2019/03/21 by Ashkan Sharabiani, Sharabiani, Ashkan, Adam P. Bress +7
Computer Science · Mathematics · Chemistry · #Computational Drug Discovery Methods #Statistical Methods in Clinical Trials #Analytical Methods in Pharmaceuticals
paper · pdf · doi:10.48550/arxiv.1903.09267
Determining the optimal initial dose for warfarin is a critically important\ntask. Several factors have an impact on the therapeutic dose for individual\npatients, such as patients' physical attributes (Age, Height, etc.), medication\nprofile, co-morbidities, and metabolic genotypes (CYP2C9 and VKORC1). These\nwide range factors influencing therapeutic dose, create a complex environment\nfor clinicians to determine the optimal initial dose. Using a sample of 4,237\npatients, we have proposed a companion classification model to one of the most\npopular dosing algorithms (International Warfarin Pharmacogenetics Consortium\n(IWPC) clinical model), which identifies the appropriate cohort of patients for\napplying this model. The proposed model functions as a clinical decision\nsupport system which assists clinicians in dosing. We have developed a\nclassification model using Support Vector Machines, with a polynomial kernel\nfunction to determine if applying the dose prediction model is appropriate for\na given patient. The IWPC clinical model will only be used if the patient is\nclassified as "Safe for model". By using the proposed methodology, the dosing\nmode's prediction accuracy increases by 15 percent in terms of Root Mean\nSquared Error and 17 percent in terms of Mean Absolute Error in dose estimates\nof patients classified as "Safe for model".\n