Aristotle University Of Thessaloniki (AUTH)
Τεχνολογίες δεδομένων μεγάλης κλίμακας και ανίχνευση ανεπιθύμητων ενεργειών φαρμάκων - φαρμακοεπαγρύπνηση
Abstract
dc:descriptionIntroduction: Post marketing drug safety surveillance mainly relies on data retrieved from spontaneous report systems and observational databases. However, due to some important limitations of these data sources, interest in exploring the use of new methods of surveillance to detect potential drug safety signals is growing. Social media and search analytics domains they are a useful source of information. In the recent years, machine learning methods developed to detect proto adverse events on Twitter posts as well as in general posts in forums or Facebook. However, there are still large unexploitable areas of data, mainly self-generated form users and patients. For example, in 2009 Google succeeded on predicting the H1N1 virus spread by analyzing phrases and terms used by users in its search engine.This study sought to detect a potential safety signal of abuse on 3 different drugs by combining different sources of surveillance, specifically Twitter posts, Google search analytics and the FDA Adverse Event Reporting System database. We introduce a new domain of data source, the search analytics domain, as well as new methods to reveal hidden patterns and correlations with existing adverse event report data sources. Methods: We collected data from sources regarding adverse events of 3 drugs (mirtazapine, pregabalin and quetiapine), from Q12004 to Q32017 and more specifically abuse related events, using eight terms from MedDRA. We used metrics, such as the search popularity score, the search interest over time and search frequencies to measure searches related to abuse. Using the same MeDRA terms we collected and analysed reports from FAERS, for the same period. We also retrieved user posts from Twitter and using an advanced Naïve-Bayes algorithm we detected references of adverse events. Results were analysed using statistics and compared to those of the official adverse event sources or clinical trials. Results: We found strong evidences of abuse liability for the selected drugs in the search analytics domain. Using two different methods strengthen the evidence toward a potential signal of mirtazapine, pregabalin and quetiapine abuse, confirming the findings from post marketing reports. Moreover, the comparison between the two domains, FAERS and the search analytics field, revealed a positive correlation between abuse-related searches and abuse-related reports in FDA Adverse Event Reporting System. Further signal validation requires clinical assessment in order to confirm causality and is a step in the signal management process performed by the regulatory authorities. Conclusion: Based on the positive correlation found between search analytics and incidents reported to the FAERS in this study, we suggest that similar methods could be invoked as supplementary pharmacovigilance tools to detect drug safety signals.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2020
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Spachos, Dimitrios
- Σπάχος, Δημήτριος
Subjects
dc:subject × 14- Δεδομένα μεγάλης κλίμακας,
- Ανεπιθύμητες ενέργειες
- Φαρμακοεπαγρύπνηση
- Big data
- Adverse events
- Pharmacovigilance
- Φυσικές Επιστήμες
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Ιατρική και Επιστήμες Υγείας
- Άλλες Ιατρικές Επιστήμες
- Natural Sciences
- Computer and Information Sciences
- Medical and Health Sciences
- Other Medical Sciences
Rights
dc:rights- Statement dc:rights
-
- BY
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/47500