Elizabeth SmalleyDirector of Product Management, Data & AnalyticsArisGlobal
In a 2022 life sciences industry report,1 70% of biopharmaceutical professionals indicated a maturity gap in the efficiency of safety signal analytics processing. Transforming signal detection, as well as signal validation, in ever smarter ways promises to close that gap, boosting pharmacovigilance (PV).
And not before time. Safety signal monitoring has barely changed in 70 years, from when the practice was first introduced in response to the international thalidomide tragedy - after GPs began to observe an increase in babies being born with disabilities including the shortening and absence of limbs, and a correlation was discovered with their mothers having taken thalidomide during pregnancy.2 Systematic monitoring of adverse drug reactions was one of the formal regulatory responses to the tragedy which affected more than 10,000 babies around the world.
Since then, other than adapting paper-based manual reporting with some form of electronic equivalent, very little has progressed in the discipline of signal detection and analysis. Outside of the clinical environment, with its inherent limitations (as well as being highly controlled, trials’ safety monitoring cannot touch the vast variety of human genomes that a drug will encounter post-market), adverse event monitoring relies too heavily on patient and clinician reporting. And, still today, too many adverse events go unreported (up to 95% in the worst cases).3
Reducing Delay and Signal ‘Noise’
Even once submitted, Individual Case Study Reports (ICSRs) take time to process before they are used in signal detection. Analyses of medical literature, another PV channel, also inevitably involve a time delay, while scouring of online forums yields too much noise. With smart, sophisticated analytics technology that can filter for causal and sensitivity to substantially reduce signal ‘noise’ (with 40%+ more accuracy than traditional signal detection methods, as demonstrated in extensive studies conducted by ArisGlobal),4 professionals will be able to distil precise adverse event insights directly using robust real-world data from electronic medical records and healthcare claims, boosting drug safety and driving new efficiency for developers. Access to this data is being democratized, thanks to strategic partnerships designed to empower drug developers and their Safety teams with timely, reliable feeds combined with smart, intuitive analytics technology.
Moving signal detection closer to the patient will help address gaps and lag time in adverse event reporting, reducing marketing authorization holder (MAH) risk. Indeed, the benefits will be widely felt right across the healthcare ecosystem – by patients, drug development companies, regulators, and clinicians.
Proactive, hypothesis-free signal detection along with improved signal strength is shown to reduce false positives and detection signals earlier. The incorporation of real-world data, meanwhile, means signals are detected even faster and with impressive precision - the equivalent of a thermometer quantifying the progression of an illness, or a financial credit score objectively assessing an individual’s economic health/risk, enabling robust new protocols and better overall outcomes.
These developments couldn’t have come at a better time. Up to now, an element of industry inertia has curbed proactive innovation in safety signal detection; typically, compliance has been the primary driver of the measures implemented.
The recent pandemic prompted swifter and more continuous vigilance as advanced vaccines were rolled out with speed, and across huge swaths of the global population. Regulators meanwhile have led the way in breakthrough innovations for signal detection and validation. Examples include EMA’s adoption of DARWIN EU, a platform to generate real-world evidence (RWE) to support the decision-making of EMA scientific committees and national competent authorities in EU Member States throughout regulatory processes. Sentinel, FDA’s national electronic system, meanwhile, is transforming the way researchers monitor the safety of FDA-regulated medical products (here real-world data is used for validating signals, though not for detection).
Large Language Models Pave the Way for AI
To keep pace with accelerating change and new waves of innovation in Life Sciences, drug developers have little choice but to ramp up their signal detection and analysis capabilities now. At a practical level, a combination of technology advances and more readily available real-world data is paving the way for much more robust and responsive safety vigilance.
At an artificial intelligence (AI) level, ‘large language models’ (next-generation neural networks) are transforming the precision with which Safety teams can distil insights from vast data sets, quickly learning and progressively honing their knowledge of what to look out for and what to discount.
The technology is so intuitive to use that Safety teams have less need for the intervention of epidemiologists or data retrieval experts, now being able to perform a deeper level of causal analysis themselves. Large language models (LLMs) are priming the pharma industry to easily embrace all kinds of AI, something that was not true even three years ago. As a result, we can expect to see extensive adoption of proactive signal detection in record time.
Provided there is an appropriate interface, and that the right data preparations have been made so that Safety teams cannot be misled by the findings, Safety professionals can perform their own investigations on the fly, in a highly repeatable way. This is ultimately much more efficient and responsive than requesting a one-off, hypothesis-based study which, as well as being labor-intensive, requires that the query parameters are known upfront.
Beyond Speed and Accuracy
The benefits of proactive signal detection, via AI-sharpened analysis of extensive and robust real-world data, in conjunction with ICSRs, are broader than simply faster speed and greater accuracy.
Firstly, as correlations are detected earlier and with improved precision, drug developers will be in a position to spend more time on higher-value activities including innovation in drug discovery, and on delivering safer drugs to patients, sooner.
Secondly, safety-based communications will become much more targeted. Instead of stating generically that a drug may increase the risk of heart attack, the advice can specify that this risk applies specifically to women between the ages of 30 and 60 who have a pre-existing heart condition, for instance - in the context of a very specific phenotype, in other words.
Developing the Commercial Potential of a Drug
Beyond safety and compliance applications, the same mathematical modelling used in adverse event monitoring also supports signal detection in drug repurposing, potentially presenting new commercial opportunities to drug developers as previously unknown and unexpected positive correlations are discovered. (A signal is any previously unknown information about the causality of a drug and event; it needn’t be a negative outcome.) In the context of a benefit-risk profile, this is an opportunity to focus as much on the benefit as on the risk profile, and to grow the commercial potential of a drug. Safety has an unprecedented opportunity to shine as a strategic partner to the business, rather than merely a cost centre that exists to contain risk.
Next-generation signal detection is emerging as an exciting field to watch and leading pharma companies are already exploring the potential to boost patient health and safety and focus more on drug repurposing.
References
1. What if you could do more with less? Learn how AI and automation can drive business change. Current State of Automation and AI Adoption in Life Sciences: https://www. arisglobal.com/industryreport/
2. How the thalidomide scandal led to safer drugs, Medical News Today, December 2020: https://www.medicalnewstoday.com/articles/how-the-thalidomide-scandal-led-to- safer-drugs#Key-regulatory-reforms
3. Pharmacovigilance: reporting requirements throughout a product’s lifecycle, NIH, September 2022: https://pubmed.ncbi.nlm.nih.gov/36187302/
4. ArisGlobal Proactive Signal Detection foundational study/webinar: https://goto.webcasts. com/starthere.jsp?ei=1632818&tp_key=1a9c698bce
Author Biography
Elizabeth Smalley is the Director of Product Management, Data & Analytics at ArisGlobal. With over 15 years of experience across data analytics, Elizabeth has a passion for exploring the intersection of data science and human reasoning, with experience bringing AI-powered software to market to drive safety and clinical outcomes for patients and clinicians. In her current role, she leads the teams managing the Data Platform, LifeSphere Clarity and LifeSphere Signals, and Risk Management product lines. www.arisglobal.com, [email protected]
Publication Detail
This article appeared in American Pharmaceutical Review:
Vol. 26, No. 8
Nov/Dec 2023
Pages: 58-59
Subscribe to our e-Newsletters
Stay up to date with the latest news, articles, and events. Plus, get special
offers from American Pharmaceutical Review delivered to your inbox!
Sign up now!