Computational Pharmacovigilance of Newly Launched Oral GLP-1 Receptor Agonists: Multi-Measure Signal Detection and Temporally Validated Machine Learning on FAERS

Purpose: Orforglipron and oral semaglutide 25 mg entered the US obesity market in 2026 without post-marketing safety data. This study aimed to detect early adverse-event reporting signals for both agents and to test whether machine-learning models that triage serious reports remain reliable after a product launch. Methods: We built a computational pharmacovigilance pipeline on deduplicated reports from the US Food and Drug Administration Adverse Event Reporting System (FAERS) received from January 2024 to September 2026. Each oral product was compared with injectable incretins reported over the same period using four disproportionality measures with false-discovery-rate control, sensitivity analyses and positive and negative controls. Four classifiers trained on 2024-2025 reports were validated on 2026 reports against simple baselines, with calibration and conformal prediction. Results: Of 212,284 reports, 3043 concerned orforglipron and 1543 oral semaglutide 25 mg. Gastrointestinal intolerance and perceived non-response were disproportionately reported for both agents (reporting odds ratios 1.4-4.0), consistently across analyses, and no pancreatic, biliary, hepatic or renal signal emerged. Gradient boosting reached an area under the curve of 0.86, but a simple reporting-process baseline reached 0.84, performance fell to 0.74 for orforglipron, and only class-conditional conformal prediction restored target coverage of serious reports. Conclusion: Early reports for both oral agents are dominated by gastrointestinal intolerance and perceived non-response; these hypothesis-generating signals cannot establish incidence or causality. Report-triage models largely learn how reports are filed and need product-specific monitoring after launch.

Authors

Institutions

Publication Details

Journal
Research Square
Published
2026-11-20
DOI
https://doi.org/10.21203/rs.3.rs-11792528/v1
Primary Topic
Pharmacovigilance and Adverse Drug Reactions
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Computational Pharmacovigilance of Newly Launched Oral GLP-1 Receptor Agonists: Multi-Measure Signal Detection and Temporally Validated Machine Learning on FAERS

Tahsien Al‐Quraishi
Research Square
Pharmacovigilance and Adverse Drug Reactions
preprint

Computational Pharmacovigilance of Newly Launched Oral GLP-1 Receptor Agonists: Multi-Measure Signal Detection and Temporally Validated Machine Learning on FAERS

Tahsien Al‐Quraishi
preprint en

Abstract

Purpose: Orforglipron and oral semaglutide 25 mg entered the US obesity market in 2026 without post-marketing safety data. This study aimed to detect early adverse-event reporting signals for both agents and to test whether machine-learning models that triage serious reports remain reliable after a product launch. Methods: We built a computational pharmacovigilance pipeline on deduplicated reports from the US Food and Drug Administration Adverse Event Reporting System (FAERS) received from January 2024 to September 2026. Each oral product was compared with injectable incretins reported over the same period using four disproportionality measures with false-discovery-rate control, sensitivity analyses and positive and negative controls. Four classifiers trained on 2024-2025 reports were validated on 2026 reports against simple baselines, with calibration and conformal prediction. Results: Of 212,284 reports, 3043 concerned orforglipron and 1543 oral semaglutide 25 mg. Gastrointestinal intolerance and perceived non-response were disproportionately reported for both agents (reporting odds ratios 1.4-4.0), consistently across analyses, and no pancreatic, biliary, hepatic or renal signal emerged. Gradient boosting reached an area under the curve of 0.86, but a simple reporting-process baseline reached 0.84, performance fell to 0.74 for orforglipron, and only class-conditional conformal prediction restored target coverage of serious reports. Conclusion: Early reports for both oral agents are dominated by gastrointestinal intolerance and perceived non-response; these hypothesis-generating signals cannot establish incidence or causality. Report-triage models largely learn how reports are filed and need product-specific monitoring after launch.

Research Square
Al-Ayen Iraqi University (IQ), Melbourne Institute of Technology (AU)
Pharmacovigilance and Adverse Drug Reactions
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon