Metastatic colorectal cancer: therapeutic options

teaser

Smitha Patiyil
MBBS

Steven R Alberts
MD
Division of Medical Oncology
Mayo Clinic
Rochester, MN
USA
E:[email protected]

Colorectal cancer is associated with significant morbidity and mortality, especially in the developed world. WHO estimates that each year, approximately one million people are diagnosed, with almost half a million deaths.(1) In the USA it is projected that, in 2006, there will be 148,610 new cases of colorectal cancer and 55,170 deaths associated with this cancer.(2) These statistics are disturbing and make it especially important that we continue to improve our management options.

Treatment of metastatic colorectal cancer (mCRC) had seen few advances in the past few years. Almost a decade ago, there was only one active drug, 5-fluoro uracil (5FU), which, when combined with leucovorin (LV), gave a response rate (RR) of 30–35%, at the cost of moderate toxicity.(3) Patients generally died within a year or less of diagnosis of metastatic disease. Surgery was the only curative modality but only a minority of patients with organ-limited metastatic disease met the criteria for resection. Since then, the therapeutic landscape has broadened considerably, with the discovery of new cytotoxic and biological agents and availability of more than a dozen drug combinations. All this has resulted in a marked improvement in overall survival (OS). However, we are left with some challenges, including appropriate sequencing of the available drugs to maximise efficacy while minimising approaches for treatment.

Systemic chemotherapy
The foundation of almost all chemotherapy combinations in colorectal cancer is 5FU, a drug that was originally discovered in the 1950s. It has a short plasma half-life (less than 15 minutes) and is active only in the S phase of the cell cycle.(4) Bolus administration of 5FU was only marginally effective, with RRs in the range of 10–15%. By adding LV, a biochemical modulator, the response rate (RR) increased to 25–35%, with an improvement in survival to 10–12 months. The introduction of ­continuous infusion 5FU represented another advance in the use of 5FU, building on the observation that prolonged exposure to the drug improved the chance of exposure of the cancer cells to the active metabolite of 5FU in their vulnerable phase, achieving a better cell kill. The toxicities were different for the infusional approach, with increased incidence of hand and foot syndrome but less haematological toxicities. Phase III trials comparing infusional regimens with bolus regimens showed significant improvement in RR, but this did not translate into a significant survival advantage.(5) Two new cytotoxic agents, ­irinotecan and ­oxaliplatin, were then introduced within the past decade. Doublets combining 5FU with oxaliplatin or irinotecan improved disease-free survival (DFS) and OS, revolutionising the way we treat mCRC. These regimens have now replaced 5FU/LV as the current standard of care. We will examine the key trials which have brought forth this change.

In 2000, Saltz et al published results of a randomised trial comparing IFL (irinotecan, bolus 5FU and LV) to either bolus 5FU/LV or single-agent irinotecan.(6) The patients who received IFL had a better progression-free survival (PFS) (median 7.0 vs 4.3 months; p=0.004), OS (median 14.8 months vs 12.6 months, p=0.04) and RRs (39% vs 21%; p<0.001) than patients who received bolus 5FU/LV. Single-agent irinotecan was similar to 5FU/LV in efficacy. IFL tends to cause more grade 3 diarrhoea, but less neutropenia and mucositis compared with 5FU/LV. In the same year, Douillard et al reported improvement in OS (median 17.4 months vs 14.1 months; p=0.03) with infusional 5FU/LV combined with irinotecan, compared with infusional 5FU/LV alone with manageable toxicities.(7) Based on these results, the combination of 5FU and irinotecan was approved by the FDA. In the USA, IFL became the standard regimen in 2000.

However, within four years, based on the results of the national phase III trial N9741 trial, FOLFOX replaced IFL as the standard of care. In this pivotal trial, published in 2004, 795 previously untreated patients were randomised to receive IFL, FOLFOX (infusional 5FU, LV and oxaliplatin) or IROX (irinotecan and oxaliplatin).(8) FOLFOX turned out to be superior to both IFL and IROX with regards to OS, PFS and RRs. The OS of 19.5 months with the FOLFOX regimen was unprecedented in mCRC. The toxicities were less in the FOLFOX arm than in the IFL arm, except for paraesthesias and neutropenia. In an interim analysis, it was found that IFL was associated with a high 60-day mortality rate, with patients dying from dehydration resulting from diarrhoea, neutropenia and sepsis.(9) This led to a reduction in the dose of IFL in N9741 and a subsequent observed improvement in IFL-related toxicities.

The results of N9741 convinced most oncologists to adopt infusional FOLFOX as the first-line regimen in mCRC. However, there are some caveats to this trial. The IFL arm used bolus 5FU that may have accounted for some of the differences in the results. Also, OS could have been affected by the fact that only 24% of patients progressing on irinotecan received oxaliplatin as it was not yet available in the USA, compared with 60% of the patients on the oxaliplatin arm who received subsequent ­irinotecan.

In clinical practice the use of FOLFOX4 is gradually being replaced by FOLFOX6 as the standard of care. FOLFOX4 entailed 5FU and LV being given as a bolus on days 1 and 2, followed by a 22-hour infusion of 5FU along with 85mg/m(2) of oxaliplatin on day 1.FOLFOX6 is potentially more convenient, as the bolus 5FU/LV is given on day 1 only, followed by a 46-hour infusion of 5FU along with oxaliplatin at 100mg/m(2) on day 1. FOLFOX6 has been further modified in current US trials by reducing the dose of oxaliplatin to 85mg/m(2). This fine tuning helps to limit the toxicities but is not based on definitive clinical evidence. Since the sensory toxicity from oxaliplatin is thought to be cumulative, interruptions in the ­oxaliplatin ­schedule was examined in the OPTIMOX1 trial.(10) Using a “stop-and-go” strategy patients were randomised to receive FOLFOX7 for six cycles, followed by maintenance 5FU/LV for 12 cycles and then reintroduction of FOLFOX7, or continuous FOLFOX4 administration until disease progression. There was no difference in survival or neurotoxicity. Because of multiple protocol violations, a modified trial, OPTIMOX2, is currently ongoing to assess fully the benefits of this approach.

After N9741, the role of irinotecan in mCRC was debated. A new infusional regimen incorporating irinotecan (FOLFIRI) was developed by ­combining the bimonthly protracted infusion 5FU/LV with ­irinotecan and it was studied in heavily pretreated patients with very acceptable toxicities.(11) The question whether FOLFOX is superior to FOLFIRI was addressed by Tournigand et al in a trial published in 2004.(12) In this relatively small-sized trial, 220 patients were randomised between two arms with patients in arm A receiving FOLFIRI followed by FOLFOX6 on progression and patients in arm B receiving FOLFOX6 followed by FOLFIRI. Both sequences provided similar OS (21.5 months vs 20.6 months, p=0.99), PFS (14.2 months vs 10.9 months, p=0.64) and RRs (56% vs 54%, p=0.26). FOLFOX6 performed better in second-line, with a RR of 15% and a PFS of 4.2 months compared with second-line FOLFIRI, with 4.4% RR and PFS of 2.5 months. The toxicities were distinct, with sensory toxicities, neutropenia and thrombocytopenia more frequent with FOLFOX and nausea, mucositis, diarrhoea and neutropenic fever observed more often in FOLFIRI. Even though grade 3/4 adverse events were more common during first-line FOLFOX (74 vs 53%, p=0.001), there were more serious or life-threatening events associated with FOLFIRI (14 vs 5%, p=0.03). In another modest-sized trial, FOLFOX4 was compared to FOLFIRI, with equal efficacy but distinct toxicities.(13) Hence, from the available evidence, we can conclude that both FOLFOX and FOLFIRI are comparable regimens and can be used as first-line therapy.

The issue of sequential chemotherapy was further explored by the FOCUS trial, a multicentre trial from the UK involving 2,135 patients.(14) There were essentially five arms in this trial, with 5FU given in all arms as per the modified de Gramont regimen (LV and infusional 5FU over 48 hours every two weeks). In arm A,patients received 5FU followed by single-agent irinotecan, while in arm Box and arm Bir, oxaliplatin or irinotecan were combined to 5FU on progression, and arm Cox and arm Cir were essentially FOLFOX and FOLFIRI, respectively. The primary endpoint was OS, with no significant differences between the arms at study conclusion. However, there was a trend towards improvement in the combination arms, OS of 16.2 months and 14.8 months in arm C and arm B, respectively, when compared to arm A (13.7 months). The OS rates were worse than other reported larger trials, likely because only 20–27% of patients could proceed to second-line therapy. These results support consideration of combination over sequential therapy, as most patients might not be eligible for second-line therapy after progression. This was also confirmed in a pooled analysis of all major trials done by Grothey et al, illustrating that OS is dependent on exposure to all three active agents, namely ­oxaliplatin, ­irinotecan and 5FU. Importantly, the survival advantage seemed to be greater when combination chemotherapy was given as first-line, especially since only 50–80% of patients received second-line chemotherapy.(15)

The use of oral fluoropyrimidines as a substitute for infusional 5FU has gained increased attention with the introduction of agents such as capecitabine. In a trial reported by Van Cutsem et al, 602 patients with previously untreated mCRC were randomised to either capecitabine or bolus 5FU/LV.(16) Capecitabine achieved results equivalent to bolus 5FU/LV with generally fewer side-effects. The side-effects of capecitabine were similar to those of infusional 5FU, with diarrhoea and hand foot syndrome being the more prominent.

Combinations of capecitabine with oxaliplatin or irinotecan (CapOx, XELOX, CapIri, XELIRI) have been assessed in a variety of phase II trials and limited phase III trials. Several phase II trials suggested efficacy(17–19) comparable to that of combination with infusional forms of 5FU. However, the safety of combining capecitabine with ­irinotecan remains uncertain. The EORTC trial 40015 compared FOLFIRI to capecitabine and irinotecan. The trial was closed prematurely because of excessive deaths in the capecitabine and irinotecan arm.(20) The safety of other schedules of capecitabine and irinotecan are being accessed in other trials, including the Dutch CAIRO trial.

The combination of capecitabine and ­oxaliplatin continues to be evaluated in clinical trials. When assessed in North American trials, such as the randomised phase II TREE trial, a dose reduction of capecitabine was required from 1,000mg/m(2) to 850mg/m(2) in order to reduce toxicities.(21) In a randomised phase III trial from Germany ­comparing capecitabine and oxaliplatin to the regimen FUFOX, the toxicity, RR and PFS were similar in the two arms.(22) However, both in the TREE1 and in the German trials, there was a trend toward inferior PFS with capecitabine and oxaliplatin. Neither of these trials was sufficiently powered to address noninferiority of the capecitabine and oxaliplatin combination. As such, this combination requires further assessment in an adequately powered trial.

In summary, the overriding theme in the treatment of mCRC is that all three drugs (5FU, ­oxaliplatin and irinotecan) are important and combination therapy is preferred, as many patients may not be able to proceed to second-line therapy in the event of a progression. Table 1 summarises the results of the major trials using these three agents. Hence, it should come as no surprise that a new regimen, FOLFOXIRI, which combines irinotecan, ­oxaliplatin and 5FU, is superior to FOLFIRI, with manageable toxicities. The RR of 60% (compared with 34% in the FOLFIRI arm) enabled more patients with liver-only metastases to undergo a metastasectomy. These promising results were presented in the 2006 GI cancer symposium by Falcone et al,(23) and require further evaluation, particularly as a potential first-line regimen for patients in whom curative resection is desired.

[[HPE28_table1_42]]

Biological therapy
Bevacizumab, a humanised antibody against vascular endothelial growth factor (VEGF), has become an integral component of therapy in mCRC. In a pivotal phase III trial, the combination of bevacizumab and IFL increased RR (44.8 vs 34.8%, p=0.004), PFS (10.6 vs 6.2 months, p<0.001) and OS (20.3 vs 15.6 months, p<0.001) over IFL alone.(24) The improvement in OS by 4.7 months was comparable to the 4.5-month survival advantage with FOLFOX4 in the N9741 trial. The most frequent adverse event ­attributable to bevacizumab was hypertension (22.4%, grade 3 in 11.4%). Bowel perforation occurred in 1.5% of patients in the bevacizumab arm, with no events reported in the chemotherapy-alone arm. The incidence of other previously reported side-effects such as bleeding, thromboembolic complications and proteinuria with bevacizumab, were not statistically different between the two groups. Overall, the adverse events appeared to be manageable and did not lead to a significant increase in the number of hospitalisations, 60-day mortality or discontinuation of the drug.

There is currently only limited information on the potential benefit of combining FOLFOX with bevacizumab. Data from the trial E3200, released in 2005, showed a significant OS advantage when ­bevacizumab was combined with FOLFOX (12.5 months) compared with FOLFOX alone (10.5 months) in second-line therapy (p=0.0024).(25) The results from the randomised phase II TREE 2 trial showed a significant advantage of using bevacizumab with FOLFOX as first-line.(26) Table 2 summarises some of the key trials.

No data are currently available to assess the potential benefit of adding bevacizumab to FOLFIRI. There is also a lack of evidence to support the use of bevacizumab after progression of a regimen ­containing this agent and, hence, it would not currently be recommended pending supporting clinical data, particularly given the cost of bevacizumab.(27) Since bevacizumab can affect wound healing and hepatic regeneration, it is suggested that elective surgery be postponed for at least eight weeks from the time of last treatment.(28)

[[HPE28_table2_43]]

Cetuximab, an epithelial growth factor receptor (EGFR)-inhibiting antibody, was also recently approved for use in mCRC. Initially, it was tested in patients with mCRC refractory to irinotecan, with the theory that EGFR blockade would improve the responsiveness of resistant cancer cells. When the combination of irinotecan and cetuximab was compared with single-agent cetuximab, the RR increased from 10.8% to 22.9%.(29) Cetuximab was reasonably well tolerated except for a rash in 80% of patients, a common side-effect of EGFR inhibitors. Less than 10% of patients developed a grade 3 or 4 rash. In an ongoing study (EPIC trial), ­cetuximab plus FOLFIRI is being compared with FOLFIRI alone as second-line after progression on oxaliplatin-based chemotherapy.(30) Cetuximab has shown activity in combination with FOLFOX in a phase II trial as well, yielding a 72% RR and PFS of 10.2 months in patients with previously untreated mCRC.(31) Ten of the 62 patients enrolled in the trial had a response that allowed surgical resection of their metastatic disease to be attempted. With a growing body of data from clinical trials, it has become clear that current techniques for assessing EGFR overexpression do not correlate with the RR to cetuximab.(32)

Based on the individual activity of bevacizumab and cetuximab when combined with ­chemotherapy, there is now growing interest in the potential use of combining these two antibodies. The Bowel Oncology with Cetuximab Antibody (BOND)-2 trial addressed the feasibility and efficacy of bevacizumab added to either cetuximab alone or to cetuximab plus ­irinotecan in patients with irinotecan-refractory disease.(33) Preliminary results were encouraging, with a RR of 23% and 38%, respectively. The time to progression was 6.9-months and 8.5-months, respectively. These data appear to show significantly better results than those of historical controls from the BOND-1 trial (RR, 11% and 23%; time to progression, 1.5 months and 4 months). However, given the small number of patients enrolled in BOND-2, confirmation of the results will require a formal comparison in a randomised phase III trial. An intergroup trial (CALGB/SWOG 80405) is currently addressing the role of cetuximab and bevacizumab in first-line therapy for colorectal cancer. Investigators will choose the initial chemotherapy regimen to be used (modified FOLFOX6 or FOLFIRI), and patients will then be randomised to receive either cetuximab, bevacizumab or both agents together in conjunction with their chemotherapy. A follow-up study to the BOND-2 study, known as the BOND-2.5 trial, will assess the activity of the cetuximab/bevacizumab/irinotecan combination in patients who have previously progressed through a bevacizumab-containing regimen.

Surgical resection
The liver is the most common site of metastasis from colorectal cancer, with liver-only involvement occurring in approximately 40% of patients. Surgical resection of metastases is currently the only potentially curable modality of therapy. The five-year OS survival rates with surgery are generally in the range of 30–40%. However, approximately 80% of patients will have unresectable disease based on the location, size or number of metastases. Defining specific criteria for resectability remains ­challenging. Some of the contraindications are the presence of unresectable extrahepatic disease or widespread involvement of the liver defined as metastases in more than six segments or 70% of the liver, or involvement of all three hepatic veins.(34)

A growing body of literature supports the use of neoadjuvant chemotherapy in selected patients with initially unresectable or not optimally resectable liver metastases. In a retrospective study, Adam et al noted that out of 701 patients with unresectable liver metastases at presentation, 95 patients (13.5%) became resectable after undergoing treatment with oxaliplatin-based therapy. Their median survival was comparable to that of patients undergoing upfront resection.(35) When the records of the intergroup trial N9741 were analysed, it was found that 24 (3.3%) of the 795 patients enrolled underwent resection, 22 of whom had received FOLFOX.(36) In one of the few prospective trials, the North Central Cancer Treatment Group (NCCTG) assessed the potential ability of FOLFOX4 to downsize initially unresectable liver-only metastases and allow resection.(37) To be potentially eligible for participation in this trial, patients were required to have liver-only ­metastases from colorectal cancer deemed unresectable by a surgeon experienced in liver surgery. Patients received FOLFOX4 until they achieved adequate shrinkage to allow surgery, developed progressive disease or became intolerant of the therapy. Of the 42 evaluable patients who received FOLFOX4, 25 had a reduction of tumour burden (21 complete or partial responses, four regressions) by preoperative imaging. Seventeen of these patients underwent surgery, with 14 having a complete resection and one a partial resection. Two patients were found to be unresectable based on carcinomatosis in one patient and miliary spread of disease through the liver in the other patient. With a median postsurgical follow-up of 22 months (range 13–32), 11 patients have had recurrences (73% of resected patients), primarily involving the hepatic remnant (64%). Median time to disease recurrence following surgery was 19 months. At three years, 67% of patients undergoing resection were still alive.

Regional chemotherapy
Hepatic artery infusion (HAI) is an infrequently used modality in the treatment of isolated hepatic metastases. It is based on the principle that the liver has a dual blood supply, with the portal circulation supplying the normal tissue and the hepatic artery supplying areas of metastases. Hence, infusion of ­chemotherapy into the hepatic artery has the potential to deliver higher concentrations of drug to the metastases in the liver and potentially minimise extrahepatic exposure. HAI has been evaluated in multiple clinical situations in isolated colorectal metastases, including neoadjuvant, adjuvant and as first-line therapy, either singly or in combination with systemic chemotherapy. Floxuridine (FUDR) has been the most widely used drug in this setting. When given as HAI, over 90% of FUDR is extracted by the liver, resulting in high levels of its active metabolite within the liver and little systemic exposure.(38) Several previously completed trials of HAI administration of FUDR versus systemic 5FU for patients with unresectable colorectal liver metastases demonstrated significantly higher tumour RRs for HAI of FUDR compared with systemic 5FU.(39,40) These trials also showed significantly longer time to hepatic progression with HAI of FUDR compared with systemic 5FU. Despite the higher RR with FUDR, conflicting results were seen in OS. A meta-analysis showed no improvement in OS in patients receiving FUDR.(41)

Given interest in liver-directed therapy, several phase I and II trials have assessed HAI of agents other than FUDR. One recent trial treated 26 patients with unresectable liver-only metastases with HAI of ­oxaliplatin together with systemic 5FU and ­leucovorin; RR was 64%, with two complete responses.(42) Five patients were able to undergo surgical resection of the residual disease. The median overall survival was 27 months. While this trial demonstrated feasibility of this approach, the rationale for giving FUDR by HAI is based on the near complete extraction of the drug into the liver. This does not appear to be the case for ­oxaliplatin or other drugs active in ­colorectal cancer. It is therefore important to question the wisdom of such an approach, given the increased ­morbidity associated with the placement and use of a HAI catheter.

In a randomised trial assessing the potential efficacy of combining HAI FUDR with systemic 5FU/LV as adjuvant therapy after resection of liver-only metastases, an improvement in two-year OS (86% vs 72%, p=0.03) was observed in patients randomised to the combination arm over those randomised to the systemic therapy-alone arm.(43) However, there was no statistically significant improvement in five-year OS between the two groups. Recently, updated results from this trial showed a median survival of 68.4 months in the combination arm, compared with 58.8 months in the systemic therapy arm (p=0.10).(44) In a separate multi-institutional trial, 135 patients with unresectable hepatic metastases were randomised to either HAI with FUDR or to systemic 5FU/LV.(45) The RRs were higher in patients receiving HAI (47% vs 24%, p=0.012) and OS was a significantly increased (24 months vs 20 months, p=0.034). However, the time to extrahepatic progression was much shorter with HAI (7.7 vs 14.8 months) and there was more biliary toxicity.

For a variety of reasons, HAI has not gained widespread use. Even though HAI was found to produce superior responses in the liver, most trials have not shown meaningful improvements in overall survival. Furthermore, the use of HAI FUDR alone does not provide meaningful therapy for extrahepatic sites of disease. The current availability of less invasive and potentially more effective options, with agents such as oxaliplatin, irinotecan, bevacizumab and cetuximab, has further lessened the use of HAI-based therapy.

Since all the previous combinations trials were with bolus 5FU/LV, it would be important to determine whether using systemic oxaliplatin or ­irinotecan concurrently with HAI would improve RRs and chances of resection. In a recent phase I trial, HAI was combined with systemic irinotecan and/or ­oxaliplatin in mostly pretreated patients (89%), with excellent RR, enabling a few patients to undergo surgical resection.(46) It is unclear whether combining chemotherapy with HAI would have a synergistic effect, making it superior to chemotherapy alone in improving chances of resection. We will need randomised trials to test this hypothesis. Until then, considering the toxicities and the inconvenience, HAI has a limited role in mCRC.

Conclusions
There is a fresh optimism in the management of mCRC with the discovery of new cytotoxic and biological agents. We have made great strides in improving the DFS and OS. This has to be tempered by the fact that colorectal cancer continues to be a major killer and OS rates are still only in the range of two years. With the expense of current therapies that combine chemotherapy with targeted agents, rational approaches to treatment requires evidence-based guidance from well-designed randomised clinical trials. There are many ongoing trials addressing these issues and the results are eagerly awaited.

References

1. WHO. Global cancer rates could increase by 50% to 15 million by 2020. Available from: http://www.who.int/mediacentre/releases/2003/pr27/en/print.html
2. CA Cancer J Clin 2006;56:106-30.
3. J Clin Oncol 1994;12:14-20.
4. Dorr TR, Von Hoff DD. Drug monographs. In: Dorr TR, Von Hoff DD,  editors. Cancer chemotherapy handbook. 2nd ed. East Norwalk, CT: Appleton and Lange; 1994. p. 500-15.
5. J Clin Oncol 1998;16:301-8.
6. N Engl J Med 2000;343:905-14.
7. Lancet 2000;355:1041-7.
8. J Clin Oncol 2004;22:23-30.
9. N Engl J Med 2001;345:144-5.
10. J Clin Oncol 2006;24:394-400.
11. Eur J Cancer 1999;35:1343-7.
12. J Clin Oncol 2004;22:229-37.
13. J Clin Oncol 2005;23:4866-75.
14. ASCO GI Cancers Symposium, Hollywood, FL, 27–29 January 2005;Abstract 165.
15. J Clin Oncol 2005;23:9441-2.
16. J Clin Oncol 2001;19:4097-106.
17. J Clin Oncol 2004;22:2084-91.
18. J Clin Oncol 2004;22 270s Suppl 14:Abstract 3602.
19. J Clin Oncol 2004;22 253s Suppl 14;Abstract 3534.
20. J Clin Oncol 2005;23 252s Suppl 16;Abstract 3525.
21. J Clin Oncol 2005;23 249s Suppl 16;Abstract 3515.
22. J Clin Oncol 2005;23 247s Suppl 16;Abstract 3507.
23. Falcone A, et al. ASCO GI Cancers Symposium, San Francisco, CA, 26–28 January 2006;Abstract 199.
24. N Engl J Med 2004;350:2335-42.
25. Giantonio BJ, et al. GI Cancers Symposium; Hollywood, FL, 27–29 January 2005;Abstract 169a.
26. Hochester HS, et al. ASCO GI Cancers Symposium, Hollywood, FL, January 27–29 January 2005;Abstract 241.
27. N Engl J Med 2004;351:317-9.
28. J Clin Oncol 2005;23:4853-5.
29. N Engl J Med 2004;351:337-45.
30. J Clin Oncol 2004;23:266s Suppl 16;Abstract 3580.
31. J Clin Oncol 2004;23:254s Suppl 16;Abstract 3535.
32. J Clin Oncol 2005;23:1803-10.
33. Saltz L, et al. ASCO GI Symposium, Hollywood, FL, 27–29 January 2005;Abstract 169b.
34. J Clin Oncol 2005;23:7125-34.
35. Ann Surg Oncol 2001;8:347-53.
36. Ann Oncol 2005;16:425-9.
37. J Clin Oncol 2005;23:9243-9.
38. Semin Oncol 2002;29:119-25.
39. J Clin Oncol 1992;10:1112-8.
40. Arch Surg 1990;125:1022-7.
41. J Natl Cancer Inst 1996;88: 252-8.
42. J Clin Oncol 2005;23:4881-7.
43. N Engl J Med 1999;341:2039-48.
44. N Engl J Med 2005;352:734-5.
45. J Clin Oncol 2006;24:1395-403.
46. J Clin Oncol 2005;23:4888-96.