331P Predicting outcomes following colorectal cancer resection: Using real-world data to empower adjuvant treatment decision making

Abstract

Guidelines advocate adjuvant therapy for selected patients with resected stage II-III colorectal cancer. Risks however remain poorly defined. ESMO guidelines describe 5-year survival after resection alone of 68-83% for stage II and 45-65% for stage III; rates can be improved by 3-5% in high-risk stage II colon cancer and 10-15% in stage III with 5FU, and a further 4-5% with oxaliplatin. These windows are wide and do not take many patient factors into account. Informed decision making with discussion of the risk to benefit ratio of chemotherapy is paramount. Grade 3-4 toxicities can occur in as many as 25-55% of patients, depending on regimen and duration of treatment. Such wide ranges make balancing potential benefit and risk very challenging. In breast cancer, the widely used PREDICT online tool allows clinicians and patients to quantitively evaluate risk in a simple and user-friendly manner. There is a clear need for a similar tool in GI oncology. The CORECT-R database is a curated, linked dataset of over 600,000 patients with colorectal cancer in England. Data is available on patient and tumour characteristics, treatments and outcomes. Using overall survival as our primary outcome we performed Cox proportional hazards modelling to compare patients receiving chemotherapy with those who did not. Multivariate modelling including both patient factors, such as age and frailty, and tumour factors, such as staging information, was used. A prototype web interface to view these data was built. We present a prototype web-based tool to predict the benefits of adjuvant chemotherapy, with appropriate confidence intervals, based on a large dataset of real-world data. Quantitative, user-friendly tools such as PREDICT are invaluable for empowering patients to make informed decisions regarding adjuvant treatment. Here we use real-world data at unprecedented scale to provide a proof-of-concept tool for colorectal cancer. Following further validation, this could be transformative for adjuvant treatment decision making.