759 Background: Based on results from phase III clinical trials chemotherapy treatment benefit in patients ≥70 years old with stage III and IV colorectal cancer (CRC III-IV) is a subject in question. The ELDERLY trial aims to explore in a pilot program viability and convenience of an evidence based protocol for Comprehensive Geriatric Asessment (CGA) as well as the contribution of further evaluations. Methods: Patients ≥70 years old with CRC III-IV were prospectively and consecutively included. After initial treatment recommendation was made, study protocol for CGA (functional, nutritional, cognitive, social, psychological, and comorbidities) was carried out. Based on CGA, early death risk prediction model and chemotherpy toxicity risk stratification subsequent final treatment recommendation was obtained. Follow up of patients and comparison to historical controls is underway. Results: 20 patients (10 CRC III, 10 CRC IV), median age 79.2 years (71-85), 65% males, were included. ECOG PS was: 0 in 35% of patients, 1 in 35%, 2 in 20%, and 3 in 10% of patients. Initial treatment recommendation was: standard for 60% of patients, adapted standard for 25%, and no treatment for 15% of patients. According to Köhne et al. (Oncologist 2008) fitness categories 25% of patients were in group I, 55% in group II, and 20% of patients in group III. Early death risk, according to Soubeyran et al. (J Clin Oncol 2012), was high for 5% of patients. Chemotherapy toxicity risk, according to Hurria et al. (J Clin Oncol 2011) was low for 15% of patients, intermediate for 60%, and high for 25% of patients. Final treatment recommendation was: standard for 25% of patients, adapted standard for 55%, and no treatment for 20% of patients. Initial and final treatment recommendation diverged in 55% of patients (more conservative final treatment recommendation in 45%, and more intense final treatment recommendation in 10%). Conclusions: Evidence based protocol CGA modified initial treatment recommendations. Follow up of patients and comparison to historical controls could contribute to improve the knwoledge and convenience of this and further evaluations.