Multimodal therapy with intravenous biweekly leucovorin, 5-fluorouracil and irinotecan combined with hepatic arterial infusion pirarubicin in non-resectable hepatic metastases from colorectal cancer (a European Association for Research in Oncology trial)

L. Zelek1, R. Bugat2, D. Cherqui3, G. Ganem4, P. Valleur5, R. Guimbaud4, O. Dupuis4, T. Aziza2, P. L. Fagniez3, J. Auroux6, H. Kobeiter6, C. Tayar3, A. C. Braud1, E. Haddad1, A. Piolot1, M. Buyse7 and P. Piedbois1,+

1 Department of Oncology, 3 Department of Surgery, 6 Department of Radiology and Department of Hepato-Gastroenterology, CHU Henri-Mondor, Assistance Publique-Hôpitaux de Paris, Créteil; 2 Department of Oncology and Department of Radiology, Centre Claudius Régaud, Toulouse; 4 Centre Jean Bernard, Le Mans; 5 Department of Surgery, Hôpital Lariboisière, Assistance Publique-Hôpitaux de Paris, Paris, France; 7 International Drug Development Institute, Brussels, Belgium

Received 25 February 2003; revised 16 May 2003; accepted 4 June 2003


    Abstract
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
Background:

The purpose of this study was to evaluate the tolerance and efficacy of combining i.v. irinotecan, 5-fluorouracil (5-FU) and leucovorin (LV) with hepatic arterial infusion (HAI) of pirarubicin in non-resectable liver metastases from colorectal cancer.

Patients and methods:

Thirty-one patients were included in a phase II trial with i.v. irinotecan/5-FU/LV administered every 2 weeks, combined with HAI pirarubicin 60 mg/m2 on day 1 every 4 weeks. In most cases HAI was administered via a percutaneous catheter.

Results:

The main grade 3/4 toxicity was neutropenia, encountered in 78% of the patients. When all patients were considered in the analysis, tumour response rate was 15 out of 31 [48%; 95% confidence interval (CI) 32% to 65%]. Liver resection was made possible in 11 patients (35%; 95% CI 21% to 53%). There were no toxic death. Median overall survival was 20.5 months, and median progression-free survival was 9.1 months. In patients with completely resected metastases, median overall survival was not reached and median progression-free survival was 20.2 months.

Conclusion:

The multimodality approach used in the present study was well-tolerated and yielded dramatic responses. An aggressive approach combining i.v. and HAI chemotherapy deserves further investigation.

Key words: colorectal cancer, hepatic artery infusion, irinotecan, liver resection, pirarubicin


    Introduction
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
Colorectal cancer is the second major cause of cancer deaths worldwide. At the time of diagnosis, ~25% of patients present with overt metastases and eventually ~50% will develop metastatic disease. The liver is the most common visceral site for disease spread [1]. Systemic 5-fluorouracil (5-FU) has been the mainstay of therapy for colorectal cancer patients for over 30 years. In the 1990s, our group confirmed through meta-analyses that objective tumour response rates for i.v. single-agent 5-FU bolus were poor (10–15%), with a median survival of only 1 year, whereas continuous infusion 5-FU or its modulation with leucovorin (LV) or methotrexate, led to a doubling of tumour response rates with a moderate impact on survival [2]. As a result, De Gramont et al. [3] developed, in France, a combined regimen with 5-FU and LV bolus followed by continuous 5-FU over 2 days every 2 weeks. More recently, irinotecan has demonstrated activity in patients relapsing after 5-FU, cost-effectiveness [4] and activity in the treatment of advanced colorectal cancer following failure with 5-FU [5]. Subsequently, following the observation of a significant improvement in survival compared with 5-FU/LV alone [6, 7] the systemic administration of irinotecan together with 5-FU/LV has been recommended in both the USA and EU for the first-line treatment of advanced colorectal cancer. On the other hand, hepatic arterial infusion (HAI) of drugs with predominant hepatic metabolism such as 5-fluoro-deoxyuridine (FUDR) or pirarubicin achieves maximum intra-tumour concentrations of chemotherapy without increasing systemic toxicity [8, 9]. A meta-analysis based on individual patient data from all randomized studies comparing HAI FUDR with i.v. fluoropyrimidines in patients with metastases confined to the liver showed that HAI FUDR dramatically enhanced tumour response rates (41% versus 14%, P <10–10) [10]. However, HAI FUDR has several disadvantages, such as the cost of the procedure requiring a pump [11], its local toxicity and its high extraction rate lowering systemic blood concentrations thereby impairing activity against extra-hepatic metastases [8]. Moreover, most recent studies [12, 13] failed to provide evidence of any significant advantage over i.v. fluoropyrimidines given with an optimal schedule, thus stressing the need for more innovative HAI approaches.

Pirarubicin (THP-Adriamycin, Theprubicine®; Aventis Pharma, France) is an hemi-synthetic anthracycline with a pharmacokinetic profile favouring locoregional therapy and thus HAI [9]. From a phase I study, its recommended dose as a single agent for HAI is 60 mg/m2 over 40 min every 3 weeks with an excellent feasibility allowing further dose escalation according to tolerance [14]. Moreover, pirarubicin can be given on an outpatient basis or during a short hospital stay and is therefore more convenient and cost-effective than protracted FUDR. In a pilot study, pirarubicin administered through the hepatic artery yielded a 33% objective response rate in patients with liver metastases from colorectal cancer [15].

Because extra-hepatic recurrence is a major issue in patients treated by HAI chemotherapy [10], a combination of i.v. chemotherapy (IVC) and HAI chemotherapy is a logical treatment strategy for patients with metastases confined to the liver. If this strategy proves optimal, substantial improvements in survival might be expected, with a high degree of conversion from initially non-resectable to resectable patients after therapy. Based on this rationale, the European Association for Research in Oncology conducted a phase II trial with i.v. 5-FU/LV and irinotecan combined with HAI pirarubicin in patients with non-resectable liver metastases from colorectal cancer.


    Patients and methods
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
Patients
Eligible patients were to have pathologically confirmed colorectal carcinoma with non-resectable liver metastases assessed by an expert surgeon. Non-resectable metastases were defined by the impossibility of performing the resection of all the lesions with clear margins (e.g. ill-located, multinodular or large lesions) and/or the likelihood of liver insufficiency due to a residual amount of liver parenchyma <40% [16]. Patients with central nervous system metastases, peritoneal involvement, mediastinal, retroperitoneal or pelvic nodes, and/or pleural effusion were not eligible. At least one bidimensionally measurable lesion larger than 2 x 2 cm documented by computed tomography (CT) scan was required. Adequate haematological (white blood cell count >4000/mm3, neutrophil count >2000/mm3, platelet count >100 000/mm3 and haemoglobin >10 g/dl), liver [serum bilirubin <1.5x upper limit of normal (ULN), serum transaminases and alkaline phosphatases <3x ULN], renal (serum creatinine <120 µmol/l or clearance >60 ml/min) and coagulation (prothrombin time and partial thromboplastin time) profiles were required. Except for adjuvant chemotherapy ending at least 4 months before inclusion and pelvic irradiation ending at least 8 weeks before inclusion (the upper limit of which should not be above vertebra S1), no prior therapy was allowed. Concurrent antithrombotic treatment, serious illnesses such as congestive heart failure or angina pectoris and pregnancy were exclusion criteria. All patients were >18 years old, had a World Health Organization (WHO) performance status of 0–2 and a life expectancy of at least 12 weeks. All patients signed an informed consent.

Procedures and treatment
Patients received IVC and HAI concurrently. Each 28-day cycle consisted of i.v. irinotecan given over 60 min at a dose of 150 mg/m2 on days 1 and 15; followed on days 1, 2, 15 and 16 by i.v. LV 200 mg/m2 over 2 h, i.v. bolus 5-FU 400 mg/m2 and i.v. infusional 5-FU 600 mg/m2 over 22 h. Appropriate antiemetic premedication, including a setron and a corticosteroid, was given. Irinotecan-related cholinergic syndrome was prevented by atropine 0.25 mg given subcutaneously before infusion in the absence of contraindications. HAI pirarubicin was given at day 16 of each cycle at a dose of 60 mg/m2 via a 60-min infusion. Hepatic artery catheters were inserted using standard interventional radiological techniques via a femoral route. The proper situation of the catheter was verified by abdominal radiography and its permeability was controlled before intra-arterial infusion. After removal of the catheter, the puncture site was compressed for 12 h and was monitored for signs of bleeding, haematoma or sepsis. Permanent intra-arterial devices with a subcutaneous infusion chamber were allowed as an alternative to hepatic artery catheterisation.

Toxicity and response assessment
In the case of an objective tumour response, treatment was to be continued for up to nine cycles. Therapy was to be stopped in the case of tumour progression, unacceptable toxicity or the patient’s refusal to continue. Tumour measurements were carried out every two cycles according to WHO criteria. In the case of an objective response, a confirmatory CT was required at 4 weeks. Blood counts were carried out weekly and liver chemistry, ionogram and serum creatinine were done before each cycle. Coagulation was controlled before each HAI. Toxicity was assessed according to the National Cancer Institute Common Toxicity Criteria.

Dose adjustments
In the absence of toxicity during the first two cycles, the dose of irinotecan was subsequently increased to 180 mg/m2. Doses were to be reduced in the case of febrile neutropenia or grade 4 neutropenia, grade 4 thrombocytopenia, grade 3/4 diarrhoea, grade 3/4 mucositis, grade 3/4 bilirubinaemia, increase of transaminases or alkaline phosphatases >5x ULN, grade 3/4 nausea or vomiting and severe hand–foot syndrome. The dose of irinotecan was reduced to 120 mg/m2 at days 1 and 15, and pirarubicin was reduced to 50 mg/m2 at day 16. In the case of toxicity occurring after a first dose reduction, subsequent doses were modified as follows: days 1 and 15, irinotecan 100 mg/m2; days 1, 2, 15 and 16, bolus 5-FU 300 mg/m2; days 1, 2, 15 and 16, infusional 5-FU 500 mg/m2; day 16, intra-arterial pirarubicin 40 mg/m2 (LV was not reduced). No further dose reductions were allowed and if severe toxicity occurred after the second dose adjustment, patients were removed from the study. In the case of any dose adjustments, further escalation was not allowed.

Chemotherapy was re-instituted after haematological recovery (neutrophils and platelet counts >1500 and >100 000/mm3, respectively) and recovery of all other toxicities to grade 0 or 1. Cycle delay >2 weeks was not permitted, and in this case patients were removed from the study.

Statistical considerations
The study was conducted using the two-stage design of Simon [17] with a planned accrual of 20 patients in step I and up to eight patients in step II. Six objective responses were to be observed in the step I accrual to proceed to the step II assuming the following hypotheses: null hypothesis: response rate <30%; alternative hypothesis: response rate >60%; type I error = 0.05; type II error = 0.05.

The primary end point was the objective tumour response rate and secondary end points were duration of response, time to progression, the proportion of patients benefiting from liver resection after tumour reduction, overall survival and treatment tolerance. Time to progression and survival were recorded from the day of inclusion. The Kaplan–Meier method was used to estimate survival [18].

Results of the study are strictly expressed following the intention-to-treat principle without excluding any patients.


    Results
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
Patient characteristics
From March 1998 to December 2000, 31 patients from three French centres were entered into the study. Their characteristics are listed in Table 1. Four patients were improperly included for the following reasons: retroperitoneal lymph nodes (n = 1), peritoneal metastases (n = 1), inadequate cardiac function (n = 1) and perturbations of liver chemistry (n = 1). Another patient with an arterial infusion chamber did not receive pirarubicin because of catheter obstruction before the first cycle. Four patients were not evaluable for response. At the time of analysis 19 patients were dead and minimum follow-up was 17 months among patients still alive.


View this table:
[in this window]
[in a new window]
 
Table 1. Patient characteristics
 
Objective tumour responses and liver surgery
Partial tumour responses were observed in 15 patients. Tumour response rate was 15 of 31 [48%; 95% confidence interval (CI) 32% to 65%] when all patients were considered in the analysis. No complete responses were observed. Metastatic disease was stabilised in seven other patients (23%). Overall, metastatic disease was therefore substantially decreased or temporally stabilised in 22 patients (71%; 95% CI 53% to 84%).

Liver resection was made possible according to the criteria mentioned above [16] in 11 responding patients initially considered as inoperable (35%; 95% CI 21% to 53%). Patient characteristics and surgical procedures are given in Table 2. The margins of liver resections were free of any tumour deposit (R0 resection) in nine patients. After surgery most patients received consolidation chemotherapy with the same i.v. regimen without HAI for 2–3 months.


View this table:
[in this window]
[in a new window]
 
Table 2. Resected patients (n = 11): characteristics and surgical procedures
 
At the time of analysis, 12 of the 15 initially responding patients had metastatic progression, and the liver was the site of first progression in all cases but two.

Overall survival and time to progression
The median overall survival was 20.5 months for the entire population. In patients with R0 resection, median overall survival was not reached, whereas it was 13.9 months in the other patients. Two- and 3-year survivals were 100% and 65%, respectively in patients with R0 liver resection, and 22% and 15% in the other patients (Figure 1). Median progression-free survival was 9.1 months for the entire population. In patients with R0 resection, median progression-free survival was 20.2 months, whereas it was 4.2 months in the other patients. Two-year progression-free survival was 42% in patients with R0 liver resection, versus 0% in the other patients (Figure 2).



View larger version (9K):
[in this window]
[in a new window]
 
Figure 1. Overall survival time in months for the entire population (left), and by liver resection (right).

 


View larger version (10K):
[in this window]
[in a new window]
 
Figure 2. Progression-free survival time in months for the entire population (left), and by liver resection (right).

 
Toxicity and complications
A total of 143 cycles were given with a median of five cycles per patient (range one to nine). No toxic deaths occurred but one patient died after two cycles from a non-treatment-related pulmonary embolism. Toxicity per cycle and period, expressed as intention-to-treat results, is given in Table 3. Median delays between days 1 and 15 of each cycle and days 15 and 29 (i.e. day 1 of the following cycle) were 15 days (range 13–30) and 22 days (range 13–57). This was mainly due to neutropenia following the second period (days 15, 16 plus HAI) of each cycle. Grade 3/4 neutropenia occurred in 11% of day 1–14 periods versus 42% of day 15–29 periods. Neutropenia was responsible for 8% and 39% of delays over the day 1–14 and day 15–29 periods of each cycle, respectively. A total of 48% of cycles were associated with a grade 3/4 neutropenia, which resulted in fever in only 5% of cycles. Taken together, 77% of patients (24 of 31) experienced grade 3/4 neutropenia, but only 19% neutropenic fever.


View this table:
[in this window]
[in a new window]
 
Table 3. Toxicity per cycle and per period (percentage of cycles)
 
Other complications were diarrhoea (22% of cycles), mucositis (15% of cycles) and non-neutropenic fever (13% of cycles), which were grade 1/2 except for three patients: one patient experienced grade 4 diarrhoea, another grade 3 vomiting and another grade 3 fever. Cholinergic symptoms were seen in 2% of cycles. No grade 3/4 thrombocytopenia, liver toxicity or hand–foot syndromes were observed. Median durations of toxicity (days 1–15 and 15–29 periods, respectively) were 3 days (range 1–7) and 3.5 days (1–12) for diarrhoea, 2 days (1–9) and 2.5 days (1–11) for nausea and vomiting, 5.5 days (1–10) and 7 days (3–24) for mucositis, and 2 days (1–17) and 3 days (1–9) for fever. Chemotherapy was given at the planned dose in 65% of cycles (n = 93). Toxicity or delays required dose reductions to the first and second level in 27% (n = 38) and 8% (n = 12) of cycles, respectively.

A total of 141 arterial catheterisations were carried out, using interventional radiological techniques in 120 cycles and surgery in 21 cycles. Severe complications were uncommon and included non-functional catheter (n = 3), failure of hepatic artery catheterisation (n = 2), pain during HAI (n = 2), extravasation (n = 2), arterial thrombosis (n = 1), haematoma (n = 1) and non-complicated haemorrhage (n = 1).


    Discussion
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
The prognosis of patients with non-resected liver metastases from colorectal cancer is poor. In a meta-analysis published in 1996 [10], our group showed that HAI chemotherapy with FUDR significantly increased the probability of tumour response compared with i.v. fluoropyrimidine chemotherapy. In that study, based on data from 654 patients, tumour response rates were 41% for patients allocated to HAI versus 14% for patients allocated to i.v. fluoropyrimidines. Survival was also better in patients allocated to HAI than in patients not allocated to HAI. Isolated extra-hepatic failures were five times higher in patients receiving HAI than in those allocated to IVC (45% versus 8%). However, the most recent studies, both using HAI 5-FU rather than FUDR, were not in favour of the use of HAI outside clinical trials [12, 13]. Moreover, major improvements in advanced colorectal cancer survival were mainly due to the widespread use of novel agents such as oxaliplatin or irinotecan [19], and the integration of such compounds into a new generation of HAI trials is a critical issue.

Based on these considerations, the present study is an attempt to combine HAI and IVC in order to achieve an antiproliferative action on both hepatic and non-hepatic sites. A medico-economic study carried out on the meta-analysis dataset showed that most of the extra costs associated with HAI FUDR were related to the use of implantable pump devices [11]. A secondary objective of the present study was therefore to improve the feasibility of the HAI procedure and to decrease its cost.

The combination of irinotecan with 5-FU/LV regimens is now widely considered as one of the most active chemotherapy regimens in patients with advanced colorectal cancer. Such regimens have yielded tumour response rates ~50% in several phase III studies, and median survivals between 20 and 24 months with tolerable toxicity [6, 7].

In the present study, the choice of pirarubicin for the HAI chemotherapy was based on its pharmacological properties, and efficacy in early trials. Previous studies have demonstrated the hepatic metabolism of pirarubicin [9, 14, 15], and in a pilot study HAI pirarubicin yielded a 33% response rate in liver metastases from colorectal cancer [15]. Moreover, both feasibility with a temporary percutaneous catheter and tolerance of HAI pirarubicin are excellent.

The multimodal therapy used in the present study led to a 48% tumour response rate (15 of 31 patients), and to a 20.5 months median survival. When analyses are restricted to eligible and evaluable patients, the tumour response rate was 56% (15 of 27). The feasibility of the investigational regimen was generally good, with no toxic deaths, no severe complications related to repeated catheter placement, and manageable haematological toxicity.

Many of the goals of this study were achieved. The treatment used safely combined one of the most active i.v. regimens and an active HAI drug. The vast majority of the patients were treated using a repeated radiological insertion of an HAI catheter without significant complications. Finally, this approach resulted in a dramatic decrease in the risk of extra-hepatic failure as compared with ‘HAI only’ approaches, and liver surgery was made possible in one-third of the patients, most of them having major liver involvement before chemotherapy. Interestingly, the survival of patients who remained inoperable after treatment was relatively poor compared with usually reported survival for patients receiving i.v. 5-FU/LV plus irinotecan [6, 7], whereas patients with R0 resection had a favourable outcome. Although this kind of comparison must be considered with caution, this result indicates that the patients included in the trial had pejorative prognostic factors.

The results presented here compare favourably with recently reported studies using HAI plus IVC [2022]. For example, in the trial reported by Copur et al. [21] using i.v. 5-FU, oral LV and HAI FUDR, the response rate was 45%, median progression-free survival 6 months, and median overall survival 13 months. In the trial recently reported by Kemeny et al. [22] using i.v. irinotecan and HAI FUDR, the response rate was 74% but median progression-free survival was 8.1 months and median overall survival was 17.2 months.

In conclusion, our study confirms that a relatively aggressive approach, combining i.v. and HAI chemotherapy, deserves further investigations. However, even though the combination used in this trial allowed metastasectomy in one-third of the patients, overall response rate and survival were not dramatically superior to best results reported with most recent IVC regimens. A synergistic or even additive effect between the i.v. and HAI regimens chosen must therefore be questioned, and further research should also address this issue.


    Acknowledgements
 
This study was supported in part by Aventis Pharma and was presented in part at the 2000 American Society of Clinical Oncology meeting (May 2000) and at the 2002 European Society for Medical Oncology meeting (October 2002).


    Footnotes
 
+ Correspondence to: Dr P. Piedbois, Department of Medical Oncology, CHU Henri Mondor, Assistance Publique-Hôpitaux de Paris, 51 Avenue Delattre de Tassigny, 94010 Créteil CEDEX, France. Tel: +33-1-49-81-25-82; Fax: +33-1-49-81-25-79; E-mail: pascal.piedbois{at}hmn.ap-hop-paris.fr Back


    References
 Top
 Abstract
 Introduction
 Patients and methods
 Results
 Discussion
 References
 
1. Cohen MC, Minsky BD, Schilsky RL. Cancer of the colon. In DeVita VT, Hellman S, Rosenberg SA (eds): Cancer: Principles and Practice of Oncology, 5th edition. Philadelphia, PA: Lippincott 1993; 1144–1197.

2. Zelek L, Piedbois P, Buyse M. Contribution of meta-analyses to the evaluation of treatments for advanced colorectal cancer. Expert Rev Anticancer Ther 2002; 2: 417–425.[Medline]

3. De Gramont A, Louvet C, Andre C et al. A review of GERCOD trials with bimonthly leucovorin plus 5-fluorouracil 48 h continuous infusion in advanced colorectal: evolution of a regimen. Eur J Cancer 1998; 34: 619–626.[CrossRef][ISI][Medline]

4. Levy-Piedbois C, Durand-Zaleski I, Juel H et al. Cost-effectiveness of second-line treatment with irinotecan or 5-FU in metastatic colorectal cancer. Ann Oncol 2000; 11: 157–161.[Abstract]

5. Rougier P, Bugat R, Douillard JY et al. Phase II study of irinotecan in the treatment of advanced colorectal cancer in chemotherapy naive patients and patients pretreated with fluorouracil-based chemotherapy. J Clin Oncol 1997; 15: 251–260.[Abstract]

6. Douillard JY, Cunningham D, Roth AD et al. Irinotecan combined with fluorouracil compared with fluorouracil alone as first-line treatment for metastatic colorectal cancer: a multicentre randomized trial. Lancet 2000; 355: 1041–1047.[CrossRef][ISI][Medline]

7. Saltz LB, Cox JV, Blanke C et al. Irinotecan plus fluorouracil and leucovorin for metastatic colorectal cancer. N Engl J Med 2000; 343: 905–914.[Abstract/Free Full Text]

8. Chen GHG, Gross JF. Intra-arterial infusion of anticancer drugs: theoretic aspects of drug delivery and review of responses. Cancer Treat Rep 1980; 64: 31–40.[ISI][Medline]

9. Robert J, Gianni L. Pharmacokinetic and metabolism of anthracyclines. Cancer Surv 1993; 17: 219–252.[ISI][Medline]

10. Meta-analysis Group in Cancer. Reappraisal of hepatic arterial infusion in the treatment of nonresectable liver metastases from colorectal cancer. J Natl Cancer Inst 1996; 88: 252–258.[Abstract/Free Full Text]

11. Durand-Zaleski I, Roche B, Buyse M et al. Economic implications of hepatic arterial infusion chemotherapy in treatment of non resectable colorectal liver metastases. J Natl Cancer Inst 1997; 89: 790–795.[Abstract/Free Full Text]

12. Lorenz M, Müller HH. Randomized, multicenter trial of fluoruracil plus leucovorin administered either via hepatic arterial or intravenous infusion versus fluorodeoxyuridine administered via hepatic arterial infusion in patients with non resectable liver metastases from colorectal carcinoma. J Clin Oncol 2000; 18: 243–254.[Abstract/Free Full Text]

13. Kerr DJ, McArdle CS, Ledermann J et al. Intrahepatic arterial versus intravenous fluorouracil and folinic acid for colorectal cancer liver metastases: a multicentre randomised trial. Lancet 2003; 361: 368–373.[CrossRef][ISI][Medline]

14. Munck JN, Rougier P, Chabot GG et al. Phase I and pharmacological study of intra-arterial administration of pirarubicin in patients with advanced hepatic metastases. Eur J Cancer 1994; 30A: 289–294.

15. Rougier P, Munck J, Elias D et al. Intra-arterial hepatic chemotherapy with pirarubicin. Preclinical and clinical studies. Am J Clin Oncol 1990; 13 (Suppl 1): S1–S4.[ISI]

16. Bismuth H, Adam R, Lévi F et al. Resection of nonresctable liver metastases from colorectal cancer after neoadjuvant chemotherapy. Ann Surg 1996; 224: 509–522.[CrossRef][ISI][Medline]

17. Simon R. Optimal two-stage designs for phase II clinical trials. Controlled Clin Trials 1989; 10: 1–10.[ISI][Medline]

18. Kaplan E, Meier P. Nonparametric estimation from incomplete observation. J Am Stat Assoc 1958; 53: 457–481.[ISI]

19. Piedbois P, Zelek L, Cherqui D. Chemotherapy of nonoperable colorectal liver metastases. Hepatogastroenterolgy 2001; 48: 711–714.[ISI][Medline]

20. Bonetti A. Hepatic artery infusion for liver metastases from colorectal cancer. Lancet 2003; 361: 358–359.[CrossRef][ISI][Medline]

21. Copur MS, Capadano M, Lynch J et al. Alternating hepatic arterial infusion and systemic chemotherapy for liver metastases from colorectal cancer: a phase II trial using intermittent percutaneous hepatic arterial access. J Clin Oncol 2001; 19: 2404–2412.[Abstract/Free Full Text]

22. Kemeny N, Gonen M, Sullivan D et al. Phase I study of hepatic arterial infusion of floxuridine and dexamethasone with systemic irinotecan for unresectable hepatic metastases from colorectal cancer. J Clin Oncol 2001; 19: 2687–2695.[Abstract/Free Full Text]