The process of fitting a patient’s adverse event into standardized medical terminology can resemble putting a square peg in a round hole. Patients speak in layman’s terms, but MedDRA (the standardized medical terminology for all regulatory submissions) is more technical, and the patients, doctors, and nurses reporting aren’t versed in MedDRA coding.
Jane Reed, director of life science at Linguamatics (an IQVIA company), connected with Outsourcing-Pharma to talk about how pharmaceutical companies and their research partners are putting AI to work in order to elevate patient safety reporting, reduce manual coding processes, and cull important insights from information gathered.
OSP: Could you please tell us a bit about Linguamatics—who you are, what you do, and what sets you apart from others operating in the same sphere?
JR: Linguamatics delivers a market-leading natural language processing (NLP) based AI platform for high-value knowledge discovery and decision support from text. We empower our customers to speed up drug development and improve patient outcomes by breaking down data silos, boosting innovation, enhancing quality, and reducing risk and complexity.
Our award-winning NLP platform is proven across multiple real-world use cases. Linguamatics has been trusted for over 15 years to deliver actionable insights that address pharma’s most pressing bench-to-bedside challenges with quantifiable ROI.
OSP: Could you discuss some of the key challenges in ensuring/monitoring drug safety?
JR: Drug safety is of critical importance at all stages in drug discovery, development, and delivery. Only around 15% of drugs successfully make it from clinical trials to FDA approval and most failures (over 70%) are due to safety and efficacy concerns. And, once a drug is on the market, safety concerns mean that monitoring has to continue.
However, there is a growing volume of safety events in increasingly varied formats, including safety case reports, emails, call center feeds, scientific literature, internal safety document, social media, and more. This is leading to unsustainable increases in the costs of traditional safety operations.
Finding and extracting the adverse event itself, plus the relevant context around it, is becoming increasingly burdensome for safety assessment and pharmacovigilance teams using traditional manual processes.
OSP: What is MedDRA coding, and how is it useful?
JR: MedDRA refers to the Medical Dictionary for Regulatory Activities. MedDRA is terminology for medical coding and communication that includes adverse events, symptoms, diseases, medical device malfunctions, and medication errors.
MedDRA is used as the adverse event reporting terminology by many drug regulatory authorities and the pharmaceutical industry worldwide. MedDRA PTs can be used to describe adverse events and adverse drug reactions, as well as coding information about patient condition or medical history.
OSP: Why is it challenging to capture a holistic view of a product’s safety profile, even with MedDRA?
JR: Capturing a holistic view of a product’s safety profile has never been particularly straightforward, and it has only gotten more challenging in recent years. The reason is simple: adverse events are reported in natural language, and the people reporting these events aren’t MedDRA coders; they are nurses, physicians, and patients, each with a unique way of expressing themselves.
To add to the complexity, these reporters have more reporting routes available than ever before, such as patient forums and social media, creating a deluge of natural language safety events that must be fully captured and understood. Even in scientific literature, the vocabulary to describe adverse events is not standardized.
Here's an example: Say a patient describes her experience with a new drug by telling her physician, “I had a horrible headache and couldn’t sleep for two days.” The word “headache” is a one-to-one match with MedDRA so coding for it is easy. In contrast, “couldn’t sleep” would not be understood automatically because it needs to be coded as “sleeplessness” in MedDRA. Finding that correct code requires a manual database search of adverse events reports that take up valuable time.
OSP: Please describe how a typical adverse might be reported by a patient or clinician and how that information makes its way to pharma companies.
JR: There are a number of core pathways for adverse event reporting. Globally, most drug regulators have report call lines or websites. For example, the FDA requires the following statement to be distributed with dispensed new and refill prescription drugs: “Call your doctor for medical advice about side effects. You may report side effects to FDA at 1-800-FDA-1088.”
For web submissions, patients, pharmacists, or other healthcare professionals may use FDA’s Medwatch or the UK’s MHRA Yellow Card site, for example. Drug manufacturers can also then access the publicly available information around adverse event reports in FAERS.
Pharmaceutical companies also receive adverse event reports directly from patients or healthcare professionals via emails, call centers, adverse event reporting websites, or social media. Any potential adverse event or safety concern, from any of these different paths, need to be fed into a pharmaceutical company’s safety vigilance system for processing, review, analysis, and reporting.
OSP: Could you please talk about NLP, and why that technology is well-suited for pharma companies seeking to understand product safety issues?
JR: NLP technology can free pharma companies from manual coding by effectively using artificial intelligence to “read” adverse event reports in their natural language and standardize them to MedDRA codes. That means that even if a single medical concept is expressed in multiple different ways across reports, NLP can, as a human would, understand the nuance in language and properly code the event.
Using NLP, pharma companies can extract the appropriate context of each event, for example, understanding the difference between “drug causing disease” vs. “drug treating disease” or the differences between an adverse event, patient indication, or medical history. NLP solutions can understand what we call morphological variants, or predictable changes a word undergoes as a result of syntax, such as “patient/patients/patients’/patient’s” or “takes/took/is taking.” Additionally, leading NLP solutions can capture common spelling mistakes and code them appropriately without manual intervention, as well as match across conjunctive words, such as capturing “head and neck pain” as two distinct events.
OSP: Can you provide any examples of specific ways pharma companies are using this type of technology to improve their coding of adverse events?
JR: One top 50 pharma company has used NLP to solve a common challenge today in the industry: coding adverse event verbatims into MedDRA, during case report processing. Initially, this company was only able to auto-code about 30% of its verbatims, meaning 70% had to be manually coded. The company works in the rare disease space, so about 90% of the verbatims they encountered were unique, or something that manual coders had not come across before and likely would not see in the future.
Prior to advances in NLP, most companies in this situation would create a synonym list to try to capture words that are different but have the same meaning. But for this client, with 90% of verbatims never recurring, creating such an extensive list would be time-consuming and costly. Clearly, the company needed another solution.
We worked with this client to leverage NLP to improve their auto-coding capabilities. We collaborated with their medical team to create MedDRA mappings and compare those to manual coding for both the adverse event and the indication.
In the end, we were able to double the level of auto-coding, moving from 30% to 60% with a very low level of mismatch. As a result, we not only improved the coding consistency but also reduced risk for case processing in medical evaluation.
OSP: In addition to advancing product safety initiatives, can you discuss some other ways these tools are improving pharma efficiencies and insights?
JR: Leading pharma companies are using NLP to unlock insights from unstructured data at virtually every stage of the drug development lifecycle. For example, during the early discovery phase of research and development, companies leverage NLP to access literature sources for landscapes of gene-disease associations, a process that reduces time spent on manual curation.
Additionally, pharma companies are using NLP during the development phase to optimize trial design and capture valuable clinical competitive intelligence by rapidly analyzing information such as clinical trial site, eligibility criteria, study characteristics, and patient numbers. Separately, in the post-market phase, pharma companies are using NLP to look across scientific literature and prescription databases to enable a better understanding of drug-drug interactions and co-prescription trends, or across social media to understand patient perceptions of disease and drugs.