The Cancer Care Treatment Outcomes Ontology (CCTO): A computable ontology for profiling treatment outcomes of patients with solid tumors.

Abstract

e18137 Background: Systematic enhancement of knowledge about outcomes of cancer therapy is a central goal for oncology research, practice and audit. However, there is no resource standardizing the disparate vocabulary used to describe treatment outcomes, which prevents translating the data acquired into value-adding information for better cancer care. Methods: To build an application ontology that organizes high-level concepts describing treatment outcomes in solid tumors, provisionally named Cancer Care Treatment Outcomes (CCTO), we extracted all trial endpoints listed in ClinicalTrials.gov, queried it using the keyword ‘cancer’, followed by an expert appraisal. Term definitions and synonyms were automatically imported from the National Cancer Institute thesaurus, and extra terms were imported from Common Toxicity Criteria for Adverse Events. Logical relationships between concepts were represented by forward chaining rules, aiding conversion between concepts. The coverage of CCTO was tested against MEDLINE abstracts, and the applicability of 1825 rules was tested in a case of metastatic colon cancer (index case). Results: After removing duplicated terms from 54705 trial entries, an ontology (CCTO), holding 1145 terms and organized into 13 concept groups (including but not limited to efficacy, safety, and quality of life), was built. CCTO not only captures a comprehensive taxonomic hierarchy (through is_a relationships) but also an evaluative hierarchy ( is_assessed_by) to aid the acquisition of treatment outcome data. At least one CCTO term was mentioned in 29493 of 29951 (98%) MEDLINE abstracts on phase 1-3 cancer trials; concepts about efficacy were mentioned in 7208 (79%) phase 1, 15051 (92%) phase 2, and 3884 (86%) phase 3 trials. We show that, by applying the relationship rules in an expert system shell, the event sequence of the index case can be readily converted into a comprehensive profile that incorporates survival measures, treatment responses, and adverse events. Conclusions: CCTO efficiently captures and categorizes high-level concepts of treatment endpoints in oncologic practice and enables rapid profiling of clinical outcomes for cancer patients.