(128) Agreement Between Interaction Alerts in Different Sources Regarding Classification and Suggested Clinical Management

Sunday, August 28, 2022

8:00 AM – 1:45 PM CEST

Location: Hall C

Publication Number: 1173

Presenting Author(s)

Carina Tukukino
– Midwife, University of Gothenburg, University of Gothenburg, Gothenburg, V�stra G�talands l�n, Sweden

Co-Author(s)

Walaa Hamdan
– Pharmacist student, MSc Pharm, Department of Pharmacology, Sahlgrenska Academy University of Gothenburg, Ludvika, Dalarnas Lan, Sweden
Frida Wallin Hybelius
– Pharmacist student, MSc Pharm, Department of Pharmacology, Sahlgrenska Academy University of Gothenburg, Sweden
Anna L. Eriksson
– PhD/MD, Department of Internal Medicine, Sahlgrenska Academy, University of Gothenburg, Gothenburg, Sweden, Vastra Gotaland, Sweden
Susanna M. Wallerstedt
– University of Gothenburg, Department of Pharmacology, Sahlgrenska Academy, University of Gothenburg, G�teborg, V�stra G�talands l�n, Sweden

Background: Interaction alerts are used to guide clinicians’ decision making and to describe drug interactions in populations, but the evidence regarding agreement between sources is sparse.

Objectives: To analyse the agreement, regarding classification and clinical management, between Janusmed (JM) drug interaction alerts and two other alert sources: Lexicomp (LC) and the Micromedex (MM).

Methods: All drug pairs that triggered an interaction alert in JM (n=200), in 274 consecutive primary care patients being treated with ≥2 drugs (65–99 years, 59% female), were included. We compared JM alerts with LC and MM alerts regarding the classification of clinical relevance and recommendations for clinical management, with focus on the JM categories D, indicating that the drug combination should be avoided, and C, indicating that the interaction can be managed by dose adjustments. JM D and C alerts were assumed to correspond to LC / MM categories X / Contraindicated and C/D / Moderate/Major, respectively. Agreement was assessed using kappa statistics. Recommendations provided to manage the interactions were categorised and discussed in consensus.

Results: A total of 170 (85%) drug pairs also elicited an alert in LC, and 139 (70%) drug pairs elicited an alert in MM. In JM, 9 (5%) out of 200 drug pairs were categorised as D, and 100 (50%) as C. The most frequently occurring substances in C/D alerts were diclofenac (n=12), warfarin (n=10), citalopram (n=9), levothyroxine (n=8), and spironolactone (n=8). Of the 109 C/D alerts in JM, none were categorised as X in LC, and 89 (82%) as C or D. In MM, none of the 109 JM C/D alerts were categorised as Contraindicated, and 75 (69%) as Moderate/Major. The agreement was minimal between JM and LC (kappa: 0.30) as well as between JM and MM (kappa: 0.22). For the 109 JM C/D alerts, a total of 246 recommendations for clinical management were provided, most frequently to monitor laboratory values (n=59, 24%), to monitor the clinical condition (n=56, 23%), and to switch to another drug (n=51, 21%). Out of 109 C/D alerts in JM, 16 (15%) drug pairs in LC, and 9 (8%) in MM, elicited the same recommendations, according to our categorisations.

Conclusions: Drug interaction databases, eliciting alerts for potentially harmful drug combinations, differ regarding classification and clinical management. This needs to be considered when used to guide clinicians and in research where interaction alerts are described.