There is a growing emphasis on signal detection and signal management in Pharmacovigilance. Companies must be able to manage signals detected: evaluate them to understand the clinical risk in light of multiple impacting factors, in a demonstrably controlled manner. Benefit/risk balance for a drug, vaccine, or treatment is impacted by the disease the product is intended to treat, distribution of the disease and product, clinical trial data, environmental influences, and other epidemiological factors. Signal management requires considered workflow and rigorous process control.
Pharmacovigilance software is changing
This is not news. The evolution of Pharmacovigilance is impacting the way we approach drug safety management of potential safety signals.
Pharmacovigilance initially was a reactive function. Only after the infamous thalidomide tragedy, and others, did regulators and industry become attuned to their shared responsibility to be good stewards of approved products.
And look at how far we have come!
Thanks to the work of dedicated experts worldwide, committed to ensuring data quality, compatibility and utility, we now have:
- Pages of standard code lists
- Common standard file formats
- Globally aligned systems
- …and more and more data collected.
Not only are the systems and processes in place, we have moved the needle from a purely required practice to a truly value added activity – where patients, practitioners, medical product manufacturers and regulators alike view Pharmacovigilance for what it is: a means to take data from a variety of environments – be it controlled clinical trials or “glorified gossip” from the field – and evaluate it for trends or signals, that can be validated or discarded.
For various reasons, such as regulatory compliance in some regions, good manufacturing practices, and responsible product monitoring in others, collection of adverse event (AE) data with approved, marketed product is an important part of product stewardship.
An effective Pharmacovigilance software (PV) system should not only efficiently collect this information, but also include a proactive risk management strategy with methods for identifying, analyzing, evaluating, communicating and mitigating risk; as well as an escalation plan for potentially serious issues.
The problem is, Industry is under a crunch:
- Increasing approved product AE caseloads (8-10% animal health, 10-12% human health) year on year, driven by new product launches, increased PV awareness, and evolving PV regulations worldwide
- Tighter budgets and fewer resources
So far, the focus has been on automating accurate case capture, entry and processing, but what about making the second part more efficient – proactive signal detection, risk identification and risk management.
What is a signal and how do we find them?
In the simplest terms, a signal in Pharmacovigilance software is an instance when the data is trying to alert us that something is different compared to prior or expected data.
Effective signal detection is in fact an interplay between the safety reviewers or medical experts and the data, via continuous data monitoring and detection of disproportionate reporting patterns among product-event pairs. Accurate signal detection depends on quality data, knowledge of the database, and tools designed to visualize, summarize, and evaluate the data.
#1 Quality data in, Quality surveillance out:
First, it is critical to have an organized, standard process as well as a controlled system for collecting and managing adverse event data, independent of the source.
Ideally it is best to collect data at the most granular level, stored in data fields that can be queried via standard code-lists. Collecting individual data points within a case may seem tedious but proves highly valuable during data analysis and evaluation. It is much easier to aggregate data points than to tease it apart later.
- Collecting data in discrete fields rather than in free texts allows for more structured querying, classification, and summarization.
- Utilizing standard code-list ensures that the data is consistent across systems, categories, and time. It also prevents masking of potential signals due to disorganized data.
- Normalizing datasets to comparable units – e.g. converting patient age data collected at source in units of weeks, months, and years to a ‘lowest common denominator’ unit.
Consider this: Data collected in free text or in other non-standardized format requires individual case review, which can be time consuming and resource intensive. Alternatively, case data where patient, product, clinical sign, causality, seriousness, and other case elements are captured via standard code lists (e.g. ICH standard, MedDRA coding dictionary) can be queried, tabulated, graphically displayed, and evaluated not only by safety reviewers, but also integrated into algorithms to facilitate artificial intelligence (AI).
#2 Know your Data
Second, it is imperative that those who are evaluating the data have a good understanding of the database –i.e.
- Patient groups
- The environment in which the product is used
- Market trends
- Product distribution
- Data source
- …and other relevant factors that may impact the frequency and nature of adverse events reported.
Additionally, it is helpful to categorize data to account for case quality and relatedness. Without this knowledge, there is a risk of mistaking otherwise expected data changes as disconcerting trends or signals. What better way to get to know your data than to “see” it in real time, learn its normal patterns. Then, develop early detection indicators to pick up on disconcerting trends and take action, before they have significant impact to patients.
Did you know: At certain times it is expected to see elevated frequency of adverse event reports, depending on the product’s life cycle stage , seasonality of distribution, or because of publicized events. Differences also occurs in trends based on product categories – e.g. biological products versus pharmaceutical products, due to their intrinsic attributes and manufacturing processes.
#3 Have the right tools to detect true signals:
The main goal of signal detection is to identify either new risks with a product in a patient population or identify a worrisome trend which may negatively impact patient health.
No one wants to create more, non-value-added work for themselves or their team. That’s why we must have the right tool to do the job. Otherwise, important signals may be missed or erroneous signals found. Ideally, a customizable computerized system, designed to fit the specific needs of your product and market which accurately identifies valid signals, aligned with a risk management strategy, are the best tools to have.
Identification of a signal does not mean that there is a problem, but is an indicator that needs to be evaluated to determine its risk level. By working up potential PV risks in the context of a well-defined risk management program, medical product manufacturers and drug sponsors can systematically evaluate the level of impact, define a communication plan, risk mitigation actions, and risk review timelines.
Kristen is a pharmacovigilance expert and holds a Doctorate of Veterinary Medicine, serving as our Executive Sales Consultant. As the Head of the Global PV team for Merial Animal Health, she developed innovative statistical signal detection tools and risk management strategies to support marketed pharmaceutical and biological products worldwide. Kristen brings more than a decade of global PV experience, and is committed to working with customers to develop innovative solutions to meet their compliance and surveillance needs. In her free time, Kristen enjoys spending time with family, training for long-distance races, exploring new recipes,and networking at every opportunity.