Aggregation and analysis of indication-symptom relationships for drugs approved in the USA
Abstract Purpose
Drug indications and disease symptoms often confound adverse event reports in real-world datasets, including electronic health records and reports in the FDA Adverse Event Reporting System (FAERS). A thorough, standardized set of indications and symptoms is needed to identify these confounders in such datasets for drug research and safety assessment. The aim of this study is to create a comprehensive list of drug-indication associations and disease-symptom associations using multiple resources, including existing databases and natural language processing.
MethodsDrug indications for drugs approved in the USA were extracted from two databases, RxNorm and Side Effect Resource (SIDER). Symptoms for these indications were extracted from MedlinePlus and using natural language processing from PubMed abstracts.
ResultsA total of 1361 unique drugs, 1656 unique indications, and 2201 unique symptoms were extracted from a wide variety of MedDRA System Organ Classes. Text-mining precision was maximized at 0.65 by examining Term Frequency Inverse Document Frequency (TF-IDF) scores of the disease-symptom associations.
ConclusionThe drug-indication associations and disease-symptom associations collected in this study may be useful in identifying confounders in other datasets, such as safety reports. With further refinement and additional drugs, indications, and symptoms, this dataset may become a quality resource for disease symptoms.
- Sponsoring Organization:
- USDOE
- OSTI ID:
- 1631784
- Journal Information:
- European Journal of Clinical Pharmacology, Journal Name: European Journal of Clinical Pharmacology Journal Issue: 9 Vol. 76; ISSN 0031-6970
- Publisher:
- Springer Science + Business MediaCopyright Statement
- Country of Publication:
- Germany
- Language:
- English
Similar Records
Informing selection of drugs for COVID-19 treatment through adverse events analysis
Structured Prompt Interrogation and Recursive Extraction of Semantics (SPIRES): a method for populating knowledge bases using zero-shot learning