Research

Data-Driven Assessment of the Association of Polymorphisms in 5-Fluorouracil Metabolism Genes with Outcome in Adjuvant Treatment of Colorectal Cancer

Read the article

Abstract

A major challenge in the assessment of medicines, treatment options, etc., is to establish a framework for the comparison of risks and benefits of many different types and magnitudes, a framework that at the same time allows a clear distinction between the roles played by the statistical analyses of data and by judgements based on personal experience and expertise. The purpose of this study was to demonstrate how clinical data can be weighted, scored and presented by the use of an eight-step data-driven benefit–risk assessment method, where two genetic profiles are compared. Our aim was to present a comprehensive approach that is simple to apply, allows direct comparison of different types of risks and benefits, quantifies the clinical relevance of data and is tailored for the comparison of different options. We analysed a cohort of 302 patients with colorectal cancer treated with 5-Fluorouracil (5-FU). Endpoints were cure rate, survival rate, time-to-death (TTD), time-to-relapse (TTR) and main adverse drug reactions. Multifactor dimensionality reduction (MDR) was used to identify genetic interaction profiles associated with outcome. We have been able to demonstrate that a specific MDR-derived combination (the MDR-1 group) of dihydropyrimidine dehydrogenase and thymidylate synthase polymorphisms is associated with increased and clinically significant difference for cure and survival rates, TTD and probably also for TTR, which are seen as the most important endpoints. An inferior profile was observed for severe myocardial ischaemia. A probably inferior profile was seen for severe arthralgia/myalgia and severe infections. A clear superior profile was seen for severe mucositis/stomatitis. The proposed approach offers comprehensive, datadriven assessment that can facilitate decision processes, for example, in a clinical setting. It employs descriptive statistical methods to highlight the clinically relevant differences between options.