Abstract LB-025: A decision support system for the treatment of esophageal cancer

Abstract

Abstract Introduction In oncology deciding which treatment to use and when to apply them in an optimal way can be challenging. Currently there are no global approaches available to optimise the treatment path for an individual patient. Focusing on esophageal cancer, we have developed a prototype of a model-based Decision Support System (DSS). It integrates a novel range of drug, tumour and patient data to better plan clinical treatment, optimise the patient care path, and ultimately deliver improved cancer care. Methods This tool was developed using data from an historical cohort of patients with operable esophageal cancer. This cohort comprised 465 patients who were treated at Oxford University Hospitals NHS Foundation Trust between 2007 and 2014. An initial data analysis focussed on using un-supervised learning methods to identify key variables that differentiated groups of patients. These variables were then used to analyse their correlation to survival and key decision points using parametric statistical models. The final models were then used to create a web-based DSS prototype. The prototype allows users to see throughout a patient's treatment journey what factors influence prognosis including key interventions such as surgery. Results The key results within the analysis centered around treatment effects on FDG-avid nodal status, using positron emission tomography-computed tomography (PET-CT). We found that change in nodal avidity due to chemotherapy is a key prognostic factor. Furthermore, this change correlated with the change in avidity and size of the primary tumour i.e. there is a degree of correlation between the primary tumour and nodes. No difference was found between the four chemotherapy regimens analysed in their ability to reduce tumour size and also change nodal status. However a hypothesis that might be tested through further research is that titrating duration of treatment and doses to the achievement of change in avid nodal status may have a significant impact on individual patient prognosis. The above key results as well as others pertaining to attempted and successful surgery can all be explored using the dynamic web-based interface developed. Conclusion We have created a DSS tool using historical data to support Cancer Multidisciplinary Teams in understanding how key variables affect treatment choices and how these in turn relate to disease outcomes. The tool has the potential to be used to support the optimisation of treatment for individual patients based on the characteristics of their disease. Thus the DSS tool developed is a first step to personalised treatment of cancer. Citation Format: Hitesh Mistry, Fernando Ortega, Frances A Brightman, John M Findlay, Mark R Middleton, Christophe Chassagnole. A decision support system for the treatment of esophageal cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr LB-025.