Enhancing patient safety: Innovations in advanced signal detection for pharmacovigilance – A narrative review

D. Suryaprabha

Assistant Manager – Clinical Research, Kauvery Hospitals, India

Abstract

Background: Signal detection and signal management in pharmacovigilance involves ongoing monitoring to identify case reports or case series reports of adverse events (AE) that are worthy for further exploration and potentially requires safety actions such as a safety signal investigation.

Methods: A summary of the types and process of signal detection are explained in this article.

Conclusion: The specialty of pharmacovigilance has undergone significant extension on their role, especially in the methods used for signal detection. From its origins in spontaneous reporting systems to the integration of advanced data mining algorithms, pharmacovigilance remains a dynamic and crucial discipline within healthcare.

Keywords: Data Mining, Clinical Trials, AI-driven pharmacovigilance, artificial intelligence, Central Drugs Standard Control Organization (CDSCO)

Background

The etymological roots of pharmacovigilance are pharmacy and vigilance, from the Greek words Pharmakon = Drug and Latin = Vigilare, which mean “To keep awake or alert, to keep watch” and “To keep watch on drugs, in particular, their safety.”

Every pharmaceutical product, approved to be used in a marked setting, has proven benefits, but also associated adverse effects. Timely detection of such unknown risks is pivotal to ensure the patient’s safety. The detection process applies to all medicinal products, covering their entire life cycle, specifically including clinical development and post-market phases, for identifying any type of adverse event, serious or non-serious. Signal detection and signal management in pharmacovigilance involve ongoing monitoring to identify case reports or case report series of AE that are worthy for further exploration and potentially requires safety actions such as a safety signal investigation.

Traditionally, signals are detected either qualitatively or quantitatively. The former involves the qualitative analysis through the manual assessment of Individual Case Safety Reports (ICSR) in an individual or cumulative manner.

The latter, on the other hand, involves the more common quantitative approach that makes use of statistical techniques – the most often used being disproportionality analysis. To identify the disproportionate reporting ratios the most common technique used is data mining. Data mining is the process of extracting useful and actionable information from large datasets.

Thalidomide and the lesson for drug safety

The Thalidomide case forever changed pharmaceutical safety.

Today, every drug undergoes rigorous testing before approval. But pharmacovigilance does not end there: monitoring continues even after commercialization, and the participation and contribution of healthcare professionals and pharmaceutical companies are essential. Reporting an adverse drug reaction is not just a regulatory obligation, it is an act of responsibility. 

The “Thalidomide tragedy (1961)”, where thousands of babies were born with severe congenital abnormalities after their mothers took thalidomide during pregnancy. taught us a crucial lesson: every report can make a difference.

Similar was the “The Chloroform Disaster (1848) – Hannah Greener, a 15-year-old girl, died after chloroform anaesthesia during a surgical procedure,

“The Sulphanilamide Elixir Tragedy (1937) – containing a toxic solvent (diethyl glycol), led to the deaths of over 100 people in the United States.

 

Fig: 1. Drug Development Phases and Post marketing surveillance

Fig: 2 Signal Detection Overview

Types of Signals

1. Qualitative signals

It involves a subjective assessment of drug safety data, focusing on individual case reports and expert judgment to identify potential safety concerns. This approach contrasts with quantitative methods, which use statistical techniques to analyze data.

Eg. Case Review and Expert Judgment, Disproportionality Analysis (Qualitative), Spontaneous Reporting

2. Quantitative Signals

It involves using statistical methods to identify drug-event combinations that appear more frequently than expected in large databases.

Eg. Disproportionality Analysis, Reporting Odds Ratio (ROR), Information Component (IC), Bayesian Data Mining, Empirical Bayes

Fig:2 Qualitative signals Review Panel and Governance Structure

3. Spontaneous Reporting Signals

These are crucial for identifying and understanding potential drug-related problems after a drug has been marketed.

4. Data Mining Signals

These are identified by analyzing large datasets, including electronic health records, to uncover patterns or associations.

5. Literature-Based Signals

These emerge from systematic reviews, meta-analyses, and publications, providing insights into the safety profile of drugs

6. Statistical Analysis Signals

These are detected using statistical methods, such as disproportionality analysis, to identify potential signals of adverse drug reactions.

7. Prescription Event Monitoring (PEM) Signals

PEM involves tracking the use of specific drugs in a cohort of patients, allowing for the identification of adverse events that may be associated with the drug.

8. Registry Signals

Disease and drug registries can also generate safety signals by tracking specific outcomes and exposures.

9. Periodic Safety Update Reports (PSUR) Signals

PSURs, which are mandatory reports submitted to regulatory authorities, can also identify new safety signals.

10. Anecdotal Reporting Signals

Individual practitioners can report anecdotal experiences, which can contribute to the early identification of adverse drug reactions

Methods

The process of signal management in pharmacovigilance, involves several key activities aimed at identifying and assessing risks associated with medicinal products. Here is a detailed breakdown of each step in the signal management process: Fig 2

  1. Signal Detection: This is the initial and most crucial step where potential signals are identified. Traditional methods often rely on solitary algorithms which may not be sufficient due to the lack of qualitative data mining features. Efforts are ongoing to develop more robust quantitative and qualitative detection algorithms.
  2. Signal Validation: Once a potential signal is detected, it needs to be validated to ensure that it is a true signal and not a false positive. This involves checking the data for accuracy and relevance.
  3. Signal Prioritization: After validation, signals are prioritized based on their potential impact on public health. This step ensures that the most critical signals are addressed first.
  4. Signal Assessment: In this phase, the validated and prioritized signals are thoroughly assessed to understand the risk they pose. This involves a detailed analysis of the available data and may include further investigations.
  5. Recommendation for Action: Based on the assessment, recommendations are made for regulatory action or other measures to mitigate the risk. This could involve updating product labels, issuing safety warnings, or conducting further studies.
  6. Exchange of Information: Finally, information about the signals and the actions taken is exchanged with relevant stakeholders, including regulatory authorities, healthcare professionals, and the public.

Fig: 3 WHO UMC Scale

Fig 4: Monthly Drug Safety Alert by IPC

To address the limitations of traditional signal detection methods, a hybrid approach has been developed. This approach combines both quantitative and qualitative methods to improve the accuracy and effectiveness of signal detection and management. By integrating these methods, the hybrid approach provides a more comprehensive assessment of signals, allowing for better grading and management based on qualitative insights.

This detailed description of the hybrid approach in signal management highlights the importance of combining different methodologies to enhance the overall pharmacovigilance process.

Some Examples

  • Quinolones and tendon rupture:The use of quinolone antibiotics has been associated with the development of tendon rupture in animal studies. This risk was identified largely based on postmarketing rather than clinical trial data and illustrates an additional important point. Many clinicians may not consider reporting a tendon rupture to the manufacturer of an antibiotic, believing that it is “biologically not possible,” yet this adverse event was first reported in medical journals as case reports seen in the postmarketing setting. This adverse event has led to boxed warnings in all quinolone labels. This is of particular relevance in dialysis patients since quinolone use is not uncommon in this population. [1]
  • Drug-induced hepatic failure:It is a frequent cause of withdrawals of drugs from the market. Troglitazone used in diabetes is an example. It was withdrawn following cases of hepatic toxicity found in postmarketing surveillance. Hepatic reactions are rare, and therefore difficult to detect prior to marketing of a drug. A trial involving 30,000 patients would be required to detect with reasonable certainty a reaction occurring in one patient in 10,000.  As a result, postmarketing surveillance for hepatic reactions to new drugs is crucial. Drug-induced liver disease is a serious reaction and should be reported to the pharmacovigilance center immediately. This applies also to new drugs with known history of hepatotoxicity.[1]

Navigating Key Challenges in Signal Detection [3]

  • Data completeness, particularly from Spontaneous Reporting Systems (SRS)
  • Lack of denominator data
  • Not able to assign a correct data mining algorithm
  • High volume of false positives
  • Signal leakage and masking

Exploring AI-Driven Approaches in Signal Detection

AI is revolutionizing pharmacovigilance by enhancing signal detection through advanced algorithms and data analysis. It can process large datasets, identify patterns, and predict potential safety issues, resulting in faster and more accurate detection of adverse drug reactions. This transformation improves efficiency, accelerates reporting, and enables more timely risk management strategies. [2]

Challenges in AI-Driven Signal Detection

Despite the numerous benefits of AI-driven signal detection, there are several challenges that need to be addressed. One of the main challenges is the lack of standardized data sources and formats, which can hinder the interoperability and integration of AI systems in pharmacovigilance. Additionally, the complexity and variability of real-world data can pose challenges for AI algorithms, leading to potential biases and inaccuracies in signal detection. Furthermore, the regulatory landscape surrounding AI in pharmacovigilance is still evolving, creating uncertainty around the validation and acceptance of AI-driven signal detection methods. [2]

Future Directions in AI-Driven Signal Detection

To address concerns about AI adoption, healthcare organizations should occupy with regulatory bodies early to ensure compliance and help shape practical guidelines. They should invest in seamless integration strategies, including change management and staff training, to minimize interruption. Ensuring interpretability and transparency through explainable AI techniques and thorough documentation is decisive. Additionally, implementing string strict data protection measures and ethical guidelines will address data security and ethical concerns. [2]

The way forward

  • Pharmacovigilance is essential for identifying and acting upon adverse drug reactions.
  • Signal detection faces challenges from data volume and the need for precise analysis.
  • Regulatory compliance and stakeholder communication are key in managing drug safety.

Conclusion

The specialty of pharmacovigilance has undergone significant extension on their role, especially in the methods used for signal detection. From its origins in spontaneous reporting systems to the integration of advanced data mining algorithms, pharmacovigilance remains a dynamic and crucial discipline within healthcare.

Key advancements in signal detection methodologies have not only increased the efficiency and accuracy of identifying potential risks associated with medicinal products but have also expanded the scope of surveillance to include diverse data sources like electronic health records and social media. Future directions point towards further refinement of these methods, with an emphasis on leveraging computerized healthcare databases and enhancing communication strategies. This will undoubtedly improve drug safety surveillance, ensuring a higher standard of public health protection. The efforts of regulatory bodies, healthcare professionals, and pharmaceutical companies all play a pivotal role in this ongoing endeavor. As pharmacovigilance continues to adapt to technological advancements and global collaborations, it is poised to become even more integral in safeguarding patient health and advancing medical knowledge in the 21st century.

AI-driven signal detection in pharmacovigilance represents a significant advancement in ensuring drug safety. By addressing current challenges and embracing future opportunities, healthcare systems will be better equipped to detect, assess, and manage adverse drug reactions (ADRs) effectively, ultimately leading to improved patient care and public health outcomes. [2] Thus, it can be concluded that pharmacovigilance is an important tool in ensuring patient safety as by reporting the ADRs, the patient morbidity and mortality can be reduced. This also enhances the knowledge of prescribers about drug-related events, and thus appropriate modification in the treatment can be done to benefit the patients.

References

  • https://www.ctdt.co.in/abstractArticleContentBrowse/CTDT/10744/JPJ/fullText#B38
  • Anjali R Chavhan et al. AI-Driven Signal Detection in Pharmacovigilance: Advancements, Challenges, and Future Directions, IJPPR. May 2024 Vol.:30, Issue:5
  • https://www.indegene.com/what-we-think/reports/advancing-pv-signal-detection-with-emerging-technologies
  • Role of Social Media for Drug Safety Signal Detection – Scientific Figure on ResearchGate. Available from: https://www.researchgate.net/figure/Types-of-Signals-in-Pharmacovigilance_fig1_378809050 [accessed 28 Apr 2025]

Kauverian Bookshelf

Applied Medical Statistics

Journals

Kauvery Hospital
Find Doctor
Book Appointment