function not only of drug clearance but also of hepatic arterial flow. Because hepatic arterial blood flow has a high regional exchange rate (100 to 1500 mL/min), drugs with a high clearance rate are needed (Collins, 1984). If a drug is not cleared rapidly, recirculation through the systemic circulation mitigates the advantage of intraarterial therapy over systemic therapy (Collins, 1986).
5 Another rationale for hepatic arterial chemotherapy, especially for patients with metastatic CRC, is the concept of a stepwise pattern of metastatic progression (Weiss, 1989; Weiss et al, 1986). This theory states that hematogenous spread occurs first via the portal vein to the liver, then from the liver to the lungs, and then to other organs. Aggressive treatment, either resection or hepatic infusion, of metastases confined to the liver yields prolonged survival for some patients.
Table 86.1 Estimated Increase in Hepatic Exposure for Drugs Given by Hepatic Arterial Infusion
| Estimated Increase by Drug | Half-Life (min) | Hepatic Arterial Exposure |
|---|---|---|
| Fluorouracil | 10 | 5- to 10-fold |
| Floxuridine | 10 | 100- to 400-fold |
| Bis-chloroethyl-nitrosourea | 5 | 6- to 7-fold |
| Mitomycin C | 10 | 6- to 8-fold |
| Cisplatin | 20 to 30 | 4- to 7-fold |
| Doxorubicin | 60 | 2-fold |
| Dichloromethotrexate | — | 6- to 8-fold |
The development of an implantable infusion pump allowed for the safe administration of hepatic arterial chemotherapy in the outpatient setting (Blackshear et al, 1972). Early trials using an implantable pump and continuous FUDR therapy produced a median response rate of 47% (Johnson & Rivkin, 1985) and a median survival of 17 months (Weiss et al, 1983). To further show that HAI had a therapeutic benefit, several randomized studies were conducted.
First-Line Therapy in Unresectable Liver Metastases
One of the first randomized trials was conducted at Memorial Sloan-Kettering Cancer Center (MSKCC) (Kemeny et al, 1987). Before randomization, patients were stratified for extent of liver involvement by tumor and baseline lactate dehydrogenase level, two factors that have been shown to be important prognostic indicators of survival (Table 86.2; Kemeny & Braun, 1983; Kemeny et al, 1989). This prospective randomized trial compared HAI with SYS using the same chemotherapeutic agent (FUDR), drug schedule (a 14-day continuous infusion), and method of administration (internal pump) in both groups.
Table 86.2 Randomized Studies of Hepatic Arterial Infusion (HAI) vs. Systemic Chemotherapy (SYS) for Unresectable Hepatic Metastases
Two subsequent European randomized trials using HAI therapy in this setting have been published. The first was conducted by the Medical Research Council (MRC) and European Organization for the Research and Treatment of Cancer (EORTC) groups, which compared HAI 5-FU/leucovorin (LV) with intravenous 5-FU/LV. Crossover from the intravenous to the HAI arm was not allowed. Of 290 patients randomized, only 66% on the HAI arms received treatment. Response rates were 22% for HAI and 19% for intravenous 5-FU/LV. This trial used subcutaneous ports rather than implantable pumps and had significant catheter-related problems (36% of HAI patients) (Kerr et al, 2003). The median survival was 14.7 months and 14.8 months in the HAI and SYS groups, respectively (P = .79).
The second trial was by a German cooperative group that randomized 168 patients with unresectable colorectal liver metastases to HAI of FUDR, HAI of 5-FU/LV, or intravenous 5-FU/LV (Lorenz & Muller, 2000). Response rates were higher in the two HAI arms, with no significant differences in time to progression, the primary end point, or overall survival (OS) between the arms. Only 70% of patients on the HAI arms actually received assigned treatment. The study also used ports instead of pumps.
The Cancer and Leukemia Group B (CALGB) (Kemeny et al, 2006) completed trial 9481, which compared systemic 5-FU/LV via the Mayo Clinic regimen, considered standard of care at the time of trial design, with HAI of FUDR, LV, and dexamethasone, a regimen that had produced high response rates (78%) and lower toxicity (3% biliary sclerosis) in a phase II study (Kemeny et al, 1994). An earlier trial randomized patients to HAI of FUDR with or without dexamethasone, which demonstrated less biliary toxicity and a trend toward median OS (23 months vs. 15 months, respectively, P = .06) in the dexamethasone-containing arm (Kemeny et al, 1992). In CALGB 9481, no crossover was permitted, and a total of 134 patients were randomized. Most patients (70%) had greater than 30% liver involvement, synchronous metastases (78%), and were chemotherapy naive (97%). Response rates were higher in the HAI group (47% vs. 24%; P = .012), although time to progression was not significantly different (5.3 months vs. 6.8 months; P = .8); time to hepatic progression was better in the HAI arm (9.8 months vs. 7.3 months; P =.017) and in the systemic arm (7.7 months vs. 14.8 months; P = .029). Median OS was significantly better in the HAI arm (24.4 months vs. 20 months; P = .0034; Kemeny et al, 2006). Resource use, quality of life, and molecular markers of prognosis, such as thymidylate synthase (TS) and p21 gene expression, were examined prospectively in this study, and final analysis of these factors were presented. At 3- and 6-month follow-up, physical functioning was improved in the HAI group. TS levels correlated with survival in HAI patients (24 months if TS >4, 14 months if TS ≤4), but these differences were not significant (P = .17).
A total of 10 randomized phase III trials have been done, most of which demonstrate a higher response rate with HAI versus SYS in patients with hepatic metastases from CRC (see Table 86.2). Whether this increase in response rate translates into increased survival is controversial. Several factors complicate this issue. Importantly, most of the trials contain relatively few patients; therefore the power to observe differences in survival rates is low. Also, because of early successes with intrahepatic infusion, some of these studies allowed patients in the systemic arm to cross over to intrahepatic therapy after tumor progression on systemic therapy. This crossover may have negated any difference in survival between the two groups. The studies show a survival advantage for those who received subsequent HAI, with a mean 1-year survival of 69% for the patients who had crossed over from SYS to HAI versus 35% for the patients who did not cross over (Table 86.3). Additionally, some trials included patients with extrahepatic disease, complicating survival analyses in the absence of systemic treatment. Other factors to consider are that some patients randomized to HAI did not receive it but were included in the survival analysis, and an absence of modern toxicity-based dose-titration schema may have resulted in fewer cycles of treatment.
Table 86.3 Summary of 1- and 2-Year Survival in Randomized Trials Comparing Hepatic Arterial Infusion (HAI) and Systemic Chemotherapy (SYS)
Two meta-analyses of the original seven trials were conducted and included more than 600 patients. The Meta-Analysis Group in Cancer (MAGIC, 1996) confirmed the increased response rates seen with HAI over SYS (41% vs. 14%); however, survival differences were deemed only statistically significant in the trials that included a control group treated only when symptoms arose. With the statistical methodology used, the authors were unable to include the data from Hohn and colleagues (1989); however, a second meta-analysis published the same year did incorporate the results from that study and found an absolute survival difference of 12.5% at 1 year (P = .002) and 7.5% at 2 years (P = .026) in favor of HAI (Harmantas et al, 1996). The authors omitted the trial by Allen-Mersch and colleagues (1994) given the fact that most patients in the SYS arm of that study received best supportive care only. Furthermore, they analyzed their results to show that, when omitting the data from Rougier and colleagues (1992)—because, like the study from Allen-Mersh and others (1994), only half of the patients in the SYS arm received chemotherapy—survival was only statistically significantly higher at the 1-year time point. Nonetheless, factoring only those six studies without crossover, the survival differences at 1 and 2 years were accentuated, and actually both significantly favor of HAI.
As mentioned above, CALGB 9481 permitted no crossover, and with the incorporation of dexamethasone, it demonstrated survival with HAI that compares well to published results with modern SYS regimens; however, a recent and well-orchestrated third meta-analysis that accounted for the design flaws discussed above examined all 10 randomized trials to date and failed to demonstrate a survival benefit with HAI (Mocellin et al, 2007). This report and the published commentary that followed demonstrated that, although there may be no clear survival benefit in using HAI instead of SYS, combining HAI and SYS has led to high response rates with acceptable toxicity profiles.
Toxicity of Intrahepatic Therapy
Table 86.4 summarizes the GI toxicities noted by investigators using the implantable pump. The side effects of SYS are almost never observed with HAI, and myelosuppression does not occur with intrahepatic FUDR (Kemeny et al, 1984). Although intrahepatic mitomycin C or bis-chloroethyl-nitrosourea may depress platelet counts, the absolute depression and frequency of depression are less than with systemic administration. Nausea, vomiting, and diarrhea do not occur with HAI of FUDR. If diarrhea does occur, shunting of the chemotherapeutic agent to the bowel should be suspected (Gluck et al, 1985).
The most common problems with HAI are ulcer disease and hepatic toxicity (Hohn et al, 1985; Kemeny et al, 1984). Severe ulcer disease results from inadvertent perfusion of the stomach and duodenum with drug via small, unligated branches from the hepatic artery, but this can be prevented by meticulous dissection at the time of pump placement (Hohn et al, 1985); however, even without radiologically visible perfusion of the stomach or duodenum, mild gastritis and duodenitis can occur. This toxicity can be reduced by careful dose reductions when any GI symptoms arise. Hepatobiliary toxicity is the most problematic toxicity seen with HAI. Although some evidence of hepatocellular necrosis and cholestasis has been found on liver biopsy specimens (Doria et al, 1986), most studies point to a combined ischemic and inflammatory effect on the bile ducts as the most important cause of this toxicity. The bile ducts are particularly sensitive to HAI; this is because the bile ducts derive their blood supply almost exclusively from the hepatic artery, similar to hepatic tumors (Northover & Terblanche, 1979).
In patients with severe toxicity, endoscopic retrograde cholangiopancreatography (ERCP) shows lesions resembling primary sclerosing cholangitis (Kemeny et al, 1985). Because the ducts are sclerotic and nondilated, ultrasound usually is not helpful. In some patients, the strictures are more focal, and they are usually worse at the bifurcation, and drainage procedures by endoscopic or by percutaneous transhepatic intubation may be helpful. Duct obstruction from metastases should be excluded first by computed tomography (CT) of the liver.
Close monitoring of liver function tests is necessary to avoid biliary sclerosis. If the serum bilirubin becomes elevated, no further treatment should be given until the bilirubin returns to normal, and then treatment should proceed with only a small test dose (0.05 mg/kg/day). In patients who cannot tolerate even a low dose for 2 weeks, it may be possible to continue treatment by giving the FUDR infusion for 1 week rather than the usual 2 weeks. At MSKCC, we modify treatment as outlined in Table 86.5.
In older trials, cholecystitis occurred in 33% of patients receiving HAI (Kemeny et al, 1986a). In more recent series, the gallbladder was removed at the time of catheter placement to prevent this complication and to avoid the confusion of these symptoms with other hepatic side effects of chemotherapy.
Surgical Technique
Hepatic Artery Pump Placement
The arterial anatomy of the liver varies (Table 86.6), and a normal anatomy is present in only about two thirds of patients (Allen et al, 2002; see Chapter 1B). Before consideration of pump placement, it is imperative to carry out a careful review of the arteriogram with the radiologist and formulate a plan for the management of aberrant anatomy. In the past, direct arteriography was required, but excellent definition now can be ascertained from CT angiography. In most cases, a pump with a single catheter is adequate to provide access to the entire hepatic arterial inflow. It is best not to place the catheter directly into the hepatic artery, which risks thrombosis of the vessel; instead, it should be placed into an accessible side branch. The gastroduodenal artery (GDA) is the preferred conduit for the catheter and provides the most reliable method of catheter implantation (Allen et al, 2002). As previously mentioned, cholecystectomy should be performed to prevent chemotherapy-induced cholecystitis. Numerous incisions have been employed for this operation with success, including an upper midline incision, right subcostal incision, or limited right subcostal hockey-stick incision. Preoperative intravenous antibiotics close to the time of incision are imperative.
For patients with unresectable disease, a staging laparoscopy is advisable to rule out occult extrahepatic disease, which in our experience occurs in approximately one third of patients (Grobmyer et al, 2004). A thorough examination of the abdomen is performed at laparoscopy and at laparotomy to look for extrahepatic disease. The most common sites of extrahepatic metastases are the peritoneum and portal lymph nodes; therefore a biopsy should be done if the lymph nodes look suspicious, because their involvement generally would preclude the use of the pump. The extent of liver involvement should be assessed and documented. Any radiographically occult hepatic tumors should be noted, and the potential for future resection should be specifically addressed in the operative note.
Standard Hepatic Arterial Anatomy
A standard cholecystectomy is performed, and the hepatic artery and its branches are circumferentially dissected. The common hepatic artery and the GDA are palpable superior to the body of the pancreas and the first portion of the duodenum. The GDA runs parallel to and lies immediately to the left of the common bile duct, and it is advisable to start by dissecting the common hepatic artery to minimize the risk of injuring the bile duct. The right gastric artery is ligated and divided. The distal common hepatic artery, the entire GDA, and the proximal proper hepatic artery are dissected away from their attachments. It is important to mobilize the full length of the extrapancreatic GDA to facilitate insertion of the catheter. Suprapyloric side branches of the GDA are often encountered and must be ligated. Frequently, branches to the pancreas and duodenum arise from many of these dissected vessels, and it is essential to identify and ligate these branches to avoid perfusion of the pancreas, stomach, or duodenum. The common hepatic artery is mobilized 1 cm proximally, and the proper hepatic artery is mobilized about 2 cm distally from the origin of the GDA. Branches to the retroperitoneum from the right or left hepatic artery are common and should be ligated. Review of preoperative angiography to look specifically for these branches is important, because they are often found in retrospect. At this point, a complete circumferential dissection of the common hepatic artery, GDA, and proper hepatic artery should be ensured such that no vessels to the pancreas, stomach, or duodenum remain (Fig. 86.1). The GDA should be temporarily occluded with palpation of the proper hepatic artery to rule out retrograde flow to the liver through the GDA secondary to celiac stenosis. No attempt at dissection of the common bile duct is necessary, because this risks devascularization and ischemic stricturing.
An arteriotomy is made in the distal GDA, and the catheter is inserted up to, but not beyond, the junction with the hepatic artery (Fig. 86.2). If the catheter protrudes into the common hepatic artery, turbulence of blood flow can lead to thrombosis of the vessel. Failure to pass the catheter to the junction leaves a short segment of the GDA exposed to full concentrations of FUDR without the diluting effect of blood flow, potentially resulting in sclerosis, thrombosis, or late dislodgment. When positioned, the catheter should be secured two or three times with nonabsorbable ties proximal to the tying rings on the catheter. Perfusion of both lobes of the liver and lack of extrahepatic perfusion is confirmed by infusing 2 to 3 mL of half-strength fluorescein through the pump and visualizing it with a Woods lamp. Half-strength methylene blue injection is an alternative method of ensuring proper perfusion. After the perfusion test, the catheter is flushed with heparinized saline, and the wounds are closed. Antibiotic coverage should be continued for 2 to 3 days postoperatively because of the seriousness of a pump infection. Any sign of erythema indicates a wound infection postoperatively and should be treated immediately and aggressively.
Aberrant Hepatic Arterial Anatomy (See Chapter 1B)
As discussed earlier, aberrant hepatic arterial anatomy is common, and numerous variations occur. Each anatomic situation is specifically addressed here, but first, general principles in managing variant anatomy are discussed. In analyses of our extensive experience with this operation, the factor most consistently associated with catheter-related complications and decreased durability of the catheter is cannulation of a vessel other than the GDA. The overall preferred technique is placement of the catheter in the GDA with ligation of the isolated variant vessel. This method relies on intrahepatic collateral development and cross-perfusion to the liver fed by the ligated vessel. Although concerns have been raised over incomplete hepatic perfusion with this technique, it is rare; in our published experience with this operation for variant anatomy, incomplete hepatic perfusion occurred once in 52 cases. This cross-perfusion sometimes can take 4 weeks to occur, and early perfusion scans may be abnormal initially; however, these should be rechecked in a few weeks to assess for the cross-perfusion, because most normalize (Allen et al, 2002; Curley et al, 1993). The only exception to this rule may be patients with central tumors so large that they impede cross-collateralization. Lastly, although cross-perfusion after ligation of aberrant vessels is highly reliable, it has not been proven that this results in equal blood flow for chemotherapy delivery.
Aberrant Origin of the Gastroduodenal Artery
The GDA can arise from the right or left hepatic artery, or there can be a trifurcation in which the GDA, right hepatic artery, and left hepatic artery all arise simultaneously from the common hepatic artery; this anomaly occurs 6% to 11% of the time (Allen et al, 2002). In general, it is preferable to place the catheter into the GDA and ligate vessels that are not receiving catheter-directed flow, because this is the technique associated with the lowest rate of catheter-related complications (Allen et al, 2002, 2005). In the case of a trifurcation, the catheter should be placed in the GDA, and perfusion should be tested with fluoroscein or methylene blue. If bilobar perfusion is adequate, ligation of vessels is unnecessary. If no perfusion of one lobe of the liver is seen at testing, the hepatic artery to that lobe should be ligated, most commonly the left hepatic artery, thereby relying on cross-perfusion of the left liver.
Replaced Left Hepatic Artery
A replaced left hepatic artery arises from the left gastric artery and supplies the left liver; no native left hepatic artery is found, an abnormality present in 4% to 16% of patients. Once again, the preferred technique is to place the catheter in the GDA and ligate the replaced left hepatic artery. Initial reports on this specific situation suggested rates of incomplete cross-perfusion of 40% (Cohen et al, 1987). More recent reports, including our experience, show that incomplete cross-perfusion is uncommon in this situation and occurred in only 1 of 10 of our patients at last analysis (Allen et al, 2002; Curley et al, 1993). Other techniques, such as placing catheters in the GDA and in a branch of the replaced left hepatic artery, can be considered in patients with bulky disease in the left liver or in those with a large central tumor that may impede cross-perfusion.
Replaced Right Hepatic Artery
A replaced right hepatic artery originates from the superior mesenteric artery, runs in the portacaval space, and supplies the whole right liver. No branches to the right liver originate from the proper hepatic artery, an anatomic situation that occurs 6% to 16% of the time. If the surgeon ligates the replaced right hepatic artery, cross-perfusion from the left hepatic artery occurs almost uniformly (Allen et al, 2002; Cohen et al, 1987; Curley et al, 1993). The catheter should be placed in the GDA, and the replaced right hepatic artery should be ligated. Other techniques have been described, such as placing a second catheter in the replaced right hepatic artery, either directly through a small arteriotomy, as there are no significant side branches from this vessel, or using a vascular graft, but these are rarely if ever indicated. This technique requires that the catheter be trimmed flush just beyond the tying ring and that it be placed such that the ring lies just inside the vessel, and the arteriotomy is closed over the ring.
