AI-assisted warfarin dose optimisation with CURATE.AI for clinical impact: Retrospective data analysis

Area of Science:

• Pharmacogenomics
• Artificial Intelligence in Medicine
• Clinical Decision Support Systems

Background:

• Standard warfarin dosing relies on physician guidance, often leading to suboptimal International Normalized Ratio (INR) control.
• Inter- and intra-patient variability in warfarin response can result in adverse events like bleeding or thrombosis.

Purpose of the Study:

• To evaluate CURATE.AI, an AI platform, for optimizing warfarin dose titration.
• To compare the performance of CURATE.AI against existing models and physician-guided dosing.

Main Methods:

• A personalized CURATE.AI response profile was created using patient warfarin doses and INR changes.
• CURATE.AI's predictive performance was assessed using metrics like Percentage Absolute Prediction Error and Percentage Prediction Error.
• Retrospective analysis of 127 patients' data.

Main Results:

• CURATE.AI showed improved prediction accuracy (lower Percentage Absolute Prediction Error and Percentage Prediction Error of 20%) compared to other models.
• The AI demonstrated negligible underprediction bias, potentially reducing bleeding risk.
• Modeled time in therapeutic range with CURATE.AI was comparable to physician-guided dosing.

Conclusions:

• CURATE.AI offers a systematic approach to warfarin dosing, reducing physician guesswork.
• This AI platform can support effective warfarin use, potentially avoiding expensive novel anticoagulants.
• Further prospective studies are warranted to establish CURATE.AI as a clinical decision support tool.