THE MODEL FOR DRUG DEVELOPMENT IN SOLID TUMOR malignancies in general, and in colon cancer in particular, has been first to identify a new agent that is effective in patients with refractory disease, and then to use that agent (alone if it is active enough, or in combination with the older agent or regimen if it is not) for patients with metastatic disease and demonstrate improved overall survival, progression-free survival, or both in a randomized trial against the previous standard. That new agent or new agent-containing combination is then used in a randomized trial in the adjuvant setting with the hope of improving the cure rate after definitive surgery. This approach is time-honored and is based on rational expectations, an acceptable apportionment of risk tolerance among patients at various stages of their diseases, and even some positive test results. But what if this strategy is wrong? In this issue of JAMA, Alberts et al report the results of the National Cancer Institute (NCI) intergroup study N0147, a randomized phase 3 trial assessing the usefulness of cetuximab, an anti-epidermal growth factor receptor monoclonal antibody, as adjuvant treatment for patients with resected stage III colon cancer. Cetuximab has been shown to be active against metastatic colon cancer, both in combination with chemotherapy and as a single agent. During the conduct of N0147, it became recognized that colon cancers with mutations in exon 2 of the KRAS gene were refractory to cetuximab and related agents, and therefore only patients with tumors that have a wild-type KRAS gene have the potential to respond. Thus, KRAS genotyping of the tumor permitted a rational selection of appropriate patients for cetuximab therapy. With the availability of this information, the N0147 trial was modified in mid course and powered to assess the effect of cetuximab for adjuvant treatment of patients with stage III colon cancer and with wild-type KRAS tumors only. Prior investigations had established the combination regimen of fluorouracil, leucovorin, and oxaliplatin (FOLFOX) as the standard postoperative regimen for stage III colon cancer; therefore, the N0147 trial became FOLFOX with or without cetuximab following resection of stage III colon cancer in patients with wild-type KRAS tumors only. How could the enhanced regimen not be successful? Could the addition of an agent with substantial activity in the metastatic setting combined with a highly reliable marker that identifies up to 40% of the total patient population that would not benefit, possibly fail to increase the cure rate in the adjuvant setting? To the surprise of many and the disappointment of all, the findings from the N0147 trial were negative. The percentage of patients alive and free of their cancer at 3 years from the date of their surgery was not improved, and the toxicity of the regimen was substantial. In fact, patients 70 years or older who received cetuximab had worse outcomes than those who received placebo. Is there precedent for such an outcome? Can agents with reliably demonstrated activity in the metastatic setting fail to provide benefit in the adjuvant setting? In colon cancer, this scenario has occurred before and in the history of adjuvant therapy for colon cancer, it is all too common. Fluorouracil, often biomodulated by the reduced folate leucovorin, was the only active drug for colon cancer until 1996, when irinotecan, a topoisomerase I inhibitor, received accelerated regulatory approval because of its demonstrated activity, including a survival benefit, in patients with fluorouracilrefractory metastatic colorectal cancer. Although irinotecan was shown to be no more active than fluorouracil, the combination of irinotecan and fluorouracil conferred a modest but statistically significant survival benefit compared with fluorouracil alone. Appropriately, on both sides of the Atlantic, investigators began the expensive and laborious task of testing irinotecan-fluorouracil combinations in stage III colon cancer. Many clinicians and patients objected to these studies, suggesting that they were unnecessary and even unethical, because the answer was already “known.” The precarious assumption was that if irinotecan alone conferred a survival benefit as a single agent in second-line treatment for meta-