Exploitation of semantic methods to cluster pharmacovigilance terms

Area of Science:

• Pharmacovigilance and drug safety research.
• Computational linguistics and natural language processing.
• Medical informatics and terminology management.

Background:

• Pharmacovigilance involves collecting and analyzing adverse drug reactions (ADRs) using controlled terminologies like MedDRA.
• Standardized MedDRA Queries (SMQs) are manually created by experts to group MedDRA terms for specific safety topics.
• Existing SMQs may not cover all important safety topics, necessitating new methods for their creation.

Purpose of the Study:

• To propose an automatic method for assisting the creation of Standardized MedDRA Queries (SMQs).
• To leverage semantic approaches for clustering semantically close MedDRA terms.
• To enhance the systematic and efficient development of pharmacovigilance tools.

Main Methods:

• Utilized semantic distance and similarity algorithms.
• Employed terminology structuring methods.
• Applied unsupervised term clustering techniques to MedDRA terms.

Main Results:

• The proposed unsupervised methods demonstrated complementarity in assisting SMQ creation.
• The approach can generate subsets of existing SMQs.
• The method offers a systematic and less time-consuming process for SMQ development.

Conclusions:

• The automatic method effectively assists in the creation of Standardized MedDRA Queries.
• Semantic clustering of MedDRA terms provides a valuable tool for pharmacovigilance.
• This approach can help expand coverage of critical safety topics in pharmacovigilance databases.