Enhancing Post-Marketing Surveillance with Generative AI
March 23, 2026 Susan Shiff
A diagnostics company that designs, manufactures, and markets molecular diagnostic assays and the automated analyzer platforms that run them. The Company’s core product line is a family of multiplex real-time PCR panels for respiratory pathogen detection, with additional products in earlier stages of development and a declining legacy line of single-analyte rapid antigen tests.
The Medical and Regulatory Affairs department was under pressure from four directions simultaneously.
EU IVDR transformed literature review from optional to mandatory. Under the old EU IVD Directive, manufacturers could self-certify most devices without systematic literature surveillance. The new EU IVDR changed this fundamentally. Manufacturers of higher-risk devices must maintain a Post-Market Surveillance Plan that includes systematic and planned gathering and review of experience gained from devices placed on the market. The associated regulatory guidance specifies that this includes continuous monitoring of scientific literature, databases of adverse events, and published clinical studies.
For this company, “continuous monitoring” fell on one person. The literature surveillance specialist was responsible for conducting and documenting systematic literature reviews across all marketed products on an ongoing basis. Not annually. Not when a regulatory submission was due. Continuously. She had to monitor PubMed, Embase, and the Cochrane Library for any publication mentioning the company’s products, competitors’ products, or the clinical conditions the products address. She had to assess each publication for safety signals, performance concerns, and changes in the state of the art. She had to document her search methodology, her inclusion and exclusion criteria, and her conclusions in a format that would survive a notified body audit.
She estimated she was spending 25-30 hours per week on literature monitoring alone, essentially her entire working week, and still falling behind. The Company’s flagship panel covers more than 20 pathogen targets; a comprehensive literature search for respiratory pathogen diagnostics generates hundreds of new publications per month. She was running queries manually against PubMed and Embase, exporting results to Excel, deduplicating, screening titles and abstracts, reading the included papers, and writing summary assessments. By the time she finished one quarterly cycle for the flagship product, the next cycle was due.
The evidence landscape was shifting faster than the company could track. The post-COVID era produced an explosion of publications on respiratory pathogen detection: co-infection patterns, emerging variants that might affect PCR target regions, head-to-head comparisons between multiplex panels, and the clinical impact of rapid molecular testing on antimicrobial stewardship and patient outcomes.
This literature directly affected the Company’s commercial positioning. When a paper showed that a competitor’s panel had higher sensitivity for a specific pathogen in a specific patient population, the MSLs needed to understand the study design, assess whether the difference was clinically meaningful, and prepare a scientific response before a laboratory director brought it up in a meeting. The Medical Affairs Manager experienced this as a timing failure. Her MSLs were walking into hospital laboratory director meetings where the director had already read a competitor-favorable paper that the Company’s team had not seen yet.
Pharmacovigilance reporting was outgrowing manual methods. The Quality and Regulatory Compliance team processes vigilance reports, the IVD equivalent of adverse event reporting. The EU IVDR outlines that the regulatory obligation is to conduct trend reporting to identify patterns that might constitute a safety signal. The team was reviewing the literature manually, searching PubMed for relevant published data on the specific pathogen and specimen type, and writing narrative assessments.
Clinical Evaluation Reports (CER) were becoming a full-time writing project. Under EU IVDR, every device requires a Clinical Evaluation Report that synthesizes all available clinical evidence into a comprehensive assessment of the device’s safety and performance. The CER must be updated whenever new evidence becomes available and must be consistent with the device’s technical file, the post-market clinical follow-up plan, and the summary of safety and performance.
The VP approved a technology evaluation with requirements reflecting three sides of the department: regulatory audit traceability, scientific exchange, and pharmacovigilance support.
There was a need for audit defensibility. The platform had to run predefined searches against PubMed and Embase on a scheduled basis, deduplicate results, and present new citations for screening. The process had to produce an auditable trail showing search dates, queries used, results returned, and screening decisions, sufficient to satisfy an auditor reviewing IVDR post-market surveillance compliance.
The VP’s wanted the platform to apply AI-based assessments to distinguish results that were genuinely relevant from results that matched the query terms but were not useful: animal studies, bibliography citations, etc.
The Medical Affairs Manager did not need a separate MSL tool; rather, she needed the surveillance infrastructure that surfaced newly identified publications within days of publication, bridging the gap between the rigorous quarterly cycles and the MSLs’ need for timely competitive intelligence.
The Company evaluated Qoniq alongside two other vendors: a large enterprise clinical evidence management platform and a general-purpose AI research tool.
Literature surveillance automation. Qoniq ingested the company’s existing search strategies and ran them against PubMed and Embase on a scheduled daily cycle, deduplicating results across sources. The literature specialist reviewed new citations through the platform’s screening interface rather than manually exporting to Excel.
The initial weeks revealed that her existing queries were missing publications. Publications were appearing in one provider’s results but not the other’s for the same query, and the specialist could not explain why without spending time investigating the indexing differences for each specific paper.
Qoniq’s AI-assisted triage layer addressed the filtering problem directly. The platform classified incoming results by relevance to the company’s surveillance objectives. The triage model learned from the specialist’s screening decisions and improved over the first several weeks. More importantly for audit purposes, the platform documented why each result was flagged, creating an auditable trail connecting search terms to indexing metadata to screening decisions.
The specialist’s quarterly surveillance cycle moved to a weekly cadence. She was reviewing the same volume of literature in less time because the platform was handling deduplication, initial relevance classification, and documentation automatically. Her time shifted from running searches and managing spreadsheets to evaluating the publications.
Literature retrieval. The team wanted Qoniq to support on-demand literature searches. The Qoniq platform rapidly surfaced all relevant published evidence: performance data, known limitations, and competitor comparisons, keyed to that pathogen and specimen type combination. This was not the same as the scheduled surveillance queries. It was on-demand retrieval that returned results in minutes rather than the days a manual search had previously required.
CER support. The medical writer began using Qoniq’s structured literature summaries (study design, patient population, comparator, key findings, quality assessment) as input for Clinical Evaluation Report updates. The summaries were formatted for direct incorporation into the CER’s evidence synthesis sections, reducing the time he spent on information retrieval and reformatting. This did not eliminate the writing work. CERs require regulatory judgment about how to characterize evidence, how to weigh conflicting results, and how to frame the overall benefit-risk narrative. But it removed the 70% of his effort that had been going into finding and summarizing the evidence in the first place.
MSL evidence alerting. As the surveillance infrastructure matured, the Medical Affairs Manager saw that the Qoniq system could also surface newly identified publications relevant to specific products, pathogens, or clinical claims to MSLs within days of publication. This was not a separate implementation. It was the same filtered, quality-assessed evidence stream that served regulatory purposes, made accessible to the field team through a different view. MSLs stopped running their own ad hoc PubMed alerts.
After the first year of operation, the department reported the following outcomes:
Audit readiness. When the notified body returned for its next assessment, the Company was able to demonstrate a reproducible, documented process for continuous literature monitoring with full traceability from search query to screening decision to summary assessment.
Surveillance cadence. Literature surveillance moved from quarterly batch cycles to a weekly cadence across all marketed products.
Specialist time reallocation. The literature specialist estimated that she had reclaimed approximately 15 hours per week from search execution, deduplication, and spreadsheet management. She was spending that time on the parts of her job that required human judgment: evaluating borderline publications, updating strategies based on evolving terminology, and supporting the medical writer with evidence quality assessments.
CER production efficiency. The medical writer estimated that CER update cycles shortened from three to four weeks per product to approximately two weeks. The time savings came almost entirely from the evidence retrieval and summarization phase. The regulatory writing itself took the same amount of time, because it required the same level of clinical and regulatory judgment as it always had.
MSL preparedness. The Medical Affairs Manager tracked the gap between publication date and MSL awareness for competitor-relevant papers. Before the implementation, the average gap was six to eight weeks (driven by the quarterly surveillance cycle). After, it dropped to less than one week. She cited a specific instance where the platform surfaced a head-to-head comparison study within days of publication, and the MSL covering that territory was able to discuss the study’s methodology and clinical significance in a meeting with a hospital laboratory director the following week. Under the previous process, that paper would not have been flagged until the next quarterly cycle.
The VP observed that the technology investment had solved a problem he thought was a staffing problem. His initial instinct had been to hire a second literature surveillance specialist, and he had been delayed by budget constraints and a thin candidate market for regulatory-grade systematic reviewers. In retrospect, a second person doing manual surveillance would have doubled the throughput but not fixed the methodology. The reproducibility and auditability gaps that the notified body had flagged were process problems, not capacity problems. The platform addressed both.
The literature surveillance specialist offered a more nuanced view. The platform had freed her from the mechanical parts of her work, which she appreciated. But it had also surfaced the degree to which her existing strategies had been quietly degrading. The Qoniq platform’s broader search and cross-provider deduplication made these gaps visible in a way that her manual process never could.
The Medical Affairs Manager was satisfied but measured. The MSL alerting worked because it was a byproduct of the regulatory surveillance infrastructure, not a standalone tool. Her MSLs did not have to learn a new system or configure anything. They received relevant publications through the same platform the specialist was using for regulatory purposes, filtered by the same quality-assessed criteria. The simplicity of that arrangement, she noted, was the reason adoption was not a problem. If it had required MSLs to set up profiles or curate preferences, she would still be fighting the same battle she had fought with their ad hoc PubMed alerts.
Related Content
Explore other case studies, blog entries, and news articles.
How One Pharmaceutical Company Transformed Medical Affairs from Reactive to Proactive
A pharmaceutical company with a marketed product in a competitive therapeutic area. The drug requires companion diagnostic testing, which adds a layer of complexity to every Medical Science Liaison…
Ready to let your scientists science?
Book a demo today to see firsthand the power and capability Qoniq can bring to your team.