Generative AI is accelerating transformation across the life sciences, and pharmaceutical companies and clinical research organizations face the dual challenge of modernizing pharmacovigilance while maintaining uncompromising standards for patient safety, data integrity, and global regulatory compliance.
This delicate balance is paramount in the detection and notification of adverse events, where a single missed risk can lead to a public health crisis, loss of confidence in a treatment, and regulatory action. OECD data shows that avoidable adverse drug events alone cost $54 billion annually.
A 2023 McKinsey Global Institute report estimates that generative AI could generate $60 billion to $110 billion (approximately 2.6-4.5% of revenue) annually across the pharmaceutical and medical products industry, driven primarily by productivity gains in research and development and knowledge work. The report further notes that generative AI and related technologies have the technological potential to automate activities that account for 60-70% of employee work time across the economy.
For global organizations managing safety data at population scale, the opportunities go far beyond efficiency. All automated decisions must be traceable and consistent with patient outcomes.
Emerj Editorial Director Matthew DeMello sat down with IQVIA's Marie Flanagan on the “AI in Business” podcast to continue the conversation about enhancing drug safety through AI and automation technologies and how these technologies are redefining pharmacovigilance.
In the next article, I'll focus on three key takeaways from the conversation.
- Use social media data to drive proactive safety: Leverage social media data to develop predictive pharmacovigilance models that predict adverse events and enable early intervention.
- Automating drug safety transformation: Combine optical character recognition (OCR), robotic process automation (RPA), and natural language processing (NLP) to identify adverse events faster and more accurately across massive unstructured safety data sets.
- Collaborating with regulators in AI-powered pharmacovigilance: Build a collaborative oversight model between life sciences businesses and regulators to align digital innovation and patient safety.
Listen to the full episode below.
guest: Marie Flanagan, Director of Product Management, Digital Projects and Solutions, IQVIA
Expertise: Drug safety, life cycle safety, regulatory reporting
recognition: Marie has held a number of positions at IQVIA, starting with Drug Safety in 2007. Over the next 20 years, Marie worked her way up through various account manager roles to director level, culminating in her current role. She also earned a bachelor's degree in microbiology and immunology from College Cook University in 2004.
Focus on preventive safety
When asked about how AI, and NLP in particular, has changed safety signal monitoring and detection from IQVIA’s perspective, Flanagan responds: Flanagan said IQVIA has been in the AI space for a long time.
Flanagan explains that the product she supports, Vigilance Detect, uses NLP trained on a proprietary bank of keywords and patterns trained in safety over the past 13 to 14 years to detect safety events in the data pipeline earlier than traditional safety workflows.
Flanagan goes on to explain how NLP is used to sift through vast amounts of unstructured data, including highly diverse and multilingual datasets, to find adverse events and product complaints. Flanagan said this is an important example of using technology to eliminate extremely tedious manual work.
She details how IQVIA works closely with regulators and influential industry players to maintain an open dialogue. Flanagan also added that the industry's regulatory environment is changing to become more patient-centric in how it deploys patient-facing technology and digital channels.
Flanagan tells listeners of the Emerj Executive Podcast that IQVIA accomplished two things by developing its patient channel.
- More information is coming in
- Gives healthcare workers who interact with patients more time to empathize, have conversations, and convey more clinically robust information to regulators.
Until recently, Flanagan said, the use of social media for patient information was ridiculed. Also, there were no safety signals or critical incidents reported on social media because the burden was too great.
Flanagan points out how that bias is changing, especially around the ability to use information in different ways. She explains how that information can be used for advance notice.
Companies like IQVIA can identify hotspots of influenza incidents in specific areas outside Manhattan and correlate that data with data from the FDA's Adverse Event Reporting System (FAERS) months later.
Flanagan argues that linking data with official resources through FAERS opens the door to the use of information by applying analytics that help companies determine what signaling activities can be used on their products before problems occur.
Integrating natural language processing and automation to transform drug safety
When asked about the opportunities and challenges of leveraging automated technologies such as OCR and RPA to detect adverse events at IQVIA, Mr. Flanagan provided important insights. She argues that if IQVIA relied solely on AI or NLP in its purest form, it would not be able to achieve as good results as it has achieved so far.
He then explained that ensuring safety is a highly complex end-to-end process, and that IQVIA benefits from a combination of automation technologies. Flanagan emphasizes that if the company used NLP as a stand-alone approach to identifying adverse events, it might only see 20-30% positive results.
“Applying OCR, RPA, and various other automated techniques to capture data and move it in a usable way can yield up to 70-80% positive results when searching for adverse events on social media.
In short, our best projects with the most favorable results combined RPA, OCR, traditional coding, and traditional machine learning with more advanced AI techniques. That's when we see the magic happen, and that's the artistry of what we do in the safety field, especially finding adverse safety events from unstructured data. ”
– Marie Flanagan, Director of Product Management, IQVIA Digital Projects and Solutions
Regulatory cooperation in AI-enabled pharmacovigilance
When asked where he sees the dynamics of regulatory bodies such as the EMA and FDA emphasizing digitization in real-world evidence for pharmacovigilance, Mr. Flanagan provided useful insight. This conversation highlighted a major shift in how regulators interact with the life sciences industry.
For decades, regulation has been seen as a necessary safeguard, but it has traditionally slowed innovation to protect patient safety. However, Flanagan explains that long-standing regulatory dynamics are changing, as regulators increasingly encourage digital transformation in safety operations and actively participate in the modernization of pharmacovigilance.
She tells Emerj's podcast listeners that she previously had to turn off comments because IVQIA simply didn't know how to handle them on its social media channels. However, it recently published comments on social media, allowing patients to communicate directly with regulators. As a result, regulators are actively sharing that information with industry and building more collaborative relationships.
But Frangan was cautious in describing the current dynamics, telling Emerge podcast listeners: “It's not harmonized in that there's no harmonization of regulations across the major regions.” “But we're seeing them working together, and we're seeing them setting very broad standards and guidelines in our industry regarding the use of AI.”
It added that regulators have not given IQVIA or other companies detailed information about their expectations. They essentially shift that responsibility onto the company, Flanagan said. She explains how she keeps things broad enough so that regulators don't stifle the progress of her company's AI implementation.
She provides life sciences leaders with highly insightful and definitive advice that illuminates how their organizations can effectively operate in a more collaborative, digitally enabled regulatory environment. She emphasizes the need to assess and optimize workflows before pursuing AI, focusing on high-value, practical digital enhancements, choosing the right use cases, and avoiding overengineering by starting small.
