In a recent issue of The New England Journal of Medicine, Dalerba et al published an impressive article describing a novel bioinformatics approach to identifying new prognostic and predictive biomarkers in patients with stage II and III colon cancer (see summary in this issue of The ASCO Post).1 They mined an extensive gene-expression database of approximately 2,300 tumor samples for a marker that reflected an unfavorable stem cell phenotype and, simultaneously, could be easily utilized in the clinical setting.
The transcription factor CDX2 emerged as the only target for which a diagnostic assay was already in existence. The authors subsequently went on to show that CDX2 negativity is both prognostic for poor disease-free survival and predictive for adjuvant chemotherapy benefit in patients with stage II and III colon cancer. Given the clinical importance of reliably identifying subsets of patients with stage II disease who have a higher risk of recurrence, the emergence of CDX2 as a potential biomarker was met with excitement and enthusiasm. But is CDX2 ready for prime time?
Study Strengths
The analysis by Dalerba et al possesses many strengths. It is an extremely large study that serves as a model of effective national and international collaboration among different institutions and scientists. The authors should be commended for considering the real-world impact and utility of their analysis by limiting their search for a biomarker to one with a widely available diagnostic test. They took advantage of existing resources and biorepositories to evaluate the association among CDX2 status, patient outcome, and chemotherapy benefit and rigorously confirmed their findings in a validation cohort utilizing the clinically available immunohistochemical assay.
Kimmie Ng, MD, MPH
They found that patients with CDX2-negative tumors had a significantly lower 5-year disease-free survival than did those with CDX2-positive tumors after multivariable adjustment (P = .003), and this association persisted within the subgroup of stage II patients. Importantly, they also found that CDX2 negativity was predictive of greater benefit from adjuvant chemotherapy in patients with both stage II and stage III colon cancer, as evidenced by significant P values for interaction of .02 and .005, respectively.
A Word of Caution
Despite these encouraging results, it is premature to start incorporating CDX2 immunohistochemistry into standard clinical practice due to the retrospective nature of the analysis and the extremely small number of patients who have CDX2-negative tumors. Indeed, only 6.9% of tumors demonstrated negative CDX2 gene expression, and only 13% demonstrated CDX2-negative protein expression, with limited survival events. With such small numbers of patients to work with, it is extremely challenging to evaluate the relationship of CDX2 status with other relevant clinical and molecular features such as grade of differentiation, number of lymph nodes sampled, microsatellite instability, BRAF mutation status, and others. Although Dalerba et al did make a valiant attempt to distinguish CDX2-negative tumors as a distinct subset from high-grade and microsatellite instability–high tumors, the sample sizes were so small that no definitive conclusions can be drawn despite the reported P values.
Moreover, although they were able to adjust for grade of differentiation in their multivariable models, they did not control for microsatellite instability status. In addition, there did seem to be a higher rate of microsatellite instability–high tumors among patients with CDX2-negative colon cancer, leading to the conundrum of how to think about adjuvant fluorouracil-based chemotherapy in patients with stage II colon cancer with both microsatellite instability–high and CDX2-negative tumors. Finally, the question of whether CDX2 negativity is really predictive of greater benefit from adjuvant chemotherapy in patients with stage II colon cancer is still under debate, as the patient cohorts that were used to study this question were not drawn from randomized clinical trials.
Confirmatory Studies Needed
Dalerba et al are the first to admit that their results need to be confirmed in larger, prospective studies within the framework of randomized studies. An ideal design would involve randomization of patients with CDX2-negative stage II colon cancer to adjuvant chemotherapy vs surgery alone, but such a trial would be extremely difficult to conduct given the small number of CDX2-negative tumors and the overall low recurrence rates and favorable survival associated with stage II disease.
However, in light of the fact that KRAS and NRAS have become accepted predictive markers of anti-EGFR (epidermal growth factor receptor) antibody efficacy based on retrospective analyses of large randomized clinical trials, it is possible that future analyses of large, completed randomized clinical trial cohorts could provide the same validation for CDX2. Future research efforts should also be directed at the following tasks: (1) determining the rate of interobserver agreement in assessing CDX2 status by immunohistochemistry in the “real world”; (2) assessing whether CDX2 gene expression and CDX2 protein expression are truly representative of the same poor prognostic subset of patients; (3) understanding the interaction of CDX2 with other relevant clinical and molecular factors; (4) evaluating which chemotherapy regimen is most beneficial for patients with CDX2-negative stage II disease, if it is indeed a predictive marker; and (5) establishing the cost-effectiveness of CDX2 testing. ■
Disclosure: Dr. Ng reported no potential conflicts of interest.
Dr. Ng is Director, Clinical Research, and Gastrointestinal Cancer Center Physician at Dana-Farber Cancer Institute in Boston.
