A Bayesian dose-individualization method for warfarin.

Area of Science:

• Pharmacology
• Clinical Pharmacy
• Medical Informatics

Background:

• Warfarin dosing presents challenges due to significant inter-individual variability, leading to suboptimal therapeutic outcomes.
• Current warfarin management strategies result in international normalized ratios (INRs) within the target range only 40-65% of the time.
• Poor INR control increases risks of clotting (INR < 2) and major bleeding (INR > 4).

Purpose of the Study:

• To prospectively evaluate the predictive accuracy of the TCIWorks Bayesian warfarin dosing system.
• To determine the anticipated time in therapeutic range (TTR) for INR values predicted by TCIWorks.

Main Methods:

• Prospective recruitment of 55 patients initiating warfarin therapy at Dunedin Hospital.
• Warfarin doses were inputted into TCIWorks, with predicted steady-state INR (INR(ss)) values generated using 0-6 serial INR observations.
• Prediction accuracy assessed using bias (mean prediction error) and imprecision (root mean square error).

Main Results:

• Initial predictions without observed INR data showed positive bias (MPE 0.52), which resolved with data input.
• Prediction precision significantly improved with three or more INR observations (RMSE < 0.53).
• TCIWorks is predicted to achieve 65% TTR with three INR measurements and 80% TTR with six measurements.

Conclusions:

• The TCIWorks Bayesian method provides accurate and precise INR(ss) predictions for warfarin therapy.
• The system is expected to achieve 65-80% TTR with three or more INR observations, aiding clinicians.
• Further research is recommended to evaluate the long-term impact of TCIWorks on INR control.