Post-market surveillance for literature on QuantiFERON TB Gold Plus: what improvements can artificial intelligence bring?
July 28, 2026 Ed Deane
Reniewicz, J. Alagna, R. Kordylas, L. Latacz, M. Suryaprakash, V. Nowak, U. Kois-Ostrowska, A. Weleszczuk, J. Blacha, A. Nikolayevskyy, V.
This is an independent, peer-reviewed research article published in IJTLD Open, an open-access journal from The Union covering tuberculosis and respiratory disease research. The study was authored by scientists at QIAGEN (including affiliations at QIAGEN Wroclaw, QIAGEN S.p.A., QIAGEN Hilden, and QIAGEN Manchester) in collaboration with Imperial College London, and it went through the standard peer-review process before publication. The paper discloses that all co-authors are QIAGEN employees and reports no other conflicts of interest.
The platform evaluated in the study is referred to throughout as the “Huma.ai Platform,” the name of our company at the time the research was conducted and submitted. QIAGEN and our team co-developed the platform together, and the underlying system is the same one we now operate as Qoniq.
QIAGEN needed to know whether an AI-assisted approach to literature surveillance could meet the EU In Vitro Diagnostic Regulation’s (IVDR) requirement for continuous, literature-based post-market surveillance (PMS), and whether it could do so without sacrificing the precision regulators expect.
Using QuantiFERON TB Gold Plus (QFT-Plus), a mature, high-volume IVD assay, as the test case, the study ran a formal head-to-head comparison: the AI-assisted platform versus QIAGEN’s existing manual Boolean search process, both querying PubMed and PubMed Central across the same full year of 2024 publications. Five reviewers screened and evaluated the results, with an independent reviewer conducting quality control on a randomly sampled subset (12 articles from the manual output and 27 from the AI-assisted output), and a third reviewer resolving any disagreements.
Notably, the platform’s methodology is built on classical natural language processing and pattern recognition rather than generative AI or large language models. That’s a deliberate design choice for a regulated environment, where transparency and reproducible, deterministic results matter when a notified body needs to understand how a result was generated.
Coverage: The Huma.ai (now Qoniq) Platform identified 673 articles for 2024, compared to 111 from the manual search. Of those, 561 were unique to the AI-assisted method and not surfaced by manual review at all.
Precision: Despite retrieving roughly six times as many articles, the platform’s precision rate of 98.21% slightly exceeded the manual method’s 95.50%. Broader coverage did not come at the cost of more noise.
Screening time: Reviewers confirmed article relevance in about 16 seconds using the AI-assisted platform, versus roughly 60 seconds manually, while full-text expert review time remained consistent at approximately 15 minutes per article across both methods.
Safety signals: Across all 664 relevant articles identified by either method, no new or previously unrecognized safety or performance signals were found. Because the AI-assisted search was so much broader, that clean result carries more evidentiary weight than the narrower manual search could have provided on its own.
Manual-only misses: The handful of articles found only by manual search were traced back to incomplete PubMed metadata rather than any platform limitation, and none altered the risk conclusions.
The authors frame the core PMS challenge for mature assays like QFT-Plus this way: reviewer time gets consumed re-confirming what’s already known, while genuinely new signals can hide in literature that a narrower, keyword-dependent search never touches. The study’s authors conclude that AI-assisted surveillance can significantly enhance the efficiency, consistency, and comprehensiveness of PMS for mature IVDs, and that as regulatory frameworks evolve, integrating explainable, validated AI systems will be essential for ethical, transparent, and effective surveillance.
The study’s scope was intentionally bounded, restricted to 2024 English-language, human-study publications indexed in PubMed and PubMed Central, without extending to Embase, Scopus, preprints, or non-English sources. So its findings speak to relative performance between the two methods within that scope, not to absolute recall across all possible literature.
World Health Organization. Guidance for post-market surveillance and market surveillance of medical devices, including in vitro diagnostics. Geneva: WHO, 2020.
World Health Organization. WHO consolidated guidelines on tuberculosis: module 3: diagnosis. Guideline/technical guidance. Geneva: WHO, 2025.
World Health Organization. Use of alternative interferon-gamma release assays for the diagnosis of TB infection: WHO policy statement. Technical guidance. Geneva: WHO, 2022.
Prevention CfDCa. Clinical testing guidance for tuberculosis: Interferon gamma release assay. Atlanta, GA, USA: U.S. Department of Health and Human Services, Centers for Disease Control and Prevention, 2024.
Oh CE, et al. Comparing the diagnostic performance of QuantiFERON-TB gold plus to other tests of latent tuberculosis infection: a systematic review and meta-analysis. Clin Infect Dis. 2021;73(5):e1116-e1125.
Reniewicz J, et al. Artificial intelligence/machine-learning tool for post-market surveillance of in vitro diagnostic assays. N Biotechnol. 2024;79:82-90.
Elbehiry A, et al. Advancing the fight against tuberculosis: integrating innovation and public health in diagnosis, treatment, vaccine development, and implementation science. Front Med (Lausanne). 2025;12:1596579.
Menon S, Koura KG. Artificial intelligence for tuberculosis control: a scoping review of applications in public health. J Glob Health. 2025;15:04192.
Orel E, et al. An automated literature review tool (LiteRev) for streamlining and accelerating research using natural language processing and machine learning: descriptive performance evaluation study. J Med Internet Res. 2023;25:e39736.
Jin Q, Leaman R, Lu Z. PubMed and beyond: biomedical literature search in the age of artificial intelligence. EBioMedicine. 2024;100:104988.
Carey N, Harte M, Mc Cullagh L. A text-mining tool generated title-abstract screening workload savings: performance evaluation versus single-human screening. J Clin Epidemiol. 2022;149:53-59.
Garcia-Gomez JM, et al. Mitigating patient harm risks: a proposal of requirements for AI in healthcare. Artif Intell Med. 2025;167:103168.
Cubukcu HC, et al. Regulating the future of laboratory medicine: european regulatory landscape of AI-driven medical device software in laboratory medicine. Clin Chem Lab Med. 2025;63(10):1891-1914.
Kalodanis K, et al. Evaluating the impact of the EU AI act on medical device regulation. Stud Health Technol Inf. 2025;323:40-44.
World Health Organization. Good pharmacovigilance practices for the pharmaceutical industry. Geneva: WHO, 2021.
Shahinda SR, Alsayed HG. Tuberculosis: pathogenesis, current treatment regimens and new drug targets. Int J Mol Sci. 2023;24(6):1-23.
Ma J, Wu X, Huang L. The use of artificial intelligence in literature search and selection of the PubMed database. Sci Program. 2022;2022(1):8855307-8855309.
Müller AC, Guido S. Introduction to machine learning with python: a guide for data scientists. Sebastopol, CA, USA: O’Reilly Media, 2016.
Kandhare P, et al. Artificial intelligence in pharmaceutical sciences: a comprehensive review. Med Nov Technol Devices. 2025;27:100375.
Wani P, et al. Role of artificial intelligence in pharmacovigilance: a concise review. J Pharm Negat Results. 2022;13(7):6149-6156.
Farah L, et al. Are current clinical studies on artificial intelligence-based medical devices comprehensive enough to support a full health technology assessment? A systematic review. Artif Intell Med. 2023;140:102547.
Liu Z, et al. AI-based language models powering drug discovery and development. Drug Discov Today. 2021;26(11):2593-2607.
Liu F, Panagiotakos D. Real-world data: a brief review of the methods, applications, challenges and opportunities. BMC Med Res Methodol. 2022;22(1):287.