2023 FDA Science Forum
Comparison of Linguamatics and FDALabel Natural Language Processing Text-Mining to Identify Information in the OVERDOSAGE Section of Tramadol Drug Labels
- Authors:
- Center:
-
Contributing OfficeCenter for Drug Evaluation and Research
Abstract
Background: The OVERDOSAGE section of prescription drug labels require analysis for outdated information that may be inaccurate or misleading. Prescription drug labels containing the same active ingredient in the same formulation should be identical. Natural language processing (NLP) text-mining can be used to efficiently query labeling to identify differences. Purpose: Perform a comparison of results extracted through text-mining, using Linguamatics and FDALabel, to search DailyMed for targeted information in the OVERDOSAGE section of tramadol drug labeling. Identify differences among labeling versions. Methodology: The OVERDOSAGE sections of tramadol drug labeling was extracted using a custom query in Linguamatics, an NLP text-mining tool. Using the unique ingredient identifier (UNII) code, the query was able to identify drugs with the same active ingredient. The results were then compared to a search of the OVERDOSAGE sections of tramadol drug labeling performed on FDALabel. Labeling was then manually analyzed for differences. Results: The query in Linguamatics, retrieved 237 drug labeling, under 30 drug application numbers; seven (23%) drugs are New Drug Application(s) (NDA); 23 (77%) are generic drugs approved under an Abbreviated New Drug Application(s) (ANDA). Results retrieved using the FDALabel platform are identical. Depending on the link re-packagers use for labeling, a single drug application number may appear multiple times with different labeling versions (ex. ANDA201384), others may appear with the same labeling (ex. ANDA200503). Four different versions of the OVERDOSAGE section were identified for Tramadol NDAs. Clinical manifestations were similar across versions. However, two (50%) versions did not mention seizures, one (25%) did not mention QT prolongation, and only three (75%) discussed increased risk of fatal overdose with co-ingestants. Management was also similar. However, one (25%) version mentioned the use of nalmefene, three (75%) discussed monitoring patients for spontaneous respiration, potential need of additional administration of antagonist, and management of opioid dependent individuals. Conclusions: Linguamatics and FDALabel natural language processing text-mining efficiently extracted identical information of the OVERDOSAGE sections of tramadol drug labeling. Depending on the product selected, a search on DailyMed may not provide the most up to date labeling. This is due to re-packagers not linking the most recent labeling version.
