A new computer algorithm for better dosage of oral anti-coagulants

Abstract number: P1967

Kruithof C. J., Vanderschoot J. P. M., Meer Van Der F. J. M., Rosendaal F. R.

Leiden University Medical Centre, Netherlands

Variation in the sensitivity for oral anti-coagulants between patients and within the patient, and the narrow therapeutic window hinders management of oral anti-coagulant treatment. Therefore, several computer programs have been developed since the 1970s to support dosing decisions. The main problem with the current algorithms, including the one used in several anticoagulation clinics in the Netherlands (‘TRODIS’), is the low proportion of patients for whom the algorithm is able to give a proposal. To improve the number and the quality of the dose proposals, a new dosing algorithm was developed. The algorithm consists of two submodels: the first one describes the dynamic properties of the anticoagulation process, and the second one the nonlinear relation between the dose and INR. Two constant parameters are included in the model, which define the relation between the dosage and the synthesis of vitamin K dependent coagulation factors and the effective half life of the anti-coagulant. These parameters were calculated with historical data of 2000 treatment periods, for each of the two anti-coagulants in use. The first submodel calculates the effective dosage from the average dose, and the second submodel estimates the sensitivity of the patient; this is followed by a dose proposal. A comparison of the new algorithm with TRODIS, for historical proposals, showed a major decrease of the prediction error (proposed INR vs. measured INR). With use of routine collected data of patients treated at the Leiden Anticoagulation Clinic, the new algorithm was compared to the currently used dosing algorithm by means of an expert evaluation. 195 dose proposals given by the new algorithm on historic data were independently analyzed by two experts. In case of disagreement between the two experts, the lowest score was taken for analysis. In 100% of the cases, the new algorithm was able to give a dose proposal where TRODIS was only able to do this in 61%. 77% of the dose proposals from the new algorithm where evaluated as good by both experts, 17% were considered average, and only 4% was evaluated as incorrect. The new algorithm will be clinically tested in a double blind randomized control trial at the Leiden Anticoagulation Clinic, in 500 patients treated with oral anti-coagulants starting April 2003.