Advancing Drug Utilisation and Drug Safety Research Through Interdisciplinary Methods
Open access
Author
Date
2023Type
- Doctoral Thesis
ETH Bibliography
yes
Altmetrics
Abstract
Randomized clinical trials (RCTs) are fundamental for evaluating the safety and efficacy during the development of new drugs. However, RCTs are often limited in their generalisability and validity due to low diversity and number of participants, and duration of exposure. Pharmacoepidemiologic research allows the evaluation of the use and effects of drugs in a more extensive and diverse patient population, reflecting real-world conditions. Moreover, the increasing ability to electronically capture real-world data (RWD) related to patient health status and transform it into new knowledge, so-called real-world evidence (RWE), has opened new opportunities for pharmacoepidemiologists to close gaps between research and clinical care. These electronic healthcare databases can be used to study the utilisation, safety, and effectiveness of drugs at population level. The use of RWD in pharmacoepidemiologic research also poses some challenges. For example, while studies on drug utilisation provide an overview of patterns in use and adverse events, capturing complex drug utilisation patterns in large databases may be a difficult task when thousands of drugs are approved and used in routine practice. Thus, there is a need to combine multidisciplinary approaches to uncover valuable information on prescription patterns using large patient datasets. Another central challenge in pharmacoepidemiology is the investigation of adverse events with unknown mechanisms of action. Pharmacoepidemiologic studies alone are limited in the ability to study only known drug effects based on the known pharmacokinetic and pharmacodynamics principles, and thus, are limited to infer conclusions when adverse events may be due to unknown mechanisms of action. Thus, in this dissertation, we focus on leveraging methodologies from pharmacoepidemiology, data science, and medicinal chemistry to address unanswered questions in the study of the utilisation and safety of drugs used to treat chronic conditions.
In the first part of this dissertation (Chapter 3), we proposed a novel application of the Apriori algorithm, an established data mining algorithm, to overcome the inherent computational challenges of assessing real-world prescription patterns at the compound level using RWD (Chapter 3.1). The analysis revealed a high prevalence of polypharmacy (i.e., use of ≥5 concomitant medications) in patients with diabetes. Additionally, we evaluated an extensive array of individual drugs and drug combinations that are frequently prescribed, surpassing the existing knowledge on polypharmacy patterns based on drug classes. Subsequently, we shifted the focus from identifying harmful combinations of drugs to identifying opportunities for optimising pharmacotherapy in patients starting their first oral antidiabetic medication by estimating the prevalence of potentially inappropriate prescriptions (PIPs; [Chapter 3.2]). The analysis indicated that the prevalence of PIPs was higher in patients receiving polypharmacy and in older patients. Thus, starting treatment with oral antidiabetic drugs, particularly in those patients receiving polypharmacy, should involve a comprehensive review of the medications to optimise prescribing decisions.
Moving forward, in the second part of this dissertation (Chapter 4), we combined interdisciplinary approaches to investigate the safety profile of Janus Kinase inhibitors (JAKis) following rising safety concerns leading to black-box warnings, despite unknown mechanisms. First, we leveraged computational and experimental approaches to investigate whether the safety concerns on thrombosis and viral infection associated with tofacitinib and baricitinib was potentially due to off-target effects (Chapter 4.1). Although our findings did not confirm the hypothesis of elevated risk of thrombosis and viral infection explained by drug-target interactions, previously unknown off-targets of baricitinib and tofacitinib were identified, suggesting them as potential candidates for drug repurposing. Subsequently, we proposed the novel prevalent new-user cohort study to investigate the incidence and risk of major adverse cardiovascular events (MACE), venous thromboembolism (VTE), and viral infection/reactivation associated with JAKi use in the Danish rheumatologic database (DANBIO) (Chapter 4.2). The use of an advanced study design and the DANBIO aims to overcome challenges posed by confounding biases in observational studies, and to provide valuable safety knowledge on the treatment of patients with rheumatoid arthritis (RA) using JAKis.
Finally, we extended our investigation on the off-target profile of tofacitinib and baricitinib with a focus on repurposing within Alzheimer’s disease (AD; [Chapter 5]). The combined approach of a machine learning-based tool and experimental validation allowed the identification and characterisation of previously unknown baricitinib off-targets potentially relevant for AD progression. Although baricitinib is unlikely to be successfully repurposed for AD, the findings contribute to our understanding of JAKis off-target effects and their potential implications. Overall, this dissertation demonstrated the ability of pharmacoepidemiology to collaborate with other disciplines and leverage different methodologies. By integrating diverse approaches, it becomes possible to identify new risks and benefits, explore off-target effects, and potentially uncover new indications for approved drugs. This interdisciplinary collaboration has the potential to impact early stages of drug development and improve prescribing decisions, highlighting the benefits of fusing alternative approaches to tackle common goals. Show more
Permanent link
https://doi.org/10.3929/ethz-b-000644914Publication status
publishedExternal links
Search print copy at ETH Library
Contributors
Examiner: Burden, Andrea M.
Examiner: Schneider, Gisbert
Examiner: Puhan, Milo
Examiner: Grisoni, Francesca
Publisher
ETH ZurichSubject
Pharmacoepidemiology; Drug Safety; Machine Learning; Drug utilisation; Real-world data; Off-targetsOrganisational unit
09633 - Burden, Andrea / Burden, Andrea
Funding
ETH-32 18-2 - Machine Learning to Inform the Study of Adverse Drug Events in Patients with Type 2 Diabetes (ETHZ)
Related publications and datasets
Is identical to: https://doi.org/10.3929/ethz-b-000592519
Is identical to: https://doi.org/10.3929/ethz-b-000549138
Is identical to: https://doi.org/10.3929/ethz-b-000626569
More
Show all metadata
ETH Bibliography
yes
Altmetrics