The importance of adjuvant therapy for advanced-stage colon cancer was brought out by a pivotal trial by the North Central Cancer Treatment Group (NCCTG) (2) and a subsequent intergroup trial sponsored by the National Cancer Institute (NCI), (3) which demonstrated the effectiveness of fluorouracil and levamisole in patients with positive lymph nodes (stage III). This treatment, moreover, has been shown to be highly cost effective (4). These results have stimulated vigorous interest in testing other fluorouracil-based regimens in patients with stage III colon cancer. By contrast, the optimal treatment for patients with stage II colon cancer (tumor that invades muscularis propria of the bowel wall but has no lymph-node metastases) is unclear. Neither the NCCTG study nor the NCI intergroup study provided definitive direction, because they included small numbers of patients with stage II disease. Currently, no cooperative or intergroup trial is comparing adjuvant therapy with "watchful waiting" in patients with stage II colon cancer.
One reason for the uncertainty is that patients with stage II colon carcinoma have a good prognosis. A five-year survival of 78 percent was found by the Gastrointestinal Tumor Study Group in an analysis of 356 patients with stage II disease and a median follow-up of 8.5 years (5). The clinical problem is therefore straightforward: how can we select the patients with stage II disease who will benefit from adjuvant therapy and avoid unnecessary treatment of patients with a low probability of recurrence?
In this issue of the Journal, Jen et al. (6) report intriguing results concerning a new molecular marker (allelic loss of chromosome 18q) that appears to predict a poor outcome in patients with stage II colon cancer. Using a technique that works with stored tissue, the authors looked for chromosome 18q allelic loss in tumors from 144 consecutive patients with stage II or III colorectal carcinoma diagnosed during a period when adjuvant therapy was not routinely used. With appropriate multivariate statistical techniques, chromosome 18q allelic loss remained predictive for mortality even after adjustment for other important features, such as tumor differentiation, the presence or absence of vascular invasion, and disease stage. The five-year survival rate among patients with stage II disease and intact chromosomes 18q was extraordinarily good (93 percent), whereas in patients with chromosome 18q allelic loss it was comparable to the rate in patients with stage III disease.
Jen et al. (6) found that chromosome 18q allelic loss failed to predict outcomes in patients with stage III disease, and they give some explanations for that result. We suggest that their study lacked the statistical power to detect a survival difference in the patients with stage III disease. Although 70 patients with stage III disease were evaluated, only about half of them died, and in only 23 percent of them was chromosome 18q intact. Statistical calculations suggest that the study had a power of only 40 percent to detect a two-fold increase in the risk of death for those with chromosome 18q allelic loss. Thus, further investigations may be warranted in patients with stage III disease.
Other questions remain. Do patients with colon cancer differ from patients with rectal cancer in the expression of this marker? Only 25 (19 percent) of the 135 patients studied for chromosome 18q allelic loss had rectal cancer. Therefore, even though the authors found no differences between patients with colon cancer and those with rectal cancer, their results speak primarily to the usefulness of chromosome 18q allelic loss as a prognostic factor for colon cancer.
Why did the investigators fail to examine and compare DNA ploidy and cell kinetic measurements? Witzig et al. (7) suggest that patients with stage II disease with nondiploid tumors and a high index of proliferation have a poor survival rate -- 70 percent at five years, as compared with 87 percent among patients with diploid tumors and a low index of proliferation. Is the prognostic information provided by allelic loss of chromosome 18q independent of that provided by DNA ploidy? A recent report suggests that cytoplasmic accumulation of p53 protein in colorectal carcinoma is also an independent prognostic indicator of overall survival, but with additional analysis according to DNA ploidy, the p53 indicator was significant only with respect to disease-free survival (8).
Where do we go from here? Jen et al. suggest that the adjuvant therapy currently offered to patients with stage III disease may be appropriate for patients with stage II disease and chromosome 18q allelic loss. However, the reasons for the increased risk of death among these patients with stage II disease may differ considerably from those among patients with stage III disease. Therefore, we cannot assume the effectiveness of levamisole combined with fluorouracil in patients with stage II disease and chromosome 18q allelic loss. An initial step might be to test for chromosome 18q allelic loss in stored tissue from the NCI intergroup trial. This might shed light on the effectiveness of adjuvant therapy in patients with stage II disease and chromosome 18q allelic loss and establish the prognostic value of the marker in a larger sample of patients with stage III disease. Alternatively, a new trial of adjuvant therapy can be conducted in patients with stage II disease and chromosome 18q allelic loss. Because we do not know whether adjuvant therapy is beneficial in this subgroup, we can ethically randomize patients to observation or therapy. A sample of about 500 patients would be required to provide a power of 80 percent to detect a survival benefit equal to that among the patients with stage III disease studied in the intergroup trial. This model could also be used to test other promising strategies (such as passive immunotherapy with the monoclonal antibody 17-1A) (9) in an untreated control group.
We encourage American cooperative-trial groups studying cancer therapy to mount a simultaneous analysis of new biologic and molecular markers in stored tissue. Using this resource may be the only way to evaluate the relative predictive value of these markers. In a cost-conscious medical environment we must carefully assess the incremental benefit of new prognostic markers before we can make global recommendations about their usefulness in clinical practice.
Margaret Tempero, M.D.
James Anderson, Ph.D.
University of Nebraska Medical Center
Omaha, NE 68198-3330