Hurdles and Complexities of Codon 13KRASMutations

The history of biomarker development for epidermal growth factor receptor (EGFR) monoclonal antibodies for patients with colorectal cancer has been a convoluted path. Each development has provided increased recognition that there is still much to learn about biomarkers for EGFR monoclonal antibodies in colorectal cancer. In the article that accompanies this editorial, Tejpar et al provide an added chapter to this complex issue and raise important questions about the level of evidence that we should require before implementing predictive biomarkers in clinical practice. Monoclonal antibodies to EGFR have demonstrated a survival benefit in multiple randomized phase III studies, including those involving patients with previously untreated and refractory metastatic colorectal cancer. Initial studies limited enrollment of patients to those whose tumors expressed detectable levels of EGFR by immunohistochemistry, primarily on the basis of the premise that the target must be detectable for patients to derive benefit. However, subsequent studies demonstrated that activity is also observed in patients whose tumors did not stain for EGFR, probably because of the limits of detection by immunohistochemistry rather than true absence of the receptor, leading to the abandonment of this test in clinical practice. A successful biomarker came with the identification of the role of codon 12 and 13 KRAS mutations as an apparently clear and convincing negative predictive biomarker. These mutations are in the phosphate binding loop of KRAS and reduce the rate of hydrolysis of guanosine triphosphate, thereby leaving KRAS in an active state. The initial studies evaluated any mutations in codon 12 and 13 together because of their steric similarities and their ability to induce autonomous growth in transformation assays. Efforts subsequently turned to other potential biomarkers, including activating mutations in BRAF, PIK3CA, NRAS, and other mutations in KRAS such as codon 61 and 146. Studies mostly used retrospective nonrandomized cohorts and suggested a lack of benefit from EGFR monoclonal antibodies in the presence of these mutations. The limitations of this approach became clear as subsequent studies demonstrated a more complicated picture, with the prognostic effects of these mutations obscuring potential predictive effects, and with inadequate statistical power as a result of the low prevalence of many of these mutations. With these missteps along the way, it should not be surprising that the story of KRAS as a predictive marker for EGFR monoclonal antibodies is more complex than we have previously appreciated. KRAS codon 13 mutations are present in approximately 5% of tumors from patients with metastatic colorectal cancer, and are almost always a replacement of the glycine amino acid with aspartate (G13D). The initial report questioning the role of codon 13 KRAS mutations as a negative predictive biomarker was published by De Roock et al using a pooled analysis of 579 patients with chemotherapy-refractory metastatic colorectal cancer who were treated with cetuximab. This study showed that cetuximab improved median overall survival and progression-free survival in those patients whose tumor had a KRAS G13D mutation (n 32), in contrast to patients whose tumor carried the more common codon 12 KRAS mutations. However, a subsequent analysis of codon 13 KRAS mutations in patients who received panitumumab-containing regimens failed to confirm these findings; this study demonstrated that codon 13 mutations remained a statistically significant negative predictive biomarker, similar to codon 12 mutations. In this context, the report by Tejpar et al explored the question of whether cetuximab in combination with chemotherapy provided benefit to previously untreated patients with metastatic colorectal cancer whose tumors contained a mutation in codon 13 of KRAS. This retrospective analysis of 1,378 patients demonstrated that patients whose tumors carry a codon 13 KRAS mutation (n 83) have a trend toward an improved progression-free survival and response rate, but not overall survival, compared with patients whose tumors contained a codon 12 KRAS mutation. Although the primary analysis was not statistically significant, there were imbalances in the baseline characteristics that led the authors to subject the study to multivariate analysis. In the multivariate model, these differences reached significance, with a magnitude of benefit in patients with codon 13 mutations that was similar to that of patients with KRAS wild-type tumors. The conclusions from the article by Tejpar et al appropriately indicate that these results remain hypothesis generating and not ready for application in clinical practice. Despite the fact that thousands of patients have been treated with EGFR monoclonal antibodies in randomized controlled trials and extensive biomarker analyses have been conducted, there is still uncertainty about how to appropriately assess these biomarker studies. To this end, the codon 13 KRAS quandary provides a useful framework for discussion of the threshold of evidence that is needed to clinically implement a predictive biomarker. From our perspective, rationally addressing two hurdles is necessary to establish confidence in a potential predictive biomarker before it can be integrated into clinical practice: a biologic explanation for the JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L VOLUME 30 NUMBER 29 OCTOBER 1