Improving the prediction of colon cancer survival after curative resection.

Abstract

552 Background: Cancer staging systems convey valuable prognostic information to both clinicians and patients. Currently, colon cancer is staged according to the American Joint Committee on Cancer (AJCC) TNM classification system. However, survival estimates for patients with the same stage of colon cancer may vary considerably due to other factors including age, sex, grade, and number of lymph nodes sampled. The objectives of this study are to 1) assess the accuracy of the seventh edition of the TNM classification system in predicting survival of patients with primary colon cancer after curative-intent surgery, and 2) evaluate the utility of incorporating additional demographic and tumor variables beyond TNM staging in improving prognostic accuracy. Methods: Patients with curative-intent resection of a first primary adenocarcinoma of the colon at the time of referral to the Cross Cancer Institute between 2004 and 2007 were identified from the Alberta Cancer Registry. We constructed three multivariate Cox’s proportional hazard models to explore the effect of supplementing TNM staging with additional demographic and tumor variables in predicting overall survival (OS). Results: 559 consecutive patients with complete chart records were identified. 52 % (n=290) were male; median age was 74. In the first model based only on T and N elements, N2 disease was correlated with increased mortality (hazard ratio [HR], 2.546; p<0.0001). When the number of lymph nodes examined (HR, 0.980; p=0.034) and number of metastatic lymph nodes detected (HR, 1.094; p<0.0001) were substituted for the N-staging element, both variables correlated positively and negatively with outcome, respectively. Finally, when tumor grade, sex and age were incorporated into the model, number of examined lymph nodes (HR, 0.980; p=0.029) and those containing tumor (HR, 1.093; p<0.0001) remained independent predictors of OS. Conclusions: Incorporating readily available demographic and tumor variables, such as age, sex and number of lymph nodes examined, can enhance the current TNM staging system and improve prognostication in early stage colon cancer.