DDReg Pharma


Automation in Pharmacovigilance and Drug Safety

The integration of artificial intelligence (AI) and machine learning (ML) technology into the pharmaceutical and life sciences industry is demonstrating significant benefits on several fronts. Whether it is incorporation of real-world data into clinical investigations or automation of operations to streamline key processes, technology is driving the industry forward. Pharmacovigilance is part of the industry that is increasingly adopting AI and ML technology to enhance drug safety- be it in the detection of adverse drug events (ADEs), processing of safety reports, or identifying drug-drug interactions for better patient safety.

Detecting adverse drug events

Data on ADEs that have been reported by patient themselves is a key source of safety information for pharmacovigilance and post-market safety surveillance. In an era of ‘tweets’ and 140-character limit posts, to-the-point and direct safety information is reported on platforms such as Twitter straight from the patient. The use of ML in directly extracting ADE data from these sources could be beneficial. A study conducted by Alvaro and colleagues utilized various ML models to identify keywords related to selective serotonin reuptake inhibitors and cognitive enhancers from 1548 tweets to extract ADE data related to these treatment options. This is advantageous as it captures information on ADEs that may not have been collected by medical professionals and demonstrates how successful ML models are in sifting through large volumes of data for more precise results [1].
A comprehensive review by Basile and colleagues highlighted how certain ML and AI technology were applied to post-marketing surveillance activities. Systemic pharmacology qualitatively analyses the relationship and/or interaction between a biological system and drugs. It is a data-rich approach for adverse drug reaction (ADR) mining as several databases and information sources are easily available as ‘big data’. The review also identified other studies in which various ML models were trained to predict new pharmaceuticals that cause ADEs [2]. Another investigation by Menard and colleagues utilized data from over 100 clinical trials that included key information on patients, diseases areas, and vitals etc to train a ML model for predicting the number and frequency of ADEs [3].

Processing safety reports

Natural Language Processing (NLP) is a form of ML technology that can classify unstructured data and patient safety reports to generate reports which would determine the severity of outcomes. A study by Evans and colleagues found they were able to apply NLP to identify cases that led to significant harm or even death. However this does not entirely replace manual review as expertise is required when evaluating the technical information of medical text [4]. Furthermore, ML algorithms were used for screening the safety reports of patients within electronic records; they analyzed large & unstructured ADE datasets that were collected via passive surveillance methods [5]. ML models can identify key terms in patient safety reports which would indicate and highlight areas in attention is required for post-market safety surveillance.
Various studies have investigated ways in which ML models can be programmed to determine the drug-drug interactions. For example, researchers generated an ML model based on lab tests and data on treatment which could help identify patients that exhibited an ADE associated with a drug-drug interaction [6].


Indeed, the use of AI and ML technology in Pharmaceuticals and Life sciences has come a long way. Be it in regulatory affairs, medico-regulatory writing, or pharmacovigilance- the pharmaceutical sector is experiencing exponential technological growth to enhance operations and processes. Particularly in pharmacovigilance and drug safety, exploiting advanced technology is resulting in better & more accurate real-world data that provides a better understanding of a drug’s safety profile, to ensure patient safety and minimize risk of ADEs.

References and Further Reading:

[1] Alvaro N, Conway M, Doan S, Lofi C, Overington J, Collier N. Crowdsourcing Twitter annotations to identify first-hand experiences of prescription drug use. Journal of biomedical informatics. 2015 Dec 1;58:280-7.

[2] Basile AO, Yahi A, Tatonetti NP. Artificial intelligence for drug toxicity and safety. Trends in pharmacological sciences. 2019 Sep 1;40(9):624-35.

[3] Ménard T, Barmaz Y, Koneswarakantha B, Bowling R, Popko L. Enabling data-driven clinical quality assurance: predicting adverse event reporting in clinical trials using machine learning. Drug safety. 2019 Sep;42(9):1045-53.

[4] Evans HP, Anastasiou A, Edwards A, Hibbert P, Makeham M, Luz S, Sheikh A, Donaldson L, Carson-Stevens A. Automated classification of primary care patient safety incident report content and severity using supervised machine learning (ML) approaches. Health informatics journal. 2020 Dec;26(4):3123-39.

[5] Marella WM, Sparnon E, Finley E. Screening electronic health record–related patient safety reports using machine learning. Journal of Patient Safety. 2017 Mar 1;13(1):31-6.

[6] Bouzillé G, Morival C, Westerlynck R, Lemordant P, Chazard E, Lecorre P, Busnel Y, Cuggia M. An Automated Detection System of Drug-Drug Interactions from Electronic Patient Records Using Big Data Analytics. InMedInfo 2019 Aug 21 (pp. 45-49).