Variable pharmacokinetics of coagulation factor VIII in the perioperative settting complicates personalisation of treatment in patients with haemophilia A

Pharmacokinetic (PK)-guided dosing of factor concentrates in patients with haemophilia A is generally recommended for the optimisation of prophylactic treatment. PK-guided dosing can also be useful in the perioperative setting, where guidelines advise to keep factor VIII (FVIII) activity levels within tight target ranges to prevent bleeding. Previous studies suggest changes in FVIII PK following medical procedures as well as potential time-dependent effects, meaning that population PK models specific to the perioperative setting are required. In this study, we use data from haemophilia A patients collected during the prophylactic and perioperative setting to identify covariates that explain changes in FVIII PK. Additionally, we use machine learning methods to find potential time-dependent effects on FVIII clearance. Perioperative FVIII clearance was generally lower compared to the prophylactic setting. Covariates related to the complexity of the medical procedures were correlated to larger decreases in clearance. Importantly, subjects undergoing more complex procedures also depicted potentially relevant time-dependent effects on clearance. These effects could be highly variable between subjects. Directly using PK parameters obtained from the prophylactic setting resulted in relatively high mean absolute percentage error (MAPE) at 26.3%, while the perioperative model including time-dependent effects depicted markedly reduced error (10.3%). The presence of high variability between subjects and potential time-dependent effects complicates the selection of optimal dosing regimens before the start of medical procedures. Our method can be used to optimise treatment in real-time, but close monitoring of FVIII levels will likely remain necessary.