Background
Whenever a patient takes two or more medications, there is a risk for drug-drug interactions (DDI). A DDI can have a pharmacokinetic basis, i.e. plasma concentrations of one drug are influenced by another drug resulting in increased or diminished pharmacological action; or a pharmacodynamic basis, i.e. one drug influences the safety or efficacy of another drug. In both cases, when relevant, a warning (signal) should occur with a recommendation to the physician or pharmacist to take appropriate action, e.g. a dose modification, additional monitoring of the patient or selecting an alternative drug.
The basis for DDI recommendations lies in the product information (label) developed by the manufacturer and approved by regulatory authorities. This information, however, covers only the most relevant co-medications that can be expected in the specific patient population, and is not intended to be complete. As a result, scientists and committees who develop DDI databases have to make extrapolations for potential DDIs with co-medications not mentioned in the product information.
Several surveys have demonstrated that DDI recommendations in various DDI databases are variable, incomplete and inconsistent. The consequence of this variability is that a DDI signals and its associated management very much depend on the system a health care professional is using, with a high chance of missed DDIs or incorrect DDI signals, and patient harm as the outcome.
Another problem is that the development of a complete and up-to-date DDI database is a time-consuming and inefficient process, where millions of healthcare providers have to assess the clinical relevance of a DDI on a daily basis. As noted above, the product of all these hours of work is still highly variable and far from perfect.
Approach
This project aims to develop an AI-driven DDI Manager tool that is correct, complete, and continuously updated by automating and standardizing the extrapolation of drug-drug interactions for new drugs based on existing knowledge.
The project will focus on:
- Extrapolating the expected effects, actions, and mechanisms of coadministration for approximately 1,000 drugs from the database
- Combining data to fill the database for DDI management
- Developing procedures for regular database updates
- Developing procedures for assessing the correctness of DDI recommendations
Data
The dataset originates from the Comed Core Database and associated Excel templates used for DDI management.
For each primary drug, the dataset contains its DDI profile, mechanism of interaction, and management recommendations. In addition, approximately 1,000 co-medications per drug are documented for interaction assessment. The full database includes hundreds of primary drugs, resulting in thousands of interaction records.
References
Burger DM, le Comte M, Smolders EJ, Jacobs TG, Ter Heine R, Knibbe CAJ, et al. What the Product Label Does Not Tell You About Drug-Drug Interaction Management: Time for a Re-Appraisal. J Clin Pharmacol. 2023;63(11):1181-5.
Kontsioti E, Maskell S, Bensalem A, Dutta B, Pirmohamed M. Similarity and consistency assessment of three major online drug-drug interaction resources. Br J Clin Pharmacol. 2022 Sep;88(9):4067-4079.
Requirements
- A (bio)medical or (bio)pharmaceutical background
- Interest in the use AI for advancing healthcare
- Preferably proven experience with technologies and projects related to this interest
Information
Project duration: 6 months
Location: Radboud University Medical Center
The student will be embedded in the department of Pharmacy, Pharmacology and Toxicology. We are with a large group of researchers where also a lot of student conduct their research.
If you are interested in applying for this Master student project, please send an email to: rtcai@radboudumc.nl
