Skip to content
DraconiticResearch
Search Sources API
226,742 studies31 sourcesUpdated 34 min ago
Peer-reviewedCC BY 4.0Cureus

Artificial Intelligence-Enabled Pharmacovigilance for Neuropsychiatric Therapies: Advancing U.S. Drug Safety Surveillance and Regulatory Decision-Making

5 affiliations
  1. 1Applied Life Sciences, Keck Graduate Institute of Applied Life Sciences, Claremont, USA
  2. 2Information Science, University at Albany, State University of New York, Albany, USA
  3. 3Data Science, East Carolina University, Greenville, USA
  4. 4Information Systems and Analytics, Northern Kentucky University, Highland Heights, USA
  5. 5Tissue and Biomaterials Engineering, Villanova University, Villanova, USA

Abstract

Neuropsychiatric medications, including antidepressants, antipsychotics, mood stabilizers, anxiolytics, and newer agents such as esketamine, are widely prescribed, but their adverse effects are often subjective, delayed in onset, and difficult to separate from the illness being treated. Post-marketing surveillance in the United States relies on the FDA Adverse Event Reporting System (FAERS), MedWatch, the Sentinel System, and Risk Evaluation and Mitigation Strategies (REMS).

This narrative review, reported according to the Scale for the Assessment of Narrative Review Articles (SANRA), examines how well these systems detect neuropsychiatric adverse events and what artificial intelligence (AI) can add. PubMed/MEDLINE, Embase, and Web of Science were searched for literature published since January 2000, supplemented by regulatory and institutional sources.

Spontaneous reporting misses most adverse drug reactions, while stigma, diagnostic overshadowing, and uncertainty about whether a symptom reflects the drug or the disorder further reduce reporting of psychiatric effects. Disproportionality signals, such as those reported for dissociation and suicidal ideation with esketamine, need cautious interpretation because notoriety bias, solicited reporting under REMS, and confounding by indication can inflate them.

Long-latency effects such as tardive dyskinesia and antipsychotic-induced metabolic change fall outside the observation windows of premarket trials and most spontaneous reports. Natural language processing (NLP) can retrieve adverse events documented in free-text psychiatric records with good accuracy, and large language models (LLMs) show early promise for structuring adverse event narratives, but most AI studies are retrospective proofs of concept, and models trained on biased data can reproduce existing inequities.

AI is therefore best used to widen and prioritize signal detection, with causality assessment and regulatory decisions remaining with clinicians and regulators. Priorities include prospective evaluation of AI tools in live pharmacovigilance workflows, linkage of FAERS with electronic health record and claims data, point-of-care reporting tools, and efforts to reduce the stigma that discourages reporting of psychiatric adverse effects.

Related

  1. Progress in pharmacovigilance research based on artificial intelligence and real-world data: a structured narrative review.
    Xuelin Sun, Dongfang Qian, Wenjing Zhao +3
    Journal of translational medicinePeer-reviewedCC BY-NC-NDFull textPharmacology & toxicologyEurope PMCPMC
  2. Artificial Intelligence in Pharmacovigilance: Guiding Principles from the CIOMS Working Group XIV
    G. Niklas Norén, Julie Durand, Vijay Kara +64
    Drug SafetyPeer-reviewedCC BY-NCFull text5 citedPharmacology & toxicologyOpenAlex
  3. From Pattern Recognition to Decision Guidance: Artificial Intelligence in Interstitial Lung Disease Diagnosis and Surveillance
    Deliang Yang, Xue Li, David Chi Leung Lam
    Respirology (Carlton, Vic.)Peer-reviewedCC BY 4.0Full textGeneral medicinePMC
  4. Building smarter surveillance: a framework for artificial intelligence-enabled monitoring of overuse injuries in youth sports.
    Patrick Mayolo, Zachary A Blashinsky, John Wahl +4
    Frontiers in public healthPeer-reviewedCC BYFull textSurgery & anaesthesiaEurope PMCCrossref
  5. Artificial intelligence in cardiology for non-experts: decoding key terms and understanding machine learning methodologies. A Scientific Statement of the ESC Working Group on e-Cardiology, the European Heart Rhythm Association (EHRA), the Association for Acute CardioVascular Care (ACVC), the Heart Failure Association (HFA), the European Association of Cardiovascular Imaging (EACVI), the European Association of Preventive Cardiology (EAPC), and the European Association of Percutaneous Cardiovascular Interventions (EAPCI) of the ESC, the ESC Regulatory Affairs Committee, and the ESC Digital Cardiology and Artificial Intelligence Committee
    Raffaele De Lucia, Matthijs Cluitmans, Julia Ramírez +24
    European Heart Journal. Digital HealthPeer-reviewedCC BY-NC-ND 4.0Full textEpidemiology & public healthPMC
BibTeX
@article{omachonu2026artificial,
  title = {{Artificial Intelligence-Enabled Pharmacovigilance for Neuropsychiatric Therapies: Advancing U.S. Drug Safety Surveillance and Regulatory Decision-Making}},
  author = {Rebecca A Omachonu and Sandra A Uzuh and Kehinde Akinola and Kelvin Ovabor and Augustine Odibo},
  journal = {Cureus},
  year = {2026},
  month = oct,
  doi = {10.7759/cureus.117858},
  url = {https://doi.org/10.7759/cureus.117858},
  note = {CC BY 4.0},
}
RIS
TY  - JOUR
TI  - Artificial Intelligence-Enabled Pharmacovigilance for Neuropsychiatric Therapies: Advancing U.S. Drug Safety Surveillance and Regulatory Decision-Making
AU  - Rebecca A Omachonu
AU  - Sandra A Uzuh
AU  - Kehinde Akinola
AU  - Kelvin Ovabor
AU  - Augustine Odibo
PY  - 2026
DA  - 2026/10/13
JO  - Cureus
DO  - 10.7759/cureus.117858
UR  - https://doi.org/10.7759/cureus.117858
AB  - Neuropsychiatric medications, including antidepressants, antipsychotics, mood stabilizers, anxiolytics, and newer agents such as esketamine, are widely prescribed, but their adverse effects are often subjective, delayed in onset, and difficult to separate from the illness being treated. Post-marketing surveillance in the United States relies on the FDA Adverse Event Reporting System (FAERS), MedWatch, the Sentinel System, and Risk Evaluation and Mitigation Strategies (REMS).  This narrative review, reported according to the Scale for the Assessment of Narrative Review Articles (SANRA), examines how well these systems detect neuropsychiatric adverse events and what artificial intelligence (AI) can add. PubMed/MEDLINE, Embase, and Web of Science were searched for literature published since January 2000, supplemented by regulatory and institutional sources.  Spontaneous reporting misses most adverse drug reactions, while stigma, diagnostic overshadowing, and uncertainty about whether a symptom reflects the drug or the disorder further reduce reporting of psychiatric effects. Disproportionality signals, such as those reported for dissociation and suicidal ideation with esketamine, need cautious interpretation because notoriety bias, solicited reporting under REMS, and confounding by indication can inflate them.  Long-latency effects such as tardive dyskinesia and antipsychotic-induced metabolic change fall outside the observation windows of premarket trials and most spontaneous reports. Natural language processing (NLP) can retrieve adverse events documented in free-text psychiatric records with good accuracy, and large language models (LLMs) show early promise for structuring adverse event narratives, but most AI studies are retrospective proofs of concept, and models trained on biased data can reproduce existing inequities.  AI is therefore best used to widen and prioritize signal detection, with causality assessment and regulatory decisions remaining with clinicians and regulators. Priorities include prospective evaluation of AI tools in live pharmacovigilance workflows, linkage of FAERS with electronic health record and claims data, point-of-care reporting tools, and efforts to reduce the stigma that discourages reporting of psychiatric adverse effects.
SN  - 2168-8184
ER  -