Pharmacometric Model-Based Real-World Data Driven Framework to Refine Unfractionated Heparin Dosing in Pediatric Population

Abstract

Unfractionated heparin (UFH) is a commonly used anticoagulant in the pediatric population and is considered the gold standard in extracorporeal membrane oxygenation (ECMO) settings to prevent circuit thrombosis. Optimal pediatric dosing is challenging due to the paucity of clinical outcomes and pharmacokinetic/pharmacodynamic (PKPD) studies in pediatric age groups. Clinicians rely on clinical experience, local hospital protocols or extrapolation from adults to decide UFH dose in the pediatric population. Reaching therapeutic targets within 24 hours has been associated with improved clinical outcomes for UFH therapy. However, less than 20% of pediatric patients achieve this goal. The readily available PKPD information and the availability of an adequate number of subjects representing all pediatric age groups with minimal selection bias make real-world data (RWD) a promising approach for answering pediatric dosing questions. This thesis aims to leverage RWD to develop a PKPD model-based framework for the entire pediatric age spectrum (neonates - <19 years) that could be used as a platform to inform objective clinical decision making (i.e., better dose selection/titration) in pediatric patients in general and in the ECMO pediatric sub-population. RWD were curated, processed, and qualified from the electronic health records (EHR) of Texas Children’s Hospital (n = 490), Utah Children’s Hospital (n = 159) and University of Maryland Medical Center (n = 29). The retrieved data of pediatric patients treated with UFH for ECMO/ non-ECMO indications contained complete information about UFH dosing, monitoring, and patient demographics. The multicenter EHR data were integrated to develop and externally validate Bayesian UFH PKPD model. The Bayesian model leveraged prior knowledge from the literature and adequately described UFH PK, measured through anti-factor Xa assay, and PKPD relationship, where the PD measurements were activated partial thromboplastin time (aPTT) and activated clotting time (ACT). Based on simulations, optimized starting infusions were proposed to achieve the therapeutic target in 45-70% of pediatric patients after initial dose. Additionally, simulations suggested refined UFH titration nomograms that can attain therapeutic targets in > 90% of patients while minimizing the target exceeding to < 1% 24 hours post UFH treatment. The Bayesian model showed potential for personalized UFH therapeutic management. The developed Bayesian PKPD model along with the dose titration schemes could eventually be incorporated into a clinical decision support tool to objectively guide clinicians with UFH dosing decisions in pediatric anticoagulation or intensive care clinics.