A repeated time-to-event model for personalized treatment of patients with hemophilia A based on individual bleeding risk

Background: 

Pharmacokinetic-guided dosing of factor (F)VIII concentrates is widely recommended to personalize the treatment of patients with hemophilia A. However, it is well known that commonly applied target FVIII plasma levels do not necessarily improve bleeding outcomes in all individuals. It is therefore desirable to adapt treatment based on individual bleeding risk rather than only factor levels. Unfortunately, there is currently no reliable clinical marker that reliably differentiates between patients with low or high-bleeding risk. 

Objectives: 

We explore the possibility of using repeated time-to-event models to quantify individual bleeding risk by combining (historical) information on annual bleeding rates and pharmacokinetics, without requiring the measurement of novel clinical markers.

Methods: 

We improved upon existing repeated time-to-event models using data from 264 patients with severe hemophilia A followed during 3 clinical trials. The model classified patients into low-bleeding, medium-bleeding, or high-bleeding frequency cohorts to reduce between-patient variability and predicted bleeding risk for specific categories of bleeds rather than pooled bleeding information.

Results: 

The resulting model has high accuracy, with >70% of predictions being within 1 bleed of the true observed bleeding rate. We demonstrate how the proposed model can be used to compare treatment based on not only the achievement of specific factor levels and factor concentrate consumption but also on projected bleeding outcomes. Importantly, this approach does not require the measurement of novel or unconventional biomarkers, facilitating its adoption in routine clinical practice. 

Conclusion: 

The proposed method may present an exciting new treatment paradigm for patients with hemophilia A.