Optimal decision support rules improve personalize warfarin treatment outcomes

Area of Science:

• Pharmacogenomics and Computational Biology
• Clinical Decision Support Systems

Background:

• Warfarin dosing requires careful management to balance bleeding and thrombosis risks.
• Personalized treatment strategies incorporating clinical and genetic factors are crucial for optimizing warfarin therapy.

Purpose of the Study:

• To develop and evaluate optimization-based approaches for generating decision support rules for personalized warfarin treatment.
• To minimize adverse events, specifically bleeding and thrombosis, in warfarin therapy.

Main Methods:

• Simulated warfarin treatment outcomes using virtual patients (clinical avatars) and five existing treatment plans.
• Utilized Time-in-Therapeutic-Range (TTR) as the objective function to minimize overall risk, representing both bleeding and thrombosis.
• Employed a decision tree algorithm to derive specific treatment rules for patient subgroups.

Main Results:

• Optimization approaches demonstrated that tailored decision rules matched to patient characteristics yield risk-minimizing treatment plans.
• The decision tree algorithm successfully produced rules identifying optimal treatment plans for distinct patient subgroups.
• The optimization method's minimization of entropy/impurity led to rules that effectively reduce overall adverse risks for broad patient populations.

Conclusions:

• Optimization-based decision support rules can significantly improve personalized warfarin treatment by minimizing patient risk.
• Decision trees are effective in translating complex optimization outputs into actionable clinical guidelines for warfarin management.
• This approach offers a pathway to safer and more effective warfarin therapy through data-driven, subgroup-specific treatment strategies.