Colon Cancer Gene-Expression Risk Scores Contribute Little to Prognostic Ability of Established Risk Factors
In a study reported in the Journal of the National Cancer Institute, Di Narzo et al found that individual colon cancer gene-expression risk scores contributed little to the prognostic ability of traditional risk factors.
Study Details
The study involved evaluation of four gene-expression risk scores using formalin-fixed paraffin-embedded microarrays in a cohort of 688 patients with stage II/III tumors from the PETACC-3 trial. The risk scores consisted of scoring systems proposed by researchers from Almac (ALM), Genomic Health (GHS), The University of Texas MD Anderson Cancer Center (MDA), and Verdex (VDA). The established risk model consisted of T stage, N stage, and microsatellite instability status.
Prognostic Ability
All four risk scores were significantly associated (P ≤ .0167) with overall survival or relapse-free survival in univariate models, but hazard ratios (HRs) were < 1.38 per interquartile range. In multivariate analysis including age, gender, TNM staging, grade, location, treatment group, presence of lymphovascular invasion, and microsatellite instability status, three scores were significantly associated (P < .05) with relapse-free survival (GHS, VDS, and ALM; HRs = 1.20–1.30), one with survival after relapse (MDA, HR = 1.89), and three with overall survival (GHS, MDA, and ALM; HRs = 1.22-1.37). In receiver-operating characteristic area under the curve (AUC) analysis of risk models, all risk scores only slightly improved relapse-free survival or overall survival prognostic ability of the risk model based on TNM staging and microsatellite instability status, with AUC gains of < 0.025 units.
Combined Risk Score
A combined risk score derived by obtaining the average of the four risk scores was associated with greater prognostic value and increased AUC gain vs individual risk scores for overall survival (HR = 1.74, P < .001, on univariate analysis; AUC increase of 0.6918 to 0.7321) and relapse-free survival (HR = 1.56, P < .001, on univariate analysis; AUC increase of 0.6723 to 0.6945).
The investigators concluded: “The four tested gene expression-based risk scores provide prognostic information but contribute only marginally to improving models based on established risk factors. A combination of the risk scores might provide more robust information. Predictors of [relapse-free survival] and [survival after relapse] might need to be different.”
Mauro Delorenzi, PhD, of SIB Swiss Institute of Bioinformatics and University Lausanne, is the corresponding author for the Journal of the National Cancer Institute article.
The study was supported by the Swiss National Science Foundation, Krebsliga Schweiz, European Union 7th Framework Programme, and Fondation Medic. For full disclosures of the study authors, visit jnci.oxfordjournals.org.
The content in this post has not been reviewed by the American Society of Clinical Oncology, Inc. (ASCO®) and does not necessarily reflect the ideas and opinions of ASCO®.
