Determining drug exposure based on medication dispensing data: a validation study of vitamin K antagonist treatment episodes against INR records

Kempers, Eva K., Visser, Chantal, Goedegebuur, Jamilla, Chen, Qingui, Søgaard, Mette, Ording, Anne Gulbech, van den Dries, Carline, Abbel, Denise, Aldridge, Sarah J., Lifford, Kate J. ORCID: https://orcid.org/0000-0002-9782-2080, Portielje, Johanneke E. A., Nierman, Melchior C., Boetes‐Draisma, Annelies, van de Leur, Sjef J. C. M., Klok, Frederikus A., Geijteman, Eric C. T., Kruip, Marieke J. H. A. and Cannegieter, Suzanne C. 2025. Determining drug exposure based on medication dispensing data: a validation study of vitamin K antagonist treatment episodes against INR records. Pharmacoepidemiology & Drug Safety 34 (6) , e70158. 10.1002/pds.70158

PDF - Published Version Available under License Creative Commons Attribution Non-commercial. Download (2MB)

Abstract

Background: In pharmaco‐epidemiological studies using vitamin K antagonist (VKA) exposure, constructing treatment episodes based on dispensed prescriptions is challenging, particularly due to the large variability in therapeutic dose. Objectives: To validate different methods of constructing VKA treatment episodes based on dispensed prescriptions, using VKA exposure based on international normalized ratio (INR) measurements as a reference. Methods: Data from five Dutch anticoagulation clinics were linked to VKA dispensing data from Statistics Netherlands. Three random samples of 10 000 VKA users between 2013 and 2019 were used to compare the construction of VKA treatment episodes based on dispensings, applying fixed or dynamic methods, against the reference of exposure based on INR measurements. A total of 60 different methods were validated by computing the percentage of INR measurements occurring outside dispensing‐based VKA treatment episodes, the ratio of VKA‐exposed person‐time based on dispensings vs. INR measurements, and the number of dispensing‐based episodes. Results: Depending on the method used to construct treatment episodes, 14.8%–42.2% of the INR measurements were not covered by a dispensing‐based episode. The VKA‐exposed person‐time ratio ranged between 0.73 and 1.13, and there was substantial variability in the number of dispensing‐based episodes. Fixed methods resulted in a lower percentage of INR measurements outside the dispensing‐based episodes, a VKA‐exposed person‐time ratio closer to 1.0, and a lower number of constructed episodes. Conclusion: Fixed methods performed better than dynamic methods when classifying VKA exposure based on dispensing data. Our findings may guide other researchers working with VKA dispensing data, especially when the tablets dispensed or the prescribed dose are unavailable.

Item Type:	Article
Date Type:	Publication
Status:	Published
Schools:	Schools > Medicine Research Institutes & Centres > Prime Centre Wales (PRIME)
Additional Information:	License information from Publisher: LICENSE 1: URL: http://creativecommons.org/licenses/by-nc/4.0/
Publisher:	Wiley
ISSN:	1053-8569
Date of First Compliant Deposit:	21 May 2025
Date of Acceptance:	26 April 2025
Last Modified:	21 May 2025 09:15
URI:	https://orca.cardiff.ac.uk/id/eprint/178397

Actions (repository staff only)

Edit Item