A multivariate model to define risk groups among patients with curatively resected stage I and II colon carcinoma: Final report of a retrospective cohort study.

Abstract

696 Background: Patients with stage I or II colon carcinoma (CC) have a 10 or 20% risk, respectively, of presenting recurrent disease after a potentially curative surgical resection and adjuvant chemotherapy have not improved survival in this setting. In this study, we present a multivariate model to define risk groups. Methods: Consecutive cases with stage I and II CC treated at a single cancer center in Mexico City, from January 1992 to December 2016, were included in this 25-year cohort. Clinical history and biochemical data were registered, and colon resection was performed with curative intention with standard lymphadenectomy. Standard hematoxylin-eosin slides and CDX2 immunohistochemistry (IHC) slides were analyzed by two independent pathologists. The Kaplan-Meier method and Cox model were used to analyze the association of prognostic factors and overall survival (OS). Results: 3,301 cases of colorectal cancer were treated during this study, but only 556 patients with stage I and II CC were included in the database: 266 women (47.8%) and 290 males (52.2%) (mean age 57.9 years); 52 (9.4%), 431 (77.5%), 36 (6.5%) and 37 (6.7%) were pTNM stages I, IIa, IIb, and IIc, respectively. R0 resection was performed in 548 patients (98.6) and R1 in 8 (1.4%). Location in the left colon (HR 1.63), hemoglobin (HR 0.93), serum albumin (HR 2.45), prognostic nutritional index 0.915), platelet count (HR 0.998), body mass index (HR 0.94), basal carcinoembryonic antigen (HR 1.0), TNM stage [stage I reference category, IIa (HR 5.46), IIb (HR 4.42), IIc (HR 9.28)], R1 residual disease (HR 2.8), negative CDX2 IHC (HR 2.1), and use of adjuvant chemotherapy (HR 0.629) were included in the final model as independent predictors of OS (model p < 0.0001).Predicted survival functions using this model defined three distinct risk groups. Conclusions: This multivariate model adds significant prognostic value to the pTNM classification. This model can be useful for stratifying prognosis in patients with CC and will aid in the design of randomized clinical trials evaluating the usefulness of adjuvant chemotherapy in this subgroup of patients with CC.