Model-Informed Dosing Regimen Of Ticagrelor In Chinese Patients With Acute Coronary Syndrome

Background and ObjectivesTicagrelor is a reversible P2Y₁₂ inhibitor with rapid platelet inhibition and was commonly used in dual antiplatelet therapy in patients with Acute Coronary Syndrome.However,clinical evidence found ticagrelor administration regimens recommended by Western guidelines significantly increased the risk of bleeding in the East Asian population compared to clopidogrel.Therefore,it was necessary to explore a suitable dosing regimen for ticagrelor in Chinese ACS patients.Thus,this study will explore the dosing regimen of ticagrelor in Chinese patients with Acute Coronary SyndromeMethods（1）A prospective,randomized,open,triple-cross self-controlled study was conducted.Healthy volunteers with CYP2C19*1*1 were enrolled and given a single oral administration of 60 mg ticagrelor,90 mg ticagrelor and 75 mg clopidogrel according to a random sequence in three periods.Collect blood samples at 0-48 hours after a single dose,and the blood concentration of ticagrelor and its metabolite AR-C124910XX and the platelet reactivity index（PRI）were measured.Pharmacodynamic characteristics were compared among the three groups and the PK/PD characteristics of ticagrelor in the Chinese population were preliminarily evaluated.（2）Population pharmacokinetics（PPK）models of ticagrelor and AR-C124910XX in Chinese healthy subjects were established based on demographic data,physiological indicators,and pharmacokinetic data from the trial in Chinese healthy subjects.After the final PPK model was established,goodness of fit graph,bootstrap and visual predictive test（VPC）were used for internal validation to evaluate the stability and predictive performance of the model.Based on the data of ACS patients treated with ticagrelor in a previous prospective real-world observational study of our group,the PPK model was used to simulate the peak steady-state concentration(C_max,ss)of each ACS patient,which compared with the PK data of ticagrelor in Chinese ACS patients reported in papers.（3）The model for predicting the risk of bleeding events within 12months of follow-up was established by machine learning method based on C_max,ss of ACS patients and clinical factors of patients.A total of eight classical machine learning methods were selected to evaluate the prediction performance of the model.F1 score,accuracy,precision,recall rate,and area under the ROC curve（AUC）were evaluation indicators to select the best model.Based on the combination of PPK and machine learning model,individualized ticagrelor administration with a lower risk of bleeding can be obtained.Results（1）Both 60 mg ticagrelor and 90 mg ticagrelor reached the maximum antiplatelet effect at 4 h,which was（17.13±9.86）%and（12.74±9.87）%,respectively.However,there was no significant difference in the antiplatelet effect between the two groups at each time point（P>0.05）.Ticagrelor and its metabolite AR-C124910XX were both correlated with PRI,but the pharmacodynamic effect of ticagrelor was delayed relative to pharmacokinetics.Even in CYP2C19*1*1 Chinese population,the antiplatelet effect of clopidogrel was still significantly lower than that of ticagrelor（P<0.05）.（2）The PPK model of ticagrelor and its metabolite AR-C124910XX in the healthy Chinese population was established,and the structure model of primary drug and metabolite were two-compartment models.SEX,ALT,and TC were the covariables of the final model.ALT was considered to have a significant effect on the clearance of ticagrelor,while SEX and TC had a significant effect on the clearance of AR-C124910XX.C_max,ss of 151ACS patients were simulated based on this model,which was close to the data in the literature.（3）By machine learning feature selection,it was found that C_max,ss,hypertension,heart failure,andβ-blockers were significantly positively correlated with bleeding events,while diabetes and stroke were significantly negatively correlated.Among the eight machine learning algorithms,K-Nearest Neighbor（KNN）algorithm performed best,with F1score,accuracy,precision,and recall rate of the final model reaching 0.83,and the area under the ROC curve reaching 0.88.The prediction performance of this model is excellent and can be further verified prospectively.ConclusionsIn this study,the PPK model and machine learning algorithm were used to bridge real-world data to establish a risk model for predicting bleeding events based on drug exposure and clinical factors,which showed good predictive performance.This model could be used to predict ticagrelor dosing regimens with a lower risk of individual bleeding.This result is expected to be further prospectively validated in clinical practice.