Augmenting Drug Safety and Pharmacovigilance Services with Artificial Intelligence (AI)

Pharmacovigilance is a broad term that describes the collection, monitoring, analysis, and prevention of adverse effects in drugs and therapies. It is an entirely scientific and process-driven area within the pharmaceutical industry. The principal goal of pharmacovigilance is to determine the safer usage of drugs by focusing on setting up signal detection systems and using advanced data analytics to proactively monitor the introduction of new medicines to large patient populations. Today, pharmacovigilance systems face increasing pressure to reshape due to complex global regulations, increasing data volumes, multiple data sources, monitoring risks more broadly, and accurately recording patient events globally. Using the right analytics is critical, but companies tend to allocate budgets towards data processing rather than innovative analysis and new technologies, resulting in suboptimal outcomes. Embracing emerging technologies and applying artificial intelligence across the pharma value chain can bring many benefits. What are the possible applications and benefits of using AI-based solutions within the particular pharmacovigilance area? Read through our article to find out.

What is pharmacovigilance (PV)?

The etymological roots for the word “pharmacovigilance” originate from Pharmakon (Greek), which means a drug, and Vigilare (Latin), which means to be alert. The World Health Organization (WHO) describes pharmacovigilance as the science and activities relating to detecting, assessing, understanding, and anticipating adverse effects or other drug-related problems. It relies on information gathered from the collection of individual case safety reports and other pharmacoepidemiological data. Pharmacovigilance responsibilities are related to several areas:

• Timely collection of data, recording, and notification
• Safety case assessments (data completeness, seriousness)
• Expedited and periodic reporting
• Insightful data analysis
• The appropriate structure for streamlined communication
• Aggregating reports
• Literature, social media (Facebook, Twitter etc.,), and other data sources screening

The pharmacovigilance data management cycle starts with data collection in digitalized systems followed by complete data entry, including adverse event coding, drug coding, causality, expectedness assessment, narrative writing, quality control, and report submissions by data storage and maintenance.

What is an adverse event?

An adverse event (AE) is defined as any untoward occurrence in a patient or clinical investigation subject administered a pharmaceutical product and doesn’t have a casual relationship with this treatment. Therefore, an adverse event can be any unfavorable and unintended sign, including an abnormal laboratory finding, symptom, or disease temporally correlated with the use of a medicinal product, whether or not considered related to the medicinal product.

In the pre-approval clinical trial with a new medicinal product or its new usages, particularly as the therapeutic doses may not be established, adverse drug reaction (ADR) is defined as all noxious and unintended responses of a medicinal product related to any amount. For pharmaceuticals on the market, adverse drug reaction can be defined as a response to a drug that is harmful and unintended and occurs at doses generally prescribed for prevention, diagnosis, or treatment of illness or modification of physiological function.

Generally, we can divide adverse reactions into unexpected (unlisted) ones when the nature or severity of the response is not consistent with the applicable product information and the expected (listed) ones whose nature, severity, specificity, and outcome are compatible with the information provided on the product. Adverse drug reactions can have different severity from minor side effects to life-threatening, serious adverse events. ADRs can be a significant cause of deaths and emergency hospital visits. Pharmacovigilance experts are therefore monitoring and understanding ADRs can help minimize and even prevent such events. This requires pharmaceutical companies and healthcare professionals to place a greater emphasis on pharmacovigilance.

What are the main areas of pharmacovigilance?

Pharmacovigilance at the clinical development phase

Throughout drug development and the post-marketing phase, pharmaceutical companies invest heavily in monitoring and evaluating patient safety and managing the risks associated with drug exposure. Surveillance of safety data is a permanent activity designed to ensure that potential safety signals are detected early and that the chances of exposing a patient to adverse drug effects are minimized.

Clinical trials play a central role in drug development. Recognizing the possible situations relating to drug safety with the help of technology is beneficial for all stakeholders. Time is also essential, as competitive product-to-market lead times help pharmaceutical companies improve profitability. Time consumed for clinical trials involves a significant investment, and predicting ADRs is crucial for a successful market introduction of the pharmaceutical.

Post-market drug safety surveillance

Confirming drug safety during clinical trials and early product development alone is almost impossible due to data limitations. The number of subjects enrolled in the clinical studies is limited, and hence rare adverse effects of the drugs may not be identified. Due to the limited duration of clinical studies, long-term adverse effects may not be identified. Clinical studies are run on a controlled study population that yields inadequate information regarding co-existing diseases once the drug is prescribed to the general public. Additionally, unique population data, such as children, the elderly, and pregnant women, is often unavailable. What activities does PV involve once the drug is already on the market?

• mandatory reporting (Marketing Authorization Holders (MAH) record periodic safety reports)
• voluntary reporting (healthcare professionals and consumers)
• case analysis from regulatory authorities or MAH
• communications (articles, webinars, blogs, podcasts, and social media)
• post-marketing surveillance (post-marketing studies or clinical trials)
• safe use initiative (misuse or errors)
• sentinel system (system for product safety monitoring)

Recording and evaluating adverse drug reactions data

A safety signal implies a possible causal relationship between an adverse event and a drug or a new aspect of a recognized adverse event and a drug, which requires further causality assessment. Usually, more than one ADR report is necessary to generate a signal. When a signal is detected, further evaluation is warranted to determine whether an actual causal relationship exists.

The entire process of recording and evaluating ADR data is one of the most crucial steps in pharmacovigilance and is defined in numerous guidelines to various degrees and is tracked by several regulatory authorities. Before signals are published, they are first clinically assessed by the PV experts from the WHO Collaborating Centre for International Drug Monitoring at Uppsala Monitoring Center, Sweden (UMC). Signal management involves many processes, such as identifying data sources, signal detection, prioritization, evaluation, analysis, and assessment with suggestions for action. It remains at the center of pharmacovigilance and drug safety. The prerequisites for safety monitoring and signal identification include:

• the accuracy and completeness of the ADR data gathered,
• proper collection and follow-up of AE reports, including proper source data verification,
• standardized coding and assessments,
• a powerful tool for analysis.

Risk assessment and management

Evaluating and managing risks is a growing concern for pharma businesses. The pace at which information now travels means that consumers are immediately aware of safety matters with medications, even when an event occurs on the other side of the world. Business strategy must take into account the benefit-risk characteristics of each product. When effectively implemented, PV can drive competitive advantage by developing a more robust benefit-risk profile and improved identification of at-risk patients. The risk assessment and management is an infinite loop of tracking the surveillance data and evaluating risks:

• Implement risk minimization/characterization and benefit maximization
• Data collection - monitor effectiveness and collect new data
• Identify and analyze risk quantification and benefit assessment
• Evaluate benefit-risk balance and opportunities to increase and characterize
• Select and plan risk characterization/minimization and benefit maximization techniques

Artificial intelligence applications for pharmacovigilance

Like many other industries, the field of pharmacovigilance is dealing with ever-increasing volumes of data. The collection, evaluation, and reporting of adverse events, known as single case processing and continuous monitoring of adverse events and drug reactions, can benefit from artificial intelligence. Automation of repetitive and routine tasks enables quicker reporting and also helps in seamless real-time communication. Emerging companies develop AI-based solutions to improve the productivity of data reporting and the efficiency of drug development as a process.

Artificial intelligence holds great promise to address critical challenges and to provide new opportunities for the pharmacovigilance area. What are the major applications of artificial intelligence for drug safety?

Automation for improving case processing and signaling

Much of the effort in pharmacovigilance is focused on recognizing ADRs, gathering cases, evaluating them, and converting relevant data into usable information for regulators and companies to address safety concerns and inform the public. Currently, the PV companies manage increasing data volumes by adding new PV team members. But there are limits to how much growth is sustainable for the organization without hiring outsourcing services to keep up with the exponentially growing data challenge.

An essential challenge beyond volume and logistics lies in the performance limitations of human experts tasked with processing large volumes of information, affecting data interpretation accuracy and consistency (commonly referred to as the interrater agreement). As organizations grow and expand across geographies, these limitations are escalating furtherly. The good news is that nearly every process in drug safety and pharmacovigilance can be automated. For example, automation supports operations such as data entry, triage, case processing, quality review, medical review, and report submission.

Supervised and unsupervised machine learning models can automate critical stages of the safety process from intake through processing by leveraging existing safety reports, ongoing signal detection, and new unconventional sources to meet the growing scale of safety adverse reaction monitoring. AI-based solutions aid in better insights and opportunities to address the root causes of issues in the compliance metrics. Automation coupled with artificial intelligence improves the quality of safety data at the start of the process and helps streamline processing and improve signal detection and evaluation. If we let AI perform the highly repetitive manual labor of extracting and reading incoming reports, it would free up time for PV experts to review and revise the reports, confirming that the case descriptions are complete, accurate, and high quality. A streamlined process would allow for faster and more consistent safety actions than the standard manual procedure.

Cognitive case processing with machine learning

Adverse events volumes from traditional and non-traditional sources are growing, providing researchers with opportunities to develop a deeper understanding of product safety profiles. On the other hand, the cost of pharmacovigilance in both expenditure and resources is continually rising due to increasing incoming adverse event reports captured by ongoing post-marketing monitoring. Several determinants, such as an aging population, increased awareness among the public, and the number of pharma products on the market, increase reported adverse reactions. Simultaneously, the case intake and processing are faced with challenges. The case pipeline is so big that PV organizations need to move away from a manual process for managing all cases to cognitive automation of all claims and a targeted expert review of selected complex cases.

There are several tasks for machine learning applied to case processing where it can bring value. ML algorithms are great for anomaly detection. With the model’s help, it is possible to identify unusual cases or data errors that require further investigation. Machine learning is also great for searching relationships between variables and can be applied for association rule learning on safety issues. Clustering models let you discover groups that are in some way similar and can be used for detection assessment, be it by grouping similar safety reports, patients, conditions, and other surrounding adverse event information. Classification algorithms provide an option to assign new cases into previously defined groups.

The most attractive case for artificial intelligence solutions applied for case processing lies within its cognitive superpowers and ability to bring new insights or summaries that improve the quality and richness of coded case information for downstream investigation and compliance activities. Cognitive case processing application changes the focus from manual data entry and analysis activities to supervised and insight-aided quality.

The benefits mentioned above include the reduced cost per case, increased case throughput, and decreased need for expert labor through the whole safety surveillance process. The joined forces between pharmacovigilance experts and artificial intelligence systems may provide a robust solution offering enhancements in pace, scale, consistency, and information quality.

Error-free reporting with Natural Language Processing (NLP) and speech & text recognition

Automating intake and management improves the efficiency and quality of the risk management cycle significantly. How to create a system for case reporting that is scalable, user-friendly, and designed to accommodate high case volumes and large data sets? How to make the case report management automated, limiting the time needed to report? AI is here to help with several tech solutions. With the help of deep learning, pharma companies can reduce the time taken to report adverse drug reactions.

Natural language processing (NLP) solutions help digitize and normalize case reporting. Much information along the process is only available in an unstructured and free-text form, and NLP algorithms can instantly turn it into a standardized, digital version. Medical content from additional data sources such as patient and case records can be structured by applying NLP to support signal verification. Speech & text recognition further enable companies to deliver error-free reporting of various clinical and regulatory information. AI can decrease the cost of resolving each case and free up valuable resources to work on more complex and value-added tasks. Another AI solution, natural language generation (NLG), can be used for medical writing and the generation of aggregate reports, which are generated from unique case reports and the signal detection process.

These applications can free up resources, helping you to investigate anomaly cases, better manage quality, improve your understanding of product safety profiles, prepare to respond to changing regulations, and evaluate new opportunities for safety insights.

Mining literature and unconventional data sources

Another use case for AI would be applying NLP to a broad set of data, such as free text in social media, news articles, literature, or medical records to detect unexpected benefits of a pharmaceutical product. This approach uses AI and trained analysts to monitor for signals indicating incredible benefits or adverse reactions. These signals provide valuable real-world intelligence that mining data from controlled clinical situations can’t find. Auto mining of literature and other unconventional data sources could lead to an expansion of indications for an already marketed product and provides an opportunity for pharmacovigilance to improve patient care while contributing to the top-line revenues of a company. This automated first search and analysis provide space for human experts to provide further analysis and polish.

From reactive to proactive pharmacovigilance

The proactive or predictive approach in PV means that safety teams start trying to identify possible drug reactions during every stage of the drug discovery and development process, opening up the possibility of solving a problem before it becomes one.

AI and deep learning, in particular, can be applied early in the drug development process. These techniques emphasize predictive accuracy over model interpretability. Advanced learning through expanded data cohorts and cognitive innovation enables efficiencies and insights at the point of decision making, transforming the way pharmacovigilance operates. Incorporating insights, retrospective and prescriptive analysis, and decision recommendations at every stage of the process pave the way for predictive and proactive PV. With predictive analysis, drug safety teams can predict and remediate safety issues before they occur for safer patients and reduced operating costs. Predictive analytics can be used to uncover and evaluate insights and potential signals in the data that were difficult—or impossible—to identify before and resolve critical business issues at scale.

Benefits of artificial intelligence in pharmacovigilance

An expedited path to the market

Time is essential in drug development, as competitive product-to-market lead times help pharmaceutical companies win the market and improve profitability. Artificial intelligence solutions can support faster decision-making. A reduced end-to-end drug development cycle means an expedited path to the market—all these with no compromise on patient safety.

Cost-effective processes

Traditionally, pharma companies address the constantly increasing resource demands and costs by outsourcing or workforce offshoring. That’s one way to cut costs. Investing in automation and augmentation custom software solutions is another one and can bring positive ROI with tangible cost reduction. The company can reduce the financial cost of a product label update or a product recall. The overall cost of the drug development process is cut thanks to greater efficiency and PV services accuracy. Case processing has the most considerable economic impact on a pharmacovigilance budget. Automation of safety case reporting and management with machine learning algorithms can have the most significant opportunity to cut the pharmaceutical development process costs.

Expedited reporting that is error-free

Automating mundane tasks is an excellent application for artificial intelligence technologies. Time-consuming and costly direct annotation of source documents can be automated, and thanks to strict rules, can eliminate human error. NLP automates safety reporting and is the first step for centralized drug safety monitoring.

Let PV experts do high-value work

Automation also allows for better focus of PV experts by eliminating repetitive, manual tasks and focusing safety teams on high-value work. Thanks to that, resources are saved, and it guarantees valuable human assets are optimized.

Data-driven insights for safety

Enlarging data sets and sources make it impossible to process pharmacovigilance information with expert labor only. The need for streamlining is massive, and data science solutions can benefit by providing automated analysis, intelligent, valuable insights, and predictions.

Regulatory compliance

With the introduction of increased regulatory requirements globally over the past few years, the costs of pharmacovigilance operations have ballooned for pharma companies. Companies are required to maintain compliance with changing regulations across countries. Being resilient in these times is necessary and having a digitalized flexible framework that supports the drug development process and international drug monitoring is the answer.

Enhanced patient experience

The end goal of all pharmacovigilance operations is patient safety. Machine learning applied for monitoring the PV process aids in increased drug safety and treatment accuracy. More agile safety signals detection contributes towards the optimum use of therapies and enhanced patient safety. Risk minimization measures can potentially be initiated faster and result in increased accuracy. Therefore, the scientific evidence generated should be more robust. Thanks to that, the application of artificial intelligence in pharmacovigilance can further improve the ability to support and protect the health and well-being of patients and the work of healthcare professionals.

The future of pharmacovigilance

Drug safety teams are under enormous pressure to do more with less. To be more diligent and ensure the highest standards are met. As the volume of safety cases continues to rise at an exponential pace and the amount of data that must be processed grows, pharmaceutical organizations are challenged to re-think pharmacovigilance. A holistic pharmacovigilance system is not just about adverse event cases entered into a database, and endless case listings spit out for analysis. It is an iterative process that begins with the first step in the pharmacovigilance system and ends with the last, feeding back to continuous improvement and communication between data interpretation accuracy and consistency.

Artificial intelligence holds great promise for safety and pharmacovigilance and is already proven in the industry. Automation, artificial intelligence, and machine learning technologies provide opportunities to shift the pharmacovigilance function from compiling and reporting data to help to raise product quality, optimize treatment plans, reduce costs, and improve patient safety.

The shift toward AI-based pharmacovigilance management platforms can pave the way for agile pharmaceutical companies presenting compelling alternatives to traditional processes and workflow. The future of pharmacovigilance lies within digitalization, artificial intelligence analytics, and patient-centered data collection, which will likely increase overall drug safety.

References

Pharmacovigilance overview - European Medicines Agency

Pharmacovigilance (PV) - Regulation and Prequalification - WHO

What is Pharmacovigilance? - Technology Networks