Mapping Adverse Drug Reactions in Canadian Vigilance Adverse Reaction Database to various standardized terminologies and determining an ideal terminology to be...

Abstract

Canada Vigilance Program is Health Canada’s post-market surveillance program that gathers and evaluates reports of suspected adverse drug reactions to various health products. This facilitates Health Canada to monitor safety profiles of these health products once they are introduced in to the market. This helps to keep track of the products that they continue to outweigh the risk associated with them. This program is supported by seven Canada Vigilance Regional Offices [1] who provide a regional point of contact for health professionals and consumers. All the collected reports and their data are then stored in the Canada Vigilance Adverse Reaction Database. The database has a total of 12 tables consisting of information regarding Drug Products, Drug product ingredients, Indications for which they are used and reactions reported for them and their severity. Adverse reactions in the database are mostly coded using MedDRA (Medical Dictionary of Regulatory Activities) which has standardized terms for symptoms, signs, diseases, syndromes and diagnoses. For the purpose of the project, only the data related to Crohn’s Disease was analyzed. One of the focus of the project was to map these adverse reactions and drug products using BioPortal REST API to various terminologies like MedDRA, SNOMED CT, MeSH, UMLS and RXNorm (mapped to respective codes, their parents and hierarchy levels) and store this data in the database using python programming language. Other focus of the project after mapping was to evaluate the results of the mappings performed and determine which is the ideal terminology which can be used as an adjunct to MedDRA. The reason for finding a terminology as a supplement to MedDRA is that MedDRA has a shallow hierarchical structure with only 4 levels. Due to lack of details in these levels of hierarchy it restricts the scope and lessens the chances of getting more fine-grained information. The results of these evaluations showed that SNOMED CT gave the best results when used along with MedDRA. SNOMED CT has a more broader scope and very detailed hierarchal structure (14 levels) which classifies the term more finely. When a particular term was searched in MedDRA the results received were based on text matching rather than concept matching. But when the same was searched using SNOMED CT hierarchal structure additional terms were returned based on related terms of the original term. This finding can be very valuable in the process of signal detection of identifying drug-patterns in Pharmacovigilance as it will enhance the matching of the terms and giving more accurate results.